Data analysis

Firstly, two authors (A.J.S., M.A.) independently screened titles and abstracts of studies against the inclusion criteria in a blinded fashion. Potentially eligible articles moved onto the next stage. Second, authors A.J.S. and M.A. independently assessed full texts of the potentially eligible articles for inclusion in the review. Third A.J.S. and M.A. developed a data extraction table including the year and country of publication, study settings, sample size and population, study duration patient demographics, intervention details, control group details, outcome data and attrition rates. A.J.S. and M.A. independently extracted data from each included article. Disagreements at each stage were resolved by discussion with S.M. The methodological quality of included studies was evaluated independently by A.J.S. and M.A. using the Cochrane risk of bias tool⁵⁰ for randomised controlled trials and the Newcastle-Ottawa Scale (NOS) for observational studies⁵¹. Disagreements were resolved by SM. We attempted to contact study authors for unclear or missing information.

Physical activity and exercise capacity measurements were only included in the meta-analysis if they used an objective measurement tool (e.g., a pedometer/accelerometer). Subjective outcome measurements were not included in the meta-analysis.

Where meta-analysis was not possible due to significant heterogeneity, we undertook a narrative synthesis describing the included studies and their risk of bias.

Mean change scores with the corresponding standard deviation (SD) for the outcomes of interest were used in the meta-analysis to obtain the overall effect size, which was presented as either the mean difference or the standardised mean difference (SMD) with a 95% confidence interval. SMD was used where the same outcome of interest was measured by different devices. Where studies had not given the mean change scores, the mean change was calculated by subtracting the post-intervention mean from the baseline mean measure. The SD for changes from baseline was calculated using an imputed correlation coefficient of 0.80 with the following formula, derived from the Cochrane handbook (Eq. (1))⁶⁶:.

Heterogeneity was assessed by I², with a value of ≥50% indicative of significant heterogeneity. If the data were heterogenous, a random-effects model was used rather than a fixed model. All statistical analysis was performed using the Cochrane Collaboration Review Manager software (version 5.4).

To understand the source of heterogeneity between studies, meta-regression analysis was performed on the mean daily step count pooled effect. Five covariates were included: age, publication year, FEV1% predicted, type of wearable used as part of the intervention, and the outcome measurement device. We conducted a mixed-effects meta-regression using Rstudio version 4.2.3. The regression analysis used a Knapp-Hartung modification and model fit was assessed by the Bayesian information criterion.