In the comment by Dr Tripepi, “Alternative Explanation of the Results,” I respectfully disagree with assertions that we misinterpreted our findings.
Our specific aim was to determine the different effects of palivizumab on hospitalizations with an RSV diagnosis and hospitalizations without an RSV diagnosis. We clearly showed differences and a dose–response effect.1 It appears Dr Tripepi would like an analysis of pooled hospitalization rates (hospitalization with RSV diagnosis plus hospitalization for bronchiolitis without an RSV diagnosis) for infants born at 29 to 32 weeks’ gestation to determine whether there is a net benefit or harm.
When we sum RSV and non-RSV hospitalizations (Tables 3 and 4 of the article) by receipt of any palivizumab, we find a 6.4% hospitalization rate for those who were dispensed ≥1 palivizumab doses compared with 6.9% for those who were not dispensed any palivizumab. This small difference is not statistically significant (P = .7). This analysis would be appropriate because the real-world intervention is prescription of palivizumab (Table 3).
For his analysis, Dr Tripepi combined the “no palivizumab” and the “1%––25% of eligible doses dispensed,” claiming that the groups did not differ substantially. This claim is incorrect. The RSV hospitalization rate differed by 74% (4.97% vs 6.72%) between these 2 groups, hardly a trivial difference. The relevant clinical question is how a palivizumab prescription to which the patient is adherent compares with no palivizumab. It appears that in pooling the data, Dr Tripepi simply summed the percentages. This is not a correct strategy because the “RSV hospitalization: no” group included the patients who had a hospitalization for bronchiolitis without an RSV diagnosis, and the “non-RSV bronchiolitis hospitalization: no” group included those who had a hospitalization with an RSV diagnosis.
To address the concerns raised by Dr Tripepi, I compared the best-case scenario: ≥80% of eligible palivizumab doses dispensed versus no palivizumab dispensed. We found a 5.93% hospitalization (RSV + bronchiolitis without RSV diagnosis) rate for those who received ≥80% of palivizumab doses compared with a 6.90% hospitalization rate for those without palivizumab. The small difference is not statistically significant (P = .5) (Table 4).
Dr Tripepi claims that the P value for Table 4 of the article is not correct, stating it should be .06 instead of .05. Although results may differ by statistical test used, we chose to use the χ2 test for these analyses. I reanalyzed these data and confirmed that the P value is .049 by χ2 test. The more conservative Fisher test yields a slightly higher P value at .06.
In conclusion, I stand by our results and conclusions. Pooling data on hospitalization with RSV and bronchiolitis without an RSV diagnosis revealed only small differences by palivizumab administration status that were neither statistically significant nor clinically important.
- Accepted October 19, 2016.
Conflict of Interest: None declared.
- Farber HJ,
- Buckwold FJ,
- Lachman B, et al
- Copyright © 2017 by the American Academy of Pediatrics