Article Text

Download PDFPDF

Extended report
Psoriatic arthritis treatment and the risk of herpes zoster
  1. D Zisman1,2,
  2. H Bitterman2,3,
  3. G Shalom3,
  4. I Feldhamer3,
  5. D Comanesther3,
  6. E Batat3,
  7. S Greenberg-Dotan3,
  8. S Cohen1,
  9. A D Cohen3,4
  1. 1Department of Rheumatology, Carmel Medical Center, Haifa, Israel
  2. 2The Ruth and Bruce Rappaport Faculty of Medicine, Technion, Haifa, Israel
  3. 3Chief Physician's Office Central Headquarters, Clalit Health Services, Tel Aviv, Israel
  4. 4Faculty of Health Sciences, Siaal Research Center for Family Medicine and Primary Care, Ben-Gurion University of the Negev, Beer-Sheva, Israel
  1. Correspondence to Dr Devy Zisman, Department of Rheumatology, Carmel Medical Center, 7 Michal Street, Haifa 34362, Israel; devyzisman{at}


Objectives To study the association between traditional disease-modifying antirheumatic drugs (c-DMARD) or anti-TNF-α agents and herpes zoster (HZ) in patients with psoriatic arthritis (PsA).

Methods A retrospective cohort study was conducted in patients with PsA between 2002 and 2013. Patients were grouped as follows: no DMARDs (Group A); c-DMARDs (Group B); anti-TNF-α agents (Group C); anti-TNF-α agents in combination with c-DMARDs (Group D). Crude incidence rates (IR) were calculated as number of HZ episodes per 1000 patient-years. A Cox regression model was used to adjust for HZ risk factors (age, gender, steroid use, Charlson Comorbidity Index score, and previous treatment) in order to estimate their contribution to the risk of the first HZ event.

Results The study included 3128 patients, mean age 50.26±14.54 years; 46.2% male. During a period of 20 096 person-years 182 HZ events were observed. The crude IR (95% CI) of HZ in the study population was 9.06 per 1000 patient-years, and in Groups A-D 7.36 (5.41 to 9.79), 9.21 (7.5 to 11.21), 8.64 (4.84 to 14.26), 17.86 (10.91 to 27.58), respectively. In a multivariate analysis, age (HR 1.01, 95% CI 1.00 to 1.02), treatment with steroids (HR 1.08, 95% CI 1.04 to 1.13), and a combination of anti-TNF-α agents and c-DMARDs (HR 2.37, 95% CI 1.32 to 4.22) were significantly associated with HZ events.

Conclusions In our database, the risk of HZ was significantly increased with age, treatment with steroids, and combination of anti-TNF-α agents and c-DMARDs, but not with c-DMARDs or anti-TNF-α therapy alone. Time to HZ event was shorter in patients treated with anti -TNF-α agents.

  • Anti-TNF
  • Psoriatic Arthritis
  • Treatment

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Herpes zoster (HZ), results from reactivation of endogenous latent varicella zoster virus (VZV) infection within the sensory ganglia. The incidence increases sharply with age from 1–4/1000 in adults aged 40 years to 7–8/1000 above the age of 50 years, and the use of immunosuppressive therapy.1 ,2 The initiation of biological therapies in patients with inflammatory arthritis highlighted the possibility of HZ reactivation. Studies have reported increased risk of opportunistic infections, including tuberculosis and fungal infection, in patients treated with TNF-α blockers, attributed to the primary role of TNF-α in both host defence and the immune response.3–5 TNF-α inhibits VZV replication and antigen expression of human fibroblast in vitro.6 Thus, inhibition of TNF-α by biologic interventions may have a profound effect on the host’s ability to defend itself against VZV infection.

