Figure 3

Forest plots showing effect estimates and 95% confidence intervals of metabolites that were statistically significantly (FDR q-value < 0.05) related to self-reported time spent on moderate-to-vigorous physical activity (MVPA; (a)) and adherence to the physical activity guideline (≥ 150 min/week of MVPA; (b)) among colorectal cancer survivors, between 6 weeks and 2 years post-treatment, including overall, inter-individual and intra-individual longitudinal associations. Asterisk (*) denotes statistical significance after FDR-adjustment (q-values < 0.05 were considered significant). Since light-intensity physical activity (LPA) was not statistically significantly associated with any of the metabolites, these results are not shown. Full results for all physical activity variables and metabolites are included in Supplementary Table 5. AAs amino acids, ACs acylcarnitines, BAs biogenic amines, MVPA moderate-to-vigorous physical activity, PCs phosphatidylcholines. Analyzed with multivariable linear mixed regression models analyzing associations of the physical activity variables as the main independent variables and as dependent variables the batch-adjusted metabolite residuals (see “Methods” section), with a separate model for each metabolite. Models were adjusted for: sex; age (y; continuous), time since treatment (per 6 months; continuous), centre (Maastricht UMC + ; VieCuri Medical Center; Zuyderland Medical Centre), body mass index (kg/m²; continuous), smoking status (current; former; never), self-reported alcohol consumption (grams/day) and number of comorbidities (no comorbidity; 1 comorbidity; ≥ 2 comorbidities), at post-treatment time points. Models were also adjusted for post-treatment levels of MVPA for the analysis of LPA, and vice versa. The percentage difference in metabolite concentrations was calculated by subtracting the exponent of the obtained regression coefficient from 1 and multiplying the outcome with 100 (since metabolite concentrations were ln-transformed).