Search
Close
Search
Advanced Search
Search Menu
AI Discovery Assistant

Extract

To the Editor—We appreciate the interest in our study [1] by Mohsen and McKendrick [2]. We find their results interesting and believe they add useful clinical information to the literature on this topic. Nonetheless, the 2 study cohorts are not alike in some important ways

Mohsen and McKendrick [2] reported that 38 women fulfilled the criteria for their study: 19 had radiologic evidence of pneumonia, and 19 were control subjects who had varicella but no pneumonia. In our study, all 18 case subjects were pregnant women with pneumonia who had, by definition, radiologic evidence of pneumonia, in addition to symptomatic dyspnea. We compared these women with 72 control subjects, pregnant women with varicella but no evidence of pneumonia. In our study, the source of varicella was the patient's own child for 46 subjects (51.1%; P=.589, by χ2 test in univariable analysis). Whereas Mohsen and McKendrick assert that an association exists between “any recognized contact with chickenpox” and development of pneumonia (odds ratio, 4.8; 95% confidence interval, 1.0–22.1), we found no known contact with varicella in 27 (30%) of our 90 subjects (P=.145). Obviously, neither of these associations was significant in our multivariable model. With our group of 90 subjects, we were able to perform a multivariable regression analysis to control for confounding variables. In contrast, in the full text of their article in the journal Thorax [3], Mohsen and McKendrick concede that they had insufficient sample size for this essential statistical step. We wonder whether their univariable analysis might be misleading and invalid due to the presence of confounding variables for which they could not account, whereas our analysis was more complete

Issue Section:
Correspondence
You do not currently have access to this article.
Download all slides