1 EFFECTS OF CHRONIC EXERCISE ON BODY IMAGE: A META -ANALYTIC REVIEW By ANNA CAMPBELL A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTE R OF SCIENCE UNIVERSITY OF FLORIDA 2009
2 2009 Anna Campbell
3 To all w ho have helped me along the way
4 ACKNOWLEDGMENTS First I would like to thank Dr. Heather Hausenblas for all her help and encourage ment over the past few years. I would not have made it to this point without her assistance I would also like to thank my committee for their help and guidance through my masters program A special thanks to both Brian and Jessica for taking me under the ir wings, answering my countless questions, and offering their constant support. I am so grateful for all of my friends who have been there for me and made this a better journey. Finally, I would like to thank my family. They have provided me unwavering support and encouragement throughout my life and I could never thank them enough for all they have done for me and given me.
5 TABLE OF CONTENTS ACKNOWLEDGMENTS .................................................................................................................... 4 page LIST OF TABLES ................................................................................................................................ 7 LIST OF FIGURES .............................................................................................................................. 8 ABSTRACT .......................................................................................................................................... 9 CHAPTER 1 INTRODUCTION ....................................................................................................................... 11 Putative Moderators of Intervention Effects ............................................................................. 13 Participant Features ..................................................................................................................... 14 Exercise Intervention Features ................................................................................................... 17 Design Features ........................................................................................................................... 21 2 LITERATURE REVIEW ........................................................................................................... 25 Scope and Significance of Body Image ..................................................................................... 25 Body Image Interventions ........................................................................................................... 27 Cognitive Behavioral Th erapy ............................................................................................ 27 Alternative Treatments ........................................................................................................ 29 Exercise ................................................................................................................................ 34 Meta -Analysis .............................................................................................................................. 39 3 METHODS .................................................................................................................................. 43 Sample of Studies ................................................................................................................ 43 Selection Criteria ................................................................................................................. 44 Coding the Studies ............................................................................................................... 44 Effect Size Calculation and Analytic Strategy ................................................................... 47 P ublication/Dissemination Bias .......................................................................................... 50 4 RESULTS .................................................................................................................................... 55 Description of Studies ......................................................................................................... 55 Publication Bias ................................................................................................................... 57 Moderator Analyses ............................................................................................................. 58 5 DISCUSSION .............................................................................................................................. 64 Publication Bias ........................................................................................................................... 64 Moderator Analyses .................................................................................................................... 66
6 Participant features .............................................................................................................. 66 Design Features .................................................................................................................... 69 Exercise Intervention Features ............................................................................................ 72 Limitations ................................................................................................................................... 74 Future Directions ......................................................................................................................... 75 Conclusion ................................................................................................................................... 76 APPENDIX A LIST OF JOURNALS MANUALLY SEARCHED ................................................................. 78 B LETTER TO ACTIVE RESEARCHERS ................................................................................. 79 C TABLE TO ACTIVE RESEARCHERS ................................................................................... 80 D CODING SHEET ........................................................................................................................ 81 E MODERATOR TABLE ............................................................................................................. 84 F EFFECT SIZE INFORMATION ............................................................................................... 86 G OUTLIER TABLE ...................................................................................................................... 92 H DESCRIPTIVE CHARACTERISTICS OF STUDIES ............................................................ 95 I RANDOM EFFECTS MODERATOR ANALYSES ............................................................. 106 REFERENCE LIST .......................................................................................................................... 108 REFERENCE LIST OF STUDIES USED IN META-ANALYSIS ............................................. 120 BIOGRAPHICAL SKETCH ........................................................................................................... 125
7 LIST OF TABLES Table page 2 1 Cashs eight step cognitive behavioral therapy .................................................................... 42 5 1 Khoshdels potential sources of asymmetry in funnel plots ................................................ 77 E 1 Moderator table ...................................................................................................................... 84 F 1 Categorical random effect size (ES) information ................................................................. 86 F 2 Continuous mixed ES information ........................................................................................ 88 F 3 Categorical fixed ES information .......................................................................................... 89 F 4 Continuous fixed ES information .......................................................................................... 91 G 1 Outlier table ............................................................................................................................ 92 H 1 Descriptive characteristics of studies included in the meta a nalysis .................................. 96 I1 Random effects moderator analyses ................................................................................... 106
8 LIST OF FIGURES Figure page 2 1 Conceptual framework for the effects of exercise on body image. ..................................... 42 4 1 Forest plot ............................................................................................................................... 62 4 2 Funnel plot .............................................................................................................................. 63 4 3 Funnel plot with imputed values ........................................................................................... 63
9 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 EFFECTS OF CHRONIC EXERCISE ON BODY IMAGE: A META -ANALYTIC REVIEW By Anna Campbell May 2009 Chair: Heather Hausenblas Major: Applied Physiology and Kinesiology Because of the high pre valence of body dissatisfaction and its negati ve consequences a meta analy sis of the effects of exercise interventions on body image is timely. Older meta analyses of the effects of exercise on body image p ooled data from a range of studies that include correlational, quasi -experimental, and experimental exercise designs; with limited examination of moder ators. The purpose of my thesis was to meta analytically examine the impact of exercise intervention s on body image ; and participant, intervention, and design features that are associated with larger intervention effects. I conducted a systematic literature search and identified 5 7 exercise interventions (with pre and post data for both the exercise and control groups) examining the effects of chronic exercise on body image ( N Total = 6,273, Experimental N = 3,639, Control N = 2,634). The quality of studies varied. For the analyses, I used Comprehensive Meta analysis 2. I found a small random effect (effect size = 0.29) indicating that exercise interventions resulted in improved body image compared to control conditions; and that participant (i.e., age), design (i.e., year of publication), and intervention (i.e., exercise frequency and specificity) features moderated the size of the effect. The result s of my study indicate that chronic exerc ise does lead to improvements in body image. Further research is
10 needed examining the mechanisms and the exercise dose response needed for this change in body image.
11 CHAPTER 1 INTRODUCTION For both genders, negative body image is common and it has detrim ental physical, psychological, and economic consequences. More specifically, negative body image is related to emotional distress (Johnson & Wardle, 2005), smoking (Croghan et al., 2006), dramatic measures to alter appearance (e.g., steroid use; Raevuori e t al., 2006), social anxiety (Cash & Fleming, 2002), impaired sexual functioning (Wiederman, 2002), depression (Stice & Bearman, 2001), and eating disorders (Stice, Presnell, & Spangler, 2002). Widespread bodyimage disturbance is associated with U.S. cons umers spending billions of dollars annually for products aimed at changing their body size and shape such as diet pills, unnecessary cosmetic surgery, beauty products, and fitness products. Society will benefit from a better understanding of the efficacy o f interventions aimed at improving body image. Body -image change interventions typically consist of psychoeducational, cognitive behavioral, or drug therapies (e.g., weight loss pills; Gollings & Paxton, 2006). Given that many of these interventions are ex pensive, in short supply, and often not suitable for young populations other more practical strategies should be examined and promoted. Furthermore, although effective treatments have been developed, only a small proportion of those with body image problem s access treatment, and thus, evaluated treatments are underutilized. There are numerous explanations for this. Some of these relate to the practicalities of treatment delivery such as geographic distance, cost, and lack of availability. Others related to the particular nature of bodyimage problems, such as patient shame and ambivalence about change (Banasiak, Paxton, & Hay, 1998). One promising alternative mode of intervention for negative body image is chronic exercise. Chronic exercise, also known as ex ercise training, refers to cumulative, acute bouts of physical activity that are planned, structured, and repeated and result in improving or
12 maintaining of one or more physical fitness components of cardiorespiratory capacity, muscle strength, body compos ition, and flexibility (Caspersen, Powell, & Christenson, 1985). Physical activity refers to skeletal muscle activation resulting in energy expenditure beyond that of a resting level. Although evaluations of body image exercise interventions have been con ducted, their results have not been comprehensively reviewed. Meta analysis is a statistical technique that can be effectively applied to the literature examining the influence of chronic exercise on body image. Meta analysis can quantify the extent to whi ch key features of a research design, such as the specificity of exercise interventions and the type of control conditions, moderate changes in body image associated with chronic exercise. Because of the high prevalence of negative body image and its negat ive consequences a comprehensive meta analytic review of the chronic exercise and body image literature is timely. The few available reviews are narrow in scope with selection bias (Bane & McAuley, 1998; Fox, 2000; Hausenblas & Fallon, 2006; Martin & Licht enberger, 2002). One meta analytic review has been published (Hausenblas & Fallon, 2006), but it pooled data from a range of study types that included correlational, quasi experimental, and experimental exercise intervention studies that rendered it imposs ible to examine moderators of the exercise intervention effects in detail. As the authors predicted, a small effect size revealed that exercise intervention participants had a more positive body image post intervention compared to the nonexercising control participants. It would be informative, however, to control for preintervention scores, and examine the impact of moderator variables using more advanced meta analytic techniques. It is also important to eliminate the possibility that these differences ar e simply the result of publication bias, which was not directly assessed by Hausenblas and Fallon. Thus, the overarching goal of my thesis is to address this important
13 gap in the literature. The first aim of this review is to provide a statistical summary of these exercise intervention programs and their effects on body image (see Chapter 4) The second aim is to examine participant, intervention, delivery, and design features that are associated with larger intervention effects (see Chapter 4) Given the heterogeneity in the effects from these interventions, it is important to systematically consider the moderators associated with interventions that produced the largest effects. The third aim is to discuss theoretical, methodological, and statistical limita tions of the literature; and explore promising directions for future research in light of the findings (Chapter 5) In this Chapter, I will provide a brief rationale for the putative moderator variables to be examined, and I will advance hypotheses for th ese moderators. The information presented in Chapter 1 is similar to the Introduction section for a scientific manuscript. In Chapter 2, I will: (a) describe the scope and significance of body image; (b) briefly review the literature examining exercise int ervention for body image; and (c) describe the importance of meta analysis for research synthesis. In Chapter 3, I will describe the methods used for this meta analysis. In Chapter 4, I will report the meta analytic results. Finally, in Chapter 5, I will d iscuss the meta analytic results and provide suggestions for future research in this area. Putative Moderators of Intervention Effects Although researchers have found positive effects of exercise for body image, discrepancies in the literature exist. Clea rly, some of the ambiguity in the results obtained is the result of methodological factors. For example, studies have employed widely varying age groups. Studies have also differed with respect to the nature, intensity, and length of the exercise manipulat ion; the type of fitness measure employed; the general health and fitness level of the participants at the beginning of the study; subjects gender; the body image measure; and the nature of the control groups.
14 A unique feature of meta analyses is that the y permit empirical examination of factors associated with variation in effect sizes. Elucidating factors that moderate intervention program effects is informative because it highlights aspects of the participant, intervention, program delivery, and researc h design that are associated with stronger intervention effects. This information should increase future intervention effects by identifying the conditions under which optimal prevention effects occur. As well, this information might identify subgroups of individuals for whom alternative intervention programs need to be developed. Analyses of moderators should also advance general theories regarding effect ive routes to alter maladaptive health behaviors and attitudes. Accordingly, I investigated several pot ential moderators of intervention effects that were selected on the basis of theory, prior findings, and previous literature reviews. The moderators I examined are discussed in detail below. Participant Features Participant gender. Across the lifespan, female populations are at higher risk for negative body image than male populations (Altabe & Thompson, 1993; Elgin & Pritchard, 2006; Feingold & Mazzella, 1998). For example, Wang et al. (2005) found that 43% of females versus 12% of males aged 10 18 years were body dissatisfied. This finding may have emerged because sociocultural pressures for thinness are greater for females (Thompson, Heinberg, Altable, & Tantleff -Dunn, 1999), which may amplify the effects of negative body image for women. In support, mo re females than males are dissatisfied with their bodies, and most females with body image concerns are dissatisfied because they feel overweight (Thompson et al., 1999). In contrast, the reasons males give for body image concerns are more heterogeneous, w ith nearly half who indicate they are dissatisfied with their weight want to gain weight (McCabe & Ricciardelli, 2004). I hypothesized that interventions effects would be stronger for female samples versus male or mixed -sex samples.
15 Participant age. Studi es routinely find that about 40% of elementary girls and 25% of elementary boys are dissatisfied with their body, with children as young as 6 years of age expressing body dissatisfaction (McCabe & Ricciardelli, 2004; Smolak, 2002). Body dissatisfaction app ears to continually increase across adolescents for female and male populations, with girls continuing to report higher negative body image than boys (Eisenberg, Neumark -Sztainer, & Paxton, 2006; Muth & Cash, 1997; Wang et al., 2005). Body dissatisfaction, however, levels off and remains stable across the adult life span for men and women, with women continuing to report higher body dissatisfaction than men (Altabe & Thompson, 1993; Cash & Henry, 1995; Mangweth-Matzek et al., 2006; McLaren & Kuh, 2004; Tigg emann, 2004). For example, over 60% of female adults and older adults (aged 60 70 years) report body dissatisfaction (Garner, 1997; MangwethMatzek et al., 2006). I hypothesized that the intervention effects would increase with age until adulthood, and t hen remain consistent into middle and late adulthood; with larger effects evidenced for female than male populations. Participant ethnicity. There is also reason to believe that ethnicity might moderate the exercise intervention effects. Meta analyses rev eal that nonCaucasian populations (in particular African Americans ) have more favorable body image compared to Caucasians (Grabe & Hyde, 2006; Roberts, Cash, Feingold, & Johnson, 2006) ; suggesting that programs targeting Caucasians might be more effective because there is greater opportunity for intervention effects. Thus, I hypothesized that ethnicity would moderate the size of the effect whereby intervention targeting Caucasians (or having a larger % Caucasians ) would have a larger effect. Psychological r isk status of participants. I hypothesized that interventions are more effective when offered to high -risk participants (i.e., selected programs) versus all individuals in a population (i.e., universal programs). Theoretically, these highrisk people are m ore motivated
16 to engage in the intervention, and thus are more likely to benefit. It is also likely that low risk individuals have less room for change on the outcomes (floor effect; Stice, Shaw, & Marti, 2006). Intervention program for other psychological pathologies (e.g., eating disorders, body image disturbance, anxiety, depression, substance abuse) usually produce stronger effects for high risk subsamples than for the full sample of individuals enrolled in these universal programs (Clark et al., 1995; Jarry & Ip, 2005; LowryWebster, Barrett, & Dadds, 2001; Murphy et al., 2001). Thus I hypothesized that intervention effects would be larger for selected programs (e.g., eating disordered, high body dissatisfied) versus universal programs. However, the di stinction between universal and selected program is often blurred. For example, most universal programs focus solely on female populations a subpopulation at highrisk for negative body image. I considered interventions delivered to all participants in int act classrooms and trials that did not mention the intervention objective during recruitment (e.g., shape improvement) to be universal program. I considered interventions that screened participants for a risk factor or that used recruitment strategies that implicated screened participants, such as advertisements for body image intervention, to be selected programs. Participant body composition. Negative body image arises primarily from sociocultural pressures to be thin and physical deviation for the curren t thin ideal espoused for women and the lean and muscular ideal espoused for men in Western culture (Thompson et al., 1999). Elevated adiposity is theorized to promote body dissatisfaction because the current ideal physique for men and women is lean. Thus, the greater the degree of deviation from the current ideal physique, the greater the ensuing body dissatisfaction. In short, elevated body mass increases the risk for body dissatisfaction (Stice & Shaw, 2002). Of significance, overweight/obese populations are more likely to have higher bodyimage disturbance and benefit from weight reduction during an
17 exercise intervention compared to normal weight populations (Franklin, Denyer, Steinbeck, Caterson, & Hill, 2006; Hrabosky et al., 2006). Thus, I hypothesize d that larger intervention effects would be evidenced for overweight/obese populations than normal weight populations. Exercise Intervention Features Exercise dose. Exercise dose consists of the following components: duration, intensity, mode, frequency, a nd length. Although the dose -response is well -established for the physical health benefits of physical activity (ACSM, 2000); the dose response of exercise needed to obtain the psychological benefits of exercise is controversial. Thus, in an attempt to det ermine the dose response needed to obtain the effects of exercise on body image, I will examine the moderating effects of exercise intensity, duration, mode, frequency, and length. The duration of an exercise session interacts with the intensity to result in the expenditure of a sufficient number of calories to achieve health, fitness, and weight management goals. The duration of exercise recommended by the ACSM reflects that interaction (20 to 60 minutes of continuous or intermittent ( minimum of 10 -minute bouts) aerobic activity throughout the day. Consequently, exercising at 70 to 85% HRmax or 60 to 80% HRR for 20 to 30 minutes, excluding time spent warming up and cooling down, enables most people to achieve health, fitness, and weight management goals (AC SM, 2000). Although deconditioned persons may improve cardiorespiratory fitness with only twice -weekly exercise, optimal training frequency is achieved with 3 to 5 workouts per week. The additional benefits of more frequent training appear to be minimal, w hereas the incidence of lower extremity injuries increases abruptly. Consequently, the ACSM recommends an exercise frequency of 3 to 5 days week. I hypothesized that exercise interventions that meet the ACSM physical activity guidelines would result in lar ger effects than interventions that did not meet the guidelines ; because the former interventions are more likely to
18 produce physical changes in body composition (e.g., weight loss, toning, muscle development) that are directly related to improved body ima ge. Theoretically, interventions with a longer length afford a greater opportunity for presentation of information and behavioral change skills. That is, they allow participants to reflect on the intervention material between sessions, and also give parti cipants a chance to try new skills and then return to the group for troubleshooting advice. Meta analysis of substance abuse and eating disorder preventions reveal that interventions of longer length produced the largest effects (Rooney & Murray, 1996; Sti ce & Shaw, 2004). As well, exercise intervention examin ing mental health outcomes appear to produce larger effects with interventions of longer length (i.e., 12 weeks or greater; Craft & Landers, 1998). I hypothesized that intervention effect would be stronger for prevention programs with a longer versus short length (in weeks). Finally, with regard to exercise type, most of the exercise interventions used aerobic exercise (e.g., walking) or a combination of aerobic exercise plus resistance training. Resis tance training may be superior to aerobic exercise because improvements in strength emerge more quickly and tend to be larger than improvements in aerobic capacity (especially for the novice exerciser). Strength can improve 10 30% within the first 6 8 weeks of a resistance training program simply as a result of neuromuscular adaptation and learning the technique of weightlifting. In contrast, improvements in aerobic capacity are dependent upon physiological adaptations that occur at a slower rate (espec ially in a walking program) and may be less noticeable in terms of changes in body composition (Martin & Lichtenberger, 2002). Thus, I hypothesized that larger effects would be evidenced for resistance training -based interventions than aerobic -based interv entions.
19 Physical fitness. Researchers have attempted to validate the notion that improvements in physical fitness are associated with improvements in body image (Martin & Lichtenberger, 2002). Unfortunately, most of this research consists of correlational and cross -sectional studies that have compared body image between exercisers and nonexercisers. Not only have these studies produced equivocal results, but the nature of their designs is inappropriate for drawing conclusions about the effects of exercise induced fitness change on body image change (Fox, 2000; Taylor & Fox, 2005). To draw conclusions, the effects of systematic exercise interventions on both fitness and body image need to be examined. Furthermore, general findings across the exercise and psy chological well -being literature indicate that improvements in physical status such as cardiovascular fitness, strength, and weight or fat loss are not consistently related to changes in psychological well -being. For example, some elements of change, such as anxiety reduction (Taylor, 2000) and mood enhancement (Biddle, 2000), appear to occur independent of fitness change. There is evidence f ro m several studies that fitness change (as measured by standard laboratory or field tests of fitness) is not necessa ry for enhanced body image (Fox, 2000).This parallels the obesity treatment literature where amount of weight lost is not consistently reflected in the psychological benefits. Perceptions of health, physical competences, fitness and body image may arise si mple because there is a feeling that the body is improving through exercise. There is some indication that muscular fitness reflected in improved tone or strength can have a more rapid and powerful sensory effect than cardiovascular or flexibility change. Thus, because of the equivocal research results, I will examine if programs that increase physical fitness have larger intervention effects than those that do not increase activity. Intervention specificity. The effects of chronic exercise on body image ha ve been unclear in interventions that used a nonspecific intervention. Exercise interventions are specific when
20 exercise is the sole intervention and are nonspecific when another therapy (e.g., drug therapy, cognitive -behavioral therapy) is added to the exercise training. Therefore, the independent effect of chronic exercise on body image is unknown when another therapy is added. When chronic exercise was completed in conjunction with a second therapy, only a few investigators included an exercise -only comp arison condition. Consequently, it is unclear whether there is an additive effect on body image when a second therapy is added to a chronic exercise intervention. Puetz et al. (2006) in their meta analyses of the effects of chronic exercise on feelings of energy and fatigue found that the effect varied according to either the presence or absence of a placebo control or whether chronic exercise was completed alone or in combination with another therapy. Unfortunately, they did not report the effects independ ently for intervention specificity. Thus, it is not known if specificity was an independent moderator of the size of the effect. Thus, I will examine the moderating effect of intervention specificity o n the size of the effect. Theory. Theoretical issues can pertain to both the exercise intervention and the body image measure. For example, the exercise intervention can be based on a theoretical framework (e.g., self -efficacy theory, transtheoretical model, theory of planned behavior) in an attempt to increase exercise adherence and maintenance. Exercise interventions that are developed using a theoretical framework tend to produce larger effects than interventions that are not based on a theoretical framework (Sallis, 2001) Theoretical testing is needed to advance intervention efforts to persuade and enable people to make healthy behavior changes (Rothman, 2004; Sallis, 2001). One salient lesson and a future priority is to incorporate theory to reveal the intervention content and mechanisms to modify physic al activity behaviors so that future interventions are more efficacious and efficient (Blue & Black, 2005). Thus, I hypothesized that exercise interventions
21 that are develop ed based on a theoretical model will produce larger effects than nontheoretical exercise interventions. As well, the body image measure selected should be based on a theoretical framework that maps how exercise may affect body image. Unfortunately, ma n y researchers have taken atheoretical approaches to selecting the body image assessment s. Common theoretical frameworks include the Exercise and Self -esteem Model (Sonstroem & Morgan, 1989; Sonstroem, Harlow, & Josephs 1994). This model assumes that exercise first influences physical self -concept such that people develop a higher degree of physical competence and physical acceptance. This subsequently should lead to heightened feelings of global self -esteem. Physical Self perception Profile which is based on a multidimensional theoretical model of self -esteem is consistent with the hierarc hical modeling of the elements of self -esteem because several measurable levels of self -perceptions exist within the physical self. In Physical Self -perception Profile self -perceptions can be categorized as superordinated (i.e., global self -esteem), domai n (i.e., physical self -worth), subdomain (i.e., body attractiveness) and self -perception; which are general and enduring at the top of the hierarchy and increasingly specific and unstable at the lower end (Fox & Corbin, 1989). Thus, I hypothesized that body image assessments that were selected based on a theoretical model with produce larger effects than nontheoretical assessments. Design Features Control group. Variations in control conditions used in chronic exercise studies have made it difficult to inte rpret the literature on chronic exercise and body image. Controls used in the literature on chronic exercise and body image have ranged from notreatment controls (e.g., waitlist controls, assessment only controls) to placebo control conditions that involv ed participants in substantial physical or cognitive activities. Placebo controls (e.g., health education
22 classes) usually are structurally matched to the intervention in terms of contact hours, and they often contain features included in chronic exercise programs that may be therapeutic, such as opportunities for social interaction. Theoretically, the no -treatment controls will produce larger effect sizes than the placebo controls when they are compared with the intervention because the placebo control groups more effectively control for demand characteristics, participant expectancies, and other nonspecific factors that contribute to intervention effect (Stice & Shaw, 2004). Because there is no consensus as to the most appropriate control conditions for in terpreting data from chronic exercise interventions, there is a need to determine whether body image is influenced by the type of control conditions selected (Puet z et al., 2006). For example, in a meta analysis of the effects of chronic exercise on feelin gs of energy and fatigue, Peut z et al. (2006) found that placebo controls resulted in increased feelings of energy and lessening of fatigue compared to no -treatment controls in certain populations. I hypothesized that larger effects would be evidenced for notreatment controls compared to placebo controls. Intervention format. Theoretically, participants in interactive programs show greater intervention effects because this presentation format helps them become engaged in the program content, which facilita tes skill acquisition and attitudinal and behavioral change. Interactive programs are also more likely to involve the exercises that allow participants to apply the skills taught in the intervention, which should facilitate skill acquisition. Meta analysis of substance abuse and eating disorder prevention programs found that interactive programs were more effective than didactic programs (Stice & Shaw, 2004; Tobler et al., 2000). I predicted that interactive programs would be more effect than didactic progr ams. Interactive programs would include a physical activity program whereas didactic programs would be solely educational (i.e., providing the physical activity guidelines).
