In the Insulin Resistance Intervention after Stroke (IRIS) trial, pioglitazone (PIO) reduced the risk of fatal/nonfatal stroke/MI by 24% vs. placebo (PLA) in 3876 nondiabetic patients with insulin resistance (IR) and recent stroke or transient ischemic attack (TIA). We sought to determine which factor(s) modified by PIO were associated with its CV benefit.
The effect of PIO was quantified using Cox models. Starting with a bivariate model, change from baseline at year 1 in a single factor (Table) was included as a covariate for stroke/MI events beyond year 1. The resulting percent change in log HR for PIO towards the null was defined as the percent mediation of PIO’s effect by that factor. The combined effect was assessed in a model using all factors identified in bivariate models as potential mediators.
The factor with the largest impact (26%) was HDL-C, with smaller mediating effects (<15%) from CRP, TG, BP and glucose. Together, these factors accounted for 43% of the treatment benefit. We could not confirm any mediating effect from the reduction in measures of IR. Findings were similar in several sensitivity analyses.
In summary, no single factor explained the majority of PIO’s CV benefit in IRIS. PIO’s favorable effects in patients with IR and stroke/TIA is likely mediated through multiple and combined benefits on lipids, BP, glucose, and inflammation.