G Model JVAC-14945; No. of Pages 1
ARTICLE IN PRESS Vaccine xxx (2014) xxx–xxx
Contents lists available at ScienceDirect
Vaccine journal homepage: www.elsevier.com/locate/vaccine
Letter to the Editor
Selection bias in the comparison between adjuvanted and unadjuvanted influenza vaccine effectiveness In his guide to the field assessment of vaccine effectiveness (VE), Orenstein highlights that valid estimates can only be obtained if baseline disease risk and case finding are similar between groups [1]. These two conditions were not respected in the paper by Van Buynder et al. [2] in reporting results from the first (2011–12) of their two sequential study seasons [3] to assess the comparative effectiveness of adjuvanted trivalent inactivated influenza vaccine (ATIV) versus non-adjuvanted TIV in the elderly in British Columbia (BC), Canada. In this study, ATIV and TIV recipients did not emerge from the same source population [2]. Most (77%) ATIV recipients lived in long-term care (LTC) compared to 29% for TIV and 27% for unvaccinated participants (p < 0.001). Although most seniors in BC live in the community, the majority (57%) of participants in this study resided in LTC. Exposure to influenza and other respiratory viruses in LTC and the community can hardly be considered similar and the likelihood of immunization also differs in these two settings (90% versus 70–75%, respectively, in BC) [4]. Case finding also systematically differs between LTC and community patients in BC. In LTC, BC has an established process for investigating clusters of influenza-like illness (ILI), encouraging the submission of specimens from up to six patients in the event of ≥2 ILI detections within a one-week period in a given care facility [5]. Authors did not indicate how multiple patients within the same LTC ILI cluster were handled. This is an important issue because the likelihood of testing positive or negative would be highly correlated among patients within a given ILI cluster, violating the assumption of independent events. Varying contribution from LTC versus community settings may explain the otherwise surprising but significant differential in testpositivity among participants in adjacent study regions of the Lower Mainland, BC (14% in Fraser, 36% in Vancouver Coastal) despite the same program strategy. Although it was not provided, participant profiles showing case and vaccine groups additionally stratified by region and LTC status may help elucidate that variation. The systematic differences in vaccine distribution, sampling collection and exposure to influenza and other respiratory viruses in
this study represent selection biases that compromise the validity of the estimates and cannot be corrected in the analysis. Overall VE estimates based on bundled LTC and community participants are unreliable but the sample size with restriction to community-based participation alone was too limited for comparative VE analysis (30 cases in total across all study groups). The 95% confidence intervals around the point estimate of VE in the community were consequently wide for ATIV (ranging 2–93%) and were not specified at all for TIV to assess overlap. Selection bias due to non-comparability of study groups and differential case finding, as well as a lack of statistical power, seriously compromise the validity of this study [1]. Author statements that ATIV “significantly improved protection” but “TIV was ineffective” go far beyond what their data can support. References [1] Orenstein WA, Bernier RH, Hinman AR. Assessing vaccine efficacy in the field: further observations. Epidemiol Rev 1988;10:212–41. [2] Van Buynder PG, Konrad S, Van Buynder JL, Brodkin E, Krajden M, Ramler G, et al. The comparative effectiveness of adjuvanted and unadjuvanted trivalent inactivated influenza vaccine (TIV) in the elderly. Vaccine 2013;31(51):6122–8, http://dx.doi.org/10.1016/j.vaccine.2013.07.059. [3] Van Buynder P. Adjuvanted influenza vaccine effectiveness in the elderly (65+ yrs). Clinical trials.gov identifier: NCT01424371. Available at: http:// clinicaltrials.gov/ct2/show/NCT01424371 [4] BC Centre for Disease Control. Influenza immunization coverage. Available at: http://www.bccdc.ca/imm-vac/BCImmunizationCov/flucoverage/default.htm [5] PHSA Laboratories Public Health Microbiology & Reference Laboratory. Influenza Like Illness (ILI) Outbreak Laboratory Form: Specimen Collection and Shipping. Available at: http://www.phsa.ca/NR/rdonlyres/AC46FB9F44FE-487C-9BF4-5C0578F21EF3/0/ILIFormOct2012.pdf
Gaston De Serres ∗ Institut national de santé publique du Québec, Biological, Environmental and Occupational Risks, 2400 d’estimauville, Quebec City, Quebec, Canada G1E 7G9 ∗ Tel.:
+1 418 666 7000x274. E-mail address: [email protected] 6 November 2013 12 December 2013 Available online xxx
0264-410X/$ – see front matter © 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.vaccine.2013.12.032
Please cite this article in press as: De Serres G. Selection bias in the comparison between adjuvanted and unadjuvanted influenza vaccine effectiveness. Vaccine (2014), http://dx.doi.org/10.1016/j.vaccine.2013.12.032
