From 2018 through 2022, 1 996 885 persons entered the study population and contributed more than 7.6 million person-years of follow-up. At study entry, 38% of the participants were aged 65 years or older, 53% were female, and 59% were White (Table). During the study period, 761 042 persons (38%) received at least 1 dose of RZV and 576 483 (29%) received 2 doses. The average duration of follow-up was 3.8 years per person. The average duration of follow-up in unvaccinated persons was 3.2 years, which included follow-up before vaccination as well as follow-up of never-vaccinated persons. The average duration of follow-up after vaccination was 1.6 years after the first dose and 1.4 years after the second dose.

Vaccine coverage increased slowly from 2018 through 2020 and more rapidly from 2021 through 2022 (Supplement Figure 1). Among the 1 419 944 persons still in follow-up as of 31 December 2022, 50.4% were unvaccinated, 11.5% had received 1 dose only, and 38.1% had received 2 doses. Among recipients of the first dose, an estimated 68% received the second dose within the recommended 6 months, 81% received the second dose within 1 year, and 90% received the second dose within 2 years (Supplement Figure 2). Throughout the study period, coverage was higher among persons aged 65 years or older compared with those aged 50 to 64 years (Supplement Figure 1).

During follow-up, 45 333 new HZ cases were identified, of which 42 798 (94%) were in unvaccinated persons. The unadjusted incidence of HZ per 1000 person-years was 6.7 among unvaccinated persons, 2.5 among partially vaccinated persons (1 dose only), and 1.7 among fully vaccinated persons (2 doses) (Figure 1).

After adjustment for covariates, VE against HZ was 64% for partially vaccinated persons and 76% for fully vaccinated persons (Figure 1). The analysis that was restricted to the prepandemic period yielded results similar to those for the overall study period; VE was 65% for partially vaccinated persons and 77% for fully vaccinated persons (Supplement Figure 3). The sensitivity analysis that omitted time-varying covariates also yielded similar results (Supplement Tables 1a and 1b).

Among fully vaccinated persons, VE waned very little with time since vaccination; VE was 79% during the first year, 75% during the second year, and 73% both during and after the third year. However, among partially vaccinated persons, VE decreased substantially, from 70% during the first year to 45% during the second year, 48% during the third year, and 52% after the third year (Figure 2).

Among fully vaccinated persons, VE was higher in those who were vaccinated before age 65 years (81%) than in those vaccinated at age 65 years or older (74%) (Figure 3). Vaccine effectiveness was lower in persons who were immunocompromised (defined on the basis of corticosteroid use) when they were vaccinated (65%) compared with those who were not immunocompromised when they were vaccinated (77%). However, VE was similar in persons who received the second dose later than recommended; it was 76% when the interval between doses was within the recommended 6 months, 78% when it was more than 6 months to less than 1 year, and 75% when it was 1 or more years.

Finally, among the covariates, older age, female sex, corticosteroid use, and higher outpatient visit frequency were associated with substantially higher risk for HZ. Risk for HZ was substantially lower in Black persons (vs. White persons and those in other races or ethnic groups) and in persons who had previously received ZVL (Supplement Table 2).

This large cohort study found that 2 doses of RZV had effectiveness of 76% against HZ. Effectiveness for fully vaccinated persons waned very little during the 4 years after vaccination. In contrast, although VE for partially vaccinated persons was 70% initially, effectiveness decreased substantially after the first year. Effectiveness for fully vaccinated persons did not differ by whether the second dose was given after the recommended 6 months. Effectiveness was slightly higher in persons vaccinated before age 65 years and somewhat lower in those who were immunocompromised due to corticosteroid use when they were vaccinated.

Our VE estimate of 76% was lower than the estimates from the 2 clinical trials, where VE was 97% among persons aged 50 years or older (7) and 90% among those aged 70 years or older (8). Our findings are generally consistent with those of 3 observational studies that reported VE estimates ranging from 70% to 86% (10–12). One reason our VE estimate was lower than the estimates in the trials may be that our case ascertainment was less specific than in the trials, where cases were confirmed by polymerase chain reaction (PCR) testing. Izurieta and colleagues reexamined data from the clinical trial and found that the VE estimate for persons aged 50 years or older would have been approximately 76%–which is similar to our VE estimate–if it had been based on clinically “suspected cases” before PCR confirmation (10).