23 Recruitment method. Intervention effects are often larger when prevention programs are delivered solely to participants who have actively self -selected into trials in response to recruitment efforts, such as media advertisements, relative to when prevention programs are offered to all people in a defined population (e.g., a particular s chool; Stice et al., 2006). Presumably this is because the former strategy recruits people who are more motivated to achieve the exercise prevention effects, and therefore engage more effectively in the prevention program. Thus, I hypothesized that interve ntion effects would be larger for self -selecting volunteers than for participants recruited through population -based recruitment efforts. Random assignment. I theorized that trials that randomly assigned participants to conditions might produce larger inte rvention effects than trials that used alternative approaches to allocating participants to treatment conditions, such as matching. I reasoned that because random assignment is the best approach to generating groups that are equivalent on any potential con founding variables at baseline (with sufficiently large sample sizes), it should therefore minimize the chances that any of these confounding variables are correlated with treatment condition, which should thus maximize the ability to detect intervention e ffects if they really occur (i.e., randomization maximizes the signal to -noise ratio reflected in inferential tests of the intervention effects). Accordingly, I hypothesized that intervention effects may be greater for interventions that used random assign ment relative to other approaches to assigning participants to condition. However, because the proper analysis of intervention effects involves tests of differential change across conditions, that adjust for any initial differences at baseline on the outco me, I suspect that this effect might not reach statistical significance. Random assignment did not emerge as a significant moderator of effect sizes in other health interventions (Stice & Shaw, 2004)
24 Publication status. I investigated whether publication status (i.e., unpublished versus published) was related to the intervention effect size. Including unpublished studies allowed us to include a richer variety of exercise studies (Conn et al., 2003). Meta analyses that only include published studies are mor e likely to overestimate the magnitude of the true population effect, because the single biggest difference between published and unpublished research is the statistical significance of the results (Cook et al., 1993). Finally, this is a rapidly developing area of science where unpublished reports provide valuable information that may be published later. I hypothesized that published studies would produce larger effects than unpublished studies. Validated body image measure. Body image is a multidimensional construct encompassing perceptual, attitudes (emotional, feelings), cognitive (thinking, evaluation), sociocultural, and behavioral components. The term body-image disturbance represents some type of maladaptive response to the body image construct (Stewa rt & Williamson, 2004). Interventions that use psychometrically sound measures should be better positioned to detect intervention effects that do occur because these measures are more sensitive In support of this, larger effects were found for validated t han unvalidated measures in eating disorder prevention trials (Stice & Shaw, 2004). Thus, I hypothesized that interventions that used validated outcome measures would observe larger intervention effects than interventions that used measures for which relia bility and validity have not been established.
25 CHAPTER 2 LITERATURE REVIEW The purposes of this chapter are to: (a) highlight the scope and significance of body image; (b) review the body image intervention literature with a focus on exercise interventio ns; and (c) discuss the importance of meta analyses for synthesizing research Scope and Significance of Body Image Body image is a persons internal view of their outer appearance (Thompson et al., 1999); and it is characterized by cognitive, behaviora l, affective, and perceptual elements. The cognitive component consists of thought processing related to body satisfaction/appearance evaluation, social situations, and general information (Bane & McAuley, 1998; Thompson et al., 1999). The behavioral component is comprised of behaviors that are appearance related (e.g., dieting, exercising, avoiding social situations). The affective element deals with emotions, negative feelings, and anxiety related to ones appearance. The perceptual aspect represents the accuracy to which an individual perceives his or her body. Body image is often assessed using a continuum ranging from none to extreme bodyimage concerns. Higher levels of bodyimage disturbance are more likely to lead to negative psychosocial outcomes s uch as depression, sexual problems, and eating disorders (Cash & Henry, 1995; Cash & Strachan, 1999; Thompson et al., 1999). With regard to eating disorders, the Diagnostic and Statistical Manual of Mental Disorders Fourth Edition (DSM IV) includes bodyimage disturbance as a criteria for anorexia nervosa and bulimia nervosa stating that there must be a disturbance in the way in which ones body weight or shape is experienced, undue influence of body weight or shape on self -evaluation, or denial of the s eriousness of the current low body weight (APA, 1994).
26 Cash and Strachan (1999) identified at least three psychosocial problems related to negative body image including depression, anxiety, and sexual dysfunction. Studies have shown that depressed indivi duals experience increased levels of body dissatisfaction (Marsella, Shizuru, Brennan, & Kameoka, 1981; Rierdan, Koff, Stubbs, 1988). Correspondingly, individuals with greater levels of body dissatisfaction are more at risk for developing depression. Likew ise, individuals with body dissatisfaction are more likely to feel socially unacceptable leading to increased social -evaluative anxiety (Cash & Strachan, 1999) and social physique anxiety. Along with body-image disturbance, these problems can lead to sexual dissatisfaction or dysfunction through an avoidance of sexual contact and sexual situations. The quest for Western societys conception of the ideal body type is a major concern for young women and men (Fallon & Hausenblas, 2004), and a main fact or in determining body dissatisfaction. Internalization of these thin ideals are a risk factor for negative body image and eating disorders (Thompson & Stice, 2001). For example, Cash and Strachan (1999) found that the more these cultural ideals are intern alized by an individual the higher the risk for body dissatisfaction. As the number of individuals with body-image problems increases, so does the risk for eating disorders within the population. A 1997 survey in Psychology Today reported that 43% of men and 53% of women negatively evaluated their overall appearance (Garner, 1997). However this survey lacks scientific stringency, and the population surveyed may not be an accurate depiction of the general population; thus, limiting its external validity. M ore scientifically stringent research using valid body image surveys report similar rates of dissatisfaction as that found in the Garner (1997) study. For example, in Cash and Henrys (1995) national survey of womens body image, over half of the 803 women reported negative appearance evaluations and
27 concerns with being or becoming overweight; with Caucasian women reporting the most body dissatisfaction. Furthermore, a meta analysis by Feingold and Mazzella (1998) found that the rate of American women with body dissatisfaction has become increasingly more negative over the past 50 years. T he prevalence of body dissatisfaction is higher in women and girls than men and boys; with adolescent/teenage females reporting the highest levels of body dissatisfaction (Feingold & Mazzella, 1998). In summary, body -image disturbance is prevalent; and it is associated with negative physical, social, psychological, and financial outcomes. Therefore, it is necessary to find efficacious prevention and treatment programs that a re cost -effective and have the potential to reach large audiences. The next section will briefly review common interventions for body dissatisfaction with an emphasis on interventions that have used exercise as the means to either prevent or treat body ima ge concerns. Body Image Interventions Cognitive Behavioral Therapy There are several types of interventions that have been developed to prevent and treat bodyimage concerns. These include cognitive behavioral therapy, exercise interventions and various alternative treatments including psychoeducational, ecological, and experiential interventions Cognitive Behavioral Therapy (CBT) is the most widely used and empirically studied treatment for psychological and mood disorders, including negative body im age ( Cash & Lavallee, 1997; Jarry & Ip, 2005; Rosen, Reiter, & Orosan, 1995). In general, CBT is effective in a wide variety of populations including clinical and nonclinical populations with body-image disturbance (Thompson, Heinberg, Altabe, & Tantleff D unn, 1999). For example, Cash and colleagues (1996) developed a CBT for negative body image which is used in varying modes such as audiotapes and self help workbooks. Cashs program consists
28 of eight steps ( see Table 2 1 ; Cash & Strachan, 1999). A study b y Strachan and Cash (2002) investigated the efficacy of Cashs self -help CBT program. The 6 week intervention participants were 86 women and 3 men ranging in age from 18 to 63 with a mean age of 38 years. Participants were randomly assigned to one of two c onditions, both included psychoeducation and self -monitoring. Group one received steps one and two and group two received steps one, two, four, and five. At the completion of the study both groups improved significantly in body satisfaction ( ES = .33) and appearance evaluation ( ES = .23). The authors believe the absence of group differences may be a result of low compliance (attrition rate = 53%). Overall the effectiveness of Cashs self help program is promising for the prevention and treatment of bodyim age disturbances. A study by Lavallee and Cash (1997) studied the efficacy of Cashs self -help CBT program as compared to a self -esteem CBT intervention developed by McKay and Fanning (1987). The 9 week intervention consisted of 37 body dissatisfied indivi duals. Although both groups improved in body image investment, evaluation, and affect only the self -help CBT group had higher rates of functional recovery. Some positive aspects to Cashs program include the relatively low cost and anonymity. However there are some limitations as well. As seen in the Strachan and Cash study (2002) there is a high level of attrition (53%) associated with self -help programs. Another limitation to this program is the high level noncompliance and the possibility of individual difficulty carrying out self -help programs. A narrative review of stand alone body image treatments by Jarry and Berardi (2004) found that of the 18 studies reviewed, 17 studies used at least one CBT component, and 15 studies used Cashs CBT approach. Ni ne studies examined and found improvements in eating attitudes and behavior. These include restraint, overeating, and eating concerns with meals per a
29 day and binges per a week improving at follow up. Several studies also found improvements in psychologica l variables including general distress, self -esteem, and anxiety which previous research has linked with body image disturbance. Despite the positive findings of the standalone treatments, Jarry and Berardi (2004) noted several study limitations. First, the variability of body image measures across studies rendered it difficult to directly compare study outcomes. Furthermore, many studies ( N = 11) failed to include a placebo condition which would allow for comparison between groups and the effectiveness o f the intervention. Additionally, because most of the studies examined college populations, the generalizability of the studies to other populations is limited. Jarry and Berardi suggested that future research focus on exploring alternative approaches for bodyimage treatments and implementing specific components of treatment with multi-center studies to increase the intervention generalizability. Additionally, CBT is effective when offered in a variety of settings including group therapy, individual t herapy, and self -directed therapy (Cash & Lavallee, 1997; Cash & Strachan, 1999). However, Jarry and Berardi stated that while CBT is an effective treatment, it is the only stand alone empirically supported body image treatment. Thus, more research is need ed focusing on alternative treatments that are more cost effective and have the potential to reach large audiences. As well, while CBT is effective, more empirical support is needed and the measures used to assess outcomes must be streamlined to better understand the efficacy of CBT. The next section will discuss alternative treatments for improving body image. Alternative Treatments O ther methods that have been researched as promising treatments for body image including psychodynamic, experiential, ps ychoeducational, ecological, and weight loss The next
30 section will briefly review these alternative treatments and corresponding studies. The pros and cons of each treatment will also be discussed. Psychodynamic interventions Psychodynamic treatments fo cus on the integration of the body self and the psychological self through psychotherapy (Cash & Pruzinsky, 2002). This method allows the patient to recognize and articulate on basic sensations and feelings to find an internal frame of reference. The t herapist helps facilitate the individuals integration of the therapeutic process and achieves self regulation from the newly developed internal center of initiative, affects and esteem (p. 467) to establish a positive body image. A recent study by Wil tink et al., (2007) examined the effects of a psychodynamic psychotherapy treatment for severely obese individuals. Patients ( N = 267) at an inpatient rehabilitation clinic were randomly assigned to a psychodynamic or behavioral treatment group. Eighty -fiv e percent of the patients were female with a mean age of 41.3 years and a mean BMI of 44.3. Patients in the psychodynamic group received both individual and group therapy for an average of 7 weeks. Results of the study found a positive effect ( ES = .56) from post study to 3-year follow up for improvements in body image regardless of treatment group or weight regain, while the results for weight loss were small ( ES = 0.26). The authors assume that there are long term benefits to both psychodynamic and behavioral psychotherapy. They also note the absence of a control group as a limitation to their study. While these results are promising, more research is needed in order to improve the effectiveness of psychodynamic treatments as well as determine more co st effective modes of delivery. Experiential interventions The experiential approach to body image treatment involves a multidimensional approach which includes mental, sensory, and somatic components ( Cash &
31 Pruzinsky, 2002). Varying experiential techni ques include but are not limited to mental imagery, hypnosis, music therapy, art therapy, breathing/relaxation exercises, body talk, and feminism. A study by Dibbell Hope (2000) investigated the effects of a dance/movement therapy in women with breast ca ncer. Thirty three women, with a mean age of 54.7, participated in the 6 week program. Women in the Authentic Movement group attended weekly 3 hour sessions while the women randomly assigned to the control group were given the opportunity to join the Authe ntic Movement group at the end of the 6 weeks. The goal of Authentic Movement is to help the individual to become aware of the body and the self through exploration of feelings and external movement. Subjective data from this study show improvements in bot h body image and self -esteem for women in the Authentic Movement group. Most of the research in this area has been done using individuals with eating disorders (Cash & Pruzinsky, 2002). Therefore, more research is needed with varying populations including clinical and nonclinical to determine the efficacy of experiential treatments. Furthermore, future studies require well -designed research, implementation of randomized control groups, and stand alone interventions. Psychoeducational interventions P sychoeducational treatments involve programs delivered through various modes such as audiotapes, videotapes, print, or internet; without the use of a therapist (Cash & Pruzinsky, 2002). Most psychoeducational programs are theory based and vary from cogniti ve behavioral theory to psychodynamic, sociocultural, or feminist theory. Winzelberg and colleagues (2000) investigated the effectiveness of a nontheory-based psychodynamic internet -based intervention on body satisfaction and weight/shape concerns (Winzelb erg et al., 2000) They randomly assigned 60 women ( M age = 20.0 years) to either a computer assisted health education group (CAHE) or control group. The CAHE program
32 consists of an eight week internet based intervention focusing on improving body image. Components of the program included audio and video software, self -monitoring journals, behavior change exercises, and weekly assignments and discussion groups. They found no significant outcome group differences at post intervention. At the three month follow up however the intervention group had significant improvement in both body image and drive for thinness compared to the control group. Although positive effects were evidenced, the authors cited several study limitations including compliance, with m ost participants completing less than two thirds of the program. More recently a study by Gollings and Paxton (2006) compared internet and face to -face group body image interventions. They found that both groups had significant improvements in body image and eating behavior from pre to post intervention. Limitations of this mode of treatment include the decrease in level of individual adherence over time and lack self -direction and motivation to adhere to the program material. Thus, internet -based interventions show promise as a useful program for improving body image, and may be better geared towards individuals who are not inclined to participate in a group setting. Ecological interventions Ecological based interventions are guided by the idea that soci al and cultural factors play a role in causing body image disturbances (Cash & Pruzinsky, 2002). This is done through the use of primary prevention programs aimed at adolescents and children and guided by cognitive -behavioral and social learning theories. McVey and Davis (2002) investigated the effects of a prevention program using a life skills promotion approach. The study involved 282 girls in Grades 6 to 8, with a mean age of 10.88 years. Girls in the prevention group participated in a six week program focused on promoting positive self -esteem/body image,
33 uncovering unrealistic body ideals set by the media, and education on healthy eating and exercising. Results indicate that body image improved for both the intervention and control group at posttest, 6 -month, and 12-month follow ups. As demonstrated by the McVey and Davis (2002) study, r esults with ecological treatments are inconsistent and mostly short lived. Future areas of research should exami ne ma le body image disturbances and the potential role of parents in promoting body image (Cash & Pruzinsky, 2002; McVey & Davis, 2002). Weight loss interventions. Another type of body image intervention is aimed at integrating weight loss and body image treatments. Often standard weight loss programs do not address the issue of body image in the goal to lose weight (Cash & Pruzinsky, 2002). Although weight loss alone is often enough to improve body image, these results can be fleeting with even minimal weight regain. Very few studies have been done examini ng the effects of a combined weight loss and body image intervention and none of them have been long term (Cash & Pruzinsky, 2002) One of the few studies investigating the effects of a weight loss program in conjunction with a body image intervention was by Ramirez and Rosen (2001). This study consisted of 14 men and 51 women ( M age = 44.0) randomly assigned to a weight control or weight control plus body image therapy group. The weight control intervention consisted of 16 weekly 1-hour sessions focused on nutrition and eating/ exercise change. The weight control plus body image therapy consisted of the same weekly weight control sessions followed by an hour of CBT body image therapy. They found improvements in body image for both groups at posttest, 3 -mont h follow up, and 1 -year follow up. Moreover, there was no difference between groups in weight loss at
34 posttest or follow ups with an average weight loss retention of 53%. Ramirez and Rosen (2001) postulated that body image therapy did not have a greater ef fect than weight control because weight loss on its own has a strong, positive effect on body image. They conclude that while weight control may be enough to improve body image, some individuals may need the added assistance of body image therapy to achieve these same effects. These are just a few examples of alternative interventions focused on improving body image and reducing disordered eating. While each area has its shortcomings it is evident that more research is needed in all areas to determine t he effectiveness and generalizability of each. Furthermore, not all interventions are appropriate for all individuals. In order to help a broader range of individuals with varying needs, interventions should be tailored to insure that all target populations are reached. Exercise Researchers have also focused on exercise as a useful intervention for preventing and treating bodyimage disturbance. There are various reasons why people engage in physical exercise. Physical, psychological, and social benefits are all factors in determining exercise behavior (Bryan & Rocheleau, 2002). Exercise reduces depression and anxiety, enhances cognitive functioning, and can aid in treating psychiatric disorders (Callaghan, 2004). Exercise has also been used as an effectiv e intervention in individuals with negative body image, bulimia nervosa, and binge eating disorders (Fisher & Thompson, 1994; Hausenblas, Cook, & Chittester, 2008). When compared to CBT as an effective treatment for body image disturbance, Fisher and Thomp son (1994) found that the positive benefits of exercise were equivalent to that of CBT. While CBT may not readily available or cost -effective for all individuals, exercise can be done with minimal cost in a variety of settings.
35 Bearing in mind that body i mage disturbance is a criterion for eating disorders perhaps it is also important to consider exercise interventions for the prevention and/or treatment of eating disorders. Understandably, exercise is not considered a standard intervention for preventing/ treating eating disorders given that excessive exercise is attributed as a behavioral feature of both anorexia nervosa and bulimia nervosa (Thurstin, 1999). Hausenblas, Cook, and Chittester (2008) argue that exercise may be beneficial for some individuals suffering from eating disorders, once given medical clearance by a physician. They propose a conceptual framework for the effects of exercise on eating disorders which takes malleable physiological, psychological, and social risk factors into account. This framework relies on a reciprocal relationship between exercise and improvements in physiological, psychological, and social factors as well as the reciprocal relationship between improvements in these factors and a decrease in eating disorder risk factors and prevalence. Likewise, this framework can be applied to the relationship between body image and exercise ( see Figure 2 1 ). Physical exercise has been researched extensively and has been shown empirically to reduce anxiety, stress, and depression (Ca rron, Hausenblas, & Estabrooks, 2003). A narrative review of 79 studies by Fox (2000) found that exercise had a positive impact on self esteem and body image, and even more so for individuals suffering from low self -esteem. Additionally, exercise is known to reduce cardiovascular disease, improve cardiac functioning, prevent osteoporosis, aid in sleep, and alleviate pain (Carron, Hausenblas, & Estabrooks, 2003). It is evident that there is ample empirical evidence to support the positive effects of exercise Just as exercise may be a beneficial treatment for eating disorders, it should similarly be considered an effective treatment for improving body image, as it is a precursor to eating disorders
36 A recent meta analysis by Hausenblas and Fallon (2006) revi ewed the impact of exercise on body image. One hundred and twenty one studies were included in the analysis and the overall mean effect size was 0.28 Specifically, a larger effect size was found for woman ( ES = 0.45) than for men ( ES = 0.26). Comparing age groups, the largest effect was found in adolescents ( ES = 0.98) followed by adults ( ES = 0.40) and elderly adults ( ES = 0.26) while university students had the smallest effect ( ES = 0.17). When comparing type of exercise performed the combination of aerobic and anaerobic had the most significant effect size ( ES = 0.39) and anaerobic or aerobic exercise alone had comparable effects sizes, ES = 0.36 and ES = 0.34 respectively. Hausenblas and Fallon (2006) also reviewed correlational interventions (68 stud ies) and found an overall mean effect size of 0.41 indicating that exercisers had more positive body image than non-exercisers. Analysis of single -group interventions (44 studies) revealed an overall mean effect of 0.24 demonstrating that exercisers had be tter body image post intervention as compared to pre intervention. As well, analysis of experimental versus control group interventions (35 studies) had an overall mean effect size of 0.28 revealing that participants in the exercise group had more positive body image post -intervention as compared to the non-exercising group. A study by Asci (2003), examined the association between physical self concept, trait anxiety, and physical fitness in college females. Forty sedentary college females participated i n an aerobic or step dance classes three times a week for fifty minutes, for ten weeks. They found that exercise was effective in strengthening physical self -perceptions and reducing trait anxiety compared to the control group. Although, Asci cautions that while exercise may be a useful means to improving physical self -concept, more research is needed using varying intensities, duration, and types of exercise to overcome methodological weaknesses.
