Abstract
BACKGROUND: Obesity affects more than one-quarter of adults in the United Kingdom and is a leading cause of preventable disease and health care costs. Digital behavior change programs can provide scalable weight management support, but maintaining engagement is challenging, and engagement is strongly associated with weight loss success. Tailoring interventions to cognitive-behavioral phenotypes, distinct patterns of thinking and behavior, offers one strategy to improve adherence. Although such approaches show promise in controlled settings, evidence from real-world digital programs is limited. OBJECTIVE: This study evaluated whether phenotype-tailored weekly advice improved engagement and weight loss in a national digital weight management program. Secondary aims were to assess correlations between advice use and outcomes, explore moderation by socioeconomic status, and capture participants' perceptions of the advice. METHODS: We conducted a quasi-experimental study among UK adults enrolled in a free 12-week program commissioned by the National Health Service. Eligible participants were aged 18-80 years with a BMI greater than 25 kg/m². The phenotype group (n=148; mean age 48 years; 127/148, 86% female; mean BMI 39 kg/m²) completed a 17-item questionnaire, were classified into one of 4 phenotypes, and received weekly tailored advice for 7 weeks. Comparators included a historical cohort enrolled 1 year earlier without phenotype advice (n=241; mean age 44 years; 171/241, 71% female) and nonresponders who did not complete the questionnaire (n=394; mean age 44 years; 299/394, 76% female). Primary outcomes were program engagement (any in-app activity such as meal logging, activity tracking, content reading, or coach messaging) and self-reported weight. RESULTS: The phenotype group recorded a mean of 257 (SD 232) engagements over 7 weeks, significantly higher than both the historical cohort (mean 159, SD 187; P<.001) and nonresponders (mean 135, SD 198; P<.001), representing 62%-90% greater activity. All engagement types were significantly elevated (P<.001 for all). Mean weight loss was -2.23 kg (SD 7.97) in the phenotype group, compared with -1.60 kg (SD 5.39; P=.29) in the historical cohort and -0.69 kg (SD 13.23; P=.23) in nonresponders. The number of phenotype-specific advice documents opened correlated with engagement (r=0.48; P<.001) but not with weight loss (P=.42). Socioeconomic status did not moderate outcomes. Posttrial interviews (n=16) provided mixed feedback: many participants described the advice as clear, relevant, and motivating, whereas others considered it too general or poorly matched. CONCLUSIONS: Phenotype-tailored weekly advice was associated with substantially higher engagement in a real-world digital program, although short-term weight differences were not statistically significant. While limited by a nonrandomized design, short follow-up, and reliance on self-reported weight, this study suggests phenotype-based tailoring may be a scalable strategy to strengthen adherence in digital weight loss interventions. Larger randomized trials with longer follow-up are warranted to determine whether increased engagement translates into clinically meaningful weight loss.