In infants born at 29 to 32 weeks’ gestation, Farber et al1 conclude that palivizumab dispensing reduces respiratory syncytial virus (RSV)-diagnosed hospitalizations but increases hospitalizations due to bronchiolitis without RSV diagnosis. I disagree with the interpretation of study results provided by the authors. By using data reported in Table 2 and Table 5 of the article, I prepared a new table (Table 1). In Table 1, I have collapsed “no palivizumab” and “1–25% eligible doses dispensed” groups into a single category because they did not substantially differ in the frequency of RSV-diagnosed and RSV-undiagnosed admissions. The total, observed admissions (with and without RSV diagnosis) according to palivizumab dispensing are reported in the second column. If no or low (1%–25%) palivizumab eligible dosages were adopted in all infants, we would observe the same frequency of RSV-diagnosed hospitalizations (ie, 5.14%) throughout all infant groups. As a consequence, the total admitted infants (with and without RSV-diagnosed admissions) would be 8.37% (instead of 7.6%) in the group “30%–50%,” 7.51% (instead of 4.7%) in the group “60%–75%,” and 9.66% (instead of 5.9%) in the group “80%–100%.” Thus, increasing palivizumab doses dispensed is associated with a dose–response reduction in observed total admissions as compared with those expected in the case of no or low (1%–25%) palivizumab eligible doses. This result overturns the interpretation of study findings provided by the authors. Therefore, the increase in admissions without RSV diagnosis does not seem to be an adverse effect of palivizumab, but it could be the expression of “confounding by indication” (patients with more severe disease are more likely to be treated with higher dosages and then more likely to experience hospitalizations for causes other than RSV).
Furthermore, for data in Table 4 of the article (infants born at 29–32 weeks’ gestation), there is a mistake in the P value calculation when RSV-undiagnosed hospitalizations are compared between 0 and >1 doses dispensed. The correct P value is not .05 (as the authors reported) but .061. Thus, there is “no significant increase in non-RSV hospitalization” in infants with palivizumab dispensing >1 as compared with remaining infants. Finally, to see whether a given protective drug effect is counterbalanced by a negative one, the likelihood of being helped or harmed2 should be appropriately calculated (Table 2) as the ratio between the number needed to harm and the number needed to treat.
As reported in the last column (Table 2), the analysis in terms of likelihood of being helped or harmed shows that palivizumab treatment is 36% more likely to help (in terms of reduction in RSV-related hospitalizations) than to harm.
- Accepted October 19, 2016.
Conflict of Interest: Dr Tripepi has received honoraria from Abbvie and Biotest.
- Copyright © 2017 by the American Academy of Pediatrics