Effectiveness of Seasonal Influenza Vaccine against Pandemic (H1N1) 2009 Virus. Part 3
Ascertainment of Case-patients and Controls
Case-patients and controls were sampled prospectively throughout the study period. A case-patient was defined as a person with ILI for whom test results for pandemic (H1N1) 2009 were positive; a control was defined as a person with negative test results for influenza virus. Analysis of vaccine effectiveness against other influenza subtypes was not undertaken because of the almost exclusive circulation of pandemic (H1N1) 2009 virus during the season; therefore, patients with positive test results for other influenza viruses were excluded. A control could become a case-patient if another illness developed during the season, but a case-patient was no longer at risk and could not be included again.
Data Analysis and Calculation of Vaccine Effectiveness
All analyses were conducted by using Stata version 10.0 (StataCorp LP, College Station, TX, USA). The χ2 test was used to compare proportions, and the Mann-Whitney U test was used to compare time from vaccination to time seen by practitioner; p<0.05 was considered significant. Patients were excluded from the vaccine effectiveness analysis if vaccination status was unknown, if the date of symptom onset was unknown, or if the interval between symptom onset and specimen collection was >4 days (because of decreased likelihood of a positive result after this time) (21,22). Patients were considered not vaccinated if time between date of vaccination and symptom onset was <14 days. If only the month of vaccination was reported, the date of vaccination was conservatively estimated to be the last day of the month. To avoid overestimation of vaccine effectiveness arising from recruitment of controls when influenza was not circulating in the population, analysis was restricted to case-patients and controls detected within the influenza season, defined as the period during which influenza-positive case-patients were detected (weeks 26–40).
Vaccine effectiveness was defined as (1–odds ratio) × 100%; the odds ratio is the odds of laboratory-confirmed pandemic (H1N1) 2009 case-patients having been vaccinated divided by the odds of controls having been vaccinated. In the test-negative case–control design, the odds ratio estimates the incidence density (rate) ratio because controls are selected longitudinally throughout the course of the study (i.e., by density sampling). The odds ratio in test-negative case–control studies has also been shown to approximate the risk ratio under conditions of varying attack rates and test sensitivity and specificity (25). Logistic regression was used to calculate odds ratios and 95% confidence intervals (CIs) for having laboratory-confirmed pandemic (H1N1) 2009, which were adjusted for the variables of age group and month of specimen collection against the following: seasonal vaccine, monovalent vaccine, both vaccines, and any (either or both the seasonal and monovalent) vaccine. Sensitivity analyses were conducted to determine the effects of the following on vaccine effectiveness: not censoring for specimens collected from ILI patients >4 days after symptom onset, including controls recruited outside the defined influenza season, and assuming that patients with unspecified type A influenza had pandemic (H1N1) 2009.