Influenza vaccine: How effective is it in the elderly?

January 2008

One of the top 10 stories in 2007 that we presented last month was about the publication last fall of two divergent views of influenza vaccine efficacy in the elderly. I indicated in this column last month that I would devote this editorial to a discussion of these reports and possibly resolve the apparent conflict. It will not be possible to resolve these contrasting views, but it is possible to explain how they came about.

Benefits overstated?

The first of the two publications in question was written by Lone Simonsen, MD, and her colleagues, mostly from the National Institute of Allergy and Infectious Diseases and NIH, and was published in the October 2007 issue of Lancet Infectious Diseases. It was not an original research article but rather an extensive essay and literature review, arguing that the mortality benefits from influenza vaccine in the elderly have been greatly overstated. In support of this position, Simonsen and her colleagues point out that, with a single exception, there are no randomized controlled trials that conclusively demonstrate vaccine efficacy in people older than 70 years. Rather, almost all the efficacy data used to justify the current influenza vaccine policy in the United States are based on observational studies in large health maintenance organization populations in which participants were either vaccinated or not vaccinated by choice rather than by protocol and subsequent hospitalization and mortality risk measured in these two groups.

For various reasons described by the researchers, seasonal all-cause mortality is probably the best measure of the effect of influenza in the population. Since the 1968 H3N2 pandemic, influenza has accounted for about 5% of the approximately 600,000 winter deaths in the United States. In this argument, influenza deaths contribute but a small proportion of all winter mortality, and influenza vaccine could not reasonably be expected to do any more than eliminate these influenza-related deaths. (This argument could possibly be disputed since assignment of influenza relatedness may be difficult at best.) In contrast, cohort observational studies appear to demonstrate a huge benefit: up to a 50% reduction in all-cause winter mortality. Simonson and her colleagues argued that this is unlikely, if not downright preposterous! They point out further that despite an increase in vaccine coverage in the elderly from 15% to 65%, crude excess mortality in that population actually increased during the same period.

They suggest that there must be one or more selection biases operating in these observational studies and are particularly concerned about the “frailty” bias, in which it is presumed that a subset of very frail unvaccinated people account for a disproportionately large number of deaths. A second concern is that seasonal all-cause mortality, although a good estimate of the influenza burden, is in reality a very nonspecific endpoint.

The researchers were supported in their argument in an editorial prepared by Jefferson and Di Pietrantonj, two participants in the Cochran Collaboration report on influenza vaccine efficacy. The researchers from the Cochran report were deeply skeptical of the results of cohort observational studies and therefore deeply skeptical about the scientific basis for influenza vaccine policy in the United States.

Vaccine efficacy defended

Scarcely 10 days after the Simonsen report was published, the study results of another cohort observational study conducted by Kristen Nichol, MD, MPH, and her colleagues were published in The New England Journal of Medicine. The study was seemingly designed to address many of the problems raised in the Simonsen report. The Nichol study was carried out over a period covering 10 influenza seasons from 1990-1991 to 1999-2000 in a large HMO in Minnesota and Wisconsin called Health Partners. During the last four winters of the study, two additional HMOs participated: Kaiser Permanente Northwest in the Vancouver, Canada; Washington; and Oregon areas and Oxford Health Plans in the New York metropolitan area. The participants were community-dwelling (not institutionalized) elderly, aged 65 years and older. In all, there were almost 714,000 person-years of observation. There were more high-risk conditions among the vaccinated than in the unvaccinated, arguing somewhat against the “frailty” bias. Over the 10 years of observation, the vaccinated population had a 27% reduced risk for hospitalization for pneumonia or influenza and a 48% reduced risk for death (due to all causes), as compared with the unvaccinated population. These risk reductions were similar across risk and age groups. The researchers modeled the effects of various biases and confounders that might have resulted in overestimates of vaccine effectiveness; they found that although extreme examples of confounder prevalence somewhat lowered the efficacy estimates, the reductions in hospitalization and death remained significant. (Personally, I was somewhat troubled by the lack of an age-related risk reduction; data from many studies confirm that immunogenicity, and therefore likely vaccine efficacy, declines with advancing age. This study did not show such an effect.)

Note the astonishing 48% reduction in all-cause mortality in the vaccinated cohort. This effect seems far greater than could reasonably be expected by an anti-influenza effect alone. Could receipt of influenza vaccine be a proxy for some other mortality-avoidance behavior? If so, what? This is the principal concern raised in the Simonsen review, and it needs to be addressed.

An accompanying editorial by John Treanor places these results in the perspective of influenza vaccine policy in the United States and points out that they fully justify a continuation of that policy.

Conflicting data

So there you have the basis for the disparity and the conflicting data. It is important to note that Simonsen and her colleagues did not argue for a change in policy but only pointed out the great uncertainty (in their view) of the efficacy data on which that policy is based. Further, no one argues that current influenza vaccines are perfect — far from it! Everyone agrees on the need for improved influenza vaccines, and work continues toward that goal.

To delve further into this disparity gets into somewhat arcane issues of epidemiologic study design that go beyond the scope of these editorial comments. One can make a strong argument that perhaps the only way to resolve this disparity is a large randomized, controlled trial; ethical issues loom large, however, and mitigate against the possibility that such a trial will ever be carried out. If such a trial is conducted, it is critically important that the diagnosis of influenza be based on laboratory confirmation rather than the much less specific clinical diagnosis of influenza life illness, as defined and commonly used by the CDC.

Another point to consider is that vaccination of the elderly themselves may be the wrong target; perhaps, as suggested by some individuals in the past several years, we should devote much more attention to preventing influenza in those populations that transmit the disease to the elderly — namely children. There is a slowly growing body of evidence supporting just that approach.

There are many more issues to consider than could be covered in this limited space. Interested readers would do well to read the two articles in their entirety, as well as the accompanying editorials. My own conclusion is that the vaccine is probably not as good as we sometimes think but better than no vaccine at all. I will continue to take the vaccine every year until something better comes along!

For more information:

• Simonsen L, Taylor RJ, Viboud C, et al. Mortality benefits of influenza vaccination in elderly people: an ongoing controversy. Lancet Infect Dis. 2007;7:658-666.
• Jefferson T, Di Pietranonj C. Inactivated influenza vaccines in the elderly — are you sure? Lancet Infect Dis. 2007;370:1199-1200.
• Nichol KL, Nordin JD, Nelson DB, et al. Effectiveness of influenza vaccine in the community-dwelling eldery. N Engl J Med. 2007;357:1373-1381.
• Treanor JD. Influenza — the goal of control. N Engl J Med. 2007;357:1439-1441.