SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods and Procedures
  5. Results
  6. Discussion
  7. SUPPLEMENTARY MATERIAL
  8. ACKNOWLEDGEMENT
  9. DISCLOSURE
  10. References
  11. Supporting Information

The prevalence of obesity is similar for men (32.2%) and women (35.5%). It has been assumed that lifestyle weight loss interventions have been developed and tested in predominately female samples, but this has not been systematically investigated. The aim of this review was to investigate total and ethnic male inclusion in randomized controlled trials of lifestyle interventions. PUBMED, MEDLINE, and PSYCHINFO were searched for randomized controlled trials of lifestyle weight loss interventions (N = 244 studies with a total of 95,207 participants) published in the last 10 years (1999–2009). A trial must be in English, included weight loss as an outcome, and tested a dietary, exercise, and/or other behavioral intervention for weight loss. Results revealed samples were on average 27% male vs. 73% female (P < 0.001). Trials recruiting a diseased sample included a larger proportion of males than those not targeting a disease (35% vs. 21%; P < 0.001). About 32% of trials used exclusively female samples, whereas only 5% used exclusively male samples (P < 0.001). No studies in the past 10 years specifically targeted minority males. Ethnic males identified composed 1.8% of total participants in US studies. Only 24% of studies that underrepresented males provided a reason. Males, especially ethnic males, are underrepresented in lifestyle weight loss trials.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods and Procedures
  5. Results
  6. Discussion
  7. SUPPLEMENTARY MATERIAL
  8. ACKNOWLEDGEMENT
  9. DISCLOSURE
  10. References
  11. Supporting Information

The most recent obesity prevalence estimates are 32.2% in men and 35.5% in women (1). Although the increasing obesity trend has leveled off for women in the past 10 years and for men in the past 3 years, rates remain high in the entire population, regardless of sex or ethnicity (1). Millions of Americans stand to benefit from effective weight loss treatment. Treatment options include lifestyle interventions, pharmacology, and surgery. Effective pharmacological management of weight is still elusive and only a small segment of the obese population is eligible for surgical approaches. Consequently, a large literature has developed for lifestyle weight loss interventions, which typically includes nutrition and physical activity counseling and behavioral modification. The Diabetes Prevention Program was the first large study to show that a lifestyle weight loss intervention has a greater impact on clinical endpoints than placebo or medication (2). In 2003, the USPSTF assigned intensive lifestyle weight loss interventions a B grade, with fair to good evidence for modest, sustained weight loss (3).

Proceedings from the 2008 Stock Conference on sex differences in obesity suggest that the literature on lifestyle weight loss interventions is largely based on female samples (4). Although this impression may be the consensus among field leaders (e.g., 4,5), the representation of men in lifestyle weight loss intervention trials, and minority men for that matter, has never been systematically investigated. Under representation of men in the weight loss literature would be a concern because men are affected by obesity to a similar degree as women, and may be even more vulnerable to certain obesity-related diseases such as cardiovascular disease (4,6,7). Lifestyle intervention is one of very few effective treatment options. To the extent that interventions are designed and tested in women, studies and resulting clinical programs might not attract male participation or generalize to men. Consequently, our knowledge base would be insufficient about how to assist men in obesity management via lifestyle changes (4).

Effective approaches to lifestyle interventions for women may not apply to men, especially minority men, given evidence for sex and cultural differences in physiological responses to physical activity, influence of stress hormones on weight, association between depression and obesity, and norms related to body image (4). Men might also have unique food and physical activity preferences. All of these factors have implications for lifestyle interventions. For example, evidence suggesting physical activity has greater impact on body weight and fat loss in males than females (8,9) might lend to lifestyle interventions for men with a greater focus on physical activity than diet.

The present investigation had five aims. The first aim was to determine the representation of men vs. women in randomized clinical trial samples for lifestyle weight loss interventions. The second aim was to determine whether men were more or less represented in trials with certain characteristics such as (i) United States vs. international, (ii) targeting a specific disease vs. obesity in general, (iii) group vs. individual vs. self-guided interventions, and (iv) diet plus exercise vs. diet only vs. exercise only interventions. The third aim was to determine the representation of ethnic minority males in trials. The fourth aim was to tabulate the reasons provided in trials for sex distribution, and finally the fifth aim was to examine whether a sex difference in weight loss was tested.

