DETERMINANTS OF VIROLOGICAL FAILURE IN HIV POSITIVE PATIENTS RECEIVING ANTIRETROVIRAL THERAPY IN HAITI 2013 2017 By SHANNAN N. RICH A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE UNIVERSITY OF FLORIDA 2018
2018 Shannan N. Rich
3 ACKNOWLEDGMENTS I thank the chair and members of my supervisory committee for their mentoring, the staff in the Departments of Epidemiology and Health Outcomes and Biomedical Informatics for their keen research assistance, the participants in my study, the Haitian Ministry of Public Health and Population, and the National Alliance of State and Territorial AIDS Directors for its generous support. I thank my parents and close friends for their loving encouragement, which motivated me to complete my study.
4 T ABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 3 LIST OF TABLES ................................ ................................ ................................ ............ 5 LIST OF ABBRE VIATIONS ................................ ................................ ............................. 6 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ ... 9 2 METHODOLOGY ................................ ................................ ................................ 13 Study Populati on ................................ ................................ ................................ 13 Ethics Approval ................................ ................................ ................................ ... 13 Virological Failure ................................ ................................ ................................ 13 Predictors ................................ ................................ ................................ ............ 14 Statistical Analysis ................................ ................................ .............................. 14 3 RESULTS ................................ ................................ ................................ ............ 16 Characteristics of Study Population ................................ ................................ .... 16 Associations of Demographic and Clinical Characteristic s with Virological Failure ................................ ................................ ................................ .............. 17 Comp arison of Patients with Known versus Unknown Virological Status ............ 18 4 DISCUSSION ................................ ................................ ................................ ...... 24 Demographic Characteristics ................................ ................................ ............... 25 Clinical Characteristics ................................ ................................ ......................... 26 Strengths and Limitations ................................ ................................ ..................... 27 5 CONCLUSION AND FUTURE DIRECTIONS ................................ ...................... 29 LIST OF REFERENCES ................................ ................................ ............................... 30 BIOGRAPHICAL SKETCH ................................ ................................ ............................ 33
5 LIST OF TABLES Table page 3 1 Demographic and clinical characteristics of HIV patients in Haiti by virological status, 2013 2017. ................................ ................................ ......... 19 3 2 Unadjusted and adjusted model results for association of demographic and clinical character istics with virological failure in Haiti, 2013 2017. ........... 21 3 3 Comparative analysis of associations of demographic and clinical cha racteristics with virological failure in Haiti, 2013 2017. .............................. 22
6 LIST OF ABBREVIATIONS 3TC Lamivudine ART Antiretroviral therapy AZT Zidovudine CI Confidence interval EFV Efavirenz EMR Electronic medical records GHESKIO Infections HIV Human immunodeficiency virus HIVDR Human immunodeficiency virus drug resistance LLV Low level viremia MSPP Haitian Ministry of Public Health and Population NVP Nevirapine NNRTI Non nucleoside reverse transcriptase inhibitor NRTI Nucleoside reverse transcriptase inhibitor (NRTIs) OR Odds ratio PIH Partners in Health PNLS National Public Health Laboratory (French acronym) SALVH Haitian Active Longitudinal Tracking of HIV database (French acronym) STI Sexually transmitted disease TB Tuberculosis TDF Tenofovir UNAIDS Joint United Nations Programme on HIV/AIDS WHO World Health Organization
7 Abstract of Thesis Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Science DETERMINANTS OF VIROLOGICAL FAILURE IN HIV POSITIVE PATIENTS RECEIVING ANTIRETR OVIRAL THERAPY IN HAITI 2013 2017 By Shannan N. Rich August 2018 Chair: Robert Cook Major: Epidemiology Studies of early virological failure in human immunodeficiency virus (HIV) positive patients in Haiti are limited. This study aimed to examine th e prevalence and determinants of early virological failure among patients initiating first line antiretroviral therapy (ART) in Haiti between 2013 and 2016. Electronic medical records data were accessed through the Haitian Activ e Longitudinal Tracking of HIV databas e. Patients who initiated ART between January 1, 2013 and December 31, 2016 and who had a viral load test performed within 3 to 12 months after ART initiation were included in the study. Multivariable logistic regressi on was performed to predict the binary outcome of virological failure, defined as a viral load level greater than or equal to 1000 HIV 1 RNA copies/mL, based on patient demographic and clinical characteristics. The prevalence of early virological failure in this study population was 67%. In multivariable analysis, virological failure was associated with a CD4 count below 200 cells/mm 3 (Odds r atio [OR]=3.0; 95% Confidence i nterval [CI] 1.9 4.9), World Health Organization (WHO) clinical stage of IV (OR=2.6; 95% CI 1.2 5.7), ART initiation year of
8 2015 (OR=1.8; 95% CI 1.2 2.9), ART regimen consisting of lamivudine (3TC), zidovudine (AZT), and n evirapine (NVP) (OR=2.5; 95% CI 1.2 4.9), and months between ART initiation and viral load test (OR=0.9; 95% CI 0.8 0. 9). Further, patients who were diagnosed in the Artibonite Centre and Sud departments were less likely to experience virological failure (OR=0.4; 95% CI 0.3 0.7, OR=0.4; 95% CI 0.2 0.7, respectively), as compared to patients who diagnosed in the Nord depar tment. In conclusion, t his study described, for the first time, differences in early virological outcomes in persons receiving ART in Haiti. Enhanced monitoring of virological outcomes is recommended for younger individuals, individuals with high WHO clin ical stages, low CD4 counts, and individuals treated with 3TC AZT NVP Future studies should attempt to explain the departmental differences observed in this study, in addition to studying additional socio ecological factors that may be at play.
