Reply to Jones et al Free

To the Editor—Jones et al assert that our conclusions hinge on exclusion of the 1652 patients recorded as having received neuraminidase inhibitor (NAI) treatment on the day of hospitalization [1]. We did this after consultation with clinicians, on the grounds that, in such patients, the decision to hospitalize and treat with NAI was simultaneous; or, that these patients had deteriorated to a point where there was practically no window of opportunity for NAI treatment to impact on the need for hospitalization. Despite conceding that this “seems reasonable,” these commentators present an alternative analysis that mistakenly considers such patients as treated, arriving at an odds ratio (OR) of 1.97 (95% confidence interval [CI] not supplied) [2]. Placing all 1652 patients into 1 group and assuming that all received preadmission NAI treatment is simply incorrect. We set out to investigate the impact of preadmission NAI use on hospitalization; our exposure variable was “NAI treatment received in the community or in an outpatient setting,” not “NAI treatment at any time” as Jones and colleagues imply with their alternative approach. If we had retained these 1652 patients in our analysis, it would have been more reasonable to classify them as untreated, given that there was no time for the drug to have worked. This would have produced a crude OR of 0.10 (95% CI, .08–.12).

Jones et al further state it is “misleading” to call our study an individual participant data (IPD) meta-analysis [2]. We make it very clear that our article is based on observational data, and the Preferred Reporting Items for Systematic Reviews and Meta-Analyses of Individual Participant Data (PRISMA) statement acknowledges the legitimacy of the IPD approach for observational studies [3]. We have been clear about our inclusion and exclusion criteria and openly discuss the limitations of our approach [1]. It is claimed that we excluded 96% of patients because of missing data [2]. This a misleading assertion, which fails to recognize that the study was part of a cluster of studies, each with different objectives related to investigating the impact of NAI treatment on a range of public health outcomes in patients with pandemic influenza. We therefore, requested raw data for all the studies together, recognizing that not all centers would have data that could contribute on all outcomes. Therefore, although data relating to this study reflect only 2.6% of the total data obtained on pandemic influenza patients in any setting, we have included 67.1% data informative for this specific study question.

Finally, Jones et al compare our findings to 2 studies that investigated the impact of NAI treatment on patients with relatively mild seasonal influenza. We have repeatedly emphasized that our findings are from a cohort of pandemic influenza patients at high risk of hospitalization, and may not be generalizable to a broader spectrum of community patients, most of whom have mild influenza. Jones et al appear to have overlooked the sensitivity analysis we performed in patients from centers with lower rates of hospitalization [1]; these findings are not dissimilar to ones that they cite [2].

Notes

Acknowledgments. This letter was written on behalf of all of the study authors [1].

Potential conflicts of interest. J. S. N.-V.-T. reports that a grant to the University of Nottingham from F. Hoffmann-La Roche funded the current study; he also reports grants to the University of Nottingham from GlaxoSmithKline for research in the area of influenza; and nonfinancial support from the European Scientific Working Group on Influenza to lecture on influenza. P. R. M. reports grants from F. Hoffman La Roche, during the conduct of the study. All other authors report no potential conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

References