<%BANNER%>

Clinical Effectiveness and Cost of Antiviral Therapy in Patients with Hepatitis C Infection in a Managed Care Setting

Permanent Link: http://ufdc.ufl.edu/UFE0024183/00001

Material Information

Title: Clinical Effectiveness and Cost of Antiviral Therapy in Patients with Hepatitis C Infection in a Managed Care Setting
Physical Description: 1 online resource (163 p.)
Language: english
Creator: Hsu, Chien-Ning
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2010

Subjects

Subjects / Keywords: cea, hcv, hepatitis, inb, mco
Pharmaceutical Outcomes and Policy -- Dissertations, Academic -- UF
Genre: Pharmaceutical Sciences thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Whether initial combination antiviral therapy results a reduction in mortality and prevention of liver transplantation and hepatocellular carcinoma in a managed care organization (MCO) setting remained unclear. The sampling uncertainty in the incremental cost-effectiveness ratio (ICER) statistic has been argued regarding to the ambiguous interpretation of negative ICERs. To overcome the statistical problems inherited in the ratio statistic and reflect current patterns of HCV care in the practice setting, the present study are: to evaluate the effectiveness of treatment in terms of end-stage liver disease (ESLD) development; to evaluate the total health care costs of treatment; to evaluate the cost-effectiveness for treatment relative to no treatment by employing the regression method in the net benefit framework. We conducted a retrospective cohort study among managed care organization (MCO) members using the Integrated Health Care Information Services (IHCIS) National Managed Care Benchmark Database in the period January 1997 to June 2007. With the base case (?1 claims of combination prescriptions), usual care (12 months of continued combination therapy) and extended care ( > 12 months of continued combination therapy) analyses, the results of present study revealed that both estimates of treatment effectiveness, mean to ESLD development (in months) and hazard ratios agreed on the beneficial effect of antiviral therapy in patients with cirrhosis in base case and usual care analyses. Study results in extended care analyses with both measures of effectiveness found no difference in treatment effect between treated and untreated groups. Statistical evidence suggests that initial combination antiviral therapy was cost-effective for cirrhotic patients in base case analysis given at the value of willingness to pay ? $15,000. Initial combination antiviral therapy was also cost-effective for cirrhotic patients with usual care given at the value of willing to pay ? $60,000. Our study results support the current treatment strategy, regarding the continuation of antiviral therapy should take early virological responses, duration of therapy and genotype of HCV into account.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Chien-Ning Hsu.
Thesis: Thesis (Ph.D.)--University of Florida, 2010.
Local: Adviser: Kauf, Teresa.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2010-10-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2010
System ID: UFE0024183:00001

Permanent Link: http://ufdc.ufl.edu/UFE0024183/00001

Material Information

Title: Clinical Effectiveness and Cost of Antiviral Therapy in Patients with Hepatitis C Infection in a Managed Care Setting
Physical Description: 1 online resource (163 p.)
Language: english
Creator: Hsu, Chien-Ning
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2010

Subjects

Subjects / Keywords: cea, hcv, hepatitis, inb, mco
Pharmaceutical Outcomes and Policy -- Dissertations, Academic -- UF
Genre: Pharmaceutical Sciences thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Whether initial combination antiviral therapy results a reduction in mortality and prevention of liver transplantation and hepatocellular carcinoma in a managed care organization (MCO) setting remained unclear. The sampling uncertainty in the incremental cost-effectiveness ratio (ICER) statistic has been argued regarding to the ambiguous interpretation of negative ICERs. To overcome the statistical problems inherited in the ratio statistic and reflect current patterns of HCV care in the practice setting, the present study are: to evaluate the effectiveness of treatment in terms of end-stage liver disease (ESLD) development; to evaluate the total health care costs of treatment; to evaluate the cost-effectiveness for treatment relative to no treatment by employing the regression method in the net benefit framework. We conducted a retrospective cohort study among managed care organization (MCO) members using the Integrated Health Care Information Services (IHCIS) National Managed Care Benchmark Database in the period January 1997 to June 2007. With the base case (?1 claims of combination prescriptions), usual care (12 months of continued combination therapy) and extended care ( > 12 months of continued combination therapy) analyses, the results of present study revealed that both estimates of treatment effectiveness, mean to ESLD development (in months) and hazard ratios agreed on the beneficial effect of antiviral therapy in patients with cirrhosis in base case and usual care analyses. Study results in extended care analyses with both measures of effectiveness found no difference in treatment effect between treated and untreated groups. Statistical evidence suggests that initial combination antiviral therapy was cost-effective for cirrhotic patients in base case analysis given at the value of willingness to pay ? $15,000. Initial combination antiviral therapy was also cost-effective for cirrhotic patients with usual care given at the value of willing to pay ? $60,000. Our study results support the current treatment strategy, regarding the continuation of antiviral therapy should take early virological responses, duration of therapy and genotype of HCV into account.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Chien-Ning Hsu.
Thesis: Thesis (Ph.D.)--University of Florida, 2010.
Local: Adviser: Kauf, Teresa.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2010-10-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2010
System ID: UFE0024183:00001


This item has the following downloads:


Full Text

PAGE 1

1 CLINICAL EFFECTIVENESS AND COST OF ANTIVIRAL THERAPY IN PA TIENTS WITH HEPATITIS C INFECTION IN A MANAGED CARE SETTING By CHIEN-NING HSU A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2010

PAGE 2

2 2010 Chien-Ning Hsu

PAGE 3

3 To my family

PAGE 4

4 ACKNOWLEDGMENTS I express my gratitude to my supervisory committee chair, Dr. Teresa Ka uf, for her guidance and support all the time. This study would not have been completed without her guidance. I give my sincere thanks to my supervisory committee members, Drs. Almut Winterstein, Earlene Lipowski, David Nelson, Jeffery Harman, and Zhou Yang for their expe rtise, instruction in research methodology development, and encouragement. With their help, I have grea tly extended my knowledge in outcomes research area, which will greatly benefit m e in my future research. I thank Mr. Huazhi Liu and Dr. Ning Li for their help with SAS program ming. I am grateful for having them around for friendship and assistance. I express my gratitude to the Department of Pharmaceutical Outcomes and Policy for their support. I appreciate the support of Dr. Richard Segal and Dr. Carole Kimberlin from t he beginning of my study here. Finally, I thank my fellow graduate students for t heir emotional support and company in the process of completing our dissertations. I wish them all the best in their journey to success.

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ...............................................................................................................4 LIST OF FIGURES .......................................................................................................................10 LIST OF ABBREVIATIONS ........................................................................................................11 ABSTRACT ...................................................................................................................................12 CHAPTER 1 INTRODUCTION ..................................................................................................................14 Background .............................................................................................................................14 Burden of Hepatitis C Virus Infection .............................................................................14 Natural History of Hepatitis C Virus Infection ...............................................................15 Treatment of Hepatitis C .................................................................................................15 Rationale for Economic Evaluations of Antiviral Therapy .............................................16 Need for Study ........................................................................................................................17 Limitations of Randomized Controlled Trial-Based Modeling Studies ..........................17 Limitations of Traditional Cost Effectiveness Estimates ................................................19 Purpose of Study .....................................................................................................................21 Study Aims .............................................................................................................................22 2 METHODOLOGICAL REVIEW ..........................................................................................26 Net-Benefit Approach to Overcome the Ratio Statistic of Cost-Effecti veness ......................26 Net-Benefit Regression Approach of Cost-effectiveness Analysis ........................................27 3 METHODOLOGY .................................................................................................................31 Data Sources ...........................................................................................................................31 Study Population .....................................................................................................................32 Exposure to Combination Antiviral Therapy .........................................................................33 Definition of Treatment Exposure ...................................................................................34 Establishment of Comparison Groups .............................................................................35 Base case analysis ....................................................................................................35 Subgroup analysis ....................................................................................................35 Stratified analysis .....................................................................................................36 Study Outcomes and Follow-Up ............................................................................................37 Effectiveness Outcomes ..................................................................................................37 Cost Outcomes .................................................................................................................38 Confounders and Covariates ...................................................................................................38 Baseline Characteristics of Study Cohort ........................................................................38 Treatment Initiation-Related Covariates .........................................................................45

PAGE 6

6 Analysis Plan and Study Hypothesis ......................................................................................45 Descriptive Data Analysis ...............................................................................................45 Effectiveness Estimation .................................................................................................45 Costs Estimation ..............................................................................................................47 Cost-Effectiveness of Initial Combination Antiviral Therapy (Net Be nefit Regression Model) .......................................................................................................48 Primary study hypothesis .........................................................................................48 Secondary study hypothesis .....................................................................................49 Regression diagnostics .............................................................................................50 Plot of Incremental Net Benefits .....................................................................................52 4 RESULTS ...............................................................................................................................63 Descriptive Characteristics .....................................................................................................63 Study Cohort ....................................................................................................................63 Baseline Characteristics of All Study Patients in Base Case Analys is ...........................63 Baseline Characteristics of Patients in Subgroup Analysis .............................................65 Estimated Effectiveness and Costs .........................................................................................65 Study Follow-Up and Clinical Outcome Events .............................................................65 Primary Effectiveness Measure: Time to ESLD Development .......................................66 Secondary Effectiveness Measure: Rate of ESLD Development ....................................70 Summary of Effectiveness Results ..................................................................................72 Total Cost ........................................................................................................................74 Cost-Effectiveness of Initial Combination Antiviral Therapy ................................................75 Net Benefit Regression Model ........................................................................................75 Significance and Effect of Covariates .............................................................................76 5 DISCUSSION .......................................................................................................................110 Effectiveness of Antiviral Therapy .......................................................................................110 Descriptions of Patients Characteristics ........................................................................110 Estimates of Treatment Effectiveness ...........................................................................111 Incremental Net Benefit of Initial Combination Antiviral Therapy .....................................113 Cirrhotic Patients ...........................................................................................................113 Non-Cirrhotic Patients ...................................................................................................116 Limitations ............................................................................................................................117 Future Research ....................................................................................................................119 Summary and Conclusions ...................................................................................................120 APPENDIX A NET BENEFIT OF INITIAL COMBINATION ANTIVIRAL THERAPY I N BASE CASE ANALYSIS ...............................................................................................................123 B NET BENEFIT OF INITIAL COMBINATION ANTIVIRAL THERAPY I N USUAL CARE ANALYSIS ...............................................................................................................134

PAGE 7

7 C COVARIATES EFFECTS ON THE INB OF ANTIVIRAL THERAPY IN BA SE CASE ANALYSIS ...............................................................................................................138 D COVARIATES EFFECTS ON THE INB OF ANTIVIRAL THERAPY IN US UAL CARE ANALYSIS ...............................................................................................................147 LIST OF REFERENCES .............................................................................................................155 BIOGRAPHICAL SKETCH .......................................................................................................163

PAGE 8

8 LIST OF TABLES Table page 3-1 The ICD-9, CPT, and HCPCS codes used for disease diagnoses and procedures ...............53 3-2 Study variable list .................................................................................................................57 3-3 Patterns of antiviral therapy..................................................................................................59 41 Baseline characteristics of all patients between treatment and contr ol groups in base case analysis .........................................................................................................................78 4-2 Factors associated with initiation of combination antiviral therapy amon g patients in base case analysis .................................................................................................................80 4-3 Factors associated with initiation of combination antiviral therapy amon g patients with or without cirrhosis involving in base case analysis ............................................................81 4-4 Baseline characteristics of subgroup patients with usual care or extended ca re ..................82 4-5 Factors associated with initiation of combination antiviral therapy amon g patients with usual care or extended care...................................................................................................84 4-6 Primary effectiveness results in base case analysis: Time to ES LD development in patients with and without cirrhosis at baseline .....................................................................85 4-7 Factors associated with primary effectiveness of antiviral thera py among patients with and without cirrhosis in base case analysis ..........................................................................86 4-8 Primary effectiveness results in subgroup analysis: Time to ESLD deve lopment in patients with usual care and extended care ...........................................................................88 4-9 Factors associated with primary effectiveness among patients wi th usual care and extended care ........................................................................................................................89 4-10 Summary of primary effectiveness results in base case and usual care analyses among patients with or without cirrhosis at baseline .......................................................................91 4-11 Factors associated with primary effectiveness among patients wi th or without cirrhosis in usual care analysis ............................................................................................................92 4-12 Secondary effectiveness results in base case analysis: Rate of ESLD development in patients with or without cirrhosis .........................................................................................94 4-13 Factors associated with secondary effectiveness among patients wi th or without cirrhosis in base case analysis ..............................................................................................95 4-14 Second effectiveness results in subgroup analysis: Rate of ESLD development i n patients with usual care or extended care .............................................................................96

PAGE 9

9 4-15 Factors associated with secondary effectiveness among patients wi th usual care or extended care ........................................................................................................................97 4-16 Summary of secondary effectiveness results in base case and usual care analyses among patients with or without cirrhosis at baseline ...........................................................98 4-17 Factors associated with secondary effectiveness among patients wi th or without cirrhosis in usual care analysis .............................................................................................99 4-18 Total cost among patients with cirrhosis in base case and usual care analys es .................100 4-19 Factors associated with mean total cost difference between treatme nt and control among patients with cirrhosis in base case and usual care analyses ..................................101 4-20 Adjusted mean net benefit difference between treatment and control among cirr hotic patients in base case and usual care analyses .....................................................................104 4-21 Covariates effects on the INB of antiviral therapy for patients with ci rrhosis in base case analysis .......................................................................................................................105 4-22 Covariates effects on the INB of antiviral therapy for patients with ci rrhosis in usual care analysis........................................................................................................................106

PAGE 10

10 LIST OF FIGURES Figure page 1-1 Natural history of hepatitis C virus (HCV) infection. ..........................................................24 1-2 Milestones of interferon (INF)-based therapy for chronic hepatitis C .................................24 1-3 Bootstrapped confidence ellipse of ICERs on the cost-effectiveness pla ne.........................25 2-1 Net monetary benefit (NMB) as a function of the threshold cost-effective ness ratio.. ........30 3-1 Sample selection process. .....................................................................................................60 3-2 Establishment of comparisons groups. .................................................................................61 3-3 Study outcomes and follow-up. ............................................................................................61 3-4 Plot of incremental net benefit (INB). ..................................................................................62 4-1 Cumulative ESLD events in the study follow-up. ..............................................................107 4-2 Cumulative ESLD events in the study follow-up (cirrhotic patients). ...............................107 4-3 Cumulative ESLD events in the study follow-up (non-cirrhotic patients). ........................107 4-4 Plot of INB (95% CI) between treatment and control among patients with cirrhosi s in base case and usual care analyses. ......................................................................................108 4-5 P-P plot of the net benefit ( l =$15,000) for cirrhotic patients in base case analysis. .........109 4-6 P-P plot of the net benefit ( l =$60,000) for cirrhotic patients in usual care analysis. ........109 5-1 Plot of INB (95% CI) between treatment and control among patients with and without cirrhosis in base case analyses. ...........................................................................................122

PAGE 11

11 LIST OF ABBREVIATIONS CEA Cost Effectiveness Analysis ESLD End-stage liver disease EVR Early virological response HCC Hepatocellular carcinoma HCV Hepatitis C virus HMO Health Maintenance Organization HR Hazard ratio ICER Incremental cost-effectiveness ratio INB Incremental net benefit INF Interferons alpha MCO Managed Care Organization NB Net benefit NMB Net monetary benefit OLS Ordinary least square OR Odd ratio POS Point of Service PPO Preferred Provider Organization RBV Ribavirin RCT Randomized Controlled Trial SVR Sustained virological response

PAGE 12

12 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy CLINICAL EFFECTIVENESS AND COST OF ANTIVIRAL THERAPY IN PA TIENTS WITH HEPATITIS C INFECTION IN A MANAGED CARE SETTING By Chien-Ning Hsu May 2010 Chair: Teresa Kauf Major: Pharmaceutical Sciences Whether initial combination antiviral therapy results a reduction in mortalit y and prevention of liver transplantation and hepatocellular carcinoma in a managed care or ganization (MCO) setting remained unclear. The sampling uncertainty in the incremen tal cost-effectiveness ratio (ICER) statistic has been argued regarding to the ambiguous interp retation of negative ICERs. To overcome the statistical problems inherited in the ratio statisti c and reflect current patterns of HCV care in the practice setting, the present study are: to evalua te the effectiveness of treatment in terms of end-stage liver disease (ESLD) development; to eva luate the total health care costs of treatment; to evaluate the cost-effectiveness for treat ment relative to no treatment by employing the regression method in the net benefit framework. We conducted a retrospective cohort study among managed care organization (MCO) members using the Integrated Health Care Information Services (IHCIS ) National Managed Care Benchmark Database in the period January 1997 to June 2007. With the base case ( 1 claims of combination prescriptions), usual care (12 months of continued combination therapy) and extended care ( 12 months of continued combination therapy) analyses, the results of present study revealed that both estimates of treatment effectiveness, mean tim e to ESLD development (in months) and hazard ratios agreed on the beneficial effect of antiviral therap y in patients with

PAGE 13

13 cirrhosis in base case and usual care analyses. Study results in extended car e analyses with both measures of effectiveness found no difference in treatment effect between treated and untreated groups. Statistical evidence suggests that initial combination antiviral the rapy was cost-effective for cirrhotic patients in base case analysis given at the value of willing ness to pay $15,000. Initial combination antiviral therapy was also cost-effective for cirr hotic patients with usual care given at the value of willing to pay $60,000. Our study results support the current treatment strategy, regarding the conti nuation of antiviral therapy should take early virological responses, duration of therapy and genotype of HCV into account.

PAGE 14

14 CHAPTER 1 INTRODUCTION Background Burden of Hepatitis C Virus Infection Approximately 130 million people worldwide are infected with HCV. 1 HCV accounts for the majority of cases of viral hepatitis in the United States, and it has been estimated that nearly 4 million people are chronically infected with the virus. 2 A national survey showed that HCV infection was most prevalent in those 30 to 49 years of age and among African-Amer icans and Hispanics. 3, 4 Chronic HCV infection is the leading cause of chronic liver disease, cirrhosis and hepatocellular carcinoma (HCC). 5 In the United States, 40% of chronic liver disease is HCVrelated, accounting for 8,000 to 10,000 deaths annually. 6 And now HCV represents the most common indication for liver transplantations. 5 Use of healthcare services among HCV-infected patients increased 25 to 30% every year during 1994 to 2001, as the population of patients with HCV aged. 7 Mathematical projections, based on the known prevalence of HCV-related liver disea se and natural disease progression, estimate the direct medical costs of HCV inf ection from the years 2010 to 2019 to be $10.7 billion ($6.7 to 14.1 billion); societal costs related to premature death from decompensated cirrhosis and HCC were projected to be approximately $21 to $54 billion. 8 Mortality and morbidity related to HCV infection are expected to increase 2 to 3 times over the next decade, because of the expected increase in the number of patients with a dvanced HCV infection. 9 Improvements in the use and effectiveness of antiviral therapy have the potential to reduce HCV-related liver complications, mortality, and healthcar e utilization associated with the disease.

PAGE 15

15 Natural History of Hepatitis C Virus Infection HCV was identified in 1989 as a blood-born RNA virus with 6 distinct major genotypes. Genotype is not a factor in the natural history of disease but does impact treatment response and the length of therapy. According to the Third National Health and Nutrition Exam ination Survey (NHANES III), nearly 80% of HCV RNA positive patients were infected wi th genotypes 1A and 1B in the United State. 10 HCV infection can be detected in blood as soon as 1 to 3 weeks after initial exposure. After acute HCV infection, most of persons are asymptomatic or only mildly sy mptomatic; persistent infection occurs in approximately 85% of cases. 11, 12 Among those with chronic hepatitis C, approximately 20% progressed to cirrhosis within 20 years of ini tial infection as shown in Figure 1-1. 13, 14 15 When cirrhosis decompensation becomes severe enough to cause liver failure, a liver transplant may be the only way to save the life of a pe rson with chronic hepatitis C. HCV-related complication is currently the leading reason for l iver transplants in the U.S. 5 Factors related to increase the risk of progressive liver disease include olde r age at time of infection, male gender, race, alcohol consumption (>30g/day), co-infection with H IV or hepatitis B (HBV) diabetes, and obesity. 16, 17 Treatment of Hepatitis C The goal of HCV treatment is to prevent complications of HCV infection; this is principally achieved by eradication of the virus. Initially, interferon alf a (INF) monotherapy was first approved to treat HCV in 1991; since then, there have been substantial improvements i n the success of HCV treatment 18 (Figure 1-2). In randomized controlled trials, the highest overall sustained virological response (SVR) rates have been achieved with the combinat ion of weekly subcutaneous injections of long-acting pegylated interferon alfa (PEG-IN F) and oral ribavirin (RBV). 18, 19

PAGE 16

16 Current standard of care for treatment of previously untreated (nave) patie nts with HCV infection is combination PEG-INF with RBV for 48 weeks for patients with genoty pe 1 HCV infection and 24 weeks for patients with genotype 2 or 3. Patients are only offered the full duration of therapy if they meet early viral response (EVR) criteria, def ined as at least a 2-log drop in viral load after the first 12 weeks of therapy. Current studies suggest a tre atment stopping rule (discontinuation) in patients with genotype 1 not reaching a 2-log drop in HCV R NA at week 12 and in those who remain HCV RNA positive at week 24. 19 Patient-level factors associated with SVR rates include age younger than 40 years, lower weight, non -Black race, and absence of cirrhosis; virus factors include lower baseline virus RNA levels, non-genotype 1, rapid virus level (RVR) decline after the first 4 weeks of therapy, and adhere nce to treatment. 19 Rationale for Economic Evaluations of Antiviral Therapy HCV poses a substantial clinical, economic, and health-related quality of lif e (HRQoL) burden to the individual and the health care system in the United State. 7-9, 20-22 Economic evaluations of antiviral therapy using decision analytic models are useful for t he following reasons: 1. Randomized controlled trials (RCTs) have shown that an average of 54to56% of patients achieve SVR with PEG-INF combination therapy 23, 24 SVR is an intermediate outcome measure. It is unclear if a successful response to treatment (SVR) is pre dictive of a positive effect on HCV-related morbidity and mortality. 2. Antiviral therapy is expensive and associated with considerable adverse eff ects, the management of which is commonly associated with additional medical expenditures. Severe interferon-related adverse events include depression and marrow suppr ession. 19 Ribavirin is contraindicated in pregnancy, and recommended to be avoided in patients with ischemic cardiovascular and cerebrovascular diseases and renal insufficie ncy. Also, hemolytic anemia may require ribavirin dose reduction or additional therapy. 19 Treatment decisions, therefore, should be recommended to patients when the potential benefits of treatment outweigh the potential risks and costs of therapy. 3. Only a small proportion of HCV-infected patients will progress to cirrhosis or HC C, and treatment is not 100% effective as stated above. In addition, patients who achieve SV R are not homogenous with respect to their risk of developing progressive liver disease.

PAGE 17

17 Allocation of sufficient resources to cover treatment costs for those most in nee d and best able to benefit is imperative to health policy decision makers. Need for Study To offer decision makers guidance on the economically optimal course of action for a given patient group, the needed information is not only the efficacy, but all-important r elevant cost, treatment outcomes, and consequences for each alternative identified. A de cision analytic framework can synthesize all available sources of information to assis t decision makers in the efficient allocation of health care resources. 25 To date, most of the published hepatitis C cost-effectiveness analyses (CEAs) applied Markov modeling to evaluate the prospective long-term (over 20 years, or lifetime) costeffectiveness of antiviral regimens. Findings from current decision ana lytic CEA modeling studies show that two forms of interferon (i.e., interferon and pegylated interferon) in combination with ribavirin are cost-effective in nave patients in terms of reduci ng liver complications or improving quality-adjusted survival. 26-31 Although decision analytic CEA has been widely used to assist policy formation on the adoption of antiviral combination therapy the limitations of the target population in the modeling studies and the statistical limitations of the cost-effectiveness estimates may not fully satisfy the needs of decis ion makers who must consider the use of therapy among a much broader population than that considered to date i n randomized clinical studies. Limitations of Randomized Controlled Trial-Based Modeling Studies Current recommendations for treatment of HCV-infected persons are derived f rom data gathered in previous RCTs. 19, 32 Although RCTs represent the gold standard for establishing efficacy, the methodological approach employed in RCTs often limits the general izability of study findings. Study inclusions and exclusions used in patient selection result i n a non-

PAGE 18

18 representative subset of the HCV-infected population. Patients who were involved in the se trials were carefully selected so as to exclude those with conditions that might potenti ally compromise treatment response. Many of the exclusion criteria in these RCTs included gr oups at high risk for hepatitis C, such as patients with HIV infection, hemophilia, renal disease, and subs tance abuse problems (i.e., alcohol and injection drug use). In addition, study inclusion/exclusion cr iteria limited enrolled populations to a homogenous cohort of patients with a minimum number of complications and who were able and willing to complete the therapy and follow-up t hat assist the assessment of efficacy. 33 As a result, the majority of persons with HCV in the general population were not eligible for enrollment in these studies. 34-39 Similarly, in clinical trial settings, the treatment and the close monitoring approach limit the generaliz ability to routine clinical settings. Finally, the intermediate study endpoints examined in mo st RCTs (i.e., SVR), further limit the ability to address issues of long-term cost-effective ness. Evidence from community settings has shown the discrepancies from the clini cal trial settings on the use of medical services for HCV care. Rates of antiviral the rapy in HCV-infected patients were 10.7% to 30.3% in some populations. 34-39 Older patients or those with psychiatric diseases, HIV co-infection, alcohol and drugs use disorders, and with public-funded insur ance programs were more likely to be ineligible for antiviral therapy. 36, 40-42 These findings indicate that today’s major issues surrounding the treatment of HCV infection in clinic al practice are (1) a broad spectrum of patients remain untreated, (2) the patient’s characteris tics and co-morbidities are key components of the treatment decision, and (3) the duration of treatment and medi cation adherence in community settings may not achieve the same level of treatment responses in clinical trials and adherent with consensus recommendations.

PAGE 19

19 Limitations of Traditional Cost Effectiveness Estimates Additional limitations of existing decision analytic cost-effective model ing studies are two major uncertainties surrounding the outcomes measure (incremental cost effec tiveness ratio, ICER) in the interpretation of CEA results. First is the uncertainty of deci sion rule that describes if the new treatment or intervention is cost-effective in the analysis. 43 The ICER, a traditional outcome measure of CEA, generates an estimate of the extra cost for an additi onal unit of benefit when a CEA involves an intervention of new technology (T 1 ) compared to no treatment or standard care (T 0 ). The expected values of mean cost and mean effect for Ti (i=0 or 1) as u Ci and u Ei respectively, the ICER comparing T 1 to T 0 is defined in Equation 1-1. E C E E C C u u u u u u ICER D D = = 0 1 0 1 (1-1) Determination of whether a new treatment intervention is cost effective (w orthwhile) relies on the decision maker’s willingness to pay ( l ), an unknown value from the cost and effectiveness data. Decision should be made to adopt the new technology if the ICER i s less than the maximum amount of l in Equation 1-2. l< =D D E Cu u ICER (1-2) Where l is the ceiling ratio, that decision maker’s maximum acceptable willingnes s to pay per unit of health gain. Because that l is left entirely to the decision maker, and will presumably vary by decision makers’ various preferences for health relative to other goods among the population or available budgets for health care, the precision of maximum amount of will ingness to paid is taken as a uncertainty surrounding the observed ICER.

PAGE 20

20 Second is the statistical uncertainty in estimation of confidence interval f or ICER. Because the true mCi and mEi in the population are unknown, the un-observerable ICER parameter is estimated using the “analogy” estimator (Equation 1-3). 44 E C E E C C ER C I D D = =0 1 0 1ˆ (1-3) Due to uncertainty in these estimates, the bootstrapping method is most widely used t o estimate the confidence interval (or variance) of ICER by multiple re plications of cost and effect differences in the study samples. It denotes the joint probability (typicall y 0.95) of containing the true ICER parameter, and is shown like an ellipse (called confidence ellipse ) on the costeffectiveness plane (Figure 1-3). Particular concerns occur when construct ing the confidence interval (CI) for the ICER, if the joint probability distribution of cost and effe cts extend more than one quadrant on the CE plane, the ICER confidence interval can be problematic. One example is the study results from an economic evaluation of crisis residential car e (T 1 ) for people who have serious mental illness in need of hospital-level care. 45 A 5,000 bootstrap estimates for ICER, a negative ICER occurs 36% of time (=8%+28%), and 77% (=28/36) of the negative ICERs represent cases in which the crisis residential costs less and pr ovides more. However, 64% of the ICER bootstrap estimates suggest that crisis residential care c osts less and provides less (Figure 3). Consequently, these study results cannot provide meaningful infor mation on the probability of implementing a program which is cost effective for the decision m akers. Because negative ICER may be economically efficient, decision makers might not want t o adopt intervention associated with reduced health. However, some analysts suggested that that magnitude of a negative ICER coveys no useful statistical information, 43 and that the confidence interval of an ICER is meaningful only

PAGE 21

21 when uncertainty is restricted to one of the positive quadrants of the CE plane. Recentl y, the net benefit regression framework of cost effectiveness was developed to manage the se uncertainties surrounding the economic evaluation of treatment intervention. 46 The net benefit (i.e., net monetary benefit, NMB) approach rearranges the formulation of the cost-effe ctiveness ratio with a threshold value (decision rule or willingness-to-paid) to overcome the ratio proble ms with traditional ICER. Furthermore, in the regression framework, factors attr ibutable to ICER can be identified and quantified the magnitude of impact on the ICER. This practical adva ntage of net benefit regression framework, therefore, is able to identify important subgroups of pa tients for the HCV therapy In summary, regardless of the limitations of the traditional cost-effective ness estimate, the results of CEA offer policymakers important information of a rank of a rank of ICE R of antiviral therapy by their maximum amount of willingness to pay. Purpose of Study HCV infection poses an increasingly significant clinical, economic, and heal th-related quality of life burden to the individual, the health care system, and society in gene ral, as the U.S. population with chronic HCV continues to age. In response, many innovative anti-HCV agents will be developed to meet the growing demands for more effective treatment i nterventions, which continue to increase costs for the treatment and prevention of liver complications There is an increasing demand among healthcare policymakers for economic evaluations of c urrent antiviral treatment intervention strategies to the healthcare system. Existing decision analytic models based on clinical trial settings conducted f or a managed care organization viewpoint to compare antiviral treatment strategies h ave served as a guide to efficient resources allocation. 30 Given the uncertainties created by the application of clinical trial data in the estimate of the cost-effectiveness ratio as mentioned above, ret rospective economic

PAGE 22

22 evaluation will provide valuable information to the stakeholders in the system about the i mpact of previous decisions and the cost-effectiveness of current HCV treatments a nd prevention of HCV-related complication strategies. To achieve this purpose, this study used a managed care organization (MCO) dataset to estimate the cost-effectiveness of initial combination antiviral therapy among newly-diagnosed, HCV-infected patients and to identify what factors contribute to economicall y attractive antiviral therapy. This thesis will evaluate the impact of patient-level factors, provi der-level factors, treatment tolerability, and treatment interruption on the efficient use of a ntiviral therapy. MCOs are key sites for analysis of the treatment of HCV. First, the age gr oup most frequently diagnosed as having HCV, ages 30 to 49 years, is likely to be employed and covered by employer -provided health insurance. Approximately 93% of privately insured persons receiving coverage from their employer are enrolled in managed care. 47 Second, a MCO setting automated claims dataset will allow for assessing the possibly best efficiency of antiviral therapy that may differ in a publicly funded healthcare system (e.g., fee-for-service or Medicaid and Medicare programs). To enable efficient use of resources, in a MCO health care environment, the health plan desig ns involve cost-containment strategies, cost sharing and benefit coverage that may aff ect health service utilizations. In this thesis, employing a net benefit regression framew ork will be able to incorporate the impact of patient-level variability on the cost effectivenes s of antiviral therapy which may not be fully explored in previous modeling studies. Study Aims To answer these study questions, three individual analyses were performed to addre ss the following specific aims: Objective 1: To estimate the effectiveness of combination antiviral therapy relative to no treatment in newly-diagnosed, HCV-infected patients. Delay in time to firs t occurrence of end-stage liver disease (ESLD) was the primary effectiveness mea sure, including HCC,

PAGE 23

23 liver transplantation, decompensated cirrhosis and a proxy of death. Furthermore, the secondary effectiveness measure was rate of ESLD progression. Objective 2: To estimate difference in average total health care cost between patie nts with and without treatment during the study period. Objective 3: To assess the cost-effectiveness of combination antiviral therapy relat ive to no treatment among newly-diagnosed, HCV-infected patients while controlling for patients baseline characteristics, including social demographics, comorbid conditions and heal th service use. The incremental net benefit (net monetary benefit, NMB) was employed as a measure of the value for the cost of antiviral therapy relative to no treatme nt in the regression framework. Effects of covariates on the incremental net benefi t of treatment will be assessed by the interaction with treatment in the regression fram ework.

PAGE 24

24 Figure 1-1. Natural history of hepatitis C virus (HCV) infection. (Source: Chen SL, Morgan TR. The natural history of hepatitis C virus (HCV) infection. Int J Med Sci 2006;3(2):4752). Figure 1-2. Milestones of interferon (INF)-based therapy for chronic hepat itis C. (Source: Strader DB, Wright T, Thomas DL, et al. Diagnosis, management, and treatment of hepatitis C. Hepatology 2004;39(4):1147-1171).

