The risk of herpes zoster in patients with rheumatoid arthritis in the United States and the United Kingdom

Abstract

Objective

To determine whether the incidence of herpes zoster is elevated in patients with rheumatoid arthritis (RA) and whether herpes zoster is associated with use of disease-modifying antirheumatic drugs (DMARDs) in patients with RA.

Methods

Two retrospective cohort studies were conducted using data from a US integrated managed care database (PharMetrics claims database) from 1998–2002 and the UK General Practice Research Database (GPRD) between 1990–2001. Rates of herpes zoster among patients with RA and randomly sampled non-RA patients were compared. A nested case–control analysis was performed within each RA cohort to examine the effect of current treatment on herpes zoster risk.

Results

A total of 122,272 patients with RA from the PharMetrics database and 38,621 from the GPRD were included. The adjusted hazard ratios of herpes zoster for patients with RA compared with non-RA patients were 1.91 (95% confidence interval [95% CI] 1.80–2.03) in the PharMetrics database and 1.65 (95% CI 1.57–1.75) in the GPRD. In the PharMetrics database, current use of biologic DMARDs alone was associated with herpes zoster (odds ratio [OR] 1.54, 95% CI 1.04–2.29), as was current use of traditional DMARDs alone (OR 1.37, 95% CI 1.18–1.59). In the GPRD, current use of traditional DMARDs was associated with herpes zoster (OR 1.27, 95% CI 1.10–1.48). In both data sources, use of oral corticosteroids was associated with herpes zoster regardless of concomitant therapies.

Conclusion

Data from 2 large databases suggested that patients with RA are at increased risk of herpes zoster. Among patients with RA, DMARDs and/or use of oral corticosteroids appeared to be associated with herpes zoster.

INTRODUCTION

More than 90% of adults in the US have serologic evidence of prior infection with the varicella-zoster virus and are therefore at risk for herpes zoster (1). Although the precise pathogenesis is not fully understood, it involves reactivation of latent virus activity in the cranial and dorsal root ganglia (2–4). Herpes zoster is associated with a risk of disseminated infection and often considerable pain and morbidity. Identified risk factors for herpes zoster include older age, ethnicity, genetic susceptibility, underlying cell-mediated immune disorders, mechanical trauma, psychological stress, and immunotoxin exposure; exogenous boosting of immunity from varicella contacts has been shown to be protective (5). Herpes zoster commonly occurs in otherwise healthy individuals; however, immunocompromised patients appear to be particularly susceptible to both herpes zoster and its complications (2, 6–8).

Although it has been suggested that herpes zoster is increased in patients with rheumatoid arthritis (RA), there is little evidence to support this assertion. It is biologically plausible that having RA may result in an increased risk of herpes zoster due to the dysregulation of the immune system in patients with RA (9–11). In addition, because many of the medications used to treat RA function by modulating the immune system, they may be associated either directly or indirectly with an increased risk of herpes zoster. Anecdotal case reports and small studies have indicated the possibility that certain RA treatments might increase the risk (12–23). A recent study estimated that the annualized incidence rate of herpes zoster was 13.2 per 1,000 patient-years and found that various treatments, but not methotrexate or biologic agents, were risk factors for herpes zoster (24).

The objective of the present study was to estimate the rate of herpes zoster in persons with RA and to determine whether individuals with RA are at increased risk of herpes zoster compared with individuals without RA. In addition, we aimed to explore the relationship between current medication use and herpes zoster.

PATIENTS AND METHODS

Study design.

Cohort and case–control studies were conducted in 2 separate data sources using similar methods. In each data source, a cohort study was performed to compute incidence rates of herpes zoster and to compare these rates between RA and non-RA patients. A nested case–control analysis within each RA cohort was then performed to determine the effect of recent medications on the risk of herpes zoster; this is an efficient approach to assess risk because it addresses complex patterns of drug exposure over time with insignificant loss of power (25).

Data sources.

US PharMetrics claims database.

