Demographic and clinical characteristics of the study groups included in the various analyses.

The three virulence analyses could be performed on different, albeit overlapping, subsets of the cohort. In the early years, RNA measurements were not yet available, while recently enrolled patients do not have sufficient data points for the calculation of the CD4 slope. Statistics of the linear regression model of the CD4 slope refer to the subset stripped of outliers below the 5% and above the 95% percentiles; this subset was used in the regression analyses. The mixed-effect model accommodated more patients due to less stringent inclusion criteria. The small number of IDUs in the viral setpoint analysis reflects their diminished proportion in the cohort by the time RNA assays have become available.
IDU, intravenous drug users; MSM, men having sex with men.