Fourteen cohort studies were included. Out of 4,564 participants, 1,205 were classified as non-responders and 3,359 as responders. Study size ranged from 60 to 804 participants. Follow-up ranged from 14 days to one year.
Compared to those with an optimal response to clopidogrel, those with a poor response had an increased risk of MACE (OR 5.67, 95% CI 2.97 to 10.84, I2=86%; 14 studies). Removal of four studies that had large risk ratios or that used VASP to assess responsiveness removed heterogeneity (I2=0%), the effect size was slightly decreased but remained statistically significant.
In sensitivity analyses stratified by loading dose, length of follow-up, method of determining non-responsiveness and adjustments for confounders, all results were statistically significant for all groups showing an increased risk in non-responders. Those with lower loading dose, shorter follow-up and use of VerifyNow to assess responsiveness were at greater risk than their counterparts.
After excluding studies that contributed to heterogeneity in the main analysis, a funnel plot showed no evidence of publication bias.