The PharMetrics integrated claims database includes information from fully adjudicated pharmacy, provider, and facility claims for members enrolled in 61 health plans across the US. Health plan members in the PharMetrics database are representative of the national commercially insured population on a variety of demographic measures including age, sex, geographic region, and plan type. Each claim in the database contains a unique encrypted patient identifier that can be used to assemble a longitudinal record of medical services and prescription drug use for each health plan member. Dates of eligibility, including interruptions, are available for the majority (∼80%) of members, enabling the calculation of person-time at risk. Data available for each medical claim include dates and location of service; International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) diagnosis codes; physician specialty; and procedure codes in the Physicians' Current Procedural Terminology (version 4) and the Health Care Financing Administration Common Procedure Coding System. Multiple diagnoses and procedures may be recorded for a single outpatient office visit or hospital admission. Data captured with each pharmacy dispensing include the drug dispensed in National Drug Code format, dosing, the date of dispensing, and the quantity and number of therapy-days dispensed. Data entered are subjected to rigorous data quality checks to ensure a standardized format and to minimize error. PharMetrics data have been used for studies in pharmacoepidemiology (26–29).

UK General Practice Research Database.

The UK General Practice Research Database (GPRD) contains computerized medical information entered by general practitioners (GPs) in the UK. At any time, approximately 3 million residents are registered with GPs who participate in the GPRD, representing a nearly 6% sample of the population of England and Wales from ∼400 general practices. Information recorded includes demographics, details of every GP consultation, a summary of specialists' clinical notes and hospital letters, and a register of written prescriptions. Significant laboratory and radiology findings, patient characteristics, patient history, and other relevant medical data may also be recorded. Medical diagnoses are coded using Oxford Medical Information System (OXMIS) and Read codes. Prescriptions are generated with office computers at the general practices and automatically recorded into the patient's computerized file using codes issued by the Prescription Pricing Authority of the UK National Health Service. Data for a given practice are coded as being up to standard when the practice shows competency at entering data into the electronic database and recording is verified to meet certain quality criteria. Data from the GPRD have been used extensively in the area of epidemiology; the overall high quality and completeness of the data has been documented in validation studies of the GPRD (30–34).

Study patients.

US PharMetrics claims database.

In the PharMetrics database, the RA cohort included all patients at least 18 years of age at the time of cohort entry who had at least one ICD-9-CM code for RA (714; excluding 714.3, 714.4, 714.9). Patients who were enrolled in Medicare gap plans, were missing a value for age or sex, had <90 days of continuous enrollment, or had a diagnosis of herpes zoster in the 90 days prior to cohort entry were excluded. The non-RA cohort comprised a random sample of 1 million individuals who met the same inclusion criteria but did not have an ICD-9-CM code for RA at any time during followup. Cohort entry was defined as the date of the first RA diagnosis for the RA cohort or the first medical claim for the non-RA cohort that occurred after health plan enrollment for at least 90 days or January 1, 1998, whichever was later. All patients were followed from the date of cohort entry until they developed herpes zoster, were no longer enrolled in the health plan, or December 31, 2002, whichever was earliest. For patients without enrollment data, the date of their last claim was used as a proxy for the end of enrollment.

UK General Practice Research Database.

We defined our GPRD study population as patients at least 18 years of age at cohort entry who had at least 90 days of up-to-standard data. The RA cohort was identified as patients who had a physician-recorded diagnosis of RA at any time in their medical record; the non-RA cohort included a random sample of 500,000 patients who had no OXMIS or Read code consistent with RA at any time. The start of followup in the RA cohort was the date of the first RA diagnosis, the patient's registration date, the date when their practice began submitting up-to-standard data, or January 1, 1990, whichever was latest. For the non-RA cohort, we defined cohort entry as the latest of the patient's registration date, the up-to-standard date for their practice, or January 1, 1990. Patients with a diagnosis of herpes zoster or a complication of herpes zoster prior to cohort entry were not included in either cohort. Patients in the GPRD were followed until they transferred out of the database, were diagnosed with herpes zoster, died, November 30, 2001, or the date when the data from their practice were no longer considered up to standard (if this occurred), whichever came first. We used 2001 as the end date to ensure that RA patients could not have been receiving biologic disease-modifying antirheumatic drugs (DMARDs) that would not have been captured in the general practice database.

Drug exposure.

