Eleven RCTs (n=5,103) were included.
The studies were randomised, with adequate concealment of treatment allocation and clinical follow-up rates of more than 90%. Eight RCTs were double-blinded, two single-blinded and one unblinded.
All DES.
There was no statistically significant difference between DES and BMS in mortality (OR 1.11, 95% CrI: 0.61, 2.06) or myocardial infarction (OR 0.92; 95% CrI: 0.65, 1.25). DES was associated with statistically significantly lower rates of major adverse cardiac events (7.8% with DES versus 16.4% with BMS; OR 0.42, 95% CrI: 0.32, 0.53) and angiographic restenosis rates (8.9% with DES versus 29.3% with BMS; OR 0.18, 95% CrI: 0.06, 0.40).
Sirolimus DES.
There was no statistically significant difference between DES and BMS in mortality or myocardial infarction. Sirolimus DES was associated with statistically significant reductions in major adverse cardiac events (6.8% with DES versus 21.0% with BMS; OR 0.28, 95% CrI: 0.17, 0.41), target lesion revascularisation (3.5% with DES versus 18.5% with BMS; OR 0.15, 95% CrI: 0.02, 0.46) and rate of angiographic restenosis (6.2% with DES versus 36.9% with BMS; OR 0.06, 95% CrI: 0, 0.34).
Polymeric paclitaxol DES.
There was no statistically significant difference between DES and BMS in mortality or myocardial infarction. Polymeric paclitaxol DES was associated with statistically significant reductions in major adverse cardiac events (8.7% with DES versus 6.7% with BMS; OR 0.47, 95% CrI: 0.25, 0.71), target lesion revascularisation (3.3% with DES versus 12.2% with BMS; OR 0.23, 95% CrI: 0.10, 0.42) and rate of angiographic restenosis (7.1% with DES versus 23.5% with BMS; OR 0.23, 95% CrI: 0.07, 0.40).
Non-polymeric paclitaxol DES.
There was no statistically significant difference between DES and BMS in mortality, myocardial infarction, target lesion revascularisation or angiographic restenosis. Non-polymeric paclitaxol DES reduced major adverse cardiac events in comparison with BMS (7.7% with DES versus 9.5% with BMS; OR 0.64, 95% CrI: 0.42, 1.00).
The results were similar after using a range of low-information prior distributions.