[Skip to Content]
[Skip to Content Landing]
Views 274
Citations 0
Editor's Correspondence
June 24, 2013

Patient-Level vs Group-Level Data to Adjust Meta-analysis on Transfusion and Mortality—Reply

Author Affiliations

Author Affiliations: Division of Cardiology, Department of Medicine, Brown University and Providence Veterans Affairs Medical Center, Providence, Rhode Island (Dr Chatterjee); Copenhagen Trial Unit, Centre for Clinical Intervention Research, Rigshospitalet, Copenhagen, Denmark (Dr Wetterslev); and Division of Cardiology, Texas Tech University, El Paso (Dr Mukherjee).

JAMA Intern Med. 2013;173(12):1157-1158. doi:10.1001/jamainternmed.2013.6852

In reply

We thank Drs Jolicœur and Tuinman et al for their interest in our article and thoughtful comments. Indeed, reporting of both adjusted and unadjusted analyses have been recommended in prior publications,1 and we have adhered to the same practice. The meta-analysis on mortality using adjusted estimates reported by the individual studies (assuming constant hazards for individual studies over the length of the follow-up period)2 has been published as an online supplement.3 A test for statistical interaction4 performed to identify differences between the adjusted and unadjusted risk estimates did not identify any significant differences between the 2 obtained estimates (interaction odds ratio, 1.29; 95% CI, 0.96-1.75) (P = .10). Furthermore, a diversity-adjusted study sequential analysis indicated a low risk of random error in our findings; however, increased risk of bias cannot be excluded.5 Shishehbor et al6 reported adjusted and unadjusted mortality outcomes for transfused and nontransfused patients with ST-segment elevation myocardial infarction, as reported in our article3; however, the citation in the original article3 is inaccurate. We thank Dr Jolicœur for identifying the same. We thank Tuinman et al for postulating further possible pathophysiologic mechanisms for explaining the association seen with blood transfusion and mortality, as enumerated in our article,3 and further research on the pathophysiologic basis for the clinical phenomena appears essential.