December 20, 1985

Comparing Antibiotics: Statistical Rigor vs Practicality

Author Affiliations

National Institute of Allergy and Infectious Diseases Bethesda, Md

JAMA. 1985;254(23):3311. doi:10.1001/jama.1985.03360230041014

To the Editor.—  Lin et al1 have concluded that seven days of ceftriaxone therapy is as effective as ten days of therapy in children with bacterial meningitis, based on the failure to find statistically significant differences in various measures of outcome between two randomized groups of 35 children each. While this conclusion may be correct, the evidence reported is not sufficient to support it. To illustrate, let us consider one of their findings: in each of the two treatment groups, four (11.4%) of 35 children had two or more neurological complications. This result suggests that the two treatments may have a similar effect. However, a 90% confidence interval for the difference between the true percentages is —12.5% to 12.5%; a statistical test at a two-sided 10% significance level would therefore not reject the hypothesis that the true percentages differ by as much as 12.5%. Thus, the study was too