Beyond Antigenic Match: Moving Toward Greater Understanding of Influenza Vaccine Effectiveness

influenza, vaccine effectiveness, H1N1pdm09

(See the major article by Skowronski et al, on pages 1487–500.)

“We have not reached solutions; we have only begun to discover how to ask questions.”

-Lewis Thomas

In the not-too-distant past, influenza vaccine effectiveness (VE) was a simple concept. If the vaccine and the viruses were a good match, then effectiveness was high. If not, then effectiveness was low. No one really knew how high or how low, because it was not feasible to routinely measure effectiveness against laboratory-confirmed influenza after a vaccine was licensed. That began to change in 2005 when Canadian investigators published the first study of influenza VE in the outpatient setting using the “test-negative design” [1]. Over the past several years, VE studies have become increasingly sophisticated to address deeper questions. We now know that VE can vary by subtype/lineage, genetic clade, product type, and prior season vaccination status [2–7]. Yet the underlying mechanisms involving host immune response, virus characteristics, and vaccine composition remain poorly understood.

In the current issue of the Journal of Infectious Diseases, Skowronski and colleagues [8] add several new pieces to this puzzle with a thorough analysis of VE in Canada during the 2015–2016 season. This was a late-arriving season dominated by H1N1pdm09 and B/Victoria. To underscore the importance of factors beyond antigenic match, VE was higher for the lineage-mismatched B/Victoria (54%) than it was for the antigenically matched H1N1pdm09 (43%), although confidence intervals (CIs) overlapped. VE against H1N1pdm09 was the lowest in Canada since the 2009 pandemic, and substantially below the VE of 71% in the 2013–2014 season. VE against H1N1pdm09 was nearly identical in the United States and Canada during 2015–2016, although there was only a modest decline in the United States from 2013–2014 (54%) to 2015–2016 (45%) [9, 10].

VE against B/Victoria was also similar in Canada and the United States despite widespread use of trivalent vaccine containing B/Yamagata in Canada [10]. Prior VE studies in 2007–2008 (Canada) and 2012–2013 (United States and Canada) demonstrated significant cross-lineage protection [11–13], and this phenomenon was seen in a randomized clinical trial [14]. However, cross-lineage protection is not consistently observed in all seasons or populations [15, 16]. The 2 influenza B lineages are genetically and antigenically distinct, and the immune mechanism for cross-lineage protection is unknown. A study of B lineage priming effects in infants and toddlers demonstrated that priming with B/Yamagata led to a dominant Yamagata response after later immunization with B/Victoria lineage [17]. The B/Victoria antigen strongly recalled antibodies to B/Yamagata and elicited a low antibody response to B/Victoria. The reverse sequence (B/Victoria priming followed by B/Yamagata immunization) has not been assessed in children. However, in mice primed with 2 doses of B/Victoria antigen, administration of B/Yamagata generates a substantial antibody response to both lineages [18]. The temporal pattern of lineage exposure may play a role in cross-lineage protection and merits further research.

Repeated annual vaccination has been associated with reduced VE in some populations and seasons [3]. In Canada, VE against H1N1pdm09 was substantially lower (41%) among persons who were vaccinated in both current and prior season compared with those vaccinated in 2015–2016 only (75%). The odds of vaccine failure (polymerase chain reaction–confirmed H1N1pdm09) was 2-fold higher in repeat than in single-season vaccinees. No repeated vaccination effects were seen for B/Victoria. In the United States, there was no evidence of repeated vaccination effects for H1N1pdm09 during 2015–2016 [10]. According to the antigenic distance hypothesis, repeated vaccination can reduce VE when vaccine 1 (v1, prior season vaccine) and vaccine 2 (v2, current season vaccine) are antigenically equivalent to each other and antigenically distinct from the current season circulating viruses [19]. This effect was originally called “negative interference,” but “negative antigenic interaction” has been suggested as a better description because the role of antibody interference is unclear [20].

On the surface, the Canadian results seem surprising given that the H1N1pdm09 vaccine strain was the same in 2014–2015 and 2015–2016 and matched to circulating viruses based on HI antigenic characterization with postinfection ferret antisera. However, the authors highlight the genetic evolution of H1N1pdm09 viruses with recent epitope substitutions and glycosylation changes that might affect antigenicity. The antigenic distance hypothesis remains a useful framework to help understand repeated vaccination effects, but its predictive utility is limited by the inconsistent findings across studies. This may improve with increasing capacity to predict antigenic effects based on genetic sequencing.

