Raw Kaplan-Meier curves can be misleading in observational studies: a treatment group might just look worse because it's older or sicker at baseline, not because the treatment itself is harmful. Rino Bellocco (Univ. Milano-Bicocca & Karolinska) currently walking us through fixing this in Stata: from unadjusted KM curves, to Cox-adjusted hazard ratios, to fully standardized survival curves that compare "what if everyone got treated" vs. "what if no one did" — giving a real, interpretable causal contrast on the survival scale, not just a hazard ratio. Presentation available at bit.ly/4yO2Vjf