Background: Published rates of coronary restenosis have fallen below 10% in drug-eluting stent trials. Early evaluations of new stents have used continuous end points that are presumed surrogates for restenosis, but the generalizability and power of such end points have not been examined systematically.
Methods and results: We examined the relationship between incremental changes in observed late loss in lumen diameter and the probability of restenosis using reported late loss from 22 published trials of various types of stents (bare-metal, drug-eluting, and small-vessel stents). Next, the power of late loss differences was compared with that of corresponding binary restenosis rates. The relationship between mean late loss and its SD was linear and did not vary with stent type (drug-eluting or bare-metal) or vessel diameter. At all levels of late loss examined (0 to 1 mm), incremental changes were associated with increasing restenosis risk (with an increasing magnitude of effect at higher levels of late loss). The power to detect a treatment effect was greater for late loss than for binary angiographic restenosis (> or =32% relative increase in power, > or =24% absolute increase for late loss between 0.2 and 0.6 mm).
Conclusions: Late loss is monotonically related to restenosis risk in published stent trials. It is a generalizable and powerful angiographic end point in early or small trials of new drug-eluting stents.