37 Using a different population, Lindwall and Lindgren (2005) examined the effects of a six month exercise intervention on physical self -perceptions and social physique anxiety in sedentary adolescent girls. The intervention group met twice a week for forty five minutes. The girls participated in a variety of aerobi c activities such as kick boxing, spinning, dancing, and water aerobics. Results indicate overall that physical self -perceptions and social physique anxiety improve with exercise in comparison to control group. In 2001 Williams and Cash did a study investi gating the effects of a circuit weight training program on body image in undergraduate males and females (Williams & Cash, 2001). Participants engaged in strength training exercises for a total of three hours a week over a period of six weeks. In compariso n to the control group, the strength training participants significantly improved on appearance evaluation, body dissatisfaction, social physique anxiety, and physical self -concept. Williams and Cash conclude that even an exercise program of short duration is successful in improving body image. Yoga has also started to gain recognition as a useful method of exercise intervention to treat body image concerns. A study done by Daubenmier (2005) examined the effect of yoga practice on body satisfaction, self -ob jectification, and eating disordered attitudes. Aside from yogas many health benefits, the main goal of yoga is to focus on greater body awareness and responsiveness to bodily sensations. Daubenmier hypothesized that increases in bodily awareness and resp onsiveness through yoga practice can lead to a reduction in the value placed on physical appearance and self -objectification. Participants included adult (Study 1, M age = 37.16) and undergraduate (Study 2, M age = 20.46) women participating in yoga or aer obic exercise classes as well as a base line group. Participants were recruited out of classes practicing Iyengar or Astanga yoga which were each an hour and a half long. Daubenmier found that
38 women who practiced yoga had greater satisfaction with physical appearance, and less self objectification as opposed to nonyoga groups. However, there are some shortcomings to this study. Because the study only consisted of a onetime survey given as individuals were leaving a yoga class, it is hard to say how long or often yoga must be practiced to achieve these benefits. Furthermore, most college and adult women sampled were either European American or Asian American making it difficult to generalize the findings to other ethnicities and age groups. Daubenmiers stu dy is one of the first studies to examine yoga as an intervention for body image and eating disturbances but more research is needed to determine its effectiveness as a treatment/prevention program. On the other hand, some researchers have found that exerc ise does not improve aspects of body image. For instance, McCabe, Ricciardelli, & Salmon (2006) examined the effects of a prevention program to improve body image and negative affect in adolescent boys and girls. The program consisted of eight weekly, fort y minute sessions with varying types of psychical activity and sports. The intervention showed no significant change in body dissatisfaction for boys or girls. In fact, only the boys levels of negative affect were positively affected. Additionaly, severa l studies (Loland, 2000; Katz, 1986; Silberstein et al., 1988; Tiggeman & Williamson, 2000) suggest that exercise has a negative impact on body satisfaction. Tiggemann and Williamson (2000) found that as exercise increased, body satisfaction and self este em decreased. Zabinski et al. (2001), in a study of three hundred and thirty-eight undergrad males and females determined that womens drive for thinness actually increased while body image did not change over the course of a fifteen week aerobic and anaer obic intervention. Caution must be taken when implementing exercise interventions in populations that may be at risk for developing eating disorders or exercise dependence. Much the same as
39 internalization of the thin ideal, exercise dependence can also be seen as a risk factor for the development of eating disorders (Hechler et al., 2005). Silberstein et al. (1988) and Katz (1986) both found that exercising for reasons related to appearance could perpetuate the risk for eating disorders. Additionally, Loland (2000) argued that the positive effects of physical activity on body image are linear with age. Physical activity in younger women (under 25 years of age) is not related to body satisfaction, but is in older women. It is evident that the studies expl oring body image and exercise interventions vary widely in age, gender, ethnicity of target population, length of intervention, measures assessed, as well as type and intensity of exercises performed. It is important to look at these moderators in order to understand the effect that exercise has on body image, as described in detail in chapter one. The solution to improving body image can not be boiled down to one specific treatment and some approaches are easier to self -determine than others. Additionally, in order to gain a better understanding of the impact of exercise on body image more empirical evidence is needed to determine if exercise can be considered as effective or more effective than CBT as a treatment in improving or decreasing negative body im age. The next section will cover the importance of meta analysis as a tool in exploring exercise interventions as well as its advantages and disadvantages. Meta -Analysis Meta analysis is a statistical analysis of a large collection of analysis results fro m individual studies for the purpose of integrating the findings (Glass, 1976, p.3). Meta analytic techniques enable examination of moderator, design, and methodological variables that may explain the effect size heterogeneity (Biddle, 2006; Hagger, 2006) Effect sizes are the strength of the association between variables, within a study (Rosenthal, et al., 2006). With such an
40 abundance of research, using a meta analytic approach allows for a more concise and credible conclusion then would be possible by a ny one study alone (Rosenthal & DiMatteo, 2001). Narrative reviews provide a good understanding of a body of literature. However, with studies ranging so vastly with regard to methods, measures, and operationalizations it is impossible to gain a clear pic ture through qualitative analysis alone. Through the use of quantitative review it is possible to both account for these inconsistencies as well as identify moderators and mediators within a body of research (Rosenthal & DiMatteo, 2001). Meta analysis offe rs many advantages over stand alone studies or narrative reviews. Foremost, meta analysis requires a thorough search of the relevant research including published and unpublished data to allow for both significant and non -significant findings (Rosenthal & D iMatteo, 2001). Furthermore, to obtain data from individual studies meta analyst must be comprehensive in their reading of an article. To effectively calculate effect sizes, the meta analyst must evaluate all aspects of an articles methods, measures, and operationalizations for inclusion/exclusion criteria. This allows for a more thorough understanding of the literature. Third, as Rosenthal and DiMatteo (2001) stated, emphasizing exploration instead of confirmation of important patterns of correlation be tween moderators and effect sizes allows for examination and reconciliation of differences among studies and adds to theory development and increases the richness of empirical work (p. 66). On the other hand, while there are numerous advantages to meta analysis there are also disadvantages. Due to the method of review and inclusion/exclusion chosen for a meta analysis some bias sampling is likely to occur (Rosenthal & DiMatteo, 2001). Not only is it possible to have bias in sampling but also bias due to lack of sufficient data provided by researchers to compute effect sizes. Another criticism is the issue of the quality of studies included in a meta -
41 analysis. With countless variations between studies it is also likely there is variation in the quality of the studies (e.g., published vs. unpublished data). This issue can be resolved with the use of a weighting technique which accounts for the methodological strengths and weakness of individual studies (e.g., journal impact factor, published vs. unpublished, experimental vs. quasi experimental designs; Rosenthal, 1991). Regardless of its disadvantages, meta analysis is a valuable method that can be used to examine the effects of exercise interventions on body image. A review of the exercise and body image li terature is certainly warranted as there is only one meta analytic review available (Hausenblas & Fallon, 2006). However, the Hausenblas and Fallon study consisted of mostly correlational studies ( N = 68). There is also a need to focus on interventions usi ng both random and fixed effect sizes and controlling for pre intervention scores. With the abundant amount of research literature available in the field of body image research synthesis of selected areas are necessary for research to gain an understanding of the impact of body image and future research directions. A meta analysis of the exercise and body image literature can be useful in examining how the key features of the research design, exercise intervention types, and group condition types, all moder ate the effect of exercise on body image. This synthesis of the literature will help to gain a better understanding of the body image literature, the effects of exercise on body image, moderators of these effects, and generalizability of the findings.
42 Table 2 1. Cashs eight step cognitive behavioral therapy Goals Step 1 B ody image self assessment questionnaires and goal setting Step 2 E ducation and self discovery of participants body image Step 3 D esensitization through mental imagery and relaxa tion skills Step 4 Identifying appearance assumptions or core bel iefs about appearance, and working to challenge these beliefs Step 5 Identifying cognitive distortions and using cognitive restructuring to modify these behaviors Step 6 R ecognizing and al tering avoidant behaviors and obsessive compulsive patterns Step 7 B ody image affirmation activities to enhance and reinforce positive body image Step 8 M aintenance and relapse prevention Reference: Cash, T.F., & Strachan, M.D. (1999). Body images, e ating disorders, and beyond. In R. Lemberg & L. Cohn (Eds.), Eating disorders: A reference sourcebook (pp. 27 37). New York: Greenwood. Figure 2 1. Conceptual framework for the effects of exercise on body image. Modified from Hausenblas, Cook, & Chittes ter (2008). Can exercise treat eating disorders? Exercise and Sport Sciences Reviews, 36(1), 4347.
43 CHAPTER 3 METHODS Sample of Studies Research related to body image and exercise was retrieved by February 2008 using the following four procedures to avoid bias retrieval of searching only major journals and to obtain fugitive studies ( Barber & Milrod, 2004; Rosenthal, 1991 ). First, Dr. Hausenblas and myself conducted computer -based searches in PubMed, Cochrane Controlled Trials Register, Cochrane Database o f Systematic Reviews (Clinical Trials) Online Journal s Search Engine Dissertation Abstracts International, and PsycINFO using the following key words: body image, exercise, physical activity, eating disorder, eating pathology, and body satisfaction /dissa tisfaction Second, ancestry searches, sometimes called treeing backward, were conducted using the references lists of all intervention research. We also manually searched all available issues of the pertinent journals in the field (e.g., Body Image: An International Journal; International Journal of Eating Disorders; see Appendix A for complete list) Third, we contacted active researchers in the field to retrieve either current or unpublished research (e.g., in press, in review). Active researchers in the field were defined as all authors with published articles within the last year (Hopewell, Clarke, Mallett, 2005) Researchers of studies already included in the meta analysis as well as authors of excluded studies due to insufficient data were contacted t hrough email for relevant data ( see Appendix B & C ). Of the 29 active researchers in the field contacted to retrieve either current or unpublished research 9 responded resulting in the retention of 7 studies and exclusion of 2. Fourth, computerized searche s were conducted on all authors of retrieved studies meeting the inclusion criteria. Finally, often because of the magnitude of controlled trials, multiple publications will occur. In an attempt to effectively code the studies, we also retrieved other
44 publ ications related to the trial in an attempt to comprehensive record the moderator variables (e.g., Sallis et al., 1999). Selection Criteria Criteria for inclusion were that: (a) the independent variable involved a chronic exercise program of at least 3 we eks (Puet z et al., 2006); (b) the dependent variable was a measure of body image that was assessed before and after an intervention involving chronic exercise; and (c) the design was experimental. We excluded studies that were cross -sectional, correlationa l, or did not have a control group because it is impossible to differentiate between spontaneous changes in body image over time as opposed to the effects of exercise. Studies of behavioral interventions which offered either education and/or advice on incr easing physical activity, or structured supervised/unsupervised physical activity exercise programs were considered for inclusion. If a study had pre, mid, and post data (e.g., Asci 2003), we used the pre and post data only to compute effect sizes. Thus, we focused exclusively on studies that tested whether the change in the outcomes over time was significantly greater in the intervention group versus the control group. It is necessary to control for initial levels of the outcome variable because otherwise the analyses are not providing a test of differential change over time across conditions (Stice et al., 2006). A total of 77 studies were excluded from the analysis because they did not meet the inclusion criteria, with the majority of studies excluded du e to lack of control group ( n = 24), as well as no true control ( n =16), not enough information to compute an effect size ( n = 15), no measure of body image ( n = 12), limited exercise component ( n = 7), and post data only ( n = 3). Coding the Studies Coding was performed by Dr. Hausenblas and myself independently ( see Appendix D for coding sheet ). Disagreements were resolved by discussion and by further examining the studies (Orwin, 1994). All the coded characteristics were used as descriptions of the studie s retrieved
45 and as potential moderator variables (Rosenthal, 199 1 ), in addition to the main moderators examined. Appendix E lists the numeric values used to code each moderator, the operationalization of each moderator, and relevant descriptive statistics describing the distribution of the moderators. We coded certain moderators two ways in an effort to ensure that we were not missing the effects of a moderator, because we did not operationalize it optimally. A priori lists of outcome measures to select whe n multiple measures were present in a study were developed to minimize the impact of coder bias on the selection process. Participant features. We coded the following participant features: age, gender, ethnicity, risk status, body composition and preinter vention fitness level. In addition to coding the average age of participants at baseline, we also coded for age category of elementary school, middle school, high school, university, adults, or older adults. With regard to participant ethnicity, we coded b oth the percentage of participants who were Caucasian (a continuous variable), because this group is at high risk for negative body image compared to non Caucasian and the dominant ethnic group represented in the sample (nominal variable). Exercise intervention features. We coded the exercise type (i.e., aerobic, resistance training, both), duration (i.e., minutes per session), length (i.e., length of intervention in weeks), frequency (i.e., number of sessions per week), and intensity (i.e., strenuous, moderate, mild). The intensity of aerobic exercise was coded using the classification system of the American College of Sports Medicine (2000, p. 150). In this system, intensity can be classified as a percentage of oxygen uptake reserve (%VO2R), a relative m easure of intensity, which permits consistent coding of intensity whether expressed as percent oxygen uptake (%VO2max), heart rate, or perceived exertion (Howley, 2001). The database allowed for the formulation of light (20 39% VO2R), moderate (40 59% VO2R), and strenuous (60 84% VO2R) categories based on these
46 guidelines. In cases where only a verbal description regarding the intensity of exercise rather than specific information such as maximum oxygen consumption, maximal heart rate, or hear rate re serve was provided, two experts in exercise physiology made a subjective assessment of the exercise intensity referring to the ASCM definitions. Intensity of resistance training was assessed in terms of repetitions and workload using standard tables (Bompa 1999). Two experts in resistive training made a subjective assessment in cases where only verbal information was provided. Following the procedures of Colcombe and Kramer (2003) the duration was also coded categorically as short (15 30 min), moderate ( 31 45 min), and long (> 45 min). The length was also coded as short (1 3 months), medium (4 6 months), and long (6 + months). The intervention specificity variable was coded into two categories: exercise only and exercise in addition to another treat ment. Finally, we coded if the experimental and control participants showed improvements on physical fitness. Design and study features. We coded for the type of control group, recruitment method, publication year (as a measure of rec ency), random assignment, type of exercise intervention (i.e., exercise based or lecture based), and publication status. Studies which did not report randomization or group assignment methodology were assumed to have used nonrandomized protocol. Although b ody image is generally conceptualized as a broad, multifaceted construct, most research in this area has focused on the narrower construct of weight/body dissatisfaction (Grogan, 2006). Each of the body image measures was coded independently. If we did not know what the measure assessed in particular for body image, we retrieved the article that contained the scale and examined the individual items to determine what type of body image the scale assessed.
47 Effect Size Calculation and Analytic Strategy Effect size calculations. Using random and fixed effects modeling procedures, I calculated effect sizes by subtracting the mean change for a control group or condition from the mean change for an experimental group or condition and dividing this difference by the pooled standar d deviation of pretest score. I adjusted all effect sizes using Hedges and Olkins (1985) small size bias correction before entering them into the analysis. When the N at the pretest differed from the N at the posttest, the smaller N was u sed. If the results were available for more specific subgroups (e.g., men and women; high school and middle school), effects were computed for the most specific group for which data were available. This procedure enabled a more comprehensive examination of the moderator variables of interest. When precise mean and standard deviation data were not reported, effect sizes were estimated (Rosenthal, 1991) from F tests, t tests, p values, or figures. For studies in which precise standard deviations were not repo rted, the standard deviation was drawn from published norms (e.g., Garner, 1991). I also estimated effect sizes when a report contained inexactly described p values such as when the authors indicated that a given finding was not significant at 0 .05. Thus, a reported nonsignificant finding was estimated to have a probability of 0 .99, whereas a significant finding was estimated to have a probability at the level of the cutoff value used in the study (e.g., 0 .05 or 0 .01). However, because the use of such repor ts may l ead to incorrect estimations, I conducted separate analyses on the set of exactly reported effect sizes and all the effect sizes (including the ones estimated on the basis of inexactly reported p values). Because these sets of anal yses yielded simi lar results, I report ed only the results that included all the effect sizes. The effect size computation methods were coded to examine if the use of different effect size estimation methods moderated the effect size (Ray & Shadish, 1996).
48 Along with the w eighted average effect sizes, I computed the 95% confidence intervals. If the confidence interval does not include zero, then the mean effect size is statistically significant at the p < .05 level. I also graphed a forest plot, which is a graph of each stu dy as a point estimate bounded by its confidence intervals. The forest plot shows at a glance the following: (a) if the overall effect reported in the analysis is based on many studies or a few; (b) if the overall effect is based on studies that are either precise or imprecise; (c) whether the treatment effects for all studies tend to line up in a row, or whether they vary substantially from one study to the next; and (d) if outliers exist (Borenstein, 2005). To determine heter ogeneity of the effect sizes, I reviewed the actual disp ersion on the forest plot and calculated both the Q -statistic and Isquared. Q tests the hypothesis that the observed variance in effect sizes is no greater than that expected by sampling error alone. Under the null hypothesis tha t all studies derive from the same population, Q will follow a chi -square distribution for df = k where k is the number of outcomes minus one. The p -value for Q like all p values, should not be interpreted by rote (Borenstein & Rothstein, 1999). That is, the absence of a significant p -value cannot be taken by itself as evidence of homogeneity as it could reflect a low power rather than actual consistency. Because of the technical and conceptual problems with the Q -Statistic, I also calculated the I -Square d, defined as variance (between -studies)/variance (total), to quantify the variance (i.e., heterogeneity; Higgins & Thompson, 2002; Higgins, Thompson, Deeks, & Altman, 2003). That is, the Q -statistic indicates whether or not there is evidence of dispersion the I -squared quantifies the dispersion. For interpretation, the I -squared values of 25, 50, and 75 are considered low, moderate, and high, respectively. Thus, an I -squared in the low range suggests that the effect sizes are homogeneous relative to the p recision of the individual studies (Higgins et al., 2003). For moderator analyses, I used QB using mixed effects
49 analysis, to explore the impact of categorical vari ables on the effect size; and I used meta regression to explore the impact of continuous variables on the effect size. Mixed -effect models provide a more stringent test of moderators and help diminish the possibility of Type I errors, which can become inflated in the fixed-effects with moderators approach to meta analysis (Overton, 1998). Both the random and fixed effect size information is reported in the appendix (see appendix F ) but for brevity in the text only the random effect size information will be reported. Furthermore, moderator analysis will only be undertaken when there are a minimum of 3 effect sizes (Wolf, 1986). Data were analyzed using SPSS 15 and Comprehensive Meta analysis 2 (BioStat, Englewood, New Jersey). Dependence. A fundamental assumption of most standard analyses used in meta analysis is the independence of effects. When multiple outcomes have been measured for the same individuals, or when the same outcome is measured at several time points for the same individuals, effect sizes computed for those multiple outcomes or time points will not be independent. A number of ways exist to deal with dependence (Becker, 2000). The primary approach used in our synthesis was to separate effects into groups that included primarily independent effect. For example, some studies had two control groups (i.e., waitlist control and placebo co ntrol). In these instances we would code for both of the control group s It was common for studies to include multiple assessments of body image. In these instances, each independent group was limited to one body image effect size. The removal order was fi rst unstandardized questionnaires. In studies with two or more effect sizes still remaining, the rule was to give priority to body dissatisfaction measures ( Groesz, Levine, & Murnen, 2002; Stice, 2000). Body dissatisfaction is one of the most consistent and robust risk and maintenance factors for eating pathology; thus, it was selected over other body image measures. If, at this point, there
50 was still two or more remaining body standardized dissatisfaction measures, random removal was conducted until one ef fect remained per independent group. If studies reported on midexperiment data or follow up data; we recorded the immediately pre and post intervention scores only. Publication/Dissemination Bias Publication bias is the term for what occurs whenever the research that appears in the published literature is systematically unrepresentative of the population of completed results (Rothstein, Sutton, & Borenstein, 2005) To assess publication bias, I undertook the following graphical and statistical methods: fo rest plots, funnel plots, Fail Safe N or file drawer analysis using both Rosenthals (1979) and Orwins (1983) procedures ( Nfs) Eggers test of intercept, and Duval and Tweedies trim and fill These procedures are described in more detail below (see Effe ct Size Calculation section for information regarding the forest plot). Funnel plot. The funnel plot is a plot of a measure of study size (i.e., standard error) on the vertical axis as a function of effect size on the horizontal axis (Borenstein, et al., 2007). Large studies appear toward the top of the graph, and tend to cluster near the mean effect size. Smaller studies appear toward the bottom of the graph and tend to disperse across a range of values due to the sampling variation in effect size estima tes. In the absence of publication bias the distribution of studies would be symmetrical about the combined effect size. On the other hand, in the presence of bias the bottom of the plot would show a higher concentration of studies on one side of the mean than the other. This would reflect the fact that smaller studies are more likely to be published if they have larger than average effects, which makes them more likely to meet the criterion for statistical significance. This will result in an overestimatio n of the treatment effect in a meta analysis (Sutton et al., 2000). The funnel plot offers a visual sense of
51 the relationship between effect size and precision, but the interpretation of the plot is largely subjective. Fail safe N. Fail Safe N a ddresses t he possibility that studies are missing from the analysis and that these studies, if included in the analysis, would shift the effect size toward the null. The classic Rosenthals Nfs indicates the number of missing studies (with M effect of 0) that would need to be added to the analysis before the combined effect would no longer be statistically significant (Rosenthal, 1991 1979). The Nfs analyses were calculated for each statistically significant relation reported. According to Rosenthals (1979) conserv ative guidelines, Nfs should exceed 5k (k = number of studies) + 10 to effectively overcome the file drawer problem. Rosenthals Nfs is limited in two ways (Borenstein, et al., 2007) First, it assumes that the effect in the hidden studies is nil, rather t han considering the possibility that some of the studies could have shown an effect in the reverse direction. Therefore, the number of studies required to nullify the effect may be smaller than the Nfs. Second, this approach focuses on statistical signific ance rather than clinical or substantive significance. So while it allows for the assumption that the treatment is not nil, it does not address the question of whether or not it remains clinically important after the missing studies have been included. Be cause of the limitations of Rosenthals Nfs I also computed Orwins Nfs, which addresses the limitations of Rosenthal (Borenstein, et al., 2007) First, it is possible to specify the mean effect in the hidden studies as being a value other than n il Second the criterion value is an effect size rather than a p value. Orwins Nfs is the number of missing studies that when added to the analysis, will bring the combined effect size below a specified threshold The minimal effect size chosen was 0.1 (VanderWerf 1992)
52 Eggers test of the intercept. Eggers test of intercept is a linear regression method used to quantify the bias captured by the funnel plot (Borenstein, 2005). The standard normal deviate is regressed on precision, defined as the inverse of the s tandard error. The intercept in this regression corresponds to the slope in a weighted regression of the effect size on the standard error. In the absence of asymmetry, the points on a funnel plot will scatter about a line that runs through the origin at s tandard normal deviate zero, intercept 0 = 0, with slope 1 indicating the size and direction of the effect (Sterne & Egger, 2005). If there is funnel plot asymmetry, the regression line will not run through the origin, so that the intercept 0 provides a measure of asymmetry. A test of the null hypothesis (i.e., 0 = 0) can be derived from the usual regression output from CMA, reporting the two tailed p -value. The larger the deviation from zero the more pronounced the asymmetry. The power for this test is usually low unless there is severe bias or a substantial number of studies. Duval and Tweedies trim and fill. The trim and fill procedure is a nonparametric technique that examines the symmetry and distribution of effect sizes plotted against the inverse of the standard error (Borenstein, et al. 2007; Duval & Tweedie, 2000). T he method initially trims the asymmetric studies off the asymmetric outlying part of the funnel plot to locate the unbiased effect and then fills the plot by re inserting the trimmed studies on the right as well as their imputed counterparts to the left the mean effect. First, the technique estimates the number of studies that may be missing as a result of publication bias, with publication bias meaning studies with effect sizes that are low or near zero r elative to the average effect. Then, the trim and fill calculates hypothetical effects for potentially omitted studies and then re -estimates the average effect size and confidence intervals on the basis of the influence of studies tha t would have been included in the analyses if they had been published.
53 Missing data. Missing information to calculate an effect size will occur when studies provide no statistics or an inadequate amount of information about the outcome scores to calculate an effect size. Some researchers fill in a conservative estimate, such as zero or the mean effect, for missing effect sizes. Imputing a single value for missing effect sizes may lead to biased results, which may be compounded when those imputed values are used to estimate the variance of the missing effect size (Pigott, 1994). Some researchers use the variance of the effect size as weights in the estimated mean effect size and in weighted least squares estimation of the linear model of effect sizes. Both o f these methods, however, result in biases in the mean effect size, and therefore, are not reco mmended (Pigott, 1994). Thus, I will undertake the following two procedures to control for this type of missing data. First, I will contact the authors of studie s that fail to provide adequate information to calculate an effect size in an attempt to obt ain this information. Second, I will use the procedures described by Bushman and Wang (1995, 1996) that combine sample effect sizes and vote -counts to examine missi ng data biases between: the included studies with the excluded studies that provide insufficient information to compute an effect size but do provide the direction of the effect (Bushman, 1994). Hence, these methods combine sample effect sizes and vote cou nts to estimate the population effect size. These methods allow all the data available from each study to be used in the final analysis, and hence are superior to simpler approaches such as omitting studies where no effect size is given, or imputing values such as zero, or the mean effect for the outcome in studies where no outcome is reported. Outliers. Potential outliers were identified by examining the effects sizes graphically and then omitting one case at a time and checking for large externally standa rdized residuals or substantially reduced measures of heterogeneity (Hedges & Olkin, 1985). These analyses were
54 conducted for each outcome variable ( see Appendix G) The identified outliers were examined to determine if characteristics unique to the study may have produced the extreme scores. The effect of outliers is often a notable increase in observed variance and a distortion of the mean. When sample sizes are small to moderate (the usual case), extreme values can occur because of large sampling errors. Such values are not true outliers and should not be eliminated from the data, because the formula for sampling error variances assumes and allows for such occasional large sampling errors. Eliminating such nonoutlier extreme values can result in overcorre ction for sampling error and underestimation of standard deviation. Because of this, I did not remove any outliers in the analyses (Hunter & Schmidt, 2004).
