RT Journal
A1 Osnabrugge RJ, Kappetein A, Janssens AW
T1 CArriage of reduced-function cyp2c19 allele among patients treated with clopidogrel
JF JAMA
JO JAMA
YR 2011
FD February 2
VO 305
IS 5
SP 467
OP 468
DO 10.1001/jama.2011.77
UL http://dx.doi.org/10.1001/jama.2011.77
AB 
First, the authors used an inappropriate strategy to reduce the substantial heterogeneity in their meta-analysis (I2 = 73%). They removed the 2 studies that caused heterogeneity (AFIJI2 and FAST-MI3) and found that the summary HR remained significant (HR, 1.42; 95% CI, 1.19-1.69; I2 = 0%). Removing the AFIJI study seems justified since the mean age, smoking status, and sex of the patients were substantially different compared with those in the other studies.2 For example, the mean age of patients in the AFIJI study was 40.1 years, whereas the overall mean age across all studies was 64.2 years. Yet excluding the FAST-MI study cannot be justified. The study is very comparable and should not be removed for the mere fact of introducing heterogeneity.3 Second, the authors did not report on the assessment of potential bias that may have impacted their results. It is known that meta-analyses are subject to bias, including first-study, selective reporting, and publication bias.4- 5 The presence of bias can be examined using several plots and tests that are available. One indication for the presence of publication bias is a difference in effects between the large and small studies.5 According to Figure 2A and 2B of the article, the smallest effects were observed in the 4 largest studies and 3 of these 4 were not statistically significant. It would be interesting if the authors compared the summary HR of the 4 smallest and 4 largest studies.