Supplementary Materials Determining the optimal vaccination schedule for herpes zoster: a cost-effectiveness analysis. Phuc Le, PhD, MPH and Michael B. Rothberg, MD, MPH. Long term efficacy of the live attenuated herpes zoster vaccine To model efficacy against HZ incidence over time, we fit a linear function using 6 yearly efficacy data points reported in the Short-Term Persistence Substudy (STPS) (1). Due to relatively small numbers of patients in the Long-Term Persistence Substudy (LTPS) (2), confidence intervals around efficacy were wide for individual years. To increase estimate stability, we aggregated efficacy for the whole LTPS study period and plotted it at 7.7 years (a weighted average for the follow-up period from 4.7 to 11.6 years). The resulting linear function (y = 0.6478 0.0544 x year) produced the best fit with the smallest root mean squared error values compared to exponential and logarithmic models. The plot of residuals against predicted values also showed a randomly scattered distribution of points, confirming the choice of linear model. To model the long-term efficacy for BOI, we again ran a linear regression on the three aggregated data points reported for SPS, STPS and LTPS (1-3). The trials suggested that the efficacy for BOI reduced over time although the duration was a little longer than efficacy for HZ incidence. The final BOI efficacy function, which had the best fit with observed data, was: y = 0.0437 x year + 0.7083. In the SPS, the efficacy for PHN incidence was approximately 66% and did not differ by age (3). However, we re-estimated this efficacy to reflect the fact that a higher than expected percentage of subjects in the placebo group experienced HZ and PHN in the first year (4). The adjusted efficacy for PHN incidence was 63.2% (95% CI, 42.2%-76.5%). Because the yearly 1
efficacy estimates in the STPS did not show a clear pattern and had wide confidence intervals, we assumed a stable efficacy for the first 5 years. At year 6, the efficacy for PHN incidence declined to 60.1% (1) and fell to 35.4% by year 9 post-vaccination (2). We described this decline after 5 years with the following function: y = 0.1 x year + 1.218. The efficacy against PHN became nil at year 12. As noted in the methods section, the efficacies for BOI and PHN incidence reported in clinical trial were for the whole population, so incorporated the vaccine effects on HZ incidence. In order to estimate the additional efficacies for BOI and PHN incidence among HZ cases, we used the following equation: x = Eff Eff HZ 1 Eff HZ where x is the efficacy for BOI (or PHN incidence) among HZ cases, Eff is the overall efficacy for BOI (or PHN incidence) reported in the trial, and EffHZ is the efficacy for HZ as estimated by the linear functions above. In our model, the vaccine efficacy for HZ incidence was applied first to reduce the probability of experiencing HZ. The additional efficacy for BOI was then applied to reduce the number of quality-adjusted life years lost among vaccinated HZ cases compared to unvaccinated HZ cases. Finally, the probability of developing PHN was reduced in proportion to the additional vaccine efficacy for PHN incidence among HZ cases. For example, an unvaccinated 60-year old man would have an annual probability of HZ of 0.008. If he experienced the disease, he would lose 0.0129 QALYs from BOI, and have a 0.054 probability of developing PHN. If he instead was vaccinated, the probability of HZ would be 0.008 x (1 EffHZ). Because the vaccine is not 100% efficacious, he might still get the disease. If he did the QALY loss due to HZ would be 0.0129 x (1 EffBOI) QALYs. The probability of developing 2
PHN would also be lower, 0.054 x (1 EffPHN). Appendix Figure 2 presents the reduction of vaccine efficacy over time. Appendix Table 1. Vaccine efficacy estimated by the model versus reported by the clinical trial Follow up period Vaccine efficacy for incidence of HZ (%) Vaccine efficacy for incidence of PHN (%) Vaccine efficacy for HZ BOI (%) Model Trial, point estimate (95% CI) Model Trial, point estimate (95% CI) Model Trial, point estimate (95% CI) SPS 52.7 51.3 63.2 64.6 61.1 60.3 (44.2 57.6) (42.2 76.5) (51.1 69.1) STPS 40.6 39.6 60.1 50.1 60.4 50.3 (18.2 55.5) (-8.8 86.7) (14.1 71.0) LTPS 22.3 21.1 35.4 37.3 38.6 35.7 (10.9 30.4) (8.8 55.8) (26.7 46.4) Note: a We re-estimated this efficacy to reflect the modification of the percentage of subjects with HZ who experienced PHN in the placebo group in year 1 to 11% (4). The original efficacy reported in the trial was 66.5% (95% CI: 47.5-79.2) (3). HZ, herpes zoster; PHN, postherpetic neuralgia; BOI, burden-of-illness; SPS, Shingles Prevention Study; STPS, Short-term Persistence Sub-study; LTPS, Long-term Persistence Sub-study; CI, confidence interval. 3
Go to Sub-tree Appendix Figure 1: Markov model. HZ, Herpes zoster; PHN, postherpetic neuralgia 1.a. Decision node and Markov states. The model begins with a decision node representing the choice to vaccinate at different ages including booster(s) or not. The cohort then moves to a chance node, represented by a circle, of male or female depending on the sex distribution of general population, and enters the Markov node afterwards, denoted by the letter M inside a circle. For the first cycle, the entire cohort enters the Healthy state, then moves between Markov health states depending on transition probabilities in subsequent cycles until everyone is subsumed by the Dead state, at which point the model terminates (5). 4
1.b. Sub-tree. This represents the chance events occurring within each annual cycle for people starting at the Healthy state. For any subsequent cycles beginning in the Monocular Blindness, Monaural Deafness, or Monocular Blindness & Monaural Deafness states, the tree is identical, but cohort members return to the corresponding initial states instead of Healthy after recovering from HZ and any short-term complications (hospitalization and PHN). Those who acquire a second disability move to the combined disability state. Bilateral blindness and deafness are extremely rare events and are not considered (5) 5
Vaccine efficacy Vaccine efficacy Vaccine efficacy a) Efficacy against herpes zoster incidence (5) 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 2 4 6 8 10 Number of years after vaccination b) Efficacy against burden of illness 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 2 4 6 8 10 Number of years after vaccination c) Efficacy against post-herpetic neuralgia incidence 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 2 4 6 8 10 12 14 Number of years after vaccination Appendix Figure 2: Vaccine efficacy over time 6
References 1. Schmader KE, Oxman MN, Levin MJ, Johnson G, Zhang JH, Betts R, et al. Persistence of the efficacy of zoster vaccine in the shingles prevention study and the short-term persistence substudy. Clin Infect Dis. 2012 Nov 15;55(10):1320-8. 2. Morrison VA, Johnson GR, Schmader KE, Levin MJ, Zhang JH, Looney DJ, et al. Long-term persistence of zoster vaccine efficacy. Clin Infect Dis. 2015 Mar 15;60(6):900-9. 3. Oxman MN, Levin MJ, Johnson GR, Schmader KE, Straus SE, Gelb LD, et al. A vaccine to prevent herpes zoster and postherpetic neuralgia in older adults. N Engl J Med. 2005 Jun 2;352(22):2271-84. 4. Brisson M, Pellissier JM, Levin MJ. Cost-effectiveness of herpes zoster vaccine: flawed assumptions regarding efficacy against postherpetic neuralgia. Clin Infect Dis. 2007 Dec 1;45(11):1527-9. 5. Le P, Rothberg MB. Cost-Effectiveness of Herpes Zoster Vaccine for Persons Aged 50 Years. Ann Intern Med. 2015 Oct 6;163(7):489-97. 7