To the Editor:

We appreciate the comments shared by Wang and colleagues.1 Here we provide responses to the key comments that were raised and highlight relevant aspects in our original publication2 and from our research.

The first point highlighted by Wang et al was regarding the limitations associated with using diagnosis codes for the identification of rheumatoid arthritis (RA) and herpes zoster (HZ) cases.1 As mentioned in the Discussion of our original publication,2 although the large sample sizes provided by the administrative claims database is a benefit of using such a data source, it is important to acknowledge there are limitations associated with this.2 One of these limitations, which was noted in the original publication, is that the study relied on “diagnosis codes that were used previously by other researchers to define HZ cases, [although] it is possible that some HZ cases may not have been identified or some cases may not have been true cases of HZ.”2,3 Additionally, the case definition for RA was based on multiple studies using administrative claims data to study this population, as cited in the Methods section of the original publication.2-7 This definition included the requirement of multiple medical claims with an RA diagnosis, specifically at least 2 claims a minimum of 6 weeks apart, and evidence of prescription medication use for treatment of RA in medical or pharmacy claims for at least 3 months.2 Although we aimed to use the best available evidence to inform exposure and outcome definitions, we generally agree that combining administrative claims data with more detailed clinical data from electronic health records could provide further certainty around case identification; however, this was beyond the scope of our work.

Another point noted as a limitation was the reported differences in the distribution of certain demographic and clinical characteristics across the RA and non-RA cohorts in the study. It is accurate that the mean age of the RA cohort was higher than that of the non-RA cohort and that a number of comorbidities measured in the study were more prevalent in the RA cohort than in the non-RA cohort. The distribution of sex was also uneven between the RA and non-RA cohorts. Although these represent important imbalances in confounders between the cohorts, we adjusted for these differences using a doubly robust approach, as described in the Methods.2 This approach adjusted for confounding variables by estimating propensity scores, which were used, along with each variable considered in the propensity scoring model, to adjust the outcome model for confounders. Using this approach, it was possible to control for differences observed between the demographic and clinical characteristics noted by Wang and colleagues.1 Additionally, by using a doubly robust approach, the results were robust to model misspecification.8 Beyond adjusting for these important confounders and reporting adjusted effect estimates in the publication, adjusted age-stratified analyses were also reported, further supporting the results of the adjusted analyses in the overall study population.

The additional risk factors noted by Wang et al1 were measured in the study and accounted for in the adjusted analyses. However, the original publication did not report full details on some of these measures. To provide clarity on the relevant confounders, the Table presented here provides additional data to address these comments, expanding on details reported in Supplementary Table S3 of the original publication.2 Of note, some of the key immunosuppressive conditions mentioned are also measured in the updated Charlson Comorbidity Index9 and, therefore, using this measure, are considered to have been presented in the original publication.2 The full details on the distribution of covariates presented in Supplementary Table S3 of the original publication2 are shown (Table).

Full CCI component comorbidities, comorbidities potentially associated with HZ, and additional immunosuppressive conditions.

We hope that highlighting these aspects of our research and the relevant content in the original publication,2 as well as providing these additional details on our results, will help to support the interpretation of the study findings.

GlaxoSmithKline Biologicals SA funded this study (GSK study identifier: VEO-000042) and was involved at all stages of study conduct. GlaxoSmithKline Biologicals SA also funded all costs associated with the development and publication of this manuscript, in accordance with Good Publication Practice 2022 guidelines (https://www.ismpp.org/gpp-2022).

DS and SP are employees in the GSK group of companies and report holding shares in the GSK group of companies. PTL, DG, WYC, and MSD are current employees and FD is a former employee of Analysis Group Inc., which received funding from the GSK group of companies to conduct the study disclosed in this publication. SM has received grant support for vaccine studies from the GSK group of companies. JRC is a member of the American College of Rheumatology Vaccine Guideline Committee.