From the annual scientific assembly of the American College of Cardiology, Orlando
Many of the biggest ooh’s and ah’s at this meeting were directed not at new blockbuster megatrial findings–those were in limited supply–but rather at groundbreaking study methodologies so dazzling even nonstatisticians were awed. It’s been a rare chance for the statisticians to take center stage and accept accolades for their work, the most extraordinary of which was hailed as “robust”–among the highest possible praise in the typically understated medical lexicon.
A case in point drawn from interventional cardiology: A comparison of 3-year clinical outcomes of drug-eluting and bare metal stents in 262,700 (!) Medicare patients, far and away the largest-ever stent study. This behemoth, commissioned by the FDA and Agency for Healthcare Research and Quality, was accomplished by heroically linking the ACC’s National Cardiovascular Data Registry with the Medicare patient database.
Most remarkably, the investigators sidestepped insurmountable confidentiality issues by tracking outcomes without any access to patient names or other individual identifiers. How did they do it? By coupling birthdates and other indirect identifiers with super-sophisticated clustering methods reliant upon inverse probability weighted estimators incorporating 102 covariates.
Statistical bling was also in evidence on the surgical side, most notably in the 1,000-patient Surgical Treatment for Ischemic Heart Failure (STICH) study, the largest-ever randomized trial comparing two surgical strategies. Its partisans hope STICH’s clear findings will help convince surgeons to move beyond their traditional reliance on anecdotal evidence and case series into the modern era of comparative effectiveness research.