Herpes Zoster Risk in Patients with Rheumatoid Arthritis and Its Association with Medications Used

Rheumatoid arthritis (RA) was associated with the risk of incident herpes zoster (HZ), which might be influenced by medication use by RA patients. We aimed to investigate the association of RA with the risk of incident HZ and how the HZ risk effected by RA medications in CIC RA patients. We conducted an observational study including population-based representative insurance claims data of 19,673 patients with RA and 39,346 matched patients without RA during 1997–2010 from the Taiwan National Health Insurance Research Database; we identified 1651 patients with catastrophic illness-certified (CIC) RA and 11,557 matched patients with non-CIC RA. Exploratory analyses assessed the association between RA/CIC RA and risk of incident HZ and its complications. The association of prescribed medications with HZ risk in CIC RA patients was also estimated. The incidence rates of HZ were higher in CIC RA patients and non-CIC RA than in the matched people without RA (21.95 and 14.03 vs. 7.36 events per 1000 person-years, respectively). The adjusted incidence rate ratio (95% confidence interval (CI)) for HZ was 1.74 (1.65–1.84) in RA patients vs. matched non-RA and 1.65 (1.44–1.89) in CIC RA patients vs. non-CIC RA. For HZ complications, RA had a 2.85-fold higher risk than non-RA, and CIC RA had a 1.78-fold higher risk than non-CIC RA. Moreover, in CIC RA patients, prednisolone use was associated with incident HZ risk compared with prednisolone nonuse (adjusted odds ratio 1.48, 1.08–2.03); prolonged prednisolone use (approximately 5 years) increased the risk (adjusted odds ratio 2.16, 1.46–3.19). Our results suggested that RA was positively associated with HZ risk, particularly in RA patients with prednisolone use.


Introduction
Herpes zoster (HZ) is a neurocutaneous disease caused by the reactivation of latent varicella-zoster virus (VZV) [1]. Early treatment is the best method to improve the quality of life and prevent postherpetic complications. VZV belongs to the human alphaherpesviruses and can cause two disease manifestations: chickenpox (i.e., varicella) and shingles (i.e., HZ) [1]. Chickenpox is a common contagious childhood disease characterized by itchy blisters, but it rarely causes serious problems [2]. VZV persists in the sensory ganglia of the cranial nerves and in the spinal dorsal-root ganglia after varicella resolves and can become reactivated after many decades of latency [1,2]. HZ may have the direct all individuals insured under Taiwan NHI. In the NHIRD, diagnosis coding is performed according to the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) diagnostic criteria. The study protocol was reviewed and approved by the Institutional Review Committee of Kaohsiung Medical University Hospital, Taiwan (KMUHIRB-EXEMPT (I)-20190011).

Inclusion and Exclusion Criteria
We included patients aged ≥20 years who received a new diagnosis of HZ. The patients were followed for a maximum of 14 years. For patients with RA, the exclusion criteria were as follows: RA or HZ diagnosed before January 1, 1997; prior diagnosis of HZ or AIDS/HIV; and time between the date of first diagnosis of RA to index date of HZ < 1 month. In the compared group (non-RA), the exclusion criteria were as follows: rheumatoid disease (ICD-9-CM 446. 5

Source of Data and Study Population
In this retrospective cohort study, the NHIRD was used. This dataset represents a nationally representative group of 1 million individuals randomly selected from among all individuals insured under Taiwan NHI. In the NHIRD, diagnosis coding is performed according to the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) diagnostic criteria. The study protocol was reviewed and approved by the Institutional Review Committee of Kaohsiung Medical University Hospital, Taiwan (KMUHIRB-EXEMPT (I)-20190011).