ARTICLE IN PRESS Vaccine xxx (2014) xxx–xxx
Contents lists available at ScienceDirect
Vaccine journal homepage: www.elsevier.com/locate/vaccine
Letter to the Editor
Selection bias in the comparison between adjuvanted and unadjuvanted influenza vaccine effectiveness In his guide to the field assessment of vaccine effectiveness (VE), Orenstein highlights that valid estimates can only be obtained if baseline disease risk and case finding are similar between groups [1]. These two conditions were not respected in the paper by Van Buynder et al. [2] in reporting results from the first (2011–12) of their two sequential study seasons [3] to assess the comparative effectiveness of adjuvanted trivalent inactivated influenza vaccine (ATIV) versus non-adjuvanted TIV in the elderly in British Columbia (BC), Canada. In this study, ATIV and TIV recipients did not emerge from the same source population [2]. Most (77%) ATIV recipients lived in long-term care (LTC) compared to 29% for TIV and 27% for unvaccinated participants (p < 0.001). Although most seniors in BC live in the community, the majority (57%) of participants in this study resided in LTC. Exposure to influenza and other respiratory viruses in LTC and the community can hardly be considered similar and the likelihood of immunization also differs in these two settings (90% versus 70–75%, respectively, in BC) [4]. Case finding also systematically differs between LTC and community patients in BC. In LTC, BC has an established process for investigating clusters of influenza-like illness (ILI), encouraging the submission of specimens from up to six patients in the event of ≥2 ILI detections within a one-week period in a given care facility [5]. Authors did not indicate how multiple patients within the same LTC ILI cluster were handled. This is an important issue because the likelihood of testing positive or negative would be highly correlated among patients within a given ILI cluster, violating the assumption of independent events. Varying contribution from LTC versus community settings may explain the otherwise surprising but significant differential in testpositivity among participants in adjacent study regions of the Lower Mainland, BC (14% in Fraser, 36% in Vancouver Coastal) despite the same program strategy. Although it was not provided, participant profiles showing case and vaccine groups additionally stratified by region and LTC status may help elucidate that variation. The systematic differences in vaccine distribution, sampling collection and exposure to influenza and other respiratory viruses in
this study represent selection biases that compromise the validity of the estimates and cannot be corrected in the analysis. Overall VE estimates based on bundled LTC and community participants are unreliable but the sample size with restriction to community-based participation alone was too limited for comparative VE analysis (30 cases in total across all study groups). The 95% confidence intervals around the point estimate of VE in the community were consequently wide for ATIV (ranging 2–93%) and were not specified at all for TIV to assess overlap. Selection bias due to non-comparability of study groups and differential case finding, as well as a lack of statistical power, seriously compromise the validity of this study [1]. Author statements that ATIV “significantly improved protection” but “TIV was ineffective” go far beyond what their data can support. References [1] Orenstein WA, Bernier RH, Hinman AR. Assessing vaccine efficacy in the field: further observations. Epidemiol Rev 1988;10:212–41. [2] Van Buynder PG, Konrad S, Van Buynder JL, Brodkin E, Krajden M, Ramler G, et al. The comparative effectiveness of adjuvanted and unadjuvanted trivalent inactivated influenza vaccine (TIV) in the elderly. Vaccine 2013;31(51):6122–8, http://dx.doi.org/10.1016/j.vaccine.2013.07.059. [3] Van Buynder P. Adjuvanted influenza vaccine effectiveness in the elderly (65+ yrs). Clinical trials.gov identifier: NCT01424371. Available at: http:// clinicaltrials.gov/ct2/show/NCT01424371 [4] BC Centre for Disease Control. Influenza immunization coverage. Available at: http://www.bccdc.ca/imm-vac/BCImmunizationCov/flucoverage/default.htm [5] PHSA Laboratories Public Health Microbiology & Reference Laboratory. Influenza Like Illness (ILI) Outbreak Laboratory Form: Specimen Collection and Shipping. Available at: http://www.phsa.ca/NR/rdonlyres/AC46FB9F44FE-487C-9BF4-5C0578F21EF3/0/ILIFormOct2012.pdf
Gaston De Serres ∗ Institut national de santé publique du Québec, Biological, Environmental and Occupational Risks, 2400 d’estimauville, Quebec City, Quebec, Canada G1E 7G9 ∗ Tel.:
+1 418 666 7000x274. E-mail address: [email protected] 6 November 2013 12 December 2013 Available online xxx
0264-410X/$ – see front matter © 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.vaccine.2013.12.032
Please cite this article in press as: De Serres G. Selection bias in the comparison between adjuvanted and unadjuvanted influenza vaccine effectiveness. Vaccine (2014), http://dx.doi.org/10.1016/j.vaccine.2013.12.032
