Moving Toward Improved Influenza Vaccines

(See the major article by Shay et al on pages 510–7)

In 1943, a placebo-controlled trial of inactivated influenza vaccine (IIV) produced in fertile eggs was carried out among members of the Army Specialized Training Program at the University of Michigan [1]. The vaccine was approximately 70% efficacious in preventing laboratory-confirmed clinical influenza and the study also demonstrated that protection against infection was correlated with titers of hemagglutination-inhibiting (HAI) antibody. For the next 25 years, the US Military continued regularly to evaluate influenza vaccines, mainly in young adults. The trial designs varied, but the reported vaccine efficacy (VE) ranged from 70% to 90% [2]. Unlike the 1943 trial, many of the subsequent military studies relied exclusively on a serologic endpoint, comparing 4-fold rise in HAI antibody titers in the vaccinated with unvaccinated rather than identifying the infecting virus, at that time a laborious procedure. We now know that using this endpoint overestimates vaccine efficacy of IIV, as those with high HAI titers produced by vaccination are less likely to have a rise in titer when infected [3].

In contrast with the situation in the military, until relatively recently, vaccine use in the US general population was mainly confined to older adults and those with underlying conditions [4]. This was based on the recognition that complications, including hospitalization and death, were more common in this population who actually had lower infection rates. Over a number of years, a variety of studies suggested that VE was much lower than 70%–90% in these older adults, and sometimes could not be demonstrated at all [5]. Many of the reports were anecdotal, and some involved nursing homes, so actual VE in those living independently was difficult to quantify. A study reviewing articles on HAI antibody vaccination response in older compared to younger adults as a way of predicting protection found that many factors made it hard to document actual differences [6]. A meta-analysis of studies with various clinical endpoints in older individuals concluded that, while heterogeneity did exist, in a subset, the VE was 56% in preventing respiratory illnesses [7]. However, that conclusion was challenged based on the nonspecific illness endpoints used.

The development and general use of the polymerase chain reaction (PCR) technique to identify actual influenza infection has now established a level playing field for various VE studies across ages. VE is rarely higher than 60%–70%, and sometimes lower. These findings came initially from placebo-controlled trials, mainly in younger adults where randomization is ethically possible [8–10]. Similar results have come recently from observational studies, using mainly the so-called test-negative design, comparing vaccination status in those with respiratory illnesses in the influenza season who test positive for influenza by PCR with those who test negative [11–13]. Because the studies are observational, they can ethically involve individuals of all ages. There has not been a consistent drop-off in VE at age 60 or 65 years; often there are also variations by age in other groups. Factors such as the virus type or subtype involved and prior year vaccination may have a significant effect of VE. In particular, VE has been relatively high for A(H1N1) and type B, in the range of 60%, but much lower for A(H3N2) at all ages, even in years without major antigenic drift [14]. An issue in interpreting the results is that older individuals, especially the oldest of the old, are relatively underrepresented in these studies.

Various approaches over the years have attempted to improve vaccine performance overall. In the days when the military dominated research on influenza, adjuvanted vaccines were extensively evaluated, and clearly demonstrated higher antibody titers which were more broadly reactive. Adjuvants studied were water-in-oil including Freund’s, and use was abandoned because of reactogenicity [15–17]. Later, the approach involving increasing the quantity of hemagglutinin (HA) in the vaccine began to be examined. A limitation was that the increased antibody response to increased HA content was far from linear, requiring much more than the 15 µg contained in the standard vaccine. The first studies were carried out in young adults and the HA content of each strain went up to 135 µg [18]. Significant increases were seen not only in serum antibody responses, but also in secretory antibody, known to correlate better with protection. Similar results were demonstrated in older adults, with antibody responses increasing 2-fold for a 9-fold increase in HA content [19].

