Perceived stress as a predictor of eating behavior during the 3-year PREVIEW lifestyle intervention

The study addresses the challenge of long-term weight loss maintenance, which often fails due to a gradual return to previous lifestyle habits. Eating behavior, measured by the Three Factor Eating Questionnaire (TFEQ), is a key target in lifestyle interventions. Prior short-term studies show that increased cognitive restraint and decreased disinhibition and hunger associate with greater weight loss, but factors predicting the long-term maintenance of these behavioral changes are less known. Stress, defined as a state where environmental demands exceed adaptive capacity, may compromise behavior maintenance through biological (e.g., HPA axis activation elevating cortisol and appetite) and cognitive self-regulation pathways. Cross-sectional evidence links higher perceived stress with lower flexible restraint, higher rigid restraint, and greater disinhibition and hunger. This study aimed to examine whether perceived stress measured after an active behavior change stage (6 months) predicts the maintenance of changes in eating behavior over the subsequent 2.5 years and whether frequently experienced high stress during maintenance associates with changes in eating behavior. A secondary objective examined associations between 3-year weight reduction success and changes in eating behavior.

Short-term intervention studies (≤1 year) consistently report that increases in cognitive restraint and decreases in disinhibition and hunger are associated with greater weight loss. Sustained decreases in uncontrolled eating have been linked to better weight maintenance up to 3 years. Prior PREVIEW analyses showed cognitive restraint negatively and disinhibition/hunger positively associated with BMI across 3 years. Cross-sectional research indicates higher perceived stress relates to less cognitive restraint in general but more rigid restraint, and to greater tendencies toward disinhibition, hunger, uncontrolled and emotional eating. Flexible restraint, characterized by moderate, non-deprivational control, appears more beneficial for weight maintenance than rigid, all-or-nothing restraint, which correlates with disinhibition. However, longitudinal evidence on whether perceived stress affects maintenance of improved eating behaviors during extended lifestyle interventions has been limited, motivating the present study.

Design and participants: Secondary observational analysis within the PREVIEW 3-year multicenter randomized lifestyle intervention (ClinicalTrials.gov NCT01777893) conducted in eight countries (Denmark, Finland, The Netherlands, UK, Spain, Bulgaria, Australia, New Zealand). Adults aged 25–70 years with BMI ≥ 25 kg/m² and pre-diabetes (ADA criteria) were recruited (2013–2015). After pre-screening (n = 5472) and screening, participants commenced a 2-month low-energy diet (LED) aiming for 3.4 MJ/day using commercial products. Continuation to a 34-month weight maintenance phase required ≥8% weight loss during LED. Eligible participants (n = 1857) were randomized to one of two diets (moderate-protein, moderate-GI vs high-protein, low-GI) and two physical activity programs (high-intensity 75 min/week vs moderate-intensity 150 min/week). Behavior modification followed the PREMIT toolbox with group visits more frequent in the first 6 months (active behavior change) and less frequent thereafter (behavior maintenance). Analytical sample: 1311 participants who attended at least one visit after month 6 and provided at least one eating behavior measure. Measures and timepoints: Assessments at baseline and at 2, 6, 12, 18, 24, and 36 months. Anthropometry: standardized weight (fasted, light clothing) and height (screening). Eating behavior: TFEQ-51 produced scales for disinhibition (0–16), hunger (0–14), and cognitive restraint subdivided into flexible and rigid restraint (each 0–7) per Westenhoefer. Higher scores indicate stronger tendencies. Cronbach’s alpha across six time-points: flexible restraint 0.65–0.72; rigid restraint 0.43–0.55; disinhibition 0.77–0.82; hunger 0.81–0.84. Perceived stress: PSS-10 (0–4 per item; total 0–40; higher indicates more stress), Cronbach’s alpha 0.78–0.87. High stress definition: Because no standard PSS cut-off exists, the top 20% at baseline defined high stress (PSS ≥ 20). Frequent high stress during maintenance was defined as high stress at least two of four assessments between 6 and 36 months. Weight reduction success over 3 years: percent change from baseline, categorized as successful (>8% loss), partially successful (1–8% loss), or unsuccessful (<1% loss). Statistical analysis: Linear mixed effects models (R 4.0.3; lme4, lmerTest) with maximum likelihood used to assess associations between perceived stress (continuous PSS at 6 months; and frequent high stress group) and eating behavior trajectories from 6 to 36 months, and between weight reduction success groups and eating behavior from 0 to 36 months. Models included fixed effects for age, sex, intervention diet, baseline BMI, and baseline eating behavior, and random effects for participant and intervention center. Main effects estimated overall associations; predictor-by-time interactions tested differential change. Non-significant interactions were dropped. P-values for fixed effects used Satterthwaite approximation; interaction p-values from ANOVA. Effect sizes: standardized betas for continuous predictors and Cohen’s d for group comparisons and change scores. Significance threshold p < 0.05.

