Pretreatment Weight Change Is Associated With Obesity Treatment Outcomes
Clinical experience suggests some individuals begin obesity treatment weighing more than they did at pretreatment assessment. Weight fluctuations between baseline screening and the first treatment session were examined among individuals enrolling in a group behavioral obesity treatment outcome study. Participants (N = 480, 94% female; 28% African American; M BMI = 35.7) were classified into those who started treatment having gained weight (≥+1.15% above screening weight), lost weight (≤−1.15% below screening weight) or remained weight stable. The majority of participants were weight-stable (61%) during the waiting period, but 23% lost weight (−2.36 ± 1.26 kg) and 16% gained weight (+2.11 ± 1.04 kg) between baseline screening and initiating treatment. Those who lost during the pretreatment period went on to have the greatest losses at 6-months (−8.9 ± 4.9 kg), with significantly greater weight losses than either the weight-stable (−6.1 ± 5.8 kg) or the weight-gain (−5.7 ± 5.8 kg) groups. Further, those who lost weight during the waiting period went on to attend a significantly higher proportion of treatment sessions and submitted more self-monitoring diaries than those who gained weight and those who stayed weight stable while waiting. Thus, pretreatment weight change was associated with treatment outcomes and may be relevant for research screening. Further, pretreatment weight change may be a clinical marker for likely success in behavioral weight control and as such warrants additional investigation to inform potential methods for enhancing outcomes for individuals in obesity treatment.
Behavioral obesity treatment programs generally obtain 7–9% weight loss at 6-months (1,2,3). These average weight losses mask significant variability, with standard deviations often as large as the mean weight losses achieved. Clearly, some individuals are very successful and others are less successful. Investigators have sought early indicators of who might be more successful within behavioral programs and who might not (1,4,5). Early identification of individuals who are likely to lose at the lower end of the weight loss distribution offers the potential for intervening in hopes of enhancing the weight losses of these individuals. Studies that have examined methods of augmenting standard behavioral intervention for individuals who were having difficulty achieving weight loss targets have shown some promising initial results (6,7). Therefore, knowledge of earlier indicators of those who are likely to struggle might be helpful in enhancing overall mean outcomes.
Clinical experience suggests that there are individuals who begin treatment sessions at a higher weight than their initial pretreatment assessment. These individuals often anecdotally report that they engaged in overeating, particularly eating high calorie foods they thought would be “off limits” once the weight management program had begun. With this overeating, they gained weight before initiating the program. This “last supper effect” has been examined with respect to emotions associated with overeating among binge eaters (8) and in individuals awaiting bariatric surgery (9), but has not been studied in a behavioral obesity treatment program. Therefore, it is unknown what proportion of individuals gain weight before starting a lifestyle weight loss program and whether there are any associations between weight gain “before really starting the program” and eventual weight loss. The current study examined pretreatment weight change among participants in a group-based behavioral lifestyle weight loss research program to determine whether any relationship might exist between pretreatment weight change and 6-month weight loss outcomes.
Methods and Procedures
Overweight and obese adults volunteering to participate in a randomized clinical trial examining different modalities of delivering a behavioral lifestyle weight loss treatment (10) were recruited in two clinical centers by advertisements placed in local newspapers, emailed notices, community events and word-of-mouth. To be eligible, individuals were required to be over 18 years of age, be overweight or obese as defined by a BMI (weight (kg)/height (m2)) between 25 and 50, be generally healthy and able to walk for exercise and to have access to a computer with an Internet connection. Individuals were excluded if they reported pregnancy in the previous 6-months or current breast feeding, a history of major medical problems for which weight loss was contraindicated, were currently engaged in other weight loss treatment, had a history of gastric bypass surgery, or reported recent significant weight loss. The study was approved by the Committee on Human Research in the Behavioral Sciences at the University of Vermont and the institutional review board at the University of Arkansas for the Medical Sciences.
