Patterns of Weight Change in a Weight Gain Prevention Study for Young Adults

Abstract

Objective:

Recovery from weight regain is uncommon during weight loss treatment. This study examined whether participants in a weight gain prevention intervention similarly struggle to recover following weight gains and what factors predict transitions.

Methods:

This is a secondary analysis of data from the Study of Novel Approaches to Weight Gain Prevention (SNAP), a randomized controlled trial comparing two weight gain prevention interventions to control. Young adults (n=599; 18-35 years) were followed over 3 years. Markov models identified transition rates in going above and returning below baseline weight across follow-up. Logistic regressions identified predictors of transitions.

Results:

At each time point, approximately double the number of participants who transitioned from below to above baseline transitioned from above to below. Magnitude of weight changes from baseline and the number of weight loss strategies used predicted transitions from below to above and above to below baseline weight (with opposite relationships). Infrequent self-weighing and lower dietary restraint predicted transitions below to above baseline weight. Treatment arm, demographics, calorie consumption and physical activity generally did not predict transitions.

Conclusions:

Young adults engaging in weight gain prevention struggle to lose gained weight. Alternative strategies are needed to address weight gains in weight gain prevention interventions.

Introduction

Weight loss maintenance is the most pressing issue in obesity treatment (1). A review of studies focusing on individuals who successfully lost 5% or more of their weight showed individuals, on average, regained all of their weight by 4 years without continued intervention (2). High rates of weight regain indicate that weight losers have trouble re-instating weight loss efforts despite having used them effectively during weight loss. Existing studies suggest that once even small amounts of weight are regained, it is unlikely that individuals will lose it. For instance, a study by Phelan and colleagues (2003) showed that only 11% of successful weight loss maintainers who gain any weight over a year are able to return to their baseline weight or below the following year (3).

Weight gain prevention is similar to weight loss maintenance in that the primary goal is to maintain current weight over time; however, the populations differ. Weight gain prevention participants begin treatment without obesity and do not experience the same biological drive towards weight regain as individuals who are weight suppressed (4). Conversely, they often are not taught the knowledge or skills necessary for weight loss. It is unclear how many participants in weight gain prevention programs are able to recover from periods of weight gain and what factors may predict these successful transitions. Knowledge regarding patterns of losses and gains could allow for optimal time points for intervention.

The current study is a secondary analysis of the Study of Novel Approaches for Weight Gain Prevention (SNAP) trial, which tested a large change (creation of a 5-10 lb. weight loss buffer) and a small change (small consistent changes to eating and activity) against a self-directed control (6). Both interventions were shown to be more effective at preventing weight gain than control (7). Distinct from the primary trial, the aims of the current study were to determine the prevalence of and the rates at which participants transitioned above and below their baseline weight across 3 years. A second aim was to identify demographic and treatment-related predictors of these transitions. We hypothesized that a relatively small percentage of participants would reverse weight gain once above baseline weight and that use of weight control strategies would be related to these transitions.

Methods

Design and Primary Outcomes of the SNAP Trial

The Study of Novel Approaches for Weight Gain Prevention (SNAP) was a three-arm randomized-controlled trial designed to test two approaches to weight gain prevention in young adults (6, 7). Both experimental conditions followed a self-regulation framework for weight control, although differed in their recommendations. The large change intervention was designed to create a 5-10 pound weight loss to buffer against weight gain over time while participants in the small change condition were asked to make small, consistent changes to diet and activity to prevent weight gain. Participants in both conditions received a 4-month in person group program (8-weekly sessions followed by 2 monthly sessions), with two optional 4-week booster sessions offered remotely in each subsequent year in the follow-up period. The control condition received a one-session, self-directed self-regulation intervention with no further intervention contact. The primary outcome of the study was weight change across the mean of 3 years of follow-up (similar to area under the curve) and included weights measured at baseline, 4 months, 12 months, 24 months; some participants (depending on time of recruitment) also had a 36 and 48 month assessment. There were significant differences among all 3 groups (large change: −2.37, small change: −0.56, and control: 0.26) for this primary outcome. A secondary outcome was weight change from baseline to year 2 (the final time point reached by all participants during SNAP). For this outcome, the large and small change interventions both differed from control but not from each other. After the ending of SNAP, a study extension permitted collection of weights at 3-6 years follow-up. Weight data from the 3-year assessments were used in the current analyses. The Institutional Review Board at each institution approved the study (Lifespan-The Miriam Hospital; UNC Chapel Hill Non-Biomedical). Additional information regarding the protocol and the primary outcomes can be found in manuscripts by Wing and colleagues (6, 7).

