Patel and colleagues1 conducted a well-designed unmatched case-control study assessing whether recent use of corticosteroid drugs was associated to complicated varicella in children. The authors' preoccupation in including in their study only children who were immunocompetent except for the use of a corticosteroid drug made them end up with only 3 exposed subjects, 2 case patients, and 1 control subject. As the exposure was very rare (1%) in the study population, the odds ratio estimate became unstable owing to potential random error, and the study did not have enough power to detect a significant association of low magnitude. In fact, the power of this study to detect the observed odds ratio of 1.6 at a significance level of 5% is only around 2%! Therefore, the study was unable to demonstrate an increased risk of complicated varicella associated to recent corticosteroid drug use.
Absence of evidence is not evidence of