Once the logistics of producing high-dose (HD) vaccine and other issues related to safety were overcome, the randomized controlled trial on the effect of the HA increase on antibody response in older individuals took place. The content of each of the 3 HD components was limited to 60 µg. In comparison with standard vaccine, the US Food and Drug Administration predefined superiority could be claimed for the A(H3N2) and A(H1N1) components but not for type B [20]. There was uncertainty on how much added protection against actual illness would result in older individuals with demonstrated increases in HA content until a trial with clinical outcomes could be conduted. Such a controlled trial is a major undertaking in terms of size especially when, as was ethically necessary, the comparator is not placebo but standard vaccine. It was carried out in 2 years, 2011–2012 and 2012–2013 and involved 31989 participants. Relative efficacy against laboratory-confirmed clinical influenza of all types was 24.2% higher in recipients of the HD formulation than in those who received regular dose; absolute VE was impossible to determine since placebo could not be used [21]. Although point estimates suggested improved VE against all types and subtypes, statistical significance was achieved only for A(H3N2). In addition, secondary analyses suggested that the HD vaccine was more effective in preventing serious events possibly related to influenza, including hospitalization, but these events were relatively infrequent [22–23].

The only way to study infrequent outcomes is to use a much larger sample size. However, that can only be done using observational studies, in which there will be self-selection for vaccination. Then there is the question of the specificity of outcomes. Even with spreading use of PCR to identify influenza infections, the test is not regularly used enough, so unbiased identification of cases through this means alone is not possible. Rapid tests are more commonly used, but while specific, most tests are not sensitive, making results difficult to interpret.

Shay et al, in this issue of The Journal of Infectious Diseases, used the Medicare database to access the required very large number of cases necessary to examine influenza-related mortality [24]. In addition to examining deaths following an International Classification of Diseases diagnosis of influenza, they also examined as secondary analyses all cases with such a diagnosis, as well as cases in which a rapid test for influenza with was ordered and oseltamivir treatment was dispensed. The latter methods were used, in part, to reproduce earlier findings from this group, which confirmed that this approach could replicate results from the randomized trial [25]. Because the study was observational, they controlled in the analysis for various factors such as frailty and socioeconomic status likely to confound the results. Well over a million recipients of the standard and of HD vaccine were studied in each of 2 years. In 2012–2013, the HD was 36.4% more effective in preventing death, but in 2013–2014 it was only 2.5% more effective, which was not statistically significant. This variation is not surprising and is actually in line with expectations. In 2013–2014 A(H1N1) predominated, but in 2012–2013 A(H3N2), known to cause higher mortality in the elderly, was most common, with some cases of type B. There is also the issue of basic differences in VE by subtype, using studies involving ambulatory visits as the benchmark. Overall, VE of standard vaccine in 2013–2014 in all age groups was approximately 54%, but in 2012–2013 it was only 39% for A(H3N2), leaving much room for improvement [26].

The demonstration that the HD vaccine has an enhanced effectiveness against influenza-associated mortality fits nicely with previous data that using this vaccine results in improved VE against uncomplicated illness and likely hospitalizations. This indicates that improvement in our 70-year-old influenza vaccines is possible, and to get there more quickly we should not ignore older technologies while working on more dramatic advances. We know adjuvants result in better vaccine responses and we now have safer formulations than those used years ago in the military experiments Such an adjuvanted vaccine for older individuals is now available in the United States [27]. It has been shown in observational studies to reduce pneumonia and influenza hospitalizations by 25% compared to standard vaccine [28]. The biggest payoff of better vaccines in terms of reduced mortality will be in the elderly, but much of the vaccine in the United States is used at younger ages and improvement, especially of A(H3N2) vaccines, is needed for all age groups. The HD and adjuvanted approaches are welcome additions to our ability to counter effects of influenza and they should be further investigated and used. They should be viewed as not the end, but the beginning, of programs leading to vaccines of higher effectiveness, longer duration, and broader protection.

Note

Potential conflicts of interest. The authors reports grants and personal fees from Sanofi-Pasteur, and personal fees from Novartis outside the submitted work. Author certifies no potential conflicts of interest. 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.

Author notes

Correspondence: A. S. Monto, MD, Department of Epidemiology, University of Michigan School of Public Health, 1415 Washington Heights, Ann Arbor, MI 48109-2029 ([email protected]).