Participants

Participants (n=599) aged 18-35 years with a BMI of 21.0 - 30.9 were recruited at two clinical sites to participate in SNAP. Inclusion criteria required participants to be English speaking with Internet access and to be able to participate fully with intervention recommendations (e.g., engage in physical activity). Exclusion criteria included a recent weight loss of >10 pounds in the prior 6 months, bariatric surgery, history of anorexia or bulimia, hospitalization for or history of severe psychiatric disorders, and currently or planning to become pregnant. A total of 507 had weight data at 3-year assessments.

The Traffic Light Zone System for Weight Self-Regulation

At the last sessions of the initial treatment programs, both the large change and small change interventions conditions were counseled in self-regulation strategies following a traffic light zone system, which defined 3 zones: green, yellow and red. The green zone was designed to keep participants close to or below their final weight following the initial 8 weeks of in-person sessions. Green zone instructions in the large change treatment were to continue with healthy eating and high levels of physical activity and in the small change treatment were to continue making 2 small changes a day. If participants were in the yellow zone, which signified they were gaining back to their baseline weight, they were asked to problem solve weight gain (large change) or to start monitoring small changes (small change). The red zone was defined as any weight greater than baseline weight. Red zone instructions in the large change intervention were to return to the recommendations for calorie consumption and physical activity provided during the weight loss phase (e.g., self-monitoring, calorie restriction). In the small change intervention, participants were instructed to return to planning and monitoring their small changes and to increase the daily number of small changes if no weight loss was experienced. In addition, both conditions were offered twice-monthly optional phone counseling if they were in the red zone to help them return to their green zone (6). Only 44 participants (out of 379 who completed the 4 month large or small change interventions) availed themselves of red zone counseling with a median of 2 sessions (range 1-46).

Measures

Masked study staff administered each assessment indicated below at baseline, 4 months, and 1, 2, and 3 years, unless an exception is noted in measure description.

Demographics

Participants self-reported age, gender, race/ethnicity, and income at baseline. Race/ethnicity was consolidated into White, Black, Hispanic, and other.

Anthropometrics

Weight was measured with a calibrated digital scale and height was measured with a wall-mounted stadiometer. Participants were measured without shoes in light clothing. The average of two measurements was used.

Self-Weighing

Participants reported their frequency of self-weighing on a 5-point Likert scale from 1 (never) to 5 (every day). The scale was dichotomized in accordance with previous research (8) and individuals who reported weighing several times per week or more were considered “frequent self-weighers”

Dietary Restraint and Disinhibition

The 51-item Three-Factor Eating Questionnaire was administered and subscales measuring dietary restraint and disinhibition were used in the current study (9). Higher scores indicated greater restraint and disinhibition.

Weight Management Strategies

The frequency of use of 45 different weight management strategies over the past 4 months was assessed using a 5-point Likert scale ranging from 1(never/hardly ever) to 5 (always/almost always). The strategies were compiled from Pound of Prevention, NHANES and the Weight Loss Maintenance trial and used in the EARLY consortium trials (10). The number of strategies a participant reported using “always or almost always” was summed.

Diet

Diet was measured using the Block Food Frequency Questionnaire, which assesses typical dietary intake. The questionnaire was given at baseline and two years only. Total daily kilocalorie consumption was used in the current study (11).

Physical Activity

The validated SenseWear Pro Armband (Body Media, Pittsburgh, PA) measured physical activity (12). Participants were instructed to wear the device during all waking hours (except swimming and bathing) for a full week and were included in the analysis if they monitored for at least 10 hours per day on at least 4 days. Moderate to vigorous physical activity (MVPA) was defined as ≥3 METs. Additional physical activity analysis information can be found in Unick et al., 2017 (13). Physical activity data were collected at all but the 3-year time point.