9 CHAPTER 1 INTRODUCTION For people living with human immunodeficiency virus (HIV), careful adherence to effective antiretroviral therapy (ART) leads to suppression of plasma viral loads to undetectable levels. The failure to achieve viral suppression is referred to as virological failure and is associated with higher rates of mortality, immunosuppression and the accumulation of resistance mutations [1,2] Individuals who are not suppressed also have a higher probability of tra nsmitting the virus to others  Virological failure has been linked to ind ividual and ecological factors including treatment nonadherence and pretreatment drug resistance, and hinders the success of HIV programs worldwide  The prevalence of virological failure, accompanied by the emergence of HIV drug resistance ( HIVDR), is increasing globally and has recently been termed the fourth HIV epidemic  Virological failure creates an obstacle for the final HIV elimination goal of achieving 90% viral suppression rates among patients receiving ART  Epidemiological monitoring for virologic al failure, formally defined by the World Health Organization ( WHO ) as the failure to achieve a plasma viral load level below 1000 HIV 1 RNA copies/mL after three months on antiretroviral therapy (ART)  can provide insight into population level HIVDR development and adherence to ART  As antiretroviral therapy (ART) continues to scale up, epidemiologic studies are needed to elucidate the correlates and predictors of virologic al failure. Drug resistance to non nucleoside reverse transcriptase inhibitors (NNRTIs) is increasing across all WHO geographic regions  This i s of greatest concern in low to middle income countries where NNRTIs, such as nevirapine (NVP) or efavirenz (EFV), are included in most first line combination ART regimens  Recent studies have
10 demonstrated a disproportionately higher burden of virologic al failure in resource poor settings, as compared to those consid ered resource dense  In Haiti, which ranks in 163 rd place on the Human Development Index  the adult HIV prevalence is the highest in the Western hemisphere [9 ] M ore than 70,000 individuals have been enrolled on ART as of 2016  While it is estimated that NNRTIs comprise 89% of patient drug regimens in Haiti  few studies h ave been conducted to estimate the prevalence of virologic al failure or pre treatment NNRTI drug resistance in this population  Viral load assessments were introduced in 2013 but were not scaled up nationally until 2016 and not at all treatment facilities  In a study conducted on patients who initiated ART at a large HIV clinic in the capital of Port au Prince between 2003 and 2008 virological failure was reportedly detected in 38% of patients tested with resistance to NNRTIs observed in 70% of these ind ividuals  Another study conducted in five hospitals in Port au Prince reported that one third of 2,313 patients failed to achieve virologic al suppression (<1000 copies/mL) after at least six months on ART  The vast majority of these patients ( 84%) were receiving ART for more than 12 months however, and it is unclear whether these patients had experienced any changes to their treatment regimens -an indicator of treatment failure prior to viral load assessment  That study found male sex, lower CD4 (<500 cells/mm 3 ), poor adherence to ART, and tuberculosis co infection were negatively associated with virological suppression  Another study which assessed the risk factors for major regimen changes in Haiti found significant associations with female sex, younger adult age (20 39 years old ), ART initiation year after 2006, an d lower CD4 (<350 cells/mm 3 ) at ART initiation 
11 G iven increasing trends in pre treatment drug resistance observed worldwide  it is becoming increasingly important to understand the ind ications for treatment failure as early as possible after treatment begins. Th e prevalence of pr etreatment NNRTI resistance in Latin America and the Caribbean was recently estimated to be 9.4% (6.6 13.2) with a predicted annual increase of 0.9%  The WHO recommends countries with a prevalence of pretreatment NNRTI resistance greater than 10% should change their first line therapies  A study of the early indications for virological failure in Haiti is treatment guidelines for first line ART. To our knowledge, no studies have assess ed the determinants of early virological outcomes of patients living in Haiti. This study used the data available from t he national HIV case based surveillance system, called the Haitian Active Longitudinal Tracking of HIV database (SALVH, French acronym), operated by the Ministry of Public Health and Population (MSPP) since 2008 and previously detailed elsewhere  Briefly, there are three main health systems that provide treatment services to persons living with HIV in Haiti and report iSant clinics (electronically administer ed by the International Training and Education Zanmi Lasante (PIH)  Using the data available from the iSant clinics, the objectives of the current study were to 1) estimate the prevalence of early virologic al failure in HIV positive individuals receiving ART 2) determine key demographic and clinical characteristics associated with viro logic al failure after controlling for covariates, and 3) identify opportunities for
12 public health action to inform current national treatment guidelines We hypothesized that the likelihood of virological failure is a function of demographic and clinical c haracteristics informed by the socio ecologic model  As a secondary objective and to evaluate potential reporting biases due to many cases with missing lab values we also sought to compare the demographic and clinical characteristics of patients who had an indication t hat a viral test was completed but lack ed lab results in the surveillance system