For analyses within the RA cohorts, each cohort member's person-time was classified on a person-day basis according to RA medication exposure. Issued prescriptions were used as a proxy for medication exposure. Each person's sequence of prescriptions was evaluated to make the best determination of exposure status to both DMARDs and non-DMARDs on each day. For parenteral medications, the duration of exposure was determined based on the labeling information for that medication. For oral medications, exposure was approximated based on the number of days supplied, which was obtained from each database. If this value was missing, the value was calculated from the prescription quantity and the dose of medication or determined based on the median days prescribed for other dispensings of the medication. We assumed full medication adherence such that the length of exposure equaled the number of days of medication the patient received. As a sensitivity analysis, we extended the length of exposure for oral medications by 20%.

For the purpose of comparison, medications were grouped into DMARDs and non-DMARDs. DMARDs were further divided into traditional DMARDs and biologic DMARDs for the PharMetrics analysis. No RA patients in the GPRD were taking biologic DMARDs at the time the data were collected. Traditional DMARDs included methotrexate, azathioprine, tacrolimus, leflunomide, cyclosporine, cyclophosphamide, hydroxychloroquine, sulfasalazine, gold thiomalate, aurothioglucose, auranofin, and penicillamine. Biologic DMARDs included infliximab, etanercept, and anakinra.

Outcome.

Incident cases of herpes zoster were identified by the first herpes zoster diagnosis occurring after the patient's cohort entry date. In the PharMetrics database, cases of herpes zoster were identified from inpatient and outpatient encounters that included ICD-9-CM code 053 (herpes zoster). In the GPRD, cases were identified through OXMIS and Read codes. For all analyses, cases of herpes zoster were considered exposed to the drugs prescribed to the patient at the time of the event, the assumption being that herpes zoster is an acute event and therefore has a close temporal link to recent but not remote exposures.

Statistical analysis.

Rates of herpes zoster for the RA and non-RA cohorts were calculated as the number of events divided by the at-risk person-time and are presented as events per 1,000 person-years. Cox proportional hazards models were used to compare the rate of herpes zoster between the RA and non-RA cohorts, adjusting for age at cohort entry, sex, calendar year of cohort entry, and comorbid conditions (diabetes, chronic lung disease, cancer, and cardiovascular disease).

All incident cases of herpes zoster were included in the nested case–control analyses, and their event dates were used as the index dates. For each case, we formed a risk set of potential controls consisting of all RA cohort members with the same year and quarter of cohort entry and with followup at least as long as that of the case. A random sample of 7 controls was selected from each risk set. Controls had to be at risk, that is, still in the cohort and without a prior diagnosis of herpes zoster when selected to be a control. The index date for the controls was defined such that the controls had a followup time that was equal to that of the case. Drug exposure and patient characteristics for each case and the matched controls were evaluated on the index date. To account for the matched study design, we used conditional logistic regression models to examine the effect of current exposure to DMARDs and oral corticosteroids on the risk of herpes zoster. The reference group was no current exposure to these medications. With this study design, the odds ratios (ORs) were valid estimates of rate ratios. The following covariates were included: age, sex, current use of nonsteroidal antiinflammatory drugs (NSAIDs) and cyclooxygenase 2 (COX-2) selective inhibitors, comorbidities (diabetes, chronic lung disease, and cancer), orthopedic procedures, number of visits to a health care provider between cohort entry and the index date, and whether or not the patient saw a rheumatologist during followup.

RESULTS

US PharMetrics claims database.

A total of 122,272 patients were included in the PharMetrics RA cohort. The RA patients were older than the 1 million sample of patients without RA, and a higher percentage were female (Table 1). The RA cohort was followed for a mean of 16.5 months, whereas the non-RA cohort had a mean of 21.7 months of followup. Methotrexate was the most frequently used traditional DMARD (24.4%), and etanercept was the most frequently used biologic DMARD (4.9%). There were 1,611 cases of herpes zoster identified in the RA cohort, resulting in a rate of 9.83 per 1,000 person-years. In the non-RA cohort, we found 6,581 cases of herpes zoster, resulting in a rate of 3.71 per 1,000 person-years. Cox proportional hazards models suggested that RA patients had a higher risk of herpes zoster compared with non-RA patients. Controlling for age, sex, and calendar year, the hazard ratio (HR) between RA and non-RA patients was 1.94 (95% confidence interval [95% CI] 1.83–2.05) for herpes zoster; after further control for selected comorbid conditions, the adjusted HR was essentially the same (HR 1.91, 95% CI 1.80–2.03).