55 CHAPTER 4 RESULTS Description of Studies A total of 57 publications ( M study year = 1997, range = 1972 2007) with 98 separate comparisons were included in the meta analysis ( see appendix H for individual study information). Reasons for multiple effect size s per study were that studies reported results separate by age ( n = 2 ) and/or gender ( n = 6 ), an d studies with either multiple control groups ( n = 7 ) or exercise groups ( n = 15; two studies reported results by both gender and age ). The studies had a total 6,273 participants (Experimental N = 3,639, Control N = 2,634) Most studies were conducted in t he United States ( n = 38), followed by United Kingdom ( n = 4), Canada ( n = 3), Turkey ( n = 3), Germany ( n = 3), Australia ( n = 2), Norway ( n = 2), Sweden ( n = 1), and Switzerland ( n = 1). Most studies were disseminated as journal articles ( n = 49); the re m ainder were dissertations/these s ( n = 8). All but one study included risk status, with most of the populations being universal ( n = 42), followed by selected ( n = 12), breast cancer participants (n = 2), and participants with psychological disorders ( n = 2). The average participant age was 30.04 ( SD = 15.35, range = 10.02 to 63.40), with 9 studies not reporting the participant age. For age group category, most participants were university students (35.7%), followed by adults (26.5%), older adults (12.2%), elementary (7.1%), middle (8.2%), and high school (2.0%); with 3 studies not reporting age group information. Most studies included female populations ( n = 31), followed by both genders (n = 22), and male populations ( n = 4; 1 study did not report gender i nformation). Only 18 studie s reported ethnicity, with Caucasian participants being the dominant ethnic group in 15 studies, followed by African American ( n = 2) and Hispanic ( n = 1). Within these studies, t he percentage of Caucasian participants ranged fro m 19 to 100% with an average of 64.8% ( SD =
56 25.67). Participants body composition was described in 32 studies; with 18 studies consisting of normal weight participants and 14 studies consisting of overweight/obese participants. Finally, 28 studies described preintervention fitness level, with 27 studies reporting sedentary/low activity participants, followed by active participants ( n = 1). Fifty -four studies described the exercise mode. Most of the interventions used aerobic exercise ( n = 32), followed by interventions with a combination of resistance training and aerobic exercise ( n = 14), resistance training only ( n = 7), and one study which had participants engage in either aerobic exercise or tai chi. The average exercise duration was 49.09 min utes (SD = 14.29, range = 20 to 75 min utes ) with 12 studies not reporting exercise duration. A total of 54 studies described the length of intervention in weeks with an average length of 12.69 ( SD = 8.14, range = 4 to 52 weeks). The average exercise f requency pe r week was 2.81 (SD = 1.04, range = 1 to 5 times per week ), with 10 studies not reporting exercise frequency. Intensity of exercise was described in 31 studies, with most participants performing at a moderate intensity (n = 20), followed by hard or very ha rd ( n = 9), light ( n = 1), and one study that had participants exercise at both moderate and hard intensity. Only 10 interventions met the ACSM (2000) physical activity guidelines (20 to 60 minutes of continuous or intermittent exercise 3 to 5 days week). All but one study reported exercise specificity, resulting in 40 exercise -only interventions and 16 exercise -in addition to another treatment interventions. A total of 12 studies based the exercise intervention on a theory. T he most frequently used theorie s were the Transtheoretical Model ( n = 4) and the Exercise and Self -Esteem Model ( n = 4 ; two studies used both the Transtheoretical Model and the Exercise and Self Esteem Model ), followed by Social Cognitive Theory ( n = 2), Penders Health Promotion Lifest yle Profile ( n = 1), Roy A daptat ion M odel ( n = 1), Conservation of Resources Theory ( n = 1), and Self -Concept Theory ( n = 1). Additionally, 24
57 studies included an objective measure of exercise or fitness, 8 studies used a self -report measure, and 4 studies used both objective and self -report measures. The most common type of measures used were exercise logs/questionnai res/7 -day recall s ( n = 11) followed by r un/walk/treadmill test s (n = 10 ), and VO2 max/heart rate test s (n = 8 ). Most of the studies used an interactive exercise intervention format ( n = 42), followed by 11 studies that used both an interactive and didactic format, and only 4 studies used didactic intervention alone Twenty -eight studies had no treatment control groups, 19 studies had placebo c ontrol groups, 6 studies had multiple control groups, and only 4 studies had an activity control group. A total of 35 studies described the randomization process. Attrition was described in 31 studies, with a mean attrition of 22.69% ( SD = 16.84, range = 0 to 67%). Almost all studies ( n = 55) used a standardized measure of body image. Finally, the most common measure of body image was the Body Dissatisfaction subscale of the Eating Disorder Inventory ( n = 10 studies), followed by Body Cathexis Scale ( n = 9 studies), Physical Self -Perception Profile ( n = 6), Weight Concern subscale of the Body Esteem Scale ( n = 5), and the Physical Self subscale of the Tennessee Self Concept Scale ( n = 5) Publication Bias The overall mean random effect size ( ES) was 0.29 (SE = .04, CI = + .07; Q (97) = 206 18, p < .001, I2 = 52. 95; and the overall mean fixed ES was 0.23 (SE = .03, CI = .05 ; see figure 4 1 for forest plot ) Rosenthals Nfs was 2,871, indicating that 2,871 null studies (or 29 missing studies fo r every observe d study) would need to be located for the combined p value to exceed .05 In comparison, Orwin Nfs was 128, revealing that 128 studies would need to be located with a mean standard difference in means of 0 to bring the combined standard difference in means under 0.1.
58 Examination of the funnel plot (with the ES on the X axis and the standard error on the Y axis) revealed that the distribution of the ES showed a pattern suggestive of publication bias ( see figure 4 2 for funnel plot ). A statistical test of sy mmetry in the plot indicated significant asymmetry (intercept = 1.55, + = 0.71, t (96) = 4.36, p < .001; Egger et al., 1997). The Duval and Tweedies trim and fill procedure pointed at the possibility of some publication bias. More specifically, this method suggested 28 studies were missing and that the point estimate and 95% confidence interval using the Trim and Fill imputed point estimate was 0.12 ( + .08 ; see figure 4 3 for funnel plot with imputed values ). Moderator Analyses Participant features. Althou gh a larger effect size ( ES ) was evidenced for female populations ( M ES = 0.32) compared to male populations ( M ES = 0.19) the ES difference was not statistically significant, QB(1) = 1.11, p = .29 ( see Appendix I for random effects moderator analyses ). Me an age moderated the size of the effect, with larger ESs evidenced for older compared to younger participants ( z = 2.07, p = .038). A significant difference was found for age category, QB(3) = 8.87, p = .031. Examination of the M ES scores indicated that t he largest effects were evidenced for adults ( M ES = 0 44) compared to older adults ( M ES = 0 33) followed by university students ( M ES = 0.23). Additionally, because of the small number of studies examining elementary ( n = 3 studies representing 7 ES ), m iddle ( n = 5 studies representing 8 ES ), and high school ( n = 2 studies representing 2 ES ), these groups were combined for the age category comparison resulting in a combined high school/middle/elementary school students group ( M ES = 0.16). A significant effect was found for ethnicity, QB(1) = 10.55, p = .001, with largest ES evidenced for nonCaucasians (Hispanic & African American; M ES = 0.69) followed by Caucasians ( M ES = 0.20). The percentage of Caucasian participants did not moderate the size of
59 the effect, z = 0.07, p = .94 A nonsignificant effect was found for risk status, QB(1) = 0.26, p = .61 with larger ESs evidenced for universal ( M ES = 0.29) compared to selected participants ( M ES = 0.23). I examined the moderating effect of psychological r isk using only universal and selected categories because of the low n in the breast cancer and psychological disorders categories. The psychological disorder category included schizophrenia, bipolar disorder, and depression but the study specifically excluded patients with previous eating disorders (Evans, Newton, & Higgins, 2005). For body composition, a nonsignificant larger effect was found for overweight/obese participants ( M ES = 0.34) compared to normal weight participants ( M ES = 0.18), QB(1) = 3.2 5 p = .07. Finally, for fitness level at preintervention, a moderator analysis was not feasible since most of the ES were for sedentary/low activity. However, a comparable effect was found for sedentary/low activity ( M ES = 0.42) and sedentary and active ( M ES = 0.41). Design features. A negative relationship was found between publication year and the size of the effect ( z = 1.49, p = .14; i.e., recency effect). No significant ES difference was found for published ( M ES = 0.29) versus unpublished studies ( M ES = .32), QB(1) = 0.10, p = .75. I did not examine the moderating effects of country of origin because most of the effect sizes were from interventions conducting in the United States (68%; see table 1 in Appendix F for the individual countries mean ES ). The type of control group (i.e., no treatment placebo or active control) did not moderate the size of the effect, QB(2 ) = 4.24, p = 12. A nonsignificant larger effect was found for no treatment controls ( M ES = 0.37), followed by placebo control ( M ES = 0.23), and active control ( M ES = 0.14) A significantly larger effect was found for exercise based interventions (M ES = 0.38) compared to lecture based ( M ES = 0.13) or combined lecture and exercise based interventions ( M ES = 0.12) QB(2) = 1 1.88, p = .00 3 Type of recruitment method moderated the size of the effect, QB(1) = 3.91, p = .05, with a larger effect found for self-
60 selected recruitment methods ( M ES = 0.38) compared to nonself -selected ( M ES = 0.22). A nonsignificant effect was evidenced f or studies with randomly assigned participants ( M ES = 0.26) compared to nonrandomly ( M ES = 0.34), QB(1) = 0.93, p = .34. Although validation of the body image measure did not moderate the size of the effect QB(1) = 0.23, p = .64, I found a larger effect size for standardized measures ( M ES = 0.30) compared to unstandardized measures (M ES = 0.20). Finally, no significant difference in effect size, QB(2) = 4.02, p = .13, was found for self-report measures of fitness ( M ES = 0.23), objective measures of fit ness ( M ES = 0.39), or the use of both self report and objective measures ( M ES = 0.17). Exercise intervention features. No significant difference in the size of the effect was found for exercise interventions that met exercise guidelines (M ES = 0.30) ver sus interventions that did not meet the guidelines (M ES = 0.32) QB(1) = 0.23, p = .63. E xercise specificity was significant, with larger effects for exercise -only interventions (M ES = 0.37) compared to exercise -in addition to another treatment ( M ES = 0 .14) QB(1) = 9.23, p = .002. A nonsignificant effect was evidenced for experimental intervention type, QB( 2) = .55, p = .76 (individual based M ES = 0.34, group-based M ES = 0.26, combination of individual and groupbased M ES = 0.32). A nonsignificant e ffect was found for exercise intensity, QB(2) = 0.637, p = 0.73, with larger effects evidenced for light intensity ( M ES = 0.38), followed by moderate intensity ( M ES = 0.36), and strenuous ( M ES = 0.29). With regard to the dose response for exercise, exercise duration ( M duration = 49.09 min, range = 20 to 75 min; z = 25, p = 80), length of intervention in weeks ( M length = 12.69, range = 4 to 52; z = 0.11, p = .92 ), and follow up in weeks ( M length = 21.85, range = 2 to 72; z = 0.24, p = .81) did not moderate the size of the effect. Mode of exercise (aerobic M ES = 0.29, resistance M ES = 0.38, or both aerobic and resistance M ES = 0.27) did not moderate the size of
61 the effect, QB(2 ) = 1.37, p = 50. Frequency per week of exercise ( M frequency = 2.81, range = 1 to 5), however, did moderate the size of the effect, with larger effects evidenced for interventions of higher frequency per week ( z = 2.42, p = .016). No difference in the size of the effect was evidenced for experimental group participants wh o improved on either fitness or body composition from pre to post intervention compared to those who did not improve on either fitness (QB(1) = 0.004, p = .95) or body composition ( QB(1) = 0.12, p = .73) from pre to post intervention. I was not able to ana lyze the moderating effects of control group fitness or body composition due to a lack of studies reporting this information. However, a larger effect was evidenced for control group participants who did not improve on fitness from pre to post intervention ( M ES = 0.40) versus participants who did improve ( M ES = 0.15). Furthermore, control group participants had larger effects for no improvement on body composition ( M ES = 0 .42) versus improvement ( M ES = 0 .10) from pre to post intervention. Finally, no di fference in the size of the effect was found for use of a theory in either measure selection ( QB(1) = 0.95, p = .33) or exercise intervention ( QB(1) = 3.30, p = .07).
62 Figure 4 1. Forest plot
63 -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 0.0 0.2 0.4 0.6 0.8 Standard ErrorStd diff in meansFunnel Plot of Standard Error by Std diff in means Figure 4 2 Funnel p lot -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 0.0 0.2 0.4 0.6 0.8 Standard ErrorStd diff in meansFunnel Plot of Standard Error by Std diff in means Figure 4 3 Funnel plot with impu ted values
64 CHAPTER 5 DISCUSSION The main purpose of my review was to provide a statistical summary of exercise intervention programs and their effects on body image. A small overall mean effect size was found indicating that exercise interventions result ed in improvements in body image compared to a control group. Thus, my results revealed that exercise interventions are effective in reducing body image concerns, and may be considered as an alternative efficacious treatment for body image concerns along w ith the typical therapies (e.g., psychoeducational, cognitive behavioral). Given the heterogeneity in the effe cts from these interventions, I examined participant, design, and exercise intervention features that may account for the heterogeneity. Below I d iscuss the findings with regard to the publication bias, moderators, the health implications, study limitations, and directions for future research. Publication Bias Because not all studies, especially those with nonsignificant findings or small treatment effects, are published this can lead to a bias in the studies included in the meta analysis. To avoid overestimating the true size of the treatment effect and subsequent inferences, it is necessary to assess the likelihood of publication bias (Borenstein, Hedges, Higgins, & Rothstein, 2007). Thus, I examined the following methods to examine publication bias: funnel plot, classic fail -safe N, Orwin fail -safe N Eggers test of intercept, as well as Duval and Tweedies trim and fill. Below I discuss each of t hese publication bias findings and the implications of these finding on result interpretation. Examination of the funnel plot distribution was suggestive of publication bias. More specifically, the bottom of the funnel plot showed a higher concentration of studies to the right of the mean indicati ng that smaller studies were more likely to be published if they had larger than
65 average effects (Borenstein et al., 2007) For this reason, both the Rosenthals classic and Orwins fail -safe N were calculated to f urther determine if the treatment effect was an artifact of bias. The large number of unpublished null trials (Rosenthals ( Nfs = 2, 871) and Orwins ( Nfs = 128) lead to the conclusion that the findings were unlikely to be biased by the file drawer problem. The reason for the disparity in Rosenthals and Orwins fail -safe N is that Rosenthals N assumes that the effect size of the st udies that would be added is zero while Orwins N allows specification of a value for the effects of the studies added. Because the interpretation of a funnel plot is largely subjective, I also calculated Eggers linear regression method to quantify or test the relationship between sample size and effect size (Egger et al., 1997). Eggers test of intercept indicated significant as ymmetry ( p < .001) confirming the funnel plot interpretation of the presence of bias. Thus, the body of evidence suggests that smaller studies reported a larger association than did the larger studies While Eggers test of intercept provides a more powerf ul test of asymmetry than the funnel plot the statistical power of the test is also limited by the number of studies included in the meta analysis. T he Duval and Tweedies trim and fill procedure also indicated the possibility of some publication bias Si nce there are more small studies on the right side of the funnel plot this leads to the assumption that there are studies missing on the left side of the funnel plot (Borenstein et al., 2007). Trim and fill suggested that a total of 28 studies are missing that would need to be imputed in order to yield an unbiased estimate of the ES (0.12). It should be noted, however, that CMA 2 states that the present release of trim and fill can occasionally produce an incorrect result which is a potential confound when interpreting its results ( BioStat, Englewood, New Jersey). In summary, although these tests provide some evidence of potential publication bias in the data, there are other sources of asymmetry in funnels plots. Table 5 1 shows the potential sources of
66 asy mmetry in funnel plots (Khoshdel et al., 2006). Among those is true heterogeneity in the effects (e.g., gender, age). Thus, further analyses investigating the sources of heterogeneity via moderator analyses was warranted. Moderator Analyses Participant f ea tures Participant gender. First, in contrast to my hypothesis I did not find that gender moderated the size of the effect. A lthough I did not find significant gender differences, it is important to note that larger effects were evidenced for female ( M ES = 0.3 2 ) compared to male (M ES = 0.19) populations. The nonsignificant findings may be due to the large confidence intervals and the small number of studies exclusively examining male populations. I t is not surprising however, that most studies included in my revie w focused on female populations because women are more likely than men to report body image concerns, and thus most in need of intervention ( Altabe & Thompson, 1993; Elgin & Pritchard, 2006; Feingold & Mazzella, 1998). With the rise in male body i mage concerns, and the pressure for men to achieve a fit physique, further research is needed examining the gender effects of exercise interventions on body image ; in particular with a focus on resistance exercise because men tend to want to increase mus cle mass to achieve their ideal physique (McCabe & Ricciardelli, 2004) Participant age. Consistent with my hypothesis age moderated the size of the effect (using both a categorical and continuous moderator analysis). For the categorical analysis, I found that intervention effects increased with age until adulthood, and then remained consistent into middle and late adulthood. Comparing the combined youth (elementary, middle, and high school), with university, adult, and older adults provides evidence of la rger effects for older compared to younger populations. Second, to corroborate this evidence, m ean age moderated the size of the effect (M = 30.04, SD = 15.35; p < .001) indicating that the interventions had a
67 greater impact on older compared to younger participants This age effect may be due to adults reporting higher body image concerns than younger populations (Striegel -Moore & Franko, 2002) Thus, it is not surprising that most of the studies focused on college and adult aged participants. However, wi th the increased interest and rise in body image concerns at young ages, further research is needed focusing on preadolescent populations ( McCabe & Ricciardelli, 2004; Smolak, 2002). For example, f urther research is needed examining the effects of exercise on body image in youth considering the high percent that report body image concerns coupled with the increase in obesity. Participant ethnicity. Evidence was found for the hypothesis that ethnicity would moderate the size of the effect however, not in t he direction predicted. While there were only a few studies that looked at nonCaucasian populations (African American or Hispanic; N = 3; 10%), this group had a significantly larger effect than interventions targeting Caucasian populations. I found no support for the hypothesis that Caucasians or having a larger percent Caucasians would have a larger effect. E thnicity (based on the percentage of Caucasians in a study) did not moderate the size of the effect Furthe r research is needed examining and reporti ng ethnic differences to further examine its moderating effect. The low number of studies examining ethnic minorities calls for further research with ethnic minorities. More specifically, only two studies exclusively examined African Americans and one stud y examined Hispanics. Interestingly, a previous meta analyses of body image and ethnicity, based on 17,000 participants, found that Asian women reported significantly more eating disturbances/body dissatisfaction than Caucasian women (Wildes, Emery, & Simo ns, 2001). Furthermore, it was concluded that ethnicity played a role in the influence and development of eating disturbances however the lack of studies examining nonCaucasian populations (i.e., African American, Asian,
68 Hispanic, Arab, Native American) ma de it impossible to determine how and why these differences exist. Clearly, more research is needed examining minorities as well as research directly comparing Caucasians to ethnic minorities. Future studies should include this descriptive information for better examination of its moderating effect. Psychological risk status of participants. In contrast to my hypothesis, I found larger intervention effects for universal programs ( M ES = 0.29) versus selected programs (e.g., eating disordered, high body dis satisfied; M ES = 0.23). In contrast to this, research has shown greater intervention effects for selected program participants due to the floor effects for low risk participants and increased likelihood for selected participants to engage more fully in the intervention (Stice & Shaw, 2004). It is possible that while selected participants may be more likely to benefit from interventions they may also be more at risk of harm from the interventions than universal participants (Trikalinos & Ioannidis, 2005). F urther research exploring universal versus selected programs should be done to better identify programs that are better suited to provide maximal benefits for these populations. Although there were not enough ES to run a moderator analysis for psychologic ally disordered participants or breast cancer patients a preliminary evaluation of the effect of exercise on body image of these populations may prove useful. Of importance, exercise training is associated with increased body satisfaction among women recov ering from breast cancer, men and women with paraplegia and quadriplegia, adolescents with postural deformities, and women classified as obese. These findings speak to the robustness of exercise for improving body image within a variety of populations (Mar tin & Lich t enberger, 2002). Participant body composition. I did not find support for my hypothesis that larger intervention effects would be evidenced for overweight/obese populations than normal weight
69 populations, however it was approaching significance (p = .07). For body composition, nonsignificant larger effects were found for participants who were overweight/obese ( M ES = 0.34) at the beginning of the intervention compared to normal weight participants ( M ES = 0.18) This may be because overweight/obe se individuals are at a higher risk for body image disturbances compared to normal weight individuals, and therefore more likely to gain benefits from an exercise intervention. Further research is still needed examining the moderating effect of weight status, as well as changes in body composition, in exercise interventions. Preintervention fitness level. There was no support for the hypothesis that low activity groups would evidence greater effects than high activity groups. Only 28 of the studies reported preintervention fitness level, and 27 of these studies had participants of low PA levels. Because no studies reported moderate activity or highly active participants exclusively it is difficult to ascertain whether sedentary participants would have produc ed larger effects. Although moderator analysis was not possible i t is commendable that most studies had low activity groups because this group is most in need of and most likely to benefit from interventions compared to high activity groups. Future studies should continue to report th e preintervention fitness level and further investigate the influence of sedentary versus active participants on the size of the effect. Design Features Control group. In contrast to my hypothesis, t he type of control group di d not moderate the size of the effect Although the hypothesis was not supported the effect was in the predicted direction, no -treatment controls had a larger nonsignificant effect size than placebo controls. Similar to these findings, Puetz et al. (2006) found that placebo controls were associated with smaller effect sizes compared to other study designs. Furthermore, they suggest that future studies more clearly report control group procedures to allow for better replication and interpretation of the resu lts. With so much variation in control conditions from study to study and