Data regarding VZV reactivation in patients treated with biological drugs is inconsistent. A number of observational studies reported an increased risk of HZ in patients receiving TNF-α blockers for inflammatory diseases, especially rheumatoid arthritis (RA).7–10 Treatment with TNF-α blockers is associated with more aggressive disease, entails multidermatomal involvement, persistent postherpetic neuralgia, and carries a higher rate of hospitalisations.11 ,12 Risk factors identified in studies include treatment with monoclonal anti-TNF-α (as opposed to etanercept), older age, use of prednisone, and comorbidities.1 ,8 ,12–14

By contrast, Dreiher et al14 observed a non-significant increased risk of HZ during treatment with biologic agents in a cohort of psoriasis patients. A US database study15 of patients with RA, inflammatory bowel disease, psoriasis, PsA and ankylosing spondylitis confirm those findings; the authors found a comparable risk of HZ in patients initiating anti-TNF-α therapies and non-biologic treatment regimens.

The aim of the present study is to assess the incidence of HZ in a cohort of psoriatic arthritis (PsA) patients, and its relationship to treatment with: traditional disease-modifying antirheumatic drugs (c-DMARD) versus anti-TNF-α agents. Data collected from surveys and registries of patients with inflammatory arthritis are now accepted worldwide as valid indicators of the long-term safety and efficacy of biologic drugs. These sources are considered to more reliably reflect the common practice of treating patients with complex comorbidities and older age3–5 ,16 than clinical trials.17–19 Hence, we screened a national health-fund database for our data.


Study population

Our cohort study used the database of Clalit Health Services (CHS), Israel’s largest health fund with approximately 4 million members, 52% of Israel's population. The computerised database has continuous real-time input from pharmaceutical, medical and administrative operating systems. It has precise information on each prescription used by the patient, including the date of the prescription, dose and mode of administration. CHS makes available its database of medical information and has been used in numerous studies.14 ,19 ,20 Medication use is coded according to anatomical therapeutic chemical (ATC) classification, and disease codes employ a modification of the International Classification of Diseases Ninth Revision (ICD-9). The treatment of PsA with biological agents was introduced in the Israeli health basket programme in January 2002. We collected data on all patients diagnosed with psoriatic arthritis, psoriatic arthropathy, psoriasis with arthritis, or arthritis psoriatic diagnoses consistent with ICD-9 code 696.0, made by a rheumatologist; or one of these diagnoses listed on their hospitalisation discharge summary. The list of patients diagnosed with psoriatic arthritis was checked for inconsistent data, using both logical algorithm as well as individualised review of medical files when needed. Thereafter, patients with unconfirmed PsA diagnosis were excluded.

The study was conducted from 1 January 2002 to 31 December 2012 with extended follow-up till the end of December2013.

Drug exposure

Data was collected on medication use, with emphasis on drug prescriptions for rheumatic disease: c-DMARDs, anti-TNF-α agents and steroids. Treatment periods were classified into the following groups according to the DMARDs prescribed:

  • Group A: not receiving c-DMARDs or anti-TNF-α drugs.

  • Group B: treated by c-DMARDs: salazopyrine, methotrexate (MTX), leflunomide, cyclosporine A, azathioprine, hydroxychloroquine, or a combination of these drugs.

  • Group C: anti-TNF-α agents including infliximab, adalimumab or etanercept.

  • Group D: anti-TNF-α agents and c-DMARDs prescribed in combination.

  • A subset of the cohort treated with anti-TNF-α (groups C and D) was also analysed according to the following categories:

    1. Group E: monoclonal anti-TNF-α agents (mab-anti TNFα): infliximab or adalimumab;

    2. Group F: monoclonal anti-TNF-α agents and c-DMARDs;

    3. Group G: etanercept;

    4. Group H: etanercept and c-DMARDs.

In addition to these drug groups, the use of MTX was recorded separately. During the study period, patients contributed exposure time to different treatment groups according to the medications they received at each exposure period. In order to be considered as a treatment-related period, the patients had to be treated with the same agent for at least six consecutive months. The follow-up period was calculated from the initiation of a new therapeutic agent until 30 days after the last administered dose, the end of the follow-up (31 December 2013), death or drop-out due to a change in healthcare provider, whichever came first. Calculation of total follow-up period was performed for each therapeutic group. Steroid exposure was defined as the annual average number of prescriptions of oral steroid-containing preparations that were purchased by the patients at the pharmacy from the start of a treatment regimen with a specific agent until the HZ event, end of the study or being lost to follow-up.