13 CHAPTER 2 METHODOLOGY Study Population This study employed a retrospective cohort design to assess the virologic al outcomes of treatment nave HIV positive individu als newly initiated on antiretroviral therapy (ART) in Haiti. The study population included patients who received their first ART prescription at one of the participating iSant treatment clinics between January 1, 2013 (when the earliest viral load assessments began in Haiti) and December 31, 2016 This date was chosen to allow for lab data to be included until December 31, 2017. Patients must have also ha d viral load testing performed between 3 to 12 months (90 to 365 days) after ART therapy initiation. Bivariate analyses were performed for all patients meeting this criterion (n=5128) whereas inferential logistic regression analyses were performed only for patients with viral load test results available in the SALVH dat abase (n=1151) Ethics Approval Th is study was reviewed by, and received ethics approval from the National Bioethics Committee in Haiti and the Institutional Review Board at the University of Florida (IRB201702830) Virological Failure P lasma blood samples were collected for every patient across the participating iSant treatment clinics and transported to the National Public Health Laboratory (PNLS) located in Port au Prince. Plasma v iral load levels were assessed at PNLS using the Generic HIV Viral Load Assay (Biocentric, Bandol, France ) which has a limit of detection of 300 10 7 copies/mL. Minimum and maximum functions were performed on
14 the available viral load results to ensure lab values were within the expected range. Patients whose earli est viral load test, assessed 3 to 12 months after initiation on ART, HIV 1 RNA copies/ mm 3 were classified as failing virologically (or virologically non suppressed) Patients whose earliest viral load test results were <1000 copies/mm 3 after 3 t o 12 months on ART were classified as virologically suppressed. Predictors Demographic characteristics were accessed at the time of the HIV case report and included age, sex, marital status, and clinic department (Haiti is organized into 10 administrative departments). Patient clinical characteristics included dates of diagnosis and ART initiation, months between viral load test date and diagnosis and ART initiation dates, earliest recorded CD4 T cell count (cells /mm 3 ) ART drug regimen [combinations of 3 possible drugs, including 1 NNRTI: nevirapine (NVP) or efavirenz (EFV) and 2 nucleoside reverse transcriptase inhibitors (NRTIs): lamivudine (3TC), zidovudine omprised less than 3 drugs or contained other less frequently prescribed drugs] WHO clinical stage of infection severity (I IV), transmission risk factors identified on the case report form, history or presence of tuberculosis (TB) o r sexually transmitted diseases (STD) and whether patients were prescribed ART by one or multiple clinics during the study period (termed Patients with missing or erroneous data on age (n=16) department (n=3), or ART regimen (n=1) were excluded from the a nalysis. Large numbers of missing or unknown values were categorized and retained for analysis. Statistical Analysis Demographic and clinical characteristics of individuals with and without virological failure were compared using Chi square test for catego rical variables and two
15 sample t tests or Wilcoxon Mann Whitney test for continuous variables with normal and skewed distribution, respectively. Bivariate analyses were also used to assess whether the patients included in the study were systematically diff erent from those who could not be included in the study due to the absence of viral load test results. L ogistic regression was used to identify significant predict ors of virologic al failure using variables found to be statistically significant in the bivar iate analyses at the p <0.05 level The full prediction model included the baseline patient demographic factors of continuous age (years), sex (male or female; [ref erence =female]) marital status (single, cohabiting, married, or divorced/widowed; [ref erence =single]) and clinic department (Artibonite Centre, Nord, Nord Est, Nord Ouest, Ouest, or Sud ; [ref erence =Nord]) The full prediction model additionally included the clinical factors: WHO clinical stage (I II, III, or IV; [ref erence =I]), year of AR T initiation (2013 2014, 2015, or 2016; [ref erence =2016]), ART regimen (3TC EFV TDF, 3TC AZT NVP, 3TC EFV AZT, or other; [ref erence =3TC EFV TDF]), earliest CD4 count (<200, 200 499, or 500+ cells/mm 3 ; [ref erence =500+ ] ), whether a transmission risk factor was identified on the case report form (yes, no, or unknown [ref erence =no]), and continuous months between viral load test date of ART initiation date. A parsimonious model was also estimated to select variables using a stepwise backwards elimination proce dure The results of the parsimonious model were similar to the results of the full model, and therefore only the full model results were presented as odds ratios (OR) with their respective 95% confidence intervals (CI). SAS version 9.4 was used for all an alyses.
16 CHAPTER 3 RESULTS Characteristics of Study Population Viral load test dates and results were available for 1151 patients. Of those, 774 had evidence of a viral load test HIV 1 RNA copies/ mm 3 which indicates prevalence of virologic al failure in this study population was 67.2%. Descriptive analyses indicated that patients failing to achieve virological suppression were on average younger than patients with suppression (31.9 years vs 35.3 years, p<0.01) and when examining age as a cat egorical variable, more likely to be under the age of 20 years (6.2% vs 15.3%, p=<0.01) (Table 3 1). Those experiencing virological failure were more likely to be diagnosed in the Ouest Department (37.9% vs 49.9%, p=<0.01), have a more severe WHO clinical stage of HIV at diagnosis (III or IV) (23.6% vs 31.9%, p=0.01), and have a CD4 cell count test result <200 cells/mm 3 (9.0% vs 19.4%, p=<0.01) (Table 3 1). Non virologically suppressed patients were less likely to be married or cohabitating as compared to s uppressed patients (49.8% vs 58.3%, p=0.03) (Table 3 1) Non virologically suppressed patients were also less likely to have a transmission risk factor identified (58.6% vs 51.0%, p=0.02) on a case report form (Table 3 1) Non virologically suppressed pati ents had longer intervals between viral load test and HIV diagnosis (32.3 vs 27.2 months, p=<0.01) and ART initiation (8.2 months vs 7.5, p=<0.01) (Table 3 1) The majority of patients in both populations were prescribed 3TC EFV TDF regimens, however patie nts experiencing virological failure were more likely to receive an ART regimen other than 3TC EFV TDF (12.7% vs 23.0%, p=<0.01) (Table 3 1) Groups did not differ with respect to sex, year of diagnosis, clinic transfer, and history or presence of tubercul osis or sexually transmitted
