IJE Advance Access originally published online on April 17, 2006
International Journal of Epidemiology 2006 35(3):799-800; doi:10.1093/ije/dyl069
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Letter to the Editor |
Bias in studies of influenza vaccine effectiveness: the authors reply to Hak et al.
1 Center for Health Studies, Group Health Cooperative, Seattle, WA, USA
2 Department of Epidemiology, University of Washington, Seattle, WA, USA
* Corresponding author. E-mail: Jackson.L{at}ghc.org
In our study of elderly members of a large health care plan, we found that seniors who had received influenza vaccine had a lower risk of death in periods before the start of the influenza season, during the influenza season, and in the summer than seniors who had not been vaccinated.1 The greatest reduction in risk was before the onset of the flu season—during that period the death rate among the vaccinated was 60% lower than among the unvaccinated. Influenza causes an estimated 5% of deaths in seniors during the influenza season,2,3 and so makes at most a negligible contribution to deaths outside the influenza season. The association we observed must therefore be spurious, probably due to confounding. The important question is, what is the source of the confounding?
Hak et al.4 have hit close to the crux of this issue when they suggest that we ‘should have excluded persons likely to die during the pre-influenza season’ from our study population. We agree that there is a healthy vaccinee bias due to factors, such as severe illness and frailty, which are associated with an increased risk of short-term mortality and a decreased likelihood of vaccination. This bias influences the estimates obtained in analyses of both influenza and non-influenza time periods. To reduce the influence of bias we must be able to accurately identify the confounding factors. Our results suggest that covariates defined by diagnosis codes do not allow us to identify the presence of severe illness or frailty.5 Including the diagnosis code variables in analytic models, therefore, does not produce estimates of effect that are less biased; in fact, this method appears to result in estimates of effect that are, if anything, more biased than the unadjusted estimates.
We agree that there are many published studies reporting a decreased risk of death and hospitalization in influenza vaccinated compared with unvaccinated seniors. Had we restricted our analyses to the influenza season, our findings could be similarly interpreted. However, reproducibility is not proof of validity. Our results clearly demonstrate that the differences in risk are not specific to the influenza season and so are biased. As we stated in our paper, we do not believe one can conclude that influenza immunization has no benefit with respect to mortality during the influenza season. Rather, we believe that because of the limitations of the data sources used, the non-randomized studies conducted to date do not provide a reliable basis for estimating the presence and/or size of this benefit.
References
1 Jackson LA, Jackson ML, Nelson JC, Neuzil KM, Weiss NS. Evidence of bias in estimates of influenza vaccine effectiveness in seniors. Int J Epidemiol doi:10.1093/ije/dyi274.
2 Simonsen L, Reichert TA, Viboud C, Blackwelder WC, Taylor RJ, Miller MA. Impact of influenza vaccination on seasonal mortality in the US elderly population. Arch Intern Med 2005;165:265–72.
3 Thompson WW, Shay DK, Weintraub E et al. Mortality associated with influenza and respiratory syncytial virus in the United States. JAMA 2003;289:179–86.
4 Hak E, Hoes AW, Nordin, J, Nicol KL. Benefits of Influenza vaccine in US elderly—appreciating issues of confounding bias and precision. Int J Epidemiol 2006;35:800–2.
5 Jackson LA, Nelson JC, Benson P et al. Functional status is a confounder of the association of influenza vaccine and risk of all cause mortality in seniors. Int J Epidemiol doi:10.1093/ije/dyi275.
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||