Inclusion and Exclusion Criteria
We included patients aged ≥20 years who received a new diagnosis of HZ. The patients were followed for a maximum of 14 years. For patients with RA, the exclusion criteria were as follows: RA or HZ diagnosed before January 1, 1997; prior diagnosis of HZ or AIDS/HIV; and time between the date of first diagnosis of RA to index date of HZ < 1 month. In the compared group (non-RA), the exclusion criteria were as follows: rheumatoid disease (ICD-9-CM 446. 5

Ascertainment of RA and CIC RA
Non-RA patients (compared group) were selected from a random sample of the ambulatory care data file of the 2010 Longitudinal Health Insurance Database (LHID2100) with the inclusion period of 1 January 1997 to 31 December 2010. The primary case definition of RA was a physician-recorded primary diagnosis (ICD-9-CM 714.0) at an outpatient or inpatient visit. In step one, 19,673 patients with RA and 39,346 matched patients without RA (1:2 ratio) were identified. However, this study is based on claims data, so there is always the inherent information bias; hence, to evaluate the robustness of case assessment, we stratified patients based on whether patient had catastrophic illness-certified (CIC) RA with prescribed medication use. In Taiwan, patients with RA can apply for catastrophic illness certification to be exempted from copayments for healthcare costs related to RA [33,34]. Patients with CIC RA are required to have thorough clinical and

Ascertainment of RA and CIC RA
Non-RA patients (compared group) were selected from a random sample of the ambulatory care data file of the 2010 Longitudinal Health Insurance Database (LHID2100) with the inclusion period of 1 January 1997 to 31 December 2010. The primary case definition of RA was a physician-recorded primary diagnosis (ICD-9-CM 714.0) at an outpatient or inpatient visit. In step one, 19,673 patients with RA and 39,346 matched patients without RA (1:2 ratio) were identified. However, this study is based on claims data, so there is always the inherent information bias; hence, to evaluate the robustness of case assessment, we stratified patients based on whether patient had catastrophic illnesscertified (CIC) RA with prescribed medication use. In Taiwan, patients with RA can apply for catastrophic illness certification to be exempted from copayments for healthcare costs related to RA [33,34]. Patients with CIC RA are required to have thorough clinical and laboratory evaluations, fulfillment of appropriate classification criteria, and review by a physician commissioned by the NHI Bureau, and thus, their data are highly accurate and reliable [33]. Once one is evaluated as having CIC RA, the status is life long, and there is no need for further evaluations. In step two, 1651 RA patients with CIC RA and 11,557 matched patients with non-CIC RA (1:7 ratio) were included. In step three, the associations of prescribed medications and HZ were estimated among 1651 patients with CIC RA.

HZ Outcomes Assessment
HZ events were defined as new HZ events (ICD-9-CM 053.x-including phenotypes of HZ with meningitis, 053.0; HZ with nervous-system complications, 053.1; and HZ with ophthalmic complications, 053.2) identified at an outpatient or inpatient visit.

Comorbidities
In addition to the demographic risk factors of age, sex, and region, we evaluated other potentially confounding factors: lipid metabolism disorders, obesity, alcohol abuse, hypertension, myocardial infarction, congestive heart failure, peripheral vascular diseases, cerebrovascular diseases, dementia, chronic pulmonary disease, peptic ulcer disease, mild liver disease, diabetes without chronic complications, diabetes with chronic complications, hemiplegia or paraplegia, renal diseases, any malignancy (including leukemia and lymphoma and excluding malignant neoplasms of skin), moderate or severe liver disease, and metastatic solid tumor-all conditions diagnosed according to ICD-9-CM [34].

Statistical Analysis
Propensity score-matched analysis was performed on the association between RA and non-RA and CIC RA and non-CIC RA. Continuous and categorical variables were analyzed using t or Wilcoxon rank sum tests and chi-squared tests, respectively; the values obtained for the RA and matched compared group were compared. The incident rate ratios (IRRs) were calculated using the generalized log-linear model and by performing Poisson regression analysis. For follow-up analysis, we calculated RA and incident HZ risk at 1-year follow-up. Potential risk factors, such as comorbidities, were incorporated into the model. Interactions between RA and sex were tested using the generalized linear model with an added interaction term and potential risk factors. The associations of patients with CIC RA and prescribed medications and HZ were estimated, and adjusted odds ratios (ORs) were calculated after adjustments for covariates by using a multiple logistic regression model. All statistical analyses were performed using SAS (version 9.4, SAS Institute, Cary, NC, USA).