Participants went through a series of screening visits to qualify for the randomized controlled trial (10). Initially, all volunteers logged on to a study recruitment website to provide basic entry qualifications and those who appeared likely to be eligible were contacted for a more comprehensive phone screening. Individuals were then invited to a group orientation meeting at which the study protocol was reviewed and those who were interested were scheduled for an individual screening visit, at which informed consent was obtained and baseline data collection began. A second screening visit allowed confirmation of interest in participating and review of the self monitoring diary recording dietary intake for a 1-week period (food type and estimate of portion size) which was required to establish final study eligibility. Only those individuals who successfully completed the second screening visit were randomized. Weight from this second screening visit (the final baseline data collection visit) was considered as the pretreatment weight for the current analyses. See Figure 1 for a flow chart of the data collection process.
Participants were randomized to one of three treatment conditions that differed only in delivery channel: In-person, Internet, or a Hybrid condition that combined in-person and Internet sessions. All treatment conditions were offered the same behavioral weight control program with the identical dietary and physical activity goals and the same behavioral strategies to achieve the targeted habit changes. Calorie restrictions and dietary fat intake goals of 25% of calories from fat were provided and graded physical activity goals of 200 min/week of moderate to vigorous physical activity were prescribed. Self-monitoring daily in a pencil-and-paper diary (In-person) or an online diary (Internet and Hybrid) was a central focus of the intervention and weekly self-weighing was recommended. All conditions met in weekly facilitator-led groups (either in-person or by “chat” online) for 24 sessions that lasted ∼60 min. Those randomized to In-person attended closed group sessions (i.e., starting with a cohort of fellow group members and tracking with that group throughout 6-months with no change in group membership). Those randomized to Internet attended virtual closed group sessions in the form of an interactive “chat” on a secure website with other members from their assigned group. All interaction with the group leader was done electronically for Internet participants. The Hybrid group received some sessions online in a chat format (75%) and some In-person sessions (25%), including the first session. The trial and the intervention have been described in greater detail (10).
After participants completed their final screening visit, they were informed that they had been enrolled in the study and would need to wait until the entire cohort had been enrolled before randomization occurred. Recruitment proceeded until a sufficient number of eligible individuals were accrued to provide 12–18 individuals in each study arm at each site. Therefore, during the pretreatment waiting period, participants (and study staff) did not have any knowledge of the experimental condition to which they would eventually be assigned. Individuals learned of their experimental assignment by email and letter 2 weeks before treatment session 1, with all participants informed of their treatment modality assignment the same time.
Body weight. Weight was measured in street clothes, without shoes, on a calibrated digital scale (Tanita, Arlington Heights, IL). Weight was measured at each assessment visit (baseline and 6-months) and before in-person treatment sessions in a private, confidential setting.
Height. Height was measured using a wall mounted stadiometer (Tanita) at baseline. BMI was calculated as weight (kg)/height (m2).
Treatment adherence. Weekly session attendance was recorded by the intervention group leader, as was submission of weekly self-monitoring diaries in which participants recorded their dietary intake, physical activity, and body weight.
Subjects were classified into groups of those who remained weight-stable, gained weight, or lost weight in the period between the final screening visit and the first treatment session. Individuals were considered to have had a weight gain if their weight at the first intervention session was >1.15% of their screening weight. Participants were considered to have lost weight if their weight at the first treatment session was 1.15% lower than their screening weight. Weight stability was defined as weighing within 1.15% of screening weight, allowing for some minor weight liability but relatively consistent body weight. The criterion of a 1.15% weight change to define weight gain or lose was selected as approximately half that used in studies that have defined weight maintenance over a 6-month period as 2.3 kg or 5 lb (11). The selected criterion is similar to the one employed by Carels and colleagues to define a clinically relevant weight loss (7). A weight change of 1.15% in the typical 100 kg participant that presents for behavioral weight control treatment would represent a weight gain or loss of ∼3 lbs.
Participants were predominantly female, obese, and well educated (Table 1). The length of the pretreatment waiting period averaged 50 ± 30 days, with no significant differences across the three weight change groups. Further, weight change groups also did not differ with respect to baseline BMI, gender, age, or education. Although the overall χ2 examining race differences between the weight change groups was significant, pairwise comparisons failed to detect significant differences in the proportion of white American and African American participants among the three weight groups after adjusting for multiple comparisons.