17 diseases. Though not statistically significant, the variable sex was retained in the multivariable analyses due to its observed significance in previous studies conducted in this population. Associations of Demographic and Clin ical Characteristic with Virological Failure In the multivariable analysis, clinic department, WHO clinical stage at diagnosis, year of ART initiation, ART regimen, value of the earliest CD4 count, and time between ART initiation and viral load test remain ed significant. Patients with virological failure were 59% less likely to be diagnosed in the Artibonite Centre combined department ( Odds Ratio [ OR ] =0.41; 95% Confidence Interval [ CI ] 0.25 0.66) and Sud department (OR=0.41; 95% CI 0.24 0.69), as compared t o patients who were treated in the Nord department (Table 3 2). Continuous age, sex and marital status were not statistically significantly associated with the outcome of virological failure. Compared to patients with a WHO stage of I at HIV diagnosis, pat ients who received a WHO clinical stage of IV were 2.6 times more likely to experience virological failure (OR=2.62; 95% CI 1.22 5.66) (Table 3 2). Patients who initiated therapy in 2015 were 1.8 times more likely to experience virological failure (OR=1.84 ; 95% CI 1.17 2.89), as compared to patients who began treatment in 2016 (Table 3 2). Patients whose treatment regimen consisted of 3TC AZT NVP were 2.5 times more likely to experience virological failure (OR=2.45; 95% CI 1.22 4.92), as compared to patient s with a treatment regimen consisting of 3TC EFV TDF (Table 3 2). Compared to patients in the highest CD4 count category (500+ cells/mm 3 ), patients in the lowest category (<200 cells/mm 3 ) had were 3 times more likely to experience virological failure (OR=3 .04; 95% CI 1.88 4.94) (Table 3 2). The time between ART initiation and viral load test was inversely correlated with virological failure. For each month increase in time on ART, the odds of virological
18 failure decreased by 13% (OR=0.87; 95% CI 0.82 0.92) (Table 3 2). The elapsed number of months between the date of HIV diagnosis and the viral load test date was not significantly associated with the outcome. The concordance statistic (c) for the multivariable model was 0.709, indicating that the model could accurately predict a and clinical characteristics included in this study Com parison of Patients with Known versus Unknown Virological Status In the analysis of the differences in the demographic characteristics of patients who had tests results available in the SALVH database vs patients who did not, patients included in the study were slightly younger (35.1 vs 33.0 years, p=<0.01) and were mor e likely to be diagnosed in the Artibonite Centre department ( 13.6% vs 8.6% p= < 0.0 1 ) compared to patients whose virological status was unknown (Table 3 3 ). In the comparison of clinical characteristics, patients included in the study were less frequently diagnosed between 2014 2015 (28.3% vs 38.3%, p=0.03), less likely to be initiated on treatment in 2015 (14.5% vs 25.1%, p=<0.01), and more likely to receive an ART regimen other than 3TC EFV TDF (19.4% vs 15.0%, p=<0.01 ), as compared to patients with know n virological status (Table 3 3 ). Additionally, although not clinically significant, patients included in the study had greater average months between diagnosis and viral load test (30.8 vs 29.7 months, p=0.02) and greater average months between ART initia tion and viral load test (7.8 vs 7.6 months, p=0.02) (Table 3 3 )
19 Table 3 1 Demographic and c linical c haracteristics of HIV p atients in Haiti by v irological s tatus, 2013 2017. Characteristic Not v irally s uppressed ( n =774) Virally s uppressed ( n =377) p for difference Mean (STD) Age at HIV d iagnosis (years) 31.9 (14.3) 35.3 (12.8) <0.001 Frequency (%) Age at HIV d iagnosis (years) <0.001 <5 49 (6.3) 7 (1.9) 5 14 45 (5.8) 6 (1.6) 15 19 25 (3.2) 10 (2.7) 20 29 196 (25.3) 106 (28.1) 30 39 241 (31.1) 129 (34.2) 40 49 137 (17.7) 66 (17.5) 50 59 65 (8.4) 38 (10.1) 60+ 16 (2.1) 15 (4.0) Sex 0.280 Female 486 (62.8) 249 (66.0) Male 288 (37.2) 128 (34.0) Marital s tatus 0.030 Single 138 (17.8) 64 (17.0) Cohabitating 284 (36.7) 160 (42.4) Married 101 (13.1) 60 (15.9) Divorced or w idowed 74 (9.6) 35 (9.3) Missing 177 (22.9) 58 (15.4) Clinic d epartment <0.001 Artibonite Centre 88 (11.4) 69 (18.3) Nord 177 (22.9) 88 (23.3) Nord Est 25 (3.2) 14 (3.7) Nord Ouest 52 (6.7) 18 (4.8) Ouest 386 (49.9) 143 (37.9) Sud 46 (5.9) 45 (11.9) Year of HIV d iagnosis 0.644 <2010 51 (6.6) 31 (8.2) 2010 11 83 (10.7) 43 (11.4) 2012 13 139 (18.0) 70 (18.6) 2014 15 229 (29.6) 97 (25.7) 2016 272 (35.1) 136 (36.1) WHO stage at diagnosis 0.011 I 103 (13.3) 49 (13.0) II 229 (29.6) 138 (36.6) III 195 (25.2) 78 (20.7) IV 52 (6.7) 11 (2.9) Missing 195 (25.2) 101 (26.7)