Results
For step one, Table 1 presents the demographic characteristics of the study population. We identified 19,673 RA cases (mean age: 46.2 ± 13.6 years) and 39,346 matched non-RA group individuals (mean age: 46.0 ± 13.6 years). Patients with RA had significantly higher frequencies of comorbidities (p < 0.05), except for dementia, moderate or severe liver disease, and metastatic solid tumor (p > 0.05).
In univariate and multivariate analyses, RA was associated with HZ ( Table 2). The incidence rate of HZ was higher in patients with RA than in the matched non-RA group (14.28 versus 7.36 events per 1000 person-years, adjusted IRR 1.74, 95% CI 1.65-1.84 with p <0.0001). We also found that RA increased HZ risk at the 1-year follow-up (adjusted IRR 30.42, 95% CI 18.02-51.36 with p < 0.0001). In the RA patient group, the RA with CIC RA matched with RA without CIC RA (adjusted IRR 1.54, 95% CI 1.35-1.77 with p < 0.0001); see Table 3. Table 1. Characteristics of patients with rheumatoid arthritis (RA) and RA with a catastrophic illness certificate (CIC).
Step One Step Two   For step two, Table 1 presents the demographic characteristics of the study population among patients with RA. We identified 1651 patients with CIC RA (mean age: 46.2 ± 12.9 years) and 18022 patients with non-CIC RA (mean age: 46.2 ± 13.6 years), from which we selected 11557 (1:7) matched patients (mean age: 46.2 ± 12.8 years). Patients with CIC RA had lower frequencies of lipid metabolism disorders, obesity, diabetes without and with chronic complications, hemiplegia, and paraplegia; patients with CIC RA had higher frequencies of chronic pulmonary disease, peptic ulcer disease, renal disease, and any malignancy comorbidities (p < 0.05). In univariate and multivariate analyses, CIC RA was associated with HZ (Table 3); the incidence rate of HZ was higher in patients with CIC RA than in matched patients with non-CIC RA (21.95 vs. 14.03 events per 1000 person-years, adjusted IRR 1.65, 95% CI 1.44-1.89). Moreover, CIC RA and the related incident HZ risk at the 1-year follow-up were analyzed (adjusted IRR 1.79, 95% CI 1.21-2.57; Table 3). CIC RA and the related incident HZ risk levels at the ≤1 year, >1 to 3 years, and >3 to 6 years of follow-up were analyzed (adjusted IRR 1.76, 1.73, 1.85, respectively; Table 3).