Table 1. Baseline and treatment outcome characteristics by weight change group
During this ∼7-week waiting period, the majority of subjects remained weight-stable (61%). Twenty three percent of randomized participants lost weight during this time, with average weight losses of 2.36 ± 1.26 kg (range = −6.44 to −0.82 kg) over this period. A smaller proportion of the sample gained weight (16%), starting treatment sessions 2.11± 1.04 kg (range = 0.82 to 8.07 kg) heavier than they were when screened at enrollment. Because the weight change groups did not differ in the number of days to start treatment, a longer pretreatment waiting period could not account for differences in weight change during the waiting period.
Those who lost weight in the pretreatment period went on to have the greatest losses at the 6-month data collection visit (−8.9 kg), with significantly greater weight losses at that point than either the weight-stable (−6.1 kg) or the weight gain (−5.7 kg) groups (P < 0.001). Controlling for length of pretreatment waiting time did not change this pattern of results. Weight change during the waiting period was significantly correlated with weight change at 6-months (P < 0.001), but accounted for only 4% of the variance in 6-month weight change. Further, those who lost weight during the waiting period went on to attend a significantly higher proportion of the group/chat sessions than either those who stayed weight stable or who gained weight during the pretreatment waiting period (Table 1). Those who lost weight pretreatment also submitted significantly more self-monitoring diaries during intervention than those who did not lose weight before treatment (P < 0.03). Thus, those individuals who lost weight in the pretreatment period went on to be more adherent and ultimately more successful in their weight control efforts than those who remained weight-stable or gained weight before treatment start.
There were significant differences in the representation of the randomized treatment conditions across the three weight change groups. In the weight gain group, there was a significantly lower representation of individuals who had been randomized to In-person delivery of weight control compared to the proportion of weight gainers in both the Internet and the Hybrid conditions (P < 0.001). Primary analyses of the randomized controlled trial indicated that In-person delivery of the behavioral program produced significantly better weight loss and better attendance at treatment sessions than Internet delivery (10); therefore, we explored what, if any, contributions the pretreatment weight change may have made to these outcomes. Because significantly fewer individuals who gained weight ended up in the In-person treatment arm than were in the other two arms, and because individuals who gained weight pretreatment, lost less weight throughout treatment and attended fewer sessions, there was a possibility that the greater weight loss observed in the In-person condition may have reflected the pretreatment weight loss (or the lack of weight gain) that occurred before treatment rather than exposure to the treatment condition per se. To explore this possibility we considered group differences with pretreatment weight change included in the model as a covariate. We found the same pattern of results as we did in the primary analyses with In-person achieving significantly greater weight losses at 6-months than the Internet and the Hybrid groups (10), indicating that pretreatment weight gain did not confound the results in this particular case. However, the potential for threat exists when pretreatment weight gain is not equally distributed across conditions and therefore may merit examination as a standard practice.
The pretreatment waiting period produced by the process of participant accrual in a research study is dynamic with respect to weight change among overweight and obese individuals queued up and ready to begin a behavioral weight control research program. Although a majority of participants remained weight-stable during this period, there was a substantial proportion for whom weight fluctuations were fairly marked and had clinical significance. Almost a quarter of enrolled participants lost an average of 2.4 kg during this waiting period despite not yet receiving health education or intervention guidance. Further, these individuals who lose weight independently go on to fare better once they start treatment than do those who did not lose weight before initiating the behavioral intervention (regardless of the intervention delivery channel to which they were exposed). Those who lost weight independently before treatment not only lost more weight at 6-months, but they also attended more treatment sessions and submitted more self-monitoring diaries. The constellation of increased self-monitoring, greater group session attendance, and better weight loss outcomes has been repeatedly reported (1,12,13) and therefore this is not a surprise. However, the potential for identifying individuals likely to fall into this group of adherent and successful participants through weight loss during a pretreatment waiting phase is a novel finding, with significant implications for research and clinical applications.
Pretreatment weight loss may serve as a behavioral indicator of a “readiness to change” or motivation and commitment to lifestyle change. Studies that have used questionnaire measures of readiness to change have not proven particularly successful in identifying individuals who will lose more weight (14,15,16) but behavioral measures such as pretreatment weight change may offer more robust predictive power. Many years ago, Brownell suggested a very modest weight loss along with self-monitoring as a treatment program requirement to identify individuals likely to succeed in behavioral weight control (17). Although a behavioral run-in of self-monitoring is often included as part of clinical trial screening procedures, a modest weight loss is not usually a criterion for study entry and it may merit reconsideration. If a goal of screening in controlled trials is to identify individuals likely to engage in or adhere to the offered treatment (and thus allow evaluation of efficacy), our experiences suggest that small weight losses on the order of 1 kg during screening may help achieve this goal.
