Rheumatoid arthritis, its treatments, and the risk of tuberculosis in Quebec, Canada

Authors


Abstract

Objective

To determine the risk of tuberculosis (TB) among a cohort of patients with rheumatoid arthritis (RA) in Quebec and assess whether this risk is associated with exposure to nonbiologic disease-modifying antirheumatic drugs (DMARDs).

Methods

We studied a cohort of patients with RA identified from the Quebec provincial physician billing and hospitalization databases for 1980–2003. TB incidence rates were determined for the period 1992–2003 and compared with the general population, standardized for age and sex using the standardized incidence ratio (SIR). Conditional logistic regression was used in a nested case–control analysis to estimate the rate ratio (RR) of TB related to nonbiologic DMARD exposure during the year before the index date.

Results

Of the 24,282 patients with RA in the cohort, 50 cases of TB were identified. The standardized incidence rate was 45.8 cases per 100,000 person-years compared with 4.2 cases per 100,000 person-years in the general population of Quebec (SIR 10.9, 95% confidence interval [95% CI] 7.9–15.0). The adjusted RR of TB was 2.4 (95% CI 1.1–5.4) with corticosteroid use and 3.0 (95% CI 1.6–5.8) with nonbiologic DMARD use.

Conclusion

The age- and sex-standardized incidence rate of TB in RA patients is 10 times that of the general population. At least some of this risk may be related to nonbiologic DMARD and corticosteroid therapies. Our data support the role of TB screening before initiation of any immunosuppressive therapy.

INTRODUCTION

Recently, there has been increasing attention given to the risk of Mycobacterium tuberculosis (TB) in patients with rheumatoid arthritis (RA) (1–4). It is becoming evident that immunosuppressants used to treat this disease, notably disease-modifying antirheumatic drugs (DMARDs) and corticosteroid agents, may modulate the observed increase in the risk of TB (5–7). However, data from countries with a low burden of TB are scarce (2, 8). In this study, we quantified the incidence rate of TB in an RA cohort and compared that rate with general population rates. We also assessed the extent of the risk of TB in patients with RA after exposure to nonbiologic DMARDs.

SUBJECTS AND METHODS

Data source.

The Canadian province of Quebec, with a population of >7 million people, has a health insurance plan for all its residents with all data computerized (9). Our study used the computerized provincial administrative databases that capture all physician visits, procedures, and hospitalizations, as well as patient demographics. The International Classification of Diseases, Ninth Revision (ICD-9) (10) was used to assign billing and hospitalization diagnosis codes (1 per physician visit and up to 15 per hospitalization). Study subjects were limited to those registered in the provincial pharmacy claims database, which covers residents age ≥65 years, those who receive social assistance, and those who do not have private drug insurance through their employer. This covers ∼50% of Quebec residents (including all those age ≥65 years). Access to the denominalized data was granted by the provincial Commission d'accès á l'information du Québec, and ethical approval was obtained from the McGill University Institutional Review Board. For Quebec, clinically confirmed TB incidence rates by age groups and sex were available for the period 1992–2003 from the Fichier Central des MADO, which is a central file of all reportable diseases in the province (11).

Case definitions.

For the period between January 1, 1980 and December 31, 2003, a cohort was formed of all subjects with ≥1 occurrence of a diagnosis of RA using the physician billing code for RA (ICD-9 code 714) during an inpatient or outpatient visit. Cohort subjects were further restricted to those who were dispensed ≥1 prescription for DMARD therapy. Cohort entry was defined by the date of the first DMARD prescription after 1 year of eligibility in the provincial health insurance system. They were observed until the earliest date of either the date of termination of enrollment in the health plan, the date of death, the end of the study period, or the date of the outcome of interest. TB cases were identified during followup if they had TB as a primary hospitalization diagnosis (ICD-9 codes 010–018) or if TB was part of a medical claim and they had at least 2 classes of the usual first-line anti-TB medications prescribed (isoniazid, rifampin, pyrazinamide, or ethambutol), and if treatment lasted at least 6 months. TB cases that occurred before the date of cohort entry were excluded from the cohort.

