Primary outcome (V̇O2peak)

We conducted a preliminary analysis, using only baseline data, to determine how age and sex affect V̇O₂peak. Previous research has found that V̇O₂max is 27% lower in females compared with males and declines about 10% per decade [50]. We found that V̇O₂peak varied by age and sex, but there was no significant age by sex interaction, so this was not included in subsequent models.

To estimate the effect of the intervention on V̇O₂peak we fit a segmented mixed effect linear regression model. Linear segments were fit for each phase of the study (baseline, intervention and follow-up), with random effects for the baseline intercept and slope. Age and sex were included as model covariates. An intention to treat analysis was used, whereby we analyzed all participants as receiving six months of intervention, as intended. The outcome of interest was the expected change in V̇O2peak over the six month intervention, in excess of what was observed during the baseline phase. This was calculated as 6 × (β_intervention − β_baseline), where β_intervention and β_baseline represent the expected monthly change in V̇O₂peak during the intervention and baseline periods, respectively and the difference represents the change attributable to the intervention. Following the recommendation of Peters et al. [51] missing data were not imputed. Model fit was assessed by examining residual plots and normal Q-Q plots. The complete model is detailed in S1 File. The R statistical programming language [52] was used for all analyses and modelling was performed with the nlme package [53].