Statistics

Descriptive continuous data are presented as mean and standard deviation and categorical variables as proportion and number of participants. We used linear mixed models with the outcome HDL-C and secondary outcomes (body weight and Vo₂peak), 1 at a time, as dependent variables. The individual was included as a random effect. We included time, intervention group (CON vs MICT vs HIIT and MICT vs HIIT), and sex and their 2-way (time × group, time × sex, and group × sex) and 3-way (time × group × sex) interactions as categorical covariates. All analyses were adjusted for age and cohabiting status. Because of non-normally distributed residuals, we used bootstrapping with the bias corrected and accelerated method, using 2000 bootstrap samples. The analyses were carried out by the per protocol approach and included participants in MICT and HIIT with 50% adherence to the prescribed exercise program and participants in CON who adhered to the national recommendations for physical activity for the entire duration of the study. We performed subanalyses whereby participants with CVD or using LMAs at inclusion were excluded. Forward linear regression was used to examine the association between changes in Vo₂peak (ΔVo₂peak), fat mass (ΔFM), and body weight (Δweight) and changes in HDL-C (ΔHDL-C) in older men and women. Delta (Δ) values were calculated by subtracting baseline values from the year 5 values. The dependent variable was ΔHDL-C, and ΔVo₂peak, ΔFM, and Δweight were used as independent variables, and we adjusted for smoking status (yes/no).22 A separate model was run for men and women. As some participants changed between treadmill and ergometer bike when performing the Vo₂peak test, these were excluded from the analysis (n=3) to ensure that changes in Vo₂peak were not due to changed modality. A 2-sided P value of less than .05 was considered statistically significant. Because of multiple comparisons, P values between 1% and 5% must be interpreted with caution. All statistical analyses were performed using SPSS version 25 (IBM). Sample size for the main study was calculated on the basis of mortality.15