Assessing Varicella Vaccine Efficacy

Reply: Wood et al have questions regarding the statistical methods for the article by Kuter et al. The varicella vaccine efficacy rate reported in the article by Kuter measures the relative reduction of the incidence of varicella after administration of varicella vaccine. For this purpose, the average (or mean) annual varicella incidence rate was used to calculate the vaccine efficacy. Specifically, the vaccine efficacy based on the long-term follow-up of vaccine recipients was calculated as 1 − (observed averaged annual varicella incidence rate in vaccinated individuals)/(age-adjusted expected annual varicella incidence rate in susceptible unvaccinated individuals). The observed average annual varicella incidence rate was defined as the ratio of the total number of varicella events observed to the total person-years at risk in the follow-up period. The age-adjusted expected annual varicella incidence rate was calculated on the basis of historical age-specific rates among susceptible unvaccinated individuals.1,2 This method follows a similar approach commonly used to evaluate the relative risk reduction of event rates over a period of follow-up in epidemiologic studies.3 Throughout the 10 years of follow-up in the study, the annual varicella rate after receipt of either 1 or 2 doses of varicella vaccine remained very low. The rate in years 7–10 was particularly low (0.2–0.6%/yr after 1 dose and 0% in all 4 years after 2 doses). The efficacy presented in the article represents an average efficacy over a 10-year period. If one were to calculate the yearly efficacy for each of the 10 separate years of the study, the yearly efficacy would range from 87% to 98% after 1 dose and 94% to 100% after 2 doses with higher efficacy in years 7–10. Thus, the average estimate of approximately 94% efficacy after 1 dose and approximately 98% efficacy after 2 doses provides an accurate representation of vaccine efficacy over the 10-year period, with no evidence of waning efficacy. Wood et al have suggested that calculation of the vaccine efficacy in this study using a cumulative varicella event rate over the entire follow-up period is more appropriate. The estimated vaccine efficacy based on cumulative incidence that they have calculated is lower than the estimated vaccine efficacy calculated based on the annual incidence rate. This is expected, because the vaccine efficacy calculated based on cumulative incidence depends on the length of the follow-up period and the number of susceptible individuals remaining at the end of the follow-up. Because the number of individuals at risk usually decreases during the follow-up period, this method generally produce an efficacy estimate that is lower than the true vaccine efficacy and it tends to decline as the length of follow-up increases regardless of the true vaccine efficacy. These 2 undesirable properties are significant drawbacks associated with an efficacy calculation based the cumulative incidence method, but not with the method based on the average annual incidence rate. For example, if one assumes that the true disease incidence rates for a vaccine among vaccine recipients and unvaccinated controls are constant throughout a 10-year period at 2% and 10% per year, respectively, then the true vaccine efficacy should be a constant of 80% (as it would be if it was calculated based on average annual incidence) throughout the follow-up period. However, the use of cumulative incidence rates in such an example would yield vaccine efficacy of 80% ([1 − 0.02/0.10] × 100%) at year 1, 79.2% ([1 − 0.02 (1 + (1 − 0.02))/[(0.10 (1 + (1 − 0.10)))] × 100%) at year 2, 76.5% at year 5, and 71.9% at year 10. Therefore, vaccine efficacy calculated based on the cumulative incidence is a less accurate measure of the true relative reduction of the incidence rate than a calculation based on the average annual incidence rate. As noted in the original manuscript, this study was conducted in both the prevaccine era (before 1995) when the probability of exposure to natural varicella remained high in this cohort and also in the postvaccination era when increased use of the vaccine may have decreased the possibility of community exposure. However, the exact role played by exogenous versus endogenous boosting in the persistence of varicella antibody in this population cannot be clearly established. We also understand that use of the historic rate of varicella disease in the household4 was based on prevaccine experience and could change with increased vaccine uptake. Vaccine efficacy estimates generated in clinical trials are generally higher than those seen in vaccine effectiveness studies conducted under real world conditions. We recognize that in routine practice, the effectiveness of a single dose of varicella vaccine has been ∼80-85%.5 If we apply the 3.3-fold reduction in varicella cases seen in this 10-year study with the use of a second dose, this would translate into a 2-dose varicella vaccine effectiveness in routine practice of ∼93% if effectiveness after 1 dose is approximately 80%. It is important to note that regardless of the method used, the differential effectiveness for 2 doses observed in either case supports the current ACIP 2 dose recommendation. We do not believe that the concerns of Wood et al are appropriate.