Study on flu shots and heart-attack risk questioned

Sep 21, 2010 (CIDRAP News) – A case-control study using records on tens of thousands of people in the United Kingdom suggests that influenza vaccination reduces the risk of heart attack in people older than 40, but other researchers who have studied the benefits of flu immunization have raised doubts about the findings.

A British research team gathered data on about 16,000 Britons who had acute myocardial infarctions (AMI) over a 6-year period and matched each of them with four controls who were free of heart attacks during that time. They calculated that, after adjustment for differences in risk factors for AMI, those who had a flu shot within the previous year had a 19% lower risk of AMI than those who didn't, according to their report in the Canadian Medical Association Journal.

But two other researchers who were asked to comment on the study said it is flawed, mainly because the authors did not conduct separate assessments of the effects of vaccination on heart attack risk during flu season and at other times of the year. They said such an assessment is a key test of whether apparent reductions in risk instead actually reflect the fact that people who get vaccinated tend to be in better health than those who don't (the "healthy user effect").

Without such separate analyses, they said, the statistical adjustment approach the authors used to deal with differences in risk factors actually makes the conclusions less accurate.

Large case-control study
The British researchers, led by A. Niroshan Siriwardena of the University of Lincoln as first author, write that several previous studies, including two randomized controlled trials, suggest that flu vaccination helps reduce the risk of AMI, but they provide "insufficient and conflicting evidence."

The researchers used the UK General Practice Research Database (which includes 5% of the English population) to find records on patients who were at least 40 years old and suffered an AMI between November 2001 and May 2007. They then tried to match each of these with four other patients who were the same age and gender, went to the same clinic, and saw a general practitioner the same month as when the case-patient had an AMI.

The investigators determined whether the patients had received a flu vaccination, defined as vaccination in the year before the date of the case-patient's heart attack—the "index date." They also assessed whether patients had been vaccinated early or late in the flu vaccination season—between Sep 1 and Nov 15 or between Nov 16 and the end of February.

Also, to look for any effect of pneumococcal vaccination, the authors examined whether patients had ever received pneumococcal vaccine before the index date.

The team identified 16,012 cases and matched them with 62,964 controls from 379 practices. Flu vaccination rates within the two groups were nearly the same: 52.9% of cases and 51.2% of controls. People with risk factors for AMI were more likely to have been vaccinated, which was not surprising since many such risk factors are also indications for flu vaccination.

Before adjusting for potential confounding variables, the researchers found that those who were vaccinated had an 8% higher risk of heart attack (odds ratio [OR], 1.08; 95% confidence interval [CI], 1.04 to 1.13).

But this changed when they adjusted for a number of potential confounding variables, such as vaccination target group status, other cardiovascular risk factors, and number of medical consultations in the past 5 years. After adjustment, the vaccinated patients were found to have a 19% lower risk of AMI (OR, 0.81; 95% CI, 0.77 to 0.85). The most important confounders were consultation rate, previous heart disease, and diabetes mellitus.

The researchers also found that vaccination within the same vaccination season as the index date was linked with a 20% decrease in AMI risk, compared with no vaccination. Further, vaccination early in the flu season was associated with a greater reduction in risk than vaccination later in the season: 21% (OR, 0.79; 95% CI, 0.75 to 0.83) versus 12% (OR, 0.88; 95% CI, 0.79 to 0.97).

After adjustment for confounders, the analysis did not show a significant effect of pneumococcal vaccination on heart attack risk: OR, 0.96; 95% CI, 0.91 to 1.02. Also, the combination of pneumococcal and flu vaccination did not reduce the risk as compared with flu vaccination alone.

Conclusions questioned
Lisa A. Jackson, MD, MPH, senior researcher with the Group Health Research Institute in Seattle, said the main finding of a link between flu vaccination and reduced AMI risk is questionable because the authors did not conduct separate assessments of the effect on AMI risk during flu season and non-flu season. If vaccination does lower the risk, "it should only happen when flu is around," she said.

Further, without those separate seasonal assessments, adjusting for the potential confounding variables actually can have the effect of making the answer "more wrong," Jackson said, adding, "You see this in all these kinds of studies."

In a 2005 study, Jackson and colleagues examined the effects of flu vaccination on the risk of hospitalization and death among 72,000 seniors before, during, and after flu season for several years. They found the largest risk reductions occurred in the pre-flu season periods, which pointed to a healthy user effect—"preferential receipt of vaccine by relatively healthy seniors." They concluded that this bias was large enough to explain the reductions in risk seen during flu season.

Jackson also found that adjusting for confounding variables did not eliminate this bias in the analysis of risk in pre-flu season periods, but instead made it worse.

Lone Simonsen, PhD, research director in George Washington University's Department of Global Health, another researcher who has studied flu vaccine effectiveness, concurred with Jackson. She noted that the unadjusted data showed no effect or even a negative effect of flu vaccination on AMI risk and that a benefit was apparent only after adjustment for confounders.

Referring to Jackson's research, Simonsen commented by e-mail, "Given the recent elegant demonstration of how that sort of analytic adjustment can actually move unadjusted data away from the truth and create profound mismeasurement, it is clear that this paper may have been profoundly affected by the same problem." She said she would be convinced by the findings only if the benefits were confined to the flu season.

Jackson also critiqued the study for counting flu vaccinations anytime in the 12 months preceding the index event, rather than counting only vaccinations received during the September-to-December vaccination season of the relevant year. If someone had an AMI on Nov 1, a vaccination they received in previous November or December would have been counted, even though it would've been for the previous flu season, she said.

Siriwardena AR, Gwini SM, Coupland CAC. Influenza vaccination, pneumococcal vaccination and risk of acute myocardial infarction: matched case-control study. Can Med Assoc J 2010 Sep 20 (Early online publication) [Abstract]

See also:

Jackson LA, Jackson MI, Nelson JC, et al. Evidence of bias in estimates of influenza vaccine effectiveness in seniors. Int J Epidemiol 2005 (published online Dec 20) [Full text]

This week's top reads