Design: randomized, two-phase, parallel-group clinical trial conducted at NYU Langone Health. Participants: 204 adults aged 18–80 with BMI 27–50 and abnormal glucose metabolism/obesity (prediabetes or moderately controlled type 2 diabetes, HbA1c 5.7–8.0%). Exclusions included kidney disease or use of medications other than metformin. Assessments occurred at baseline, 3, and 6 months; analyses followed an intention-to-treat approach with linear mixed models. Randomization was blinded to participants until week 5 of intervention. Baseline characteristics: mean age 58 years (SD 11); 66.8% female; mean BMI 33.9 (SD 4.8); HbA1c 5.8% (SD 0.6%).
Personalized diet estimated postprandial glycemic response (PPGR) to meals using a machine-learning algorithm that integrated anthropometrics, HbA1c, lifestyle data, and gut microbiome; Standardized low-fat diet with fat <25% of total energy. Both groups received 14 group counseling sessions via WebEx; self-monitoring of diet and body weight through the PNP smartphone app with real-time feedback on energy balance and macronutrients; PPGR feedback presented as color-coded meals scores (green/excellent, yellow/medium, red/bad); targets included 7% weight loss from baseline, a 500 kcal/day energy deficit, and 30 minutes of physical activity daily; duration: 6 months of active intervention followed by 6 months of maintenance/observation.
Weight change at 6 months: standardized group −4.31% (95% CI −5.37 to −3.24); personalized group −3.26% (95% CI −4.25 to −2.26); between-group difference 1.05% (95% CI −0.40 to 2.50; P = .16). From 4 to 6 months, standardized group achieved greater weight loss than personalized by 1.66% (95% CI 0.13 to 3.18; P = .03). No significant between-group differences in fat mass, fat percentage, fat-free mass, respiratory quotient, or adaptive thermogenesis; resting energy expenditure declined more in the standardized group (−92.3 kcal/d difference; 95% CI 0.9 to 183.8; P = .05). Adherence to counseling sessions was similar; self-monitoring adherence was higher in the personalized group (days with >0 kcal logged: 54.3% vs 41.1%; P = .01; days with ≥50% of caloric target logged: 42.0% vs 29.9%; P = .01). Conclusion: A precision-nutrition approach targeting PPGR did not produce greater weight loss than a standardized low-fat diet over 6 months. Future work should improve dietary self-monitoring adherence and intervention exposure and refine weight-loss–specific predictive algorithms for precision nutrition.
COVID-19 disruptions reduced data collection and follow-up, with some secondary outcomes missing and 12-month measurements not uniformly performed. Delays related to stool sampling and app adjustments may have limited intervention exposure. Retention was imperfect (lower in the standardized group) and the sample was largely well-educated, English-speaking, drawn from a single health system, limiting generalizability. Physical activity data were limited to baseline measurements; dietary intake was assessed with a single automated 24-hour recall at each time point, reducing capture of day-to-day variability.
