Summary: In this secondary analysis of a Swedish primary-care randomised trial (n=119) of a 12-week smartphone dietary-education app for type 2 diabetes, participants with higher app engagement showed modestly larger dietary improvements than low-engagement users. Directly after education the high-engagement group differed from the low-engagement group for whole grains (beta=20.4, 95% CI 0.57 to 40.3, P=.03), with scattered macronutrient differences at 6 to 9 months. No significant cardiometabolic marker improvements were reported, and the engagement groups were defined post hoc rather than randomised.
PICO Summary
| Element | Detail |
|---|---|
| Population | 119 adults with type 2 diabetes recruited in primary care, Sweden; 60.5% (72/119) men, mean age 63.2 years (SD 10.3), mean BMI 30.1 kg/m2 (SD 5.1). Secondary analysis of a 1:1 randomised controlled trial. |
| Intervention | High app engagement, defined post hoc as 100% of in-app activities completed: 53.8% of users (64/119). Moderate engagement (50% to 99.9%): 21.8% (26/119). Mean activities completed across the cohort 77.1%. |
| Comparison | Low app engagement, defined as fewer than 50% of activities completed: 24.4% of users (29/119). Engagement categories were observational, not randomised allocations. |
| Outcome | Diet quality and intake (food frequency questionnaire) plus clinical markers, by linear regression of between-group change. Directly after education: whole grains beta=20.4 (95% CI 0.57 to 40.3, P=.03) for high versus low engagement. At 6 to 9 months: significant differences for fibre, whole grains, carbohydrate, saturated fat, sodium and total energy in moderate versus low, and for carbohydrate in high versus low. No significant cardiometabolic clinical-marker improvements were reported; no ARR/NNT applicable. |
Expert Commentary
This is a secondary, associational analysis rather than a test of whether the app works, and it should be read as such. Because the engagement categories were defined after the fact by how much of the app each person actually completed, the comparison is confounded by the very traits that drive engagement: motivation, health literacy and baseline dietary interest. People who finish every module are plausibly the ones who would have improved their diet regardless, so the modest signals seen here, a whole-grain difference immediately after education and a handful of macronutrient differences at 6 to 9 months, cannot be attributed to the app itself. The verdict is that higher engagement tracks with slightly healthier eating, but causation is not established and the cardiometabolic markers did not move significantly. The single most important limitation is this post-hoc, non-randomised grouping, which converts a randomised trial into observational data and invites reverse causation. The sample is also small (n=119) and the findings are scattered across nutrients rather than anchored to a prespecified primary outcome, raising the risk of chance findings. Can I use this with my patients? Not yet as evidence that the app improves outcomes, though it is reasonable to tell an engaged, primary-care patient with type 2 diabetes that completing a structured dietary programme is consistent with better eating. Future trials should compare randomised arms and prespecify clinically meaningful endpoints.
References
Sjoblom L, Stenbeck F, Trolle Lagerros Y, Hantikainen E, Bonn SE. Engagement With a Smartphone-Delivered Dietary Education Intervention and Its Relation to Dietary Intake and Cardiometabolic Risk Markers in People With Type 2 Diabetes: Secondary Analysis of a Randomized Controlled Trial. JMIR Form Res. 2025;9:e71408. doi:10.2196/71408