RA
The cumulative incidence rates of herpes zoster of those with and those without RA, and those with RA with a CIC and without a CIC (using the log-rank test and Kaplan-Meier analysis), are significantly different, with log rank p < 0.0001 (Figure 2a,b). In Table 4 are the results of the control group and the RA patients for HZ complications (the adjusted hazard ratio (HR) was 2.85, 95% CI 2.53-3.21). For the RA groups, the HR of CIC RA associated with getting HZ complications (the adjusted HR was 1.78 95% CI 1.39-2.29), and both relations are statically significant at p < 0.0001. Hazard ratio (HR) was calculated by using a Cox proportional hazards regression model. In Table 4 are the results of the control group and the RA patients for HZ complications (the adjusted hazard ratio (HR) was 2.85, 95% CI 2.53-3.21). For the RA groups, the HR of CIC RA associated with getting HZ complications (the adjusted HR was 1.78 95% CI 1.39-2.29), and both relations are statically significant at p < 0.0001.
In step three, the associations of medications prescribed to CIC RA patients and HZ were estimated. Table 5 presents the HZ patients with medications used for CIC RA. Note that 27.4% of HZ patients had nervous-system complications. HZ incidence was significantly higher in women (86.5%; p = 0.0018) and in patients with prednisolone use (76.1%; p = 0.0095). We also tried to estimate prednisolone use's association with the risk of HZ, but the difference was only marginally significant between HZ and non-HZ in non-CIC RA patients (p = 0.0874; Supplementary Tables S1 and S2).  To further evaluate the effect of prednisolone, use on HZ risk in patients with CIC RA, we classified the average prednisolone dose by stratifying the prednisolone exposure into yes or no and categorizing the total days and total dose (in mg) according to a quartile method. Synergistic effects of DMARDs or biopharmaceuticals with prednisolone were observed. Table 6 presents the relationship between prednisolone use or not and HZ risk (76.1% vs. 68.0%; adjusted OR 1.50, 95% CI 1.10-2.04). Patient prednisolone use analyzed by days and total dose revealed an increased HZ risk in the respective last quartiles (OR 2.25, 95% CI 1.54-3.30, and OR 2.04, 95% CI 1.39-3.00, respectively). The aforementioned results suggest a positive relationship between long-term prednisolone use (approximately 4.92 years (i.e., 1795 days)) and HZ risk, indicating that prednisolone was a major risk factor of HZ in patients with CIC RA. DMARDs combined with prednisolone showed an increased HZ risk compared with untreated patients (73.7% vs. 65.7%; OR 1.66, 95% CI 1.07-2.60 p = 0.0249), suggesting that prednisolone is often used in combination with DMARDs for treating RA in Taiwan. However, combining biopharmaceuticals with prednisolone did not significantly increase the risk of HZ (OR 1.31, 95% CI 0.81-2.12 p = 0.2740). Additionally, the biopharmaceuticals etanercept and adalimumab (anti-TNF-α agents) and rituximab (B-cell-depleting agent) had no significant association with HZ risk (Tables 5 and 6) DMARDs: disease-modifying antirheumatic drugs. We classified the average prednisolone dose by using three approaches: stratifying the prednisolone exposure into yes or no and categorizing the total days and total dosage (mg) according to a quartile method. Adjusted odds ratio (OR) was calculated after adjustment for age group, sex, and comorbidities by using a multiple logistic regression model. * Per 1825 mg/year was derived as follows: 5 mg/days × 365 days.