- Sample characteristics (n = 1311; 65% women): mean (SD) age 54 (10) years, BMI 34.3 (5.7) kg/m². During the first 6 months, flexible and rigid restraint increased and disinhibition and hunger decreased (all p < 0.001); perceived stress did not change. - Perceived stress at 6 months and maintenance of eating behavior: Time interactions were not significant, indicating stress at 6 months did not predict subsequent changes from 6 to 36 months. However, higher perceived stress at 6 months was associated with overall levels during maintenance: lower flexible restraint (standardized beta −0.16, 95% CI −0.20 to −0.12; p < 0.001), higher disinhibition (0.13, 0.10 to 0.17; p < 0.001), and higher hunger (0.13, 0.09 to 0.17; p < 0.001). - Frequent high stress during maintenance: Group × time interaction for flexible restraint p = 0.026, indicating a greater decline among those with frequent high stress (n = 132) versus those without (n = 588). Mean (SD) change in flexible restraint from 6 to 36 months: −1.1 (2.1) with frequent stress vs −0.7 (1.8) without; Cohen’s d for change = 0.24 (95% CI 0.04–0.43). Between-group difference at month 36: Cohen’s d = 0.50 (0.31–0.70). For disinhibition, no differential change, but frequent high stress was associated with higher overall disinhibition from 6 to 36 months (beta 1.00, 95% CI 0.58–1.43; p < 0.001); between-group Cohen’s d at months 6, 12, 24, 36 were 0.66, 0.56, 0.64, 0.67, respectively. No significant group × time interactions for rigid restraint or hunger. - Weight reduction success (completers n = 962): 30% successful (>8% loss), 42% partially successful (1–8%), 27% unsuccessful (<1%). Group × time interactions for all eating behaviors p < 0.001. Flexible and rigid restraint increased during the first 6 months in all groups; after 6 months, flexible restraint remained stable in the successful group but decreased in partially and unsuccessful groups (largest decline in unsuccessful). Disinhibition and hunger decreased to 6 months in all groups; after 6 months they increased slightly in partially and unsuccessful groups. Effect sizes for total change (0 to 36 months) between successful and unsuccessful: flexible restraint d = 1.16 (0.96–1.35), rigid restraint d = 0.55 (0.37–0.72), disinhibition d = 0.61 (0.42–0.80), hunger d = 0.51 (0.32–0.70).

Perceived stress at the end of the active behavior change stage did not predict subsequent trajectories of eating behaviors, suggesting that a single time-point stress assessment may be insufficient to capture its impact on behavior maintenance. However, both higher perceived stress at 6 months and frequent high stress over the maintenance period were linked to adverse eating behavior profiles: lower flexible restraint and higher disinhibition and hunger. Frequent stress was specifically associated with a greater lapse in flexible restraint, a form of cognitive control that has been linked to successful weight maintenance. These findings align with theoretical and empirical literature indicating that stress, via biological pathways (elevated cortisol and changes in appetite-regulating hormones) and reduced cognitive self-regulation, can undermine adaptive eating behaviors and promote overeating. The stronger association of flexible versus rigid restraint with weight reduction success underscores the potential importance of fostering flexible control strategies within weight management programs. The observed relationships between stress and disinhibition/hunger further highlight mechanisms through which stress may compromise long-term weight maintenance, including stress-induced eating and preference for palatable foods. Clinically, identifying participants with persistent high stress and integrating stress management techniques may help sustain beneficial eating behavior changes and support weight loss maintenance.

Perceived stress was not a prospective predictor of change in eating behavior from 6 to 36 months; however, higher stress was associated with overall lower flexible restraint and higher disinhibition and hunger during the maintenance period, and frequent high stress was linked to greater declines in flexible restraint. Flexible restraint appeared particularly relevant for successful long-term weight reduction. Interventions should consider incorporating stress reduction and management strategies, especially for individuals experiencing high or persistent stress, to enhance maintenance of beneficial eating behaviors and improve weight loss maintenance outcomes. Future research should test stress-management components within long-term lifestyle interventions and refine methods for identifying individuals with sustained high stress.

- Secondary observational analysis within an intervention trial limits causal inference and may be subject to unmeasured confounding. - Selection bias: continuation required ≥8% initial weight loss; overall drop-out was substantial (only 43% completed), and higher perceived stress was associated with higher likelihood of drop-out, potentially attenuating associations. - Stress levels in the sample were moderate on average; results may not generalize to highly stressed populations. - High stress and frequent high stress cut-offs (PSS ≥ 20; at least two high-stress time-points) were arbitrary due to lack of established thresholds. - Eating behavior (TFEQ) and perceived stress (PSS) were self-reported and susceptible to reporting bias. - Reliability of the rigid restraint scale was modest (Cronbach’s alpha 0.43–0.55), potentially affecting precision. - Diet and physical activity randomization arms were pooled; although prior analyses found no differences in eating behavior changes, residual effects cannot be entirely excluded.