Reply to: ESTIMATING THE FULL VALUE OF THE HIGH-DOSE INFLUENZA VACCINE

To the Editor

We appreciate the interest from Abbvie Pharmaceuticals and Sanofi Pasteur as expressed by Drs. Ertmer and Chit in a response to our policy recommendations for influenza vaccination strategies in the elderly.^2,3 They state that, it is “difficult to use models to appropriately predict the full public health impact of HD-IIV3.” We agree. However, when population-level data are sparse, models are an excellent option to test the potential impact of alternative policy strategies. Though Ertmer and Chit agree that high-dose trivalent influenza vaccine (HD-IIV3) is currently the most economically favorable vaccination strategy for US seniors, they argue that the findings of our model are excessively conservative. In comparing our generalized cost-effectiveness model² with a similar model derived from experimental results¹, Ertmer and Chit contend that our results underestimate potential savings from averted hospitalizations. We believe, however, that the results reported from our National Institutes of Health funded project appropriately reflect the caution required as a potential influence on national influenza policy decisions while responsibly reflecting available data.

Interestingly, the similarities between the two models outweigh differences. Though structural assumptions are cited as a weakness, both models operate on the same structural phenomenon and utilize similar parameters and parameter values. The notable exception driving the differing results is not averted hospitalizations, as Ertmer and Chit suggest, but rather the estimated effectiveness of standard dose IIV3. Contrary to Ertmer and Chit’s critique, our estimate for flu-related hospitalizations specifically captures both cardiorespiratory hospital admissions and pneumococcal disease. The derivation of this parameter is described by Molinari et al.⁴ who analyzed flu-related ICD-9 and ICD-10 codes as well as excess hospitalization and mortality rates potentially attributable to seasonal influenza. In addition to correctly accounting for all flu-related hospitalizations, we believe this surveillance-based analysis to be a more reliable estimate of historical trends than the single randomized trial results proposed by Ertmer and Chit. Because both models utilize similar cost and effectiveness parameters, differences in the estimation of vaccine effectiveness (VE) can be highlighted as the primary driver of the more conservative results.

Because of the extraordinary adaptability of the influenza virus, VE can change radically from year to year. Both models anchor the VE of the alternative strategies to the effectiveness of standard dose IIV3 to standardize the comparison of alternate vaccination strategies across a variable range of annual scenarios spanning hypothetical seasons with a poor serotype match to those with an excellent match. In this way, the relative effectiveness of IIV4 and HD-IIV3 compared with standard dose IIV3 remain constant while allowing for sensitivity testing across a range of values of standard dose IIV3 effectiveness. Therefore, the estimation of standard dose IIV3 effectiveness heavily influences the range of population outcomes. As standard dose IIV3 effectiveness increases, so does the number of averted cases of disease.

Chit et al¹ computed cost-effectiveness outcomes using a mathematically derived aggregate measure whose methodologic assumptions were challenged by Beyer et al.⁵ Moreover, the value of 49% IIV3 VE is considerably higher than what has been observed in recent history. The CDC’s U.S. Flu Vaccine Effectiveness Network surveillance data from the prior three seasons (2011–2014) report VE in individuals ≥65 years as 43%,⁶ 32%,⁷ and 39%.⁸ Our model used the median value of 39% and computed a probabilistic sensitivity analysis for VE values ranging from 0% to 65%. Under these conditions, HD IIV3 remains cost-effective but not cost-saving.