Discussion
In this observational study, patients with RA had a 1.94-fold increased HZ risk compared with the matched non-RA group in Taiwan, which is consistent with a Japanese study (2.19-fold) showing an around two-times-higher chance of getting HZ for RA patients [28]. Among the RA group, patients with CIC RA had a 1.56-fold increase in HZ risk compared with patients with non-CIC RA. Notably, patients with RA during 1-year of follow-up were highly associated with HZ compared with the matched non-RA group (adjusted IRR 30.42, 95% CI 18.02-51.36 with p < 0.0001). Taken together, patients with RA were more likely to develop HZ (Tables 2 and 3). Our results also showed that the nervous-system complications in CIC RA patients occurred 27.4% of the time with HZ infection, which might impair the quality of life of those patients. We also revealed that patients with CIC RA taking prednisolone had a dose-dependent increase in HZ risk (5 mg/day per year, OR 1.04, 1.01-1.07; Table 6). By contrast, the use of the biopharmaceuticals etanercept, adalimumab, and rituximab led to no significant increase in the risk of HZ in patients with CIC RA. HZ risk decreased with the use of biopharmaceutical agents only (the adjusted OR was 0.76 95% CI 0.17-3.39 p = 0.7217) when compared to using both biopharmaceutical agents and prednisolone. Then, with the use of only prednisolone (the adjusted OR was 1.50 95% CI 1.08-2.07 p = 0.0146) compared to combined treatment, the OR slightly dropped, though not significantly (the adjusted OR was 1.31 95% CI 0.81-2.12 p = 0.2740) ( Table 6). The use of new treatments might reduce the hazard ratio of steroid use, but the numbers of patients on those new treatments is still very low compared to those on DMARDs and prednisolone, but we can pay more the attention to the new treatment outcomes in the further study.
Accumulating evidence from epidemiology studies and clinical trials has revealed that RA is associated with increased HZ risk [17,18,35]. Declining virus-specific cellular immunity to VZV infection occurs both naturally by aging and by immunosuppressive treatments, predisposing RA patients to an increased HZ risk within a few years of disease diagnosis [4,6]. In the prebiologic era, in addition to disease itself, several drugs used to treat RA, particularly corticosteroids, were implicated as potentially increasing the HZ risk [22]. Corticosteroids have been reported to increase the HZ risk not only in patients with RA, but also in patients with other autoimmune diseases, such as systemic lupus erythematosus [16,18,35,36]. In agreement with previous observations, our data also indicate that corticosteroids (prednisolone) increased the HZ risk in patients with CIC RA. Moreover, the increase in risk was dose-dependent: an approximately 4% to 6% increase in risk if the daily prednisolone dose was >5 mg in patients with CIC RA.
Among the DMARDs, hydroxychloroquine and MTX have been most frequently mentioned to increase the HZ risk [18,37]. In our data, we did not find a significant difference in the percentage of patients with CIC RA, with or without HZ, taking hydroxychloroquine (79.9% vs. 74.2%) or MTX (61.8% vs. 60.1%). In our study, DMARDs played a notable role: when used alone, they were not associated with an increased risk, but when combined with corticosteroids (prednisolone), they increased the HZ risk 1.67-fold compared with untreated patients (Tables 5 and 6). This indicates synergistic effects of DMARDs with prednisolone. As DMARDs are the most indispensable drugs used for RA treatment and corticosteroids are often used in combination with DMARDs, the risk-benefit balance should be adjusted individually.
Since the introduction of biologics, the risks associated with TNF inhibitors have been widely discussed, but the data are conflicting [15,16,35,[37][38][39]. A meta-analysis revealed a significantly increased risk of up to 61% [38]. By contrast, a large US multi-institutional collaboration study did not find an association between TNF inhibitor initiation and HZ risk [15]. Our data showed that TNF inhibitors (and rituximab, a B-cell-depleting agent) were not significantly associated with HZ risk in patients with CIC RA. Notably, biopharmaceuticals combined with prednisolone did not significantly increase the HZ risk compared with untreated patients. Therefore, low dose prednisolone and early taping of prednisolone need to be considered to reduce the complications. Our results showed that the risk of HZ was increased in CIC RA patients. Therefore, HZ vaccinations may be required in these patients to improve their life quality [39,40].
There are several limitations to our study. First, the disease activity of each patient could not be identified from the NHIRD. Disease activity affects the patient's susceptibility to HZ, as discussed in previous reports [19][20][21][22]. Second, some of these patients may have sought treatment by using alternative medicines, which could have suppressed their immune function. For instance, triptolide is one such alternative medicine that is very often used to treat autoimmune diseases [41]. Finally, although we adjusted our data for age, sex, and comorbidities, some unknown confounding factors might still have impacted our conclusions.

Conclusions
Patients with RA are more likely to develop HZ early, particularly CIC RA patients at 1-year follow-up. We revealed that long-term use of corticosteroid (prednisolone; approximately 4.92 years) is strongly associated with HZ. Early tapering is therefore recommended to lower the risk of HZ. We found that in patients with CIC RA, the biopharmaceuticals had no significant association with HZ risk. Additionally, combining biopharmaceuticals with prednisolone did not significantly increase the risk of HZ. Our results might help physicians be alert to the possibility of early HZ in patients with RA, particularly in the first year of follow-up in patients with a daily prednisolone dose of >5 mg over 5 years. Our study might provide some important clinical messages to the physicians responsible for the decision making of RA patients' treatment to reduce the risk of HZ.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/ijerph20032123/s1. Supplementary Table S1. Characteristics of rheumatoid arthritis patients with a non-catastrophic illness certificate (non-CIC) with or without herpes zoster. Supplementary Table S2. Prednisolone use associated with HZ risk in non-CIC RA patients.