Statistical Analysis

Latent transition analysis (LTA) was used to apply Markov modeling, in which longitudinal change is modeled as transitions between discrete states (14, 15). Above baseline (i.e., the red zone) or at/below baseline (i.e., non-red zone) weights were the states of interest and were measured using a single indicator at each time point with the assumption of no measurement error (scales for weight measurements were calibrated and multiple weights were taken to provide highly-accurate weight measurements). LTA provides estimates of stage membership at baseline, change between stage membership from time t to t + 1, and can be used to identify predictors of stage membership and transitions over time in stage membership. The state at t+1 depends only on previous time t; prior states have no influence. Predictors are assessed using binary logistic regression within the Markov model. Logistic regression results provide an estimate of the effects of a one-unit increase in a predictor on the log odds of transitioning from above baseline to below baseline (or vice versa) between two assessment time points. If a predictor was significant, follow-up testing identified whether it was significant for each type of transition (above to below and below to above baseline). Likelihood ratio tests are used to determine significance of effects with an alpha of <0.05.

For the analysis, a baseline model was fit initially to determine prevalence rates of being above baseline or below baseline over time and the rates of transitions between categories. Prevalence rates were also assessed by treatment condition at each time point. Predictors were then added to the baseline model individually to determine effects of predictors on rates of transitions at each time point. For treatment, two separate models were run to test large and small changes against control as well as large and small changes against each other. If a predictor was found to be significant, follow-up testing of individual logistic regressions was used to determine which transition (i.e., above baseline to below baseline, below baseline to above baseline, or both) was significant. All analyses were completed using PROC LTA, which uses a full information maximum likelihood-based estimation procedure that handles missing data on observed indicators using the expectation-maximization algorithm. PROC LTA is not able to accommodate missing predictor values, thus in models with predictors, cases are restricted to those with available data.(15)

Results

Demographics and 3-Year Weight Change

Demographics of participants who contributed to the present analyses are presented in Table 1. The retention rate at the 3-year assessments was 84.6% and did not differ by condition or by prior weight loss (7). Percent weight change from baseline to 3 years did not differ significantly by condition F(2,500)=1.18, p=0.306 (large change: 0.34±8.00%, small change: 0.74±7.14%, and control: 1.64±8.68%).

Table 1.

Baseline Demographics of Study Participants

Variable	Participants
Gender, n (%)
Male	130 (21.7)
Female	469 (78.3)
Age (years; mean ± SD)	28.2 ± 4.4
Race/Ethnicity, n (%)
White	438 (73.1)
Black	66 (11.0)
Hispanic	46 (7.7)
Other	49 (8.2)
BMI (mean ± SD)	25.4 ± 2.6
Incomea
<$25,000	189 (33.0)
$25,000-$49,999	213 (37.2)
$50,000-$74,999	110 (19.2)
>$75,000	61 (10.6)
Student Status
Non-student	390 (65.1)
Student (full or part-time)	209 (34.9)

^a26 participants did not report income

Prevalence Rates and Transition Rate Below and Above Baseline Weight

Prevalence rates.

At 4 months, 24% of participants were above their baseline weight. Prevalence of individuals above baseline weight increased over time, with 52% of participants above their baseline weight at 3 years [Figure 1]. Among individuals who had data at all five time points (n=446), 35.7% stayed below their baseline weight, whereas 10.8% were above their baseline weight at all time points.

Figure 1.

Percentage of Participants Above and Below Baseline Weight (indicated after category) and Transitions Over Three Years

Analyses by treatment condition showed that relative to the control group, a smaller number of individuals were above baseline in the large change group at each time point (p’s<0.021) and in the small change group at 4 months (p=0.007) and 2 years (p<0.001) [Figure 2]. A significantly smaller number of individuals were above baseline in the large change group at 4 months (p<0.001) and 1 year (p=0.007) compared to the small change group.

Figure 2.

Percentage of Participants above Baseline Weight by Treatment Group over Three Years. The reference group is the control intervention at each time point.

