Instrumental variable analysis to compare effectiveness of stents in the extremely elderly.

BACKGROUND: Evaluating novel therapies is challenging in the extremely elderly. Instrumental variable methods identify variables associated with treatment allocation to perform adjusted comparisons that may overcome limitations of more traditional approaches. METHODS AND RESULTS: Among all patients aged ≥85 years undergoing percutaneous coronary intervention in nonfederal hospitals in Massachusetts between 2003 and 2009 (n=2690), we identified quarterly drug-eluting stent (DES) use rates as an instrumental variable. We estimated risk-adjusted differences in outcomes for DES versus bare metal stents using a 2-stage least squares instrumental variable analysis method. Quarterly DES use ranged from 15% to 88%. Unadjusted 1-year mortality rates were 14.5% for DES versus 23.0% for bare metal stents (risk difference, -8.5%; P<0.001), an implausible finding compared with randomized trial results. Using instrumental variable analysis, DES were associated with no difference in 1-year mortality (risk difference, -0.8%; P=0.76) or bleeding (risk difference, 2.3%; P=0.33) and with significant reduction in target vessel revascularization (risk difference, -8.3%; P<0.0001). CONCLUSIONS: Using an instrumental variable analysis, DES were associated with similar mortality and bleeding and a significant reduction in target vessel revascularization compared with bare metal stents in the extremely elderly. Variation in use rates may be useful as an instrumental variable to facilitate comparative effectiveness in groups underrepresented in randomized trials.