PAGE 25

25 Figure 1-3. Bootstrapped confidence ellipse of ICERs on the cost-effectivene ss plane.

PAGE 26

26 CHAPTER 2 METHODOLOGICAL REVIEW Net-Benefit Approach to Overcome the Ratio Statistic of Cost-Effect iveness To quantify this sampling uncertainty in ICER measure, NMB was proposed by Sti nnett and Mullaphy in 1998. 43 The net benefit (NB) was proposed as an alternative summary measure of value for money of health interventions. The NB approach employs a simple scale i n order to overcome the problems with cost-effectiveness (CE) ratios. When the CE rati o comparing a new treatment (T 1 ) with an alternative intervention (T 0 ), the threshold value (R T ) in the analysis was defined in Equation 2-1. ( ) () E C E E C C T R D D = m m m m m m 0 1 0 1 (2-1) New treatment will be adopted if the net monetary benefit (NMB) is greate r than 0 in Equation 2-2. T E CR D D C E T R m m (2-2) In the model, m C and m E are mean cost and mean health effect, respectively, of treatment T i (i=1 or 0). C E T R D D m m is called net monetary benefit (NMB) of the health intervention. It is the increase in effectiveness ( 0 1 E E m m ) multiplied by the amount of money the decision maker is willing to pay per unit of increased effectiveness (R T = l ), less the increase in cost ( 0 1 C C m m ). Therefore, the new health intervention is cost effective (Equation 2-3), if 0 > = D D C E NMB m m l (2-3)

PAGE 27

27 Another form of NB is net health benefit (NHB), which defines the new health intervention as cost effective (Equation 2-4), if 0 > = D Dl m mC E NHB (2-4) Advantages of net benefit compared to the traditional ICER are summarized below. First, NB is better to manage uncertainty in CEA. Unlike the ICER, variance of net benefits can be estimated from sample mean costs and mean effects. Variance of NMB is deter mined by ) cov( 2 ) var( ) var( ) ˆ var( 2 C E C E B M N D D D + D =l l and the variance of NHB is determined by ) cov( 2 ) var( 1 ) var( ) ˆ var( 2 C E C E B H N D D D + D =l l Based on the estimated variance and central limit theorem, a (1a ) % confidence interval (CI) can be constructed as: NB z B N CI 2 2 / ˆs a = Second, it avoids the ambiguity of traditional ICERs. As model 2 showed, NMB is a function of the threshold value of ICER. When l or R T is equal to zero, the negative ICER will be represented as an intercept on the y axis ( a ). Also, NMB is 0 when the ICER of health intervention (T 1 ) is equal to the threshold value ( l or R T ). See Figure 2.1. Net-Benefit Regression Approach of Cost-effectiveness Analysis Hoch et al. first published the net benefits statistic which can be used to estimat e incremental cost-effectiveness within the regression framework. 46 They incorporated the net benefit in the standard linear regression framework to assess the impact of e xplanatory variables on the cost effectiveness.

PAGE 28

28 In the net benefit approach, the difference in the mean net benefit of the new intervent ion (T 1 ) and mean net benefit of standard care treatment (T 0 ) can provide the overall incremental net benefit by ( ) ( ) B M N C E C C E E C E C E NMB NMB ˆ ) ( ) ( 0 1 0 1 0 0 1 1 0 1 D = D = = = -l l l l Therefore, cost effectiveness is estimated by using the net-benefit f ramework to define a net-benefit value for individual subject, i i i C E NMB = l Where Ei and Ci are the observed effect and cost for subject i. A linear model for s ubject i’s netmonetary-benefit (NMB) is formed as in Equation 2-5: = + + + = p j i i ij j i T X NMB 1e d b a (2-5) Where a is the intercept term, p covariates X, T is a treatment dummy variable (taking 1 for new treatment intervention under consideration, 0 for standard care or no treatment), and e is a stochastic error term. Regression coefficient ( d ) on the treatment dummy variable provides the estimate of the incremental net-benefit (cost effectiveness), 0 1NMB NMB Significance of this net-benefit regression framework is to add additional explanatory variabl es in order to directly examine their impact on cost-effectiveness. In the model, d gives the incremental netbenefit of implementing new treatment intervention controlling for confounding vari ables. In addition, the interaction terms can be added into the model for examining marginal e ffects influenced by the covariates:

PAGE 29

29 == + + + + = p j i p j ij j i i ij j i X T T X NMB 11e g d b a (2-6) Where the magnitude and significance of the coefficient g j on the interaction between treatment and covariates indicate how the cost-effectiveness of T 1 is expected to vary at the margin. Advantages of a regression model 4 are the ability to examine the impacts of c ovariates on the incremental cost-effectiveness. Current CEA results pay less attention to the impact of a patient’s comorbid condi tions and medication adherence on the effectiveness of treatment. Furthermore, most studie s were conducted in the earlier years when many techniques were under development; ther efore, decision making uncertainty was not undertaken in the estimation of cost effective ness. There are two distinct advantages of the net-benefit regression model, which are the purpose of using this method of CEA. First, it is able to provide the ability to evaluate the importance of cova riates on the marginal cost-effectiveness of antiviral therapy, thus allowing the i dentification of patient subgroups for HCV therapies that are more cost-effective. Second, it provides att ractive statistical properties to summarize the uncertainty surrounding the outcome m easure of ICER in the cost-effectiveness analysis.

PAGE 30

30 Figure 2-1. Net monetary benefit (NMB) as a function of the threshold cost-eff ectiveness ratio. (Source: Drummond MF, SM TG, O'Brien BJ. Methods for the Economic Evaluation of Health Care Programmes, Chaper 5: Cost Effectiveness Analysis. Third e d. New York: Oxford University Press; 2005).

PAGE 31

31 CHAPTER 3 METHODOLOGY Data Sources Our study sample was constructed using the Integrated Health Care Inform ation Services (IHCIS) National Managed Care Benchmark Database between January 1997 and June 2007. The IHCIS is a national managed care database that includes more than 80.1 mill ion patients from a total of 46 health plans in September 2007. Data elements used in the study include eligibility records, patient demographics, inpatient and outpatient medical se rvices, and pharmacy claims. The IHCIS cost figures are standardized across health plans, reflecting he alth plan payments for all provider services. There are several different approaches (e.g., algorithms, multivariate models) for standardizing pricing for each of the following services categories: inpatient and outpatient facilities, professional services, pharmacy claims, and ancillary services. Adjustments performed by the IHCIS are designed to create standard proc ess reflecting allowed payments. Therefore, price comparisons across patients and geographical area s can be made in a consistent manner. All expenditures for each study sample were collected and a djusted to 2007 US dollars using the Consumer Price Index for Medical Care Services. All data was linked by de-identified person identifiers. According to personal contact, the IHCIS prescription drug information is reli able for the date of prescription and amount paid, but is not reliable for quantity or days of supply for an antiviral prescription dispensed. Patient’s age is only available as year of birth in the dataset. Moreover, information on death and causes of death are not reliable in the IHCIS datas et. In place of such a dataset, operational definitions and assumptions to classify antivi ral therapy exposure and effectiveness of antiviral treatment were derived based on cli nical treatment

PAGE 32

32 guidelines and the availability of dataset. This study was approved by the Ins titutional Review Board of the J. Hillis Miller Health Science Center at the University of Florida on 17 January 2008. Study Population Our study cohort consisted of patients with newly-diagnosed, hepatitis C virus (HC V) infection based on the presence of ICD-9 codes (70.44, 70.41, 70.54, 70.51) for HCV. To include patients with equal exposure to risk factor information for HCV progression betwe en treated and non-treated patients, the newly-diagnosed HCV design allows for minimizing the difference in duration of HCV history and controlling co-existing comorbid conditions. For instance, one might expect patients who had a longer history of HCV infection may be more likely to receive treatment for HCV infection, and perhaps may have failed to respond to previous treatme nt. Moreover, duration of HCV infection is associated with probability of progression fr om mild disease to ESLD. While it is impossible to confirm the accurate time of HCV infection, an operationally defined “newly-diagnosed HCV-infected” cohort would provide the abilit y to adjust for the potential influence of these factors on treatment effect. Inclusion criteria for the study cohort were as follows (Figure 3-1): Step 1. Patients with 1 ICD-9codes for a HCV infection in inpatients, or patients with 2 outpatient encounters 30 days apart with 1 ICD-9 codes for a HCV diagnosis, or patients with <2 outpatient encounters 30 days apart with 1 ICD-9 codes for a HCV diagnosis, but with 1 interferon claims between January 1997 and June 2007. Date of the first claim in which an ICD-9 code for HCV is observed was defined as the diagnosis date. Step 2. Patients must have continuous enrollment for 365 days before the diagnosis date with no gap in enrollment greater than 30 days. Newly-diagnosed HCV patients wer e defined as those who had no HCV claims during the 365 days in the pre-diagnosis period. Step 3. Exclusion criteria for patient selection included the following: o Patients with prior organ transplantation (except for liver) and alcoholism, ide ntified by the presence of ICD-9 codes in the pre-index date period. (See Table 3-1)

PAGE 33

33 o Patients re-infected with HCV after liver transplantation. These patie nts were identified by the first occurrence of ICD-9 codes for liver transplantation i n the preindex date period, and earlier than the first occurrence of HCV-related cirrhosi s claim. Coding system for liver transplantation was described in “Effectivene ss Outcome” section. o Patients with co-existing HCC, identified by the presence of ICD codes for HCC in the pre-index date period. Coding system for liver transplantation was described in “Effectiveness Outcome” section. Step 4. Included patients must be aged between 18 to 64 years at the index date. Age at HCV diagnosis was calculated from the year of diagnosis date to the year of bi rth. Step 5. Additional study criteria for comparisons of study outcomes were describe d in the following sections. Exposure to Combination Antiviral Therapy Exposure to combination antiviral therapy was determined by the presence of Nat ional Drug Code (NDC) in pharmacy claims and Healthcare Common Procedure Coding System (HCPCS) codes (Table 3-1) in outpatient claims for all forms of interferon, inc luding interferon alfa 2a, interferon 2b, Rebetron (interferon alfa 2b combined with ribavirin), PEGINTERFERON (interferon alfa 2a combined with ribavirin), pegylated interferon alfa 2 a and pegylated interferon alfa 2b, and NDC for ribavirin available in the dataset. 32, 48 Only newly-diagnosed HCV patients with initial antiviral combination therapy were included in the analysis. In other words, all treated patients in the study had recei ved at least one claim of interferon in combination with ribavirin after the diagnosis date (i.e., post diagnosis period). Patients were excluded if they were monotherapy users or had a prior ant iviral prescription (including any form of interferon and ribavirin) in the pre-diagnosis pe riod. Rationales of using new treatment users are (1) inclusion of prevalent treat ment users can lead to a treatment group with difference duration of prior therapy, which can lead to ove ror underestimating treatment effect; (2) inclusion of prevalent treatment us ers may lead to fail to

PAGE 34

34 adjust for baseline confounders between treatment groups because these factors are plausibly affected by treatment. 49 Definition of Treatment Exposure Because the reliability of the pharmacy claims is restricted to date of service (prescription fill date) and amount paid, the date of the first interferon prescription was define d as the index date. Operational definitions of treatment exposure were defined by the numbers of continuous prescription refills in each month (30 days) after the index date. Continuous refills were constructed based on the following findings in the pharmacy claims: Over 60% of interferons were dispensed every 24 to 32 days among treated patients ; the days supply of antiviral medications was assumed 30 days; Medication interruption was defined by an allowable gap (i.e., 30 days) between 2 consecutive refills. Continuous refills, therefore, were defined as a £ 60-days period between 2 consecutive refills. If a patient had >60 days between 2 consecuti ve refills it indicated the presence of a medication gap (i.e., >30 days without treatment) before the subsequent refill. If the patients had a medication gap >60 days (i.e., 90 days between 2 consecutive refills), then the patient was considered to continue another new course of treatment on the subsequent refill. Ribavirin was assumed to be dispensed concurrently with interferon for a treate d patient, because the majority of treated patients (>90%) had ribavirin total refills less than 20% interferon total refills, and ribavirin and interferon had similar patterns on di spensing frequency among treated patients in Table3-3. Therefore, information related to interferon refilling in each month was use d to indicate exposure to antiviral combination therapy in the analyses. Ribavirin refilling and pr escribing duration was assessed in the follow-up to examine the level of combination therapy per sistency. In the base case analysis, patients who ever received at least one refi ll of interferon combination with ribavirin were compared to those who never received treatment. To expl ore the influence of treatment duration on the cost-effectiveness of antiviral therapy subgroup analysis was performed in the comparisons of net benefit of treatment among patients wit h “usual care”

PAGE 35

35 or “extended care” relative to no treatment. We used “usual treatment” to indica te patients who ever received 11to14 continuous refills of antiviral therapy after the index date (r eferring to 12 months or 48 weeks of treatment duration within 15 months). Treated patients who received 15 continuous refills of antiviral therapy were classified as “extended care .” Based on the operational definition using continuous prescription refills, we assumed that patient s in the subgroup analysis remained in the study cohort at least 12 months. Establishment of Comparison Groups Base case analysis Study treatment (treated patients) and control (untreated patients) groups were generated by a pseudo-index date randomly assignment. Because there is no information on dates of interferon refills among untreated patients, a pseudo-index date was assi gned to a subset of nontreated patients (controls) based on the distribution of length of time between the dia gnosis date and the index date (i.e., date of first interferon prescription) among treated pa tients. Study controls were randomly selected with treated patients with a 3:1 ratio. For exa mple, for a patient who initiated antiviral therapy 60 days after the HCV diagnosis date, one of the e ligible untreated patients (control) was randomly selected and assigned a pseudo-index date, 60 days af ter this control patient’s diagnosis date, as shown in Figure 3-2. Subgroup analysis In the subgroup analysis, controls were individually selected from a set of untre ated patients based on the distribution of length of time between the HCV diagnosis date and t he end of treatment (i.e., date of last interferon prescription plus 30 days) in patient s with usual care or extended care. Similarly, subgroup controls were randomly selected with trea ted patients with a 3:1 ratio. Similarly, decision on the index date in the subgroup analysis used the same appro ach in the base case analysis described above (based on the length in time between da te of HCV

PAGE 36

36 diagnosis to the date of first interferon prescription among patients with usual care or extended care and their matched controls). Stratified analysis According to the natural history of HCV infection, once an advanced liver diseas e (e.g., cirrhotic fibrosis) is established, the risk of hospital readmission for the init ial or other decompensation, HCC development, and liver transplantation or death may vary across individual patients (Figure 1-1). Since treatment was initiated in patients wi th HCV-related cirrhotic symptoms in clinical practice (as compared to patients with asym ptomatic chronic HCV in the clinical trial setting), a dimension of risk heterogeneity may also under lie treatment effect heterogeneity. For this reason, differences in treatment effectiveness and costs were analyzed between HCV-infected patients with and without compensated cirrhosis. Furthermor e, considering low doses of interferon combined with ribavirin may be initiated to pati ents with mild degree of cirrhosis decompensation who have been a candidate of liver transpl antation in practice. Treatment effect in the cirrhotic group could be confounded by the severi ty of HCV infection (e.g. portal hypertension, variceal bleeding, encephalopathy) are t reated with antiviral therapy. To account for the impact of cirrhosis decompensation on the ESLD development, patients with severe decompensated cirrhosis were categorized and analyz ed in the stratified cirrhotic group both in base case and subgroup analyses. Uses of the ICD-9, CPT, and HCPCS codes for cirrhosis diagnoses and procedures a re shown in Table 3-1. Operational definitions of cirrhosis diagnoses are shown in the “Effectiveness Outcomes” and “Confounders and Covariates” sections. In the anal ysis phase, an interaction term that evaluates the difference in treatment effective ness and cost attributed to cirrhosis was examined by an interaction with treatment in the effectiven ess and cost estimation model.

PAGE 37

37 Study Outcomes and Follow-Up Patients were followed from the index date up to when the outcomes of interest occurre d, the date of the last data in the database, or they turned 65 years of age, whichever came first. Clinical outcomes were censored at the patient’s last follow-up visit or when they turned 65 years of age after the index date. The clinical effectiveness and cost outco mes are defined in the following sections. (Figure 3-3. Study outcomes and follow-up) Effectiveness Outcomes Study primary effectiveness measure was defined as a delay in ESLD m easured by survival time in months. ESLD was defined as the occurrence of severe decompensa tion, HCC, LT, or a proxy of death in the study follow-up period (i.e., the period of observation after t he index date). Study secondary effectiveness measure was the Kaplan-Me ier estimate of ESLD occurrence rate during the study follow-up. Clinical end point of study interest (ESLD) was categorized over the study foll ow-up, including 1 ICD-9 codes of HCC, severe decompensated cirrhosis (e.g., ascites, varicea l bleeding, hepatic coma), and LT in the first 3 diagnostic and procedure fields in outpat ient claims, or 1 ICD-9/CPT codes of HCC, severe hepatic decompensation and LT in first 4 diagnostic and procedure fields in inpatient claims, using ICD-9, CPT, and HCPCS codes as shown in Table 3-1. To establish the validity of ICD-9 codes of severe hepatic decompe nsation status, patients were required to have 2 outpatient encounters 7 days apart with 1 ICD-9 codes for severe decompensated cirrhosis in the outpatient claims file. For pati ents with a diagnosis of severe decompensated cirrhosis in both pre-index and after the index dat e were excluded from the analysis. In addition, the status code V58.69 (in Table 3-1) is commonly coded for long-term use of medication monitoring in patients who received a liver trans plant and with other medical conditions. Patients with V58.69 in the first 3 diagnostic and procedure fields

PAGE 38

38 in outpatient claims file or the first 4 diagnostic and procedure fields in inpatient claims were required to have a concomitant diagnosis of liver transplant. A proxy of death was defined by loss of health insurance eligibility within 32 days after receiving nursing home and hospice services. Information on the nursing home and hospice service was obtained from the provider type code and service type code in the IHCI S user manual in 2007. Cost Outcomes From the MCO perspective, direct medical costs among individual patients including inpatient, outpatient, and pharmacy services incurred in each follow-up month were ex amined. All costs were adjusted to 2007 dollars using the medical service component of the Consum er Price Index. Confounders and Covariates Baseline Characteristics of Study Cohort To control for those patients at high risk for complications with either antiviral t herapy or no treatment, baseline characteristics of cohort members and changes in s ome of these during the study follow-up were identified. The known risk factors for HCV-related complic ations could be obtained and allowed discrete categorizations operationalized from the databas e, as shown in Table 3-2. These are hypothetical potential confounders: Patients’ social demographic characteristics. Patient’s age, gender, year of HCV diagnosis, and social economic status were identified for all study subjects. Studies of the natural history of chronic hepatitis C have shown male patients and those over age 40 at the time of infection were associated with more rapid HCV-induced fibrosis progression. 50 In this study, patient’s age was retrieved in each month during study follow-up followed by the age at HCV diagnosis year, and was treated as a time-dependent covariate in the Cox reg ression model and in

PAGE 39

39 the inverse probability weighting process for total cost estimation. It is important to note that gender is not only associated with disease progression but also tied to poor treatme nt response. A binary variable created for gender was analyzed; female patients were coded by 0 and male patients were coded by 1. Social economic conditions, including type of insurance, type of health plan, and census region, are associated with patient access to antiviral therapy in clinical practice and the willingness to initiate and complete treatment. 34-36, 38 The social economic indicators were subtracted from the eligibility duration covering the HCV diagnosis date, including t ype of insurance, type of health plan, and geographic region. Two binary variables were c reated for the type of insurance (private/public). Patients with public insurance, including Medica re and Medicaid, were coded by 0, and those with private insurance were coded by 1. To further account for the differences among types of health plans, HMO (health maintenance organization) was categorized as plan1, PPO (preferred provider organization) as plan2, and POS (point of service) as plan3. Three binary variables were created for those patients with the type of health plan (i.e. 1/0=yes/no). As a general rule, POS and PPO offer more freedom and choices t han an HMO. Even if patients go out-of-network for their medical needs, they are still c overed to a certain degree. HMOs, for example, do not cover members if they go outside of the HMO network of providers. Unlike PPO, POS plans have no deductibles and limited co-payments for in-network coverage. With a PPO, patients are required to meet deductibles and pay co payments. 51 In a chronic health condition, like HCV infection, PPO and POS offer a greater choice, yet a higher copayment on the access or referral to specialists or the doctor who patients trust may have impacts on the choice of antiviral therapy initiation. Although geographic area may not directly influence HCV care, the practice patterns and patient educ ation levels vary in

PAGE 40

40 different geographic regions. Census regions were categorized by 1=New Engl and and Middle Atlantic, 2=East North Central and West North Central, 3=South Atlantic, East Sout h Central, and West South Central, 4=Mountain and Pacific, and 5=National and Other in the IHCIS dataset. And, 5 binary variables were created for each category of census reg ion was analyzed; patients resided in the census region were coded by 1 and outside the region were code d by 0. Since the National Institute of Health (NIH) initiated their second consensu s guideline regarding the management and treatment of HCV infection in 2002, 52 HCV identification and management have been disseminated among health professionals through updated cli nical guidelines for the HCV-infected subpopulation (e.g., HIV and HBV con-infection, psychi atric/or drug disorders). 18, 19, 53, 54 To examine the potential impact of practice pattern variation on HCV care, the patient’s year of HCV diagnosis was categorized as 1998=0, 1999=1, and on up to 2007=9 in this study, and 10 binary variables created for the individual year of HCV diagnos is was analyzed. Patient comorbid conditions. Risks for more rapid HCV-induced fibrosis progression, such as individuals co-infected with HBV or HIV were identified in all study s ubjects. 55, 56 Even though HIV accelerates HCV-related liver disease, studies showed HCV antivir al therapy appear to be protective in most instances, including those with HIV co-infection and delay hepat ic decompensation. 57-59 Patients with co-infection were identified by the presence of 1 claims of ICD-9 codes for HIV/HBV during the pre-index period. A b inary variable was created in patients co-infected with/without HBV infection (1/0) and with/without HIV infection (1/0). There is increasing information showing that obesity and type II diabetes a re associated with liver fibrosis progression and poor response to hepatitis C treatment. 32 60, 61 Obesity and the

PAGE 41

41 metabolic syndrome, including insulin resistance and hyperlipidemia are assoc iated with steatosis (fatty liver), which is linked with a greater risk of cirrhosis and H CC. 60 To account for the fact that current ICD-9 codes may not be sufficient to captur e patient’s diabetic and/or obese condition, Food and Drug Administration (FDA)-approved hypoglyce mic agents and anti-obesity drugs were used. Patients with obesity and diabetes we re identified by the presence of 1 claims of ICD-9 codes for diabetes or obesity in inpatient or outpatient claim s file, or 1 claims of NDC for FDA-approved prescriptions for obesity and diabetes therapy in pharmacy file during the pre-index period. FDA-approved prescriptions for diabetes included in this study were insulin; metformin and Metaglip (glipizide and metformin), Janumet (sitagliptin and metformin) Glucovance (glyburide and metformin), Actoplus Met (pioglitazone and metformin), Avandamet (rosiglitazone and metformin); sulfonylurea (i.e., glipizide, glyburide, glimepiride), thiazolidinediones (i.e., rosiglitazone, piogl itazone) and Avandaryl (rosiglitazone and glimepiride), Duetact (glimepiride and pioglitazone); alphaglucosidase inhibitors (i.e., acarbose, miglitol), Prandimet (repaglinide and metformin); and meglitinides (i.e., repaglinide, nateglinide). FDA-approved prescriptions for obes ity used in this study were sibutramine and orlistat. A b inary variable was created in patients with/without obesity (1/0) and with/without diabetes (1/0). Patients with mental illness (including depression) and drugs disorders (inje ction drug users) have a greater risk of HCV infection. However, active psychiatric/ drug disorders once were considered as relative contraindications for combination antiviral thera py because of HCV treatment-related psychiatric side effects and lack of successful coll aborating care between experts in HCV and healthcare providers specializing in substance-abuse. 33, 52 Research has shown it is feasible and effective to treat patients with HCV and comorbid psychi atric and

PAGE 42

42 substance use disorders through a multidisciplinary team. 62 In addition, the depression side effects may lead to dose reduction and treatment discontinuation affecting the effica cy of antiviral therapy. The FDA-approved selective serotonin reuptake inhibitors (SS RIs) have been concomitantly prescribed with interferon-based antiviral therapy to redu ce the risk of unwanted effect and complete treatment. 63 Patients with psychiatric/drugs disorders were identified by the presen ce of 1 claims of ICD-9 codes for psychiatric/drugs disorders in inpatient or outpatient claims file during the preindex period. Patients with depression were identified by the presence of 1 claims of ICD-9 codes for depression in inpatient or outpatient claims file or 1 claims of NDC for FDA approved SSRIs during the pre-index period. A hepatologist suggested the common SSRIs us ed in HCV-infected patients with depression were fluoxetine, paroxetine, sertr aline, citalopram, escitalopram, fluvoxamine, venlafaxine, and duloxetine. Additionally, to explore the assoc iation between antidepressant use and treatment outcome, patients prescribed with anti depressants (SSRIs) were identified by the presence of 1 claims of NDC for FDA-approved SSRIs in the follow-up period. Three binary variables were created for patients with/wit hout psychiatric disorder (1/0), with/without drug dependence (1/0), and use/no-use antidepressants (1/ 0). Other comorbid conditions, end-stage renal disease (ESRD) or dialysis, heart dis eases, chronic obstructive pulmonary disease (COPD), cerebral vascular disease (CV D), and hyperthyroidism, were reported to possibly impact treatment decisions. 52 Some of the important factors that need to be taken into consideration of appropriate timing for antivira l therapy in patients with the comorbid conditions above include severity of liver disease, comorbid conditions, and motivation. Patients with comorbid conditions were identified by the pres ence of 1 claims of ICD-9, CPT, and HCPCS codes for ESRD/dialysis, COPD, CVD, heart dis ease, and

PAGE 43

43 hyperthyroidism either in inpatient or outpatient claims filed during the preindex period. Binary variables were created for patients with/without ESRD (1/0), with/without COPD (1/0), with/without CVD (1/0), with/without heart disease (1/0), and with/without hyperthy roidism (1/0). All comorbid conditions were identified using ICD-9, CPT, and HCPCS codes, shown in Table 3-1. Prior medical services uses. Annual medical expenditure, hospitalization, outpatient visits, and emergency room visits within the same pre-index period were used as proxies of pa tient health status. Patients with higher levels of medical service utilization can be expected to have many health issues and greater severity of co-morbidities might adversel y affect the initiation of antiviral therapy and thereby increase the possibility of poor outcome. Patients’ prior annual medical expenditure was calculated by the total amount of payment per year, including outpatient, inpatient, and pharmacy services. Total healthcare expenditure was adjusted to 2007 dollars using the medical service component of the Consumer Price I ndex. To be comparable to the general population, “National Health Care Expenses in the U.S Civilian Noninstitutionalized Population, 2003” data available at the time this study conducte d 64 were used in this study. On average, the annual medical expenditure per patients aged <65 years was approximately $5,670 in 2003 ($6,634 adjusted to 2007 dollars). A binary variable therefore, was created for patients having an annual medical expenditure $6700 higher than the national adult population and patients having an annual medical expenditure $6700 lower than the national adult population (1/0). Liver biopsy once was recommended to be routinely performed prior to treatment initi ation for unsuspected cirrhosis diagnosis, which may change the prognosis and considerati ons regarding therapy or follow-up. 52 There are increasing arguments related with the effectiveness

PAGE 44

44 of biopsy on unsuspected cirrhosis, 65-67 and social costs of biopsy for treatment decision among patients with cirrhosis. 65 Some clinicians are tending to reserve biopsy for circumstances in which the biopsy would influence the decision regarding either initiation or continuati on of therapy. 68 Therefore, it is importance to examine the influence of pre-treatment biopsy on treatment initiation and treatment outcome among patients without cirrhosis. A binary variable was used for patients having a liver biopsy during the pre-index period and patients w ithout a liver biopsy during the pre-index period (1/0). Characteristics of certain health professional visits in patients with HC V infection were used as proxies for the need in comorbid and HCV infection. Patients with more visits of ce rtain professionals were considered to have a greater need for their comorbid or HCV i nfection. Numbers of health professional visits were documented during the pre-index period, includi ng family practice, gastroenterology, internal medicine, and infectious diseas e specialties, identified to examine their impacts on treatment initiation and outcome. Instrument of HCV symptomatic conditions. Although patients with chronic HCV are potential candidates for treatment, each patient’s unique characteristics HCV-related symptoms, and motivation may affect the timing of treatment initiation. Time to treatme nt initiation was calculated based on the duration between index date and HCV diagnosis date, and empiric ally categorized by 6 months and <6 months (0/1). In the absence of controlled study data, no definitive recommendations can be made about the timing of treatment initiati on; however, it seems reasonable to delay treatment for 2 to 4 months after acute onset to allow for spontaneous resolution. It is suggested to repeat viral check 6 months to 1 year after sponta neous clearance due to the possibility of replication restart. 18 Therefore, a cut-point of 6 month was considered as

PAGE 45

45 a proxy to indicate that patients with a shorter time to initiation had a greater level of severity of HCV-related cirrhotic symptoms compared to patients who deferred treatment Treatment Initiation-Related Covariates To identify and account for potential bias introduced by treatment initiation-rela ted covariates on outcome measure, multivariate logistic regressions were used to development adjusted odds ratio (OR) to determine the statistically significant covar iates to be included in the effectiveness and costs estimation models. A stepwise selection method was used to create a final model with statistically significant effects of exploratory va riables on receiving antiviral therapy. There were 32 identified baseline characteristics included into a st epwise, multivariate logistic selection model with a 5% significant level. Table 3-2 shows the speci fications and descriptions of these baseline characteristics. Analysis Plan and Study Hypothesis Descriptive Data Analysis Data were expressed as frequencies, mean (standard deviation), and percent. The associations between treatment initiation and an individual patient’s baseline characteristics were examined using adjusted logistic regression model. The a priori level of significance was set at 0.05 ( a ). All data management was performed using SAS version 9.2 (SAS Institute, Cary, N C). Statistical analyses were done with the STATA IC10.0 (Stata Corp., College S tation, TX). Effectiveness Estimation One inevitable limitation to the study is that some patients were not followed until the last date of dataset (6/30/2007), so their clinical outcomes and costs are were not fully obs erved (censoring). For example, the estimation of the difference in mean cost tend s to be biased due to the fact that the distribution of healthcare costs is typically right skewe d. 69-71 Moreover, healthcare costs may be related to effects of covariates changing over t ime. One of the suggested

PAGE 46

46 statistical methods for handling both issues is to use inverse probability censoring weighting (ICPW). 72, 73 Our study uses weighted ordinary least squares (OLS) regression to estima te total costs and clinical effectiveness. 74, 75 Furthermore, the ICPW statistical technique also has been used to estimat e mean time to ESLD with censored data (Equation 3-1), where the weights are derived from surviva l model estimates from a Cox regression model conditionally on baseline characterist ics and treatment indicator, as in Equation 3-2. Patient’s age at HCV diagnosis in this model was set up a s a timedependent variable. Complete observations are weighted by their inversed probability of not being censored at the time of complication for patients with a complication withi n the study observation interval. For patients who are free of ESLD until the end of time inte rval, the observations are weighted by their inversed probability of not being censored at the end of the interval. 75 As with the ICPW method, the censoring indicator is reversed so that the event of interest is denoted as 0 and the censored event as 1. E(E i ) = E Z Ei (3-1) In Equation 3-1, i = 1,2,3,..,n (number of patients), Z Ei is a set of covariates whose effects on time to a complication occurrence and E is a set of regression parameters. When Z Ei = 1 indicates the i th patient received treatment, then E is the impact of treatment on time of complication occurrence relative to no treatment, adjusted for the other covaria tes. Therefore, the E in Equation 3-1 can be estimated by Equation 3-2. E Z n i n i i X G i X i Ei Z Ei Z i X G i E = = ¢ = 11 *) ( ˆ 1 ) *) ( ˆ ( ˆ d d b (3-2) Where; n = number of patients = 0 if i th patient is censored during entire duration of follow-up, otherwise = 1

PAGE 47

47 = min (time of event, censoring time), X i = time to event, t = censoring time = end of follow-up = Kaplan-Meier estimator of probability of not being censored at = a set of covariates whose effects in effectiveness Costs Estimation Using a similar approach as described above, 74, 75 the estimated costs for an individual patient within the study period were the sum of the product of the Kaplan-Meier probabil ity of costs incurred in each time interval and the mean total costs from observed event in tha t interval in Equation 3-3. The weights used in the linear regression model are based on survival model estimates from the inverse of probability estimates in the time-dependent Cox r egression model ( Equation 3-4). 74, 75 Accrual intervals for the cost estimation in the follow-up were defined by month. Time intervals correspond to the intervals between data collection visits E(C ki ) = Ck Z Ci (3-3) Where; C k I = observed cost for i th patient during interval k. k = 1,2,3…K, which is K intervals in the entire study follow-up (0, t and 0 = a 1
PAGE 48

48 = Kaplan-Meier estimator of probability of not being censored at Cki = observed costs for i th patient during time period k Cost-Effectiveness of Initial Combination Antiviral Therapy (Net Bene fit Regression Model) We performed the main analysis to estimate the cost-effectiveness of com bination antiviral therapy relative to no treatment among newly-diagnosed HCV-infected patie nts. The primary effectiveness measure, time to occurrence of ESLD, was applied in the net benefi t regression model. Because the treatment effectiveness analyses in the base case and subg roup cohorts was confirmed with primary and secondary effectiveness measures in patients w ith cirrhosis, the incremental net benefit was analyzed within two distinct cirrhotic patients cohorts with and without 12months antiviral therapy (i.e. base case and usual care groups of patients). Mor eover, the net benefit regression model was performed for patients without cirrhosis i n base case analysis to compare the magnitude of net benefit difference in pa tients with and without cirrhosis. Individual patient’s NMB was estimated in the Equation 3-5. The maximum willingnes sto-pay (WTP, l ), as recommended by Stinnett and Mullhay, estimated net benefit is a function of a range of values for l 43 A value of l of, for example, $0; $10,000; $25,000; $50,000; $75,000; and $100,000 was employed in Equation 3-5. i C i E i NMB ˆ ˆ =l (3-5) Where; = expected effectiveness with censoring adjustment for i th patient obtained in Equation 3-1 expected costs with censoring adjustment for i th patient obtained in Equation 3-3 Primary study hypothesis “Is antiviral therapy cost effective for HCV-infected patients with c irrhosis in a national MCO, controlling for confounders?” was explored in the construction of Equation 3-6.