Prior VE studies in the United States and Europe have found evidence of modest waning of vaccine protection during a single season with increasing time since vaccination [21–24]. Interpretation of these findings has been challenging because bias/confounding may occur with changes in the at-risk population between early and late season. Within-season waning of protection was examined in Canada during 2015–2016, and adjusted VE against H1N1pdm09 declined significantly from early season (January–February, 62%) to late season (March–April, 19%). This was not driven by age group differences. The difference between early and late season VE against H1N1pdm09 in Canada is generally consistent with results from the U.S. Flu VE Network over 4 seasons [22]. VE against B/Victoria did not decline from early to late season in Canada, in contrast to findings from the United States in 4 previous seasons. Waning protection within a season is not explained by antibody decay, because the estimated time to reach a 2-fold titer reduction is >600 days [25].

Skowronski et al [8] also explored immunological cohort effects for H1N1pdm09 protection, including the potential effects of K163 epitope specificity. Beginning in 2013–2014, there was widespread circulation of a new phylogenetic group (clade 6B) with a K163Q mutation (K166Q using H3 numbering scheme) in the Sa antigenic site [26]. Further genetic evolution led to the emergence of clade 6B.1 viruses in 2015–2016 with additional mutations, including glycosylation changes (S162N) that could shield the K163 epitope from vaccine-induced antibodies.

Results of a prior study by Linderman and colleagues [26] suggested that persons born from 1940 to 1984 (and especially those born in 1965–1979) may have impaired responses to H1N1pdm09 viruses with the Q163 mutation. These authors hypothesized that vaccination with H1N1pdm09 antigen may have reinforced (focused) the original response against K163 epitope in certain age groups who were infected with K163 viruses in early life. This would occur at the expense of de novo responses to current Q163 viruses. The analysis of VE by age group in Canada was generally consistent with this hypothesis, but there was insufficient power to distinguish subtle differences across age groups. The authors also note that vaccine-induced protection in different age groups might be influenced by the first influenza infection in childhood, especially if it was caused by a different subtype (heterosubtypic imprinting). The interaction of imprinting with later vaccine response is poorly understood, but there is epidemiologic evidence that imprinting may enhance lifelong protection against influenza A viruses of the same phylogenetic group [27].

In age-stratified analyses, 2015–2016 VE against H1N1pdm09 (hemagglutinin [HA] group 1) in Canada was only 30% (95% CI, −11% to 56%) among persons born in 1957–1976, an age group in which most individuals were primed with either H2N2 (HA group 1) or H3N2 (HA group 2). In contrast, VE was 63% (95% CI, 9%–85%) in young adults who were born in 1986–1999 when H1N1 viruses were cocirculating with H3N2 [8, supplement 15]. A model of 2015–2016 VE by year of birth demonstrated a U-shaped pattern with the lowest VE in persons born around 1967, although CIs were very wide [8, fig 4B]. These intriguing findings emphasize how little we know about the role of heterosubtypic priming, epitope focusing, and other factors that may contribute to differences in VE across age groups.

Skowronski et al [8] clearly acknowledge the limitations of these exploratory analyses. Theirs was an observational study, and the power to detect age-stratified effects was low. Interpretation of birth cohort effects is particularly difficult because year of birth is an imprecise surrogate for imprinting subtype and epitope specificity. An additional limitation of the repeated vaccination analysis is the reliance on self-report and sentinel physician documentation to classify vaccination status in prior seasons. Although many persons habitually receive the vaccine (or not) each season, there is potential for misclassification due to poor recall. Ideally, it would be helpful for VE studies to independently validate physician-reported or patient-reported vaccination status for at least a subset of participants.

Antigenic match remains an important determinant of VE, but antigenic match alone does not explain why influenza vaccines are predictably less effective than vaccines against other viral infections. It is only one part of a complex web of interactions that include genetic evolution of circulating viruses, gain or loss of glycosylation at key antigenic sites, early childhood imprinting, epitope specificity, egg-induced mutations in high-growth vaccine viruses, and the individual immune “landscape” generated by a lifetime of repeated vaccination or infection. Despite decades of influenza vaccine research, we still do not have a comprehensive understanding of these interactions and how they influence vaccine protection. Advances in these areas will be essential to achieve the holy grail of influenza prevention: a broadly protective “universal” vaccine that generates strong and sustained protection against seasonal and pandemic viruses. From this perspective, the era of antigenic match is over and the era of influenza virus-host immunoepidemiology is just beginning.

Acknowledgments. Thanks to Brendan Flannery, Thomas Friedrich, Alicia Fry, and Huong McLean for their review and comments.

Potential conflicts of interest. E. B. has received prior research support from Medimmune. The author has 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.