Transition rates are illustrated in Figure 1 and presented in Table 2. The actual numbers of participants transitioning are estimated (based on the initial n=599) to provide a sense of participant flow. From one time point to the next, approximately two times the number of participants transition to above baseline compared to the number who transition to below baseline. Transition rates were not significantly different between time points (i.e., the rates of transitioning did not increase or decrease over time; p>0.05).

Table 2.

Transition Rates (estimated Ns^a) for Weight Category Status from Previous to Subsequent Time Point Across 3-Year Follow-up

	Remain in Category		Transition to New Category
	Below BL➔ Below BL	Above BL➔ Above BL	Above BL➔ Below BL	Below BL➔ Above BL
12 Months	0.768 (348)	0.720 (105)	0.280 (41)	0.232 (105)
24 Months	0.755 (293)	0.801 (169)	0.199 (42)	0.245 (95)
36 Months	0.709 (238)	0.803 (212)	0.196 (52)	0.291 (97)

BL = Baseline

^aNs provided are based on the starting sample n=599 and do not reflect attrition.

Predictors of Transitions Above and Below Baseline Weight

Demographic variables did not predict transition rates at any time point. The only significant treatment effect on transitions occurred between the small change and control groups at 24 months. The effect indicated individuals in the small change intervention showed a 2.8 greater likelihood to transition from above to below their baseline weight (p=0.022) whereas individuals in control showed a 2.5 (1/0.4) greater likelihood to transition from below to above their baseline weight (p=0.042).

Table 3 provides odds ratios for all other significant predictors of transitions. The variables that predicted transitions in both directions (e.g., both weight loss and gain transitions) were magnitude of difference in weight from baseline weight at the previous time point (from time “t” to “t+1”) and the number of weight loss strategies used in the previous 4 months. For weight difference from baseline, all time points were significant, with the exception of transitioning below baseline at 24 months. For number of weight loss strategies used, only transitions at 12 and 24 months were significant. Individuals who transitioned from above to below baseline had gained less weight and reported using a greater number of weight loss strategies (12 mo: 5.55±5.00 vs. 3.85±4.4 strategies) compared to those who remained above baseline. In contrast, those who transitioned in the opposite direction (from below to above baseline) had lost less weight and reported using fewer weight loss strategies (12 mo: 3.64±3.85 vs. 6.46±6.57 strategies) than those who remained below. The differences in the mean number of weight loss strategies used remained similar at 24 months.

Table 3.

Odds ratios of Significant Predictors Affecting Transitions Above and Below Baseline Weight^a

	Magnitude of Weight Difference from Baseline Weight		Sum of Weight Loss Strategies		Self-Weighing		Restraint
	Above BL➔ Below BL	Below BL➔ Above BL	Above BL➔ Below BL	Below BL➔ Above BL	Above BL➔ Below BL	Below BL➔ Above BL	Above BL➔ Below BL	Below BL➔ Above BL
12 Months	0.70*	1.44*	1.12*	0.92*	2.14	0.37*	0.99	0.86*
24 Months	0.92	1.36*	1.11*	0.94*	1.97	0.29*	0.98	0.79*
36 Months	0.73*	1.42*	1.00	0.93	1.16	0.54	1.08	0.86

^aThe reference group for each odds ratio are individuals who stayed in the same weight category (i.e., did not transition) at the indicated time point. For example, for individuals who went from above baseline weight at 4 months to below baseline at 12 months, the magnitude of weight difference odds ratio of 0.70 is in comparison to individual who were above baseline at 4 months and remained above baseline at 12 months; Questionnaire measurements were from time points concurrent with weight measurements;

^*p<0.05 in follow-up post-hoc testing with binomial logistic regressions

Frequency of self-weighing and dietary restraint only predicted weight gain transitions, with individuals reporting less frequent self-weighing and lesser dietary restraint bring more likely to gain weight above baseline than those who remained below baseline at 12 and 24 months. Disinhibition, kilocalorie consumption, and MVPA did not predict transitions in either direction at any time point.