PAGE 49

49 =+ + + =p j i i ij j iT X NMB1e d b a (3-6) Where; T i =1 indicates i th patient receiving antiviral combination therapy, otherwise =0. d =estimated cost-effectiveness of antiviral therapy (i.e., incremental net-benefit) compared to no treatment (Ti=0) by controlling for covariates (Xij), including either st atistically significant effectiveness-related or total costs-related covariate s. Null hypothesis (Ho): d dd d 0 It indicates that the mean NMB of combination antiviral therapy was not statistically significant higher than no treatment among ne wly diagnosed HCVinfected patients. Rejection of the null hypothesis would suggest that the HCV-infected patients havi ng antiviral therapy have higher NMB than those who went without treatment (if d >0). Secondary study hypothesis “Which subgroup of patients with potential factors makes antiviral therapy more c ost effective?” The individual impact of a variety of covariates on the estimate d incremental netbenefit was assessed including the interactions of the treatment with covari ates which were found to have a statistically significant risk of complication in Equation 3-7. ==+ + + + =p j i p j ij j i i ij j iX T T X NMB11e g d b a (3-7) In this equation, gj indicates if the cost effectiveness of antiviral therapy statistical ly differs among subgroups. Null hypothesis (Ho): g gg gj =0. It indicates that there is no significant effect on the interaction between individual covariate and treatment. Rejection of the null hypothesis would suggest that the cost effectiveness of ant iviral therapy (mean NMB differences) differed significantly between patient s with specific covariates

PAGE 50

50 and their counterparts (if gj 0). Covariates (Xij) included in the interaction terms with treatment were from model 3-6 when their p-values for the NMB coefficient estimate were statistically significant (<0.05). Regression diagnostics To ensure the regression coefficient (d) in the primary analysis is an unbiased, consistent, and efficient OLS estimator, the assumptions of OLS were examined by the foll owing diagnostic analyses. 76 Heteroskedasticity. The assumption of homoskedasticity is the variance of error term (ei), given Xi is constant. If the variances of error terms vary with Xi there i s heteroskedasticity. Heteroskedasticity could lead to an invalid estimation of the standard error of t he coefficient (d). Although it does not bias the actual coefficient estimates, it can cause an incor rect conclusion due to invalid statistical inferences of coefficient significance. The Breusch-Pagan-Godfrey and White tests were used to detect heteroskedasticity. White’s heteroskedas ticity-consistent covariance matrix (1980) was used to correct heteroskedasticity. Normality of error term. The use of normality of error term is to ensure the distribution of the dependent variable (NMBi) conditional on the independent variables (Xj) was nor mally distributed. If there are skewed, the OLS estimates would be inefficient. A Jarque-Bera test and standardized normal probability plot was used to examine the normality of the resi duals. Omitted variable bias. Omitted variable biases were used to examine if a relevant variable correlated with treatment (Ti) is omitted in the included variables (Xj). If an omitted variable (un-measured variable) is correlated with the independent variables, t he regression coefficient will be biased. The instrumental variable approach was investigat ed to control potential unmeasured treatment selection bias. The instrumental variable appr oach involves the

PAGE 51

51 assumption that an instrument is related to the treatment (Ti), but not related to the outcome of the interest (NMBi). First, two variables, overall outpatient medication copayment rate and having a PPO health plan, were examined to ensure that they could serve as an effective instrumental variable. This factor was a valid instrument if (a) the overall medication copayment rate varies in the likelihood of treatment initiation (unadjusted F test 10), (b) the overall medication copayment rate is independent with measured confounders for treatment initiation (i.e., patient-level characteristics), and (c) the overall medication copayment rate is not rel ated with the study outcome (NMB). Outpatient medication copayment rates were measured by the pr oportion of total amount of the copayment divided by the total medication expenditures incurred during the study period (>20% =1 implied the patient has high medication copayment; 20% = 0 indicated the patient has low copayment). A binary variable, PPO/non-PPO for patients with and w ithout PPO, was examined. The relationship between treatment initiation and instrument m easures was explored using multivariate logistic regression model, individually. Because not every study patient had pharmacy claims, overall medication copaym ent rate was not able to be an instrument in this study. Type of health plan (PPO vs. non-PPO) was a strong predictor of treatment initiation. When PPO was not included in the effective ness and costs estimation model, it was not associated with the net benefit difference controlling for patient baseline characteristics. However, the incremental net benefit wa s not significantly different between patients with PPO and those who without PPO in the instrumental va riable regression model. One of possible causes was the association between PPO and tr eatment initiation is not unique. PPO was not only highly correlated with treatment initia tion, but also correlated with other baseline characteristics in the multivariable logis tic regression model.

PAGE 52

52 Therefore, type of health plan was taken as one of the patient-level charact eristic variables and included in the effectiveness, cost estimation, and net benefit regression models Plot of Incremental Net Benefits At each l value, the model containing the patient’s baseline characteristics and the treatment indicator (multivariable-adjusted model) was performed for hypot hesis 1. Included baseline characteristics in the multivariate adjusted model was either si gnificant relevant with primary effectiveness or total costs. At each level of l value, net benefit regression models were respectively performed in the patient subgroups, patients with and without cirrhosis. Plots of incremental net benefit (INB), can provide visual results for the mea n net benefit difference between treatment and no treatment as a function of WTP (l) values ranging from $0 to $100,000 (e.g., $0; $10,000; $20,000; $30,000; $40,000; and $80,000). Figure 3-4 demonstrates the incremental net benefit (INB) plots. INB is expressed as a function of WTP and mean change in NMB. Along the x-axis are estimates of the WTP to achieve an addit ional treatment success (positive NMB). The slope of each line is positive, indicating that treatment increases effectiveness (i.e., INB), and the curved lines represent the uppe r and lower 95% confidence limits of the INB plot. If the WTP is equal to zero, INB is equal to mi nus the cost difference. Any negative intercept indicates that the treatment incr eases the cost. INBs will change in a linear fashion as the WTP amounts are varied. Health policy makers can t hen make resource allocation decisions based on various WTP values. For any value of l, a net benefit regression produces a value of INB (i.e., cost-effectiveness); the 95% CI ba nd of each NMB plot is allowed to demonstrate the strength of evidence that indicate treatment is cost-effective.

PAGE 53

53 Table 3-1. The ICD-9, CPT, and HCPCS codes used for disease diagnoses and procedures HCV-related disease conditions HCV infection 70.41 Acute hepatitis C with coma 70.51 Acute hepatitis C without coma 70.44 Chronic hepatitis C with coma 70.54 Chronic hepatitis C without coma Compensated cirrhosis (non-alcoholic cirrhosis) 571.4x Chronic liver disease 571.5 Cirrhosis of liver without mention of alcohol 571.6 Biliary cirrhosis 571.8 Other chronic nonalcoholic liver disease 571.9 Unspecified chronic liver disease without mention of alcohol 273.2 Paraproteinemia (Mixed cryoglobulinemia) 446.29 Leukocytoclastic vasculitis associated with chronic Hepatitis 573.0 Chronic passive congestion of liver 573.8 Hepatoptosis 573.9 Unspecified disorder of liver Severe decompensated cirrhosis 456.0x Esophageal varices with bleeding 456.1x Esophageal varices without mention of bleeding 456.2x Esophageal varices in diseases classified elsewhere 456.8x Varices of other sites 572.2x Hepatic coma 572.3x Portal hypertension 572.4x Hepatorenal syndrome 572.8x Other sequelae of chronic liver disease 782.4 Jaundice 789.5x Ascites HCV-related disorders in other system 286.7 Acquired coagulation factor deficiency 286.9 Other and unspecified coagulation defects 289.4 Hypersplenism 416.8 Pulmonary hypertension, secondary 511.8 Other specified forms of effusion, except tuberculous Hepatocellular carcinoma (HCC)/ malignant neoplasm 155.0 Liver, primary 155.1 Intrahepatic bile ducts 155.2 Liver, not specified as primary or secondary

PAGE 54

54 Table 3-1. Continued HCV-related disease conditions Liver transplantation ICD-9 status codes V42.7 Liver transplant V58.69 Long-term (current) use of other medications (patients with V58.69 have to be associated with n 1 claims ICD/CPT codes of liver transplant in inpatient or outpatient claims files) ICD procedure codes: 50.51,50.59 Liver transplant CPT: 47135, 47136 Comorbid conditions HIV co-infection 079.53 Human immunodeficiency virus, type 2 [HIV-2] 042 Human immunodeficiency virus [HIV] disease V08 Asymptomatic human immunodeficiency virus [HIV] infection status HBV co-infection 070.2 Viral hepatitis B with hepatic coma 070.3 Viral hepatitis B without mention of hepatic coma V02.61 Hepatitis B carrier Alcoholism 291.xx Alcoholic psychoses 303.0x Acute alcoholic intoxication 303.9x Other and unspecified alcohol dependence 305.0x Alcohol abuse 357.5 Alcoholic polyneuropathy 425.5 Alcoholic cardiomyopathy 535.3x Alcoholic gastritis 571.0 Alcoholic fatty liver 571.1 Acute alcoholic hepatitis 571.2 Alcoholic cirrhosis of liver 571.3 Alcoholic liver damage, unspecified Obesity 278.x, 278.xx Diabetes 250.x, 250.xx Cerebral vascular diseases 433.xx, 434.xx Occlusion or stenosis of cerebral arteries 435.x Transient ischemic attack 430.x-432.x, 436.x Stroke

PAGE 55

55 Table 3-1. Continued Comorbid conditions Drugs dependence 304.0x Opioid-type dependence 304.2x Cocaine dependence 304.4x Amphetamine and other pscychostimulant dependence 304.7x Combinations of opioid-type drug with any other 305.5x Opioid abuse 305.6x Cocaine abuse 305.7x Amphetamine or related acting sympathomimetic abuse Depression 293.83 Transient organic psychotic condition, depressive type 296.2x Major depressive disorder, single episode 296.3x Major depressive disorder, recurrent episode 296.5x Bipolar affective disorder, depressed 298.0 Depressive type psychosis 300.4 Neurotic depression 307.44 Hypersomnia associated with depression 309.0x Brief depressive reaction 309.1x Prolonged depressive reaction 309.28 Adjustment reaction with anxiety and depression 311 Depressive disorder, not elsewhere classified Chronic mental or mood disorders (Psychiatric diseases) 290.xx Senile and presenile organic psychotic conditions 293.xx Transient organic psychotic conditions 294.xx Other organic psychotic conditions (chronic) 295.xx Schizophrenic disorders 296.xx Affective psychoses 298.9 Unspecified psychosis 300.xx Neurotic disorders excluding 300.4 Heart diseases 77, 78 411.xx, 414.xx Ischemic heart disease 413.x Angina 428.0, 428.1, 428.9, 428.2x, 428.3x, 428.4x, 398.91, 402.01, 402.11, 402.91, 404.01, 404.11, 404.91 Congestive heart failure 410.x, 412.x Myocardial infarction 785.9 Carotid bruit Organ transplantation V42.x (excluding V42.7)

PAGE 56

56 Table 3-1. Continued Comorbid conditions Dialysis or End-stage renal disease, (ESRD) 77 ICD9 codes: V42.0 Kidney transplant V45.1 Renal dialysis status V56.0 Extracorporeal dialysis V56.8 Other dialysis ICD procedure codes: 39.27, 39.42, 39.43,39.49, 39.50, 39.53, 39.93, 39.94. CPT-4: 90921, 90925, 90935, 90937, 90945, 90947, 90940, 90989, 90993, 90997, 90999, 93990, 50340, 50360, 50365, Chronic obstructive pulmonary disease (COPD) 490.x-491.x Chronic bronchitis 492.x Emphysema 494.x Bronchiectasis 496.x Chronic airway obstruction Hyperthyroidism 242.xx Liver biopsy CTP codes: 47000, 47100 Interferon injections ICD procedure codes: 50.11, 50.12 HCPCS codes: J3590, J9219, J9212, J9213, J9214, J9215, J9216, S0145, S0146

PAGE 57

57 Table 3-2. Study variable list Variables Coding of covariates Data Patient demographics Age at HCV diagnosis year 17 to 64 years Age in the month th Continuous Dichotomous Gender Female=0 Male=1 Dichotomous Census region 1 Northeast=1 Yes/No (1/0) Dichotomous Midwest=2 Yes/No (1/0) Dichotomous Southeast=3 Yes/No (1/0) Dichotomous West=4 Yes/No (1/0) Dichotomous Other=5 Yes/No (1/0) Dichotomous Year of HCV diagnosis 1998=0 Yes/No (1/0) Dichotomous 1999=1 Yes/No (1/0) Dichotomous 2000=2 Yes/No (1/0) Dichotomous 2001=3 Yes/No (1/0) Dichotomous 2002=4 Yes/No (1/0) Dichotomous 2003=5 Yes/No (1/0) Dichotomous 2004=6 Yes/No (1/0) Dichotomous 2005=7 Yes/No (1/0) Dichotomous 2006=8 Yes/No (1/0) Dichotomous 2007=9 Yes/No (1/0) Dichotomous Insurance coverage Public=0 Private=1 Dichotomous Types of health plans 2 POS=plan1 Yes/No (1/0) Dichotomous PPO=plan2 Yes/No (1/0) Dichotomous HMO=plan3 Yes/No (1/0) Dichotomous Time to treatment initiation <6 months=1 6 months=0 Dichotomous Comorbid conditions Severe decompensated cirrhosis Yes/ No (1/0) Dichotomous Compensated cirrhosis Yes/ No (1/0) Dichotomous Chronic obstructive pulmonary diseases, COPD Yes/ No (1/0) Dichotomous Cerebral vascular disease, CVD Yes/ No (1/0) Dichotomous Drugs dependence Yes/ No (1/0) Dichotomous Diabetes Yes/ No (1/0) Dichotomous Depression Yes/ No (1/0) Dichotomous Heart diseases Yes/ No (1/0) Dichotomous Hyperthyroidism Yes/ No (1/0) Dichotomous Obesity Yes/ No (1/0) Dichotomous Psychiatric disorders Yes/ No (1/0) Dichotomous

PAGE 58

58 Table 3-2. Continued Variables Coding of covariates Data Comorbid conditions HIV co-infection Yes/ No (1/0) Dichotomous HBV co-infection Yes/ No (1/0) Dichotomous Use of health services Annual medical expenditure 3 $6700=1 <$6700=0 Dichotomous Hospitalization Yes/ No (1/0) Dichotomous Number of outpatient visits 0, 1, 2, 3, …. Continuous Number of Emergency visits 0, 1, 2, 3, …. Continuous Liver biopsy Yes/ No (1/0) Dichotomous Provider-level factor Number of Family/General practice visits 0, 1, 2, 3, …. Continuous Number of Gastroenterological visits 0, 1, 2, 3, …. Continuous Number of Internal Medicine visits 0, 1, 2, 3, …. Continuous Number of Infectious Disease visits 0, 1, 2, 3, …. Continuous Effectiveness outcome 4 Time to first event of ESLD in months 0, 1, 2, 3, … Continuous Costs outcome Direct health care costs in dollar amount 0, 1, 2, 3, … Continuous 1. Census Region: 1=New England and Middle Atlantic; 2=East North Central and West N orth Central; 3=South Atlantic, East South Central, and West South Central; 4=Mountain and Pacific; 5=National and Other in the dataset. 2. Health plan: Preferred Provider Organiza tion (PPO), Health Maintenance Organization (HMO) and Point of Service (POS). 3. Health car e costs and expenditure, including inpatient, outpatient, diagnostic, procedure costs, and prescript ions, were adjusted to 2007 U.S. dollars using the medical service component of the Consumer Price Inde x. The cut-off point was based on the “National Health Care Expenses in the U.S. Civilian Noninstitutionalized Population, 2003” results, 64 the annual medical expenditure per patients aged <65 years was approximately $5,670 in 2003 ($6,634 adjusted to 2007 dollars). 4. End-stage liver disease (ESLD), including occurrences of severe decompensated c irrhosis, HCC and liver transplantation, and a proxy of death defined as loss of health insurance eligibi lity 32 days after receiving hospice services.

PAGE 59

59 Table 3-3. Patterns of antiviral therapy Total treated 1 (n=3,896) Usual care 1 (n=882) Extended care 1 (n=119) Duration of treatment 2 mean (SD) month INF use 8.3 (7.9) 13.3 (6.0) 25.2 (17.8) RBV use 7.5 (6.2) 12.2 (5.1) 17.7 (11.9) RBV/INF duration ratio, % 98.0 (38.6) 94.2 (17.5) 79.7 (30.1) Number of prescription refills, mean (SD) INF refills, 7.4 (5.6) 12.7 (2.3) 24.3 (11.7) RBV refills, 6.5 (4.4) 11.1 (3.0) 15.3 (7.8) RBV/INF refills ratio, % 94.5 (41.5) 88.2 (20.0) 68.6 (30.8) Abbreviations: INF=interferon alpha, RBV=Ribavirin. 1. Total treated (treat ment group) includes those patients who ever received 1 INF and RBV claims. Using INF claims as treatment exposure indicator, usual care is defined as those patients receivi ng 11 to 14 continuous INF refills (indicating at least 12 months or 48 weeks of treatment wit hin a 15-month period). Extended care is defined as those patients receiving 15 continuous INF refills. 2. Total treatment duration was defined as the number of days between the last and firs t prescription claims plus 30 days.

PAGE 60

60 Figure 3-1. Sample selection process.

PAGE 61

61 Figure 3-2. Establishment of comparisons groups. Note: Diagnosis date = first date of HCV diagnosis, Index date = first date of interferon prescription. Figure 3-3. Study outcomes and follow-up. Note: Diagnosis date = first date of HCV di agnosis, Index date = first date of interferon prescription, ESLD (end-stage liver disease), HCC (hepatocellular carcinoma), LT (liver transplantation), severe hepa tic decompensation, including variceal bleeding, hepatic coma, etc.

PAGE 62

62 Figure 3-4. Plot of incremental net benefit (INB).

PAGE 63

63 CHAPTER 4 RESULTS Descriptive Characteristics Study Cohort A total of 163,917 MCO members were diagnosed with HCV infection between 1997 and 2007. Eliminations of individuals without 2 medical claims of HCV infection and without an INF claim, individuals without continuous eligibility, individuals receiving monother apy, individuals receiving antiviral therapy prior to the HCV diagnosis date, and thos e with alcoholism, organ transplantation, HCV re-infection after liver transplantation, H CC, or ages <18 or >64 years old reduced the study sample to 17,076 individuals (treated 3996, untreated 13,080). Among those selected in the study samples, non-treated patients were randomly selected with a 3:1 ratio to treated patients and assigned the pseudo-index date. Untreated patie nts were randomly selected based on 3996 treated patients, and those who became 65 years old by t he pseudo-index date (n=52), had either HCC or LT before the assigned index date (n=352) withdrew at the same date with index date (n=63), and had rare comorbid conditions (<1%) i n either the treatment or control group [i.e., end-stage renal disease/hemodialysi s (n=72), and hyperthyroidism (n=317)] were excluded from the analysis. A total of 15,071 patients satisfied all requirements for inclusion into the study anal ysis. (Figure 3-1) The base case study cohort consisted of 3,896 patients (treatment group) who ever received combination antiviral therapy and 11,175 patients (control group) who never re ceived treatment in the study period. Baseline Characteristics of All Study Patients in Base Case Analysis Table 4-1 reports baseline characteristics of study patients. Briefly, the newly-diagnosed HCV-infected patient cohort had a mean age of 47.5 (8.3) years and included more ma le than

PAGE 64

64 female patients (61.4%% vs. 38.6%, respectively). More than half the cohort resided in t he New England and Middle Atlantic regions of the US. Most patients (97.7%) had private insuranc e coverage at the time of HCV diagnosis. Among all patients, 37.0% enrolled in a HMO, fol lowing by 36.2% closed a PPO plan and 26.9% in POS plan. Considerable proportions of patients had existing cirrhosis, including 1.5% of patients with severe decompensated cirrhosi s and 18% with compensated cirrhosis. Patients had mental health–related conditions, including depr ession incurred after HCV diagnosis (4.8%), and psychiatric disorders (17.5%). Of the observed baseline characteristics of the study cohort, variables st atistically associated with initiation of antiviral therapy in base case cohort were show n in Table 4-2. Treated patients (treatment group) were more likely than untreated pati ents (control group) to be male (OR=1.25, 95% CI=1.15–1.36), and enrolled in a PPO plan (OR=1.14, CI=1.04–1.25). Having a condition associated with compensated cirrhosis (OR=1.68, 95% CI=1.52–1.86), depression (OR=1.85, 95% CI=1.57-2.1), receiving a higher annual medical expenditure (OR=1.94, 95% CI=1.76-2.15), having a biopsy procedure (OR=3.30, 95% CI=2.90–3.75), and often visiting a gastroenterologist (OR=1.12, 95% CI=1.10-1.14) were associated w ith an increased likelihood of initiating antiviral therapy (Table 4-2). Factors associated with patients receiving antiviral therapy initiat ion varied by the presence of cirrhosis in the base case cohort. Table 4-3 shows patients’ baseline charac teristics of two groups of patients with and without cirrhosis at baseline. Patients without cirrhosi s were more likely to have comorbid conditions than patients with cirrhosis, including diabetes (OR= 0.83, 95% CI=0.71–0.96), drugs dependence (OR=0.40, 95% CI=0.29–0.56, and HIV co-infection (OR=0.47, 95% CI=0.34–0.65).

PAGE 65

65 Baseline Characteristics of Patients in Subgroup Analysis Table 4-4 shows the distribution baseline characteristics of qualified patients in subgroup analysis. Among patients with usual care (n=1001) in subgroup analysis, 882 patient s received a 48-weeks treatment course (indicating 12to15 continuous months of antiviral treatme nt). Among patients with extend care (n=472), 119 patients had a longer than 48-weeks therapy (indicat ing greater than 15 continuous months of antiviral therapy). Among those randomly select ed untreated patients (n=2958), 2605 patients were matched to usual care patients and 353 untre ated patients to extended care patients. Compared to the patients in the base case analysis, the year of HCV diagnosi s was more likely to incur during 2000to2006 among patients receiving either usual care or extende d care. In the following analyses, therefore, were performed for those patients fir stly diagnosed with HCV infection during the period. Of subgroup patients, approximately 20% were diagnosed with cirrhosis, which was more than those cirrhotic patients (18%) involved in the base cas e analysis. Of the observed baseline characteristics of the subgroup patients, variables statistically associated with initiation of antiviral therapy among patients with usual ca re and extended care were shown in Table 4-5. A broader spectrum of comorbid conditions were negatively associ ated with treatment initiation among patients with usual care, including COPD, depr ession, HBV and HIV co-infection than those who received extended care (diabetes and psychiatr ic disorders). Estimated Effectiveness and Costs Study Follow-Up and Clinical Outcome Events On average, the observed follow-up period from the index date to first ESLD occurrence or censoring was 18.217.0 months, ranging from less than 1 month to a maximum of 88.3 months. ESLD developed in a 1.7% (n=67) of treated patients and 1.5% (n=166) of untreated patients. Cumulative incidences of outcome events between treated and untreated patients wi th or without

PAGE 66

66 cirrhosis during the follow-up period are shown in Table 4-11. Most common event of ESLD was decompensated cirrhosis (n=127), following by HCC (n=91) and LT (n=45). The rate of ESLD development among patients with cirrhosis was lower in treatment groups than control groups (3.7% vs. 5.4%, respectively). Among patients without cirrhosis, the rate of ES LD development was higher in treatment than control (1.0% vs. 0.8%, respectively). In those patients with usual care, the average observed follow-up period from the index date to first ESLD or censoring was 30.416.7 months (95% of subgroup patients remained at least 12 months in the study cohort). ESLD developed in a 2.1% (n=21) of treated patients and 1.4% (n=42) of untreated patients during observed follow-up in the Table 4-13. Patients wit h cirrhosis in usual care analysis had 1.4% of ESLD development rate in treatment group and 1.3% in control group. Among patients with extended care, the average observed follow-up per iod from the index date to first ESLD or censoring was 40.516.7 months. The rate of ESLD development was higher in treatment than control (7.6% vs. 2.6%, respectively). Figure 4-1 to 43 depicts the ESLD event rates between patients with and without cirrhosis in base c ase, usual care and extended care analysis. Note that a lower rate of ESLD progres sion in treatment than in untreated group among patients with cirrhosis existed in base case and usual ca re analyses. Unfortunately, among patients without cirrhosis, a higher ESLD event rate in t reatment than in untreated group was in base case, usual care and extended care analyses. Primary Effectiveness Measure: Time to ESLD Development Base case analysis. Study results of estimated time to first ESLD occurrence are shown in Table 4-6. The adjusted regression coefficient comparing the treatment to the control in the base case analysis was 3.32 (se=0.06, p<.0001) months, indicating that on the average time to ES LD occurred in treated patients delayed 3.32 months as compared with untreated patients, whi le controlling for treatment initiation-related covariates and instruments re lated with symptomatic

PAGE 67

67 conditions and maintenance. The effect of cirrhosis on the time to ESLD occurrence of treatment intervention was determined by examining the coefficient of interaction wit h treatment in the multivariate adjusted regression model, which was a difference of -0.33 (se=0.13, p= 0.01). There was a significant difference in the mean time to ESLD occurrence i n patients with cirrhosis [coefficient=3.01 (se=0.15, p<.0001)] and without cirrhosis [coefficient=3.4 4 (se=0.06, p<.0001)] in the stratified analyses. Table 4-7 shows the various baseline characteristic influences on the heterogen eity of treatment effect according to the time to ESLD occurrence. In patients with and without cirrhosis, a significant increase in the mean time to ESLD occurrence was related with the year of HCV diagnosis, the performance of liver biopsy, time to treatment initiation gre ater than 6 months. The average time to ESLD occurrence was shorter in male than female in pa tients with and without cirrhosis patients [coefficient=-1.16 (se=0.14), p<.0001, coefficient=-0.36 (se =0.05), p<.0001, respectively]. Patients with depression after HCV diagnosed was signific antly related with a shorten time of ESLD occurrence in cirrhotic and non-cirrhotic groups [coef ficient=-1.41 (se=0.28), p<.0001 vs. coefficient= -1.19 (se=0.08), p<.0001]. Surprisingly, HIV co-infection was related with a significant increase in the time to ESLD occurrence in p atients without cirrhosis [coefficient=1.31 (se=0.14), p<.0001]. A possible explanation is that patient s receiving antiviral therapy were more likely to be those who at well HIV disease condi tions (e.g. CD4 counts 350 cells/mm 3 ) comparing to HIV-co-infected patients without suppressed HIV and the pre-existing risk of rapid progression of liver fibrosis were not high as expecte d among treated patients. Interpretations and discussions for this study finding were shown in the dis cussions section.

PAGE 68

68 In addition, in patients with and without cirrhosis, a significant reduction in mean tim e to ESLD occurrence was related with prior health service utilization, including prior annual medical expenditure $6700 [coefficient=-1.75 (se=0.16), p<.0001 vs. coefficient= -1.08 (se=0.07), p<.0001], and hospitalization [coefficient= -2.01 (se=0.18), p<.0001 vs. coefficient= 1.45 (se=0.07), p<.0001], and gastroenterology visit [coefficient= -0.07 (se=0.01), p<.0001 vs. coefficient= -0.12 (se=0.01), p<.0001]. Subgroup analysis. The significant increase in the average time to ESLD between treatment and control was 1.33 months (se=0.09, p<.0001) in usual care analysis. The average time to ESLD event estimated no difference between treatment and control in the e xtended care analysis [coefficient=1.27 (se=1.20), p=0.29] in Table 4-8. Various baseline characteristic influences on the heterogeneity of t reatment effect according to the time to ESLD occurrence in subgroup analysis as shown in Table 4-9. In patients with usual care and extended care, a significant mean reduction in ti me to ESLD event was associated with cirrhosis [coefficient= -3.32 (se=0.10), p<.0001 vs. coefficie nt= -3.15 (se=1.186), p=<.01]. In patients with usual care, other comorbid conditions led to a reduction i n the time to ESLD occurrence included depression [coefficient= -2.74 (se=0.10), p<.001], HB V co-infection [coefficient= -1.93 (se=0.20), p<.0001]. Comorbid condition significantly reduc ed the time to ESLD occurrence in patients with extended care was diabetes [coeff icient= -5.44 (se=1.43), p<.0001]. HIV co-infection was related with a significant increase in the time to ESLD occurrence in patients with usual care [coefficient=3.95 (se=0.22), p<.0001]. In pa tients with usual care, prior health service utilization led to a significant reducti on in the time to ESLD occurrence included prior annual medical expenditure $6700 [coefficient= -1.94 (se=0.10), p<.0001], biopsy [coefficient= -1.28 (se=0.14), p<.0001], hospitalization [coefficient= -0.59

PAGE 69

69 (se=0.11), p<.0001], outpatient visits [coefficient= -0.02 (se=0.00), p<.0001], gastroenterology visit [coefficient= -0.20 (se=0.01), p<.0001] and infectious disease visits [coefficient= 0.03 (se=0.01), p<0.01]. In patients with extended care, biopsy and antiviral therapy init iation 6 months after the diagnosed date led to a significant reduction in the time to ESLD deve lopment [coefficient= -5.90 (se=2.00), p<0.01, and coefficient= -3.05 (se=1.02), p<0.01, respectively] Similarly, the effect of cirrhosis on the time to ESLD occurrence of trea tment intervention was determined by examining the coefficient of interaction with treatme nt in the multivariate adjusted regression model, a significant difference of -0.59 (se=0.20, p< 0.01) was seen in patients with usual care and a difference of 3.12 (se=2.48, p=0.21) in patie nts with extended care. Among patients with usual care, there was a significant increase, with 0.92 months (se=0.32, p<0.001) in patients with cirrhosis and 1.56 months (se=0.07, p<.0001) in patients without cirrhosis in Table 4-10. Results of stratified analyses among patients wit h usual care, a similar, significantly delayed the ESLD progression in treated patient s as comparing to untreated patients, were consistent with patients in the base case analysis. Table 4-11 shows the influence of various patients’ characteristics on the treatm ent effect among patients with and without cirrhosis in usual care analysis. Among patient s with cirrhosis, patients’ characteristics related with a significant reduction in the tim e to ESLD occurrence in base case analysis were similar with the factors among patients with usua l care. Table 4-7 and Table 4-11 show that age, time to treatment initiation and antidepressant use among ci rrhotic patients were related with the time to ESLD development in the base case ana lysis, but not significantly relevant with patients in the usual care analysis. Among patie nts without cirrhosis, a broader spectrum of comorbid conditions was associated with a similar, signific ant decrease in the time to ESLD development as compared with patients without cirrhosis. Patient s’

PAGE 70

70 characteristics associated with heterogeneous treatment effect i n non-cirrhotic patients in the base case analysis were most likely to be seen in the patients with non-cirr hosis in the usual care analysis; despite the drugs dependences and antidepressant use were associat ed with noncirrhotic patients in the base case analysis and COPD was associated wi th those patients in the usual care analysis. Secondary Effectiveness Measure: Rate of ESLD Development Base case analysis. Another effectiveness measure, the Kaplan-Meier survival estimate for ESLD-related events (HCC, liver transplantation, severe decompensated cir rhosis, or proxied death) after a start of treatment initiation was shown in Table 4-12. The KM est imate of ESLD progression rate (hazard ratio) comparing treated patients to untreated was 0.28 (95% CI, 0.20– 0.40), controlling for treatment initiation related covariates, instruments rel ated with symptomatic condition and treatment maintenance. The adjusted hazard ratio compar ing treatment to no treatment revealed the beneficial effect among patients who received combination antiviral therapy were approximately 30% as likely to have an ESLD as patients without treatment at any point in time after a start of treatment (at 80.3 mont hs). The adjusted hazard ratio among patients with cirrhosis was 0.32 (95% CI, 0.21–0.50) for treated patients relative to untreated patients, and among patients without ci rrhosis was 0.23 (95% CI, 0.13–0.40) for treated patients relative to untreated patients controlling for the potential confounders. Table 4-13 shows the various baseline characteristics influences on the heterogeneous treatment effect according to the adjusted hazard ratio in the Cox regression model. Among patients with and without cirrhosis, older and male patients were ass ociated with a risk of ESLD progression, despite that the greater risk for male patients wa s not significant (hazard ratio=1.42, 95% CI, 0.93–2.15) in patients without cirrhosis. Additionally, patients wi th HBV co-infection had a higher risk of ESLD progression than patients without HBV co-infection;

PAGE 71

71 the hazard ratio was 2.91 (95% CI, 1.57–5.41) in patients with cirrhosis and hazard ratio was 2.55 (95% CI, 1.02–6.40) in patients without cirrhosis. In patients without cirrhosis, diabetes was related with a high risk of ESLD progression (hazard ratio=4.71, 95% CI, 2.93–7.55). Pati ents had undergone liver biopsy and initiated therapy <6 months had a lower risk of ESLD progression, although the lower risk for treatment initiation within 6 months was not significant (hazard ratio=0.67 (95% CI, 0.41–1.09) among non-cirrhotic patients. Subgroup analysis. Table 4-14 shows the results of secondary effectiveness in subgroup patient cohort. Among patients with usual care, the ESLD developed in 1.4% (n=12) of trea ted patients and 1.3% (n=33) of untreated patients. The KM estimate of ESLD progressi on (hazard ratio) revealed no difference between treated patients and untreated patients (hazard ratio=0.61, 95% CI, 0.28–1.35) controlling for treatment initiation related covariates and instrume nts related with symptomatic condition. Among patients with extended care, the ESLD developed in 7.6% (n=9) of treated patients and 2.6% (n=9) of untreated patients. The adjusted hazard rat io estimate showed no difference in ESLD development between treated and untreated patients (ha zard ratio=1.79, 95% CI, 0.58–5.54). Table 4-15 shows the various baseline characteristic influences on the heterogen eity of treatment effect according to the Kaplan-Meier estimate in the Cox reg ression model. In patients with usual care, a higher risk of ESLD progression was associated with ma le (hazard ratio=2.80, 95% CI,1.38–5.70), cirrhosis (hazard ratio=10.72, 95% CI, 5.23–21.91) and decompensated cirrhosis (hazard ratio=5.54, 95% CI, 1.80–17.08), HBV co-infection (hazard rat io=3.95, 95% CI, 1.14–13.60) and initiating treatment within 6 months (hazard ratio=2.19, 95% CI, 1.07–4.49). Among patients receiving extended care, the presence of cirrhosis (hazard r atio=18.24, 95% CI,