Nested case–control analysis.

To study the effects of the medications, we used a nested case–control analysis within the RA cohort, an efficient approach to address the complex patterns of medication exposure over time without significant loss of power (12). For each case of TB, we randomly selected 30 control subjects from the cohort after matching them on the date of cohort entry and ensuring that control subjects were at risk on the day that the case occurred. This date was designated as the index date.

Exposure measurement.

All medications received by the patients during the observation period were identified from dispensed prescription data. The type, date of filling, and quantity of each prescription were obtained from the database. For the purpose of comparisons, we considered exposure to any nonbiologic DMARD as a prescription of methotrexate, hydroxychloroquine, chloroquine, sulfasalazine, azathioprine, leflunomide, cyclophosphamide, cyclosporine, gold compounds, minocycline, or penicillamine in the year prior to the index date. Patients were considered currently exposed to DMARDs if they had filled a prescription for any DMARDs up to 45 days prior to the index date. New DMARD users were RA patients whose only prescriptions for DMARDs in the previous year occurred within 90 days prior to the index date. Chronic DMARD users were those RA patients who filled prescriptions throughout the year prior to the index date. The reference group had no DMARD use in the year prior to the index date. Since biologic response modifiers such as anti–tumor necrosis factor agents were only introduced to the provincial formulary in 2002, they were not part of our analyses because they were absent from the cohort.

Covariate information.

Age and sex were used as covariates to define the study sample. Our multivariate analyses controlled for comorbid clinical conditions known to increase the risk of TB that might also influence DMARD therapy choice (13), including diabetes (ICD-9 codes 250.0–250.9), chronic renal failure and/or hemodialysis (ICD-9 codes 585 and V56.0), solid organ transplantation such as renal or heart transplantation (ICD-9 codes V42.0 and V42.1), carcinoma of the head or neck (ICD-9 codes 173.4 and 195.0), and silicosis (ICD-9 code 502). These were based on diagnoses made at any time prior to the index date. We used the mean number of physician visits as a correlate of disease severity. Other medications commonly used as concomitant treatment for RA such as nonsteroidal antiinflammatory drugs (NSAIDs), corticosteroids, and cyclooxygenase 2 (COX-2) inhibitors were also included as covariates. For corticosteroids and COX-2 inhibitors, current exposure was defined by prescriptions dispensed in the 30-day period prior to the index date.

Statistical analysis.

For the comparison of the incidence rates of TB associated with RA, the total person-years of followup in the entire cohort were used to estimate TB incidence rates in RA, which were compared with the Quebec general population TB incidence rate. TB incidence rates were calculated in the general population of Quebec and in the RA cohort for the period 1992–2003. To adjust for the potential confounding effects of sex and age, rates were standardized using the direct method and the midyear 1996 Quebec population as the standard in order to estimate the standardized incidence ratio (SIR). The population data were made available through Statistics Canada (14).

To assess the rate ratio (RR) of TB associated with DMARD use, odds ratios and 95% confidence intervals (95% CIs) were estimated from the nested case–control analysis within the RA cohort. Conditional logistic regression was used to adjust these estimates for the concurrent use of corticosteroids, NSAIDs, and COX-2 inhibitors, as well as for age, sex, mean number of visits per year, and all comorbid clinical conditions.

In order to assess duration of exposure, we further stratified the risk among current users according to the number of nonbiologic DMARD prescriptions obtained in the year prior to the index date. Analyses were performed using SAS statistical software (15).