For each study period, the following data were recorded at the beginning of the follow-up: demographic data, such as sex and age, comorbidities calculated by the Charlson Comorbidity Index,21 treatment with c-DMARDs prior to January 2002, steroid exposure, drug administrated including MTX use, and HZ events. HZ was defined by the ICD-9 codes 053.0–0.53.9 in conjunction with the use of acyclovir prescribed as intravenous or oral for a period length of minimum 5 days. Occurrence of recurrent HZ events was allowed only if there was a difference of 180 days between data input suggesting an event. In patients with recurrent HZ, only the first HZ event was included in the regression analyses. HZ was attributed to c-DMARDs or biologic drugs if the patient was receiving these medications at the time of, or within 1 month prior to, the event. Patients on a specific drug group who switched treatment, contributed follow-up time to that group up to a month after its discontinuation; after this month, the event was attributed to the subsequent drug group, with no overlap between the periods.

Statistical analysis

For univariate analysis, χ2 tests were used for comparison of the distribution of categorical parameters, while means of continuous variables among the different groups were compared by t tests for normally distributed parameters or by Mann–Whitney tests in parameters not normally distributed, as appropriate. Crude incidence rates were calculated as the number of HZ infections per 1000 patient-years of follow-up in each group. Time to the first HZ event was calculated in treatments periods where HZ occurred. Mann–Whitney test was used to compare the time to the first HZ event between the non-treatment group and the different treatment regiments (Group B-H). Cox regression models were used to adjust for HZ risk factors (age, gender, steroid use, Charlson Comorbidity Index score, and previous treatment with c-DMARDs) in order to estimate the contribution of treatment to the risk of HZ. A multivariate analysis was used in two models. The first regression model included groups A–D. In the second regression model, the groups were divided according to the different anti-TNF-α agents: etanercept or mab-anti-TNF-α agents.

Statistical analysis was performed with SPSS for Windows software, V.18.0 (Chicago, Illinois, USA). All statistical tests were 2-sided, and a p value <0.05, or p value with Bonfferoni correction in multiple comparisons were considered significant. Data use was approved by the local ethics committee.


A total of 3131 patients who met the inclusion criteria were identified in the database; three who were treated by ustekinumab were excluded from the study. The mean age of the entire study population was 50.26±14.54 years, consisted of 1446 males (46.2% of the total). During the 12 years of follow-up, 1066 (31.78%) patients contributed 6386.54 years to the control group (A), 2156 patients (54.01%) contributed 10 854.71 patients’ years to the c-DMARDs group (B), 587 patients contributed 1735.5 patients’ years (8.64%) to the anti-TNF-α agent group (C) and 427 patients contributed 1119.8 patients year (5.57%) to the combination of biological and non-biological DMARDs group (D) as summarised in table 1.


Patients’ baseline characteristics and descriptive statistics for the study population and for each treatment group

There were no statistically significant differences among the different groups regarding sex distribution, use of c-DMARDs before 2002, and the use of MTX. Patients treated with c-DMARDs were older and had a lower Charlson Comorbidity Score Index.

A total of 182 events of HZ were recorded in 20 096 patient-years, with an incidence rate of 9.06/1000 patient-years (95% CI 7.79 to 10.47). None of the patients received vaccination with VZ. There were no significant differences in the incidence rate between untreated patients (7.36/1000 patient-years 95% CI 5.41 to 9.79) and those treated with c-DMARDs (9.21/1000 patient-years 95% CI 7.5 to 11.21) and anti-TNF-α agents (8.64/1000 patient-years). Combination therapy of anti-TNF-α and c-DMARDs (Group D) revealed higher incidence rate (17.86/1000 patient-years, 95% CI 10.91 to 27.58), compared to the untreated group, which was attributed mainly to the highest rate in the group of patients treated by etanercept and c-DMARDs (27.80/1000 patient-years, (95% CI 12.71 to 52.78) (table 2). The time to HZ event was shorter in patients treated with biologic agents: (15.85±12.08 months in the combined anti-TNF-α agent and c-DMARDs patients, (Group D) and 14.00±9.76 months in patients treated by anti-TNF-α agents alone (Group C), as compared with 46.09±34.85 months in patients treated by c-DMARDs (Group B) and 55.64±34.97 months in PsA patients not treated by DMARDs.

