Background: Development of new therapeutics for severe influenza disease is a priority considering emerging resistance to neuraminidase inhibitors (NAIs). However, development of effective non-NAI therapeutics has been hampered by the lack of robust clinical endpoints and is the focus of a newly formed DHHS and academic working group. This retrospective pilot study evaluated ordinal scale endpoint as part of an US DHHS and academic working group effort.
Methods: 215 patients who received NAIs within 24 hours of admission and who had a baseline National Early Warning System score > 3 were selected from collected data from 703 adult patients admitted to Northwestern Memorial Hospital between April 2009 and March 2014 with PCR-confirmed influenza illness for assessment of the candidate ordinal scale. Components of a novel ordinal scale endpoint, in order of decreasing severity included: death > ICU with mechanical ventilation > ICU w/o mechanical ventilation > hospital floor > hospital discharge. Each patients most severe outcome was recorded daily for 14 days following admission.
Results: Based on time of influenza symptom onset to initial NAI treatment, these patients were divided into 3 groups: < 48 hrs. (N= 58), > 48 to < 96 hrs. (N=77), and > 96 hrs. (N=80); demographics are outlined in Table 1. Figure 1 demonstrates the relative frequency distribution of ordinal components for the three groups. Figure 2 demonstrates the mean daily score assuming a unit decrease from 5 for death to 1 for discharge. Both figures visually indicated a more rapid transition to progressively less severe outcomes in patients with earlier NAI treatment.
Conclusion: The ordinal scale endpoint revealed positive treatment effects attributable to earlier NAI treatment, consistent with published data on NAI use for severe influenza disease. Future studies will need to assess the optimal analysis methodology, which could include ordinal logistic regression and longitudinal analysis techniques, and other ordinal endpoint. Prospectively collected data should also be used to validate this endpoint as well.
J. Katzen, None
Y. Gao, None
J. Houk, None
M. Willis, None
J. Tegeris, None
M. Ison, None