Pooling and the Potential Dilution of Repeat Influenza Vaccination Effects

To the Editor—In their recent article, Cheng et al explored repeat vaccination effects in patients enrolled in a hospital-based surveillance system during the 2010–2015 seasons in Australia, concluding no impact or higher vaccine effectiveness (VE) against influenza-associated hospitalization with vaccination in both current and prior seasons, compared to either season alone [1]. Repeat vaccination effects are complex and anticipated to vary with the antigenic relatedness between successive multivalent vaccine components and their corresponding epidemic viruses [2, 3]. Where heterogeneity is expected, pooling across influenza seasons or types/subtypes, as undertaken by these authors, is inappropriate. While it may be tempting to aggregate analyses to increase statistical power, such pooling will mask underlying heterogeneity (Figure 1), diluting repeat vaccination effects and obscuring insights into potential mechanisms and program implications. Hypothetical illustration of how pooling data across seasons or types/subtypes may mask heterogeneity in negative/positive interference from prior season’s vaccination. *Versus those unvaccinated both prior/current season. Abbreviations: V+, number of influenza positive/total vaccinated; V-, number of influenza positive/total unvaccinated. Patterns of positive or negative interference (ie, enhanced or reduced protection) with repeat vaccination are also expected to differ between the Northern and Southern hemispheres given seasonal differences in recommended vaccine components and their relatedness to circulating variants. In the Southern Hemisphere for the 2010–2015 seasons (Table 1), negative interference from prior season’s vaccination should have been pronounced in 2012 following 3 consecutive seasons of homologous A/Perth/16/2009(H3N2)-like vaccine and dominant circulation of A(H3N2) antigenic drift variants [4, 5]. This pattern was somewhat visible in pooled type/subtype findings presented by Cheng et al for 2012, but was utterly obscured in their overall pooled-season results [1]. Conversely, Australia may have been spared the specific conditions associated with historically low VE during the 2014–2015 season in the Northern Hemisphere [3, 6–8], given that A/Texas/50/2012(H3N2)-like vaccine was used in only a single season (vs 2 consecutive seasons) and influenza B viruses instead predominated in 2015 in the Southern Hemisphere. World Health Organization Recommendations for Trivalent Influenza Vaccines and Seasonal Activity in the Southern Hemisphere, 2010–2015 Seasons Component changes to Southern Hemisphere seasonal trivalent influenza vaccines from the prior season are indicated in bold font. Dominant circulating influenza strains for each season are also indicated in bold font. Abbreviations: NNDSS, National Notifiable Disease Surveillance System; WHO, World Health Organization. aA/Texas/50/2012-like viruses when propagated in eggs are considered antigenically similar to the cell-propagated prototype virus A/Victoria/361/2011. World Health Organization Recommendations for Trivalent Influenza Vaccines and Seasonal Activity in the Southern Hemisphere, 2010–2015 Seasons Component changes to Southern Hemisphere seasonal trivalent influenza vaccines from the prior season are indicated in bold font. Dominant circulating influenza strains for each season are also indicated in bold font. Abbreviations: NNDSS, National Notifiable Disease Surveillance System; WHO, World Health Organization. aA/Texas/50/2012-like viruses when propagated in eggs are considered antigenically similar to the cell-propagated prototype virus A/Victoria/361/2011. Although authors attempted to stratify VE analyses by subtype (or season—but not both), subtyping was only performed on about one-third of influenza A specimens. The finding of higher VE in the subtype-stratified analyses compared to the overall influenza A estimate signals selection bias in their findings—a limitation acknowledged by authors. Furthermore, vaccination status was missing for approximately 40% of patients. Partial reliance on medical record review for vaccination history may have introduced misclassification bias if patients with incompletely documented data were more likely to be unvaccinated in at least one season or systematically differed from included patients on key characteristics, as suggested in their Table 1 [1]. Finally, influenza testing was performed at clinician discretion. It is unclear whether a standardized case definition for acute respiratory illness was systematically applied to minimize the differential pretest likelihood of influenza positivity between vaccinated and unvaccinated patients. Given anticipated differences in vaccine responses by influenza season or type/subtype, pooling across these considerations can lead to potentially misleading results—particularly in trying to unravel emerging concerns related to interseason, cross-strain interactions and their variable impact. Ultimately, sum total benefit does not negate the importance of potentially deleterious effects of serial vaccination some seasons. We therefore urge caution in the interpretation of repeat vaccination effects predicated on pooled analyses, or their generalization to guide or reinforce program recommendations pending more suitably stratified investigation. Potential conflicts of interest. All authors: No reported conflicts of interest. Both 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.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.