In Reply:
We thank Dr Gusmao-Flores for the interest in our recent article “Corticosteroid therapy for severe community-acquired pneumonia: a meta-analysis.”1 We appreciate the opportunity to respond to the concerns raised with regard to our article.
According to Cochrane recommendations, the chi-square test measures the heterogeneity of observed effect sizes from an underlying overall effect. This test has low power in detecting true heterogeneity when studies have a small sample size or are few in number.2,3 Hence, we used a P value of .10 to determine statistical significance. When P was > .10, a fixed-effects model was used.
The Peto odds ratio (OR) method works well when intervention effects are small, events are rare, and the studies have similar numbers in 2 groups.2,4 Bradburn et al4 found that the Mantel-Haenszel OR method using a 0.5 zero-cell correction produces great bias, whereas the Peto method needs no correction. Therefore, mortality was analyzed using the Peto method to calculate the OR and 95% CI in our meta-analysis. Even if the OR was used, the P value from the heterogeneity test was .19 (I2 = 36%), and the fixed-effects model could be used. A significant difference still existed between the corticosteroid and control groups (OR = 0.39, 95% CI 0.16–0.94) (Fig. 1).
Footnotes
The authors have disclosed no conflicts of interest.
- Copyright © 2014 by Daedalus Enterprises