Methods and Procedures

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods and Procedures
  5. Results
  6. Discussion
  7. SUPPLEMENTARY MATERIAL
  8. ACKNOWLEDGEMENT
  9. DISCLOSURE
  10. References
  11. Supporting Information

Data sources

Three authors (S.P., J.S.B., and K.L.S.) conducted literature searches for randomized clinical trials for behavioral weight loss treatment conducted in the last 10 years (1999–2009) using PUBMED, MEDLINE, and PSYCHINFO search engines. Search terms included “weight,” “loss,” “randomized,” “trial,” and “overweight” or “obese” in the title, abstract, or keywords of the article. Each author conducted a search of a specific time period (e.g., 1999–2001, 2002–2005, and 2006–2009) and initially screened the article for exclusion based on the contents of the title and abstract. If no reasons to exclude were evident in the abstract, the full article was obtained and inclusion and exclusion criteria reviewed. Only published journal articles were included in the search.

Study selection criteria

Eligibility criteria included: (i) randomized trial, (ii) weight was an outcome (primary or secondary), (iii) intervention included instruction on diet, exercise, or both, (iv) study population was adults 18 years and older. Exclusion criteria included (i) primary goal was to test the efficacy of a pharmacological or surgical intervention, (ii) studies where all food is provided to participants, (iii) the intervention goal was weight gain prevention or weight loss maintenance, (iv) published in language other than English.

The initial search generated 514 papers. Of those, 156 were excluded after reviewing the abstract, leaving 358 articles that were distributed to coders. Coders then excluded 114 that were discovered to not meet inclusion criteria during the coding process (see Figure 1). If an article was a secondary analysis of a parent trial published elsewhere, the parent trial was reviewed if it met the inclusion criteria. A total of 244 articles were eligible and subsequently coded (see Supplementary Appendix online).

image

Figure 1. Flow diagram of article inclusion and exclusion.

Download figure to PowerPoint

Development of coding criteria

Three authors (S.P., J.S.B., and K.L.S.) initially generated the coding variables and coding form. After the creation of the initial coding form, five of the authors (S.P., K.L.S., J.S.B., J.O., and J.L.) coded five articles, calculated inter-rater reliability, discussed discrepancies, and made modifications to items that generated a >20% rate of discrepancies. This process was repeated seven times until 100% agreement was reached among the five coders and the final coding form was set. A reference document was created for use by coders that described the specific rules of coding for each item. Inter-rater agreement was calculated for each item across coders and reported by item below. Inter-rater agreement was also calculated for each trial and the mean inter-rater agreement across trials was 87%.

Coding variables

Sample size. Coders recorded the total randomized sample size as reported in the article. Inter-rater agreement for this item was 86.5%.

Sex distribution. The total number of males and females was also recorded, if reported, and it was noted if the article did not provide sex information. When sex information was provided but not based on the total randomized sample, this was noted. Inter-rater agreement for this item was 88.1%.

Justification for sex distribution. Articles were searched for a justification for the sex distribution. If authors made an attempt to explain the sex distribution, they were coded as having done so and the specific justification provided was recorded by the coder verbatim. Inter-rater agreement was not calculated for this item because it was open-ended.

Ethnic distribution of the sample. For studies conducted in the United States, coders first indicated whether or not the ethnic distribution of the sample was reported. If it was reported, the coder then recorded the total number and percent of each ethnic group (white, African American, Latino, Asian, native Hawaiian/Pacific Islander, Native American, or other) that were included in the sample. If provided, the coder also recorded the sex distribution across ethnic groups. If the ethnic distribution reported in the article was not on the total randomized sample, this was noted. Studies conducted outside of the United States were coded as “international” and ethnic distribution was not coded given the variability in ethnic categories across countries. Inter-rater agreement for this item was 91.4%.

Recruited sample with disease vs. obesity in general. Whether recruitment specifically targeted people with a medical condition or simply obesity in general was coded because many medical conditions are sex specific which could explain the sex distribution. Samples were characterized as obese in general if inclusion criteria did not require a medical condition or were characterized as diseased if they required participants to have a specific medical condition (e.g., type 2 diabetes, breast cancer, and prostate cancer). Inter-rater agreement for this item was 95.5%.

Intervention format. The intervention delivery format was coded into one of the following categories: individual, group, couples, mail, phone, or e-mail/internet. Inter-rater agreement for this item was 67.6%, lower than other items because many trials used multiple formats some of which were low in frequency (e.g., mailings). To increase the reliability of this item, the six categories were collapsed into three. An intervention was considered individual if it was delivered entirely on a one-on-one basis. An intervention was considered group if it included group meetings, even if individual meetings or other contacts were planned. An intervention was considered mail, phone, and e-mail/internet if entirely delivered by that modality. Because, we hypothesized that group-based interventions might not be preferred by men, we removed group-based interventions from other intervention format categories (e.g., internet) so that those categories would reflect that format unaffected by groups. For example, if an intervention included both internet and group contacts, it was considered as group-based intervention.

