(A) Decision tree representing the outcomes in the first 30 days after the index procedure. (B) Decision tree representing the outcomes in the period 30 days to 6 months. The tree for the period 6 months to 1 year is similar. (C) Bubble diagram shows the initial therapeutic decision between drug-eluting stent (DES) and bare-metal stent (BMS) followed by the first year of life after the index procedure. The Markov health states are represented in the ovals with arrows indicating movement between the states from 1 year to next. Each of the health states were further subdivided on the basis of prior myocardial infarction but is not shown for simplicity and brevity. CABG = coronary artery bypass graft surgery; MI = myocardial infarction; ST = stent thrombosis; TVR = target vessel revascularization.

This graph demonstrates the predicted difference in quality-adjusted life expectancy (QALE) between drug-eluting stents (DES) and bare-metal stents (BMS) over a range of excess risk for very late stent thrombosis (VLST) with DES. At an annual risk of 0.14%/year (over 4 years), the predicted QALE was equal for the 2 strategies, and above this level of risk BMS was the optimal treatment strategy.

As the theoretical risk period for stent thrombosis was extended, the annual incremental risk of VLST in DES over BMS that preserved DES benefit decreased. Abbreviations as in Figure 2.

This graph demonstrates the relationship between plausible variations in key model parameters and predicted maximum tolerable excess risk of VLST in DES, below which DES would be preferred over BMS. The model was most sensitive to variations in the stent thrombosis case fatality rate, the relative risks of TVR and late stent thrombosis of DES versus BMS, and the disutility associated with repeat revascularization. *The higher range value of the estimate for this variable results in a smaller threshold value of the VLST in DES and vice versa. CAD = coronary artery disease; QALY = quality-adjusted life years; RR = relative risk; other abbreviations as in Figures 1 and 2.

Under the base case assumptions of annual BMS TVR rate of 14%, the incremental risk of VLST in DES over 3 years was 0.14%/year, over which the preferred strategy would be BMS. As the population risk of restenosis (and TVR) increases over the expected rate of 14%, as observed in diabetic patients, small vessels, and long lesions, the annual incremental risk of VLST in DES over BMS that preserves DES benefit increased. Abbreviations as in Figures 1 and 2.

Current DES reduce target vessel revascularization (TVR) by 65%. If the relative risk (RR) reduction in TVR with DES was lower than 65%, then the annual incremental risk of VLST in DES that preserved DES benefit over BMS was decreased. Abbreviations as in Figures 1 and 2.

This graph demonstrates the impact of the disutility associated with repeat revascularization on the predicted maximum tolerable excess risk of VLST. Abbreviations as in Figures 1 and 2.

The difference in observed yearly rates of VLST for sirolimus-eluting stents (SES) and paclitaxel-eluting stents (PES), versus BMS, are shown as point estimates taken from the recent pooled analyses (confidence intervals not shown) (10,12). The DES versus BMS differences fall just below the VLST threshold we identified (0.14%/year), signifying preserved overall benefit of DES versus BMS; however, a small absolute increase in DES VLST would exceed the threshold above which no net benefit would exist for DES over BMS. Abbreviations as in Figures 1 and 2.