*2.4. Covariates*

Data on age, sex, body fat percentage, height and mobility were collected at all assessment phases. Barefoot standing height ( ±0.1 cm) was measured using a wall-mounted stadiometer [26]. Measurements of body fat percentage were obtained from whole body DXA scans. Mobility was self-reported and divided into seven categories ranging from "very active" to "bedfast". For these analyses, two categories of mobility were considered; sedentary (included "sedentary", "limited", "inactive", "chair or bedridden" and "bedfast") and active (included "very active" and "active").

#### *2.5. Statistical Analyses*

All statistical tests were performed using Minitab 17 (Minitab, LLC, State College, PA, USA). The DII was analysed as a continuous variable. Kolmogorov–Smirnov test was used to investigate normality of the data. Independent sample *t* test was used to compare continuous characteristics between sex or other dichotomised factors. If necessary, a non-parametric Mann–Whitney U test was used for this purpose, and a Chi-square test was used for categorical variables. The natural log-transformation was used to normalise TUG scores (used to assess muscle function), which were positively skewed.

Separate linear regression models were used to examine the association between DII and muscle mass and muscle function. A logistic regression model was used to examine the association between DII score and these components combined. Bivariable regression models with no adjustment for participant characteristics were presented (model 1), followed by multivariable regression models that accounted for age (years), sex (male/female) and body fat percentage (%) (model 2). Further adjustments were made for mobility (active/inactive) for ALM/height2, and height (m) for ln (TUG). Interaction between co-variables were tested and retained in the final model (model 3) if the interaction term was statistically significant (*p* < 0.05). To test for interaction terms, DII was dichotomised according to the median. Daily nutritional energy intake was not included in the multivariable models as a covariate because energy is already included as a constituent of the DII [36]. Results are presented as standardised beta coefficient (β) and standard error (SE), or as an odds ratio (OR) and 95% confidence interval (95% CI).