We found that the effectiveness of 2 RZV doses waned very little during the 4 years of follow-up. This is broadly consistent with the long-term follow-up of the available trial participants, which found that VE waned only modestly during the 8 years, from 98% in year 1 to 90% in year 4 and 84% in year 8 (13).

Recombinant zoster vaccine is recommended as a 2-dose vaccine series. Although 1-dose VE was 70% in the first year, it decreased to 45% to 52% in the next 3 years, which supports the importance of the second dose. Notably, VE increased after the second dose regardless of whether it was administered within the recommended 6 months after the first dose. This finding provides reassurance about the effectiveness of the second dose even if receipt of it was delayed due to vaccine shortage or other factors.

Subgroup analyses found that VE was lower but still substantial among persons who might have been immunocompromised at the time of vaccination due to recent use of corticosteroids. A recent clinical trial among patients who were immunocompromised after an autologous stem cell transplant similarly reported substantial VE (68%) (16). Additional studies found VE of 87% for patients with hematologic cancer and 91% for those with potential immune-mediated diseases (16–18). Although we found that VE was lower in the subgroup who used corticosteroids in the 3 months before vaccination, we also found that corticosteroid users were at higher risk for HZ. Taken together, these findings suggest that the number of HZ cases prevented per 100 vaccine recipients in our study population was similar in corticosteroid users and nonusers.

Our study was strengthened by the large size, demographic diversity, and high vaccine coverage of the study population at the 4 VSD sites. Nearly 50% of the study population had received at least 1 dose of RZV by the end of the study period. In addition, our research design and analytic approach facilitated estimation of VE by time since vaccination, age at vaccination, immunocompromise status at vaccination, and time between the first and second doses. We were able to assess vaccine protection for 4 years after vaccination; previous real-world studies included only 1 to 2 years of follow-up (10, 11). The specification of Cox regression with a calendar timeline reduced the potential for confounding arising from the effects of the COVID-19 pandemic on seeking care for shingles; analyses were stratified by “risk sets” comprising persons who were at risk for HZ on the same date at the same VSD site.

The study also had limitations. First, our case ascertainment required an HZ diagnosis code together with an antiviral prescription. With this approach, HZ status was not identified as specifically as in the clinical trials, which required PCR confirmation. This may have lowered our VE estimates. Second, our case ascertainment likely missed some persons with relatively mild cases who did not seek care or presented too late for an antiviral prescription. If RZV was less likely to prevent these undetected milder cases, our VE estimate would be biased upward. Third, our study population was limited to people with health insurance and may not be representative of other populations. Fourth, approximately 27% of the study population was lost to follow-up before the end of the study–7% due to death and another 20% due to disenrollment from their health plan. Fifth, some potential confounders were measured imperfectly or not at all. Although many factors can contribute to immunocompromise, we only measured use of corticosteroids and thereby missed a substantial proportion of immunocompromised persons. However, in our study population, the prevalence of corticosteroid use during follow-up was similar in fully vaccinated and unvaccinated persons. If the prevalence of unmeasured immunocompromise was also similar in fully vaccinated persons and unvaccinated persons, it is unlikely that unmeasured immunocompromise biased our estimate of the effectiveness of full vaccination in the overall study population. On the other hand, if unmeasured immunocompromise was actually more prevalent in vaccinated persons than in unvaccinated persons, this would likely bias the VE estimate downward.

In conclusion, 2 doses of RZV showed high effectiveness that waned very little over 4 years. Our finding of substantial VE among corticosteroid recipients underscores the value of vaccination for these persons, who are at increased risk for HZ. Our finding that the effectiveness of 1 dose decreased after a year further supports the current recommendation for a second dose.