RESULTS

The cohort consisted of 24,282 patients with RA; 70.1% were women. The mean ± SD subject age at the time of cohort entry was 61.7 ± 14.6 years. There were 50 cases of TB identified during the followup period. The overall sex- and age-standardized rate of TB in the RA cohort for the period 1992–2003 was 45.8 cases per 100,000 person-years of followup compared with 4.2 cases per 100,000 person-years in the general population of Quebec (SIR 10.9, 95% CI 7.9–15.0).

Table 1 shows the baseline characteristics of the case patients with TB compared with the control subjects. The proportion of men was significantly higher in the TB case group than in the controls. There was a higher proportion of case patients with TB that had been exposed to corticosteroids and nonbiologic DMARDs (specifically, exposure to methotrexate, leflunomide, and cyclosporine) as compared with the control subjects during the 1-year period prior to the index date.

Table 1. Baseline characteristics of RA case patients with TB and control subjects during the year prior to the index date for the period 1980–2003*
 Case patients with TB (n = 50)Control subjects (n = 1,500)P
  • *

    Values are the percentage unless otherwise indicated. RA = rheumatoid arthritis; TB = tuberculosis; DMARDs = disease-modifying antirheumatic drugs; COX-2 = cyclooxygenase 2; NSAIDs = nonsteroidal antiinflammatory drugs.

  • Other comorbid clinical conditions and risk factors are silicosis, chronic renal failure/hemodialysis, solid organ transplantation, and carcinoma.

Age, mean ± SD years65.6 ± 13.167.6 ± 14.30.06
Men50.030.00.002
Diabetes12.010.10.67
Other conditions2.01.90.62
Use of all DMARDs72.045.60.0004
 Methotrexate58.030.3< 0.0001
 Leflunomide6.00.50.003
 Cyclosporine6.00.80.01
 Other20.018.50.78
Current use of corticosteroids18.08.10.03
Current use of COX-2 inhibitors8.05.70.53
Use of NSAIDs56.049.90.39

Table 2 shows that the adjusted RR of TB associated with any nonbiologic DMARD exposure in the year prior to the index date is 3.0 (95% CI 1.6–5.8). Current use of corticosteroids was also significantly associated with the occurrence of TB (adjusted RR 2.4, 95% CI 1.1–5.4). We then restricted the analysis to those age ≥65 years to reflect the risk profile for TB. The overall rate ratios observed in the older cohort for DMARDs and corticosteroid exposure were 3.5 (95% CI 1.5–8.3) and 3.5 (95% CI 1.4–8.7), respectively.

Table 2. Crude and adjusted rate ratios (RRs) of developing TB, according to nonbiologic anti-RA medication use in the previous year*
 Case patients with TB (n = 50)Control subjects (n = 1,500)Crude RRAdjusted RR95% CI
  • *

    95% CI = 95% confidence interval; see Table 1 for additional definitions.

  • Adjusted for age, sex, diabetes, other comorbid conditions, use of DMARDs, current use of corticosteroids, current use of COX-2 inhibitors, and use of NSAIDs.

Any DMARDs366963.23.01.6–5.8
 Methotrexate294543.43.41.8–6.4
 Leflunomide3717.111.72.1–65.1
 Cyclosporine3127.73.80.9–16.6
 Other102771.11.60.7–3.6
Corticosteroids91222.52.41.1–5.4
COX-2 inhibitors4851.51.40.5–4.4
NSAIDs287481.31.20.6–2.3

Long-term DMARD users were significantly at risk of developing TB (adjusted RR 2.8, 95% CI 1.4–5.7), whereas the increased risk among new users failed to reach significance (adjusted RR 2.3, 95% CI 0.5–9.8). We evaluated the number of prescriptions filled for nonbiologic DMARDs among the current users. The RR was significantly elevated for those receiving >6 prescriptions in the year prior to the index date (adjusted RR 3.0, 95% CI 1.2–7.7) and did not vary substantially for those receiving ≥11 prescriptions (Table 3).

