Exercise Snacks Enabled by an mHealth Intervention Delivered in General Practice to Promote a Sustainable Active Lifestyle in Adults With Obesity

Background and rationale.

Physical activity is a cornerstone of obesity management, yet most interventions fail to embed activity into everyday life and their effects fade once supervised contact ends. A major obstacle is low initial fitness, which makes conventional exercise unpleasant and reinforces inactivity. Exercise snacks, that is brief structured bouts of activity distributed across the day, lower that barrier and lend themselves to being embedded in an ordinary routine. Sustaining the change beyond the supervised phase, however, requires continued support, which is expensive to deliver by human counsellors at scale. This trial evaluates a lifestyle physical activity programme centred on exercise snacks, initiated in general practice, and separately evaluates whether an automated conversational agent can take over the supporting role once the supervised phase ends.

Objectives and endpoints.

The trial has two objectives. Each is supported by its own randomisation, in its own population, over its own time window, and each therefore has its own primary endpoint.

The first objective is to determine the effect of the six-month programme compared with an active control condition. The first primary endpoint is the change in device-measured physical activity from baseline to six months.

The second objective is to determine the effect of automated maintenance support delivered by a conversational agent, compared with no further support, among participants who received the programme. The second primary endpoint is the change in device-measured physical activity from six to eighteen months. This comparison is not conditional on the result of the first.

Testing structure.

The two primary endpoints are treated as two separate confirmatory families. Each is tested at a two-sided alpha of 0.05, and no adjustment for multiplicity is applied between them. The justification is that the two comparisons rest on two physically separate randomisations, address different scientific questions, are estimated in different populations and cover different time windows. They are not co-primary endpoints, and success is not defined as statistical significance on both.

Within each family, the change in peak oxygen uptake over the same window is a secondary endpoint included in a fixed-sequence testing procedure, and is therefore tested only if the primary endpoint of that family reaches significance. All remaining outcomes are exploratory and are reported with effect estimates and confidence intervals rather than formally tested.

Design and randomisation.

Participants are randomised twice. At baseline, participants are randomised in a 2 to 1 ratio to the six-month programme or to the active control condition. At six months, once the six-month assessment has been completed for that participant and after the last programme contact contributing to it, all participants who were allocated to the programme are randomised 1 to 1 to the conversational agent or to no further support. Participants allocated to the active control condition are not randomised a second time and continue unchanged to eighteen months. Both randomisations are stratified and use blocked allocation sequences; the stratification factors and block sizes are specified in the statistical analysis plan. Randomisation is performed centrally in REDCap.

The second randomisation is unconditional. It does not depend on response, adherence, or any variable measured during the first six months, and the trial therefore does not embed adaptive interventions to be compared. What the design embeds is three fixed sequences, represented by the three arms.

Immediately after the second randomisation, every participant in the programme receives a single transition contact from their own counsellor, communicating the allocation and what happens next. In the arm allocated to the conversational agent, this contact continues into counsellor-led onboarding of the tool; in the arm allocated to no further support, it ends as a non-therapeutic check-in. This contact is not counted as counselling dose.

Setting, participants and recruitment.

The trial is conducted in 15 to 25 general practices within commuting distance of the two study sites, in Prague and in Ostrava. The general practice network is coordinated by the Institute of General Practice, First Faculty of Medicine, Charles University, which contracts and supports the participating practices. General practitioners invite potentially eligible patients from their own practice records. Interested patients register their interest online and are prescreened by telephone. Preliminarily eligible patients are then sampled at random and invited to an entry visit with their own general practitioner, who confirms eligibility, obtains written informed consent and takes blood samples. All remaining assessments take place at the study sites. Recruitment proceeds in four waves of three months.

Intervention.

The six-month programme comprises two supervised group sessions and six individual counselling sessions delivered remotely by video call, based on motivational interviewing and delivered by trained counsellors. The first group session introduces the concept of exercise snacks, sets up the activity tracker and practises technique; the second, two weeks later, works on exercise intensity and execution. Individual counselling is spread across the six months and focuses on embedding activity into the participant's own daily routine. All participants in the programme receive a wrist-worn activity tracker showing daily intensity minutes, calibrated to their individual heart-rate thresholds derived from their own cardiopulmonary exercise test. The target given to participants is at least 150 intensity minutes per week.

From six to eighteen months, participants allocated to maintenance support receive an automated conversational agent based on a large language model, grounded in motivational interviewing and the Health Action Process Approach, delivered through a widely used messaging application and drawing on the participant's own activity data. Participants allocated to no further support receive no programme contact during this period and keep their tracker.

Active control condition.

Participants allocated to the active control condition receive the same wrist-worn activity tracker, which they also keep, together with a short standardised leaflet on physical activity recommendations and on using the tracker. They receive no supervised sessions and no counselling, and attend the same three assessment visits.

Assessments.

Assessments take place at baseline, at six months and at eighteen months. At each assessment, participants wear a research-grade accelerometer on one wrist and a second wrist-worn device on the other for seven consecutive days. No programme contact takes place during the fourteen days preceding an assessment, so that measurement is not reactive.

Sample size.

A total of 306 participants are randomised at baseline. Assuming 10 per cent attrition to six months and 20 per cent to eighteen months, a two-sided alpha of 0.05 and 80 per cent power, the trial can detect a standardised effect size of 0.36 for the first primary endpoint and 0.44 for the second.

Statistical analysis.

Both primary endpoints are analysed on an intention-to-treat basis, using linear mixed models adjusted for the baseline value and prespecified covariates. Both primary endpoints use a treatment-policy strategy for all intercurrent events, including non-adherence, withdrawal from the programme, and initiation of a glucagon-like peptide-1 receptor agonist or bariatric surgery during follow-up. Group assignment is not changed by these events and assessments are sought regardless. The statistical analysis plan for each primary endpoint is finalised before the last data contributing to that endpoint are collected. There is no interim analysis for efficacy and no stopping rule; follow-up to eighteen months proceeds irrespective of the six-month result.