Table 2

Crude incidence rates of herpes zoster (HZ) events/1000 treatment-years

In a multivariate Cox regression analysis (table 3, Model 1) after adjustment for sex, age, Charlson Comorbidity Index score, steroid treatment and previous non-biologic treatment, a statistically significant risk for HZ was observed in the combined treatment group of anti-TNF-α agents and conventional DMARDs (Group D) (HR 2.37, 95% CI 1.32 to 4.22, p=0.004). Other factors associated with HZ infection were age (HR 1.01, 95% CI 1.00 to 1.02, p<0.05) and steroid treatment (HR1.08, 95% CI 1.04 to 1.13, p<0.001). Sex, previous treatment with c-DMARDs, and Charlson Comorbidity Index score did not affect HZ rates (table 3 Model 1). The effects of age and steroids treatment in the second model were similar to the first model, and among the different anti-TNF-α combination regiments only the use of etanercept and c-DMARDs (Group H) was significant (HR 3.60, 95% CI 1.67 to 7.75) associated with HZ event (table 3, Model 2).

Table 3

Time to herpes zoster event according to the Cox proportional hazard model


Our study includes 3128 PsA patients with a total of 20 096 patient years. This large sample size enabled us to evaluate the impact of various treatments in patients with PsA in real-life conditions. Increased risk of HZ was found in patients treated with a combination of anti-TNF-α agents and c-DMARDs, but not in those treated with anti-TNF-α agents and c-DMARDs separately.

The risk of HZ in our PsA cohort (9.21/1000 patient-years) is higher than the expected risk in the general population, which is around 7–8/1000 after age 50 years.1 Dreiher et al14 found a lower risk of HZ (5.7/1000 patient-years) in patients with psoriasis in the CHS population. Possible explanations for this difference are the complex treatment regimens for PsA that include combinations of biologic agents and c-DMARDs, and higher expected comorbidities rates in a PsA cohort compared with a psoriasis cohort.

The overall risk of HZ in patients with PsA in our cohort is comparable to the risk in patients with RA of 9.96 cases/1000 patient-years reported in a retrospective cohort study from the US Veterans Administration health database.7

Several studies reported conflicting results on the risk of HZ among patients with RA treated with biologic drugs. A significant increase of HZ in patients treated with anti-TNF-α compared with conventional DMARDs was reported in six studies, including a meta-analysis,9 ,12 ,22–25 while six other studies found no significant association.13 ,15 ,16 ,26–28

Comparing those results with our findings is not straightforward. Treatment regimens for RA and PsA differ: for example, combination therapy of c-DMARDs and anti-TNF-α agents is more common in RA. Additionally, the gender distribution was almost equal between men and women in our study while most RA studies have a predominance of females, who have higher risk of developing HZ.

Our data is consistent with that of Winthrop et al15 whose study based on a large database in the USA also found that patients with PsA and other inflammatory diseases given anti-TNF-α therapies were not at higher risk of HZ compared with patients given non-biologic treatment regimens. The findings were similar despite the different study methodologies. We recorded patients over time and differentiated between those treated with a combination of c-DMARDs and biologic agents and those treated with biological c-DMARDs alone; Winthrop’s group15 investigated only new users and did not differentiate between those treated with anti-TNF-α alone or in combination with c-DMARDs.