70 a need to identify how control conditions effect body image, f urther research is needed examining the ideal type of control group for comparison purposes. Intervention format. I n support of my hypothesis, the intervention format significantly moderated the size of the effect. I nteractive programs were found to be more effect ive than didactic programs or a combination of both interactive and didactic, for improving body image This is in support of previous research which found interactive programs more beneficial for eating disorder preventions (Stice & Shaw, 2004; Tobler et al., 2000). Stice and Shaw (2004) propose that interactive programs are more effective because participants m ore readily engage in the program facilitating skill acquisition and attitudinal and behavioral change. Future researchers should consider designing exercise -based interventions to further explore the benefits of interactive programs. Recruitment method. In support of my hypothesis, significant larger effects were found for self-selecting volunteers than for participants recruited through population based recruitment efforts. This lends further support to the notion that participants who are self -selected are more motivated and engaged more actively in the intervention and will therefore gain more benefits (Stice et al., 2006 ). It is likely that these self -selecting participants represent the preparation or action stage of change, and are thus intending and motivating to make changes (Prochaska et al, 1994). Participants in the precontemplation and contemplation stage, however, are more likely in need of intervention. Thus, future researchers are encouraged to examine which stage of change their participant s are in, and target individuals in the preaction stages. Random assignment. Contrary to my hypothesis, intervention effects were not found to be larger for interventions that used random assignment relative to other approaches. However, as hypothesized t his effect did not reach statistical significance similar to previous health
71 interventions (Stice & Shaw, 2004) Previous research has shown that nonrandomized trials are more likely to both under and over estimate the size of the effect (Conn & Rantz, 2003). Future research should continue to use randomized trials as they are useful in ruling out possible confounds, as well as allow for more precise inferences with regard to intervention effects (Stice & Shaw, 2004) Publication status. In contrast to the hypothesis and the extant literature, I found that publication status (i.e., unpublished versus published) was not related to the intervention effect size ( Hopewell et al., 2007). However, of interest unpublished studies produced nonsignificant larger effe cts than published studies My finding provides support for publication status being an inadequate indication for study quality, in part because authors often do not submit studies unless they have statistically significant treatment effects (Conn & Rantz, 2003); illustrating the need to examine publication bias via multiple methods (e.g., funnel plot, fail safe n ) to obtain a more accurate picture of publication bias in the extant literature. Validated body image measure. The hypothesis that interventions using validated outcome measures would observe larger intervention effects than interventions that used unstandardized measures was not supported. A nonsignificant larger effect was found for standardized measures, which is consistent with the hypothesis, however this may be due to the large confidence i nterval and minimal number of studies that used unvalidated measures. I t is commendable that most researchers employed a standardized measure (95%) and future studies should continue to do so. It was my ori ginal intent to examine body image categories. However, most of the ES represented the attitudinal component of body image (i.e., body dissatisfaction), which is the most common type of body image examined (Thompson &Van Den Berg, 2002). Since body image i s a multidimensional construct encompassing perceptual, affective, cognitive,
72 sociocultural, and behavioral components future researchers are encouraged to examine the effects of exercise on body image using a multidimensional body image approach to gain a more comprehensive view of body image Exercise Intervention Features Exercise Dose. I found no support for my hypothesis that exercise interventions that met the ACSM physical activity guidelines would result in larger effects than interventions th at did not meet the guidelines. It should be noted that most studies ( 70%) did not meet the ACSM physical activity guidelines, and future interventions should strive to meet these guidelines to more clearly determine a standard for future exercise interventions I also found no support for my hypothesis that the intervention effect would be stronger for p revention programs with a long versus short length (in weeks). Furthermore, a lthough the exercise dose response is well established for the physical health bene fits of physical activity (ACSM, 2000) I did not find a moderating effect for exercise duration, intensity, or length. I did, however, find that exercise frequency moderated the size of the effect, with greater exercise frequency per week resulting in larg er effect sizes. Additionally I did not find significant support for the hypothesis that larger effects would be evidenced for resistance training based interventions than aerobic -based interventions It should be noted, however, that larger effects were evidenced in the predicted direction for resistance interventions ( M ES = 0.38) versus aerobic interventions ( M ES = 0.29). Finally, no moderating effect was found for follow ups, most likely because only a few studies conducted follow ups (18%), so it is uncertain whether fitness interventions pr oduce lasting body image change F urther research into the issue of the dose response is clearly important and much needed, and should be systematically examined in future intervention studies. Physical fitness. E xamination of the hypothesis that programs that increase physical fitness have larger intervention effects than those that do not increase activity revealed that the
73 fitness level improvements moderate d the size of the effect for the control group, but not for the experimental group. That is, for the experimental groups the size of the effect did not differ for participants whose fitness level improved from pre to post intervention versus participants whose fitness level did not improv e from pre to post int ervention. In the control group, the size of the effect for participants whose fitness level did not improve from pre to post was larger than for those that did improve. However, this may be due to the small number of studies that found improvements in fit ness of the control groups. Intervention specificity. An e xamination of the moderating effect of intervention specificity on the size of the effect found that exercise specificity moderated the size of the effect, with larger effects evidenced for specifi c compar ed to nonspecific interventions On the other hand, Puetz et al. (2006) found that specific interventions yielded smaller effects than other interventions, but this also may have been due to type of control group used. Clearly it is not known wheth er specificity independently moderates the size of the effect. Because little is known about how or why exercise specificity moderates this effect future research should focus on further exploring the exercise interventions both alone and in conjunction wi th other therapies. Theory. My hypothesis that body image assessments that are selected based on a theoretical model will produce larger effects than nontheoretical assessments was not supported It should be noted that although the results were nonsignif icant, studies that used theory for measure selection produced larger effects than those that did not. Similarly, my hypothesis that exercise interventions that are developed based on theoretical models will produce larger effects than nontheoretical exerc ise interventions was not supported, however approaching significance (p = .07). These nonsignificant findings may be due to a lack of studies employing theories for
74 both measure selection and exercise intervention development. Future studies should consi der the use of theoretical models in both intervention design and measure selection, as they tend to produce larger effects (Sallis, 2001). Limitations Limitations of the extant literature include the lack of studies examining male populations as well as nonCaucasian populations. Without further research examining these two subgroups it is not possible to generalize the findings of this study to such populations. Furthermore, many studies did not adequately prescribe, monitor, and control the frequency, du ration, and intensity of the exercise to ensure fitness gains. Despite this limitation it is still possible to examine the psychological effects of participation in an exercise program independent of changes in fitness. However, it does not permit the exam ination of the psychological effects of increased fitness. Similarly, researchers have not adequately described the physiological measures that were used to assess change in fitness and often assumed that participation in an exercise program was synonymous with increased fitness. Finally, while many studies demonstrated the effects of exercise on body image over limited periods of time, very few studies have been designed to examine the long term intervention phase of a study in a nonclinical population. The question as to whether the effects of exercise are transitory (i.e., only lasting as long as the intervention) or long term (i.e., continuing after cessation of the intervention) remains virtually unexplored. Limitations specific to my meta analysis incl ude search limitations. That is, although an attempt was made to exhaustively ident ify eligible studies through the search strategies, limitations do exist. For example, the computer -based searches were limited to MEDLINE. Because none of the electronic da tabases include all published studies, searching multiple databases is recommend ed For example, searching EMBASE can add up to 30% more references, mainly from European journals that are not indexed on MEDLINE (Khoshdel et al.,
75 2006). Of note, omission of EMBASE electronic database references does not appear to bias the results of meta analysis, but only reduces precision (Suarez et al., 2000). Further limitations of this meta analysis are similar to criticisms of meta analyses in general (Borenstein et al ., 2007) For example, t he file drawer problem can occur if the collection of studies is based on a biased sample of studies that will be reflected in the overall effect size Several methods exist to assess and determine the extent of bias and meta analys is allows us to quantify this bias. Another criticism of meta analyses is the possibility of combining studies with different characteristics that can overlook important differences between the studies. However, meta analysis allows for the synthesi s of va rying studies through exc lusion criteria to determine the similarity of included studies. Meta analysis also allows for generalizability of the findings based on the studies included. Lastly, critics claim that meta analyses are inevitably performed poorly because of the complexity of meta analysis, which leads to mistakes. As Bore nstein et al. (2007) states primary studies are performed poorly and have flaws but these flaws are in the application of the method, rather than problems with the method itself. Researchers should consider these flaws and their impact on the analysis in order to prevent them in subsequent meta analyses. Future Directions In addition to the future directions discussed in the individual moderator analyses above, o ne important nex t step toward science -based practice is the need to better understand which biological, psychological, and social aspects of chronic exercise contribute to improved body image. For example, exercise interventions often involve substantial social interaction. Thus, if the effect of chronic exercise on body image is to be understood, it is critical to investigate or control for variables that are independent of exercise itself, such as social interaction (Puetz et al., 2006). If the factors that derive body i mage change can be identified, then exercise
76 interventions can be designed to be more effective by targeting these mechanisms. It is most likely that objective physical characteristic changes (e.g., weight loss) are just one set of variables that can influ ence body image. Indeed, Cashs (2002) cognitive behavioral model of body image identifies physical characteristics as one of four developmental influences on body image (in addition to cultural socialization, interpersonal experiences, and personality var iables). Of relevance to understanding the effects of exercise, Cash noted that changes in body weight, muscularity, and physical competence can all influence body image. Another set of variables that might explain the effects of exercise on body image ar e changes in peoples perceptions of their physical characteristics. For example, strength training can make exercisers feel stronger, thinner, and more toned. These perceived physical changes may elicit improvements in body image independent of objective physical changes (Martin & Lich t enberger, 2002). In support of this, Martin Ginis et al. (2005) found that body image improvements were related to subjective physical changes in participants of a 12 -week strength training program. In short, although there is sound evidence that exercise produces positive changes in well -being through improved physical self -perceptions, the question still remains as to the main mechanisms underpinning such change. For the fine -tuning of intervention design, it is important not only for mechanisms to be determined but also for the conditions under which they optimally function to be identified. C onclusion In summary, as Borenstein et al. (2007) stated, the goal of meta analysis is to broaden the base of studies in some way, expand the question, and study the pattern of answers. As such, this meta analysis provides further evidence that exercise represents an innovative, practical, and widely disseminable intervention for negative body image. Although small effect sizes for the effects of exercise on body image were found it is important to emphasize the advantages of
77 exercise over other types of therapy, such as cognitive -behavioral therapy. For example, exercise has the ability to reach and benefit large audiences. Other p ractical advantages of exercise are that compared to other interventions, exercise has a relatively low cost, negligible negative side effects, and is a socially acceptable behavior; which may result in greater treatment acceptance. Finally, exercise is se lf -sustaining because it can be maintained once the basic skills are learnt. Further research is needed comparing exercise to other body image interventions to determine its effectiveness in randomized controlled trials. Table 5 1. Khoshdels potential so ur ces of asymmetry in funnel plots Selection bias Publication and reporting bias Biased inclusion criteria True heterogeneity: size of effect differs according to study size Intensity of intervention Differences in underlying risk Data irregularities Poor methodological design of small studies Inadequate analysis Fraud Artefact: due to poor choice of effect measure Chance Reference: Khoshdel, A., Attia, J., & Carney, S. L. (2006). Basic concepts in meta analysis: A primer for clinicians. International Journal of Clinical Practice, 60 1287 1294.
78 APPENDIX A LIST OF JOURNALS MAN UALLY SEARCHED List of Journals Manually Searched : Body Image: An International Journal Eating and Weight Disorders Eating Disorders: A Journal of Treatment and Prevention Health Psychology International Journal of Eating Disorders International Journal of Sport and Exercise Psychology Journal of Applied Sport Psychology Journal of Sport & Exercise Psychology Perceptual and Motor Skills Psychology of Sport and Exercise
79 APPENDIX B LETTER TO ACTIVE RES EARCHERS Dear Colleague The Exercise Psychology lab here a t the University of Florida is conducting a meta analysis on the relationship between exercise interventions and body image. We are looking for data from experimental studies which are not readily available through search engines or the UF school libraries (e.g. dissertations, masters theses, unpublished data, poster presentations, in review articles, etc.). We would appreciate any relevant data to add to our meta a nalysis. Please only include findings from experimental studies with both an exercise and control group. We will be more than happy to cite your research in the paper if it is used. Studies or relevant data can be mailed to the postal address below, sent a s a reply to this email, or emailed to Dr. Heather Hausenblas at email@example.com Exercise Psychology Lab University of Florida P.O. Box 118205 Gainesville, FL 326118205 We have attached a table detailing the information needed (e.g., population, type of intervention, sample size, etc.). If all the given information is available in the research you send it is not necessary to fill out the table. Any information you can provide us with would be greatly appr eciated. If there are any question or concerns regarding the study please feel free to contact us. Thank you. Sincerely, Dr. Heather Hausenblas, Associate Professor, Director of Exercise Psychology Laboratory, University of Florida Anna Campbell, M.S. S tudent, Sports and Exercise Psychology, University of Florida
80 APPENDIX C TABLE TO ACTIVE RESE ARCHERS Author Name(s) and Country Publication Year Type of Study (published article, thesis, dissertation, unpublished data, etc.) Sample Size (include se x of population or # of males/females) Sample Population (university students, children, eating disordered, etc.) Participant Ethnicity (and % if available) Participant Age (mean and range) Participant Body Composition (normal, overweight, obese) a nd Activity Level Pre -intervention (sedentary, active) Exercise Intervention Frequency (per week), Duration (in minutes), and Length (in weeks) Type of Exercise Performed (aerobic, resistance, both, etc.) Intervention Type (individual/home based, group based, etc.) Theory Used for Exercise Intervention Type of Control (no treatment, placebo control, activity control, etc.) Body Image Measure(s) Results (e.g., M, SD for EXP and CONTROL group, F value, sign level, etc.)
81 APPENDIX D CODING SHEET GENERAL CHARACTERISTICS 1) Study #____________ 2) Authors Names: ____________________________________________________________________ 3) Year of Publication: ____________ 4) Country of Origin: 1 = United States 2 = Canada 3 = Australia 4 = United Kingdom PARTICIPANT FEATURES 5) Age: Mean= ____ Group: 1 = Elementary, 2 = Middle, 3 = High, 4 = Univer, 5 = Adults, 6 = Older Adults 6) Sex: 1 = Female 2 = Male 3 = Both 7) Ethnicity: A) Caucasian ________% (if given) B) Dom inant Ethnic Group: 1 =Caucasian, 2 = African American, 3 = Hispanic, 4 = Other 8) Risk Status: 1 = Universal 2 = Selected [ High Risk Population (e.g. eating disordered)] 9) Body Composition: 1 = Normal weight 2 = Overweight/Obe se 10) Fitness Level at Preintervention: 1 = Sedentary/Low Active 2 = Active EXERCISE INTERVENTION FEATURES 11) Type: 1 = Aerobic 2 = Resistance Training 3 = Resistance and Aerobic 12) Duration (in minutes): ____________ 13) Length of intervention (in weeks): ____________ 14) Frequency: ____________ 15) Intensity of exercise performed: 1 = Light 2 = Moderate 3 = Hard/Very Hard 16) Length of follow up (in weeks): ____________ 17) Fitness Exp: 1 = Improve 2 = No Improv e 18) Body Composition Exp: 1 = Improve 2 = No Improve 19) Fitness Control: 1 = Improve 2 = No Improve 20) Body Composition Control: 1 = Improve 2 = No Improve 21) Specificity: 1 = Exercise Only 2 = Exercise in Addition to Another Treatment 22) Experimental Intervention Type: 1 = Individual based 2 = Groupbased 3 = Both 23) Meet Exercise Guidelines: 1 = Yes 2 = No 24) Theory Used for Exercise Intervention: 1 = Yes, 2 = No Describe ____________________________ Theory Used for Measure Selection: 1 = Yes, 2 = No Describe ____________________________
82 DESIGN AND STUDY FEATURES 25) Control Group: 1 = No treatment (usual care); 2 = Placebo control (Health Ed. class); 3 = Activity control 26) Recruitment Method: 1 = Self selected 2 = Not self selected 27) Publication Status: 1 = Published 2 = Unpublished 28) Attrition Rate: ________% Dropout 29) Random Assignment: 1 = Yes 2 = No 30) Exercise Intervention: 1 = Exercise based 2 = L ecture based 3 = Both MEASURE FEATURES 31) Measure of exercise/fitness: 1 = Self Report; 2 = Objective Describe_________________________ 32) Body Image Outcome Measure: ______________________________________________________________________________________ ______________________________________________________________________________________ 33) Standardized Measure with established validity and reliability: 1 = Yes 2 = No 34) Number of Body Image Measures: ___________ 35) Correction Factor Needed: 1. Yes; 2. No ( If higher scores = less pathology, correction factor is needed) EFFECT SIZE INFORMATION: Continuous (Ms) Unmatched Groups, Pre and Post Data (select 1) A. Means, SD pre and post, N, in each group, pre/post corr Exercise Group: Pr e M_____ Pre SD______ Post M ______ Post SD ______ N _____ Control Group: Pre M_____ Pre SD______ Post M ______ Post SD ______ N _____ Pre Post correlation: _________ B. Means, SD difference, N, in each group, pre/post corr Exercise Group: Pre M_____ Post M ______ Difference SD _____ N _____ Control Group: Pre M_____ Post M ______ Difference SD ______ N _____ Pre Post correlation: _________ C. Means pre and post in each group, t within groups, N Exercise Group: Pre M_____ Post M ______ Paired t for change _____ N _____ Control Group: Pre M_____ Post M ______ Paired t for change ______ N _____ Pre Post correlation: _________ D. Means pre and post in each group, p within groups, N Exercise Group: Pre M_____ Post M ______ Paired p for change ______ N _____ Control Group: Pre M_____ Post M ______ Paired p for change ______ N _____ Tails for p value: _______ Pre Post correlation: _________ E. Means pre and post in each group, F for differences between changes, N Exercise Group: Pre M_____ Post M ______ N _____ Control Group: Pre M_____ Post M ______ N _____ F for difference: _______ Pre Post correlation: _________ F. Mean change, SD pre and post, N, in each group, Pre/post corr
83 Exercise Group: M Change_____ Pre SD ______ Post SD ______ N _____ Control Group: M Change_____ Pre SD ______ Post SD ______ N _____ Pre Post correlation: _________ G. Mean change, SD difference, N, in each group, pre/post corr Exercise Group: M Change _____ Difference SD ______ N _____ Control Group: M Change _____ Difference SD ______ N _____ Pre Post correlation: _________ H. Mean change in each group, t within groups, N Exercise Group: M Change_____ Paired t for change ______ N _____ Control Group: M Change_____ Paired t for change ______ N _____ Pre Post correlation: _________ I. Mean change in each group, p within groups, N Exercise Group: M Change_____ Paired p for change ______ N _____ Control Group: M Change_____ Paired p for change ______ N _____ Tails for p values______ Pre Post correlation: _________ J. Mean change in each group, F for difference between changes, N Exercise Group: M Change_____ N _____ Control Group: M Change_____ N _____ F for difference: ________ Pre Post correlation: _________ K. F for difference between changes, N Exercise Group: N _____ Control Group: N _____ F for difference: ________ Pre Post correlation: _________ NOTE:
84 APPENDIX E MODERATOR TABLE Table E 1. Moderator table Moderator Value Coding Description and Criteria Participant Features Age Mean Range Continuous Continuous Mea n age of participants at baseline. Number of values in the age range (e.g., 18 22 = 5). When age range was not reported.grade range was substituted (+2 because most grade levels include students from 2 age levels). Gender 1 = Female 2 = Male 3 = Both = 3 Categorical variable representing whether intervention assessed females only, males only, or both genders. Ethnicity % Caucasian Dominant Ethnic Group % Caucasian 1 = Caucasian 2 = African American 3 = Hispanic, 4 = Ot her Percentage of participants from the entire sample (at baseline) who were Caucasian because Caucasians are at highest risk for bodyimage disturbance. Ethnic group representing most of the sample. Risk Status 1 = Universal, 2 = Selected Categorical variable representing whether the study was universally implemented or whether study participants were selected because they are a group at increased risk for bodyimage disturbance (e.g., eating disordered, overweight) Body Composition 1 = Normal Weight, 2 = Overweight/obese Categorical variable representing whether the population was overweight/obese or normal weight using the classification of overweight and obesity by BMI. Because there were so few studies that examined underweight populations, we com bined normal and underweight populations. Similarly, because of the low number of obese populations, we combined overweight and obese populations. Fitness Level Preintervention 1 = Sedentary, 2 = Low Active; 3 = Active Categorical variable representin g the fitness level of participants. Sedentary represents no physical activity; low active represents exercising less than the ACSM guidelines; active represents meeting the ACSM guidelines.
85 Table E 1. Continued Moderator Value Coding Description and Cri teria Exercise Intervention Features Type 1= Aerobic, 2 = Resistance Training, 3 = Both Categorical variable representing whether the exercise intervention consisted of aerobic, resistance training, or both. Duration Continuous Number of minutes per session excluding warmup and cooldown. Length Continuous Length of exercise intervention in weeks. Frequency Continuous Number of sessions per week. Intensity 1 = Light, 2 = Moderate, 3 = Hard/Very Hard Categorical variable representing the intensity of exercise based on the ACSM (2000, p. 150) classification table. Physical Fitness 1 = Improvement, 2 = No Improvement, 3 = Not Reported Categorical variable representing whether the intervention group showed improvement from pre to post intervention on physical fitness. Specificity 1 = Exercise Only, 2 = Exercise in Addition to Another Treatment Categorical variable representing whether the intervention was an exercise only group, or exercise in addition to another treatment (cognitive -behavioral theory, weight management). Meet Guidelines 1= Yes; 2 = No Categorical variable representing whether the exercise intervention met the ACSM (2000) guidelines of exercising 3 5 times a week, at a moderate intensity level for 30 minutes per session. Design a nd Study Features Control Group 1 = No treatment, 2 = Placebo control, 3 = Activity control The literature allowed for a meaningful statistical c omparison when control condi tion variable was coded into 2 categories: notreatment (e.g., usual care) place bo controls (e.g., cognitively oriented or health education classes), and activity controls (e.g., yoga) Recruitment Method 1 = Self selected, 2 = Not self -selected Categorical variable representing whether participants were recruited through a populati on based strategy (e.g., at particular school) or self selected in response to broader recruitment efforts (e.g., media advertisement). Publication Status 1 = Published, 2 = Unpublished Categorical variable representing whether report was published (peer reviewed journal article) or unpublished (e.g., masters thesis, dissertation). If 2 separate reports were used to code a single study (e.g., dissertation and published report in scientific journal), we coded the type of the more formally published report (i.e., journal article).
