Untangling the Effects of Prior Vaccination on Subsequent Influenza Vaccine Effectiveness

Influenza A(H1N1)pdm09, Influenza Vaccine, Repeat Vaccination, Vaccine Effectiveness.

(See the major article by Martínez-Baz et al on pages 847–5.)

Influenza vaccines are currently our best tool for the prevention of the substantial morbidity and mortality resulting from annual influenza epidemics. In the United States, annual vaccination recommendations began in the 1960s and have gradually expanded in scope so that, as of 2010, annual vaccination is recommended for all individuals ≥6 months of age [1]. Many other countries also have recommendations for vaccination, which are more limited in scope and are generally focused on high-risk groups, such as elderly individuals and people with chronic health conditions. Although the populations recommended to receive vaccine vary by country, a common characteristic of all immunization policies is that annual vaccination is recommended. Annual vaccination is necessary because of antigenic drift among viruses that circulate each year and waning protection over time.

It has long been known that the immunologic response to influenza vaccination may be modified by past influenza virus infection. However, the effect of past vaccination on the ability of subsequent vaccination to prevent infection has been less clear. In his classic article, “On the Doctrine of Original Antigenic Sin,” Thomas Francis noted that the antigenic properties of the influenza virus with which a child is first infected can shape their response to exposures later in life [2]. Vaccination was actually proposed by Francis as the way to induce broadened immunity against current strains and previously circulating strains that emerged through past episodes of antigenic drift. These observations on the imprinting of the first infection on later response continue to be valid, as evidenced by an age cohort effect, based on year of birth, that has been seen in vaccine-induced immunity to 2009 pandemic influenza A(H1N1) viruses (A[H1N1]pdm09) [3]. It has also recently been reported that the severity of avian A(H5N1) and A(H7N9) infections in humans may vary, based on whether a person’s first exposure to influenza virus was to an A(H1N1) or A(H3N2) virus, and this has been termed “first flu is forever” [4, 5]. There have been examples other than those involving influenza virus in which imprinting of early infections on immunologic memory has been observed [6].

In the 1970s, Hoskins et al reported a phenomenon not related to imprinting by early infection in which boys at the Christ’s Hospital school in England who were protected by vaccination in a first study year were more likely than those infected in the first year to be infected in the next year [7]. This initially was interpreted to mean that infection was more protective than vaccination, for which there is a general consensus. However, following the third year, it was found that the overall 3-year risk of influenza did not vary by vaccination status, and the conclusion made by Hoskins was that there was no long-term advantage of annual influenza vaccination. There were a number of critiques of the methods of this study, and the National Institute for Allergy and Infectious Diseases responded by funding a clinical trial to attempt to address the issue experimentally. That study and a subsequent meta-analysis found no consistent pattern of reduced vaccine effectiveness (VE) and concluded that the policy of annual vaccination was appropriate [8, 9].

Controversy around repeated vaccination remained relatively dormant until data from a VE study performed in a household cohort during the 2010–2011 season demonstrated reduced VE among individuals vaccinated in the prior year [10]. Similar findings were subsequently observed in various other studies, primarily using the case-test-negative design to assess VE in healthcare settings [11–13]. However, these reductions in VE have not been observed each season and have primarily been associated with outbreaks of A(H3N2) infection. This points to a potentially complicated immune mechanism underlying the phenomenon. One hypothesis, proposed by Smith et al, suggests that patterns of reduced VE could be explained by the antigenic distances between past and current vaccine antigens and the viruses that circulate [14]. This model was used to explain apparent discrepancies between data from the study by Hoskins et al and those from the subsequent clinical trial.

Clearly, these observations over many years have implications for vaccination policy in terms of frequency of vaccination, vaccine strain selection, and research priorities for the development of novel influenza vaccines. It is therefore imperative to identify the specific instances in which this phenomenon occurs and to understand the immunologic and methodologic issues that underlay it. Only then can there be a rational approach to addressing the problem.

In this issue of The Journal of Infectious Diseases, Martínez-Baz et al estimate annual influenza VE against A(H1N1)pdm09 infection, with consideration of prior vaccination, over 4 influenza seasons, from 2010–2011 through 2015–2016 [15]. Their study takes advantage of the fact that the A(H1N1)pdm09 component of the influenza vaccine remained unchanged from 2009 through 2016. In season-specific analyses, there were no significant differences in VE between those vaccinated in the current season only and those vaccinated in both the current and prior seasons. In pooled analyses, there was some suggestion that, among individuals vaccinated in the current season, VE was lower among those with >2 prior doses than among those with no or 1–2 prior doses. However, the most meaningful finding of this study is that influenza vaccine reduced the odds of infection by nearly 50% on average across the 4 influenza seasons studied. Although we would like the VE to be >50%, it is estimated that tens of thousands of hospitalizations are averted by vaccination even in years when VE is lower than this [16].

The authors also notably report a dose-response trend in decreasing VE with time since last vaccination. Despite this, significant protection was demonstrated for those whose most recent vaccination was in the current season, as well as for those whose most recent vaccination was in the immediately previous season. This is consistent with studies that have suggested that influenza vaccine–induced protection wanes relatively slowly, thus providing season-long protection, assuming there is little antigenic change in circulating viruses [17, 18]. Obviously, vaccines that provide greater initial protection, as has been the case for A(H1N1)pdm09, will provide protection for a longer duration.

Although the results of this study are encouraging as to the effectiveness of annual vaccination programs in preventing A(H1N1)pdm09 infection, the situation is likely different for A(H3N2). VE has consistently been lower against A(H3N2), and antigenic changes in circulating viruses have occurred more rapidly, necessitating updating of strains included in the vaccine [19]. Reduced VE among previously vaccinated individuals has also been observed most commonly in seasons when A(H3N2) predominates. Untangling this phenomenon for A(H3N2) is much more challenging, given extensive histories of prior exposure to various strains through vaccination and natural infection. Longitudinal studies with excellent documentation of vaccination and infection and collection of immunologic specimens can help, although definitive studies are likely cost prohibitive. Therefore, mathematical modeling may also play a role. Continued efforts toward the development of universal or longer-lasting vaccines would also simplify the issue.

In terms of policy regarding annual influenza vaccination, we must ultimately ask whether individuals who are vaccinated annually are better off than those who are not. For A(H1N1)pdm09, the results of Martínez-Baz et al suggest that the answer is almost certainly yes. Further, results from most studies that demonstrate reduced VE with sequential A(H3N2) vaccination also demonstrate that those vaccinated in both years are better protected than those unvaccinated or vaccinated only in the prior year. Therefore, especially given the unpredictability of the observed effect, our current policy should continue. Meanwhile, it is imperative to work on understanding the immunologic mechanisms underlying the effect so we can move toward correcting the problem. Vaccines of longer duration of protection would also solve the problem, but that long-term goal should be pursued as we simultaneously attempt to understand the impact of sequential vaccination using current vaccines.

Note

Potential conflicts of interest. A. S. M. reports grant support from Sanofi Pasteur and consultancy fees from Sanofi, GSK, and Novavax. J. G. P. certifies no potential 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.

Author notes

Correspondence: J. G. Petrie, Department of Epidemiology, University of Michigan School of Public Health, 1415 Washington Heights, Ann Arbor, Michigan 48109-2029 ([email protected]).