PAGE 72

72 4.25–78.25), decompensated cirrhosis (hazard ratio=13.88, 95% CI, 3.05–63.08), and diabetes (hazard ratio=7.58 (95CI, 2.01–28.59) were associated with a risk of ESLD development. Additionally, the stratification on the presence of cirrhosis was performed in pa tients with usual care and extended care as shown in Table 4-14. Hazard ratios for the interact ion term between treatment and prior cirrhosis revealed a significant reduction am ong patients in usual care analysis (hazard ratio=0.17, 95% CI, 0.04–0.81), and no difference of 1.92 (95% CI, 0.62– 5.96) in patients with extended care. As shown in Table 4-16, cirrhotic patients with usual c are having a significant estimated 24% (hazard ratio=0.24, 95% CI, 0.06-0.94) reduction in the hazard for ESLD development, while no significant reduction on the risk of ESLD devel opment among non-cirrhotic patients with usual care (hazard ratio=0,88, 95% CI, 0.27–2.87). Summary of Effectiveness Results In the prior sections presents the effectiveness results individually obtaine d in the base case, usual care and extended care analyses. Table 4-10 shows a summary of resul ts obtained by using the primary effectiveness estimate (mean time to ESLD event) for the effect of antiviral therapy in base case and usual care analyses. Likewise, Table 4-16 summar izes the secondary effectiveness estimate, hazard ratios for the treatment effect in base case and usual care analyses. Both estimates of treatment effect revealed a significant benefici al effect on the prevention of ESLD event among patients with cirrhosis. Although, among patients without cirr hosis in the usual care analysis did not reach statistical difference in hazard of ESLD event between treated and untreated patients (hazard ratio=0.88, 95% CI, 0.27-2.87), the trend for the beneficial eff ect using primary effectiveness estimate versus hazard ratios were consi stent (coefficient=1.56, se=0.07, p <.0001). Effectiveness results among patients with extended care, the two estim ates in Table 4-8 and Table 4-14 for the treatment effect revealed consistent no diff erence in ESLD

PAGE 73

73 progression between treatment and control among a number of different fitted mode ls with various important baseline characteristics. Data from randomized, controlled trials and current treatment recommendations on H CV management suggested that the cost-effectiveness of initial antiviral the rapy for patients with cirrhosis has important interpretations to the decision makers in a MCO health deli very system. First, although individuals with advanced fibrosis typically responded less well to antiviral therapy in terms of SVR rate, 23, 24 additional benefits of antiviral therapy were seen among patients with advanced fibrosis and cirrhosis on more prolonged and potent viral suppressi on activity. 23, 79, 80 The histological response may lead to a reduction in the rate of ESLD progression. In practice, this group of patients had a greatest need for effective therapy in order to avoid HCC development and liver transplant. Second, several recent studies involving in responses rates have established that the low likelihood of achieving SVR with fur ther antiviral therapy for HCV genotype 1 infected patients who lack of virological respons e after 24 weeks of combination therapy. More recently, discontinuation of antiviral therapy for pat ients remaining viral positive with qualitative PCR testing after 12-weeks of initial ther apy has been suggested as a means of reducing antiviral treatment related adverse effects and costs 23, 81 As the treatment was more likely initiated to HCV-infected patients wit h cirrhosis in this patient cohort, the question is raised as to whether the continuous 12 months (48 weeks) antivira l therapy is more cost-effective than a shorter duration of treatment in cirrhoti c patients, assuming that treatment discontinuation was mainly caused by the lack of early virologi cal response at the end of 6months therapy. Of note, the average duration of antiviral therapy was 7.5-8.3 months in all treated patients receiving at least one claim of combination therapy i n Table 3-3. To determine the net benefit of initial antiviral therapy for patients wit h cirrhosis from a MCO

PAGE 74

74 perspective, the following total cost and cost-effectiveness analysis wer e shown for patients with cirrhosis in the base case and usual care analyses. Total Cost Total direct medical costs during the study period were assessed for patie nts with cirrhosis in base case and subgroup analyses. As described in chapter 3.6, the total c osts were obtained by the sum of weighted monthly costs in the follow-up by inverse probability of censoring. All baseline characteristics in the primary effectivenes s estimation model were included in the inverse probability of censoring weighting procedure. Table 4-18 summa rizes the estimated total cost comparisons of patients with cirrhosis in base case and usua l care analyses. As expected, the mean difference in total costs between the treatment and control group was higher for patients with cirrhosis in the usual care (coefficient=$32,953, se=5642.31, p<.0001) than those patients in base case analysis (coefficient= $25,722, se=2365.28, p<.0001). All measured baseline characteristics and created instruments were a djusted in the multivariate regression model to examine its influence on total cost differen ce between treatment and control, except for the variable of prior annual medical expenditure. The various ba seline characteristics impact on the mean total costs difference among patient s with cirrhosis is shown in Table 4-19. Social-demographic characteristics, including census regions, PO S versus HMO health plan played a statistically significant role in the mean total cost difference between treatment and control in the base case analysis, despite POS versus HMO was ins ignificantly associated with total cost in the usual care analysis. The role of patient’s co -morbidity conditions and prior health care services utilization in the total cost were varie d between base case and usual care analyses. For example, patients with drug dependence eme rgency room visits were associated with a reduction in total cost during the initial treatment p eriod in the usual care analysis [coefficient=$86,435, se=19783, p<.0001, and coefficient=$5796, se=19783, p<.0001,

PAGE 75

75 respectively], yet diabetes and psychiatric disorders were significantl y associated with the difference in the mean total cost in the base case analysis [coefficient= $7025, se=2969, p=0.02, and coefficient=-$6670, se=2924, p=0.02, respectively]. Cost-Effectiveness of Initial Combination Antiviral Therapy Net Benefit Regression Model Results of the net benefit regression were shown at selected values of WTP (l=$0; $10,000; $20,000; and up to $80,000). Table 4-20 shows the mean net benefit difference (i.e., INB) between treatment and control for patients with cirrhosis in base case and usual c are analyses. On average, the net benefit of treatment was higher among patients involving in the ba se case analysis compared to patients with the usual care at the same value of WTP (Ta ble 4-20). In the base case analysis, the hypothesis that antiviral therapy is cost-effec tive was found in the multivariate adjusted regression model where minimum value of WTP at $15,000 (INB=11471, 95% CI=16822-22173, one-sided test with p<0.001). In patients with usual care, however, the hypothesis that antiviral therapy is cost-effective was found in the multivar iate adjusted regression model where minimum value of WTP at $60,000 (INB=24553, 95% CI= -3819-52925, one-sided test with p=0.04). Given the values of WTP, the strength of statistical e vidence about treatment efficiency (i.e., INB estimate) supported by 95% CI are shown in regression models in Figure 4-2. Plots for the adjusted estimates of INB and its 95% CI by WTP depict the evidenc e that the antiviral therapy has a larger effect of net benefit for patients in the ba se case analysis than those who with usual care in Figure 4-2. The 95% lower limit of INB for patients with cir rhosis in the usual care analysis crosses the x-axis at WTP $60,000 indicating that only a WTP as high as $60,000 would lead us to reject the null hypothesis ( 0) in favor of treatment, INB ( )>0 at

PAGE 76

76 the 5% level significance. In addition, the greater variance of INB also inc reased as WTP became higher in patients with usual care compared to those patients in the base ca se cohort. Appendix A and B show the adjusted net benefit difference for patients in base case and usual care at selected WTP values. Significance and Effect of Covariates Significances and effects of risk factors and/or covariates on the INB e stimate for the antiviral therapy intervention including factors that were either signif icantly relevant with primary effectiveness estimate or difference in average total cost s are summarized in Table 4-21 and Table 4-22. The covariates effect on the INB between treatment and control w ere examined by the interaction terms with treatment in a multivariate adjusted model a t various selected values of WTP (Appendix C and D). As Table 4-21 and Appendix C show, among patients in base case analysis, year of HCV diagnosis during 2002 and 2006, severe decompensated cirrhosis, diabetes, HIV co-infec tion, and antidepressant use were potential important covariates on incremental costeffectiveness of treatment, although not all of interaction terms across the regression model a t WTP=$15,000; $20,000; and $30,000 were consistently significant at the 5% level. Among patients with usual care, year of HCV diagnosis at 2001 and 2002 relative to 2006 were potential important on a decrease in the cost-effectiveness of treatment, although not all of the inter action terms across the regression model at WTP=$60,000; $70,000; and $8,000 were consistently significant at the 5% level, as in Table 4-22 and Appendix D. Lower treatment efficiency in base case cohort was associated with dec ompensated cirrhosis, diabetes and HIV co-infection among cirrhotic patients. The primary effectiveness estimate for the initial antiviral therapy was lower in cirrhotic pati ents having decompensated cirrhosis and diabetes compared to their counterparts (coefficient=-7.86, se=0.24, p<.0001, and

PAGE 77

77 coefficient=-2.66, se=0.18, p<.0001, respectively). On the other hand, although the presence of HIV co-infection did not have a significant influence on the treatment effective ness, total costs was substantially higher than those patients without HCV-HIV co-i nfection (coefficient=$40,509, se=7386.52, p<.0001). Visual inspection for the assumptions of error term normality confirmed the heter ogeneity of treatment effect in the group of patients with cirrhosis between base case and usual care analyses. Figure 4-3 depicts an example of standardized normal probability (PP) plot of net benefit at the value of WTP=$15,000 for patients in base case analysis. The P-P plots showed that patients in the base case analysis were less sensitive to deviation from nor mality in the middle range of data than those patients in the usual care analysis as shown in Figu re 4-4 at the value of WTP=$60,000. Possible explanations could be related with greater diversity o f study samples in the usual care analysis than those in the base case analysis that di dn’t receive antiviral therapy but had good outcomes and relatively low costs.

PAGE 78

78 Table 41. Baseline characteristics of all patients between treatme nt and control groups in base case analysis Variable All patients (15071) Treatment (3896) Control (11175) Social demographics Age at HCV diagnosis, years Mean ( D SD) 47.5 (8.3) 47.5 (7.5) 47.5 (8.5) Gender, % Female 38.6 35.5 39.7 Male 61.4 64.6 60.3 Census region, 1 % 1 52.5 54.3 51.9 2 8.3 8.7 8.2 3 26.1 26.0 26.1 4 9.3 7.7 9.9 5 3.8 3.4 3.9 Insurance, % Public 2.3 1.5 2.5 Private 97.7 98.5 97.5 Health Plan, 2 % HMO 37.0 35.8 37.4 PPO 36.2 38.7 35.3 POS 26.9 25.6 27.3 Year of HCV diagnosis, % 1998 1.3 0 1.7 1999 1.3 0 1.7 2000 7.0 4.9 7.8 2001 8.8 9.8 8.5 2002 14.2 17.2 13.2 2003 10.0 11.9 9.4 2004 12.4 13.6 11.9 2005 24.4 24.8 24.3 2006 16.7 14.9 17.3 2007 3.9 3.0 4.3 1. Census Region: 1=New England and Middle Atlantic; 2=East North Central and West North Central; 3=South Atlantic, East South Central, and West South Central; 4=Mountai n and Pacific; 5=National and Other in the dataset. 2. Health plan: Preferred Provider Organization (PPO), Health Maintenance Organization (HMO) and Point of Servi ce (POS).

PAGE 79

79 Table 4-1. Continued Variable All patients (n=15071) Treatment (n=3896) Control (n=11175) Prior comorbid conditions,% Severe decompensated cirrhosis 1.5 1.1 1.6 Cirrhosis 18.0 25.4 15.4 Chronic obstructive pulmonary diseases (COPD) 12.7 11.5 13.2 Cerebral vascular disease (CVD) 2.8 2.4 2.9 Depression after HCV diagnosis 4.8 7.4 13.8 Diabetes 12.3 11.2 12.7 Drug dependence 3.7 1.8 4.4 HBV co-infection 3.6 2.7 3.9 Heart diseases 9.5 8.9 9.8 HIV co-infection 3.2 1.9 3.6 Obesity 5.0 4.9 5.0 Psychiatric disorders 17.5 16.2 18.0 Prior medical services use Annual medical expenditure, % $6700 66.6 63.1 67.8 n $6700 33.4 36.9 32.2 Liver biopsy, % 8.1 17.5 4.9 Hospitalization, % 20.9 13.8 23.4 Emergency room, visits Mean D (SD) 1.2 (3.3) 1.1 (3.2) 1.3 (3.3) Outpatient, visits Mean D (SD) 21.6 (26.0) 21.9 (23.6) 21.6 (26.8) Family/general practice, visits Mean D (SD) 3.5 (6.4) 3.4 (6.1) 3.5 (6.6) Gastroenterology, visits Mean D (SD) 2.1 (3.6) 3.0 (3.2) 1.8 (3.6) Infectious disease, visits Mean D (SD) 0.3 (2.9) 0.2 (1.0) 0.4 (3.3) Internal Medicine, visits Mean D (SD) 5.3 (9.1) 5.2 (8.0) 5.3 (9.5) Instruments related with symptomatic conditions and maintenance Time to treatment,% 6 months 64.2 64.3 63.7 n 6 months 35.8 35.7 36.3 Antidepressant use,% 15.9 29.3 11.3

PAGE 80

80 Table 4-2. Factors associated with initiation of combination antiviral therapy among patients in base case analysis Variable Adjusted OR 1 (95% CI) Social demographics Age at HCV diagnosis 0.99 (0.99–1.00) Gender Male vs. Female 1.25(1.15–1.36) Health Plan 2 POS vs. HMO 1.00 (0.90-1.11) PPO vs. HMO 1.14 (1.04–1.25) Year of HCV diagnosis 1998 vs. 2007 0 1999 vs. 2007 0 2000 vs. 2007 0.68 (0.52–0.89) 2001 vs. 2007 1.45 (1.13–1.86) 2002 vs. 2007 1.65 (1.31–2.10) 2003 vs. 2007 1.35 (1.06–1.73) 2004 vs. 2007 1.41 (1.11–1.79) 2005 vs. 2007 1.25 (1.00–1.57) 2006 vs. 2007 1.14 (0.90–1.44) Prior comorbid conditions Severe decompensated cirrhosis 0.32 (0.22–0.48) Cirrhosis 1.68 (1.52–1.86) Chronic obstructive pulmonary diseases (COPD) 0.88 (0.78–0.99) Depression 1.85 (1.57–2.19) Diabetes 0.81 (0.72–0.92) Drug dependence 0.50 (0.38–0.65) HBV co-infection 0.53 (0.42–0.67) HIV co-infection 0.53 (0.40–0.70) Prior medical services use Annual medical expenditure $6700 vs. <$6700 1.94 (1.76–2.15) Liver biopsy 3.30 (2.90–3.75) Hospitalization 0.38 (0.34–0.43) Outpatient visits 0.996 (0.994–0.998) Emergency room visits 0.985 (0.971–0.999) Gastroenterologist visits 1.12 (1.10–1.14) Infectious disease visits 0.95 (0.92–0.98) 1. Adjusted OR was obtained by multivariate logistic regression analysis w ith stepwise selection procedure, including 27 baseline characteristics. 2. Heath plan: PPO=Pre ferred Provider Organization, non-PPO= Health Maintenance Organization (HMO), Point of Service (POS), and Other labeled in the dataset

PAGE 81

81 Table 4-3. Factors associated with initiation of combination antiviral therapy among patients with or without cirrhosis involving in base case analysis Variable 1 Cirrhosis Non-cirrhosis Social demographics Age at HCV diagnosis NS 3 0.99 (0.98–0.99) Gender Male vs. Female 1.28 (1.07–1.52) 1.23 (1.12–1.35) Health Plan 2 POS vs. HMO 1.15 (0.92–1.44) NS PPO vs. HMO 1.32 (1.09–1.61) NS Year of HCV diagnosis 1998 vs. 2007 0 0 1999 vs. 2007 0 0 2000 vs. 2007 1.55 (0.76–3.15) 0.56 (0.41–0.77) 2001 vs. 2007 2.44 (1.21–4.91) 1.36 (1.04–1.78) 2002 vs. 2007 2.89 (1.46-5.72) 1.60 (1.25–2.06) 2003 vs. 2007 2.93 (1.46–5.88) 1.20 (0.92–1.56) 2004 vs. 2007 2.35 (1.17–4.69) 1.34 (1.04–1.74) 2005 vs. 2007 2.12 (1.08–4.13) 1.17 (0.92–1.50) 2006 vs. 2007 2.11 (1.07–4.18) 1.02 (0.79–1.31) Prior comorbid conditions Severe decompensated cirrhosis 0.46 (0.31–0.67) -Depression 2.23 (1.60–3.12) 1.71 (1.41-2.08) Diabetes NS 0.83 (0.71–0.96) Drugs dependence NS 0.40 (0.29–0.56) HBV co-infection 0.36 (0.24–0.55) 0.62 (0.47–0.82) HIV co-infection NS 0.47 (0.34–0.65) Prior medical services use Annual medical expenditure $6700 vs. <$6700 1.58 (1.29–1.93) 2.04 (1.82–2.89) Liver biopsy 2.61 (2.03–3.36) 3.37 (2.90–3.91) Hospitalization 0.38 (0.30–0.48) 0.39 (0.34–0.45) Outpatient visits 0.994 (0.991–0.998) 0.996 (0.993–0.998) Emergency room visits NS 0.97 (0.96–0.99) Gastroenterologist visits 1.04 (1.02–1.06) 1.20 (1.18–1.23) Infectious disease visits 0.94 (0.89–1.00) 0.94 (0.92–0.96) 1. Adjusted OR (odds ratio) was obtained by multivariate logistic regression ana lysis with stepwise selection procedure, including 26 baseline characteristics in pat ients with cirrhosis, and 25 baseline characteristics in patients without cirrhosis. 2. Health plan: P referred Provider Organization (PPO), non-PPO=Health Maintenance Organization (HMO) and Point of Service (POS). 3. NS (not selected), indicating the variable failed to reach statisti cal significance in the multivariate logistic regression analysis with stepwise selection.

PAGE 82

82 Table 4-4. Baseline characteristics of subgroup patients with usual care or extended care Variable 1 Usual care Extended care Treatment (n=882) Control (n=2605) Treatment (n=119) Control (n=353) Social demographics Age at HCV diagnosis, years Mean ( D SD) 47.2 (7.3) 47.8 (8.2) 48.5 (6.5) 47.6 (8.1) Gender % Female 33.8 42.2 27.7 40.8 Male 66.2 57.8 72.3 59.2 Census region 1 % 1 64.9 62.7 69.8 64.9 2 6.8 7.1 3.4 4.0 3 20.0 19.3 17.7 17.3 4 4.7 7.2 4.2 6.0 5 3.7 3.8 5.0 7.9 Insurance, % Public 0.7 2.7 1.7 0.9 Private 99.3 97.3 98.3 99.1 Health Plan 2 % HMO 37.3 39.5 31.9 37.1 PPO 40.8 37.1 47.1 41.1 POS 21.9 23.4 21.0 21.8 Year of HCV diagnosis,% 1998 0.0 0.8 0.0 0.9 1999 0.0 0.0 0.0 0.0 2000 6.5 10.2 21.0 17.6 2001 14.5 11.1 21.0 18.4 2002 17.5 15.8 14.3 20.7 2003 14.5 13.7 12.6 11.3 2004 14.7 16.0 7.6 12.8 2005 25.7 27.1 21.0 16.2 2006 6.6 5.2 2.5 2.3 2007 0.0 0.0 0.0 0.0 1. Census Region: 1=New England and Middle Atlantic; 2=East North Central and West North Central; 3=South Atlantic, East South Central, and West South Central; 4=Mountai n and Pacific; 5=National and Other in the dataset. 2. Health plan: Preferred Provider Organization (PPO), Health Maintenance Organization (HMO) and Point of Servi ce (POS).

PAGE 83

83 Table 4-4. Continued Variable 1 Usual care Extended care Treatment (n=882) Control (n=2605) Treatment (n=119) Control (n=353) Prior comorbid conditions,% Severe decompensated cirrhosis 0.3 0.7 2.5 2.0 Cirrhosis 29.0 16.7 37.0 21.0 Chronic obstructive pulmonary diseases (COPD) 10.2 13.9 16.0 11.6 Cerebral vascular disease (CVD) 1.6 3.0 3.4 1.4 Depression after HCV diagnosis 8.1 3.4 5.0 2.6 Diabetes 10.3 13.1 9.2 15.9 Drug dependence 1.7 3.5 0.8 1.7 HBV co-infection 2.5 4.0 0.8 3.1 Heart diseases 9.0 10.5 10.1 8.2 HIV co-infection 2.0 3.6 3.4 2.8 Obesity 5.0 5.2 1.7 3.4 Psychiatric disorders 15.1 17.9 7.6 15.6 Prior medical services use Annual medical expenditure, % <$6700 66.7 67.3 56.3 71.7 n $6700 33.3 32.7 43.7 28.3 Liver biopsy, % 18.3 5.2 13.5 4.0 Hospitalization, % 11.7 22.4 13.5 17.3 Emergency room, visits Mean D (SD) 1.1 (2.3) 0.8 (1.9) 0.9 (2.1) 0.9 (1.9) Outpatient, visits Mean D (SD) 22.7 (26.4) 22.5 (22.5) 19.9 (13.4) 21.3 (28.8) Family/general practice, visits Mean D (SD) 3.3 (6.2) 3.4 (5.8) 3.6 (6.1) 2.8 (6.2) Gastroenterology, visits Mean D (SD) 2.1 (3.6) 3.2 (3.7) 3.0 (2.5) 1.8 (3.5) Infectious disease, visits Mean D (SD) 0.3 (3.4) 0.2 (1.1) 0.2 (1.2) 0.2 (0.9) Internal Medicine, visits Mean D (SD) 5.6 (9.1) 5.5 (8.0) 5.1 (6.2) 5.3 (7.4) Instruments related with symptomatic conditions and maintenance Time to treatment,% 6 months 60.7 61.1 67.2 67.7 6 months 39.3 38.9 32.8 32.3 Antidepressant use 33.5 14.7 40.3 17.3

PAGE 84

84 Table 4-5. Factors associated with initiation of combination antiviral therapy among patients with usual care or extended care Variable 1 Usual care Extended care Social demographics Age at HCV diagnosis 0. 98 (0.97–0.99) NS Gender Male vs. Female 1.52 (1.27–1.82) 1.80 (1.11–2.92) Private insurance 3.05 (1.19–7.82) NS Health Plan 2 POS vs. HMO 0.87 (0.69–1.10) NS PPO vs. HMO 1.21 (1.00–1.46) NS Year of HCV diagnosis 2000 vs. 2006 0.33 (0.21–0.53) NS 2001 vs. 2006 0.87 (0.58–1.30) NS 2002 vs. 2006 0.69 (0.47–1.03) NS 2003 vs. 2006 0.55 (0.37–0.83) NS 2004 vs. 2006 0.56 (0.37–0.83) NS 2005 vs. 2006 0.68 (0.47–0.98) NS Prior comorbid conditions Severe decompensated cirrhosis 0.12 (0.02–0.58) NS Depression 1.65 (1.60–2.01) 2.03 (1.25-3.30) Chronic obstructive pulmonary diseases (COPD) 0.71 (0.55–0.94) NS Depression 2.39 (1.67–3.42) NS Diabetes NS 0.42 (0.20–0.87) HBV co-infection 0.37 (0.22–0.63) NS HIV co-infection 0.46 (0.26–0.83) NS Psychiatric disorders NS 0.40 (0.18–0.87) Prior medical services use Annual medical expenditure $6700 vs. <$6700 1.78 (1.43–2.20) 2.47 (1.46–4.16) Liver biopsy 3.50 (2.66–4.60) 3.40 (1.54–7.54) Hospitalization 0.34 (0.26–0.46) 0.45 (0.22–0.92) Outpatient visits 0.996 (0.992–0.999) NS Emergency room visits 0.92 (0.88–0.97) NS Gastroenterologist visits 1.17 (1.31–1.20) NS Infectious disease visits 0.96 (0.93–0.99) NS 1. Adjusted OR (odds ratio) was obtained by multivariate logistic regression ana lysis with stepwise selection procedure, including 26 baseline characteristics in pat ients with cirrhosis, and 25 baseline characteristics in patients without cirrhosis. 2. Health plan: P referred Provider Organization (PPO), non-PPO=Health Maintenance Organization (HMO) and Point of Service (POS). 3. NS (not selected), indicating the variable failed to reach statisti cal significance in the multivariate logistic regression analysis with stepwise selection.

PAGE 85

85 Table 4-6. Primary effectiveness results in base case analysis: Time t o ESLD development in patients with and without cirrhosis at baseline Time to ESLD outcome All patients Cirrhosis Non-cirrhosis Treatment (n=3896) Control (n=11175) Treatment (n=991) Control (n=1721) Treatment (n=2905) Control (n=9454) Mean (SD) months 20.1 (8.1) 15.7 (9.0) 19.4 (8.0) 14.3 (8.8) 20.3 (8.2) 16.0 (9.0) Adjusted mean difference (se) [p value] 1 3.32 (0.06) [<.0001] 2 3.01 (0.15) [<.0001] 3.44 (0.06) [<.0001] 1. Various baseline characteristics included in the multivariate adjusted re gression models to examine the heterogeneity of treatment on time to ESLD were listed in Ta ble 4-5. 2. The coefficient of interaction term between cirrhosis and treatment interventi on was -0.33 (se=0.13, p=0.01).

PAGE 86

86 Table 4-7. Factors associated with primary effectiveness of antiviral the rapy among patients with and without cirrhosis in base case analysis Variable 1 All patients Cirrhosis Non-cirrhosis Social demographics Age at HCV diagnosis 0.04 (0.00) [<.0001] -0.01 (0.01) [ 0.30] 0.05 (0.00) [<.0001] Gender -0.49 (0.05) [<.0001] -1.16 (0.14) [<.0001] -0.36 (0.05) [<.0001] PPO health Plan POS vs. HMO -0.06 (0.06) [0.32] 0.22 (0.16) [0.16] -PPO vs. HMO 1.08 (0.06) [<.0001] 0.16 (0.18) [0.36] -Year of HCV diagnosis 1998 vs. 2007 8.28 (0.24) [<.0001] 7.82 (0.67) [<.0001] 8.05 (0.26) [<.0001] 1999 vs. 2007 2.79 (0.24) [<.0001] 2.27 (0.77) [<0.01] 2.84 (0.25) [<.0001] 2000 vs. 2007 29.86 (0.15) [<.0001] 25.17 (0.49) [<.0001] 30.45 (0.16) [<.0001] 2001 vs. 2007 27.01 (0.14) [<.0001] 23.42 (0.49) [<.0001] 27.22 (0.15) [<.0001] 2002 vs. 2007 22.16 (0.14) [<.0001] 19.01 (0.47) [<.0001] 22.42 (0.14) [<.0001] 2003 vs. 2007 22.07 (0.14) [<.0001] 19.48 (0.48) [<.0001] 23.27 (0.15) [<.0001] 2004 vs. 2007 17.80 (0.13) [<.0001] 14.21 (0.47) [<.0001] 17.96 (0.13) [<.0001] 2005 vs. 2007 12.37 (0.13) [<.0001] 9.98 (0.45) [<.0001] 12.58 (0.13) [<.0001] 2006 vs. 2007 5.57 (0.13) [<.0001] 4.31 (0.46) [<.0001] 5.68 (0.13) [<.0001] Prior comorbid conditions Severe decompensated cirrhosis -7.93 (0.20) [<.0001] -8.00 (0.26) [<.0001] -Cirrhosis -1.26 (0.06) [<.0001] --Depression -1.52 (0.11) [<.0001] -1.41 (0.28) [<.0001] -1.19 (0.08) [<.0001] Drugs dependence -1.32 (0.13) [<.0001] --1.31 (0.13) [<.0001]

PAGE 87

87 Table 4-7. Continued Variable 1 All patients Cirrhosis Non-cirrhosis Prior comorbid conditions Diabetes -1.54 (0.07) [<.0001] --1.19 (0.08) [<.0001] HBV co-infection -0.94 (0.13) [<.0001] -1.79 (0.79) [<.0001] -0.51 (0.14) [<.0001] HIV co-infection 1.41(0.14) [<.0001] -1.31 (0.14) [<.0001] Prior medical services use Annual medical expenditure -1.24 (0.06) [<.0001] -1.75 (0.16) [<.0001] -1.08 (0.07) [<.0001] Liver biopsy 1.46 (0.09) [<.0001] 2.19 (0.21) [<.0001] 1.29 (1.00) [<.0001] Hospitalization -1.46 (0.07) [<.0001] -2.01 (0.18) [<.0001] -1.45 (0.07) [<.0001] Outpatients visits 0.01 (0.00) [<.0001] 0.01 (0.00) [<.0001] 0.02 (0.00) [<.0001] Emergency room visits -0.12 (0.01) [<.0001] --0.01 (0.01) [<.0001] Gastroenterology visit -0.09 ( 0.01) [<.0001] -0.07 (0.01) [<.0001] -0.12 (0.01) [<.0001] Infectious disease 0.01(0.01) [0.35] 0.02 (0.02) [0.42] -0.01 (0.01) [0.54] Instruments related with symptomatic conditions and maintenance Time to treatment 6 vs. <6 months 2.82 (0.05) [<.0001] 2.74 (0.14) [<.0001] 2.77 (0.05) [<.0001] Antidepressants use 0.20 (0.07) [<0.01] 0.71 (0.18) [<.001] 0.09 (0.07) [0.23] 1. Values shown in the tables are coefficients (mean (se), [p-value]) associa ted with treatment initiation relevant covariates and instruments related with sy mptomatic conditions and maintenance in the multivariate adjusted regression models. Treatment initi ation related covariates in base case analysis were shown in Table 4-2, Table 4-3 for patients w ith or without cirrhosis in base case analysis.

PAGE 88

88 Table 4-8. Primary effectiveness results in subgroup analysis: Time to ESL D development in patients with usual care and extended care Time to ESLD Outcome Usual care Extended care Treatment (n=882) Control (n=2605) Treatment (n=119) Control (n=353) Mean (SD) months 29.8 (10.6) 29.4 (11.0) 33.6 (11.3) 33.1 (11.2) Adjusted mean difference (se) [p value] 1 1.33 (0.09) [<.0001] 2 1.27 (1.20) [0.29] 3 1. Various baseline characteristics included in the multivariate adjusted re gression models to examine the heterogeneity of treatment on time to ESLD are listed in Tabl e 4-9. 2. The interaction between treatment and prior cirrhosis was -0.59 (0.20) [p<0.01]. 3. The interaction between treatment and prior cirrhosis was 3.12 (2.48) [p=0.21].

PAGE 89

89 Table 4-9. Factors associated with primary effectiveness among patients with usual care and extended care Variable 1 Usual care Extended care Social demographics Age at HCV diagnosis 0.05 (0.00) [<.0001] -0.00 (0.06) [0.98] Gender (Male vs. Female) -0.68 (0.08) [<.0001] 0.27 (1.00) [0.78] Private insurance 1.64(0.26) [<.0001] -Type of health Plan POS vs. HMO 0.06 (0.10) [0.52] -PPO vs. HMO 0.00 (0.08) [0.95] -Year of HCV diagnosis 2000 vs. 2006 34.15 (0.20) [<.0001] -2001 vs. 2006 28.20(0.19) [<.0001] -2002 vs. 2006 25.46 (0.19) [<.0001] -2003 vs. 2006 21.58 (0.19) [<.0001] -2004 vs. 2006 13.16 (0.18) [<.0001] -2005 vs. 2006 5.74 (0.17) [<.0001] -Prior comorbid conditions Severe decompensated cirrhosis -0.69 (0.48) [0.15] -8.02 (3.48) [0.02] Cirrhosis -3.32 (0.10) [<.0001] -3.15 (1.18) [<0.01] Chronic obstructive pulmonary diseases (COPD) 0.02 (0.11) [0.85] -Depression -2.74 (0.18) [<.001] -Diabetes --5.44 (1.43) [<0.001] HBV co-infection -1.93 (0.20) [<.0001] -HIV co-infection 3.95 (0.22) [<.0001] -Psychiatric disorders --2.37 (1.44) [0.10]

PAGE 90

90 Table 4-9. Continued Variable 1 Usual care Extended care Prior medical services use Annual medical expenditure -1.94 (0.10) [<.0001] -2.28 (1.21) [0.06] Biopsy -1.28 (0.14) [<.0001] -5.90 (2.00) [<0.01] Hospitalization -0.59 (0.11) [<.0001] -2.57 (1.50) [0.09] Outpatient visits -0.02 (0.00) [<.0001] -Emergency visits -0.01 (0.01) [0.44] -Gastroenterology visits -0.20 (0.01) [<.0001] -Infectious disease visits 0.03 (0.01) [<0.01] -Instruments related with symptomatic conditions and maintenance Time to treatment 6 vs. <6 months 0.00 (0.08) [1.00] -3.05 (1.02) [<0.01] Antidepressants use 0.06 (0.10) [0.54] -0.88 (1.22) [0.47] 1. Values shown in the tables are coefficients (mean (se), [p-value]) associa ted with treatment initiation relevant covariates and instruments related with sy mptomatic conditions and maintenance (as shown in Table 4-5) in the multivariate adjusted regression models

PAGE 91

91 Table 4-10. Summary of primary effectiveness results in base case and usual c are analyses among patients with or without cirrhosis at baseline Time to ESLD outcome Cirrhosis Non-cirrhosis Treatment Control Treatment Control Base case cohort (n) 991 1721 2905 9454 Mean (SD) months 19.4 (8.0) 14.3 (8.8) 20.3 (8.2) 16.0 (9.0) Adjusted mean difference (se) [p value] 3.01 (0.15) [<.0001] 3.45 (0.06) [<.0001] Usual care (n) 256 462 626 2169 Mean (SD) months 27.8 (10.4) 27.2 (10.9) 30.6 (10.6) 29.9 (10.9) Adjusted mean difference (se) [p value] 1 0.92 (0.32) [<.001] 1.56 (0.07) [<.0001] 1. Various baseline characteristics related to the heterogeneity of treat ment on time to ESLD for patients with or without cirrhosis in usual care analysis are listed in T able 4-11.