Intervention type. Studies were coded as to whether the intervention targeted diet, exercise, or both. Inter-rater agreement for this item was 93%.

Data extraction. Coders (authors S.P, J.S.B., K.L.S., J.O., J.L., and M.C.W.) were randomly assigned to pairs each week throughout the data extraction phase of the study. Each pair was given five articles to code. Coders compared their extracted data for each article, and discrepancies were recorded for reliability analyses. Discrepancies were reconciled between coders and when not possible, consensus was obtained in the entire group of coders.

Statistical analysis. Means for quantitative variables and percentages for categorical variables were calculated. Trial characteristics are described in Table 1. Comparisons were made using either tests for proportions, χ2, or F-tests with statistical significance set at a P < 0.05 (two-tailed). Statistical analyses were conducted using SPSS version 17.0 (SPSS, Chicago, IL).

Table 1.  Characteristics of participants across 244 trials
inline image

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods and Procedures
  5. Results
  6. Discussion
  7. SUPPLEMENTARY MATERIAL
  8. ACKNOWLEDGEMENT
  9. DISCLOSURE
  10. References
  11. Supporting Information

Male representation

Of 244 articles, 237 (97%) reported the sex of the sample. Samples were on average 27% male vs. 73% female (z(236) = −0.83; P = 0.00). About two-thirds (63%; n = 150) included both males and females, 32% (n = 75) recruited exclusively female samples, but only 5% (n = 12) recruited exclusively male samples (χ2(1) = 45.62; P <.001). US trials had a significantly smaller mean proportion of males than international trials (21% vs. 34%; F(236) = 11.56; P < 0.001; d = 0.45). Trials recruiting participants with a specific medical condition (e.g., type 2 diabetes, hypertension) included a significantly larger proportion of males than those with no specific disease criteria (36% vs. 22%; F(236) = 14.87; P < 0.001, d = 0.53).

Reasons for under representation of males

Among the 165 studies that underrepresented males (defined as <40% male), 39 (24%) provided an explanation for the sex distribution. Of these 39 studies, 34% justified exclusion or disproportionate representation of males due to focus on a female-specific disease (e.g., breast cancer patients), 13% stated it was based on interest in postmenopausal females, 13% stated that it was necessary to exclude men to enhance power, 10% focused on ethnic minority women, 8% stated that recruitment methods did not attract many men, 8% stated that previous research has excluded women, 5% stated that women are disproportionately affected by obesity, 3% stated that the sample was representative of adults with type 2 diabetes, 3% stated that women are more likely to diet, 3% focused on women on hormone replacement therapy, and 3% stated that men were excluded because knee osteoarthritis disproportionately affects women.

Intervention format

Most interventions were delivered either via group (47%) or individual meetings (46%), with a small number that were delivered solely via mail, e-mail, or the internet (4%), and 3% not reporting format. No statistically significant differences appeared for proportion of males in samples by intervention format, however, the trend was toward lower representation of males in group (24%) compared to individual (29%) or mail/e-mail/internet (34%) (F(229) = 1.52; P = 0.22).

Intervention type

Most interventions included both diet and exercise (67%), whereas 28% included diet only and 5% included exercise only. The proportion of men did not vary by intervention type (F(235) = 1.24; P = 0.29). Men composed 26, 30, and 17% of diet/exercise, diet only, and exercise only trial samples, respectively.

Ethnic male representation

Of the 140 U.S.-based studies (n = 79,369 participants), 96 (69%) reported the ethnicity of the sample, but only 88 of the 96 included ethnic minorities in the sample (which means the remaining eight were 100% white). In the 88 studies with ethnic minorities in the sample, 29 (33%) were all-female samples, 0% were all-male samples, and 59 (67%) included both males and females. Of the 59 studies that included both ethnic minorities and males, only three reported ethnicity by sex. All three of these studies exclusively recruited a minority group (no whites), two recruited African Americans and one recruited Hispanics. To obtain data on the remaining 56, we first searched for published preliminary reports of the trial by reviewing the reference sections of each trial. If data could not be obtained via a preliminary report, we then searched for published secondary reports of the trial using cited reference search. If data could not be obtained via a secondary report, we contacted the corresponding authors to retrieve ethnicity by sex information. Using these procedures, we received complete ethnicity by sex data for 24 studies, which increased the total number of studies with complete ethnicity by sex data to 27. These 27 studies included a total of 15,356 participants, of which 5,284 (34%) were male. Of 15,356 total participants, 3,377 (22%) were white males, 595 (4%) were African-American males, 546 (4%) were Latino males, 117 (1%) were Asian males, 0 (0%) were Native Hawaiian/Pacific Islander males, 81 (0.5%) were Native American males, and 77 (0.5%) were other. In sum, of the 79,369 study participants in U.S. studies, we could only confirm with certainty the presence of 1,416 ethnic males (from 27 of 59 studies that included males and ethnic minorities), which is only 1.8% of the total number of participants in U.S. studies. It is both possible and likely that in the 32 other studies that included males and ethnic minorities, some participants were both male and of ethnic origin. It is also possible that in the 44 studies that did not report ethnicity data, some participants were minority males.