20 Table 3 1. Continued. Characteristic Not v irally s uppressed ( n =774) Virally s uppressed ( n =377) p for difference Frequency (%) Year of ART initiation 0.009 2013 84 (10.9) 34 (9.0) 2014 19 (2.5) 14 (3.7) 2015 129 (16.7) 38 (10.1) 2016 542 (70.0) 291 (77.2) ART regimen <0.001 3TC AZT NVP 68 (8.8) 12 (3.2) 3TC EFV AZT 50 (6.5) 20 (5.3) 3TC EFV TDF 596 (77.0) 329 (87.3) Other 60 (7.8) 16 (4.2) Earliest CD4 count <0.001 <200 150 (19.4) 34 (9.0) 200 499 182 (23.5) 89 (23.6) 500+ 175 (22.6) 111 (29.4) Missing 267 (34.5) 143 (37.9) History/presence of TB 0.645 No 16 (2.1) 11 (2.9) Yes 393 (50.8) 193 (51.2) Unknown 365 (47.2) 173 (45.9) History/presence of STIs 0.106 No 158 (20.4) 71 (18.8) Yes 301 (38.9) 128 (34.0) Unknown 315 (40.7) 178 (47.2) Any risk factor identified 0.015 No 379 (49.0) 156 (41.4) Yes 395 (51.0) 221 (58.6) Clinic transfer 0.868 No 729 (94.2) 356 (94.4) Yes 45 (5.8) 21 (5.6) Mean (STD) Months between diagnosis & test 27.2 (29.5) 32.3 (33.0) 0.012 Months between ART & viral test 7.5 (2.6) 8.2 (2.2) <0.001
21 Table 3 2 Unadjusted and a djusted m odel r esults for a ssociation of d emographic and c linical c haracteristics with v irological f ailure in Haiti, 2013 2017. Characteristic Unadjusted ORs (95% CI) Adjusted ORs (95% CI) Age at HIV d iagnosis (years) 0.98 (0.97;0.99) 0.99 (0.97;1.00) Sex Female Ref erence Ref erence Male 1.15 (0.89;1.49) 1.19 (0.90;1.57) Marital s tatus Single Ref erence Ref erence Cohabitating 0.82 (0.58;1.17) 1.04 (0.71;1.54) Married 0.78 (0.51;1.21) 0.94 (0.58;1.54) Divorced/ w idowed 0.98 (0.60;1.62) 1.15 (0.66;2.03) Clinic d epartment Nord Ref erence Ref erence Artibonite Centre 0.63 (0.42;0.95) 0.41 (0.25;0.66) Nord Est 0.89 (0.44;1.79) 0.74 (0.35;1.57) Nord Ouest 1.44 (0.79;2.60) 1.01 (0.54;1.90) Ouest 1.34 (0.98;1.85) 0.84 (0.55;1.28) Sud 0.51 (0.31;0.83) 0.41 (0.24;0.69) WHO s tage at d iagnosis I Ref erence Ref erence II 0.79 (0.53;1.18) 0.82 (0.54;1.26) III 1.19 (0.77;1.83) 1.26 (0.80;2.00) IV 2.25 (1.08;4.69) 2.62 (1.22;5.66) Year of ART i nitiation 2013 1.33 (0.87;2.03) 0.78 (0.46;1.32) 2014 0.73 (0.36;1.48) 0.58 (0.26;1.27) 2015 1.82 (1.24;2.69) 1.84 (1.17;2.89) 2016 Ref erence Ref erence ART r egimen 3TC EFV TDF Ref erence Ref erence 3TC AZT NVP 3.13 (1.67;5.86) 2.45 (1.22;4.92) 3TC EFV AZT 1.38 (0.81;2.36) 1.37 (0.75;2.49) Other 2.07 (1.17;3.65) 1.61 (0.86;3.03) Earliest CD4 c ount <200 2.80 (1.80;4.35) 3.04 (1.88;4.94) 200 499 1.30 (0.92;1.84) 1.42 (0.96;2.08) 500+ Ref erence Ref erence Any r isk f actor i dentified No Ref erence Ref erence Yes 0.74 (0.57;0.94) 0.81 (0.61;1.07) Months between d iagnosis and t est 1.00 (0.99;1.00) 0.99 (0.99;1.00) Months between ART and t est 0.90 (0.85;0.94) 0.87 (0.82;0.92)
22 Table 3 3 Comparative a nalysis of a ssociations of d emographic and c linical c haracteristics with v irological f ailure in Haiti, 2013 2017. Characteristic Virologic s tatus k nown ( n =1151) Virologic s tatus u nknown ( n =3977) p for difference Mean (STD) Age at HIV d iagnosis (years) 33.0 (13.9) 35.1 (13.5) <0.00 1 Frequency (%) Age at HIV d iagnosis (years) 0.001 <5 56 (4.9) 117 (2.9) 5 14 51 (4.4) 128 (3.2) 15 19 35 (3.0) 113 (2.8) 20 29 302 (26.2) 941 (23.7) 30 39 370 (32.2) 1279 (32.2) 40 49 203 (17.6) 825 (20.7) 50 59 103 (9.0) 423 (10.6) 60+ 31 (2.7) 151 (3.8) Sex 0.254 Female 735 (63.9) 2466 (62.0) Male 416 (36.1) 1511 (38.0) Marital s tatus 0.088 Single 202 (17.6) 663 (16.7) Cohabitating 444 (38.6) 1482 (37.3) Married 161 (14.0) 580 (14.6) Divorced or w idowed 109 (9.5) 491 (12.4) Missing 235 (20.4) 761 (19.1) Clinic d epartment <0.001 Artibonite Centre 157 (13.6) 341 (8.6) Nord 265 (23.0) 971 (24.4) Nord Est 39 (3.4) 187 (4.7) Nord Ouest 70 (6.1) 161 (4.1) Ouest 529 (46.0) 1879 (47.3) Sud 91 (7.9) 438 (11.0) Year of HIV d iagnosis <0.001 <2010 82 (7.1) 324 (8.2) 2010 11 126 (11.0) 235 (5.9) 2012 13 209 (18.2) 594 (14.9) 2014 15 326 (28.3) 1523 (38.3) 2016 408 (35.5) 1301 (32.7) WHO s tage at d iagnosis 0.032 I 152 (13.2) 602 (15.1) II 367 (31.9) 1238 (31.1) III 273 (23.7) 1029 (25.9) IV 63 (5.5) 154 (3.9) Missing 296 (25.7) 954 (24.0)
23 Table 3 3 Continued. Characteristic Virologic s tatus k nown ( n =1151) Virologic s tatus u nknown ( n =3977) p for difference Frequency (%) Year of ART i nitiation <0.001 2013 118 (10.3) 176 (4.4) 2014 33 (2.9) 145 (3.7) 2015 167 (14.5) 1000 (25.1) 2016 833 (72.4) 2656 (66.8) ART r egimen 0.001 3TC AZT NVP 80 (7.0) 185 (4.7) 3TC EFV AZT 70 (6.1) 223 (5.6) 3TC EFV TDF 925 (80.4) 3381 (85.0) Other 76 (6.6) 188 (4.7) Earliest CD4 c ount <0.001 <200 184 (16.0) 681 (17.1) 200 499 271 (23.5) 1030 (25.9) 500+ 286 (24.9) 1138 (28.6) Missing 410 (35.6) 1128 (28.4) History/ p resence of TB 0.273 No 586 (50.9) 2069 (52.0) Yes 27 (2.4) 123 (3.1) Unknown 538 (46.7) 1785 (44.9) History/ p resence of STIs 0.003 No 429 (37.3) 1601 (40.3) Yes 229 (19.9) 894 (22.5) Unknown 493 (42.8) 1482 (37.3) Any r isk f actor i dentified 0.875 No 535 (46.5) 1859 (46.7) Yes 616 (53.5) 2118 (53.3) Clinic t ransfer 0.084 No 1085 (94.3) 3798 (95.5) Yes 66 (5.7) 179 (4.5) Mean (STD) Months between d iagnosis and t est 28.9 (30.8) 26.6 (29.7) 0.016 Months between ART and t est 7.8 (2.5) 7.6 (2.5) 0.018
24 CHAPTER 4 DISCUSSION The purpose of this study was to estimate the prevalence and determine the demographic and clinical characteristics associated with virologic al failure in this population of HIV positive individuals receiving ART in Haiti. As a result, we identified opportunities for public health action to inform current treatment guidelines. The prevalence of virological failure in this study was 67% which differs markedly from previous studies in this population on ART The Joint United Nations Programme on HIV/AIDS ( UNAIDS ) reported in 2016 that among those who were diagnosed and receiving therapy, 31% were reported to be virologically failing  Further, another study found that virological failure was detected in 32% of patients tested  Comparisons of these estimates are difficult because case definition s of treatment failure differ ( viral load level 1000 copies/mL after 6 to 60 months or more on ART vs 1000 copies/mL after 3 to 12 months on ART applied in this study )  Most patients in the study conducted by Jean Louis et al. (84%) were receiving ART for at least 12 months before their viral loads were assessed, and it is unknown whether patients had experienced any changes to their initial treatment regimens during the follow up period, which could indicate virological failure  The number of months on ART was negatively correlated with the outcom e of virological failure in the current study, inferring that patients were more likely to achieve viral suppression the longer they are on ART. This suggests that patients in the current study population may have been prematurely classified as faili ng therapy in this study, because of the case definition applied. However, previous studies have reported that earlier indications of high viral load results are positively correlated with virological failure [19,20] According to a study conducted