PAGE 92

92 Table 4-11. Factors associated with primary effectiveness among patients with or without cirrhosis in usual care analysis Variable 1 Cirrhosis Non-cirrhosis Social demographics Age at HCV diagnosis 0.10 (0.02) [<.0001] 0.04 (0.00) [<.0001] Gender (Male vs. Female) -1.56 (0.29) [<.0001] -0.20 (0.06) [<0.001] Type of health Plan POS vs. HMO -1.14 (0.39) [<0.01] 0.44 (0.07) [<.0001] PPO vs. HMO -0.14 (0.32) [0.65] 0.01 (0.06) [0.85] Year of HCV diagnosis 2000 vs. 2006 31.52 (0.78) [<.0001] 34.84 (0.15) [<.0001] 2001 vs. 2006 25.34 (0.75) [<.0001] 28.94 (0.14) [<.0001] 2002 vs. 2006 21.60 (0.72) [<.0001] 26.45 (0.14) [<.0001] 2003 vs. 2006 20.00 (0.75) [<.0001] 21.94 (0.14) [<.0001] 2004 vs. 2006 11.58 (0.73) [<.0001] 13.49 (0.14) [<.0001] 2005 vs. 2006 4.69 (0.69) [<.0001] 6.13 (0.13) [<.0001] Prior comorbid conditions Severe decompensated cirrhosis 0.33 (0.84) [0.69] -Chronic obstructive pulmonary diseases (COPD) 0.00 (0.43) [0.99] -0.24 (0.08) [<0.01] Depression -1.37 (0.57) [0.02] -3.41 (0.14) [<.0001] HBV co-infection -4.22 (0.63) [<.0001] -0.74 (0.16) [<.0001] HIV co-infection 6.45 (0.98) [<.0001] 3.46 (0.16) [<.0001]

PAGE 93

93 Table 4-11. Continued Variable 1 Cirrhosis Non-cirrhosis Prior medical services use Annual medical expenditure -4.00 (0.33) [<.0001] -1.86 (0.07) [<.0001] Biopsy -1.70 (0.43) [<.0001] -1.20 (0.11) [<.0001] Hospitalization -1.14 (0.42) [<0.01] -0.46 (0.08) [<.0001] Emergency visits 0.39 (0.08) [<0.01] -0.08 (0.01) [<.0001] Gastroenterology visits -0.23 (0.03) [<.0001] -0.25 (0.01) [<.0001] Infectious disease visits 0.11 (0.05) [0.02] 0.02 (0.01) [0.01] Instruments related with symptomatic conditions and maintenance Time to treatment 6 vs. <6 months -0.30 (0.30) [0.32] 0.24 (0.06) [<.0001] Antidepressants use 0.24 (0.35) [0.49] -0.19 (0.07) [0.01] 1. Values shown in the tables are coefficients (mean (se), [p-value]) associa ted with treatment initiation relevant covariates and important baseline included in the multivariate adjusted regression models.

PAGE 94

94 Table 4-12. Secondary effectiveness results in base case analysis: Rate of ESLD development in patients with or without cirrhosis Outcome All patients Cirrhosis Non-cirrhosis Treatment (n=3896) Control (11175) Treatment (n=991) Control (n=1721) Treatment (n=2905) Control (n=9454) Overall, n (%) 67(1.7) 166 (1.5) 37 (3.7) 92 (5.4) 30 (1.0) 74 (0.8) ESLD HCC 31 60 18 37 13 23 LT 11 34 6 28 5 6 Decompensated cirrhosis 1 38 89 20 42 18 47 Proxy 2 10 2 1 0 9 Adjusted HR 2 (95% CI) 0.28 (0.20–0.40) 3 0.32 (0.21–0.50) 0.23 (0.13–0.40) Abbreviations: ESLD=end stage liver disease, HCC=hepatocellular carcino ma, LT=liver transplantation, HR=Hazard Ratio. 1. Decompensated cirrhosis=variceal blee ding, hepatic coma and other decompensated conditions in Table 3-1. 2. Various baseline characteristic s, including treatment initiation relevant covariates and important baseline char acteristics included in Cox regression models to examine the heterogeneity of treatment eff ect are shown in Table 4-13. 3. The interaction between treatment and prior cirrhosis was 0.61 (0.34–1.10).

PAGE 95

95 Table 4-13. Factors associated with secondary effectiveness among patients with or without cirrhosis in base case analysis Variable 1 All patients Cirrhosis Non-cirrhosis Social demographics Age at HCV diagnosis 1.03 (1.01–1.04) 1.03 (1.00–1.05) 1.02 (0.99–1.05) Gender ( Male vs. Female) 1.81 (1.35–2.43) 2.30 (1.52–3.49) 1.42 (0.93–2.15) PPO health Plan POS vs. HMO 0.92 (0.66–1.26) 0.97 (0.64–1.49) 0.99 (0.65–1.50) PPO vs. HMO 1.05 (0.74–1.47) 1.12 (0.70–1.79) 0.97 (0.64–1.49) Year of HCV diagnosis 1998 vs. 2007 0.00 (0.00–0.04) 0.01 (0.00–0.09) 0.00 (0.00– ) 1999 vs. 2007 0.20 (0.04–1.05) 0.40 (0.05–2.92) 0.00 (0.00– ) 2000 vs. 2007 0.00 (0.00–0.00) 0.00 (0.00–0.00) 0.00 (0.00–0.00) 2001 vs. 2007 0.00 (0.00–0.00) 0.00 (0.00–0.00) 0.00 (0.00–0.00) 2002 vs. 2007 0.00 (0.00–0.00) 0.00 (0.00–0.01) 0.00 (0.00–0.00) 2003 vs. 2007 0.00 (0.00–0.00) 0.00 (0.00–0.03) 0.00 (0.00–0.00) 2004 vs. 2007 0.00 (0.00–0.01) 0.01 (0.00–0.05) 0.00 (0.00–0.00) 2005 vs. 2007 0.02 (0.00–0.02) 0.05 (0.01–0.23) 0.00 (0.00–0.02) 2006 vs. 2007 0.04 (0.01–0.10) 0.10 (0.02–0.49) 0.03 (0.01–0.18) Prior comorbid conditions Decompensated cirrhosis 59.22 (33.82–103.69) 31.48 (17.72–55.91) – Cirrhosis 8.69 (6.11–11.73) – – COPD 0.57 (0.38–0.87) – – Depression 1.24 (0.66–2.31) 1.44 (0.71–2.94) 0.68 (0.16–2.81) Diabetes 4.24 (3.07–5.86) – 4.71 (2.93–7.55) Drug dependence 1.62 (0.78–3.36) – 0.90 (0.22–3.77) HBV co-infection 3.28 (1.97–5.46) 2.91 (1.57–5.41) 2.55 (1.02–6.40) HIV co-infection 0.58 (0.27 – 1.23) – 0.25 (0.06–1.07) Prior medical services use Annual medical expenditure 2.45 (1.69–3.26) 2.49 (1.62–3.82) 2.44 (1.45–4.08) Biopsy 0.30 (0.17–0.53) 0.39 (0.20–0.75) 0.18 (0.06–0.52) Hospitalization 1.83 (1.27–2.62) 1.52 (0.96–2.41) 2.96 (1.69–5.17) Outpatient visits 0.99 (0.99–1.00) 1.00 (0.99–1.00) 0.99 (0.98–1.00) Emergency room visits 1.05 (1.02–1.08) – 1.01 (0.90–1.12) Gastroenterology visits 1.00 (0.98–1.03) 1.03 (1.00–1.06) 1.02 (0.95–1.10) Infectious disease visits 1.01 (0.99–1.04) 0.92 (0.81–1.05) 1.03 (1.00–1.05) Instruments related with symptomatic conditions and maintenance Time to treatment <6 vs. 6months 0.67 (0.50-0.91) 0.65 (0.44–0.96) 0.67 (0.41–1.09) Antidepressants use 1.44 (1.02–2.02) 1.26 (0.78–2.02) 1.66 (1.01–2.71) 1. Values shown in the tables are adjusted hazard ratios, HR (95% CI) associated wit h treatment initiation relevant covariates and important baseline characte ristics included in the Cox regression models

PAGE 96

96 Table 4-14 Second effectiveness results in subgroup analysis: Rate of ESLD deve lopment in patients with usual care or extended care Outcome Usual care Extended care Treatment (n=882) Control (n=2605) Treatment (n=119) Control (n=353) Overall n (%) 12 (1.4) 33 (1.3) 9 (7.6) 9 (2.6) ESLD HCC 8 11 4 3 LT 1 10 2 5 Decompensated cirrhosis 5 14 5 2 Proxy 0 3 0 1 Adjusted HR (95% CI) 1 0.61 (0.28–1.35) 2 1.79 (0.58–5.54) 3 Abbreviations: ESLD=end stage liver disease, HCC=hepatocellular carcino ma, LT=liver transplantation, HR=Hazard Ratio. 1. Various baseline characteristics, includi ng treatment initiation relevant covariates and instruments related with symptomatic condi tions and maintenance included in the Cox regression models to examine the heterogeneity of t reatment effect are shown in Table 4-15. 2. The interaction between treatment and prior cirrhosi s was 0.17 (0.04–0.81). 3. The interaction between treatment and prior cirrhosis was 1.92 (0.62–5.96)

PAGE 97

97 Table 4-15. Factors associated with secondary effectiveness among patients with usual care or extended care Variable 1 Usual care Extended care 3 Social demographics Age at HCV diagnosis 0.97 (0.93–1.01) 0.89 (0.82–0.95) Gender ( Male vs. Female) 2.80 (1.38–5.70) 1.52 (0.44–5.27) Private insurance (0.00– ) -PPO health Plan POS vs. HMO 2.14 (0.98–4.66) -PPO vs. HMO 0.86 (0.41–1.82) -Year of HCV diagnosis 2 2000 vs. 2006 0.00 (0.00–0.01) -2001 vs. 2006 0.01 (0.00–0.12) -2002 vs. 2006 0.02 (0.00–0.25) -2003 vs. 2006 0.01 (0.00–0.17) -2004 vs. 2006 0.07 (0.01–0.96) -2005 vs. 2006 0.47 (0.05–4.43) -Prior comorbid conditions Decompensated cirrhosis 5.54 (1.80–17.08) 13.88 (3.05–63.08) Cirrhosis 10.72 (5.23–21.91) 18.24 (4.25–78.25) Chronic obstructive pulmonary diseases (COPD) 1.22 (0.46–3.20) 1.33 (0.39–4.61) Depression 0.59 (0.08–4.55) 0.75 (0.18–3.10) Diabetes -7.58 (2.01–28.59) HBV co-infection 3.95 (1.14–13.60) -HIV co-infection 0.00 (0.00– ) -Psychiatric disorders -1.25 (0.27–5.81) Prior medical services use Annual medical expenditure 4.87 (2.25–10.56) 1.47 (0.44–4.89) Biopsy 1.90 (0.60–6.06) 3.66 (0.60–22.29) Hospitalization 1.21 (0.53–2.77) 1.00 (0.26–3.95) Outpatient visits 1.01 (0.99–1.02) -Emergency visits 0.86 (0.67–1.09) -Gastroenterology visits 1.01 (0.92–1.11) -Infectious disease visits 0.66 (0.34–1.26) -Instrument related with symptomatic condition and treatment maintenance Time to treatment <6 vs. 6 months 2.19 (1.07–4.49) 1.44 (0.49–4.26) Antidepressants use 1.53 (0.73–3.23) 0.86 (0.27–2.75) 1. Values shown in the tables are adjusted hazard ratios, HR (95% CI) associated wit h treatment initiation relevant covariates and important baseline characte ristics (as shown in Table 4-5) included in the Cox regression models. 2. Due to the majority of patients in subgroup analysis were firstly diagnosed with HCV infection during 2000 and 2006, all analyses were performed within the period. 3. Additional baseline covariates added int o the Cox regression model (same covariates in the usual care analysis), the HR= 2.34 (0.58–9.41) for patients with extended care.

PAGE 98

98 Table 4-16. Summary of secondary effectiveness results in base case and usual c are analyses among patients with or without cirrhosis at baseline Rate of ESLD development Cirrhosis Non-cirrhosis Treatment Control Treatment Control Base case cohort, n 991 1721 2905 9454 ESLD event, n (%) 37 (3.7) 92 (5.4) 30 (1.0) 74 (0.8) Adjusted HR(95% CI) 0.32 (0.21-0.50) 0.23 (0.13-0.40) Usual care, n 256 436 626 2169 ESLD event, n (%) 3 (1.2) 22 (5.1) 9 (1.4) 11 (0.5) Adjusted HR (95% CI) 1 0.24 (0.06-0.94) 0.88 (0.27-2.87) 1. Various baseline characteristics included in the Cox regression models to ex amine the heterogeneity of treatment effect are shown in Table 4-17 for patients invol ving in usual care analysis.

PAGE 99

99 Table 4-17. Factors associated with secondary effectiveness among patients with or without cirrhosis in usual care analysis Variable 1 Cirrhosis Non-cirrhosis Social demographics Age at HCV diagnosis 0.94 (0.90–0.99) 1.00 (0.93–1.06) Gender 3.72 (1.38–10.05) 1.95 (0.69–5.52) PPO health Plan POS vs. HMO 2.51 (0.83–7.58) 1.69 (0.54–5.26) PPO vs. HMO 0.93 (0.33–2.60) 0.87 (0.29–2.65) Year of HCV diagnosis 2 2000 vs. 2006 0.00 (0.00–0.08) 0.00 (0.00– ) 2001 vs. 2006 0.01 (0.00–0.16) 0.00 (0.00– ) 2002 vs. 2006 0.02 (0.00–0.47) 0.00 (0.00– ) 2003 vs. 2006 0.00 (0.00– ) 0.00 (0.00– ) 2004 vs. 2006 0.05 (0.00–0.94) 2.06 (0.00– ) 2005 vs. 2006 0.22 (0.02–2.54) (0.00– ) Prior comorbid conditions Decompensated cirrhosis 3.87 (1.16–12.89) 13.88 (3.05–63.08) Chronic obstructive pulmonary diseases (COPD) 0.92 (0.19–4.47) 1.33 (0.39–4.61) Depression 1.06 (0.12–8.77) 0.75 (0.18–3.10) HBV co-infection 6.53 (1.61–26.40) -HIV co-infection 0.00 (0.00– ) -Prior medical services use Annual medical expenditure 5.31 (1.75–16.11) 5.18 (1.57–17.02) Biopsy 2.25 (0.43–11.91) 2.93 (0.56–15.50) Hospitalization 0.88 (0.27–2.90) 1.51 (0.43–5.32) Emergency visits 0.81 (0.54–1.21) 0.93 (0.68–1.27) Gastroenterology visits 1.01 (0.91–1.12) 1.00 (0.75–1.35) Infectious disease visits 1.00 (0– ) 0.85 (0.48–1.53) Instrument related with symptomatic condition and treatment maintenance Time to treatment <6 vs. 6 months 2.19 (0.81–5.91) 2.46 (0.78–7.78) Antidepressants use 1.21 (0.36–4.05) 1.81 (0.66–4.93) 1. Values shown in the tables are adjusted hazard ratios, HR (95% CI) associated wit h treatment initiation relevant covariates and important baseline characte ristics included in the Cox regression models

PAGE 100

100 Table 4-18. Total cost among patients with cirrhosis in base case and usual care analyses Patients n Total costs per patient Mean (SD) dollar Adjusted mean difference (se) 1 P value Base case cohort Control 1721 $25,092.91 (56,000.58) 0 Treatment 991 $54,907.83 (62,656.31) 25,722(2365.28) <.0001 Usual care Control 436 $42,714.59 (81,376.59) 0 Treatment 256 $76341.62 (62,464.60) 32,953 (5642.31) <.0001 1. The statistical relevant covariates on mean total cost difference are sho wn in Table 4-19.

PAGE 101

101 Table 4-19. Factors associated with mean total cost difference between tre atment and control among patients with cirrhosis in base case and usual care analyses Variables 1 Base case Usual care Social demographics Age at HCV diagnosis 58 (139.11) [0.68] 43 (345.37) [0.90] Gender 231 (2257.23) [0.92] 2621 (5375.24) [0.63] Census region 2 1 vs. 5 -42859 (4655.73) [<.0001] -109370 (12538) [<.0001] 2 vs. 5 -51017 (6074.78) [<.0001] -116435 (16092) [<.0001] 3 vs. 5 -51289 (5313.53) [<.0001] -115558 (14627) [<.0001] 4 vs. 5 -49004 (6090.18) [<.0001] -116623 (17145) [<.0001] Insurance (Private vs. Public) 5659 (7778.04) [0.47] -36391 (23827) [0.13] Types of Health Plans POS vs. HMO 6486 (2863.74) [0.02] -7315 (7069.79) [0.30] PPO vs. HMO 47 (2532.88) [0.99] -1612 (5666.40) [0.78] Year of HCV diagnosis 3 1998 vs. 2007 (2006) 15421(10611) [0.15] -1999 vs. 2007 (2006) -558 (12180) [0.96] -2000 vs. 2007 (2006) 38928 (7841.05) [<.0001] 20608 (14495) [0.16] 2001 vs. 2007 (2006) 39718 (7721.35) [<.0001] 32861 (13874) [0.02] 2002 vs. 2007 (2006) 28228 (7474.19) [<0.001] 30128 (13272) [0.02] 2003 vs. 2007 (2006) 31800 (7631.29) [<.0001] 27133 (13568) [0.05] 2004 vs. 2007 (2006) 24501 (7452.65) [0.001] 18479 (13220) [0.16] 2005 vs. 2007 (2006) 18267 (7056.34) [0.01] 2185 (12434) [0.86] 2006 vs. 2007 9949 (7181.14) [0.17] -

PAGE 102

102 Table 4-19. Continued Variables 1 Base case Usual care Prior comorbid conditions Severe decompensated cirrhosis -2126 (4080.52) [0.60] 5430 (15051) [0.72] Chronic obstructive pulmonary diseases (COPD) -2072 (3196.42) [0.52] -11565 (7874.15) [0.14] Cerebral vascular disease (CVD) 20552 (5680.32) [<.0001] 51728 (14214) [<0.001] Depression -2336 (4471.02) [0.60] 2837 (10232.00) [0.78] Diabetes 7025 (2968.87) [0.02] 9638 (7462.10) [0.20] Drug dependence -3483 (6311.27) [0.58] 86435 (19783) [<.0001] HBV co-infection -4560 (4540.02) [0.32] 5033 (11187) [0.65] Heart diseases 11723 (3388.20) [<0.001] 19820 (7911.23) [0.01] HIV co-infection 40509 (7386.52) [<.0001] 87912 (17353) [<.0001] Obesity -4031 (4489.65) [0.37] -6306 (11674) [0.59] Psychiatric disorders -6670 (2923.69) [0.02] -4452 (6882.19) [0.52] Prior medical services use Biopsy -1196 (3344.17) [0.72] 5030 (7702.84) [0.51] Hospitalization 3469 (2717.14) [0.20] 2717 (7127.38) [0.70] Outpatient visits -179 (43.07) [<.0001] -114 (108.08) [0.29] Emergency room visits 510 (263.35) [0.05] 5796 (1460.05) [<.0001] Family practice visits 624 (148.07) [<.0001] 118 (403.68) [0.77] Gastroenterology visits -17 (193.90) [0.93] 250 (520.36) [0.59] Infectious disease visits -176 (361.44) [0.63] -1352 (899.60) [0.14] Internal medicine visits 281 (117.59) [0.02] -428 (280.02) [0.13]

PAGE 103

103 Table 4-19 Continued Variables 1 Base case Usual care Instrument related with symptomatic condition and treatment maintenance Time to treatment initiation 8858 (2262.50) [<.0001] 13985 (5483) [0.01] Antidepressants use 19784 (2898.35) [<.0001] 21558 (6272.09) [<.001] 1. Values shown in the tables are coefficients (mean (se), [p-value]) associa ted with all baseline characteristics into the multivariate adjusted regression models. 2. Referen ce group= 5 (National and Other in the dataset); =New England and Middle Atlantic; 2=Eas t North Central and West North Central; 3=South Atlantic, East South Central, and West South Central; 4=Mountain and Pacific. 3. In usual care analysis, because no patient was diagnos ed HCV in the year of 2007, the reference year was analyzed with 2006.

PAGE 104

104 Table 4-20. Adjusted mean net benefit difference between treatment and control among cirrhotic patients in base case and usual care analyses Base case 1 Usual care 2 WTP Mean Lower limit Upper limit Mean Lower limit Upper limit l =0 -25254 -29815 -20692 -33052 -42635 -23469 l =10K 2797 -2076 7669 -23451 -33132 -13771 l =20K 3 30847 24862 36831 -13850 -25782 -1918 l =30K 58897 51346 66448 -4250 -19671 11172 l =40K 86947 77600 96294 5351 -14145 24847 l =50K 114997 103734 126260 14952 -8904 38809 l =60K 4 143047 129801 156294 24553 -3819 52925 l =70K 171098 155826 186370 34154 1175 67133 l =80K 199148 181823 216472 43755 6111 81399 1-2. The value of lower and upper limit indicate 95% CI limit were varied at selec ted WTP (l) in the multivariate adjusted model. Both base case and usual care visual results in INB estimate for antiviral therapy are shown in Figure 4-2. 3. In base case analysis, the hypothe tical assumption that treatment is cost-effective was found at WTP $15,000 in the multivariate adjusted regression model (mean net benefit difference= 11471, 95% CI=16822 to 22173, one-sided test with p<0.001). 4. In usual care analysis, the hypothetical assumption that treatment i s costeffective was found at WTP $60,000 in the multivariate adjusted regression model (mean net benefit difference=24553, 95% CI= -3819 to 52925, one-sided test with p=0.04)

PAGE 105

105 Table 4-21. Covariates effects on the INB of antiviral therapy for patients with cirrhosis in base case analysis Covariate INB ( l =15,000) (se) [p value] INB ( l =20,000) (se) [p value] INB ( l =30,000) (se) [p value] Constant term -36469 (12438.11) [0.003] -36460 (13946.66) [0.01] -36440 (17622.49) [0.04] Treatment 9204 (25610.39) [0.72] 14211 (28716.54) [0.62] 24225 (36285.17) [0.50] Year of HCV diagnosis 2002 vs. 2007 45944 (22364.45) [0.04] 65096 (25076.92) [0.01] 103400 (31686.29) [0.006] 2003 vs. 2007 36245 (22588.63) [0.11] 55034 (25328.29) [0.03] 92613 (32003.91) [0.004] 2004 vs. 2007 40262 (22296.52) [0.07] 57884 (25000.75) [0.02] 93128 (31590.04) [0.003] 2005 vs. 2007 31253 (21442.88) [0.15] 45360 (24043.58) [0.06] 73573 (30380.59) [0.02] 2006 vs. 2007 35264 (21706.19) [0.10] 47596 (24338.83) [0.05] 72259 (30753.66) [0.02] Severe decompensated cirrhosis -30947 (11959.51) [0.01] -40045 (13410.02) [0.003] -58240 (16944.41) [0.001] Diabetes -7888 (7450.70) [0.29] -13237 (8354.36) [0.11] -23934 (10556.27) [0.02] HIV co-infection -37300 (18270.06) [0.04] -34539 (20485.9 4) [0.09] -29019 (25885.2 9) [0.26] Antidepressant use -12951 (6967.70) [0.06] -15809 (7812.78) [0.04] -21527 (9871.95) [0.03] R-squared (adjusted) Prob>F 0.77 <.0001 0. 83 <.0001 0.88 <.0001 Note: The full model of covariate adjusted incremental net benefit estimate w ith treatment interactions in base case analysis is shown in the Appendix C.

PAGE 106

106 Table 4-22. Covariates effects on the INB of antiviral therapy for patients with cirrhosis in usual care analysis Covariate INB ( l =60,000) (se) [p value] INB ( l =70,000) (se) [p value] INB ( l =80,000) (se) [p value] Constant term 259712 (80461.09) [<001] 284491 (92924.82) [<0.01] -309271 (105605.51) [<.01] Treatment 203740 (138586.38) [0.14] 240612 (160053.94) [0.13] 277483 (181895.20) [0.12] Year of HCV diagnosis 2001 vs. 2006 146771 (68895.53) [0.04] 166747 (79567.72) [0.04] -186723 (90425.68) [0.04] 2002 vs. 2006 -155104 (67504.53) [0.02] 186277 (77961.24) [0.02] -217450 (88599.98) [0.01] HBV 130600 (79808.36) [0.10] 156443 (92170.98) [0.09] 182286 (104748.81) [0.08] Emergency room visits 16448 (9118.66) [0.07] 18692 (10531.17) [0.08] 20936 (1968.28) [0.08] R-squared (adjusted) Prob>F 0.94 <.0001 0. 94 <.0001 0.94 <.0001 Note: The full model of covariate adjusted incremental net benefit estimate w ith treatment interactions in usual care analysis is shown in the Appendix D.

PAGE 107

107 Figure 4-1. Cumulative ESLD events in the study follow-up. Figure 4-2. Cumulative ESLD events in the study follow-up (cirrhotic patients) Figure 4-3. Cumulative ESLD events in the study follow-up (non-cirrhotic patients)

PAGE 108

108 Figure 4-4. Plot of INB (95% CI) between treatment and control among patients wit h cirrhosis in base case and usual care analyses. The values are shown in Table 4-20. Among patients with cirrhosis in base case analysis, treatment is cost-effec tive was found at WTP $15,000 (d= 11471, 95% CI=16822 to 22173, one-sided test with p<0.001). Among patients with cirrhosis in usual care analysis, treatment is cos teffective was found at WTP $60,000 (d=24553, 95% CI= -3819 to 52925, onesided test with p=0.04).

PAGE 109

109 0.00 0.25 0.50 0.75 1.00 Normal F[(nmb15-m)/s] 0.00 0.25 0.50 0.75 1.00 Empirical P[i] = i/(N+1) Figure 4-5. P-P plot of the net benefit (l=$15,000) for cirrhotic patients in base case analysis. 0.00 0.25 0.50 0.75 1.00 Normal F[(nmb60-m)/s] 0.00 0.25 0.50 0.75 1.00 Empirical P[i] = i/(N+1) Figure 4-6. P-P plot of the net benefit (l=$60,000) for cirrhotic patients in usual care analysis.

PAGE 110

110 CHAPTER 5 DISCUSSION Our study shows that, the INB estimate for the cost-effectiveness of init ial combination antiviral therapy varies depending on the duration of antiviral therapy and the pr esence of cirrhosis among patients infected with HCV in the managed care setting. Pat ients with extended care of antiviral therapy suggest no evidence in prevention of ESLD progressi on between treatment and control without considering the presence of cirrhosis. Among patients without cirrhosis, KM estimate of treatment effect suggests the potential benefi cial trend of initial combination therapy, although the evidence was lack of statistical power in patient s with usual care. It further demonstrates that using the net benefit regression frame work estimates the efficiency of antiviral therapy varies depending on the cirrhosis status and c omorbid conditions. Lastly, this study provides empirical evidence on the efficiency of initial a ntiviral therapy in practice, which facilitate better understanding and implementation of HC V care in the managed care setting. Effectiveness of Antiviral Therapy Descriptions of Patients Characteristics Descriptive analyses in the Table 4-1 and Table 4-4 of Chapter 4 present the extent of imbalance in patients’ baseline characteristics between base cas e and subgroup analyses. First, risks associated with ESLD progression at baseline vary across patients involving in different sample groups. For examples, treated patients with extended care compared to all t reated patients were more likely to be older and male patients, having compensated and decompensated cirrhosis at the time of antiviral therapy initiated. Consequently, the magnitude of pre-existing differences between treatment and control vary across three sampled groups. As shown in Table 4-3 and Table 4-5,

PAGE 111

111 potential confounders with respect to baseline characteristics associated wi th the initiation of antiviral therapy vary in different analysis groups. Second, the rationale of using sample groups for analysis based on the duration of treatment in this study was to estimate difference in treatment effec ts between patients who were non-responders or slow-responders with an extended care after a full course of antivi ral therapy and those who have a SVR after a standard of care while controlling for potential c onfounders in the sample group. Thus, as usual care and extended care analyses were performed in t his study will allow providing complementary evidence to the treatment effect obtained in t he original base case analysis. The summary of effectiveness results in Table 4-10 a nd Table 4-16 display the trend for a beneficial treatment effect of antiviral therapy in bas e case analysis was consistent with usual care analysis Additionally, a s shown in Table 48 and Table 4-14 there is no difference in treatment effect between treatment and control among patient s with extended care. Estimates of Treatment Effectiveness Study results revealed that two metrics of treatment effectiveness measures agreed on a reduction in the risk of ESLD progression among cirrhotic patients in base case a nalysis and those with usual care (Table 4-10 and Table 4-16). However, among non-cirrhotic patients in usual care analysis, both estimates (average time to ESLD event in months and ha zard ratio) had a same trend for a beneficial effect of antiviral therapy, despite the KM estimate in a timedependent Cox regression model failed to achieve a statistically significa nt difference between treatment and control (coefficient=1.56, se=0.07, p<.0001 and hazard ratio=0.88, 95% CI=0.272.87, respectively). Assuming correct model specification and no violation of assumptions necessary for the ordinary least squares (OLS) linear regression model and the Cox regression model two estimates for treatment effect vary by their structural assumptions and a pplications. First, the

PAGE 112

112 dependent variable is commonly to be a continuous variable in OLS regression model, while it is a binary variable in Cox regression. Cox regression is more likely to use at pre dicting the probability of a binary outcome incurring at any given time during the followup, and the time is measured as a “median” statistic in the model. 82 It is possible that the observed follow-up periods in the study samples are skewed and estimated magnitude of treatment e ffect distorts when the mean differs from the median estimate. Second, model specification for c ovariates included in the model may influence tests of significance for the treatment d ummy variable. For example, age at HCV diagnosis was treated as time-dependent variable in Cox regression analyses, which may provide more robust estimation for its relationship with tre atment effect than OLS does. Also, the number of ESLD event was small in the control group of patients without cirrhosis due to a relatively short follow-up to observe the development of advanc ed liver disease (average observation period, months =32.2/31.0 (treatment/control)), w hich leads to an insufficient power of statistical analysis in the Cox regression model. It has remained unclear whether the extended antiviral therapy reduces th e risk of ESLD. In the present study, we also found that of no difference in treatment effect between treat and untreated patients with extended care. Alternative low-dose peginterferon m aintenance therapy (90 mcg/wk) combination with ribavirin has been investigated prospecti vely in the HALT-C trial for patients with advanced fibrosis who had failed initial thera py. 83 After a 3.5year follow-up, there was no statistical difference in outcomes between th e treatment group and the controls, indicating that their data do not support the use of maintenance therapy i n treatment of non-responders. However, in recent data, Kaiser et al 84 showed that in a subgroup of patients with at least a one-log drop in HCV RNA levels, long-term low-dose mai ntenance therapy decreased fibrosis scores in non-responders with fibrosis and cirrhosis. F urther

PAGE 113

113 subgroup of patient analyses may identify patients who are more likely to res pond to extended duration of therapy. Incremental Net Benefit of Initial Combination Antiviral Therapy Cirrhotic Patients Examining treatment dummy coefficients in comparison to the control group coeffi cient reveals that the INB for cirrhotic patients in the base case analysis we re more efficient than those patients with usual care. Although, all analyses were assumed by the assum ptions of treatment discontinuation incurred by the lack of early virological response or treatmentrelated adverse effects for patients in the base case analysis. It is imperative to MCO decision makers in health plan design that enhancing patient identification, such as patients with developing ci rrhosis or characteristics related to likelihood of rapid treatment response will all ow for optimizing therapy to ensure that the intended goals of therapy are met. For patients with developing cirrhosis, although they have a lower likelihood of virol ogical response, 52, 85 the present study results support the cumulative evidences suggest that antivira l therapy can reduce the need of liver transplant, risk of cirrhosis decompensation and HCC development, even if it does not eradicate the hepatitis C virus. 58 86-89 Considering approximately 20% of study samples already had cirrhosis, further treatm ent goal will be focus on the health-related quality improvement, and the prolongation on HCC development and the need for liver transplants. As not all patients respond to antiviral therapy or nec essarily develop progressive liver disease, using patient’s infected HCV genotype and early virological response rate to guide duration of treatment have been shown to increase cost-effective ness of combination antiviral therapy. 26, 81 Emerging recent clinical opinions on initial treatment for HCV-infected patients, patients with cirrhosis should be assessed for virolo gical response at the