Sex differences in outcomes

Only 19 studies tested sex differences in outcomes. Six of these reported a significant interaction of intervention condition by sex. In three of these studies, men lost more weight than women in the intervention condition. In the other three, differences were found within sex between conditions (e.g., women did better in one condition than women in the other condition).

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods and Procedures
  5. Results
  6. Discussion
  7. SUPPLEMENTARY MATERIAL
  8. ACKNOWLEDGEMENT
  9. DISCLOSURE
  10. References
  11. Supporting Information

Findings revealed that males are underrepresented in lifestyle weight loss interventions, with the average sample being 27% male. In U.S. trials, the representation of males was particularly low at 22%. Very little information exists for why men are underrepresented in these trials. Only 24% of trials that under represented men (i.e., samples <40% male) provided an explanation for the sex distribution. About 1/3 excluded men because of focus on a predominantly female health condition (e.g., breast cancer), yet far fewer trials (5%) focused exclusively on men or male-specific health conditions (e.g., prostate cancer).

One possible reason for male under representation is that men are not as motivated to lose weight as women. The perception of overweight differs between the sexes, with overweight men being less likely to perceive themselves as overweight relative to women with the same BMI (5,10). Social norms about obesity are also different between the sexes. Women experience more societal pressure to lose weight, which may contribute to their overrepresentation in weight loss trials (11). The sex difference in social norms about weight might vary across different countries and cultures given that international studies tended to include a higher proportion of men than US studies. Because men were better represented in studies recruiting people with a medical condition, illness may increase weight loss motivation among men. Research identifying factors influencing motivation to lose weight in men would help to inform interventions to increase participation in formal interventions.

Another possibility is that men may be less interested in seeking outside help for their weight than women. Men may have greater self-efficacy to lose weight on their own, however, obesity prevalence estimates would suggest that men are not more successful at managing their weight than women. Alternatively, men may not be interested in the type of interventions put forth in lifestyle weight loss studies. The format of most interventions was group- or individual-based counseling which may or may not be preferred by men. Group interventions had the lowest representation of men. Self-guided interventions (i.e., mail, e-mail, and internet) had the highest representation of men, although this difference was not statistically significant. Nonsignificance is possibly due to the small number of interventions that used this format, but this trend might suggest that men are more comfortable with self-guided interventions as opposed to counselor-guided interventions. Only one intervention used a couple's intervention in spite of research suggesting that partner support has an important influence on men's health behaviors (12) and that men are not typically the primary grocery shoppers or cooks in their households (13). Couples-based lifestyle interventions should be explored further as a means to engage men in structured treatment.

The type of weight loss intervention (diet, exercise, or both) did not influence male representation. Because men have a greater tendency to lose weight via exercise (8,9), it was expected that men might have greater representation in exercise only interventions. However, a very small proportion of weight loss trials utilized exercise only interventions (i.e., 5%) and the trend was toward these trials having the lowest proportion of males relative to diet only and diet plus exercise interventions. This is surprising in light of findings of a 2003 trial that men lost clinically significant weight from exercise and no dietary intervention, but women did not (14). Exercise intervention could be a promising approach to weight loss treatment in men. A much larger literature exists for exercise trials than was reviewed here, but only those that included weight loss as an outcome were included in this review. Qualitative research methods may be helpful to determine what intervention features men find interesting and helpful. Also, randomized trials are needed comparing different types of interventions in samples that include enough males to test efficacy in men specifically.