86 APPENDIX F EFFECT SIZE INFORMAT ION Table F 1. Categorical random effect size (ES) i nformation M ean ES S tandard E rror 95% CI N umber of ES p value PARTICIPANT Gender Male Female Both 0.189 0.3 00 0.281 0 .09 2 0 .05 6 0.070 0.180 0.100 0.138 12 55 27 0.039 Age Elementary Middle High school University Adults Older Adults Combined (elementary, middle, and high school) 0.006 0.239 0.478 0.225 0.442 0.329 0.160 0.072 0.168 0.283 0.066 0.064 0.100 0.088 0.141 0.330 0.555 0.130 0.126 0.197 0.172 7 8 2 35 26 12 17 0.929 0 .15 5 0 .09 1 0 .001 0 .001 0.068 Ethnicity Caucasian NonCaucasian 0.200 0.692 0.063 0.138 0.124 0.270 26 3 0.001 Psychological Risk Status Universal Sele ct ed Breast Cancer Psychological Disorder 0.289 0.231 0.143 0.243 0.044 0.105 0.284 0.328 0.085 0.206 0.557 0.643 72 17 2 3 0.028 0.614 0.459 Body Composition Normal Overweight 0.180 0.338 0.053 0.069 0.105 0.13 6 30 24 0 .001 Preinte rvention Fitness Level Sedentary/Low Activity Sedentary & Active 0.415 0.405 0.051 0.223 0.100 0.437 42 2 0.069 DESIGN Type of Control Gr ou p No treatment Placebo Active 0.369 0.232 0.138 0.055 0.060 0.135 0.108 0.117 0.265 49 41 8 0.305 Exercise Intervention Exercise -based Lecture -based Both 0.376 0.129 0.12 0 0.050 0.095 0.06 6 0.097 0.186 0.12 9 69 6 23 0.173 0.06 6
87 Table F 1. Continued M ean ES S tandard E rror 95% CI N umber of ES p value Recruitment Method Self -selected Nonself selected 0.379 0.220 0.059 0.054 0.117 0.105 38 53 Random Assignment Yes No 0.257 0.336 0.047 0.067 0.093 0.131 60 38 Publication Status Published Unpublished 0.291 0.322 0.043 0.08 7 0.084 0.171 85 13 Country of Origin US Canada Australia UK Turkey Sweden Switzerland Norway Germany 0.330 0.166 0.041 0.334 0.362 0.170 0.119 0.314 0.154 0.056 0.087 0.141 0.108 0.131 0.257 0.221 0.098 0.131 0.109 0.171 0.275 0.212 0.257 0.503 0.434 0.192 0.2 57 66 3 5 6 5 1 2 5 5 0.057 0.770 0.002 0.006 0.508 0.591 0.001 0.241 Validated Measure Standardized Unstandardized 0.295 0.201 0.039 0.195 0.077 0.382 92 6 0.302 Measure of Fitness Self report Objective Both 0.226 0.392 0.172 0.089 0.066 0. 109 0.174 0.129 0.214 12 39 5 0.011 0.113 EXERCISE Mode Aerobic Resistance Both 0.291 0.375 0.268 0.060 0.067 0.074 0.118 0.131 0.145 53 17 23 Met Exercise G uidelines Yes No 0.301 0.316 0.113 0.04 3 0.222 0.085 14 72 0 .008 Specificity of Exercise Specific Nonspecific 0.367 0.14 1 0.053 0.05 2 0.104 0.10 2 61 33 0 .007 Experimental Intervention Type Individual -based Group -based Both 0.341 0.262 0.320 0.111 0.044 0.132 0.217 0.087 0.260 12 75 7 0.002 0.016
88 Table F 1. Continued M ean ES S tandard E rror 95% CI N umber of ES p value Intensity of Exercise Light Moderate Hard/Very Hard 0.384 0.358 0.286 0.129 0.069 0.079 0.254 0.135 0.155 4 31 13 0.003 Fitness Experimental Improv e No Improve 0.344 0.351 0.064 0.094 0.126 0.184 37 14 Body Composition Experimental Improve No Improve 0.458 0.398 0.153 0.079 0.301 0.155 9 20 0.003 Fitness Control Improve No Improve 0.152 0.397 0.102 0.054 0.200 0.105 2 42 0. 136 Body Composition Control Improve No Improve 0.099 0.423 0.214 0.071 0.420 0.138 2 26 0.643 Theory Used for Measure Selection Yes No 0.431 0.286 0.143 0.040 0.281 0.079 4 94 0.003 Theory Used for Exercise Intervention Yes No 0.181 0.320 0.060 0.047 0.117 0.093 21 77 0.002 Note : CI = confidence interval; = p < .001 Table F 2. Continuous m ixed ES i nformation z value p value Number of ES PARTICIPANT Age 2.074 0.038 81 % Caucasian 0.074 0.941 28 DESIGN Attrition Rate % 0.515 0.607 49 Year of Publication 1.490 0.136 98 EXERCISE Duration (in min.) 0.248 0.804 76 Length (in weeks) 0.107 0.915 93 Frequency (per week) 2.416 0.016 79 Follow up (in weeks) 0.243 0.808 13
89 Table F 3. Categorical fixed ES information M ean ES S tandard E rror 95% CI N umber of ES p value PARTICIPANT Gender Male Female Both 0.163 0.212 0.266 0.056 0.037 0.042 0.110 0.074 0.082 12 55 27 0.004 Age Elementary Middle High school University Adults Older Adults Combined (elementary, middle, and high school) 0.006 0.281 0.530 0.178 0.442 0.296 0.156 0.072 0.097 0.155 0.040 0.056 0.082 0.054 0.141 0.189 0.304 0.079 0.110 0.161 0.106 7 8 2 35 26 12 17 0.929 0.004 0.001 0. 004 Ethnicity Caucasian NonCaucasian 0.140 0.691 0.042 0.136 0.082 0.267 26 3 0.001 Psychological Risk Status Universal Select ed Breast Cancer Psychological Disorder 0.222 0.221 0.143 0.201 0.027 0.069 0.284 0.197 0.054 0.135 0.557 0.387 72 17 2 3 0.001 0.614 0.308 Body Composition Normal Overweight 0.161 0.242 0.043 0.046 0.085 0.091 30 24 Pre intervention Fitness Level Sedentary/Low Activity Sedentary & Active 0.408 0.405 0.043 0.223 0.085 0.437 42 2 0.069 DESIGN Type of Control Group No treatment Placebo Active 0.269 0.227 0.023 0.035 0.039 0.082 0.069 0.077 0.161 49 41 8 0.775 Exercise Intervention Exercise -based Lecture -based Both 0.330 0.089 0.1 07 0.033 0.066 0.046 0.065 0.129 0.090 69 6 23 0.177 0.0 19 Recruitment Method Self -selected Non self selected 0.338 0.192 0.045 0.033 0.089 0.064 38 53
90 Table F 3. Continued M ean ES S tandard E rror 95% CI N umber of ES p value Random Assignment Yes No 0.204 0.226 0.032 0.039 0.064 0.076 60 38 Publication Status Published Unpublished 0.222 0.322 0.026 0.08 7 0.051 0.171 85 13 Country of Origin US Canada Australia UK Turkey Sweden Switzerland Norway Germany 0.241 0.166 0.006 0.334 0.362 0. 170 0.119 0.313 0.154 0.031 0.087 0.101 0.108 0.131 0.257 0.221 0.096 0.131 0.060 0.170 0.199 0.212 0.257 0.503 0.434 0.189 0.257 66 3 5 6 5 1 2 5 5 0.057 0.956 0.002 0.006 0.508 0.591 0.001 0.241 Validated Measure Standardized Unstandardized 0.232 0.210 0.026 0.081 0.051 0.159 92 6 0.009 Measure of Fitness Self report Objective Both 0.191 0.377 0.172 0.067 0.043 0.109 0.132 0.085 0.214 12 39 5 0.05 0.113 EXERCISE Mode Aerobic Resistance Both 0.232 0.375 0.214 0.036 0.067 0.050 0.070 0.131 0.098 53 17 23 Met Exercise Guidelines Yes No 0.146 0.268 0.052 0.030 0.102 0.059 14 72 0.005 Specificity of Exercise Specific Nonspecific 0.321 0.1 11 0.034 0.037 0.067 0.073 61 33 0.003 E xperimental Intervention Type Individual -based Group -based Both 0.214 0.219 0.320 0.054 0.029 0.132 0.106 0.057 0.20 12 75 7 0.016 Intensity of Exercise Light Moderate Hard/Very Hard 0.384 0.318 0.255 0.129 0.046 0.068 0.254 0.090 0.134 4 31 13 0.003
91 Table F 3. Continued M ean ES S tandard E rror 95% CI N umber of ES p value Fitness Experimental Improve No Improve 0.321 0.315 0.039 0.083 0.077 0.163 37 14 Body Composition Experimental Improve No Improve 0.515 0.331 0.102 0.055 0.201 0.108 9 20 Fitness Control Improve No Improve 0.145 0.371 0.090 0.043 0.177 0.084 2 42 0.109 Body Composition Control Improve No Improve 0.099 0.376 0.214 0.050 0.420 0.098 2 26 0.643 Theory Used for Measure Selec tion Yes No 0.431 0.224 0.143 0.025 0.281 0.05 4 94 0.003 Theory Used for Exercise Intervention Yes No 0.138 0.269 0.045 0.030 0.089 0.058 21 77 0.002 Note : CI = confidence interval; = p < .001 Table F 4. Continuous f ixed ES i nformation z value p value Number of ES PARTICIPANT Age 3.320 81 % Caucasian 0.168 0.867 28 DESIGN Attrition Rate % 0.884 0.377 49 Year of Publication 3.546 98 EXERCISE Duration (in min.) 0.655 0.513 76 Length (in weeks) 0.018 0.986 93 Fre quency (per week) 3.594 79 Follow up (in weeks) 0.243 0.808 13 Note : = p < .001
92 APPENDIX G OUTLIER TABLE Table G 1. Outlier table Author ES (SE) Z Value M Fixed ES (with removal) M Random ES (with removal) Potential Reason for Outlier Annessi ( 2005) 0.54 (0.24) 2.28 0.22 (0.03) 0.22 (0.05) 1. Population: Sedentary 2. Sample Size: E= 48, C = 30 Brown, Morrow, & Livingston (1982) 0.47 (0.20) 2.33 0.21 (0.03) 0.28 (0.05) 1. Population: Low in self concept Brown, Wang, Hinkle, Webber, Ahlquist, P uleo, et al. (1991) 0.89 (0.44) 2.03 0.22 (0.03) 0.28 (0.05) 1. Sample size: E1 = 11, E2 = 11, E3 = 9, E4 = 14, C = 12 Collingwood (1972) 0.77 (0.29) 2.61 0.21 (0.03) 0.28 (0.05) 1. Intervention: Short length = 4 weeks DAmato (1981) 0.60 (0.28) 2.10 0.2 2 (0.03) 0.28 (0.05) 1. Intervention: Inconsistent duration due to variation of 15 to 30 minutes depending on fitness level DiLorenzo, Bargman, Stucky Ropp, Brassington, Frensch, & LaFontaine (1999) 0.47 (0.22) 2.13 0.22 (0.03) 0.28 (0.05) 1. Population: Sedentary 2. Sample size: E = 82, C = 29 Evans, Newton, & Higgins (2005) 0.94 (0.38) 2.45 0.22 (0.03) 0.28 (0.05) 1. Population: Overweight 2. Sample Size: E = 23, C = 11 Finkenberg, DiNucci, & McCune (1993) (females only) 0.37 (0.14) 2.67 0 .25 (0.03) 0.30 (0.05) 1. Population: Largely female (54%) 2. Sample size: E = 116, C = 99
93 Table G 1. Continued Author ES (SE) Z Value M Fixed ES (with removal) M Random ES (with removal) Potential Reason for Outlier Fisher & Thompson (1994) 0.87 (0.3 5) 2.49 0.22 (0.03) 0.28 (0.05) 1. Population: Low on MBSRQ scores Hilyer & Mitchell (1979) (low self concept only) 1.02 (0.34) 3.04 0.23 (0.03) 0.30 (0.05) 1. Population: Either high or low in self -concept 2. Sample size: total N = 120 King, Taylor Haskell, & DeBusk (1989) 1.14 (0.21) 5.53 0.20 (0.03) 0.26 (0.05) 1. Population: Sedentary 2. Intervention: Long length = 24 weeks Perry, Rosenblatt, Kempner, Feldman, Paolercio, & Van Bemden (2002) 0.74 (0.19) 3.79 0.21 (0.03) 0.28 (0.05) 1. Population : Mostly Hispanic (72%) 2. Sample size: E = 161, C = 33 Pinto, Clark, Maruyama, et al. (2003) 1.99 (0.60) 3.32 0.22 (0.03) 0.28 (0.05) 1. Population: Overweight, older adult, breast cancer 2. Sample size: E = 12, C = 12 Sherblom & Rust (2004) 1.04 ( 0.22) 4.87 0.21 (0.03) 0.27 (0.05) 1. Sample size: E1 = 35, E2 = 56, C = 44 Smith & Michel (2006) 1.03 (0.34) 3.07 0.21 (0.03) 0.28 (0.05) 1. Population: Largely African American (60%), overweight, pregnant women Talbot & Taylor (1998) 0.45 (0.23) 1.98 0.22 (0.03) 0.28 (0.05) 1. Population: Sedentary, older adult, at risk for heart disease Tucker (1987) 0.50 (0.13) 3.84 0.21 (0.03) 0.28 (0.05) 1. Sample size: E = 114, C = 127
94 Table G 1. Continued Author ES (SE) Z Value M Fixed ES (with removal) M R andom ES (with removal) Potential Reason for Outlier Williams & Cash (2001) 0.47 (0.23) 2.04 0.22 (0.03) 0.28 (0.05) 1. Population: Sedentary, largely African American (56%) All studies removed: 0.12 (.03) 0.12 (.03) Note : E = experimental group; C = control group
95 APPENDIX H DESCRIPTIVE CHARACTE RISTICS OF STUDIES
96 Table H 1. Descriptive characteristics of studies included in the m eta analysis Study Participants by Group (E = Experimental Number; C = Control Number) Mean Age Body Image Measure I ntervention Type Intervention Length (weeks) and Intensity Alfermann & Stoll (2000) E = 24 (Resistance + Aerobic) C = 13 (No treatment) E = 36.7 C = 39.3 Concerns About Physical Attractiveness Group 24 Moderate Alfermann & Stoll (2000) 1. E = 31 (Resistance + Aerobic) C = 16* (Placebo) E & C = 43.2 Concerns About Physical Attractiveness Group 24 Moderate 2. E = 26 (Aerobic) C = 16* (Placebo) E & C = 43.2 Anderson, Murphy, Murtagh, & Nevill (2006) 1. E = 10 (Aerobic) C = 9* (No treatment) E & C = 38.1 EDI Body Dissatisfaction Scale Individual 8 Moderate 2. E = 9 (Aerobic) C = 9* (No treatment) E & C = 38.1 Annessi (2005) E = 48 (Aerobic) C = 30 (No treatment) E & C = 41.4 Body Esteem Scale Weight Concern Group 12 Hard/Very Hard Asci (2002) Male: E = 33 (Aerobic) C = 32 (Placebo) E & C = 22.8 Physical Self Perceptions Profile Attractive Body Group 10 Hard/Very Hard Female: E = 37 (Aerobic) C = 36 (Placebo) E & C = 21.7 Asi (2003) E = 20 (Aerobic) C = 20 (Placebo) E = 21.35 C = 21.20 Physical Self description Questionnaire Body Fat subscale Group 10 Hard/Very Hard
97 Table H 1. Continued Study Participants by Group (E = Experimental Number; C = Control N umber) Mean Age Body Image Measure Intervention Type Intervention Length (weeks) and Intensity Asci, Kin, & Kosar (1998) 1. E = 15 (Aerobic) C = 15* (Placebo) Age range = 19 28 Physical Self Perceptions Profile Physical Attractiveness Group 8 Moder ate 2. E = 15 (Aerobic) C = 15* (Placebo) Age range = 19 28 Barenholtz (1995) E = 25 (Resist. + Aerobic, TX) C = 25 (No treatment) Middle school EDI Body Dissatisfaction Scale Group 6 NP Bartlewski, Van Raalte, & Brewer (1996) E = 15 (Aerobi c) C = 28 (Placebo) E & C = 22.12 Body Esteem Scale .87) Group 10 NP Ben Schlomo & Short (1986) 1. E = 5 (Aerobic) C = 5* (No treatment) Age range = 21 45 Tennessee Self Concept Scale Physical Self subscale Group 6 Hard/Very Hard 2. E = 4 (Aerobic) C = 5 (No treatment) Age range = 2145 Bowden, Rust, Dunsmore, & Briggs (2005) E = 140 (Resistance + Aerobic) C = 77 (Placebo) University SPAS 12 item Group 16 NP Brown & Harrison (1986) Young: E = 21 (Resistance) C = 21 (No treatment) E & C = 21.5 Tennessee Self concept Scale Physical Self subscale Group 12 Moderate & Hard Mature: E = 23 (Resistance) C = 18 (No treatment) E & C = 44.4 Brown, Morrow, & Livingston (1982) E = 50 (Aerobic, TX) C = 50 (Placebo) E = 24.2 C = 20.5 Tennessee Self co ncept Scale Physical Self Group 14 NP
98 Table H 1. Continued Study Participants by Group (E = Experimental Number; C = Control Number) Mean Age Body Image Measure Intervention Type Intervention Length (weeks) and Intensity Brown, Wang, Hinkle, Webber, Ahlquist, Puleo, et al. (1991) 1. E = 11 (Resistance) C = 12* (No treatment) NP Body Cathexis Scale NP 12 Light 2. E = 11 (Resistance) C = 12* (No treatment) NP 3. E = 9 (Resistance) C = 12* (No treatment) NP 4 E = 14 (Resistance) C = 12* (No treatment) NP Brown, Wang, Ward, Ebbeling, Fortlage, Puleo, Benson, & Rippe (1995) 1. E = 12 (Aerobic) C = 17* (No treatment) E = 52.7 C = 52.1 Body Cathexis Scale Group 16 Moderate 2. E = 18 (Aerobic) C = 17* (No treatment) E = 53.4 C = 52.1 16 Light 3. E = 15 (Aerobic, TX) C = 17* (No treatment) E = 53.7 C = 52.1 4. E = 7 (Martial arts, TX) C = 17* (No treatment) E = 50.9 C = 52.1 Cocklin (1989) 1. E = 22* (Aerobic) C = 24 (Act ivity) E = 25.62 C = 31.21 Body Cathexis Scale Group 8 Hard/Very Hard 2. E = 22* (Aerobic) C = 23 (No treatment) E = 25.62 C = 27.83 Collingwood (1972) E = 25 (Resistance + Aerobic) C = 25 (Placebo) Age range = 18 26 Body Attitude Scale Evalua tive subscale Group 4 NP Daley, Copeland, Wright, Roalfe, & Wales (2006) 1. E = 28* (Aerobic, TX) C = 30 (No treatment) E & C = 13.1 Children and Youth Physical Self Perception Profile Attractive Body Adequacy subscale Individual 8 Moderate 2. E = 28* (Aerobic, TX) C = 23 (Activity) E & C = 13.1
99 Table H 1. Continued Study Participants by Group (E = Experimental Number; C = Control Number) Mean Age Body Image Measure Intervention Type Intervention Length (weeks) and Intensity DAmato (1 981) E = 27 (Aerobic) C = 25 (Placebo) E & C = 26.6 Body Cathexis Scale Group 8 NP DiLorenzo, Bargman, StuckyRopp, Brassington, Frensch, & LaFontaine (1999) E = 82 (Aerobic) C = 29 (No treatment) E = 33.05 C = 29.45 Tennessee Self Concept Scale Physic al Self subscale Group 12 Hard/Very Hard Eliot (1998) 1. E = 13 (Resistance + Aerobic) C = 12* (No treatment) E = 47.4 C = 38.5 BASS of MBSRQ Both 6 NP 2. E = 16 (Resist. + Aerobic, TX) C = 12* (No treatment) E = 38.9 C = 38.5 Evans, Newton, & Higgins (2005) E = 23 (Not specified, TX) C = 11 (No treatment) E = 34.6 C = 33.6 Clinical Global Impressions Body Image Individual 12 NP Finkenberg, DiNucci, & McCune (1993) Female: E = 116 (Aerobic) C = 99 (Activity) University Body Esteem Scale .87) Group NP Male: E = 38 (Aerobic) C = 60 (Activity) University Fisher & Thompson (1994) 1. E = 18* (Resist + Aerobic, TX) C = 18 (Placebo) E & C = 23.5 EDI Body Dissatisfaction Scale Both 6 NP 2. E = 18* (Resist + Aerobic, TX) C = 18 (No treatment) E & C = 23.5
100 Table H 1. Continued Study Participants by Group (E = Experimental Number; C = Control Number) Mean Age Body Image Measure Intervention Type Intervention Length (weeks) and Intensity For d, Puckett, Blessing, & Tucker (1989) 1. E = 21 (Aerobic) C = 20* (Placebo) E & C = 19.8 Body Cathexis Scale Group 8 NP 2. E = 17 (Aerobic) C = 20* (Placebo) E & C = 19.8 3. E = 15 (Aerobic) C = 20* (Placebo) E & C = 19.8 4. E = 22 (Resistance) C = 20* (Placebo) E & C = 19.8 Ford, Puckett, Reeve, & Lafavi (1991) 1. E = 23 (Resistance) C = 35* (Placebo) University Body Cathexis Scale Group 8 NP 2. E = 35 (Resistance) C = 35* (Placebo) University 3. E = 20 (Aer obic) C = 35* (Placebo) University Fossati, Amati, Painot, Reiner, Haenni, & Golay (2004) 1. E = 25* (Aerobic, TX) C = 13 (Placebo) E = 37.4 C = 45.6 Body Dissatisfaction of EDI 2 Both 12 NP 2. E = 25* (Aerobic, TX) C = 23 (Placebo) E = 37.4 C = 42.3 Gehrman, Hovell, Sallis, & Keating (2006) Male: E = 16 (Resist. + Aerobic, TX) C = 16 (Placebo) E & C = 11.5 Body Dissatisfaction of EDI 2 Group 8 NP Female: E = 33 (Resist. + Aerobic, TX) C = 19 (Placebo) E & C = 11.5 Gilman (199 6) 1. E = 49 (Aerobic) C = 9* (Placebo) E & C = 20.5 BASS of MBSRQ Group 14 NP 2. E = 40 (Resistance) C = 9* (Placebo) E & C = 20.5
101 Table H 1. Continued Study Participants by Group (E = Experimental Number; C = Control Number) Mean Age Bod y Image Measure Intervention Type Intervention Length (weeks) and Intensity Hase (1995) 1. E = 18* (Resistance) C = 20 (Placebo) E = 14.5 C = 15.0 EDI Body Dissatisfaction Scale Group 8 Hard/Very Hard 2. E = 18* (Resistance) C = 20 (No treatme nt) E = 14.5 C = 13.9 Henry, Anshel, & Michael (2006) 1. E = 23 (Aerobic) C = 21* (Activity) E = 19.4 C = 20.1 Body Self image Questionnaire Fatness Evaluation Group 12 Moderate 2. E = 28 (Resistance) C = 21* (Activity) E = 19.1 C = 20.1 Hilyer & Mitchell (1979) (High Self -concept: 1 & 2; Low Self concept: 3 & 4) 1. E = 20 (Ae robic) C = 20* (Placebo) E & C = 19.8 Tennessee Self concept Scale Physical Self Group 10 NP 2. E = 20 (Aerobic, TX) C = 20* (Placebo) E & C = 19.8 3. E = 20 (Aerobic) C = 20* (Placebo) E & C = 19.8 4. E = 20 (Aerobic, TX) C = 20* (Placebo) E & C = 19.8 Huang, Norman, Zabinski, Calfas, & Patrick (2007) Female: E = 175 (Not specified, TX) C = 174 (No treatment) Age Range = 1214 EDI Body Dissatisfaction Scale (modified) Individual 52 NP Male: E = 166 (Not specified, TX) C = 142 (No treatment) Age Range = 1214 King, Taylor, Haskell, & DeBusk (1989) E = 57 (Aerobic) C = 52 (No treatment) E & C = 48 Satisfaction with Physical Shape and Appearance (unstandardized) Individual 24 Moderate
102 Tab le H 1. Continued Study Participants by Group (E = Experimental Number; C = Control Number) Mean Age Body Image Measure Intervention Type Intervention Length (weeks) and Intensity Lindwall & Lindgren (2005) E = 27 (Aerobic, TX) C = 35 (No treatment) E & C = 16.35 Physical Self perceptions Profile Bodily Attractiveness subscale Group 22 NP Marquez Sterling, Perry, Kaplan, Halberstein, & Signorile (2000) E = 9 (Aerobic) C = 6 (No treatment) E = 31.3 C = 27.8 Body Cathexis Scale Group 15 Hard/Very Hard McCabe, Ricciardelli, & Salmon (2006) 3/4 th grade Male: E = 44 (Aerobic, TX) C = 36 (No treatment) E = 10.25 C = 10.23 Body Image and Body Change Questionnaire for Children Weight Dissatisfaction (single item) Group 8 NP 5/6 th grade Male: E = 64 (Aer obic, TX) C = 51 (No treatment) 3/4 th grade Female: E = 41 (Aerobic, TX) C = 33 (No treatment) E = 10.08 C = 9.96 5/6 th grade Female: E = 51 (Aerobic, TX) C = 48 (No treatment) Mock, Burke, Sheehan, Creaton, Winningham, & Liebman (1994) E = 9 (Aerobic, TX) C = 5 (No treatment) E & C = 44 Body Image Visual Analogue Scale Group 20 Moderate OLoughlin, Paradia, Meshefedjian, & Kishchuck (1998) E = 82 (Not specified, TX) C = 75 (No treatment) E = 39.2 C = 37.0 Satisfaction with Appearanc e (author developed; unstandardized) Individual 8 Light
103 Table H 1. Continued Study Participants by Group (E = Experimental Number; C = Control Number) Mean Age Body Image Measure Intervention Type Intervention Length (weeks) and Intensity Perry, Rosenb latt, Kempner, Feldman, Paolercio, & Van Bemden (2002) E = 161 (Resist. + Aerobic, TX) C = 33 (No treatment) E = 16.53 C = 15.61 Body Satisfaction body silhouette (adapted from Stunkard & Sorensen) Group 24 Moderate Pinto, Clark, Maruyama, et al. (2003) E = 12 (Resistance + Aerobic) C = 12 (No treatment) E & C = 52.5 BES Weight Concern Group 12 Moderate Pinto, Frierson, Rabin, Trunzo, & Marcus (2005) E = 39 (Aerobic) C = 43 (No treatment) E = 53.42 C = 52.86 BES Weight Concern Individual 12 Moderat e Sandel, Judge, Landry, Faria, Ouellette, & Majczak (2005) E = 19 (Resistance + Aerobic) C = 18 (No treatment) E = 59.7 C = 59.5 Body Image Scale Group 12 NP Scanlon (1991) E = 18 (Aerobic) C = 15 (Activity) E & C = 12.6 EDI Body Dissatisfaction Scale Group 5 Moderate Shaw, Ebbeck, & Snow (2000) E = 18 (Resistance) C = 22 (No treatment) E = 64.2 C = 62.5 Physical Self concept Profile Physical Appearance subscale Group 36 NP Sherblom & Rust (2004) 1. E = 35 (Aerobic) C = 44* (Placebo) E & C = 12 .1 Body Image Satisfaction Questionnaire Group 6 NP 2. E = 56 (Aerobic) C = 44* (Placebo) E & C = 12.1 Smith & Michel (2006) E = 20 (Aerobic) C = 20 (No treatment) E = 25.1 C = 24.8 Pregnant Body Shape Group 6 Moderate