Discussion

The current study found the prevalence of young adults who had successfully maintained their weight below baseline in a weight gain prevention study went from 76% at 4 months to 48% at 3 years. Rates of transitioning either above or below baseline weight were steady over follow-up, with approximately double the number of participants transitioning from below to above baseline compared to the number transitioning from above to below baseline. Less weight loss from baseline weight, fewer number of weight loss strategies used, less frequent self-weighing frequency, and less dietary restraint were related to transitions from below to above baseline and less weight gained from baseline weight and greater number of weight loss strategies predicted transitions from above baseline to below.

The transition rates indicate that about 75% of participants remain either above or below their baseline weight at each time point, suggesting the majority of individuals continue on a similar weight trajectory over time. However, approximately 25% of these individuals will transition to either above or below their baseline weight at each time point. Given that a much larger number of participants are below their baseline weight at each assessment, this equates to approximately twice the number of participants transitioning from below to above baseline compared to above to below, explaining the increasing prevalence rates over time. The relatively low rates of recovery below baseline (20-28% or an estimated 40-50 participants) suggest that once an individual gains weight, it is difficult to lose. These rates are higher than the 11% found by Phelan and colleagues (2003), who focused on successful weight loss maintainers. Differences in outcomes may stem from population differences. A number of biological and behavioral challenges are associated with maintaining large weight losses, likely making it even more difficult for the weight loss maintainers to lose regained weight (4). However, results from both studies indicate weight that is gained following a time period of successful weight management is challenging to lose with little follow-up intervention.

Despite knowledge of the challenges associated with maintaining behavior change (e.g., waning motivation, decreasing self-regulation)(16), the barriers to effectively combating weight regain, or reinitiating behavior change efforts, are not well-understood. Similarly, weight loss maintenance interventions have shown some success in preventing weight regain (17, 18), but known strategies or interventions to treat weight regain are limited. Potential strategies to investigate could be those that decrease burden of reinitiating weight control efforts, such as a simplified self-monitoring system (19), or that increase motivation, such as motivational interviewing (20). Recent literature also indicates that individuals who regain weight may have a negative bias in remembering their experience in a weight management intervention (21), which may contribute to a lack of desire to reinitiate behavior change techniques. Future studies should consider how best to encourage participants to re-initiate weight control efforts when needed.

As such, the timing of such intervention strategies is relevant. Results of the current study showed the proportion of people transitioning above or below baseline did not differ by follow-up time point. Thus, there may be no optimal time period to provide supplementary treatment to affect the largest number of participants, but instead, intervention may be equally useful across follow-up. One other important consideration related to timing of additional intervention is participant engagement. Wing and colleagues (2003) showed behavioral momentum to engage in weight control efforts did not deteriorate following a six-week break in treatment (22) and “chunking” treatment into modules with breaks in between has been suggested to enhance long-term maintenance of weight loss (23, 24). However, other studies have shown high attrition and low interest in booster sessions following completion of an intervention (25) or in reinitiation of the same weight loss program following weight regain (26). Booster sessions are often optional whereas treatment “chunking” is built into the structure of the program, which may provide some insight into different participant responses. Additional work is needed to identify the ideal timing and structure of supplementary content for weight regainers to harness participant interest.

A second goal of this paper was to identify characteristics of and strategies used by individuals who are successfully able to recover from weight gains. Absolute magnitude of weight change from baseline was a significant predictor of both transitions above or below baseline weight at most time points. For individuals above their baseline weight, greater amounts of weight gained would require greater losses to return to baseline, increasing the challenge of doing so. Phelan and colleagues (2003) also found weight loss maintainers who regained greater amounts of weight were less likely to return to their baseline maintenance weight. The results speak to the importance of aiding participants in addressing small increases of weight above baseline, when reversing weight gain may be more manageable. On the other hand, losing greater amounts of weight from baseline seems to provide a greater buffer against returning to baseline weight, which is also true of weight loss trials (27, 28).

