3.1 Sex differences in generalizability of HFrEF trials 87 Statistical analysis Continuous data are presented as mean with standard deviation while categorical variables were reported in absolute and relative frequencies. Mean and proportion differences between each group were calculated and reported as significant based on their corresponding 99% confidence intervals (CI). Unadjusted outcomes were calculated with cumulative incidence curves for each of the 6 subgroups outline above. The competing event for cardiovascular mortality was death from other causes whereas for first HF hospitalization, it was all-cause deaths. To test whether the RCT group was more, less, or equally likely to die than the RCT-eligible group, standardized mortality ratios (SMRs) were calculated and stratified by sex. SMRs were calculated by dividing the observed mortality count in the RCT group by expected mortality count in the RCT group. The observed mortality counts were the actual deaths recorded in the RCTs at one year. In standard SMR analysis, expected counts are the number of deaths that would be predicted if the study population (RCT group) were to have the same age and/or sex-specific rates as the standard population (RCT-eligible group).26 However, one limitation in SMR analysis is the inability to account for case-mix between populations.27 To calculate more precise expected mortality counts in the RCTs, we used a validated prognostic model to apply characteristics of the RCT-eligible group to the RCT group.28–30 We first fitted a Poisson model with 11 prognostic indicators from a validated MAGGIC HF risk score (age, sex, LVEF, NYHA class, serum creatinine, chronic obstructive pulmonary disease (COPD), diabetes, systolic blood pressure, body mass index (BMI), HF duration, smoking status) in a stepwise manner to the RCT-eligible SwedeHF group. Model 1 was the empty model, model 2 included age and sex, model 3 additionally included NYHA class, SBP, and creatinine, and model 4 was fully adjusted with all 11 prognostic variables. Each model with derived coefficients from the RCT-eligible population was then applied to each RCT to derive expected counts. If these prognostic factors and their associated risks were similar between the RCT and RCT-eligible group, then the expected deaths in the RCTs would be equal to the observed deaths leading to an SMR value of 1. Therefore, SMR ratios above 1 indicates that there are more observed deaths in the RCT population than would be expected based on characteristics derived from the RCT-eligible population, and vice versa for SMR ratios below 1.0. The SMRs for all trials were pooled using fixed effect meta-analysis and the corresponding 95% CI was determined using methods
RkJQdWJsaXNoZXIy MjY0ODMw