Dietary Patterns During Weight Loss Maintenance vs. Weight Regain: A Secondary Analysis of the Look AHEAD Trial

Background/Objectives: Limited information exists on how dietary patterns change in adults who experience weight regain vs. those who maintain weight loss after lifestyle interventions. Methods: Five hundred fifty-two adults (60 ± 1.0 years, 33.8 ± 0.4 kg/m²) with type 2 diabetes mellitus from the Look AHEAD Trial achieved ≥ 7% weight loss after Year 1, completed follow-up visits through Year 4, and provided three 134-item food frequency questionnaires. Weight "regain" (WLR) was defined as regaining ≥ 50% of the initial weight lost. Dietary patterns were determined using established DASH diet scores (scale: 0-9) and principal component analysis (PCA; higher scores = more adherent). Repeated measures linear mixed models assessed group and sex differences in dietary patterns. Results: Dietary patterns were similar between groups during weight loss (baseline to Year 1). WLR DASH scores decreased more from Year 1 to Year 4 compared to "maintain" (WLM) (WLR: Y1: 5.66 ± 0.14, Y4: 4.60 ± 0.14; WLM: Y1: 5.49 ± 0.13, Y4: 4.92 ± 0.13; difference-p < 0.01). Of the two PCA-derived dietary patterns, Pattern 1 (vegetable, fruit, and fish) decreased more in WLR (WLR: Y1: 0.12 ± 0.16, Y4: -0.14 ± 0.16, WLM: Y1: 0.06 ± 0.14, Y4: 0.25 ± 0.15; difference-p < 0.01), while Pattern 2 (low-fiber grains and high-fat animal proteins) increased in WLR (WLR: Y1: 0.40 ± 0.11, Y4: 0.61 ± 0.11, WLM: Y1: 0.34 ± 0.10, Y4: 0.21 ± 0.10, difference-p < 0.01). Sex differences showed that only WLR women and WLM men increased sweets from Y1 to Y4 (WLR women Y1: 0.26 ± 0.04, Y4: 0.41 ± 0.04; p < 0.01; WLM men: Y1: 0.23 ± 0.04, Y4: 0.38 ± 0.04; p < 0.01). Conclusions: These data demonstrate that differences in dietary patterns between WLR and WLM emerge after the initial weight loss intervention with some sex differences. This suggests that longer-term shifts in dietary patterns after lifestyle interventions may influence weight loss maintenance.