25 in India, 83% of cases of virological failure were identified at the time of the initial viral load test (assessed within 6 t o 12 months after ART initiation)  Further, a study conducted on low level viremia (LLV = viral load levels between 50 and 1000 copies/mL) in South Africa found that patients with any previous result of LLV after 5 months on ART had 2.8 times increased risk of confirmed virological failure and 5.2 times increased risk of switching to second line therapy  Further, studies have shown that rates of virological suppression tend to decline with each subsequent year on first line ART  potentially due to increasing drug resistance mutations [22 ] Earlier detection of virological failure and switch to second l ine therapy was previously found to increase the probability of survival in this population from 88% to 93%  Demographic Characteristic s The demographic factor most strongly associated with virologic al failure in this study was diagnosis at clinics located outside of the Artibonite Centre and Sud departments There was no association between age, sex, or marital status with virological failure in the multivariable analysis. It is unclear why patients who were diagnosed in the Artibonite Centre and Sud departments were more likely to achieve viral suppression than patients in some of the other departments. It is possible that since these departments border the Ouest department, where the capital Port au Prince is located, that the clinics located in these departments may have access to better resources (e.g. medications, testing services) and therefore have more robust HIV care services T he lack of a statistically significant difference in the outcome of virological failure between males and females varies from previous studies in this population. In Jean Louis et al males were 20% less likely to be virologically suppressed after at least 6
26 months on ART  and in McNairy et al., male sex was associated with higher HIV mortality after 12 months on ART  While it is possible that males may have poorer health outcomes in the long term due to issues such as ad herence or loss to follow up [9,13] this study found no difference in earlier virological outcomes by sex. Clinical Characteristics The patient clinical factors most strongly associated with virological failure in this analysis were a WHO stage of IV, a CD4 count <200 cells/mm 3 an ART regimen consisting of 3TC AZT NVP, and an ART initiation year of 2015. The finding that virological failu re was associated with lower CD4 counts and higher WHO clinical stage of HIV at diagnosis is consistent with previous studies in this population [13,15] and with a stud y conducted in a similar ly low resource population  The first line ART regimen found to be associated with increased odds of failure in this study (3TC AZT NVP) was similarly associated with reduced odds of virological suppression in another study conducted in this popula tion  In another study conducted in Haiti, 31% of people treated with this first line regimen eventually switched to a second line regimen, although this result was not statistically significant at the 0.05 level  Patients who initiated treatment in 2015 were more likely to experience virological failure suggesting a time varying effect H owever, p atients who initiated ART in 2015 were also more likely to have unknown virological status suggest in g that this significance may be attributed to stastical artifact due to the smaller frequency of patients in this category The results of the analysis indicate that ART treatment initiatives should be targeted at younger individuals, individuals with mo re severe infection stages, and individuals with significant immune status decline. M ost patients in this study were receiving the current recommended ART regimen: 3TC EFV TDF P atients who were
27 not receiving this regimen were more likely to be non virologically suppressed In particular, the ART regimen: 3TC AZT NVP was linked to the greatest odds of virological failure in this study when compared to the 3TC EFV TDF regimen potentially due to NVP which is an NNRTI In a systematic review of the effectiveness of EFV vs NVP (both NNRTIs), researchers reported little to no difference in viral suppression, however, the development of drug resistance was slightly higher for patients who received NVP  Strengths and Limitations This study had several strengths. It was the first to assess the virologi cal outcomes of individuals living with HIV outside of the capital of Port au Prince and identified regions (departments) associated with increased odds of virological failure. Associations of virological failure with low CD4 cell counts and high WHO clini cal stages are consistent with previous studies. Although not statistically significant, there was an upward trend observed in the strength of the associations with each increasing WHO HIV infection stage category and a downward trend in the strength of th e associations with each increasing CD4 count category. The study also had some limitations worth noting. First, most patient records that were queried had unknown virological status, excluding them from the primary analysis and calling into question poten tial selection bias. Patients included in the study were slightly younger, more likely to be diagnosed in the Artibonite Centre department, less frequently diagnosed between 2014 and 20 15 or initiated on ART in 2015, and more likely to receive an ART regim en other than 3TC EFV TDF. Additionally, p atients included in the study also had a greater mean number of months between viral load test and diagnosis and ART initiation While, an ART regimen other than 3TC EFV TDF, and ART initiation in 2015 were associa ted with