Our search did not produce a single study in the last 10 years that exclusively recruited ethnic men. Ethnic males were also not well-identified in studies generally as evidenced by a lack of reporting of sex by ethnic group. We could only confirm the presence of 1,416 ethnic males (1.8%) from a total pool of 95,207 participants in the identified trials. This number is an underestimate because 32 of the 59 trials that reported they included males and ethnic minorities, did not specify how many of the ethnic minority participants were male. The representation of ethnic men in the lifestyle intervention literature would still seem to be quite low. Although obesity is less prevalent in black males compared to black females (37.3% vs. 49.6%), and in Hispanic males compared to Hispanic females (34% vs. 43%), obesity rates of black and Hispanic males are as high as white females (i.e., 33%) who comprise the majority of lifestyle intervention samples (1). Ethnic males are extremely under represented in lifestyle intervention samples relative to the degree to which they are impacted by obesity. How to treat ethnic men for their obesity is a completely uninvestigated area of research, and the quality of clinical care for ethnic men likely reflects this lack of knowledge. Kumanyika (5) has called for both minority-specific studies as well as subgroup analyses in multiethnic studies of lifestyle interventions to increase our understanding of what works in minority groups. Trials should report the breakdown of sex by ethnicity to make apparent the representativeness to each specific group. Developmental work on the lifestyle habits of minority men, environmental and motivational factors influencing weight, and intervention preferences would be extremely valuable to designing appropriate interventions.

This review has some limitations. First, the search did not include pharmacological and surgical interventions. Pharmacological interventions for obesity produce only modest effects and are typically recommended as adjunctive to lifestyle interventions (3). Only a small proportion of obese adults are clinically eligible for bariatric surgery, but studies have revealed that women have twice as many bariatric surgeries as men (15). Another limitation is that the representation of ethnic males in lifestyle intervention trials could not be quantified because data were often not reported or attainable. Conclusions regarding ethnic men are limited for this reason. Furthermore, sex and ethnic differences may be tested and reported in separate publications than the main trial outcome paper which this review would not have captured. However, the under representation of males, and ethnic males especially, would seem to limit power to test sex and ethnic differences in most trials.

White females are under the greatest societal scrutiny to be thin (11) and as a consequence, their willingness to volunteer for lifestyle weight loss intervention trials has made them a convenient target of research. This phenomenon may have essentially sculpted the research literature on lifestyle weight loss interventions strongly toward white females at the expense of men who are equally affected by obesity. According to Public Law 103–43, the NIH Policy on Inclusion of Women and Minorities, the NIH must: “ensure that women and members of minorities and their subpopulations are included in all human subject research; for phase III clinical trials, ensure that women and minorities and their subpopulations must be included such that valid analyses of differences in intervention effect can be accomplished; not allow cost as an acceptable reason for excluding these groups; and, initiate programs and support for outreach efforts to recruit these groups into clinical studies.” The current status of the lifestyle weight loss intervention literature is such that men are not participating in rates that allow for valid analyses of differential intervention effects. Outreach efforts to recruit men, and especially minority men are insufficient.

Trials typically do not report details on recruitment strategies making it impossible to know to what extent targeted efforts were made to recruit males and minorities. Targeted recruitment efforts may be necessary to increase the representation of males, especially minority males, in weight loss trials, although it is not entirely clear how to do this. The literature provides little guidance on effective recruitment strategies for subgroups who may be reluctant to volunteer for studies. Reports of successful targeted recruitment approaches would assist investigators attempting to achieve balanced and representative samples.

The under representation of men in lifestyle weight loss interventions is a significant concern given that men have higher rates of cardiovascular morbidity and mortality (16) and shorter lifespan than women (17). Men stand to benefit greatly from healthier lifestyle choices and weight management. More intensive efforts are needed to engage men, especially minority men, into lifestyle interventions as well as to design interventions that suit their needs and preferences.

ACKNOWLEDGEMENT

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods and Procedures
  5. Results
  6. Discussion
  7. SUPPLEMENTARY MATERIAL
  8. ACKNOWLEDGEMENT
  9. DISCLOSURE
  10. References
  11. Supporting Information

S.P. receives funding via NIH grant R01MH078012 and K.L.S. receives funding from NIH grant R34MH086678. All authors had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods and Procedures
  5. Results
  6. Discussion
  7. SUPPLEMENTARY MATERIAL
  8. ACKNOWLEDGEMENT
  9. DISCLOSURE
  10. References
  11. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods and Procedures
  5. Results
  6. Discussion
  7. SUPPLEMENTARY MATERIAL
  8. ACKNOWLEDGEMENT
  9. DISCLOSURE
  10. References
  11. Supporting Information

supporting Information

FilenameFormatSizeDescription
oby_2411_sm_1.doc111KSupporting info item

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.