A smaller, but nonetheless noteworthy, proportion of individuals waiting to start the research-based behavioral weight control program gained weight. The 16% who gained weight did not differ from weight-stable or weight loss groups on demographic characteristics. The 2.11 kg weight gain observed in this study is comparable in magnitude to the pretreatment gain seen among obese binge eaters presenting for eating disorder treatment (8). Eldredge and colleagues reported weight gains of ∼2 kg among obese binge eaters waiting about 50 days to start treatment in their study. They report that individuals who identified depression or anger as the primary emotional trigger for overeating episodes were the ones most likely to gain weight while waiting to initiate the psychological treatment (8). This study did not have information on binge eating behaviors or emotional triggers of overeating and therefore cannot address this issue directly. Nonetheless the remarkable similarity in the magnitude of weight gain over comparable waiting periods suggests that the “last supper effect” may be a reproducible and clinically relevant phenomenon among treatment-seeking obese individuals and worthy of further study.
Weight fluctuation before randomization and treatment initiation warrants consideration from study design perspective as well. Indeed, our experiences should serve as a cautionary tale. Randomization may not equally distribute the pretreatment weight change groups across experimental conditions and this unequal distribution may represent a significant experimental threat, which has heretofore not been examined. A higher proportion of weight losers in one condition may predispose that condition to appear more successful in producing weight loss at follow-up than other conditions, which had fewer individuals who lost weight pretreatment, regardless of the intervention delivered. This suggests that obesity treatment research may need to examine weight change during the accrual period and take it into consideration when analyzing outcomes. Weight loss during the pretreatment waiting period also appears to offer an early clinical suggestion of who will likely be successful in a behavioral weight control program. Previous studies have demonstrated that individuals who lose the most weight during the initial 4–6 months of treatment are those who are most likely to go on to have the largest long-term weight losses (18,19), but these are the first data of which we are aware that indicate that this pattern may be apparent before treatment sessions have even begun. Conversely, those who gain weight or stay weight-stable before initiating treatment may warrant additional intervention to enhance eventual weight loss success. The provision of some instruction during the waiting period before treatment begins, may increase the proportion of participants who lose weight pretreatment—and perhaps improve weight losses later in the program as well. Surgical obesity treatment programs have reported that the provision of pretreatment instruction in behavioral strategies for weight loss enhances surgically induced weight losses in some (20) but not all (9) circumstances, suggesting that there may be some approaches for the waiting period that are more beneficial than others.
Limitations to this study must be acknowledged. Although the sample is fairly large, it is predominantly female and thus extrapolation to males must be done with caution. Further, all participants were volunteers for a randomized clinical trial of obesity treatments and underwent a rigorous screening process which included multiple visits and a week of self-monitoring and therefore the sample cannot be assumed to be representative of the general population of obese individuals entering nonresearch related weight management programs. Another limitation to the study design is that weight at the first treatment session in the Internet condition was self-reported by participants rather than observed. We have demonstrated a strong correlation between body weight self-reported online in the web-based obesity control treatment program and weight observed in the clinic (21). Furthermore, when we restricted the analyses to only the In-person and Hybrid conditions for which we had observed weights at the initial treatment session and thus for which this concern was not relevant, we find a similar pattern of pretreatment weight gain (data not shown). This suggests that self-report of weight in the Internet arm was not distorting the results. Nonetheless, these limitations and the novel nature of the findings merit replication before modification to either obesity treatment study design or analysis of randomized obesity treatment clinical trials would be warranted. However, the potential impact of prerandomization weight change on interpretation of important clinical trials suggests that such confirmatory research would be recommended sooner rather than later. In addition, further exploration into the pretreatment weight loss phenomenon may offer directions for enhancing clinical outcomes by proactively addressing weight management during the pretreatment period.
This study was supported by NIDDK R01 DK056746 to Drs J.H.-B. and D.S.W.
The authors declared no conflict of interest.