Table 1. Patient characteristics*

Variable	PharMetrics		GPRD
	RA cohort (n = 122,272)	Non-RA cohort (n = 1,000,000)	RA cohort (n = 38,621)	Non-RA cohort (n = 500,000)
* Values are the percentage unless otherwise indicated. PharMetrics = US PharMetrics claims database; GPRD = UK General Practice Research Database; RA = rheumatoid arthritis; IQR = interquartile range. † Obtained from International Classification of Diseases, Ninth Revision, Clinical Modification codes during the 90 days prior to cohort entry for PharMetrics and from medical history for GPRD.
Female sex	73.1	55.5	71.7	52.1
Age at cohort entry, years
18–44	30.4	60.7	18.7	59.4
45–64	59.1	36.4	39.5	24.1
≥65	10.5	2.9	41.8	16.5
Followup, months
Mean ± SD	16.5 ± 12.2	21.7 ± 17.8	52.0 ± 35.5	54.7 ± 37.5
Median (IQR)	14.1 (6.2–24.3)	17.0 (7.0–33.1)	47.7 (20.9–76.7)	50.0 (20.5–83.3)
Comorbid conditions†
Cancer	2.6	1.1	3.0	1.5
Cardiovascular disease	19.2	7.6	21.4	9.3
Diabetes	6.8	2.8	2.7	1.5
Chronic lung disease	5.5	2.2	8.4	5.3

The nested case–control analysis included all 1,611 cases of herpes zoster and 11,277 controls. The cases were more likely to be older, diabetic, have chronic lung disease, have cancer, be receiving an NSAID or COX-2 selective inhibitor, and have previously seen a health care provider more often during followup (Table 2). Compared with the reference group of nonuse of DMARDs or oral corticosteroids on the index date, the adjusted ORs suggested that exposure to biologic DMARDs alone on the index date was associated with herpes zoster (OR 1.54, 95% CI 1.04–2.29) (Table 3). Using biologic DMARDs with traditional DMARDs increased the risk, but it was not statistically significant (OR 1.38; 95% CI 0.83–2.27). Traditional DMARD use alone was associated with herpes zoster (OR 1.37, 95% CI 1.18–1.59) when compared with nonuse of DMARDs or oral corticosteroids. Oral corticosteroid use was associated with herpes zoster when used alone, combined with biologic DMARDs, and/or combined with traditional DMARDs. Results were not substantially affected when the duration of medication exposure was increased by 20% beyond the length of the prescription or when only patients with at least 6 months of data before cohort entry were analyzed.

Table 2. Cases and controls of herpes zoster from the US PharMetrics claims database*

Characteristic	Cases (n = 1,611)	Controls (n = 11,277)
* Values are the number (percentage) unless otherwise indicated. NSAID = nonsteroidal antiinflammatory drug; COX-2 = cyclooxygenase 2. † As determined by a single International Classification of Diseases, Ninth Revision, Clinical Modification code between cohort entry and the index date.
Female sex	1,215 (75.4)	8,243 (73.1)
Age at index date, mean ± SD years	55.4 ± 12.2	51.1 ± 12.4
Followup, mean ± SD months	12.3 ± 10.3	12.3 ± 10.3
Comorbid conditions†
Diabetes	219 (13.6)	1,049 (9.3)
Chronic lung disease	252 (15.6)	1,114 (9.9)
Cancer	210 (13.0)	849 (7.5)
Orthopedic procedure	17 (1.1)	70 (0.6)
Seen by a rheumatologist	853 (53.0)	5,424 (48.1)
Current NSAID/COX-2 selective inhibitor use	417 (25.9)	2,672 (23.7)
No. of visits between cohort entry and index date
<10	647 (40.2)	5,448 (48.3)
11–30	503 (31.2)	3,667 (32.5)
31–50	238 (14.8)	1,275 (11.3)
≥51	223 (13.8)	887 (7.9)

Table 3. Association between medication use and herpes zoster among patients with rheumatoid arthritis in the US PharMetrics claims database*

