We thank Dr. Radack for his interest in our meta-analysis and for his comments on it. Several issues, however, require correction. Dr. Radack noted that his study [1] included 41 patients. In fact, 45 patients were enrolled, but 41 completed the trial. Importantly, 30 patients who completed this parallel study had been randomly assigned to receive ibuprofen or placebo, and the remainder received acetaminophen. Chalmers and colleagues [2] reported that acetaminophen (called paracetamol in Australia) produced an elevation in blood pressure. Consequently, only the data obtained for the 30 patients receiving ibuprofen or placebo could be included in the meta-analysis, given that acetaminophen could not be considered equivalent to placebo with respect to its effects on blood pressure. The sex and racial distribution included in our meta-analysis [3] reflect the characteristics of these 30 patients.

Radack and Deck [4] claimed that before 1987, only eight double-blind, randomized controlled trials were available for inclusion in a meta-analysis. We have identified 50 such trials in the literature. As noted [3], these small trials were relatively homogeneous in terms of study design, objective (comparing one or more NSAIDs with each other or with placebo), and outcome measure (blood pressure). Not only were these trials relatively homogeneous in a clinical sense, but our statistical test of heterogeneity failed to reach significance, thereby showing their statistical homogeneity. The absence of methodologic heterogeneity facilitated a meaningful pooled mean effect of NSAIDs on blood pressure. The presence of multiple, small, relatively homogeneous trials with differing results has been considered one of the most appropriate scenarios for meta-analysis^[5]. The fact that our meta-analysis involved substantially more trials and included considerably more patients than the review by Radack and Deck [4] lends greater weight to our findings.

Dr. Radack's comment that all articles on NSAIDs and blood pressure provided systolic and diastolic blood pressure measurements is incorrect. Because several articles (14%) provided mean blood pressure values only, we used mean blood pressure as the main outcome to maximize the number of trials included in the meta-analysis and to reduce the chance of bias. However, regardless of the blood pressure measurement chosen for the analysis, the message is unchanged: Nonsteroidal anti-inflammatory drugs elevate blood pressure and antagonize the blood pressure-lowering effect of antihypertensive medication to an extent that may potentially increase hypertension-related morbidity.