PAGE 114

114 end of week 12 or 24 treatment depending on genotype (24 weeks for genotypes 1/4 and 12 weeks for genotypes 2/3) and treatment continued in those who are viral negative at the time. MCO decision makers should support the reassessment of virological response during t reatment to reduce substantial drug costs and minimize adverse treatment effects, th ereby improving efficiency of initial combination antiviral therapy. Significance and effect of covariates. The differences on the costs, benefits and costeffectiveness of treatment intervention varied widely across the rang e of different levels of patient’s co-morbid conditions between base case and usual care were noteworthy Among cirrhotic patients received approximately 7.5 months of antiviral therapy, cirr hosis status and comorbid conditions have more impact on the marginal cost-effectiveness compare d with patients receiving 12 months of antiviral therapy under the statistical signi ficance in Table 4-21 and Table 4-22. Although treatment of patients with decompensated cirrhosis (e.g., variceal bl eeding and hepatic coma) is feasible, 90 the observed decreased benefits or total cost increase is noteworthy in the present study. Theoretically, patients with cirrhosis decompensation ar e candidates on the waiting list of liver transplant. Treatment of patients with advanced liver c irrhosis waiting for liver transplant is costly and the likelihood of preventing recurrence is not clear. 91 92, 93 Optimum timing of treatment initiation for patients at different stage of disease severity for the prevention of HCC, liver transplant and death must be studied further. Diabetes is not just a risk factor of developing HCC in chronic hepatitis C patient s. 61 Recent studies suggest that increasing level of insulin resistance are ass ociated with impaired initial virological response and SVR. 94, 95 The results of less cost-effectiveness in the presence of diabetes compared to non-diabetic patients with cirrhosis, was mainly associ ated with less

PAGE 115

115 effectiveness in base case analysis. The prevalence of diabetes and impair ed fasting glucose is high; thus, further interventions aimed at reducing insulin resistance in chronic he patitis C through a multidisciplinary approach are warranted, including the administration of hypoglycemic agents and lifestyle changes should probably be included in the c linical management of patients with chronic hepatitis C and insulin resistance, although the potential beneficial effects on liver fibrosis progression and response to therapy remai n to be assessed. Patients co-infected with HIV/HCV infection are difficult-to-treat Although HIV coinfected patients, in particular, when CD4 count appears less than 200cells/mm 2 might accelerate HCV-related disease progression, 18 treatment efficacy of pegylated interferon combination with ribavirin in recent randomized, controlled trial showed 40% of SVR in patients with HCV /HIV co-infection 57, 59 and it was cost-effective in terms of increase in life-expectancy. 96, 97 We assumed that combination antiviral therapy was more likely to be prescribed to thos e who with well control HIV co-infected patients, as HIV-infected patients with low CD 4 cell count would suffer a higher risk of opportunistic infection and worsen quality of life in prac tice. The results of base case analysis showed that treatment of cirrhotic patients with HCV /HIV co-infection was effective; while the cost of treatment was substantial and outweigh the benef its was seen among patients co-infected with HCV/HIV, even though its effect on the marginal cos t-effectiveness was not consistent across any value of WTP. The results are limited with suffi cient data to characterize the stage and severity of HIV infection. Year of HCV diagnosis was another important factor associated with the cos teffectiveness of treatment in the group of cirrhotic patients with usual care In Table 4-22, the results of marginal cost-effectiveness in patients diagnosed with HCV in the year of 2001 and 2002 was less efficient than those who were diagnosed with HCV in year of 2006. The

PAGE 116

116 difference in cost and primary effectiveness measure in the year of 2001 and 2002 c ould be related with varied patterns of initial combination therapy in HCV-infected patients across the MCO health care system. It is also possible that pegylated interferons comb ination with ribavirin was approved and recommended as a standard of treatment from 2000 to 2002; consequently, the number of patients newly diagnosed with HCV around the period increased and patients wer e likely to introduce with this relatively expensive combination regiments. Non-Cirrhotic Patients Decision to treat or not with combination antiviral therapy should take several impor tant factors into consideration, including histological findings, symptoms, patient’s co -morbid conditions, age, and motivation. Management of patients with mild histological fibros is remains controversial because not all patients with HCV infection progress to cirrhosi s and some may delay initiation of treatment to avoid its potential side effects. 55 The cost-effectiveness of treating patients with no cirrhosis or mild fibrosis has been questioned, since the pro gnosis even without therapy is excellent, further underscoring the importance of accu rately staging the severity of liver disease. 29, 98 On the other hand, with delayed treatment, HCV-infected patients become older, develop risk for cirrhosis progression, impaired quality of life, and comp rise costeffectiveness of treatment intervention. For patients with no or only mild fibrosis in those whose treatment is deferred, liver biopsy can be used to monitor liver disease progr ession. The immediate treatment initiation for a group of patients with only mild cirrhos is has been suggested to be cost effective compared to a watchful waiting strategy with liver biopsy every 3 years and combination therapy in patients found to have cirrhosis on liver biopsy. 27 After weighting the risks, benefits and costs of existing HCV treatment periodical liver biopsy and biochemical markers monitoring could be used to guide recommendation for treatment with optimal timing. However, considering the bias of liver biopsy pe rformance, 99

PAGE 117

117 patient’s perceptions on the severity of disease manifestation 100 a efficient regular monitoring strategy involving a good referral mechanism to the gastroenterology specia lty could be costly and lack of an appropriate screening rate in the MCO setting In the present study, combination therapy was also favor for a group of asymptoma tic patients in base case analysis. It implicitly suggests that treatment might be cost-effective when the discontinuation of therapy was associated with lack of virological respons e after 7 months therapy. From the MCO perspective, it further highlights the value of confirma tory tests for early virological response at end of 24 weeks therapy (genotype 1/4 infection); t hereby increases marginal cost-effectiveness of antiviral therapy. The comparisons of INB between patients with and without cirrhosis in base case cohort shown in Figure 5-1. From the perspective of t he society, further research for asymptomatic HCV-infected patient s is required on the costeffectiveness of combination therapy compared to a periodical watchful wait ing strategy; in particular the intervention's long-term impact on HRQoL and health service c osts requires further evaluation. Limitations Our study has several noteworthy limitations. As previously discussed, the prim ary limitation and threat to the internal validity of this study was the possible pre sence of selection bias in treatment. In this study, the concern is whether patients who received com bination antiviral therapy were at a greater risk of ESLD compared to those in the cont rol group. Although benefit, costs, and cost-effectiveness of treatment intervention were stratified analyzed with an adjustment for those baseline characteristics relevant to treatm ent initiation and several created instruments, the treatment and control groups may not have equal unmeasur ed confounding, and that selection bias may present in the stratified level. It has bee n suggested that a low baseline serum viral load is associated with a significantly higher pro bability of achieving

PAGE 118

118 SVR following initial combination therapy. For example, treatment effect in cirrhotic individuals with a higher baseline serum viral load (>800,000 IU/mL) appeared less effective than in noncirrhotic patients with a low baseline serum viral load. 85 The amount of bias in effectiveness measure between patients with and without cirrhosis was not clear due to the unmea sured confounding. A limitation includes a disadvantage of the IHCIS dataset. The information r elated with prescription dosage and duration of prescription provided was not reliable in the dataset Although the operational definitions of continuous refilling were employed to categor ize patients with usual or extended care in subgroup cohort, it would loss information in a group of treated patients who reduce doses of prescription due to treatment-related side effects. This un-measured confounder is a strong predictor of achieving SVR, which may affect the precisi on of provided results. If patients receiving extended care were more likely to decrea se doses of either interferon alpha or ribavirin because of side effects, then that observed treatment variabili ty in extended care analyses would lead to a smaller extent of difference in treatment e ffect. No laboratory and histological evidence is available to confirm treatment outcome and HCV-rela ted cirrhosis progression. Study results obtained in the MCO population may not be generalizable to HCV-infe cted patients in different health care systems. Specifically, patients with fewer barriers to access to HCV care than patients with public health insurance may have better medicati on adherence to achieve a successful treatment outcome. Also more than half the study sample re sided in the Northeast and Middle Atlantic regions, and the patterns of antiviral therapy use i n this study population may differ from HCV-infected persons outside the UnitedHealth manage d care program. However, this MCO program with more than 80 million members alive represe nts a

PAGE 119

119 large and important high risk population that has not been well evaluated for HCV care a nd therefore warrants comprehensive investigation. Future Research Managed care organizations are sharing a large proportion of economic burden related with HCV infection. 14 MCOs appear to have more incentives than fee-for-service environment to slow the growth of health care expenditures. Applications of the net benefit reg ression framework in this study presented the valuable information about current initia l antiviral therapy efficiency among newly-diagnosed, HCV-infected patients to decision maker s involved in the MCO program. It demonstrates the statistical merits to decide whether a ntiviral therapy is costeffective based on the predicated value (i.e., INB) in the linear regression mode l. Although this study was not able to answer with which patient outcome that policymakers would adopt w hat value of willingness to pay for treatment cost, future research will have to e xtend the investigation, such as a cost-benefit analysis. Analysis of WTP by HCV-infe cted patients for initial therapy, or the optimal timing of treatment initiation by progres sive disease stages, will allow valuing all aspects of outcome improvement by antiviral therapy Another important area of research is the application of the net benefit regres sion framework using observational data. In order to account for the incomplete data (cens oring) issue in effectiveness and costs measure, a method of inverse probability censoring w eighting (ICPW) was employed in this study to adjust for potential bias introduced by censoring patie nts who dropout in the study follow up. One of limitation of using ICPW in the linear regression m odels for both measures of cost and effectiveness was assumed the pattern of censoring ha s to be random. With respect to the primary effectiveness estimate in this study, sur vival time data are often right skewed with a small proportion of patients surviving much longer than the res t of patients. Although the reported median survival time is commonly in the KM estimate of

PAGE 120

120 survival function, some consider it is not an efficient estimator of the expected survi val time. For instance, median survival time is estimated as a single point in time on the survival curves, and it does not take the exact magnitude of most observations into account. Furthermore, median survival time is difficult to give a meaningful interpretation. 101 There are several approaches have been discussed in the literature, including restricted mean method 101 and ICPW 75, 102 have been investigated using randomized controlled trial data. Likewise, costs data a re concerned with censoring and a highly right-skewed distribution in a small proportion of patients. The comparisons of advantages and disadvantages among different approaches have not been performed and well discussed in both survival time and cost estimation. Further res earch focus on ways to overcome incomplete data and methodological limitations in real-world obse rvation data and comparisons of patterns of censorings in expected survival time and costs would increase the utility of the net benefit regression framework for maximizi ng value of heath resources. Summary and Conclusions Estimating the cost-effectiveness of treatment interventions using realworld data is challenging. The present study shows that the net benefit estimation of initi al combination antiviral therapy within a regression framework is dependent on the richness of ava ilable patientlevel data. This study is the first empirical investigation of HCV treat ment to apply inverse probability of censoring weights to censored effectiveness data. The resul ts support that initial combination antiviral therapy during compensated cirrhosis is cost-effecti ve. The results of total cost estimation during the follow-up revealed the mean difference in total cost s between the treatment and control group was higher for patients with cirrhosis in usual care a nalysis than those patients in base case analysis. A limitation is that median Kaplan-Meie r estimate for the risk of end stage liver disease progression suffered from the lack of sufficie nt statistical power to

PAGE 121

121 demonstrate beneficial effect of treatment among non-cirrhotic patients with usual care. Moreover, the presence of potential selection bias cannot be ignored. Initial combination antiviral therapy during compensated cirrhosis is cost-e ffective compared with no therapy at a minimum willingness to pay threshold of $15,000 for patien ts receiving an average of approximately 7.5 months therapy, and has a minimum will ingness to pay threshold of $60,000 for patients with an average of 12 months therapy. Treatment initiat ion for non-cirrhotic patients is also cost-effective at a minimum willingness to pay threshold of $15,000, and yielded greater net benefit only when therapy discontinued in those who are vira l positive after approximately 7.5 months treatment. Additional work is needed to bette r understand whether these thresholds would be considered cost-effective from the MC O perspective. Future examinations of the best cost-effectiveness strate gy for initial combination antiviral therapy should consider duration of therapy, early virological response and genotype into account.

PAGE 122

122 Figure 5-1. Plot of INB (95% CI) between treatment and control among patients wit h and without cirrhosis in base case analyses. Patients with cirrhosis, treatme nt is costeffective was found at WTP $15,000 (d= 11471, 95% CI=16822 to 22173, onesided test with p<0.001). For patients without cirrhosis, treatment is cost-effe ctive was found at the same threshold as cirrhotic patients (WTP $15,000, d= 24815, 95% CI=11601 to 38030, one-sided test with p<0.001).

PAGE 123

123 APPENDIX A NET BENEFIT OF INITIAL COMBINATION ANTIVIRAL THERAPY IN B ASE CASE ANALYSIS Patients with Cirrhosis Abbreviations for variables in multivariate net benefit regression models ar e: tx (treatment indicator), age_HCV (age at the year of HCV diagnosis), plan1 (health plan: POS), plan2 (health plan: PPO), plan3 (health plan: HMO), geo1 (census region: New England and Middle At lantic), geo2 (census region: East North Central and West North Central), geo3 (census region: South Atlantic, East South Central, and Mountain and Pacific), geo4 (census region: We st South Central), geo5 (census region: National and Other in the dataset), yr_hcv0=1998, y r_hcv1=1999, yr_hcv2=2000, yr_hcv3=2001, yr_hcv4=2002, yr_hcv5=2003, yr_hcv6=2004, yr_hcv7=2005, yr_hcv8=2006, yr_hcv9=2007, DC2 (severe decompensated cirrhosis), CC (compensated cirrhosis), Ddpend (Drug Dependence), dpress_late (depression after HCV diag nosis), DM (diabetes), HBV (Hepatitis B virus infection), HIV (Human Immunodeficienc y Virus), anncost_b4 ( prior annual medical expenditure),Hx_b4 (prior hospitalization history), FG _cnt (family practice physician visits), GS_cnt ( gastroenterologist vis its), Intern_cnt (internal medicine physician visits). out_cnt (outpatient/physician visits), InMed_cnt ( internal medicine physician visits), TT_lab (time to treatment initiation), anti_dpress (anti depressants use). Table A-1. WTP=$15,000 [Regression with robust standard errors] Linear regression Number of obs = 2712 F( 34, 2677) = 331.16 Prob > F = 0 R-squared = 0.767 Root MSE = 64773

PAGE 124

124 Table A-1. Continued nmb15 Coef. Robust Std. Err. t P>|t| [95% Conf. Interval] tx 16821.7 2728.957 6.16 0.000 11470.63 22172.78 gender -20220.2 2722.791 -7.43 0.000 -25559.2 -1488 1.2 plan2 8761.94 2840.408 3.08 0.002 3192.324 14331.56 plan3 9414.779 4060.917 2.32 0.021 1451.928 17377.6 3 geo1 109218.5 11020.52 9.91 0.000 87608.96 130828.1 geo2 55715.18 11932.39 4.67 0.000 32317.55 79112.81 geo3 48726.8 11614.74 4.2 0.000 25952.02 71501.58 geo4 29257.76 11579.62 2.53 0.012 6551.855 51963.67 yr_hcv0 121568.5 10771.16 11.29 0.000 100447.9 1426 89.2 yr_hcv1 26197.69 11353.87 2.31 0.021 3934.451 48460 .92 yr_hcv2 314579.6 9836.25 31.98 0.000 295292.1 33386 7 yr_hcv3 291805.4 8700.85 33.54 0.000 274744.3 30886 6.4 yr_hcv4 245859.2 8216.327 29.92 0.000 229748.2 2619 70.2 yr_hcv5 233697.3 8191.592 28.53 0.000 217634.8 2497 59.8 yr_hcv6 171584.9 7899.448 21.72 0.000 156095.2 1870 74.5 yr_hcv7 131007.3 7330.31 17.87 0.000 116633.7 14538 1 yr_hcv8 54667.04 7580.78 7.21 0.000 39802.27 69531. 82 dc2_b4 -105626 5466.337 -19.32 0.000 -116345 -94907 .2 cvd_b4 -8032.27 11445.24 -0.7 0.483 -30474.7 14410. 14 dm_b4 -44598.8 3884.881 -11.48 0.000 -52216.5 -3698 1.1 dpress_late -18514.3 7308.986 -2.53 0.011 -32846.1 -4182.45 hbv_b4 -25504 4229.048 -6.03 0.000 -33796.5 -17211. 5 hiv_b4 -22908.5 15020.64 -1.53 0.127 -52361.7 6544. 732 heart_b4 -13590.3 5391.578 -2.52 0.012 -24162.3 -30 18.19 psych_b4 4096.716 3715.612 1.1 0.270 -3189.04 11382 .48 anncost_b4 -21234.6 2993.917 -7.09 0.000 -27105.2 15364 biopsy_b4 34476.32 3208.884 10.74 0.000 28184.17 40 768.46 hx_b4 -27465.5 3731.875 -7.36 0.000 -34783.1 -20147 .8 out_cnt 257.1462 62.32477 4.13 0.000 134.9366 379.3 557 fg_cnt -437.296 231.9167 -1.89 0.059 -892.05 17.458 24 gs_cnt -814.7 350.2614 -2.33 0.020 -1501.51 -127.88 9 intern_cnt -1017.16 209.5522 -4.85 0.000 -1428.06 606.257 tt_lab 32037.25 2628.9 12.19 0.000 26882.37 37192.1 3 anti_dpress -8111.76 3978.39 -2.04 0.042 -15912.8 310.734 _cons -33116.6 13725.67 -2.41 0.016 -60030.6 -6202. 59

PAGE 125

125 Table A-2. WTP=$20,000 [Regression with robust standard errors] Linear regression Number of obs = 2712 F( 34, 2677) = 455.28 Prob > F = 0 R-squared = 0.8284 Root MSE = 72627 Table A-2. Continued nmb20 Coef. Robust Std. Err. t P>|t| [95% Conf. Interval] tx 30846.78 3051.888 10.11 0.000 24862.49 36831.08 gender -26709.1 3036.589 -8.8 0.000 -32663.4 -20754 .8 plan2 11961.62 3230.367 3.7 0.000 5627.352 18295.89 plan3 14846.63 4441.379 3.34 0.001 6137.747 23555.5 1 geo1 131806.7 11633.42 11.33 0.000 108995.3 154618. 1 geo2 57326.92 12660.64 4.53 0.000 32501.29 82152.54 geo3 48068.33 12263.61 3.92 0.000 24021.22 72115.44 geo4 22732.16 12295.96 1.85 0.065 -1378.37 46842.7 yr_hcv0 166468.3 13230.87 12.58 0.000 140524.5 1924 12 yr_hcv1 34219.98 14372.88 2.38 0.017 6036.909 62403 .05 yr_hcv2 431578.3 12152.21 35.51 0.000 407749.6 4554 07 yr_hcv3 401601.3 10944.66 36.69 0.000 380140.4 4230 62.1 yr_hcv4 336498.7 10455.03 32.19 0.000 315997.9 3569 99.4 yr_hcv5 321516.6 10416.86 30.87 0.000 301090.7 3419 42.5 yr_hcv6 236227.6 10090.25 23.41 0.000 216442.2 2560 13.1 yr_hcv7 180173.2 9546.064 18.87 0.000 161454.8 1988 91.6 yr_hcv8 75859.32 9730.541 7.8 0.000 56779.19 94939. 46 dc2_b4 -141674 6585.536 -21.51 0.000 -154588 -12876 1 cvd_b4 -3903.61 12001.02 -0.33 0.745 -27435.8 19628 .59 dm_b4 -57360.8 4313.362 -13.3 0.000 -65818.7 -48903 dpress_late -25481.6 7966.455 -3.2 0.001 -41102.7 9860.61 hbv_b4 -35590.6 5027.923 -7.08 0.000 -45449.6 -2573 1.6 hiv_b4 -17915 15688.11 -1.14 0.254 -48677 12847.03 heart_b4 -14339.3 5769.076 -2.49 0.013 -25651.5 -30 26.95 psych_b4 2997.224 4177.419 0.72 0.473 -5194.07 1118 8.52 anncost_b4 -25552.7 3368.349 -7.59 0.000 -32157.5 18947.9 biopsy_b4 45488.52 3757.836 12.1 0.000 38119.96 528 57.07 hx_b4 -36729.6 4187.679 -8.77 0.000 -44941 -28518.2 out_cnt 275.7892 71.04098 3.88 0.000 136.4884 415.0 899 fg_cnt -381.161 253.6577 -1.5 0.133 -878.546 116.22 37 gs_cnt -1079.29 424.6812 -2.54 0.011 -1912.03 -246.558

PAGE 126

126 Table A-2. Continued nmb20 Coef. Robust Std. Err. t P>|t| [95% Conf. Interval] intern_cnt -1265.4 228.7712 -5.53 0.000 -1713.99 -8 16.817 tt_lab 45444 2913.952 15.6 0.000 39730.18 51157.83 anti_dpress -4354.88 4266.119 -1.02 0.307 -12720.1 4010.347 _cons -33864.4 15477.69 -2.19 0.029 -64213.8 -3514. 94 Table A-3.WTP=$30,000 [Regression with robust standard errors] Linear regression Number of obs = 2712 F( 34, 2677) = 638.61 Prob > F = 0 R-squared = 0.8763 Root MSE = 91785 Table A-3. Continued nmb30 Coef. Robust Std. Err. t P>|t| [95% Conf. Interval] tx 58896.94 3850.989 15.29 0.000 51345.73 66448.16 gender -39686.9 3813.648 -10.41 0.000 -47164.9 -322 08.9 plan2 18360.97 4166.253 4.41 0.000 10191.57 26530.3 8 plan3 25710.32 5395.001 4.77 0.000 15131.53 36289.1 1 geo1 176983 13247.61 13.36 0.000 151006.4 202959.6 geo2 60550.39 14553.78 4.16 0.000 32012.6 89088.18 geo3 46751.4 13961.19 3.35 0.001 19375.6 74127.2 geo4 9680.964 14152.33 0.68 0.494 -18069.6 37431.56 yr_hcv0 256267.7 18503.08 13.85 0.000 219986 292549 .5 yr_hcv1 50264.56 20632.84 2.44 0.015 9806.647 90722 .46 yr_hcv2 665575.8 17194.47 38.71 0.000 631860 699291 .5 yr_hcv3 621193 15743.06 39.46 0.000 590323.2 652062 .8 yr_hcv4 517777.6 15178.11 34.11 0.000 488015.6 5475 39.6 yr_hcv5 497155.2 15113.4 32.89 0.000 467520 526790. 3 yr_hcv6 365513.2 14696.11 24.87 0.000 336696.4 3943 30.1 yr_hcv7 278504.8 14129.13 19.71 0.000 250799.7 3062 10 yr_hcv8 118243.9 14239 8.3 0.000 90323.34 146164.4 dc2_b4 -213771 9099.484 -23.49 0.000 -231614 -19592 9 cvd_b4 4353.729 13468.07 0.32 0.747 -22055.1 30762. 6 dm_b4 -82884.9 5375.667 -15.42 0.000 -93425.8 -7234 4 dpress_late -39416.4 9669.947 -4.08 0.000 -58377.7 -20455 hbv_b4 -55763.8 6871.283 -8.12 0.000 -69237.3 -4229 0.2 hiv_b4 -7928.02 17489.89 -0.45 0.650 -42223.1 26367 .05

PAGE 127

127 Table A-3. Continued nmb30 Coef. Robust Std. Err. t P>|t| [95% Conf. Interval] heart_b4 -15837.2 6760.871 -2.34 0.019 -29094.3 -25 80.16 psych_b4 798.2403 5285.823 0.15 0.880 -9566.47 1116 2.95 anncost_b4 -34188.9 4286.413 -7.98 0.000 -42593.9 25783.9 biopsy_b4 67512.92 5042.604 13.39 0.000 57625.13 77 400.71 hx_b4 -55257.8 5307.81 -10.41 0.000 -65665.6 -44850 out_cnt 313.0752 92.60896 3.38 0.001 131.4829 494.6 676 fg_cnt -268.892 307.1223 -0.88 0.381 -871.113 333.3 289 gs_cnt -1608.48 592.9698 -2.71 0.007 -2771.21 -445. 758 intern_cnt -1761.9 277.0762 -6.36 0.000 -2305.2 -12 18.59 tt_lab 72257.5 3619.644 19.96 0.000 65159.92 79355. 08 anti_dpress 3158.897 5024.785 0.63 0.530 -6693.96 1 3011.75 _cons -35360 19685.67 -1.8 0.073 -73960.6 3240.659 Table A-4.WTP=$40,000 [Regression with robust standard errors] Linear regression Number of obs = 2712 F( 34, 2677) = 752.91 Prob > F = 0 R-squared = 0.8938 Root MSE = 1.10E+05 Table A-4. Continued nmb40 Coef. Robust Std. Err. t P>|t| [95% Conf. Interval] tx 86947.1 4766.846 18.24 0.000 77600.03 96294.18 gender -52664.7 4705.84 -11.19 0.000 -61892.1 -4343 7.2 plan2 24760.33 5215.073 4.75 0.000 14534.35 34986.3 1 plan3 36574.02 6508.996 5.62 0.000 23810.85 49337.1 8 geo1 222159.3 15237.66 14.58 0.000 192280.5 252038. 1 geo2 63773.87 16860.24 3.78 0.000 30713.45 96834.28 geo3 45434.46 16046.01 2.83 0.005 13970.65 76898.28 geo4 -3370.24 16408.61 -0.21 0.837 -35545.1 28804.5 9 yr_hcv0 346067.2 23985.72 14.43 0.000 299034.8 3930 99.6 yr_hcv1 66309.13 27016.83 2.45 0.014 13333.17 11928 5.1 yr_hcv2 899573.2 22474.96 40.03 0.000 855503.2 9436 43.3 yr_hcv3 840784.8 20714.68 40.59 0.000 800166.4 8814 03.2 yr_hcv4 699056.6 20034.95 34.89 0.000 659771 738342 .1 yr_hcv5 672793.8 19944.11 33.73 0.000 633686.3 7119 01.2

PAGE 128

128 Table A-4. Continued nmb40 Coef. Robust Std. Err. t P>|t| [95% Conf. Interval] yr_hcv6 494798.8 19423.08 25.47 0.000 456713.1 5328 84.6 yr_hcv7 376836.5 18790.79 20.05 0.000 339990.6 4136 82.4 yr_hcv8 160628.4 18859 8.52 0.000 123648.8 197608.1 dc2_b4 -285868 11782.95 -24.26 0.000 -308973 -26276 4 cvd_b4 12611.07 15291.3 0.82 0.410 -17372.9 42595.0 2 dm_b4 -108409 6598.992 -16.43 0.000 -121349 -95469. 3 dpress_late -53351.1 11697.33 -4.56 0.000 -76287.8 -30414.4 hbv_b4 -75936.9 8869.874 -8.56 0.000 -93329.4 -5854 4.4 hiv_b4 2058.962 19766.73 0.1 0.917 -36700.6 40818.5 6 heart_b4 -17335.2 7966.87 -2.18 0.030 -32957 -1713. 34 psych_b4 -1400.74 6534.09 -0.21 0.830 -14213.1 1141 1.63 anncost_b4 -42825.1 5330.885 -8.03 0.000 -53278.2 32372.1 biopsy_b4 89537.32 6450.685 13.88 0.000 76888.49 10 2186.2 hx_b4 -73786 6585.177 -11.2 0.000 -86698.5 -60873.4 out_cnt 350.3613 117.0995 2.99 0.003 120.7466 579.9 76 fg_cnt -156.623 368.9582 -0.42 0.671 -880.095 566.8 492 gs_cnt -2137.67 772.9349 -2.77 0.006 -3653.28 -622. 062 intern_cnt -2258.39 333.6491 -6.77 0.000 -2912.62 1604.15 tt_lab 99071 4432.473 22.35 0.000 90379.58 107762.4 anti_dpress 10672.67 5947.234 1.79 0.073 -988.967 2 2334.31 _cons -36855.6 24420.09 -1.51 0.131 -84739.8 11028. 55 Table A-5.WTP=$50,000 [Regression with robust standard errors] Linear regression Number of obs = 2712 F( 34, 2677) = 824.66 Prob > F = 0 R-squared = 0.9019 Root MSE = 1.40E+05 Table A-5. Continued nmb50 Coef. Robust Std. Err. t P>|t| [95% Conf. Interval] tx 114997.3 5743.88 20.02 0.000 103734.4 126260.2 gender -65642.5 5658.968 -11.6 0.000 -76738.8 -5454 6.1 plan2 31159.69 6320.856 4.93 0.000 18765.43 43553.9 4 plan3 47437.71 7714.195 6.15 0.000 32311.33 62564.0 9 geo1 267335.6 17475.63 15.3 0.000 233068.5 301602.7 geo2 66997.34 19433.41 3.45 0.001 28891.32 105103.4

PAGE 129

129 Table A-5. Continued nmb50 Coef. Robust Std. Err. t P>|t| [95% Conf. Interval] geo3 44117.53 18386.81 2.4 0.016 8063.741 80171.31 geo4 -16421.4 18922.27 -0.87 0.386 -53525.2 20682.3 1 yr_hcv0 435866.7 29561.93 14.74 0.000 377900.2 4938 33.2 yr_hcv1 82353.71 33453.92 2.46 0.014 16755.58 14795 1.8 yr_hcv2 1133571 27858.54 40.69 0.000 1078944 118819 7 yr_hcv3 1060377 25759.43 41.16 0.000 1009866 111088 7 yr_hcv4 880335.5 24947.55 35.29 0.000 831417.1 9292 53.9 yr_hcv5 848432.3 24830.81 34.17 0.000 799742.8 8971 21.8 yr_hcv6 624084.4 24200.29 25.79 0.000 576631.2 6715 37.5 yr_hcv7 475168.2 23484.28 20.23 0.000 429119 521217 .3 yr_hcv8 203013 23524.91 8.63 0.000 156884.2 249141. 8 dc2_b4 -357965 14542.39 -24.62 0.000 -386481 -32945 0 cvd_b4 20868.4 17358.84 1.2 0.229 -13169.7 54906.5 dm_b4 -133933 7908.969 -16.93 0.000 -149441 -118425 dpress_late -67285.8 13907.68 -4.84 0.000 -94556.7 -40014.9 hbv_b4 -96110.1 10938.94 -8.79 0.000 -117560 -74660 .5 hiv_b4 12045.95 22374.04 0.54 0.590 -31826.2 55918. 1 heart_b4 -18833.2 9304.149 -2.02 0.043 -37077.2 -58 9.107 psych_b4 -3599.73 7855.828 -0.46 0.647 -19003.8 118 04.38 anncost_b4 -51461.4 6440.556 -7.99 0.000 -64090.3 38832.4 biopsy_b4 111561.7 7916.552 14.09 0.000 96038.55 12 7084.9 hx_b4 -92314.2 7944.294 -11.62 0.000 -107892 -76736 .6 out_cnt 387.6474 143.0191 2.71 0.007 107.2083 668.0 864 fg_cnt -44.3535 435.615 -0.1 0.919 -898.529 809.822 3 gs_cnt -2666.86 958.0185 -2.78 0.005 -4545.39 -788. 33 intern_cnt -2754.88 394.9529 -6.98 0.000 -3529.32 1980.43 tt_lab 125884.5 5303.403 23.74 0.000 115485.3 13628 3.7 anti_dpress 18186.44 6968.728 2.61 0.009 4521.807 3 1851.08 _cons -38351.2 29427.97 -1.3 0.193 -96055.1 19352.6 4 Table A-6.WTP=$60,000 [Regression with robust standard errors] Linear regression Number of obs = 2712 F( 34, 2677) = 871.37 Prob > F = 0 R-squared = 0.9064 Root MSE = 1.60E+05

PAGE 130

130 Table A-6. Continued nmb60 Coef. Robust Std. Err. t P>|t| [95% Conf. Interval] tx 143047.4 6755.598 21.17 0.000 129800.7 156294.1 gender -78620.2 6646.87 -11.83 0.000 -91653.7 -6558 6.7 plan2 37559.04 7458.31 5.04 0.000 22934.41 52183.67 plan3 58301.4 8973.927 6.5 0.000 40704.88 75897.93 geo1 312511.9 19877.97 15.72 0.000 273534.2 351489. 6 geo2 70220.81 22180.66 3.17 0.002 26727.85 113713.8 geo3 42800.59 20897.76 2.05 0.041 1823.218 83777.97 geo4 -29472.6 21603.67 -1.36 0.173 -71834.2 12888.9 3 yr_hcv0 525666.2 35187.27 14.94 0.000 456669.2 5946 63.1 yr_hcv1 98398.29 39918.43 2.46 0.014 20124.21 17667 2.4 yr_hcv2 1367568 33295.24 41.07 0.000 1302281 143285 5 yr_hcv3 1279968 30841.44 41.5 0.000 1219493 1340444 yr_hcv4 1061614 29888.44 35.52 0.000 1003008 112022 1 yr_hcv5 1024071 29745.91 34.43 0.000 965743.6 10823 98 yr_hcv6 753370 29002.94 25.98 0.000 696499.5 810240 .4 yr_hcv7 573499.8 28193.72 20.34 0.000 518216.2 6287 83.5 yr_hcv8 245397.6 28213.97 8.7 0.000 190074.2 300720 .9 dc2_b4 -430062 17341.57 -24.8 0.000 -464067 -396058 cvd_b4 29125.73 19593.51 1.49 0.137 -9294.21 67545. 68 dm_b4 -159457 9268.93 -17.2 0.000 -177632 -141282 dpress_late -81220.5 16226.39 -5.01 0.000 -113038 49403 hbv_b4 -116283 13044.99 -8.91 0.000 -141863 -90704 hiv_b4 22032.92 25209.51 0.87 0.382 -27399.2 71465 heart_b4 -20331.1 10723.71 -1.9 0.058 -41358.7 696. 4649 psych_b4 -5798.71 9219.493 -0.63 0.529 -23876.8 122 79.34 anncost_b4 -60097.6 7586.872 -7.92 0.000 -74974.3 45220.8 biopsy_b4 133586.1 9413.246 14.19 0.000 115128.2 15 2044.1 hx_b4 -110842 9349.577 -11.86 0.000 -129176 -92509. 3 out_cnt 424.9334 169.7142 2.5 0.012 92.14926 757.71 76 fg_cnt 67.9157 505.188 0.13 0.893 -922.682 1058.514 gs_cnt -3196.05 1145.743 -2.79 0.005 -5442.68 -949. 419 intern_cnt -3251.37 459.0963 -7.08 0.000 -4151.59 2351.15 tt_lab 152698 6208.03 24.6 0.000 140525 164871 anti_dpress 25700.22 8051.657 3.19 0.001 9912.123 4 1488.31 _cons -39846.8 34590.73 -1.15 0.249 -107674 27980.4 3