28 increased odds of virological failure in the primary analysis, treatment in the Artibonite Centre department and more time on ART were associated with reduced odds. Therefore, the results of the bias assessment indicate that though the populations differed with respect to some demographic and clinical characteristics, these characteristics did not appear to be related to virological status. A second limitation of the current analysis was the use of only a single viral load measurement to classify v irological failure. Repeated viral load assessments (at least two) are more precise and increase statistical power  however, they were not sufficiently available for consideration in this analysis. Further, the proportion of missing data on some covariates ranged from 20.4% to 46.7% which limited their utility in the analyses but rep resent opportunities to improve reporting Of note, data on the history or presence of tuberculosis and sexually transmitted diseases was significantly under reported (46.7% and 42.8% respectively). This likely explains the lack of an association between v irological outcomes and comorbid tuberculosis, which is a known risk factor for virological failure. Additionally, CD4 count data was missing for approximately 35.6% of patients which is much higher than previous studies conducted in this population This may be due to its decreased use as an eligibility screener in recent years, as it is no longer required for ART eligibility. Further, no data was available to consider treatment adherence or prior treatment exposure both of which are linked to virological failure.
29 CHAPTER 5 CONCLUSION AND FUTURE DIRECTIONS In conclusion, t his study was the first to describe the differences in early virological status in Haiti. Virological status differed by age, clinic department, year of ART initiation, WHO clinical stage, CD4 counts, ART regimen, and time on ART. Patients diagnosed in the Artibonite Centre and Sud departments were less likely to experience virological failure Future studies should confirm these findings and attempt to explain the departmental differences in virological response to identify underlying factors. The alternative ART regimen: 3TC AZT NVP was strongly associated with poorer virological outcomes in the present study and in another recently conducted study  It is therefore recommended that the Ministry of Public Health and Population (MSPP) enhance their monitoring of virologica l outcomes in patients receiving these first line regimens. This study included patients who were treated in 76 of 164 (46%) HIV treatment clinics in Haiti. While there are three main electronic medical records (EMR) systems that report data to SALVH (iSan t GHESKIO, and PIH), only iSant clinics could be included in the study due to lack of data availability for the other systems. The MSPP should encourage better reporting from the other EMR systems moving forward. Continued escalation of viral load testi ng is recommended, however the turnaround time for reporting of test results should be minimized. Lastly, future studies should be conducted to evaluate the role of ART adherence and pre treatment NNRTI drug resistance on virological outcomes in Haiti.
30 L IST OF REFERENCES 1 Lynen L, Van Griensven J, Elliott J. Monitoring for treatment failure in patients on first line antiretroviral treatment in resource constrained settings. Curr Opin HIV AIDS 2010; 5 :1 5. 2 Chun HM, Obeng Aduasare YF, Broyles LN, Ellenberger D. Expansion of Viral Load Testing and the Potential Impact on Human Immunodeficiency Virus Drug Resistance. J Infect Dis 2017; 216 :S808 S811. 3 Laborde Balen G, Taverne B, Ndour CT, Kouanfack C, Peeters M, Ndoye I, et al. The fourth HIV epidemic. Lancet Infect Dis 2018; 18 :379 380. 4 World Health Organization (WHO), Consolidated Guidelines on the Use of Antiretroviral Drugs for Treating and Preventing HIV Infection: Recommendations for a Public Health Approach 2013 http://www.ncbi.nlm.nih.gov/books/NBK195400/ [ A ccessed 21 Jun e 2018 ] 5 World Health Organization (WHO) Consolidated Strategic Information Guidelines for HIV in the Health Sector. WHO Guidelines Appro ved by the Guidelines Review Committee 2015 ', http://apps.who.int/iris/bitstream/10665/164716/1/97892415087 59_eng.pdf [A ccessed 22 May 201 8] 6 Haile Selassie H, World Health Organization (WHO) Department of HIV/AIDS, Centers for Disease Control and Prevention (U.S.), Global Fund to Fight AIDS T and Malaria, et al. WHO HIV drug resistance report 2017 ', http://apps.who.int/iris/bitstrea m/10665/255896/1/9789241512831 eng.pdf [A ccessed 21 Jun e 2 018 ] 7 Gupta RK, Gregson J, Parkin N, Haile Selassie H, Tanuri A, Andrade Forero L, et al. HIV 1 drug resistance before initiation or re initiation of first line antiretroviral therapy in low income and middle income countries: a systematic review and me ta regression analysis. Lancet Infect Dis 2018; 18 :346 355. 8 United Nations Development Programme (UNDP) Human Development Indicators Haiti 2016 ', http://hdr.undp.org/en/countries/profiles/HTI [A ccessed 22 May 2018 ] 9 McNairy ML, Joseph P, Unterbrink M, Galbaud S, Mathon J E, Rivera V, et al. Outcomes after antiretroviral therapy during the expansion of HIV services in Haiti. PLOS ONE 2017; 12 :e0175521. 10 Auld AF, Pelletier V, Robin E, Shiraishi RW, Dee J, Antoine M, et al. Retention throughout the HIV Care and Treatment Cascade: from Diagnosis to Antiretroviral Treatment of Adults and Children Living with HIV Haiti, 1985 2015. Morb Mortal Wkly MMWR 11 Wolff MJ, Giganti MJ, Cortes CP, Cahn P, Grinsztejn B, Pape JW, et al. A decade of HAART in Latin America: Long term outcomes among the first wave of HIV patients to receive combination therapy. PloS One 2017; 12 :e0179769.