Medication†	Cases (n = 1,611)	Controls (n = 11,277)	Unadjusted OR (95% CI)	Adjusted OR (95% CI)‡
* Values are the number (percentage) unless otherwise indicated. OR = odds ratio; 95% CI = 95% confidence interval; DMARDs = disease-modifying antirheumatic drugs. † The categories related to current DMARD/oral corticosteroid exposure are mutually exclusive. ‡ Adjusted for age, sex, current cyclooxygenase 2 inhibitor and nonsteroidal antiinflammatory drug use, comorbidities (diabetes, chronic lung disease, and cancer), orthopedic procedures, number of visits to a health care provider between cohort entry and the index date, and whether or not the patient saw a rheumatologist during followup.
No DMARDs or oral corticosteroids	877 (54.4)	7,690 (68.2)	Reference	Reference
Biologic DMARDs only	32 (2.0)	185 (1.6)	1.52 (1.03–2.23)	1.54 (1.04–2.29)
Traditional DMARDs only	306 (19.0)	2,018 (17.9)	1.34 (1.17–1.54)	1.37 (1.18–1.59)
Oral corticosteroids only	166 (10.3)	503 (4.5)	2.95 (2.44–3.58)	2.51 (2.05–3.06)
Traditional DMARDs and biologic DMARDs	19 (1.2)	129 (1.1)	1.31 (0.81–2.14)	1.38 (0.83–2.27)
Biologic DMARDs and oral corticosteroids	12 (0.7)	43 (0.4)	2.49 (1.31–4.76)	2.44 (1.26–4.73)
Traditional DMARDs and oral corticosteroids	188 (11.7)	660 (5.9)	2.51 (2.11–3.00)	2.39 (1.99–2.88)
Traditional DMARDs, biologic DMARDs, and oral corticosteroids	11 (0.7)	49 (0.4)	1.96 (1.02–3.80)	1.96 (0.99–3.86)

UK General Practice Research Database.

The results from the GPRD were similar to those from PharMetrics. There were 38,621 patients included in the GPRD RA cohort, and a total of 500,000 controls were randomly selected. Members of the RA cohort were more likely to be female, older, and have comorbidities (Table 1). Sulfasalazine was the most frequently used DMARD (22.2%). A total of 1,719 RA patients and 9,209 non-RA patients were diagnosed with herpes zoster during followup, resulting in rates of 10.60 and 4.10 per 1,000 person-years for the RA and non-RA groups, respectively. Controlling for age, sex, and year of cohort entry, the HR between RA and non-RA patients was 1.70 (95% CI 1.61–1.79); after further adjustment for comorbid conditions recorded in the medical history, the HR was 1.65 (95% CI 1.57–1.75).

All 1,719 cases of herpes zoster identified within the RA cohort and 12,033 controls were included in the nested case–control analysis. As in the PharMetrics analysis, cases were more likely to be older, have their comorbidities examined, be receiving an NSAID or COX-2 selective inhibitor, and have seen a rheumatologist (Table 4). Cases also tended to have had more frequent visits to a health care provider prior to the index date. The strength of association was significantly higher for patients receiving DMARDs or glucocorticoids compared with those not receiving these medications (Table 5). The adjusted OR associated with DMARD use without oral corticosteroids was 1.27 (95% CI 1.10–1.48). The OR was slightly higher for oral corticosteroids alone (OR 1.46, 95% CI 1.24–1.70), and highest for oral corticosteroids and DMARDs in combination (OR 1.82, 95% 1.46–2.26). When we lengthened the exposure window by 20% for oral medications, the results were similar.

Table 4. Cases and controls of herpes zoster from the UK General Practice Research Database*

Characteristic	Cases (n = 1,719)	Controls (n = 12,033)
* Values are the number (percentage) unless otherwise indicated. NSAID = nonsteroidal antiinflammatory drug; COX-2 = cyclooxygenase 2. † Obtained from physician diagnoses and medical history.
Female sex	1,273 (74.1)	8,721 (72.5)
Age at index date, mean ± SD years	63.9 ± 12.4	60.6 ± 15.0
Followup, mean ± SD months	38.8 ± 30.0	38.8 ± 30.0
Comorbid conditions†
Diabetes	73 (4.3)	479 (4.0)
Chronic lung disease	260 (15.1)	1,496 (12.4)
Cancer	123 (7.2)	606 (5.0)
Orthopedic procedure†	246 (14.3)	1,352 (11.2)
Seen by a rheumatologist	557 (32.4)	3,412 (28.4)
Current NSAID/COX-2 selective inhibitor use	745 (43.3)	4,410 (36.7)
Number of visits between cohort entry and index date
<10	524 (30.5)	4,476 (37.2)
11–30	548 (31.9)	3,805 (31.6)
31–50	267 (15.5)	1,717 (14.3)
≥51	380 (22.1)	2,035 (16.9)