104 Table H 1. Continued Study Participants by Group (E = Experimental Number; C = Control Number) Mean Age Body Image Measure Intervention Type Intervention Length (weeks) and Intensity Sorensen, Anderssen, Hjeran, Holme, & Ursin (1997) 1. E = 48 (Resis tance + Aerobic) C = 53* (Placebo) E & C = 44.9 Harter Adult Self perception Profile Appearance subscale Group 52 Moderate 2. E = 64 (Resist. + Aerobic, TX) C = 53* (Placebo) E & C = 44.9 3. E = 48 (Resistance + Aerobic) C = 43* (No tr eatment) E & C = 44.9 4. E = 64 (Resist. + Aerobic, TX) C = 43* (No treatment) E & C = 44.9 Stein (1989) 1. E = 28 (Aerobic) C = 35* (Placebo) E & C = 20.02 Body Cathexis Scale Group 7 Moderate 2. E = 26 (Resistance) C = 35* (Placeb o) E & C = 20.02 Stock, Miranda, Evans, Plessis, Ridley, Yeh, & Chanoine (2007) E = 228 (Aerobic, TX) C = 132 (No treatment) Elementary Figure Rating Scale (modified) Group 21 Hard/Very Hard Stoll & Alfermann (2002) 1. E = 42* (Resistance + Aerobic) C = 18 (Placebo) E = 61.60 C = 59.67 Body Self concept Physical Attractiveness Group 14 Moderate 2. E = 42* (Resistance + Aerobic) C = 28 (No treatment) E = 61.60 C = 61.93
105 Table H 1. Continued Study Participants by Gr oup (E = Experimental Number; C = Control Number) Mean Age Body Image Measure Intervention Type Intervention Length (weeks) and Intensity Sundgot Borgen, Rosenvinge, Bahr, & Schneider (2002) E = 12 (Resistance + Aerobic) C = 15 (No treatment) E = 23 C = 22 EDI Body Dissatisfaction Scale Both 16 Moderate Talbot & Taylor (1998) E = 46 (Aerobic) C = 35 (No treatment) E & C = 54.2 Physical Self perceptions Profile Appearance subscale (modified) Individual 10 NP Taylor & Fox (2005) E = 97 (Aero bic) C = 45 (No treatment) E = 54.1 C = 54.4 Physical Self Perceptions Profile Body Appearance subscale Individual 16 Moderate Tucker (1987) E = 114 (Resistance) C = 127 (Placebo) University Body Cathexis Scale Group 16 NP Williams & Cash (2001) E = 39 (Resistance) C = 39 (No treatment) E & C = 21.7 BASS of MBSRQ Group 6 NP Zabinski, et al (2001) Female: E = 80 (Resist. + Aerobic, TX) C = 97 (Placebo) E & C = 24 EDI Female Concerns/Body .90) Group 15 Moderate Male: E = 79 (Resist. + Aerobic, TX) C = 66 (Placebo) E & C = 24 EDI Male Concerns/Body Dissatisfaction Note: Body image measures were standardized, unless otherwise reported, and the alpha (if reported) appears in () after the scale; Both = Individual and g roup based; NP = Not p rovided; TX = Exercise in addition to another treatment; = Same experimental or control group
106 APPENDIX I RANDOM EFFECTS MODER ATOR ANALYSES Table I 1. Random e ffects moderator a nalyses Number of ES Q B df p value PARTICIPANT Gender ( female, male, both ) Female = 55 Male = 12 Both = 27 1.072 2 0.585 Age ( combined university, adults, older adults ) Combined = 17 University = 35 Adults = 26 Older Adults = 12 8.873 3 0.031 Ethnicity ( Caucasian, NonCaucasian ) Caucasian = 26 NonCaucasian = 3 10.552 1 0.001 Psychological Risk Status ( universal, selected ) Universal = 72 Selected = 17 0.263 1 0.608 Body Composition ( normal, overweigh t ) Normal = 30 Overweight = 24 3.246 1 0.072 DESIGN Type of Control Gr ou p ( no treatment, placebo, active ) No treatment = 49 Placebo = 41 Active = 8 4.241 2 0.120 Exercise Intervention ( exercise based, lecture -based, both) Exercise based = 69 Lecture -based = 6 Both = 23 11.875 2 0.003 Recruitment Method ( self selecte d, nonself-selected ) Self selected = 38 Nonself -selected = 53 3.914 1 0.048 Random Assignment ( yes, no ) Yes = 60 No = 38 0.928 1 0.336 Publication Status ( published, unpublished) Publis hed = 85 Unpublished = 13 0.102 1 0.750 Validated Measure ( standardized, unstandardized) Standardized = 92 Unstandardized = 6 0.226 1 0.635 Measure of Fitness ( self report, objective, both ) Self report = 12 Objective = 39 Both = 5 4.021 2 0.134 EXERCISE Mode ( a erobic, resistance, both ) Aerobic = 53 Resistance = 17 Both = 23 1.374 2 0.503 Met Exercise G uidelines ( yes, no ) Yes = 14 No = 72 0.015 1 0.902 Specificity of Exercise ( specific, nonspecific ) Specific = 61 Nonspecific = 33 9.229 1 0.002
107 Ta ble I 1. Continued Number of ES Q B df p value Experimental Intervention Type (individual -based, group-based, both) Individual based = 12 Group -based = 75 Both = 7 0.553 2 0.758 Intensity of Exercise ( li ght, moderate, hard/very hard) Light = 4 Moderate = 31 Hard/Very Hard = 13 0.637 2 0.727 Fitness Experimental ( improve, no improve ) Improve = 37 No Improve = 14 0.004 1 0.951 Body Composition Experimental (improve, no improve ) Improve = 9 No Improve = 20 0.121 1 0.728 Theory Used for Measure Selec tion ( yes, no) Yes = 4 No = 94 0.951 1 0.329 Theory Used for Exercise Intervention ( yes, no ) Yes = 21 No = 77 3.298 1 0.069
108 REFERENCE LIST Altabe, M., & Thompson, J. K. (1993). Body image changes during early adulthood. International Journal of Eating Disorders, 13, 323 328. American College of Sports Medicine (2000). ACSMs guidelines for exercise testing and prescription (6th ed.). Baltimore MD : Williams & Wilkins American Psychiatric Association. (1994). Diagnostic and statistical manual of m ental disorders. (4th ed.) Washington, DC: Author. Asi, F.H. (2003). The effects of physical fitness training on trait anxiety and physical self concept of female university students. Psychology of Sport and Exercise, 4(3), 255 264. Banasiak, S. J., P axton, S. J., & Hay, P. (1998). Evaluating accessible treatments for bulimic eating disorders in primary care. Australian Journal of Primary Health, 4, 147 155. Bane, S., & McAuley, E. (1998). Body image and exercise. In J.L. Duda (Ed.), Advances in sport and exercise psychology measurement ( pp. 311 324). Morgantown, WV: Fitness Information Technology. Barber, J.P. & Milrod, B. (2004). Pitfalls of Meta analysis. American Journal of Psychiatry, 161(6), 1131. Becker, B. J. (2000). Multivariate meta analys is. In H. E. A. Tinsley & S. Brown (Eds.), Handbook of applied and multivariate statistics and mathematical modeling (pp. 499 525). San Diego CA : Academic Press Biddle, S. J. H. (2000). Emotion, mood, and physical activity. In S. J. H. Biddle, K. R. Fox & S. H. Boutcher, (Eds.), Physical activity and psychological well -being (pp. 63 88). London UK : Routledge. Biddle, S.J.H. (2006). Research synthesis in sport and exercise psychology; Chaos in the brickyard revisited. European Journal of Sport Science, 6(2), 97 102. Blue, C. L., & Black, D. R. (2005). Synthesis of intervention research to modify physical activity and dietary behaviors. Research and Theory for Nursing Practice, 19, 25 61. Bompa, T.O. (1999). Periodization: theory and methodology of tr aining ( 4th ed.). Champaign IL : Human Kinetics. Borenstein, M. (2005). Software for publication bias. In Rothstein, H.R., Sutton, A.J., & Borenstein, M. (Eds.). (2005). Publication bias in meta-analysis: Prevention, assessment and adjustment (pp. 193 220). Sussex UK : John Wiley and Sons. Borenstein, M., Hedges, L.V., Higgins, J.P.T., & Rothstein, H.R. (2007). Introduction to metaanalysis. Unpublished manuscript.
109 Borenstein, M., & Rothstein, H. (1999). Comprehensive meta -analysis. A computer program f or research synthesis. Englewood, NJ: Biostat Bryan, A.D., & Rocheleau, C.A. (2002). Predicting aerobic versus resistance exercise using the theory of planned behavior. American Journal of Health Behavior, 26(2), 83 94. Bushman, B. J. (1994). Vote -count ing procedures in meta analysis. In H. Cooper and L. V. Hedges (Eds.), The handbook of research synthesis (pp. 193 213). New York NY : Russell Sage Foundation. Bushman, B.J. & Wang, M.C. (1995). A procedure for combining sample correlation coefficients an d vote counts to obtain an estimate and a confidence interval for the population correlation coefficient. Psychological Bulletin, 117(3) 530 546. Bushman, B. J., & Wang, M. C. (1996). A procedure for combining sample standardized mean differences and vot e counts to estimate the population standardized mean difference in fixed effect models. Psychological Methods, 1, 66 80. Callaghan, P. (2004). Exercise: A neglected intervention in mental health care? Journal of Psychiatric and Mental Health Nursing, 11, 476 483. Carron, A.V., Hausenblas, H.A., & Estabrooks, P.A. (2003). The psychology of physical activity New York, NY: McGrawHill. Cash, T.F. (1996). The treatment of body image disturbances. In J.K. Thompson (Ed.), Body image, eating disorders, and obesity: An integrative guide for assessment and treatment (pp. 83 107). Washington, D.C.: American Psychological Association. Cash, T. F. (2002). Cognitive behavioral perspectives on body image. In T. F. Cash & T. Pruzinsky (Eds.), Body image: A handbook of theory, research, and clinical practice (pp. 38 46). New York NY: Guilford Press. Cash, T. F., & Fleming, E. C. (2002). Body image and social relations. In T. F. Cash & T. Pruzinsky (Eds.), Body image: A Handbook of theory, research, and clinical prac tice (pp. 277 286). New York, NY : Guilford Press. Cash, T.F., & Henry, P.E. (1995). Womens body images: The results of a national survey in the U.S.A. Sex Roles, 33(1/2), 19 28. Cash, T.F. & Lavallee, D.M (1997). Cognitive -behavioral body image therapy: Extended evidence of the efficacy of a self -directed program. Journal of Rational -Emotive & Cognitive -Behavioral Therapy, 15(4), 281 294. Cash, T.F., & Pruzinsky, T. (Eds.). (2002). Body image: A handbook of theory, research, and clinical practice New Y ork NY: The Guilford Press.
110 Cash, T.F., & Strachan, M.D. (1999). Body images, eating disorders, and beyond. In R. Lemberg & L. Cohn (Eds.), Eating disorders: A reference sourcebook (pp. 27 37). New York NY: Greenwood. Caspersen, C. J., Powell, K. E., & Christenson, G. M. (1985). Physical activity, exercise, and physical fitness: Definitions and distinctions for health -related research. Public Health Reports, 100, 126 131. Clarke, G., Hawkins, W., Murphy, M., Sheeber, L., Lewinsohn, P. M., & Seeley, J. R. (1995). Targeted prevention of unipolar depressive disorder in an at risk sample of high school adolescents: A randomized trial of group cognitive intervention. Journal of the American Academy of Child and Adolescent Psychiatry, 34, 312 321. Colcombe, S., & Kramer, A. F. (2003). Fitness effects on the cognitive function of older adults: A meta analytic study. Psychological Science, 14, 125 130. Conn, V., & Rantz, M. (2003). Research methods: Managing primary study quality in meta analyses? Research Nur sing Health, 26, 322 333. Conn, V., Valentine, J., Cooper, H., Rantz, M. (2003). Grey literature in meta analyses. Nursing Research, 52 256 261. Cook, D. J., Guyatt, G. H., & Ryan, G. (1993). Should unpublished data be included in meta analyses? JAMA, 2 69, 749 753. Craft, L. L., & Landers, D. M. (1998). The effects of exercise on clinical depression and depression resulting from mental illness: A meta analysis. Journal of Sport & Exercise Psychology, 20, 339 357. Croghan, I. T., Bronars, C., Patten, C. A., Schroeder, D. R., Nirelli, L. M., Thomas, J. L., Clark, M. M., Vickers, K. S., Foraker, R., Lane, K., Houlihan, D., Offord, K. P., & Hurt, R. D. (2006). Is smoking related to body image satisfaction, stress, and self -esteem in young adults? American J ournal of Health Behavior, 30, 322 333. Daubenmier, J.J. (2005). The relationship of yoga, body awareness, and body responsiveness to self -objectification and disordered eating. Psychology of Women Quarterly, 29, 207 219. Dibbell Hope, S. (2000). The use of dance/movement therapy in psychological adaptation to breast cancer. The Arts in Psychotherapy, 27(1), 51 68. Duval, S., & Tweedie, R. (2000 ). Trim and fill: A simple funnel-plot -based method of testing and adjusting for publication bias in meta analy sis. Biometrics, 56(2), 455 463. Egger, M., Smith, G. D., Schneider, M., & Minder, C. (1997). Bias in meta -analysis detected by a simple graphical test. British Medical Journal, 315, 629 634.
111 Eisenberg, M. E., Neumark Sztainer, D., & Paxton, S. J. (2006) Five year change in body satisfaction among adolescents. Journal of Psychosomatic Research, 61, 521 527. Elgin, J., & Pritchard, M. (2006). Gender differences in disordered eating and its correlates. Eating and Weight Disorders, 11, 96 101. Fallon, E.A ., & Hausenblas, H.A. (2005). Media images of the ideal female body: Can acute exercise moderate their psychological impact ? Body Image, 2, 62 73. Feingold, A., & Mazzella, R. (1998). Gender differences in body image are increasing. Psychological Scienc es, 9, 190 195. Fisher, E., & Thompson, J.K. (1994). A comparative evaluation of cognitive -behavioral therapy (CBT) versus exercise therapy (ET) for the treatment of body image disturbance: Preliminary findings. Behavioral Modification, (18), 171 185. F ox, K. R. (2000). The effects of exercise on self -perceptions and self -esteem. In S. J. H. Biddle, K. R. Fox, & S. H. Boutcher, (Eds.), Physical activity and psychological well -being (pp. 88 117). London, UK : Routledge. Fox, K.R. & Corbin, C.B. (1989). Th e physical self -perception profile: Development and preliminary validation. Journal of Sport and Exercise Psychology, 11(4), 408 430. Franklin, J., Denyer, G., Steinbeck, K. S., Caterson, I. D., & Hill, A. J. (2006). Obesity and risk of low self -esteem: A statewide survey of Australian children. Pediatrics, 118, 2481 2487. Garner, D. M. (1991). Eating Disorders Inventory 2: Professional Manual Odessa, FL: Psychological Assessment Resources. Garner, D.M. (1997, January/February). The 1997 bodyimage su rvey results. Psychology Today, 30(1), 30 44, 75 80, 84. Glass, G.V. (1976). Primary, secondary, and meta analysis of research. Educational Researcher, 5, 3 8. Gollings, E.K., & Paxton, S.J. (2006). Comparison of internet and face -to -face delivery of a g roup body image and disordered eating intervention for women: A pilot study. Eating Disorders, 14 1 15. Grabe, S., & Hyde, J. S. (2006). Ethnicity and body dissatisfaction among women in the United States: A meta analysis. Psychological Bulletin, 132, 6 22 640.
112 Groesz, L.M., Levine, M.P., & Murnen, S.K. (2002). The effect of experimental presentation of thin media images on body satisfaction: A meta analytic review. International Journal of Eating Disorders, 31, 1 16. Grogan, S. (2006). Body image and health: Contemporary perspectives. Journal of Health Psychology, 11, 523 529. Hagger, M.S. (2006). Meta analysis in sport and exercise research: Review, recent developments, and recommendations. European Journal of Sport Science, 6(2), 103 115. Hausenbla s, H.A., Cook, B.J., & Chittester, N.I. (2008). Can exercise treat eating disorders? Exercise and Sport Sciences Reviews, 36,1, 43 47. Hausenblas, H.A., & Fallon, E.A. (2006). Exercise and body image: A meta analysis. Psychology and Health, 21(1), 33 47. Hausenblas, H.A. & Symons Downs, D. (2002). Exercise dependence: A systematic review. Psychology of Sport and Exercise, 3, 89 123. Hechler, T., Beumont, P., Marks, P., & Touyz, S. (2005). How do clinical specialists understand the role of physical activ ity in eating disorders? European Eating Disorders Review, 13(2), 125 132. Hedges, L. V., & Olkin, I. (1985). Statistical methods for meta -analysis New York NY: Academic. Higgins, J., & Thompson, S. (2002). Quantifying heterogeneity in a meta analysis. Stat istics in Med icine 15 1539 1558. Higgins, J., Thompson, S., Deeks, J., & Altman, D. (2003). Measuring inconsistencies in meta analyses. British Medical Journal, 327, 557 560. Hopewell, S., Clarke, M., & Mallett, S. (2005). Grey literature and syst ematic reviews. In Rothstein, H.R., Sutton, A.J., & Borenstein, M. (Eds.). (2005). Publication bias in metaanalysis: Prevention, assessment and adjustment (pp. 241 259). Sussex: John Wiley and Sons. Hopewell, S., McDonald, S., Clarke, M., & Egger, M. (20 07). Grey literature in meta analyses of randomized trials of health care interventions. Cochrane Database of Systematic Reviews Apr 18: MR000010. Howley, E. T. (2001). Type of activity: Resistance, aerobic and leisure versus occupational physical activit y. Medicine and Science in Sports and Exercise, 33, S364 S369. Hrabosky, J. J., Masheb, R. M., White, M. A., Rothschild, B. S., Burke -Martindale, C. H., & Grilo, C. M. (2006). A prospective study of body dissatisfaction and concerns in
113 extremely obese gas tric bypass patients: 6 and 12-month postoperative outcomes. Obesity Surgery, 16, 1615 1621. Hunter, J. E., & Schmidt, F. L. (2004). Methods of meta-analysis. Error and bias in research findings (2 nd ed ). Sage Publications. Jarry, J.L. & Berardi, K. (2 004). Characteristics and effectiveness of stand alone body image treatments: a review of the empirical literature. Body Image, 1, 319 333. Jarry, J.L., & Ip, K. (2005). The effectiveness of standalone cognitive -behavioral therapy for body image: A meta analysis. Body Image, 2, 317 331. Johnson, F., & Wardle, J. (2005). Dietary restraint, body dissatisfaction, and psychological distress: A prospective analysis. Journal of Abnormal Psychology, 114, 119 125. Katz, J.L. (1986). Long-distance running, anore xia nervosa, and bulimia: A report of two cases. Comprehensive Psychiatry, 27(1), 74 78. Khoshdel, A., Attia, J., & Carney, S. L. (2006). Basic concepts in meta analysis: A primer for clinicians. International Journal of Clinical Practice, 60, 1287 1294. Lavallee, D. M., & Cash, T. F. (1997, November). The comparative efficacy of two cognitive behavioral self -help programs for a negative body image. Poster presented at the annual meeting of the Association for Advancement of Behavior Therapy, Miami Beach. Lindwall, M., & Lindgren, E. (2005). The effects of a 6 -month exercise intervention programme on physical self -perceptions and social physique anxiety in non-physically active adolescent Swedish girls. Psychology of Sport and Exercise, 6(6), 643 658. L oland, N.W. (2000). The aging body: Attitudes toward bodily appearance among physically active and inactive women and men of different ages. Journal of Aging and Physical Activity, 8, 197 213. Lowry Webster, H. M., Barrett, P. M., & Dadds, M. R. (2001). A universal prevention trial of anxiety and depressive symptomatology in childhood: Preliminary data from an Australian study. Behavior Change, 18, 36 50. MangwethMatzek, B., Rupp, C. I., Hausmann, A., Assmayr, K., Mariacher, E., Kemmler, G., Whitworth, A. B., & Bieble, W. (2006). Never too old for eating disorders or body dissatisfaction: A community study of elderly women. International Journal of Eating Disorders, 39 583 586. Marsella, A.J., Shizuru, L., Brennan, J., & Kameoka, V. (1981). Depression and body image satisfaction Journal of Cross -Cultural Psychology 12(3), 360 371.
114 Martin, K. A., & Lichtenberger, C. M. (2002). Fitness enhancement and changes in body image. In T. F. Cash & T. Pruzinksy (Eds ), A handbook of theory, research, and clinical practice (pp. 414 421). New York NY : The Guilford Press. Martin Ginis, K. A., Eng, J. J., Arbour, K. P., Hartman, J W., & Phillips, S. M. (2005). Mind over muscle? Sex differences in the relationshi p between body image change and subjective and objecti ve physical changes follow ing a 12-week strength training program. Body Image, 2(4) 363 372. McCabe, M. P., & Ricciardelli, L. A. (2004). Body image dissatisfaction among males across the lifespan: A review of past literature. Journal of Psychosomatic R esearch, 56 675 685. McCabe, M.P., Ricciardelli, L.A., & Salmon, J. (2006). Evaluation of a prevention program to address body focus and negative affect among children. Journal of Health Psychology, 11(4), 589 598. McKay, M., & Fanning, P. (1987). Self-esteem Oakland, CA: New Harbinger. McLaren, L. & Kuh, D. (2004). Body dissatisfaction in midlife. Journal of Women and Aging, 16, 35 54. McVey, G.L., & Davis, R. (2002). A program to promote positive body image: A one year follow up evaluation. Journal of Early Adolescence, 22(1), 96 108. Murphy, J. G., Ducknick, J. J., Vuchinich, R. E., Davison, J. W., Karg, R. S., Olson, A. M., et al. (2001). Relative efficacy of a brief motivational intervention for college student drinkers. Psychology of Addictive Behaviors, 15, 373 379. Muth, J. L., & Cash, T. F. (1997). Body -image attitudes: What difference does gender make? Journal of Applied Social Psychology, 27, 1438 1452. Orwin, R.G. (1983). A fail -safe N for effect size in meta analysis. Journal of Educati onal Statistics, 8(2) 157 159. Overton, W.F. (1998). Developmental psychology: Philosophy, concepts, and methodology. In R.M. Lerner (Ed.), Handbook of child psychology: Theoretical models of human development (5th ed., Vol. 1, pp. 107 189). New York, N Y : Wiley. Pigott, T. (1994). Methods of handling missing data in research synthesis. In H. Cooper & L. Hedges (Eds.), The handbook of research synthesis (pp. 163 176). New York NY : Russell Sage Foundation. Prochaska, J.O., Velicer, W.F., Rossi, J.S., Go ldstein, M.G., Marcus, B.H., Rakowski, W., et al. (1994). Stages of change and decisional balance for 12 problem behaviors. Health Psychology, 13(1), 39 46.