PAGE 131

131 Table A-7. WTP=$70,000 [Regression with robust standard errors] Linear regression Number of obs = 2712 F( 34, 2677) = 903.04 Prob > F = 0 R-squared = 0.9091 Root MSE = 1.90E+05 Table A-7. Continued nmb70 Coef. Robust Std. Err. t P>|t| [95% Conf. Interval] tx 171097.6 7788.495 21.97 0.000 155825.5 186369.7 gender -91598 7656.097 -11.96 0.000 -106611 -76585. 6 plan2 43958.4 8614.898 5.1 0.000 27065.87 60850.93 plan3 69165.1 10268.14 6.74 0.000 49030.81 89299.39 geo1 357688.2 22391.82 15.97 0.000 313781.2 401595. 2 geo2 73444.29 25044.77 2.93 0.003 24335.24 122553.3 geo3 41483.66 23524.43 1.76 0.078 -4644.22 87611.54 geo4 -42523.8 24397.55 -1.74 0.081 -90363.8 5316.12 1 yr_hcv0 615465.6 40841.43 15.07 0.000 535381.7 6955 49.6 yr_hcv1 114442.9 46398.91 2.47 0.014 23461.54 20542 4.2 yr_hcv2 1601566 38762.71 41.32 0.000 1525558 167757 3 yr_hcv3 1499560 35944.91 41.72 0.000 1429078 157004 3 yr_hcv4 1242893 34845.57 35.67 0.000 1174566 131122 0 yr_hcv5 1199710 34677.35 34.6 0.000 1131712 1267707 yr_hcv6 882655.6 33820.18 26.1 0.000 816339.2 94897 1.9 yr_hcv7 671831.5 32912.25 20.41 0.000 607295.5 7363 67.5 yr_hcv8 287782.1 32916.28 8.74 0.000 223238.2 35232 6 dc2_b4 -502159 20163.96 -24.9 0.000 -541698 -462621 cvd_b4 37383.08 21944.3 1.7 0.089 -5646.42 80412.57 dm_b4 -184981 10659.76 -17.35 0.000 -205883 -164079 dpress_late -95155.3 18613.01 -5.11 0.000 -131653 58657.9 hbv_b4 -136456 15172.63 -8.99 0.000 -166208 -106705 hiv_b4 32019.91 28204.4 1.14 0.256 -23284.7 87324.5 2 heart_b4 -21829.1 12196.85 -1.79 0.074 -45745.3 208 7.103 psych_b4 -7997.69 10608.93 -0.75 0.451 -28800.2 128 04.83 anncost_b4 -68733.8 8755.452 -7.85 0.000 -85901.9 51565.7 biopsy_b4 155610.5 10928.11 14.24 0.000 134182.1 17 7038.9 hx_b4 -129371 10782.99 -12 0.000 -150514 -108227 out_cnt 462.2195 196.8696 2.35 0.019 76.18764 848.2 513 fg_cnt 180.185 576.6225 0.31 0.755 -950.486 1310.85 6

PAGE 132

132 Table A-7. Continued nmb70 Coef. Robust Std. Err. t P>|t| [95% Conf. Interval] gs_cnt -3725.24 1334.994 -2.79 0.005 -6342.96 -1107 .52 intern_cnt -3747.86 525.0396 -7.14 0.000 -4777.38 2718.34 tt_lab 179511.5 7133.545 25.16 0.000 165523.7 19349 9.3 anti_dpress 33213.99 9174.292 3.62 0.000 15224.58 5 1203.41 _cons -41342.4 39848.22 -1.04 0.300 -119479 36793.9 8 Table A-8. WTP=$80,000 [Regression with robust standard errors]: Linear regression Number of obs = 2712 F( 34, 2677) = 925.33 Prob > F = 0 R-squared = 0.9108 Root MSE = 2.10E+05 Table A-8. Continued nmb80 Coef. Robust Std. Err. t P>|t| [95% Conf. Interval] tx 199147.7 8835.148 22.54 0.000 181823.3 216472.1 gender -104576 8679.212 -12.05 0.000 -121594 -87557 .2 plan2 50357.76 9783.837 5.15 0.000 31173.11 69542.4 plan3 80028.8 11585.29 6.91 0.000 57311.78 102745.8 geo1 402864.5 24983.56 16.13 0.000 353875.5 451853. 5 geo2 76667.76 27989.89 2.74 0.006 21783.76 131551.7 geo3 40166.72 26232.08 1.53 0.126 -11270.5 91603.9 geo4 -55575 27269.38 -2.04 0.042 -109046 -2103.87 yr_hcv0 705265.1 46513.9 15.16 0.000 614058.3 79647 1.9 yr_hcv1 130487.4 52889.47 2.47 0.014 26779.09 23419 5.8 yr_hcv2 1835563 44249.55 41.48 0.000 1748796 192233 0 yr_hcv3 1719152 41061.83 41.87 0.000 1638636 179966 8 yr_hcv4 1424172 39812.89 35.77 0.000 1346105 150223 9 yr_hcv5 1375348 39619.01 34.71 0.000 1297661 145303 5 yr_hcv6 1011941 38646.56 26.18 0.000 936161 1087721 yr_hcv7 770163.2 37636.46 20.46 0.000 696363.7 8439 62.6 yr_hcv8 330166.7 37626.88 8.77 0.000 256386 403947. 4 dc2_b4 -574256 23001 -24.97 0.000 -619358 -529155 cvd_b4 45640.42 24377.65 1.87 0.061 -2160.51 93441. 34 dm_b4 -210505 12070.8 -17.44 0.000 -234174 -186836 dpress_late -109090 21044.45 -5.18 0.000 -150355 -6 7825 hbv_b4 -156630 17313.9 -9.05 0.000 -190580 -122680

PAGE 133

133 Table A-8. Continued nmb80 Coef. Robust Std. Err. t P>|t| [95% Conf. Interval] hiv_b4 42006.89 31313 1.34 0.180 -19393.2 103407 heart_b4 -23327.1 13706.3 -1.7 0.089 -50203.1 3548. 957 psych_b4 -10196.7 12015.2 -0.85 0.396 -33756.7 1336 3.33 anncost_b4 -77370 9938.444 -7.78 0.000 -96857.8 -57 882.2 biopsy_b4 177634.9 12454.51 14.26 0.000 153213.5 20 2056.4 hx_b4 -147899 12234.65 -12.09 0.000 -171889 -123909 out_cnt 499.5055 224.3182 2.23 0.026 59.65112 939.3 599 fg_cnt 292.4541 649.3045 0.45 0.652 -980.735 1565.6 43 gs_cnt -4254.43 1525.204 -2.79 0.005 -7245.13 -1263 .73 intern_cnt -4244.35 592.1818 -7.17 0.000 -5405.53 3083.17 tt_lab 206325 8072.768 25.56 0.000 190495.5 222154. 5 anti_dpress 40727.76 10323.69 3.95 0.000 20484.55 6 0970.97 _cons -42838 45167.38 -0.95 0.343 -131405 45728.45

PAGE 134

134 APPENDIX B NET BENEFIT OF INITIAL COMBINATION ANTIVIRAL THERAPY IN U SUAL CARE ANALYSIS Patients with Cirrhosis Table B-1.WTP=$60,000 [Regression with robust standard errors] Linear regression Number of obs = 688 F( 29, 658) = 679.56 Prob > F = 0 R-squared = 0.9395 Root MSE = 1.60E+05 Table B-1. Continued Robust Std. Err. nmb60 Coef. t P>|t| [95% Conf. Interval] tx 24553.18 14449.28 1.7 0.090 -3819.08 52925.44 age_hcv 7765.249 771.3688 10.07 0.000 6250.608 9279 .89 gender -97516.9 11840.45 -8.24 0.000 -120767 -74267 .3 plan2 2424.332 13296.06 0.18 0.855 -23683.5 28532.1 6 plan3 -14988.1 19655.15 -0.76 0.446 -53582.5 23606. 28 geo1 751448 77222.32 9.73 0.000 599816.1 903079.9 geo2 709134.5 79850.31 8.88 0.000 552342.4 865926.6 geo3 596005.5 79044.08 7.54 0.000 440796.5 751214.6 geo4 473508.8 85739.56 5.52 0.000 305152.6 641864.9 yr_hcv2 1846462 32003.91 57.69 0.000 1783620 190930 4 yr_hcv3 1497774 33430.19 44.8 0.000 1432131 1563416 yr_hcv4 1273787 32053.95 39.74 0.000 1210847 133672 8 yr_hcv5 1123307 27492.58 40.86 0.000 1069323 117729 0 yr_hcv6 627206.1 26622.28 23.56 0.000 574931.2 6794 81 yr_hcv7 282946.5 24303.12 11.64 0.000 235225.5 3306 67.5 cvd_b4 -163934 43714.5 -3.75 0.000 -249771 -78097.2 ddpend_b4 -167609 54657.07 -3.07 0.002 -274933 -602 86 dpress_late -81898.2 24472.09 -3.35 0.001 -129951 33845.4 hbv_b4 -230982 37899.76 -6.09 0.000 -305401 -156563 hiv_b4 345897.5 58768.04 5.89 0.000 230502 461293 heart_b4 -84579.6 21986.24 -3.85 0.000 -127751 -414 08 anncost_b4 -218698 14964.41 -14.61 0.000 -248082 -1 89314 biopsy_b4 -122201 15581.73 -7.84 0.000 -152797 -916 05.3 hx_b4 -3656.45 22680.72 -0.16 0.872 -48191.8 40878. 85

PAGE 135

135 Table B-1.Continued Robust Std. Err. nmb60 Coef. t P>|t| [95% Conf. Interval] er_cnt 15310.89 4223.963 3.62 0.000 7016.819 23604. 96 gs_cnt -10777.3 1182.831 -9.11 0.000 -13099.9 -8454 .76 inf_cnt 13641.2 4361.463 3.13 0.002 5077.136 22205. 26 tt_lab -22027.4 13997.64 -1.57 0.116 -49512.8 5458. 081 anti_dpress 8039.544 15049.86 0.53 0.593 -21512 375 91.09 _cons -195392 93331.24 -2.09 0.037 -378655 -12128.9 Table B-2.WTP=$70,000 [Regression with robust standard errors] Linear regression Number of obs = 688 F( 29, 658) = 704.02 Prob > F = 0 R-squared = 0.9407 Root MSE = 1.80E+05 Table B-2. Continued nmb70 Coef. Robust Std. Err. t P>|t| [95% Conf. Interval] tx 34154.09 16795.46 2.03 0.042 1174.934 67133.24 age_hcv 9068.86 878.7297 10.32 0.000 7343.408 10794 .31 gender -112865 13692.65 -8.24 0.000 -139751 -85978. 1 plan2 2209.612 15339.71 0.14 0.886 -27911.1 32330.3 plan3 -18761.9 22911.23 -0.82 0.413 -63749.9 26226. 03 geo1 858748 89934.71 9.55 0.000 682154.4 1035342 geo2 808519.5 93146.84 8.68 0.000 625618.6 991420.4 geo3 676598.9 92047.61 7.35 0.000 495856.4 857341.3 geo4 533988 99782.48 5.35 0.000 338057.5 729918.5 yr_hcv2 2158136 37052.51 58.25 0.000 2085380 223089 1 yr_hcv3 1753124 38877.44 45.09 0.000 1676786 182946 3 yr_hcv4 1491513 37275.52 40.01 0.000 1418320 156470 7 yr_hcv5 1315166 31917.77 41.2 0.000 1252493 1377839 yr_hcv6 735039.8 30839.07 23.83 0.000 674485 795594 .7 yr_hcv7 330720.2 28252.65 11.71 0.000 275243.9 3861 96.4 cvd_b4 -183320 49300.35 -3.72 0.000 -280125 -86515. 1 ddpend_b4 -181639 58512.37 -3.1 0.002 -296533 -6674 5.9 dpress_late -95670.7 28006.91 -3.42 0.001 -150664 40677 hbv_b4 -269201 44364.02 -6.07 0.000 -356313 -182089 hiv_b4 417507.3 66161.9 6.31 0.000 287593.4 547421. 2 heart_b4 -95706.5 25384.01 -3.77 0.000 -145550 -458 63 anncost_b4 -254427 17367.86 -14.65 0.000 -288531 -2 20324

PAGE 136

136 Table B-2. Continued nmb70 Coef. Robust Std. Err. t P>|t| [95% Conf. Interval] biopsy_b4 -141851 17999.63 -7.88 0.000 -177195 -106 507 hx_b4 -4400.33 26364.36 -0.17 0.867 -56168.8 47368. 1 er_cnt 18668.84 4929.514 3.79 0.000 8989.368 28348. 32 gs_cnt -12573.8 1365.31 -9.21 0.000 -15254.7 -9892. 91 inf_cnt 15661.32 5065.108 3.09 0.002 5715.598 25607 .05 tt_lab -23118.1 16219.07 -1.43 0.155 -54965.5 8729. 229 anti_dpress 12863.32 17287.15 0.74 0.457 -21081.3 4 6807.96 _cons -212157 108340.9 -1.96 0.051 -424893 578.4189 Table B-3.WTP=$80,000 [Regression with robust standard errors]: Linear regression Number of obs = 688 F( 29, 658) = 721.16 Prob > F = 0 R-squared = 0.9413 Root MSE = 2.10E+05 Table B-3. Continued nmb80 Coef. Robust Std. Err. t P>|t| [95% Conf. Interval] tx 43754.97 19171.06 2.28 0.023 6111.142 81398.81 age_hcv 10372.47 989.5831 10.48 0.000 8429.348 1231 5.59 gender -128212 15573.27 -8.23 0.000 -158792 -97633. 1 plan2 1994.913 17418.34 0.11 0.909 -32207.3 36197.1 5 plan3 -22535.7 26213.34 -0.86 0.390 -74007.6 28936. 16 geo1 966048 102890.1 9.39 0.000 764015.4 1168081 geo2 907904.5 106671.3 8.51 0.000 698447.3 1117362 geo3 757192.2 105307.7 7.19 0.000 550412.6 963971.7 geo4 594467.3 114070.3 5.21 0.000 370481.6 818453 yr_hcv2 2469809 42145.07 58.6 0.000 2387054 2552564 yr_hcv3 2008475 44376.03 45.26 0.000 1921339 209561 1 yr_hcv4 1709239 42546.7 40.17 0.000 1625696 1792783 yr_hcv5 1507025 36387.25 41.42 0.000 1435576 157847 4 yr_hcv6 842873.5 35105.52 24.01 0.000 773941.2 9118 05.9 yr_hcv7 378493.9 32235.47 11.74 0.000 315197.1 4417 90.6 cvd_b4 -202706 55187.34 -3.67 0.000 -311071 -94341. 7 ddpend_b4 -195669 63313.89 -3.09 0.002 -319991 -713 47.8 dpress_late -109443 31675.21 -3.46 0.001 -171640 -4 7246.6 hbv_b4 -307421 50879.28 -6.04 0.000 -407326 -207515 hiv_b4 489117.1 74143.35 6.6 0.000 343531 634703.1

PAGE 137

137 Table B-3. Continued nmb80 Coef. Robust Std. Err. t P>|t| [95% Conf. Interval] heart_b4 -106833 28908 -3.7 0.000 -163596 -50070.2 anncost_b4 -290157 19791.44 -14.66 0.000 -329019 -2 51295 biopsy_b4 -161501 20442.66 -7.9 0.000 -201641 -1213 60 hx_b4 -5144.19 30104.94 -0.17 0.864 -64257.5 53969. 14 er_cnt 22026.79 5657.764 3.89 0.000 10917.35 33136. 24 gs_cnt -14370.3 1558.136 -9.22 0.000 -17429.8 -1131 0.7 inf_cnt 17681.45 5777.268 3.06 0.002 6337.342 29025 .55 tt_lab -24208.9 18475.3 -1.31 0.191 -60486.6 12068. 73 anti_dpress 17687.11 19569.97 0.9 0.366 -20740 5611 4.23 _cons -228923 123564.5 -1.85 0.064 -471551 13705.59

PAGE 138

138 APPENDIX C COVARIATES EFFECTS ON THE INB OF ANTIVIRAL THERAPY IN BASE CASE ANALYSIS Patients with Cirrhosis Table C-1. WTP=$15,000 in multivariate adjusted model Source DF Sum of Squares Mean Square F Value Pr > F Model 65 3.73E+13 5.7E+11 139.92 <.0001 Error 2646 1.09E+13 4.1E+09 Corrected Total 2711 4.82E+13 Table C-1. Continued R-Square Coeff Var Root MSE nmb15 Mean 0.774629 31.05834 64078.62 206317 Table C-1. Continued Parameter Estimate Standard Error t Value Pr > |t| Intercept -36469.3677 12438.10721 -2.93 0.0034 tx 9204.1647 25610.3859 0.36 0.7193 gender -18017.7124 3298.44668 -5.46 <.0001 plan2 8288.1011 3733.40191 2.22 0.0265 plan3 5953.159 4168.86551 1.43 0.1534 geo1 113456.6709 6804.35439 16.67 <.0001 geo2 63050.7942 9039.53957 6.98 <.0001 geo3 55463.8981 7730.71576 7.17 <.0001 geo4 32147.5459 8724.19127 3.68 0.0002 Yr_HCV0 118007.3069 12980.71633 9.09 <.0001 Yr_HCV1 23287.5459 14774.22977 1.58 0.1151 Yr_HCV2 325084.8049 10391.29195 31.28 <.0001 Yr_HCV3 301823.0893 10391.71037 29.04 <.0001 Yr_HCV4 234707.5723 10010.37824 23.45 <.0001 Yr_HCV5 225987.1133 10388.70709 21.75 <.0001 Yr_HCV6 163253.2482 10003.51003 16.32 <.0001 Yr_HCV7 126093.8127 9308.54005 13.55 <.0001 Yr_HCV8 47868.1947 9497.74716 5.04 <.0001 TT_lab 32811.8868 3348.13885 9.8 <.0001 anti_dpress -572.6307 5150.20931 -0.11 0.9115 dc2_b4 -98775.8346 5406.42585 -18.27 <.0001 CVD_b4 -16583.1798 7852.21672 -2.11 0.0348 dm_b4 -42241.8645 4237.48045 -9.97 <.0001

PAGE 139

139 Table C-1. Continued Parameter Estimate Standard Error t Value Pr > |t| dpress_late -24766.7491 7495.66356 -3.3 0.001 HBV_b4 -23122.2943 6020.84308 -3.84 0.0001 HIV_b4 -10345.5297 9979.47786 -1.04 0.3 Heart_b4 -11934.4791 4943.55556 -2.41 0.0158 Psych_b4 5247.1489 4276.28642 1.23 0.2199 anncost_b4 -26565.2432 4082.80224 -6.51 <.0001 Biopsy_b4 36111.7884 5908.52153 6.11 <.0001 hx_b4 -23284.7067 4253.11438 -5.47 <.0001 out_cnt 264.2758 58.93145 4.48 <.0001 FG_cnt -387.6183 201.21125 -1.93 0.0542 GS_cnt -744.6424 252.19984 -2.95 0.0032 Intern_cnt -1083.7346 162.28813 -6.68 <.0001 tx*gender -5402.9622 5535.79027 -0.98 0.3292 tx*plan2 1747.3051 6157.53147 0.28 0.7766 tx*plan3 11150.8333 7092.06624 1.57 0.116 tx*geo1 -9480.4046 11636.53077 -0.81 0.4153 tx*geo2 -20012.3084 14938.17917 -1.34 0.1805 tx*geo3 -15389.2383 13255.09511 -1.16 0.2457 tx*geo4 -7748.0757 15394.85756 -0.5 0.6148 tx*Yr_HCV0 0 tx*Yr_HCV1 0 tx*Yr_HCV2 -19006.3496 23472.91247 -0.81 0.4182 tx*Yr_HCV3 -9064.794 22887.3573 -0.4 0.6921 tx*Yr_HCV4 45943.8304 22364.45175 2.05 0.04 tx*Yr_HCV5 36244.7018 22588.63053 1.6 0.1087 tx*Yr_HCV6 40262.42 22296.51948 1.81 0.0711 tx*Yr_HCV7 31253.1163 21442.88246 1.46 0.1451 tx*Yr_HCV8 35264.0971 21706.1946 1.62 0.1044 tx*TT_lab -8266.7883 5731.10679 -1.44 0.1493 tx*anti_dpress -12950.6725 6967.7045 -1.86 0.0632 tx*dc2_b4 -30947.0551 11959.51191 -2.59 0.0097 tx*CVD_b4 22557.0877 14897.37978 1.51 0.1301 tx*dm_b4 -7888.1295 7450.70302 -1.06 0.2898 tx*dpress_late 13658.6629 10638.68935 1.28 0.1993 tx*HBV_b4 -9648.5506 13287.67281 -0.73 0.4678 tx*HIV_b4 -37299.5449 18270.05731 -2.04 0.0413 tx*Heart_b4 -1711.9281 8357.14464 -0.2 0.8377 tx*Psych_b4 -6987.1078 7235.11436 -0.97 0.3343

PAGE 140

140 Table C-1. Continued Parameter Estimate Standard Error t Value Pr > |t| tx*anncost_b4 9989.9821 6214.74627 1.61 0.1081 tx*Biopsy_b4 -4411.2926 8054.42177 -0.55 0.584 tx*hx_b4 -9575.7569 7592.0073 -1.26 0.2073 tx*out_cnt -6.3201 124.66759 -0.05 0.9596 tx*FG_cnt 52.2929 389.45439 0.13 0.8932 tx*GS_cnt -220.7435 583.56762 -0.38 0.7053 tx*Intern_cnt 289.9506 311.4351 0.93 0.3519 Table C-2. WTP=$20,000 in multivariate adjusted model Source DF Sum of Squares Mean Square F Value Pr > F Model 65 6.8620811E13 1.0557048E12 204.50 <.0001 Error 2646 1.3659916E13 5162477545.6 Corrected Total 2711 8.2280726E13 Table C-2. Continued R-Square Coeff Var Root MSE nmb20 Mean 0.833984 25.02755 71850.38 287085.2 Table C-2. Continued Parameter Estimate Standard Error t Value Pr > |t | Intercept -36459.6073 13946.66054 -2.61 0.0090 tx 14211.0852 28716.53640 0.49 0.6207 gender -24239.5692 3698.49812 -6.55 <.0001 plan2 11503.1705 4186.20680 2.75 0.0060 plan3 11096.6596 4674.48553 2.37 0.0177 geo1 136540.0685 7629.61914 17.90 <.0001 geo2 65679.8027 10135.89831 6.48 <.0001 geo3 56151.0881 8668.33406 6.48 <.0001 geo4 26643.9191 9782.30305 2.72 0.0065 Yr_HCV0 160146.4826 14555.07990 11.00 <.0001

PAGE 141

141 Table C-2. Continued Parameter Estimate Standard Error t Value Pr > |t | Yr_HCV1 28958.5586 16566.11926 1.75 0.0806 Yr_HCV2 440710.9484 11651.59770 37.82 <.0001 Yr_HCV3 409043.3662 11652.06687 35.10 <.0001 Yr_HCV4 321306.2793 11224.48495 28.63 <.0001 Yr_HCV5 309705.9957 11648.69934 26.59 <.0001 Yr_HCV6 224452.6458 11216.78373 20.01 <.0001 Yr_HCV7 173094.7787 10437.52446 16.58 <.0001 Yr_HCV8 67388.4747 10649.67951 6.33 <.0001 TT_lab 46253.1894 3754.21720 12.32 <.0001 anti_dpress 4496.6294 5774.85141 0.78 0.4363 dc2_b4 -133141.2397 6062.14312 -21.96 <.0001 CVD_b4 -13743.7483 8804.57124 -1.56 0.1186 dm_b4 -53362.1098 4751.42241 -11.23 <.0001 dpress_late -31892.4821 8404.77361 -3.79 0.0002 HBV_b4 -32289.7341 6751.07982 -4.78 <.0001 HIV_b4 -5923.5452 11189.83683 -0.53 0.5966 Heart_b4 -12737.2042 5543.13370 -2.30 0.0216 Psych_b4 4092.6355 4794.93496 0.85 0.3934 anncost_b4 -30993.2979 4577.98409 -6.77 <.0001 Biopsy_b4 47044.3960 6625.13537 7.10 <.0001 hx_b4 -32361.2336 4768.95251 -6.79 <.0001 out_cnt 286.4273 66.07894 4.33 <.0001 FG_cnt -326.3682 225.61511 -1.45 0.1481 GS_cnt -999.1202 282.78785 -3.53 0.0004 Intern_cnt -1301.6706 181.97121 -7.15 <.0001

PAGE 142

142 Table C-2. Continued Parameter Estimate Standard Error t Value Pr > |t | tx*gender -6183.0086 6207.19748 -1.00 0.3193 tx*plan2 1369.8456 6904.34643 0.20 0.8427 tx*plan3 12583.5437 7952.22606 1.58 0.1137 tx*geo1 -11448.1062 13047.86506 -0.88 0.3804 tx*geo2 -23708.0700 16749.95322 -1.42 0.1571 tx*geo3 -19733.5520 14862.73665 -1.33 0.1844 tx*geo4 -11364.3832 17262.01976 -0.66 0.5104 tx*Yr_HCV0 0.0000 B tx*Yr_HCV1 0.0000 B tx*Yr_HCV2 -7042.0726 26319.81993 -0.27 0.7891 tx*Yr_HCV3 6274.8907 25663.24582 0.24 0.8069 tx*Yr_HCV4 65095.8987 25076.91977 2.60 0.0095 tx*Yr_HCV5 55034.2762 25328.28804 2.17 0.0299 tx*Yr_HCV6 57884.1436 25000.74835 2.32 0.0207 tx*Yr_HCV7 45359.7958 24043.57814 1.89 0.0593 tx*Yr_HCV8 47595.8873 24338.82604 1.96 0.0506 tx*TT_lab -8823.8716 6426.20292 -1.37 0.1698 tx*anti_dpress -15809.3471 7812.78114 -2.02 0.043 1 tx*dc2_b4 -40044.8418 13410.01890 -2.99 0.0029 tx*CVD_b4 26437.9191 16704.20549 1.58 0.1136 tx*dm_b4 -13236.8288 8354.36004 -1.58 0.1132 tx*dpress_late 14896.1648 11929.00065 1.25 0.2119 tx*HBV_b4 -13672.3772 14899.26553 -0.92 0.3589 tx*HIV_b4 -34539.4245 20485.93753 -1.69 0.0919 tx*Heart_b4 -1194.5661 9370.73924 -0.13 0.8986

PAGE 143

143 Table C-2. Continued Parameter Estimate Standard Error t Value Pr > |t| tx*Psych_b4 -7596.9544 8112.62375 -0.94 0.3491 tx*anncost_b4 9785.5729 6968.50052 1.40 0.1604 tx*Biopsy_b4 -4688.9641 9031.30069 -0.52 0.6037 tx*hx_b4 -11019.3758 8512.80238 -1.29 0.1956 tx*out_cnt -12.0880 139.78787 -0.09 0.9311 tx*FG_cnt 28.5072 436.68929 0.07 0.9480 tx*GS_cnt -214.3022 654.34550 -0.33 0.7433 tx*Intern_cnt 181.6982 349.20744 0.52 0.6029 Table C-3.WTP=$30,000 in multivariate adjusted model Source DF Sum of Squares Mean Square F Value Pr > F Model 65 1.6054228E14 2.4698812E12 299.66 <.0001 Error 2646 2.180933E13 8242377217.9 Corrected Total 2711 1.8235161E14 Table C-3. Continued R-Square Coeff Var Root MSE nmb30 Mean 0.880400 20.23700 90787.54 448621.6 Table C-3. Continued Parameter Estimate Standard Error t Value Pr > |t | Intercept -36440.0865 17622.49458 -2.07 0.0388 tx 24224.9261 36285.17420 0.67 0.5044 gender -36683.2827 4673.28813 -7.85 <.0001 plan2 17933.3093 5289.53914 3.39 0.0007 plan3 21383.6609 5906.51042 3.62 0.0003 geo1 182706.8638 9640.51011 18.95 <.0001

PAGE 144

144 Table C-3. Continued Parameter Estimate Standard Error t Value Pr > |t | geo2 70937.8196 12807.35360 5.54 <.0001 geo3 57525.4683 10952.99262 5.25 <.0001 geo4 15636.6655 12360.56345 1.27 0.2060 Yr_HCV0 244424.8339 18391.27122 13.29 <.0001 Yr_HCV1 40300.5840 20932.34764 1.93 0.0543 Yr_HCV2 671963.2355 14722.53640 45.64 <.0001 Yr_HCV3 623483.9199 14723.12922 42.35 <.0001 Yr_HCV4 494503.6934 14182.85221 34.87 <.0001 Yr_HCV5 477143.7604 14718.87413 32.42 <.0001 Yr_HCV6 346851.4408 14173.12123 24.47 <.0001 Yr_HCV7 267096.7107 13188.47746 20.25 <.0001 Yr_HCV8 106429.0346 13456.54890 7.91 <.0001 TT_lab 73135.7945 4743.69274 15.42 <.0001 anti_dpress 14635.1496 7296.89286 2.01 0.0450 dc2_b4 -201872.0499 7659.90424 -26.35 <.0001 CVD_b4 -8064.8852 11125.13699 -0.72 0.4686 dm_b4 -75602.6005 6003.72508 -12.59 <.0001 dpress_late -46143.9480 10619.96719 -4.35 <.0001 HBV_b4 -50624.6139 8530.41968 -5.93 <.0001 HIV_b4 2920.4239 14139.07210 0.21 0.8364 Heart_b4 -14342.6546 7004.10276 -2.05 0.0407 Psych_b4 1783.6085 6058.70596 0.29 0.7685 anncost_b4 -39849.4073 5784.57470 -6.89 <.0001 Biopsy_b4 68909.6113 8371.28084 8.23 <.0001 hx_b4 -50514.2876 6025.87547 -8.38 <.0001

PAGE 145

145 Table C-3. Continued Parameter Estimate Standard Error t Value Pr > |t | out_cnt 330.7304 83.49496 3.96 <.0001 FG_cnt -203.8680 285.07907 -0.72 0.4746 GS_cnt -1508.0757 357.32047 -4.22 <.0001 Intern_cnt -1737.5426 229.93223 -7.56 <.0001 tx*gender -7743.1015 7843.18968 -0.99 0.3236 tx*plan2 614.9267 8724.08181 0.07 0.9438 tx*plan3 15448.9646 10048.14452 1.54 0.1243 tx*geo1 -15383.5094 16486.80921 -0.93 0.3509 tx*geo2 -31099.5934 21164.63358 -1.47 0.1418 tx*geo3 -28422.1794 18780.01514 -1.51 0.1303 tx*geo4 -18596.9982 21811.66229 -0.85 0.3939 tx*Yr_HCV0 0.0000 B tx*Yr_HCV1 0.0000 B tx*Yr_HCV2 16886.4813 33256.77017 0.51 0.6117 tx*Yr_HCV3 36954.2600 32427.14692 1.14 0.2546 tx*Yr_HCV4 103400.0354 31686.28659 3.26 0.0011 tx*Yr_HCV5 92613.4251 32003.90642 2.89 0.0038 tx*Yr_HCV6 93127.5909 31590.03915 2.95 0.0032 tx*Yr_HCV7 73573.1547 30380.59357 2.42 0.0155 tx*Yr_HCV8 72259.4678 30753.65811 2.35 0.0189 tx*TT_lab -9938.0382 8119.91701 -1.22 0.2211 tx*anti_dpress -21526.6963 9871.94697 -2.18 0.029 3 tx*dc2_b4 -58240.4152 16944.41367 -3.44 0.0006

PAGE 146

146 Table C-3. Continued Parameter Estimate Standard Error t Value Pr > |t | tx*CVD_b4 34199.5817 21106.82840 1.62 0.1053 tx*dm_b4 -23934.2273 10556.26643 -2.27 0.0235 tx*dpress_late 17371.1686 15073.05270 1.15 0.2492 tx*HBV_b4 -21720.0302 18826.17171 -1.15 0.2487 tx*HIV_b4 -29019.1839 25885.28789 -1.12 0.2624 tx*Heart_b4 -159.8423 11840.52634 -0.01 0.9892 tx*Psych_b4 -8816.6475 10250.81723 -0.86 0.3898 tx*anncost_b4 9376.7545 8805.14460 1.06 0.2870 tx*Biopsy_b4 -5244.3071 11411.62410 -0.46 0.6459 tx*hx_b4 -13906.6136 10756.46843 -1.29 0.1962 tx*out_cnt -23.6239 176.63089 -0.13 0.8936 tx*FG_cnt -19.0642 551.78475 -0.03 0.9724 tx*GS_cnt -201.4195 826.80725 -0.24 0.8076 tx*Intern_cnt -34.8065 441.24586 -0.08 0.9371

PAGE 147

147 APPENDIX D COVARIATES EFFECTS ON THE INB OF ANTIVIRAL THERAPY IN USUA L CARE ANALYSIS Patients with Cirrhosis Table D-1.WTP=$60,000 in multivariate adjusted model Source DF Sum of Squares Mean Square F Value Pr > F Model 55 2.579578E14 4.6901419E12 187.42 <.0001 Error 632 1.5815912E13 25025177897 Corrected Total 687 2.7377372E14 Table D-1. Continued R-Square Coeff Var Root MSE nmb60 Mean 0.942230 9.899500 158193.5 1597995 Table D-1.Continued Parameter Estimate Standard Error t Value Pr > |t| Intercept -259711.502 80461.0874 -3.23 0.0013 tx 203740.419 138586.3807 1.47 0.1420 age_HCV 7593.419 1028.1910 7.39 <.0001 gender -96599.750 16533.8761 -5.84 <.0001 plan2 6348.906 17791.1090 0.36 0.7213 plan3 -12813.356 22291.6185 -0.57 0.5656 geo1 778069.450 39313.3358 19.79 <.0001 geo2 719749.530 49664.2000 14.49 <.0001 geo3 634346.631 46100.4208 13.76 <.0001 geo4 528966.435 52746.5169 10.03 <.0001 Yr_HCV2 1888435.362 48331.4487 39.07 <.0001 Yr_HCV3 1558921.691 48070.3645 32.43 <.0001 Yr_HCV4 1335553.608 45851.7925 29.13 <.0001