Table 5. Association between medication use and herpes zoster among patients with rheumatoid arthritis in the UK General Practice Research Database*

Medication†	Cases (n = 1,719)	Controls (n = 12,033)	Unadjusted OR (95% CI)	Adjusted OR (95% CI)‡
* Values are the number (percentage) unless otherwise indicated. OR = odds ratio; 95% CI = 95% confidence interval; DMARDs = disease-modifying antirheumatic drugs. † The categories related to current DMARD/oral corticosteroid exposure are mutually exclusive. ‡ Adjusted for age, sex, current cyclooxygenase 2 inhibitor and nonsteroidal antiinflammatory drug use, comorbidities (diabetes, chronic lung disease, and cancer), history of an orthopedic procedure, number of visits to a health care provider between cohort entry and the index date, and whether or not the patient saw a rheumatologist during followup.
No DMARDs or oral corticosteroids	1,084 (63.1)	8,869 (73.7)	Reference	Reference
Traditional DMARDs only	273 (15.9)	1,633 (13.6)	1.36 (1.81–1.57)	1.27 (1.10–1.48)
Oral corticosteroids only	240 (14.0)	1,061 (8.8)	1.84 (1.58–2.15)	1.46 (1.24–1.70)
Traditional DMARDs and oral corticosteroids	122 (7.1)	470 (3.9)	2.13 (1.73–2.63)	1.82 (1.46–2.26)

DISCUSSION

In these 2 large population-based studies, we found that patients with RA had a nearly 2-fold increase in the risk of herpes zoster compared with individuals without RA. Among patients with RA, those treated with DMARDs and/or oral corticosteroids appeared to be at higher risk.

Several large studies have demonstrated that patients with RA have an increased risk of infection. In an inception cohort of 609 patients with RA compared with population-based patients without RA, Doran et al found an increased risk of objectively confirmed infections (HR 1.7, 95% CI 1.4–2.0), infections requiring hospitalization (HR 1.8, 95% CI 1.5–2.2), and any documented infection (HR 1.5, 95% CI 1.3–1.6) (35). Multiple studies have demonstrated significantly elevated infection-related mortality among patients with RA (36–38). In addition, some of the immunosuppressive medications that are used to treat RA, particularly azathioprine and cyclophosphamide, have been found to be associated with increased infection (39, 40). Recent data on the risk of infection related to anti–tumor necrosis factor (anti-TNF) therapy have suggested that these medications may be associated with an increased risk of serious infection (41, 42). In their meta-analysis of 9 clinical trials of infliximab or adalimumab, Bongartz et al demonstrated an increased risk of serious infections in the treatment arms compared with placebo (OR 2.0, 95% CI 1.3–3.1) (42). However, recent data from the British Society for Rheumatology Biologics Register showed no increased risk of overall serious infection for patients treated with anti-TNF therapies compared with conventional DMARDs (43), and data from the National Databank for the Rheumatic Diseases demonstrated no increased risk of pneumonia associated with anti-TNF therapy (44).

In terms of herpes zoster, declining virus-specific cell-mediated immune responses, which occur naturally as a result of aging or are induced by immunosuppressive illness or medical treatments, have been found to increase the risk of herpes zoster (4, 5). Specifically, patients with malignancy, those receiving immunosuppressive medications (including corticosteroids), organ-transplant recipients, and persons who are seropositive for human immunodeficiency virus have been shown to have an increased risk of developing herpes zoster (5). A number of cases of herpes zoster in patients with RA receiving specific medications have been described (15–23). In one small study, Antonelli et al reported an apparent increase in the incidence of herpes zoster in patients with longstanding severe RA treated with low-dose weekly methotrexate compared with the general population (12). Wolfe et al recently found that the incidence of herpes zoster was increased in RA compared with population-based rates (24). In their study, after adjustment for severity, various treatments including azathioprine, prednisone, leflunomide, and cyclophosphamide were risk factors for herpes zoster; methotrexate and biologic agents were not associated with increased risk. In the study by Wolfe et al, cases of herpes zoster were identified by self-report, and they did not formally quantify the increased risk of herpes zoster associated with RA or examine the effect of strictly current medication use.

