Influenza vaccination reduced myocardial infarctions in UK older adults: a prior event rate ratio study

OBJECTIVE
We aimed to estimate the real-world effectiveness of the influenza vaccine against myocardial infarction (MI) and influenza in the decade since adults aged ≥65y were first recommended the vaccine.


STUDY DESIGN AND SETTING
We identified annual cohorts, 1997 to 2011, of adults aged ≥65y, without previous influenza vaccination, from UK general practices, registered with the Clinical Practice Research Datalink. Using a quasi-experimental study design to control for confounding bias, we estimated influenza vaccine effectiveness on hospitalisation for MI, influenza and antibiotic prescriptions for lower respiratory tract infections.


RESULTS
Vaccination was moderately effective against influenza, the prior event rate ratio (PERR)-adjusted hazard ratios [HR] ranging from 0.70 in 1999 to 0.99 in 2001. PERR-adjusted HRs demonstrated a protective effect against MIs, varying between 0.40 in 2010 to 0.89 in 2001. Aggregated across the cohorts, influenza vaccination reduced the risk of MIs by 39% (95%confidence interval: 34%, 44%) .


CONCLUSIONS
Effectiveness of the flu vaccine in preventing MIs in older UK adults is consistent with the limited evidence from clinical trials. Similar trends in effectiveness against influenza and against MIs suggest the risk of influenza mediates the effectiveness against MIs, although divergence in some years implies the mechanism may be complex.


Weighted results and PERR adjustment
From the IPTW analysis, adjusting for measured confounders, the results were variously distributed around unity (Table 1).No results for any cohort's study (blue circles in fig 1) were significantly different from the null, but the largest protective effect was estimated for 2003 with an HR of 0.78 (95% CI: 0.59, 1.01), respectively.
The HRs from 2004 onwards were above one, indicating harmful effects with HRs as high as 1.28 (95% CI: 0.95, 1.74) in 2009.Once the PERR method had been applied to the weighted estimates, the HRs (blue triangles in fig 1) fell below one and were in closer agreement with those from the PERR-adjusted unweighted-Cox models (grew triangles in fig 1), indicating a protective effect from vaccination against MIs, although those for 1997 and 2001 were not statistically significant.The diagnostic plots (Supplementary figures 1-5) of the standardised and unstandardised mean differences of potential confounders in the study periods revealed better balance was generally achieved for the earlier cohorts, noting that not all the variables contributed to the propensity score.It was also apparent that after 2001, age became the leading variable with the greatest imbalance between vaccination groups.This seemed to coincide with the shift from risk-based vaccination prior to the policy introduction to the age-based eligibility criterion for vaccination.

Pairwise estimates
The HRs of vaccination effect from the pairwise model (

Pairwise estimates
The HRs from the pairwise method, while always further below the null, tracked very closely with the PERR-adjusted results, and ranged from 0.63 (95% CI: 0.54, 0.72) in 1999 to 0.90 (95% CI: 0.78, 1.05) in 2001 ( Table 3: Hazard ratios (95% confidence intervals) from the pairwise analysis of the effect of influenza vaccination on the composite influenza outcome for each cohort.Note that each cohort is a reduced subset of patients with an outcome in either the prior or study period as demanded by the pairwise likelihood.

Fig 1 :
Fig 1: Inverse probability treatment weighted hazard ratios for the study periods (blue circles) and PERR-adjusted hazard ratios (grey triangles) from 1997 to 2001.Error bars represent 95% confidence intervals for the HRs.

Fig 2 :
Fig 2: Hazard ratios of the estimated effect of influenza vaccination on MI hospital admissions from the PERR-adjusted model (grey dots) and the pairwise model (blue triangles) with errors bars representing 95% confidence intervals.

Fig 3 :
Fig 3: Hazard ratios for the effect of influenza vaccination on influenza outcomes from the PERR-adjusted model (grey dots) and Pairwise model (blue triangles) with errors bars representing 95% confidence intervals.

Table 1 :
Inverse probability treatment weighted hazard ratios (95% confidence intervals) for the study and prior periods of each cohort from 1997 to 2011 with the PERR method applied.

Table 2
those for the PERR method.This is a characteristic of the Pairwise method, as only those individuals, whose shortest survival time from either period ending in an outcome, contribute information to the likelihood function.
greater overall protective effect of vaccination against MI than that estimated through the PERR method.There are notably fewer subjects in each cohort compared to

Table 2 :
Pairwise-adjusted hazard ratios (95% confidence intervals) for each annual cohort for the effect of influenza vaccination on MI hospital admissions.