31 12 Rebeiro PF, Cesar C, Shepherd BE, De Boni RB, Corts CP, Rodriguez F, et al. Assessing the HI V Care Continuum in Latin America: progress in clinical retention, cART use and viral suppression. J Int AIDS Soc 2016; 19 doi:10.7448/IAS.19.1.20636 13 Jean Louis F, Buteau J, Franois K, Hulland E, Domerant JW, Yang C, et al. Virologic outcome among p atients receiving antiretroviral therapy at five hospitals in Haiti PloS One 2018; 13 :e0192077. 14 Charles M, Noel F, Leger P, Severe P, Riviere C, Beauharnais CA, et al. Survival, plasma HIV 1 RNA concentrations and drug resistance in HIV 1 infected Hai tian adolescents and young adults on antiretrovirals. Bull World Health Organ 2008; 86 :970 977. 15 Cesar C, Jenkins CA, Shepherd BE, Padgett D, Meja F, Ribeiro SR, et al. Incidence of virological failure and major regimen change of initial combination an tiretroviral therapy in the Latin America and the Caribbean: an observational cohort study. Lancet HIV 2015; 2 :e492 500. 16 Delcher C, Puttkammer N, Arnoux R, Francois K, Griswold M, Zaidi I, et al. Validating Procedures used to Identify Duplicate Reports HIV/AIDS Case Surveillance System. J Regist Manag 2016; 43 :10 15. 17 Hare AQ, Ordez CE, Johnson BA, del Rio C, Kearns RA, Wu B, et al. Gender Specific Risk Factors for Virologic Failure in KwaZulu Natal: Automobile Ownership and Financial Insecurity. AIDS Behav 2014; 18 :2219 2229. 18 UN Joint Programme on HIV/AIDS ( UNAIDS ) Haiti: HIV and AIDS Estimates 2015 ', http://www.unaids.org/en/regionscountries/countries/haiti/ [Accessed 22 May 2018] 19 Alvarez Uria G, Naik PK, Pakam R, Midde M. Early HIV viral load determination after initiating first line antiretroviral therapy for indentifying patients with high risk of developing virological failure: data from a cohort study in a resource limited setting. Trop Med Int Health TM IH 20 12; 17 :1152 1155. 20 Hermans LE, Moorhouse M, Carmona S, Grobbee DE, Hofstra LM, Richman DD, et al. Effect of HIV 1 low level viraemia during antiretroviral therapy on treatment outcomes in WHO guided South African treatment programmes: a multicentre coho rt study. Lancet Infect Dis 2018; 18 :188 197. 21 Hassan AS, Nabwera HM, Mwaringa SM, Obonyo CA, Sanders EJ, Rinke de Wit TF, et al. HIV 1 virologic failure and acquired drug resistance among first line antiretroviral experienced adults at a rural HIV clin ic in coastal Kenya: a cross sectional study. AIDS Res Ther 2014; 11 :9. 22 Stadeli KM, Richman DD. Rates of emergence of HIV drug resistance in resource limited settings: a systematic review. Antivir Ther 2012; 18 :115 123.
32 23 Charles M, Leger PD, Severe P, Guiteau C, Apollon A, Gulick RM, et al. Virologic, clinical and immunologic responses following failure of first line antiretroviral therapy in Haiti. J Int AIDS Soc 2012; 15 :17375. 24 Ruprez M, Pou C, Maculuve S, Cedeo S, Luis L, Rodrguez J, et al. Determinants of virological failure and antiretroviral drug resistance in Mozambique. J Antimicrob Chemother 2015; 70 :2639 2647. 25 Mbuagbaw L, Mursleen S, Irlam JH, Spaulding AB, Rutherford GW, Siegfried N. Efavirenz or nevirapine in three drug combinat ion therapy with two nucleoside or nucleotide reverse transcriptase inhibitors for initial treatment of HIV infection in antiretroviral nave individuals. Cochrane Database Syst Rev 2016; 12 :CD004246. 26 Rose CE, Gardner L, Craw J, Girde S, Wawrzyniak AJ, Drainoni M L, et al. A Comparison of Methods for Analyzing Viral Load Data in Studies of HIV Patients. PLOS ONE 2015; 10 :e0130090.
33 BIOGRAPHICAL SKETCH Shannan majored in epidemiology. While simultaneously working for the Departments of Epidemiology and Health Outcomes and Biomedical Informatics, she