The only behavioral strategy assessed that predicted transitions both above and below baseline weight was the sum of total weight loss strategies. A recommendation to employ a multitude of strategies (>5 based on means in the current analysis) could potentially be used in the future as a helpful intervention for participants who are struggling to cope with weight gain. This finding complements existing literature indicating that the number of weight loss strategies used is a mediator of weight loss (29). Dietary restraint and self-weighing were predictive of weight gain transitions, but not weight loss transitions. Both dietary restraint and self-weighing have been identified as important predictors of successful 3-year outcomes in SNAP (8). The current study contributes to existing findings by reinforcing the value of frequent self-weighing and greater dietary restraint across follow-up, although also suggests their effects become less robust after 2 years with little intervention. However, lack of significant findings regarding the ability of dietary restraint and self-weighing to predict transitions from above to below baseline indicate they may be less useful, at least individually, for combating weight gain.

Treatment, particularly the large change intervention, was helpful at keeping a larger number of participants below their baseline weight, but the number of individuals who were transitioning from below to above baseline (or vice versa) at each time point was similar across treatments. Both the large and small change intervention groups received relapse prevention training to prevent weight regain and facilitate losing gained weight. Participants were also offered optional twice-monthly phone counseling in the follow-up period if they were above their baseline weight, although uptake of counseling was low. As indicated above, additional or new strategies may be needed to help address weight gain/regain or to motivate a recommitment to treatment activities following deterioration of program effects in weight gain prevention interventions. One exception was that individuals in the small change intervention were better able to lose weight gained above baseline and to prevent transitioning above baseline weight at 2 years compared to the control group. While it is unclear why effects occurred only at year 2, the results demonstrate a signal that small changes may be preferable long-term approach to preventing weight gain and coping with setbacks compared to large changes, perhaps due to the less burdensome nature of treatment activities and a corresponding willingness to reinitiate them.

Baseline demographic characteristics also did not predict transitions above or below baseline weight over time. Demographic variables are non-modifiable, therefore it may be promising that various demographic groups show similar transition rates. Finally, calorie consumption and MVPA were not significant predictors of transitions. Phelan and colleagues (2003) also did not find objectively measured behavioral variables to distinguish individuals who recovered from weight regain and those who did not, although dietary variables in particular are difficult to measure with accuracy (30). However, both energy intake and minutes of MVPA were found to predict weight outcomes in the current sample at 2 years in other studies (13, 31), speaking to their importance for weight management more generally.

This study took a unique approach to analysis that includes limitations. Categorizing participants into groups characterized by current weight being above or below baseline reflected study specific recommendations; however, transitions in this weight categories encompass both small and large magnitudes of weight change. Relatedly, the categorization is not sensitive to participants who either lose or gain large amounts of weight without crossing the baseline weight threshold. Future studies may consider alternative definitions of transitions (e.g., individuals who experience a certain amount of weight change) and inclusion of additional assessment time points that may also provide valuable insight into the time course of weight gain/regain and the ability of effectively address weight increases within a weight gain prevention intervention. Additionally, the predictor questionnaire measures occurred at the same assessment time points as the weight measurements, which may reflect more recent behavior instead of behavior representative of the entire time between assessments.

Conclusion

The current study indicates that weight gain prevention intervention participants struggle to return to their initial weight following weight gains. Common predictors of weight control intervention outcomes (e.g., frequent self-weighing) were found to predict transitions in intervention follow-up, although fewer were identified that predicted recovery from weight gain above baseline than predicted weight increase above baseline. Future work may try to identify additional predictors of transitions and consider additional or alternative strategies that may aid individuals in recovering from unwanted weight gain in the maintenance phase of a weight control intervention.

What is known about this subject?

• Participants in behavioral obesity treatments struggle to recover from weight regains that they experience

What are the new findings in this manuscript?

• Participants in weight gain prevention interventions who experience a weight gain are unlikely to lose it over subsequent years

• A smaller magnitude of weight gain from baseline weight and a larger number of weight loss strategies used predicts recovery from weight gain during a weight gain prevention program

How might your results change research or clinical practice?

• Addressing weight gains when they are small in magnitude and prescription of and coaching around a greater set of strategies for weight management may be helpful in addressing weight gains

• Additional research into strategies that are effective at combating weight gain during weight gain prevention programs is needed