115 Pue tz, T. W., O'Connor, P. J., & Dishman, R. K. (2006). Effects of chronic exercise on feelings of energy and fatigue: A quantitative synthesis. Psychological Bulletin, 132, 866 876. Ramirez, E. M., & Rosen, J.C. (2001). A comparison of weight control and weight control plus body image therapy for obese men and women. Journal of Consulting and Clinical Psychology, 69(3), 440 446. Rauvuori, A., Keski Rahkonen, A., Bulick, C. M., Rose, R. J., Rissanen, A., & Kaprio, J. (2006). Muscle dissatisfaction in young adult men. Clinical Practice and Epidemiology in Mental Health, 2 1 8. Rierdan, J., Koff, E., S tubbs, M.L. (1988). Gender, depression, and body image in early adolescents. The Journal of Early Adolescence, 8(2), 109 117. Roberts, A., Cash, T. F., Feingold, A., & Johnson, B. T. (2006). Are black white differences in females' body dissatisfaction decreasing? A meta analytic review. Journal of Consulting and Clinical Psychology, 74, 1121 1131. Rooney, B. L., & Murray, D. M. (1996). A meta analysis of smoking prevention programs after adjustment for errors in the unit of analysis. Health Education Quarterly, 23 48 64. Rosen, J.C., Reiter, J., & Orosan, P. (1995). Cognitive behavioral body image therapy for body dysmorphic disorder Journal of Consulting and Clinical Psychology, 63, 263 269. Rosenthal, D.A., Hoyt, W.T., Ferrin, J.M., Miller, S., & C ohen, N.D. (2006). Advanced methods in meta analytic research: Applications and implications for rehabilitation counseling research. Rehabilitation Counseling Bulleting, 49(4), 234 246. Rosenthal, R. (1979). The file drawer problem and tolerance for nul l results. Psychological Bulletin, 86 638 641. Rosenthal, R. (1991). Meta -Analysis: A review. Psychosomatic Medicine, 53, 247 271. Rosenthal, R. (1991). Meta -analytic procedures for social research. Newbury Park, CA: Sage. Rosenthal, R. & DiMatteo, M.R (2001). Meta analysis: Recent developments in quantitative methods for literature reviews. Annual Review of Psychology, 52, 59 82. Rothman, A.J. (2004) Is there nothing more practical than a good theory? Why innovations and advances in health behavior change will arise if interventions are used to test and refine theory. International Journal of Behavioral Nutrition and Physical Activity, 1, 11 Rothstein, H.R., Sutton, A.J., & Borenstein, M. (Eds.). (2005). Publication bias in meta analysis: Preventio n, assessment and adjustment Sussex UK : John Wiley and Sons.
116 Sallis, J. F., Calfas, K. J., Nichols, J. F., Sarkin, J. A., Johnson, M. F., Capa rosa, S., Thompson, S. & Alcaraz, J. E. (1999). Evaluation of a university course to promote physical activity. Project GRAD. Research Quarterly for Exercise and Sport, 70, 1 10. Silberstein, L.R., Striegel-Moore, R.H., Timko, C., & Rodin, J. (1988). Behavioral and psychological implications of body dissatisfaction: Do men and women differ? Sex Roles, 19, 219 232. Smolak, L. (2002). Body image development in children. In T. F. Cash and T. Pruzinsky (Eds.), Body image. A handbook of theory, research, and clinical practice (pp. 65 73). New York NY : Guilford Press. Sonstroem, R.J., Harlow,L.L., & Josephs, L. (1994). Exercise and self -esteem: Validity of model expansion and exercise associations. Journal of Sport and Exercise Psychology, 16, 29 42. Sonstroem, R. J., & Morgan, W. P. (1989). Exercise and self -esteem: Rationale and model. Medicine and Science in Spor ts and Exercise, 21 329 337. Sterne, J.A.C., & Egger, M. (2005). Regression methods to detect publication and other bias in meta analysis. In Rothstein, H.R., Sutton, A.J., & Borenstein, M. (Eds.). (2005). Publication bias in meta-analysis: Prevention, assessment and adjustment (pp. 99 110). Sussex UK : John Wiley and Sons. Stewart, T. M., & Williamson, D. A. (2004). Assessment of body image disturbances. In J. K. Thompson (ed.), Handbook of eating disorders and obesity (pp. 495 514). New York, NY: John Wiley. Stice, E., & Bearman, S. K. (2001). Bodyimage and eating disturbances prospectively predict increases in depressive symptoms in adolescent girls: A growth curve analysis. Developmental Psychology, 37, 597 607. Stice, E. & Shaw, H.E. (2002). Role of body dissatisfaction in the onset and maintenance of eating pathology: A synthesis of research findings. Journal of Psychosomatic Research, 53, 985 993. Stice, E. & Shaw, H. (2004). Eating disorder prevention programs: A meta analytic review. Psychological Bulletin, 130(2) 206 227. Stice, E., Shaw, H., & Marti, C. N. (2006). A meta analytic review of obesity prevention programs for children and adolescents: The skinny on interventions that work. Psychological Bulletin, 132, 667 691. Stice, E., Presne ll, K., & Spangler, D. (2002). Risk factors for binge eating onset: A prospective investigation. Health Psychology, 21, 131 138.
117 Strachan, M.D., & Cash, T.F. (2002). Self -help for a negative body image: A comparison of components of a cognitive behavioral program. Behavior Therapy, 22, 235 251. Striegel Moore, R.H., & Franko, D.L. (2002). Body image issues among girls and women In T. F. Cash & T. Pruzinsky (Eds.), Body image: A handbook of theory, research, and clinical practice (pp. 183 191 ). New York NY: Guilford Press Stunkard, A. J., Sorenson, T., & Schlusinger, F. (1983). Use of the Danish adoption registers for the study of obesity and thinness. In S. S. Kety, L. P. Rowland, R. L. Sidman, & S. W. Matthysse (Eds.), The genetics of neurological and psychological disorders (pp. 115 120). New York, NY : Raven Press. Suarez-Almazor, M. E., Belseck, E., Homik, J., Dorgan, M., & Ramos Remus, C. (2000). Identifying clinical trials in the medical literature with electronic databases: MEDLINE alone is not e nough. Control Clinic Trials, 21 476 487. Sutton, A. J., Abrams, K. R., Jones, D. R., Sheldon, T. A., & Song, F. (2000a). Missing data In Methods for meta-analysis in medical research (pp. 199 204). New York NY : John Wiley & Sons Taylor, A. H. (2000) Physical activity, anxiety, and stress. In S. J. H. Biddle, K. R. Fox, & S. H. Boutcher, (Eds.), Physical activity and psychological well -being (pp. 10 45). London, UK : Routledge. Taylor, A.H., & Fox, K.R. (2005). Effectiveness of a primary care exercis e referral intervention for changing physical self -perceptions over 9 months. Health Psychology 24 11 21. Thompson, J.K., Heinberg, L.J., Altabe, M., & Tantleff -Dunn, S. (1999). Exacting beauty. Theory, assessment, and treatment of body image disturbance Washington, DC: American Psychological Association. Thompson, J.K. & Spana, R.E. (1988). The adjustable light beam method for the assessment of size estimation accuracy: Description, psychometric, and normative data. International Journal of Eating Dis orders, 7(4) 521 526. Thompson, J.K., & Stice, E. (2001). Thinideal internalization: Mounting evidence for a new risk factor for body image disturbance and eating pathology. Current Directions in Psychological Science, 10(5), 181 183 Thompson, J.K., & Van Den Berg, P. (2002). Measuring body image attitudes among adolescents and adults In T. F. Cash & T. Pruzinksy (Eds), A handbook of theory, research, and clinical practice (pp. 142 154 ). New York NY : The Guilford Press. Thurstin, A.H. (1999). Behavi oral, physical, and psychological symptoms of eating disorders. In R. Lemberg & L. Cohn (Eds.), Eating disorders: A reference sourcebook (pp. 12 17). New York NY : Greenwood.
118 Tiggemann, M. (2004). Body image across the adult life span: Stability and chang e. Body Image, 1 29 41. Tiggemann, M., & Williamson, S. (2000). The effect of exercise on body satisfaction and self esteem as a function of gender and age. Sex Roles, 43, 119 127. Tobler, N. S., Roona, M. R., Ochshorn, P., Marshall, D. G., Streke, A. V ., & Stackploe, K. M. (2000). School -based adolescent drug prevention programs: 1998 meta analysis. Journal of Primary Prevention, 20, 275 336. Trikalinos, T.A. & Ioannidis, J.P.A. (2005). Assessing the evolution of effect sizes over time. In Rothstein, H .R., Sutton, A.J., & Borenstein, M. (Eds.). (2005). Publication bias in metaanalysis: Prevention, assessment and adjustment (pp. 241 259). Sussex UK : John Wiley and Sons. Vander Werf, E. (1992). Lack's clutch size hypothesis: an examination of the evide nce using meta analysis. Ecology. 73, 1699 1705. Wang, Z., Byrne, N, M., Kenardy, J. A., & Hills, A. P. (2005). Influences of ethnicity and socioeconomic status on the body dissatisfaction and eating behaviour of Australian children and adolescents. Eatin g Behaviors, 6, 23 33. Wiederman, M. W. ( 2002). Body image and sexual functioning. In T. F. Cash & T. Pruzinsky (Eds.), Body image: A handbook of theory, research, and clinical practice (pp. 287 294). New York NY : Guildford Press. Wildes, J.E., Emery, R .E., & Simons, A.D. (2001). The roles of ethnicity and culture in the development of eating disturbance and body dissatisfaction: A meta analytic review. Clinical Psychology Review, 21(4), 521 551. Williams, P.A., & Cash, T.F. (2001). Effects of a circu it weight training program on the body images of college students. International Journal of Eating Disorders, 30, 75 82. Wiltink, J., Dippel, A., Szczepanski, M., Thiede, R., Alt, C., & Beutel, M.E. (2007). Long -term weight loss maintenance after inpatien t psychotherapy of severely obese patients based on a randomized study: Predictors and maintaining factors of health behavior. Journal of Psychosomatic Research, 62, 691 698. Winzelberg, A.J., Eppstein, D., Eldredge, K.L., Wilfley, D., Dasmahapatra, R., Barr Taylor, C., et al. (2000). Effectiveness of an internet -based program for reducing risk factors for eating disorders. Journal of Consulting and Clinical Psychology, 68(2), 346 350. Wolf, F. M. (1986). Meta -analysis: Quantitative methods for research synthesis Beverly Hills, CA: Sage Publications.
119 Zabinski, M.F., Calfas, K.J., Gehrman, C.A., Wilfley, D.E., & Sallis, J.F. (2001). Effects of a physical activity intervention on body image in university seniors: Project GRAD. Annals of Behavioral Medic ine, 23(4), 247 252.
120 REFERENCE LIST OF STUDIES USED IN META-ANALYSIS Alfermann, D., & Stoll, O. (2000). Effects of physical exercise on self -concept and well -being. International Journal of Sport Psychology, 30, 47 65. Anderson, A. G., Murphy, M.H., Murtagh, E., & Nevill, A. (2006). An 8 -week randomized controlled trial on the effects of brisk walking, and brisk walking with abdominal electrical muscle stimulation on anthropometric, body c omposition, and self -perception measures in sedentary adult women. Psychology of Sport and Exercise, 7, 437 451. Annessi, J. J. (2005). Relations of body esteem factors with exercise session attendance in women initiating a physical activity program. Per ceptual and Motor Skills, 100, 995 1003. Asci, F. H. (2002). The effects of step dance on physical self -perception of female and male university students. International Journal of Sport Psychology, 33, 431 443. Asi, F.H. (2003). The effects of physica l fitness training on trait anxiety and physical self concept of female university students. Psychology of Sport and Exercise, 4(3), 255 264. Asci, F. H., Kin, A., & Kosar, N. (1998). Effects of participation in an 8 week aerobic dance and step aerobics program on physical self -perception and body image satisfaction. International Journal of Sport Psychology 29, 366 375. Barenholtz, D. E. (1995). The effects of an exercise program on the eating behavior, body image and self -esteem of adolescent girls D issertation Abstracts International, 56, 2853. ( AAT 9530049) Bartlewski, P.P., Van Raalte, J.L., & Brewer, B.W. (1996). Effects of aerobic exercise on the social physique anxiety and body esteem of female college students. Women in Sport and Physcial Activity Journal, 5(2), 49. Ben Schlomo, L. S., & Short, M. A. (1986). The effects of physical conditioning on selected dimensions of self -concept in sedentary females. Occupational Therapy in Mental Health, 5 27 46. Bowden, R.G., Rust, D.M., Dunsmore, S., & Briggs, J. (2005). Changes in social physique anxiety during sixteen -week physical activity courses. Psychological Reports, 96(3), 690 692. Brown, D.R., & Harrison, J. M. (1986). The effects of a strength training program on the strength and self -c oncept of two female age groups. Research Quarterly for Exercise and Sport, 57, 315 320. Brown, D.R., Wang, Y., Hinkle, R., Webber, L., Ahlquist, L., Puleo, E., Ward, A., & Rippe, J. (1991). The effects of four strength training program on body cathexis, physical estimation, and self -esteem. Medicine and Science in Sports and Exercise 23 S82.
121 Brown, D.R., Wang, Y., Ward, A., Ebbeling, C. B., Fortlage, L., Puleo, E., Benson, H., & Rippe, J. M., (1995). Chronic psychological effects of exercise and exer cise plus cognitive strategies. Medicine and Science in Sports and Exercise, 27, 765 775. Brown, E.Y., Morrow, J.R., & Livingston, S.M. (1982). Self -concept changes in women as a result of training. Journal of Sport Psychology, 4, 354 363. Cocklin, J. C (1989). The effects of physical fitness and body cathexis on self -concept change in women after aerobic conditioning. Doctoral dissertation, Oklahoma State University, Dissertation Abstracts International, 49 (10) 4594B. Collingwood, T. R. (1972). Th e effects of physical training upon behavior and self attitudes. Journal of Clinical Psychology, 28, 583 585. DAmato, K. R. (1981). The effects of cardiovascular efficiency on psychological and physiological variables in women. Unpublished doctoral diss ertation, California School of Professional Psychology, Fresno. Daley, A. J., Copeland, R. J., Wright, N. P., Roalfe, A., & Wales, J. K. (2006). Exercise therapy as a treatment for psychopathologic conditions in obese and morbidly obese adolescents: A ran domized, controlled trial. Pediatrics, 118, 2126 2134. DiLorenzo, T. M., Bargman, E. P., Stucky Ropp, R., Brassington, G. S., Frensch, P. A., & LaFontaine, T. (1999). Long-term effects of aerobic exercise on psychological outcomes. Preventive Medicine, 2 8, 75 85. Eliot, A. B. (1998). Enhancing womens body image: A comparison of treatment interventions Unpublished doctoral dissertation, University of New Mexico. Evans, S., Newton, R., & Higgins, S. (2005). Nutritional intervention to prevent weight gain in patients commenced on olanzapine: A randomized controlled trial. The Australian and New Zealand Journal of Psychiatry, 39, 479 486. Finkenberg, M.E., DiNucci, J.M., & McCune, S.L. (1993). Body esteem and enrollment in classes with different level s of physical activity. Perceptual and Motor Skills, 76(3), 783 792. Fisher, E., & Thompson, J. K. (1994). A comparative evaluation of cognitive -behavioral therapy vs exercise therapy for the treatment of body image disturbance: Preliminary findings. Beh avior Modification, 18, 171 185. Ford, H. T., Puckett, J. R., Blessing, D. L., & Tucker, L. A. (1989). Effects of selected physical activities on health related fitness and psychological well -being. Psychological Reports, 64, 203 208.
122 Ford, H.T., Pucke tt, J.R., Reeve, T.G., & Lafavi, R.G. (1991). Effects of selected physical activities on global self -concept and body -cathexis scores. Psychological Reports, 68, 1339 1343. Fossati, M., Amati, F., Painot, M., Reiner, M., Haenni, C., & Golay, A. (2004). C ognitive behavioral therapy with simultaneous nutritional and physical activity education in obese patients with binge eating disorder. Eating and Weight Disorders, 9, 134 138. Gehrman, C. A., Hovell, M. F., Sallis, J. F., & Keating, K. (2006). The effect s of a physical activity and nutrition intervention on body dissatisfaction, drive for thinness, and weight concerns in pre adolescents. Body Image, 3, 345 351. Gilman, J. C. (1996). The effects of weight training and aerobic exercise on body image, anxie ty, locus of control, and self -esteem Unpublished doctoral dissertation, University of Southern California. Hase, J. A. (1995). Weight training effects on body image, self -esteem and risk factors for an eating disorder Unpublished doctoral dissertation, California School of Professional Psychology. Henry, R. N., Anshel, M. H., & Michael, T. (2006). Effects of aerobic and circuit training on fitness and body image among women. Journal of Sport Behavior 29, 281 303. Hilyer, J. C., & Mitchell, W. (1979 ). Effect of systematic physical fitness training combines with counseling on the self -concept of college students Journal of Counseling Psychology, 26(5), 427 436. Huang, J. S., Norman, G. J., Zabinski, M. F., Calfas, K., & Patrick, K. (2007). Body ima ge and self -esteem among adolescents undergoing an intervention targeting dietary and physical activity behaviors. Journal of Adolescent Health, 40, 245 251. King, A. C., Taylor, C. B., Haskell, W. L., & DeBusk, R. F. (1989). Influence of regular aerobic exercise on psychological health: A randomized, controlled trial of healthy middle aged adults. Health Psychology, 8, 305 324. Lindwall, M., & Lindgren, E. (2005). The effects of a 6 -month exercise intervention programme on physical self -perceptions an d social physique anxiety in non-physically active adolescent Swedish girls. Psychology of Sport and Exercise, 6(6) 643 658. Marquez -Sterling, S., Perry, A. C., Kaplan, T. A., Halberstein, R. A., & Signorile, J. F. (2000). Physical and psychological cha nges with vigorous exercise in sedentary primigravidae. Medicine and Science in Sports and Exercise, 32, 58 62. McCabe, M.P., Ricciardelli, L.A., & Salmon, J. (2006). Evaluation of a prevention program to address body focus and negative affect among chil dren. Journal of Health Psychology, 11(4), 589 598.
123 Mock, V., Burke, M. B., Sheehan, P., Creaton, E. M., Winningham, M. L. & Liebman, M. (1994). A nursing rehabilitation program for women with breast cancer receiving adjuvant chemotherapy. Oncology Nursi ng Forum, 21, 899 907. OLoughlin, J., Paradia, G., Meshefedjian, G., & Kishchuck. N. (1998). Evaluation of an 8 -week mailed healthy -weight intervention. Preventive Medicine, 27 288 295. Perry, A.C., Rosenblatt, E.S., Kempner, L., Feldman, B.B.. Paol ercio, M.A., & Van Bemden, A.L. (2002). The effects of an exercise physiology program on physical fitness variables, body satisfaction, and physiology knowledge. Journal of Strength and Conditioning Research, 16(2), 219 226. Pinto, B. M., Clark, M. M., Ma ruyama, N. C. et al. (2003). Psychological and fitness changes associated with exercise participation among women with breast cancer. Psycho oncology, 12, 118 126. Pinto, B.M., Frierson, G.M., Rabin, C., Trunzo, J.J., & Marcus, B.H. (2005). Home -based ph ysical activity intervention for breast cancer patients. Journal of Clinical Oncology, 23, 3577 3587. Sandel, S. L., Judge, J. O., Landry, N., Faria, L., Ouellette, R., & Majczak, M. (2005). Dance and movement program improves quality -of life measures in breast cancer survivors. Cancer Nursing, 28, 301 309. Scanlon, P. A. (1991). Effects of a five week aerobic exercise class on the body image of adolescent girls Unpublished Masters Thesis, The American University. Shaw, J. M., Ebbeck, V., & Snow, C. M. (2000). Body composition and physical self -concept in older women. Journal of Women & Aging, 12, 59 75. Sherblom, P. R., & Rust, D. M. (2004). Body image, figure rating, and body mass index of girls enrolled in health, physical education, or athletics classes. Perceptual and Motor Skills, 99, 473 482. Smith, S. A., & Michel, Y. (2006). A pilot study on the effects of aquatic exercises on discomfort of pregnancy. Journal of Obstetric, Gynecologic, & Neonatal Nursing, 35, 315 323. Sorensen, M., Ander ssen, S., Hjeran, I., Holme, I., & Ursin, H. (1997). Exercise and diet interventions improve perceptions of self in middle aged adults. Scandinavian Journal of Medicine and Science in Sports, 7, 312 320. Stein, P. N. (1989). The differential effects of a erobic and nonaerobic exercise on affect and self concept. Unpublished doctoral dissertation, Hofstra University.
124 Stock, S., Miranda, C., Evans, S., Plessis, S., Ridley, J., Yeh, S., & Chanoine, J. (2007). Healthy buddies: A novel, peer led health promoti on program for the prevention of obesity and eating disorders in children in elementary school. Pediatrics, 120, 1059 1068. Stoll, O., & Alfermann, D. (2002). Effects of physical exercise on resources evaluation, body self -concept and well -being among o lder adults. Anxiety, Stress & Coping: An International Journal, 15(3) 311 319. Sundgot Borgen, J., Rosenvinge, J.H., Bahr, R., & Schneider, L.S. (2002). The effect of exercise, cognitive therapy, and nutritional counseling in treating bulimia nervosa. Medicine & Science in Sports & Exercise, 34(2), 190 195. Talbot, H. M., & Taylor, A. H. (1998). Changes in physical self perceptions: Findings from a randomized controlled study of a GP exercise referral scheme. Journal of Sports Sciences, 16 105 106. Taylor, A. H., & Fox, K. R. (2005). Effectiveness of a primary care exercise referral intervention for changing physical self -perceptions over 9 months. Health Psychology, 24, 11 21. Tucker, L. A. (1987). Effect of weight training on body attitudes: Who benefits most? Journal of Sports Medicine and Physical Fitness, 27, 70 78. Williams, P.A., & Cash, T.F. (2001). Effects of a circuit weight training program on the body images of college students. International Journal of Eating Disorders, 30, 75 82. Zabinski, M.F., Calfas, K.J., Gehrman, C.A., Wilfley, D.E., & Sallis, J.F. (2001). Effects of a physical activity intervention on body image in university seniors: Project GRAD. Annals of Behavioral Medicine 23(4), 247 252.
125 BIOGRAPHICAL SKETCH Anna w as born in Germany and ra ised in Satellite Beach, Florida. After attending Satellite High School s he obtained her undergraduate degree in psychology from the University of Florida. Through volunteer research in the exercise psychology lab she became intere sted in sports and exercise psychology and decided to pursue her masters in the field. After completing her masters she plans to attend the University of Florida pursuing a doctorate degree in physical therapy.