The data regarding the risk of HZ according to the anti-TNF-α different type is contradictory. In the US database from the Veterans Administration13 and RABBIT12 registries, infliximab seemed to pose greater risk of reactivation than etanercept and adalimumab. The BSRBR study9 documented significantly higher risk of HZ with infliximab compared with adalimumab only. On the other hand, Winthrop et al15 found similar risk of HZ in different anti-TNF-α compounds. We observed no significant difference in HZ events between PsA patients treated with etanercept and monoclonal antibodies when these agents were used separately, but the risk was increased when they were combined with c-DMARDs. Confounded by indication, in which patients with more complex background diseases (reflected in the higher Charlson Comorbidity Index score) were treated with etanercept than with monoclonal antibodies is a possible explanation for the increased risk of HZ with etanercept combined with c-DMARDs.

Our findings confirm previous studies showing that older age1 ,29 and treatment with corticosteroids22 ,27 are risk factors for HZ. By contrast with Strangfeld et al,12 we did not find an increased risk in patients with comorbidity: malignancy, chronic lung disease, renal failure and liver disease. Our work did not address specific comorbidities, taking, rather, a more general approach using the Charlson Comorbidity Index; this validated score for predicting mortality by classifying or weighting comorbidities has been widely used by health researchers to measure burden of disease and case mix.

Our finding that the mean time to HZ event after beginning anti-TNF-α treatment between 8 and 19 months is in line with the results of other studies showing a peak in the incidence rate in the first 2 years after initiating biological therapies.9 ,30 A number of factors can account for this heightened risk at that time,9 including a ‘healthy user’ effect, and time-varying confounders such as steroid exposure.9

Large-scale studies are needed to clarify some still-debated issues around patients’ selection for treatment with biologic drugs and prophylactic vaccination of immunosuppressed patients. Indeed, there are vaccines available today against chicken pox and HZ that avert the complications of reactivation and postherpetic neuralgia.31 Until these issues are resolved, patients prescribed biologic drugs should be informed about the early symptoms of HZ in order to sooner, and better, combat the severity of symptoms and perhaps prevent such complications as postherpetic neuralgia.29

Study limitations include the retrospective collection of data with no recorded disease activity parameters. Thus, we were not able to directly assess the effects of disease severity and confounding by indication on the risk for HZ. The complex calculation of the different steroid regimens and preparations was simplified, as the number of prescriptions drawn from pharmacy per year. We did not investigate the severity of HZ reactivation. The identification of HZ reactivation required diagnosis and treatment with antiviral agents; therefore, milder events could have been missed. These limitations affected all study groups.

The strengths of this study are the large sample size and the data collection reflecting real life, covering the first anti-TNF treatment as well as the other treatment regimens over time. The database encompassing 12 years allowed us to compare untreated patients with patients treated by different strategies: c-DMARD and anti-TNF-α agents alone or in combination.


In conclusion, we found that the risk of HZ in patients with PsA is increased with age, treatment with steroids, and combination of anti-TNFα agents and c-DMARDs, but not with c-DMARDs or anti-TNF-α therapy administrated separately. In our database, the time to HZ event was shorter in patients treated with anti TNF-α agents. Further studies will enable more precise and individually tailored treatment.



  • Handling editor Tore K Kvien

  • Contributors DZ: conceived and participated in the design of the study, analysed and interpreted the results and drafted the manuscript. HB: conceived and participated in the design, analysed and interpreted the results and contributed substantial editing of the manuscript. GS, SG: conceived and participated in the design of the study, analysis and interpretation of results and editing of the manuscript. IF, ADC: conceived and participated in the design, collected data, analysed and interpreted results, and edited the manuscript. DC: conceived and participated in the design and collection of data. DB: conceived and participated in the design of the study, collection of data, and analysis and interpretation of results. SC: conceived and participated in the design of the study and editing of the manuscript. All authors read and approved the final manuscript.

  • Competing interests None.

  • Ethics approval Clalit Health services ethics committee.

  • Provenance and peer review Not commissioned; externally peer reviewed.