It is important to recognize the strengths of the study. We used data sources from 2 countries with potentially different practice styles, and the data were collected through different methods. Our results were consistent across the data sources, although information on biologic DMARD use was only available in the PharMetrics database. Both data sources are extremely large, representative samples of individuals with RA. Although the PharMetrics database includes only patients in managed care plans, the GPRD contains a sample of nearly all types of RA patients in the UK and includes a substantial number of patients older than 65 years of age. In both data sources, all information on exposures, outcomes, and potential confounders was recorded prior to the start of the study, eliminating the possibility of biases that result from analyzing retrospectively collected data.

A number of factors may have contributed to our apparent findings. It is possible that the observed increased risk of herpes zoster associated with RA and with certain RA medications in this study was due to detection bias. Although patients with RA or those receiving specific medications may be monitored closely, we expect that the symptoms associated with herpes zoster are severe enough that any patient would seek medical attention, and therefore the case would be documented in the computer as a claim or as a diagnosis by a GP. Herpes zoster usually has distinctive signs and symptoms such that it is not likely to be confused with another diagnosis.

A possible explanation for our finding that the risk of herpes zoster is increased with the use of DMARDs and/or corticosteroids is that the use of these drugs is a marker for increased disease severity, and it is actually disease severity that increases the risk. Although the GPRD and PharMetrics data sources have a great deal of information on each patient, neither has detailed clinical information about the severity of RA. Because the study included all patients with a diagnosis of RA in the databases, there were patients with different degrees of RA severity. Our analysis indicated that patients with RA were at a significantly increased risk for herpes zoster; therefore, it may be that disease severity increases risk. Although we attempted to control for some markers of disease severity using multivariate regression (e.g., orthopedic procedures, care by a rheumatologist) without the specific clinical variables indicative of RA severity (i.e., swollen joint count, serum rheumatoid factor level, etc.), it is possible that there was some residual confounding in our analysis.

In neither data source were data specifically collected for this study; therefore, we had no control over the variables collected or the accuracy of the data collection. There could have been misclassification of any of the variables collected in either data source. It is possible that patients could have been classified as having RA when they did not have the disease, which would have biased our estimates toward the null; this could have occurred as a random error in coding or as a nonrandom error if RA were used as a rule-out diagnosis. In both data sources, drug exposure was inferred from automated prescription data. It was not possible to determine whether patients actually consumed the prescribed medication and whether or not they adhered to the dosing instructions. In the GPRD, it is possible that all of the medications were not captured by the GP in the database; however, we anticipate that the majority of medications were included.

In summary, using data from PharMetrics and the GPRD, we found that patients with RA were at increased risk of herpes zoster. Among RA patients, we found that current use of DMARDs and/or oral corticosteroids appeared to be associated with an increased likelihood of herpes zoster compared with no exposure. Additional studies should be conducted to determine whether it is actually disease severity or exposure to these RA medications that increases risk among patients with RA. In addition, studies should be performed to examine the prognosis and risk of complications, including dissemination, associated with herpes zoster among RA patients compared with non-RA patients and also across medication groups. Given that many of the drugs used to treat RA have been associated with substantial benefit in terms of disease improvement and quality of life, any increase in the risk of herpes zoster must be considered in the context of the benefits expected from the medications.

AUTHOR CONTRIBUTIONS

Dr. Smitten had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study design. Smitten, Choi, Hochberg, Suissa, Simon, Testa, Chan.

Acquisition of data. Smitten, Simon.

Analysis and interpretation of data. Smitten, Choi, Hochberg, Suissa, Simon, Testa, Chan.

Manuscript preparation. Smitten, Choi, Hochberg, Suissa, Testa, Chan.

Statistical analysis. Smitten, Testa.

ROLE OF THE STUDY SPONSOR

This study was supported by the Harvard Pharmacoepidemiology Program and the Bristol-Myers Squibb Company. Bristol-Myers Squibb Company provided the data for the study without restriction. The Bristol-Myers Squibb author (TAS) assisted in the study design and interpretation of data. The Bristol-Myers Squibb author, in conjunction with all other authors, critically reviewed the manuscript and agreed to submit it for publication.