PAGE 148

148 Table D-1.Continued Parameter Estimate Standard Error t Value Pr > |t| Yr_HCV5 1162222.878 48353.7954 24.04 <.0001 Yr_HCV6 673258.104 45836.3536 14.69 <.0001 Yr_HCV7 319770.258 43963.8600 7.27 <.0001 TT_lab -7577.496 17016.7303 -0.45 0.6563 anti_dpress -999.866 21856.4325 -0.05 0.9635 CVD_b4 -125584.067 39756.5540 -3.16 0.0017 Ddpend_b4 -259707.002 73149.3149 -3.55 0.0004 dpress_late -82604.096 35601.0061 -2.32 0.0206 HBV_b4 -247787.557 30055.4557 -8.24 <.0001 HIV_b4 348971.477 43645.8483 8.00 <.0001 Heart_b4 -79373.805 19895.2364 -3.99 <.0001 anncost_b4 -218197.255 19242.4886 -11.34 <.0001 Biopsy_b4 -138893.642 29607.3173 -4.69 <.0001 hx_b4 -13724.934 22981.9157 -0.60 0.5506 er_cnt 12100.671 4030.0518 3.00 0.0028 GS_cnt -11282.810 1653.1250 -6.83 <.0001 INF_cnt 13066.204 2378.0170 5.49 <.0001 tx*age_HCV 45.725 1852.8664 0.02 0.9803 tx*gender -6165.051 27573.2092 -0.22 0.8231 tx*plan2 -8532.278 29508.7688 -0.29 0.7726 tx*plan3 -2602.547 36455.5171 -0.07 0.9431 tx*geo1 -76328.842 65369.9483 -1.17 0.2434 tx*geo2 -24840.183 85391.8975 -0.29 0.7712 tx*geo3 -97961.991 75736.6598 -1.29 0.1963 tx*geo4 -136104.490 91500.2253 -1.49 0.1374

PAGE 149

149 Table D-1.Continued Parameter Estimate Standard Error t Value Pr > |t| tx*Yr_HCV2 -126560.769 75475.1684 -1.68 0.0941 tx*Yr_HCV3 -146770.633 68895.5328 -2.13 0.0335 tx*Yr_HCV4 -155103.517 67504.5314 -2.30 0.0219 tx*Yr_HCV5 -98926.618 69293.0868 -1.43 0.1539 tx*Yr_HCV6 -110989.617 67399.1787 -1.65 0.1001 tx*Yr_HCV7 -81946.379 62609.8711 -1.31 0.1911 tx*TT_lab -36913.225 29067.1091 -1.27 0.2046 tx*anti_dpress 20711.654 32213.4639 0.64 0.5205 tx*CVD_b4 -70615.031 85111.5195 -0.83 0.4070 tx*Ddpend_b4 170790.707 100510.2050 1.70 0.0898 tx*dpress_late -3182.146 51420.6915 -0.06 0.9507 tx*HBV_b4 130599.804 79808.3612 1.64 0.1023 tx*anncost_b4 -12149.771 30485.0424 -0.40 0.6904 tx*Biopsy_b4 29230.771 39401.1151 0.74 0.4584 tx*hx_b4 39091.311 42700.6332 0.92 0.3603 tx*er_cnt 16448.168 9118.6581 1.80 0.0717 tx*GS_cnt 440.567 2574.5997 0.17 0.8642 tx*INF_cnt 9479.151 8328.5263 1.14 0.2555 Table D-2.WTP=$70,000 in multivariate adjusted model Source DF Sum of Squares Mean Square F Value Pr > F Model 55 3.5171446E14 6.3948083E12 191.58 <.0001 Error 632 2.1095309E13 33378654023 Corrected Total 687 3.7280977E14

PAGE 150

150 Table D-2. Continued R-Square Coeff Var Root MSE nmb70 Mean 0.943415 9.751631 182698.3 1873515 Table D-2. Continued Parameter Estimate Standard Error t Value Pr > |t| Intercept -284491.261 92924.8184 -3.06 0.0023 tx 240611.500 160053.9425 1.50 0.1333 age_HCV 8833.757 1187.4617 7.44 <.0001 gender -112091.059 19095.0369 -5.87 <.0001 plan2 6345.039 20547.0201 0.31 0.7576 plan3 -18380.388 25744.6757 -0.71 0.4755 geo1 887074.041 45403.1222 19.54 <.0001 geo2 818983.124 57357.3750 14.28 <.0001 geo3 719724.740 53241.5528 13.52 <.0001 geo4 596420.605 60917.1547 9.79 <.0001 Yr_HCV2 2207426.202 55818.1754 39.55 <.0001 Yr_HCV3 1823032.109 55516.6483 32.84 <.0001 Yr_HCV4 1566273.808 52954.4110 29.58 <.0001 Yr_HCV5 1363016.425 55843.9837 24.41 <.0001 Yr_HCV6 790095.318 52936.5805 14.93 <.0001 Yr_HCV7 375234.088 50774.0305 7.39 <.0001 TT_lab -5669.810 19652.6871 -0.29 0.7731 anti_dpress 2468.726 25242.0778 0.10 0.9221 CVD_b4 -139498.775 45914.9966 -3.04 0.0025 Ddpend_b4 -279821.133 84480.4243 -3.31 0.0010 dpress_late -98773.741 41115.7385 -2.40 0.0166 HBV_b4 -290266.144 34711.1611 -8.36 <.0001

PAGE 151

151 Table D-2. Continued Parameter Estimate Standard Error t Value Pr > |t| HIV_b4 420869.329 50406.7575 8.35 <.0001 Heart_b4 -89550.954 22977.0848 -3.90 0.0001 anncost_b4 -253315.590 22223.2238 -11.40 <.0001 Biopsy_b4 -161164.028 34193.6043 -4.71 <.0001 hx_b4 -14993.543 26541.9026 -0.56 0.5723 er_cnt 14777.082 4654.3223 3.17 0.0016 GS_cnt -13048.314 1909.2004 -6.83 <.0001 INF_cnt 14706.615 2746.3809 5.35 <.0001 tx*age_HCV 45.725 2139.8825 0.02 0.9830 tx*gender -7103.752 31844.4051 -0.22 0.8235 tx*plan2 -9397.905 34079.7902 -0.28 0.7828 tx*plan3 1579.052 42102.6165 0.04 0.9701 tx*geo1 -83385.423 75496.0040 -1.10 0.2698 tx*geo2 -26216.610 98619.4299 -0.27 0.7905 tx*geo3 -110673.296 87468.5588 -1.27 0.2062 tx*geo4 -155338.739 105673.9610 -1.47 0.1421 tx*Yr_HCV2 -145535.925 87166.5614 -1.67 0.0955 tx*Yr_HCV3 -166746.890 79567.7150 -2.10 0.0365 tx*Yr_HCV4 -186276.852 77961.2421 -2.39 0.0172 tx*Yr_HCV5 -117210.576 80026.8517 -1.46 0.1435 tx*Yr_HCV6 -129820.072 77839.5699 -1.67 0.0959 tx*Yr_HCV7 -96036.507 72308.3802 -1.33 0.1846 tx*TT_lab -44826.923 33569.7158 -1.34 0.1822 tx*anti_dpress 23869.100 37203.4529 0.64 0.5214 tx*CVD_b4 -84017.895 98295.6202 -0.85 0.3930

PAGE 152

152 Table D-2. Continued Parameter Estimate Standard Error t Value Pr > |t| tx*Ddpend_b4 182655.522 116079.6211 1.57 0.1161 tx*dpress_late -185.871 59385.9538 -0.00 0.9975 tx*HBV_b4 156443.140 92170.9823 1.70 0.0901 tx*anncost_b4 -14761.799 35207.2924 -0.42 0.6752 tx*Biopsy_b4 34357.036 45504.4989 0.76 0.4505 tx*hx_b4 40323.117 49315.1251 0.82 0.4139 tx*er_cnt 18692.177 10531.1732 1.77 0.0764 tx*GS_cnt 200.668 2973.4151 0.07 0.9462 tx*INF_cnt 13819.171 9618.6470 1.44 0.1513 Table D-3. WTP=$80,000 in multivariate adjusted model Source DF Sum of Squares Mean Square F Value Pr > F Model 55 4.6005228E14 8.364587E12 194.03 <.0001 Error 632 2.7245554E13 43110054101 Corrected Total 687 4.8729784E14 Table D-3. Continued R-Square Coeff Var Root MSE nmb80 Mean 0.944088 9.661527 207629.6 2149035 Table D-3.Continued Parameter Estimate Standard Error t Value Pr > |t| Intercept -309271.021 105605.5140 -2.93 0.0035 tx 277482.581 181895.2049 1.53 0.1276 age_HCV 10074.096 1349.5050 7.47 <.0001 gender -127582.369 21700.7816 -5.88 <.0001

PAGE 153

153 Table D-3.Continued Parameter Estimate Standard Error t Value Pr > |t| plan2 6341.171 23350.9051 0.27 0.7860 plan3 -23947.420 29257.8426 -0.82 0.4134 geo1 996078.632 51598.9178 19.30 <.0001 geo2 918216.718 65184.4704 14.09 <.0001 geo3 805102.850 60506.9953 13.31 <.0001 geo4 663874.775 69230.0243 9.59 <.0001 Yr_HCV2 2526417.041 63435.2287 39.83 <.0001 Yr_HCV3 2087142.527 63092.5546 33.08 <.0001 Yr_HCV4 1796994.008 60180.6696 29.86 <.0001 Yr_HCV5 1563809.972 63464.5589 24.64 <.0001 Yr_HCV6 906932.532 60160.4060 15.08 <.0001 Yr_HCV7 430697.918 57702.7503 7.46 <.0001 TT_lab -3762.124 22334.5298 -0.17 0.8663 anti_dpress 5937.319 28686.6592 0.21 0.8361 CVD_b4 -153413.484 52180.6435 -2.94 0.0034 Ddpend_b4 -299935.265 96008.7821 -3.12 0.0019 dpress_late -114943.385 46726.4696 -2.46 0.0142 HBV_b4 -332744.732 39447.9114 -8.44 <.0001 HIV_b4 492767.181 57285.3586 8.60 <.0001 Heart_b4 -99728.102 26112.5810 -3.82 0.0001 anncost_b4 -288433.925 25255.8468 -11.42 <.0001 Biopsy_b4 -183434.415 38859.7279 -4.72 <.0001 hx_b4 -16262.152 30163.8606 -0.54 0.5900 er_cnt 17453.493 5289.4599 3.30 0.0010 GS_cnt -14813.817 2169.7335 -6.83 <.0001 INF_cnt 16347.025 3121.1572 5.24 <.0001

PAGE 154

154 Table D-3.Continued Parameter Estimate Standard Error t Value Pr > |t| tx*age_HCV 45.724 2431.8949 0.02 0.9850 tx*gender -8042.454 36189.9526 -0.22 0.8242 tx*plan2 -10263.533 38730.3826 -0.26 0.7911 tx*plan3 5760.650 47848.0189 0.12 0.9042 tx*geo1 -90442.004 85798.3308 -1.05 0.2922 tx*geo2 -27593.037 112077.2229 -0.25 0.8056 tx*geo3 -123384.601 99404.6830 -1.24 0.2150 tx*geo4 -174572.988 120094.4287 -1.45 0.1465 tx*Yr_HCV2 -164511.081 99061.4745 -1.66 0.0973 tx*Yr_HCV3 -186723.147 90425.6752 -2.06 0.0393 tx*Yr_HCV4 -217450.188 88599.9800 -2.45 0.0144 tx*Yr_HCV5 -135494.535 90947.4665 -1.49 0.1368 tx*Yr_HCV6 -148650.526 88461.7042 -1.68 0.0934 tx*Yr_HCV7 -110126.635 82175.7179 -1.34 0.1807 tx*TT_lab -52740.620 38150.7024 -1.38 0.1673 tx*anti_dpress 27026.545 42280.3062 0.64 0.5229 tx*CVD_b4 -97420.759 111709.2256 -0.87 0.3835 tx*Ddpend_b4 194520.336 131920.0647 1.47 0.1408 tx*dpress_late 2810.403 67489.8728 0.04 0.9668 tx*HBV_b4 182286.476 104748.8081 1.74 0.0823 tx*anncost_b4 -17373.827 40011.7458 -0.43 0.6643 tx*Biopsy_b4 39483.301 51714.1285 0.76 0.4455 tx*hx_b4 41554.923 56044.7599 0.74 0.4587 tx*er_cnt 20936.186 11968.2769 1.75 0.0807 tx*GS_cnt -39.231 3379.1729 -0.01 0.9907 tx*INF_cnt 18159.191 10931.2257 1.66 0.0972

PAGE 155

155 LIST OF REFERENCES 1. Alter MJ. Epidemiology of hepatitis C virus infection. World J Gastroenterol 2007;13(17):2436-41. 2. The National Center for HIV/AIDS, Viral Hepatitis, STD, and TB Prevent ion. National Hepatitis C Prevention Strategy: Hepatitis C Infection in the United Sta tes (2008). (Accessed May 18, 2008, at http://www.cdc.gov/ncidod/diseases/hepatitis/c/plan/HCV_infection.htm. ) 3. Alter MJ, Kruszon-Moran D, Nainan OV, et al. The prevalence of hepatitis C virus infection in the United States, 1988 through 1994. N Engl J Med 1999;341(8):556-62. 4. Armstrong GL WA, Simard EP, McQuillan GM, Kuhnert WL, Alter MJ. The Prevale nce of Hepatitis C Virus Infection in the United States,1999 through 2002. Ann Intern Med 2006;144:705-14. 5. Dienstag JL, McHutchison JG. American Gastroenterological Association t echnical review on the management of hepatitis C. Gastroenterology 2006;130(1):231-64; quiz 14-7. 6. Armstrong GL, Wasley A, Simard EP, et al. The Prevalence of Hepatitis C Virus Infection in the United States,1999 through 2002. Ann Intern Med 2006;144:705-14. 7. Grant WC, Jhaveri RR, McHutchison JG, et al. Trends in health care resource use for hepatitis C virus infection in the United States. Hepatology 2005;42(6):1406-13. 8. Wong JB, McQuillan GM, McHutchison JG, et al. Estimating future hepatitis C mor bidity, mortality, and costs in the United States. Am J Public Health 2000;90(10):1562-9. 9. Davis GL, Albright JE, Cook SF, et al. Projecting future complications of chronic hepatitis C in the United States. Liver Transpl 2003;9(4):331-8. 10. Nainan OV, Alter MJ, Kruszon-Moran D, et al. Hepatitis C virus genotypes and vira l concentrations in participants of a general population survey in the United States. Gastroenterology 2006;131(2):478-84. 11. Alter HJ, Seeff LB. Recovery, persistence, and sequelae in hepatitis C vi rus infection: a perspective on long-term outcome. Semin Liver Dis 2000;20(1):17-35. 12. Hoofnagle JH. Hepatitis C: the clinical spectrum of disease. Hepatology 1 997;26(3 Suppl 1):15S-20S. 13. Seeff LB. Natural history of hepatitis C. Hepatology 1997;26(3 Suppl 1):21S-8S. 14. McHutchison JG, Bacon BR. Chronic hepatitis C: an age wave of disease burden. Am J Manag Care 2005;11(10 Suppl):S286-95; quiz S307-11.

PAGE 156

156 15. Chen SL, Morgan TR. The natural history of hepatitis C virus (HCV) infection. Int J Med Sci 2006;3(2):47-52. 16. American Gastroenterological Association. American Gastroenterolo gical Association Technical Review on the Management of Hepatitis C. Gasteroenterology 2006;130:231 -64. 17. Piasecki BA, Lewis JD, Reddy KR, et al. Influence of alcohol use, race, and vi ral coinfections on spontaneous HCV clearance in a US veteran population. Hepatology 2004;40(4):892-9. 18. Strader DB, Wright T, Thomas DL, et al. Diagnosis, management, and treatment of hepatitis C. Hepatology 2004;39(4):1147-71. 19. American Gastroenterological Association. Medical position statement on t he management of Hepatitis C. Gastroenterology 2006;130:225-30. 20. Armstrong GL, Alter MJ, McQuillan GM, et al. The past incidence of hepatitis C virus infection: implications for the future burden of chronic liver disease in the Unite d States. Hepatology 2000;31(3):777-82. 21. Bonkovsky HL, Snow KK, Malet PF, et al. Health-related quality of life in patient s with chronic hepatitis C and advanced fibrosis. J Hepatol 2007;46(3):420-31. 22. Kim WR. The burden of hepatitis C in the United States. Hepatology 2002;36(5 Suppl 1):S30-4. 23. Fried MW, Shiffman ML, Reddy KR, et al. Peginterferon alfa-2a plus ribaviri n for chronic hepatitis C virus infection. N Engl J Med 2002;347(13):975-82. 24. Manns MP, McHutchison JG, Gordon SC, et al. Peginterferon alfa-2b plus ribavirin compared with interferon alfa-2b plus ribavirin for initial treatment of chroni c hepatitis C: a randomised trial. Lancet 2001;358(9286):958-65. 25. Claxton K, Sculpher M, Drummond M. A rational framework for decision making by the National Institute For Clinical Excellence (NICE). Lancet 2002;360(9334):711 -5. 26. Younossi ZM, Singer ME, McHutchison JG, et al. Cost effectiveness of interferon al pha2b combined with ribavirin for the treatment of chronic hepatitis C. Hepatology 1999;30(5):1318-24. 27. Wong JB, Koff RS. Watchful waiting with periodic liver biopsy versus immediate empirical therapy for histologically mild chronic hepatitis C. A cost-ef fectiveness analysis. Ann Intern Med 2000;133(9):665-75.

PAGE 157

157 28. Wong JB, Poynard T, Ling MH, et al. Cost-effectiveness of 24 or 48 weeks of interfe ron alpha-2b alone or with ribavirin as initial treatment of chronic hepatitis C. Inte rnational Hepatitis Interventional Therapy Group. Am J Gastroenterol 2000;95(6):1524-30. 29. Salomon JA, Weinstein MC, Hammitt JK, et al. Cost-effectiveness of tr eatment for chronic hepatitis C infection in an evolving patient population. Jama 2003;290(2):228-37. 30. Sullivan SD, Craxi A, Alberti A, et al. Cost effectiveness of peginterferon alpha -2a plus ribavirin versus interferon alpha-2b plus ribavirin as initial therapy for trea tment-naive chronic hepatitis C. Pharmacoeconomics 2004;22(4):257-65. 31. Yeh WS, Armstrong EP, Skrepnek GH, et al. Peginterferon alfa-2a versus pegi nterferon alfa-2b as initial treatment of hepatitis C virus infection: a cost-utilit y analysis from the perspective of the Veterans Affairs Health Care System. Pharmacothera py 2007;27(6):81324. 32. National Institutes of Health (NIH) Consensus Statement on Management of He patitis C: 2002. NIH Consens State Sci Statements 2002;19(3):1-46. 33. Strader DB. Understudied populations with hepatitis C. Hepatology 2002;36(5 Suppl 1):S226-36. 34. Calvert JFJ, Goldenberg PC, Schock C. Chronic hepatitis C infection in a rural Medic aid HMO. J Rural Health 2005;21(1):74-8. 35. Falck-Ytter Y, Kale H, Mullen KD, et al. Surprisingly small effect of a ntiviral treatment in patients with hepatitis C. Ann Intern Med 2002;136(4):288-92. 36. Markowitz JS, Gutterman EM, Hodes D, et al. Factors associated with the init iation of alpha-interferon treatment in Medicaid patients diagnosed with hepatitis C. J Viral Hepat 2005;12(2):176-85. 37. Shatin D, Schech SD, Patel K, et al. Population-based hepatitis C surveillance and treatment in a national managed care organization. Am J Manag Care 2004;10:250-25. 38. Rocca LG, Yawn BP, Wollan P, et al. Management of patients with hepatitis C in a community population: diagnosis, discussions, and decisions to treat. Ann Fam Med 2004;2(2):116-24. 39. Bini EJ, Brau N, Currie S, et al. Prospective multicenter study of eligibili ty for antiviral therapy among 4,084 U.S. veterans with chronic hepatitis C virus infection. Am J Gastroenterol 2005;100(8):1772-9. 40. Hare CB, Morris JA, Chu A, et al. Comparison of characteristics of treated and non -treated patients with hepatitis C infection. Pharmacoepidemiol Drug Saf 2006;15(2):71-6.

PAGE 158

158 41. Butt AA, Justice AC, Skanderson M, et al. Rate and predictors of treatment pres cription for hepatitis C. Gut 2007;56(3):385-9. 42. Kanwal F, Hoang T, Spiegel BM, et al. Predictors of treatment in patients with chronic hepatitis C infection role of patient versus nonpatient factors. Hepatology 2007;46(6):1741-9. 43. Stinnett AA, Mullahy J. Net health benefits: a new framework for the anal ysis of uncertainty in cost-effectiveness analysis. Med Decis Making 1998;18(2 Suppl):S68 -80. 44. O'Brien BJ, Drummond MF, Labelle RJ, et al. In search of power and significa nce: issues in the design and analysis of stochastic cost-effectiveness studies in health c are. Med Care 1994;32(2):150-63. 45. Hoch JS, Dewa CS. Lessons from Trial-Based Cost-Effectiveness Analy ses of Mental Health Interventions: Why Uncertainty About the Outcome, Estimate and W illingness to Pay Matters. Pharmacoeconomics 2007;25(10):807-16. 46. Hoch JS, Briggs AH, Willan AR. Something old, something new, something borrowed, something blue: a framework for the marriage of health econometrics and cost-effectiveness analysis. Health Econ 2002;11(5):415-30. 47. Health Insurance Association of American. Source Book of Health Insurance Dat a, 19992000. Washington, DC: Health Insurance Association of American.; 1999. 48. National Institutes of Health (NIH). National Institutes of Health Cons ensus Development Conference Panel statement: management of hepatitis C. Hepatology 1997;26 (3 Suppl 1):2S-10S. 49. Ray WA. Evaluating Medication Effects Outside of Clinical Trials: New -User Designs. American Journal of Epidemiology 2003;158(9):915-20. 50. Poynard T, Yuen MF, Ratziu V, et al. Viral hepatitis C. Lancet 2003;362(9401):2095-100. 51. Reschovsky J, Kemper P, Tu H. Does type of health insurance affect health care us e and assessments of care among the privately insured? Health Serv Res 2000;35(1 ): 219-37. 52. National Institutes of Health (NIH). National Institutes of Health Cons ensus Development Conference Panel statement: management of hepatitis C. Hepatology 2002;19:1-46. 53. Strader DB. Coinfection with HIV and hepatitis C virus in injection drug users a nd minority populations. Clin Infect Dis 2005;41 Suppl 1:S7-13. 54. Soriano V, Sulkowski M, Bergin C, et al. Care of patients with chronic hepatitis C a nd HIV co-infection: recommendations from the HIV-HCV International Panel. A ids 2002;16(6):813-28.

PAGE 159

159 55. Poynard T, Ratziu V, Charlotte F, et al. Rates and risk factors of liver fibr osis progression in patients with chronic hepatitis c. J Hepatol 2001;34(5):730-9. 56. Poynard T, Mathurin P, Lai CL, et al. A comparison of fibrosis progression in chronic liver diseases. J Hepatol 2003;38(3):257-65. 57. Torriani FJ R-TM, Rockstroh JK et al. Peginterferon Alfa-2a plus ribavirin for chronic hepatitis C virus infection in HIV-infected patients. N Engl J Med 2004;351:438-50. 58. Poynard T, McHutchison J, Manns M, et al. Impact of pegylated interferon alfa2b and ribavirin on liver fibrosis in patients with chronic hepatitis C. Gastroenterolog y 2002;122(5):1303-13. 59. Carrat F B-SF, Pol S et al. Pegylated interferon alfa-2b vs standard inter feron alfa-2b, plus ribavirin, for chronic hepatitis C in HIV-infected patients: a randomized control led trial. JAMA 2004;292:2839-48. 60. Jonsson JR, Barrie HD, O'Rourke P, et al. Obesity and steatosis influence s erum and hepatic inflammatory markers in chronic hepatitis C. Hepatology 2008;48(1):80-7. 61. Leandro G, Mangia A, Hui J, et al. Relationship between steatosis, inflammati on, and fibrosis in chronic hepatitis C: a meta-analysis of individual patient data. Gast roenterology 2006;130(6):1636-42. 62. Geppert CM, Arora S. Widening the door: the evolution of hepatitis C treatment in pat ients with psychiatric disorders. Hepatology 2007;46(4):957-9. 63. Schaefer M SM, Garkisch AS, Pich M, Hinzpeter A, Uebelhack R, et al.. Prevention of interferon-alpha associated depression in psychiatric risk patients with chr onic hepatitis C. J Hepatol 2005;42:793-98. 64. Kashihara D, Carper, K. National Health Care Expenses in the U.S. Civilian Noninstitutionalized Population, 2003 Rockville, MD. : Agency for Healthcare Researc h and Quality; 2005. Report No.: Statistical Brief #103. 65. Andriulli A, Festa V, Leandro G, et al. Usefulness of a liver biopsy in the evaluat ion of patients with elevated ALT values and serological markers of hepatitis vira l infection: an AIGO study. Dig Dis Sci 2001;46(7):1409-15. 66. Saadeh S, Cammell G, Carey WD, et al. The role of liver biopsy in chronic hepatit is C. Hepatology 2001;33(1):196-200. 67. Andriulli A, Persico M, Iacobellis A, et al. Treatment of patients with HCV i nfection with or without liver biopsy. J Viral Hepat 2004;11(6):536-42.

PAGE 160

160 68. Garcia G, Keeffe EB. Liver biopsy in chronic hepatitis C: routine or selec tive. Am J Gastroenterol 2001;96(11):3053-5. 69. Raikou M, McGuire A. Estimating medical care costs under conditions of censoring J Health Econ 2004;23(3):443-70. 70. Lin DY, Feuer EJ, Etzioni R, et al. Estimating medical costs from incomple te follow-up data. Biometrics 1997;53(2):419-34. 71. O'Hagan A, Stevens JW. On estimators of medical costs with censored data. J Health Econ 2004;23(3):615-25. 72. Young TA. Estimating mean total costs in the presence of censoring: a compara tive assessment of methods. Pharmacoeconomics 2005;23(12):1229-42. 73. Doshi JA, Glick HA, Polsky D. Analyses of cost data in economic evaluations conducte d alongside randomized controlled trials. Value Health 2006;9(5):334-40. 74. Lin DY. Linear regression analysis of censored medical costs. Biosta tistics 2000;1(1):3547. 75. Willan AR, Lin DY, Manca A. Regression methods for cost-effectiveness analy sis with censored data. Stat Med 2005;24(1):131-45. 76. Wooldridge JM. Introductory econometrics: a modern approach Stamford, CT: SouthWestern College Publishing; 2000. 77. Sloan FA, Bethel MA, Ruiz DJ, et al. The growing burden of diabetes mellitus in the US elderly population. Arch Intern Med 2008;168(2):192-9. 78. Nguyen GC, Segev DL, Thuluvath PJ. Nationwide increase in hospitalizations and hepatitis C among inpatients with cirrhosis and sequelae of portal hypertension. C linical Gastroenterology and Hepatology 2007;5(9):1092-9. 79. Lindsay KL, Trepo C, Heintges T, et al. A randomized, double-blind trial comparing pegylated interferon alfa-2b to interferon alfa-2b as initial treatment for chronic hepatitis C. Hepatology 2001;34(2):395-403. 80. Bruno S, Stroffolini T, Colombo M, et al. Sustained virological response to interferonalpha is associated with improved outcome in HCV-related cirrhosis: a retrospec tive study. Hepatology 2007;45(3):579-87. 81. Wong JB, Davis GL, McHutchison JG, et al. Economic and clinical effects of evaluat ing rapid viral response to peginterferon alfa-2b plus ribavirin for the initial treat ment of chronic hepatitis C. Am J Gastroenterol 2003;98(11):2354-62.

PAGE 161

161 82. Collett D. Modelling Survival Data in Medical Research 2nd ed. Boca Raton: Chapm an & Hall/CRC; 2003. 83. Di Bisceglie AM, Shiffman ML, Everson GT, et al. Prolonged therapy of advanced c hronic hepatitis C with low-dose peginterferon. N Engl J Med 2008;359(23):2429-41. 84. Kaiser S, Lutze B, Werner Rea. Long-term low dose treatment with pegyl ated interferon alpha 2b for 6 years leads to a significant reduction in fibrosis and inflammator y score in a subgroup of chronic hepatitis C nonresponder patients with fibrosis or cirrhosis In: AASLD; 2008; San Francisco; 2008. 85. Mihm U, Herrmann E, Sarrazin C, et al. Review article: predicting response in hepatitis C virus therapy. Aliment Pharmacol Ther 2006;23(8):1043-54. 86. Shiffman ML, Di Bisceglie AM, Lindsay KL, et al. Peginterferon alfa2a and ribavirin in patients with chronic hepatitis C who have failed prior treatment. Gastroente rology 2004;126(4):1015-23. 87. Shiffman ML, Ghany MG, Morgan TR, et al. Impact of reducing peginterferon alf a-2a and ribavirin dose during retreatment in patients with chronic hepatitis C. Gastroe nterology 2007;132(1):103-12. 88. Metwally MA, Zein CO, Zein NN. Regression of hepatic fibrosis and cirrhos is in patients with chronic hepatitis C treated with interferon-based therapy. Gastroenter ology 2003;124(5):1561. 89. Mallet V, Gilgenkrantz H, Serpaggi J, et al. Brief communication: the relat ionship of regression of cirrhosis to outcome in chronic hepatitis C. Ann Intern Med 2008;149(6):399-403. 90. Wiesner R, Sorrell M, Villamil F, et al. Report of the first International Liver Transplantation Society expert panel consensus conference on liver transplantat ion and hepatitis C. Liver Transpl 2003;9(11):S1-9. 91. Forns X, Garcia-Retortillo M, Serrano T, et al. Antiviral therapy of patie nts with decompensated cirrhosis to prevent recurrence of hepatitis C after liver tr ansplantation. J Hepatol 2003;39:389–96. 92. Thomas R, Brems J, Guzman-Hartman G, et al. Infection with chronic hepatiti s C virus and liver transplantation: a role for interferon therapy before transplantation. Liver Transpl 2003;9:905–15. 93. Everson G, Trotter J, Forman L, et al. Treatment of advanced hepatitis C with a low accelerating dosage regimen of antiviral therapy. Hepatology 2005;42:255–62.

PAGE 162

162 94. D'Souza R, Sabin C, Foster G. Insulin resistance plays a significant role i n liver fibrosis in chronic hepatitis C and in the response to antiviral therapy. Am J Gastroenterol 2005;100:1509–15. 95. Romero-Gmez M, Del Mar Viloria M, Andrade RJ. Insulin resistance impairs sustained response rate to peginterferon plus ribavirin in chronic hepatitis C patients. Gastroenterology 2005;128:636–41. 96. Kuehne FC, Bethe U, Freedberg K, et al. Treatment for hepatitis C virus in huma n immunodeficiency virus-infected patients: clinical benefits and cost-effec tiveness. Arch Intern Med 2002;162(22):2545-56. 97. Campos NG, Salomon JA, Servoss JC, et al. Cost-effectiveness of treatment for hepatitis C in an urban cohort co-infected with HIV. Am J Med 2007;120(3):272-9. 98. Shepherd J, Jones J, Hartwell D, et al. Interferon alpha (pegylated and non-pegyla ted) and ribavirin for the treatment of mild chronic hepatitis C: a systematic revie w and economic evaluation. Health Technol Assess 2007;11(11):1-205, iii. 99. Bedossa P, Dargere D, Paradis V. Sampling variability of liver fibrosis in chronic hepatitis C. Hepatology 2003;38(6):1449-57. 100. Fraenkel L, Chodkowski D, Lim J, et al. Patients' Preferences for Treat ment of Hepatitis C. Med Decis Making 2009. 101. Neymark N, Adriaenssen I, Gorlia T, et al. Estimating survival gain fo r economic evaluations with survival time as principle endpoint: A cost-effectiveness anal ysis of adding early hormonal therapy to radiotherapy in patients with locally advanc ed prostate cancer. Health Econ 2002;11:233-48. 102. Willan AR, Lin DY, Cook RJ, et al. Using inverse-weighting in cost-effectivene ss analysis with censored data. Stat Methods Med Res 2002;11(6):539-51.

PAGE 163

163 BIOGRAPHICAL SKETCH Chien-Ning Hsu was born in Kaohsiung, Taiwan. She received a Master of Science de gree in clinical pharmacy from National Taiwan University, Taipei in 1996. Her thes is title was "Drug use evaluation and preliminary cost analysis of uncomplicated appendectomy cl inical path.” She continued her clinical pharmacy fellowship for two years at Nationa l Taiwan University Hospital. During that time, she gained experience and skills in pharmacotherapy and diseas e management in several areas including critical care, infectious disease, nephrology, and o rgan transplantation. She worked in the drug regulatory agency, Center for Drug Evaluation (CDE) in T aiwan for 3 years before she began her doctoral program in Pharmaceutical Outco mes and Policy at the University of Florida, College of Pharmacy. She served as a project manager in the new drug application (NDA) and investigational new drug (IND) review processes. Chie n-Ning has authored and coauthored peer-reviewed publications and presented at national and interna tional academic conferences. She is interested in evaluating the appropriatenes s of drug use and health outcomes with special regard to economic consequences.