Table 3. Crude and adjusted rate ratios (RRs) of developing TB, according to the user type and number of prescriptions among current nonbiologic anti-RA medication users*
 Case patients with TB (n = 50)Control subjects (n = 1,500)Crude RRAdjusted RR95% CI
  • *

    95% CI = 95% confidence interval; see Table 1 for additional definitions.

  • Adjusted for age, sex, diabetes, other comorbid conditions, use of DMARDs, current use of corticosteroids, current use of COX-2 inhibitors, and use of NSAIDs.

New users4872.52.30.5–9.8
Long-term users234922.92.81.4–5.7
Current users265702.82.61.3–5.2
 1–5 prescriptions71852.22.00.7–5.8
 6–10 prescriptions81633.13.01.2–7.7
 ≥11 prescriptions111652.92.81.2–6.5

DISCUSSION

Our study provides new data on TB incidence in an elderly RA population from a country with a low burden of TB (16). The age- and sex-standardized incidence rate of TB in RA patients is 10 times that of the general population. This finding is consistent, although with a stronger magnitude, with previous studies performed in Asia (3, 4) and Europe (1, 2), but inconsistent with a study in the US that found no increased risk (8) using nonstandardized rates for comparison. Unfortunately, we could not capture other factors associated with the risk of TB disease while being exposed to RA drugs, including country of birth, Aboriginal race/ethnicity, socioeconomic status, a history of recent contact with an individual with TB, and the presence of TB-associated abnormal findings on a chest radiograph (13). Information on human immunodeficiency virus infection status or smoking behavior was also lacking (17, 18). If at-risk groups were overrepresented in our cohort, it may explain the higher than previously reported observed risk of TB. The extent to which patients who have RA will demonstrate an excess risk of TB will depend on the prevalence of TB disease, latent TB infection, and potential risk factors in the community. Our findings probably reflect the life experience of this older cohort, which was more likely than the relatively younger group to have been exposed to TB in their youth when TB was highly prevalent in Canada (16).

Men are more at risk of TB in Canada (16), and this risk holds true in our findings regardless of the female predominance in our cohort of RA patients. Our observation is consistent with previous literature that specifically looked at sex as a potential risk factor for TB in patients with RA (3).

Our finding seems to confirm that an increased risk of TB was present before the introduction of biologic response modifiers. We demonstrated that the risk of TB, although increased, was nonsignificant among new users (those with a first prescription in the previous 90 days). Furthermore, to explore duration of exposure, we stratified the risk among current users according to the number of nonbiologic DMARD prescriptions obtained in the year prior to the index date. The risk of TB disease increases with the number of prescriptions filled to reach significance among those with 6 or more prescriptions and among chronic users. To what extent this risk of TB, especially in the older age range, is due to the disease itself versus exposure to immunosuppressants, particularly systemic steroids (19, 20), cannot be determined from our study because disease activity is intimately linked to medication exposures. The study was based entirely on computerized claims databases, which are extremely useful to conduct analyses with large samples of a rare disease (RA) and outcome (TB), but make it impossible to validate the diagnosis of RA or TB. However, for both disease definitions we required an exposure to a specific drug therapy, thus optimizing accuracy. Combining billing codes with DMARD prescription data enhances the validity of the RA diagnosis, as has been demonstrated by other investigators (21, 22). Some reassurance is also provided in that the demographics for our RA cohort (i.e., age and sex) were similar to those in clinical RA populations. Regarding the validity of our TB definition, we obtained a second non-RA comparison group constructed from the same data source comprised of elderly patients with respiratory diseases (23). The SIR for TB among those age ≥60 years was 2.2 (95% CI 1.1–4.4) and comparable to an odds ratio of 2.4 (95% CI 1.4–2.4) reported for a general population of patients with respiratory diseases in the UK (20). Since TB rates are comparable among non-RA populations, we are reassured that our case detection definition did not contribute significantly to the SIR observed. Nonetheless, as in all observational studies, caution must be used in interpretation of the results.

A great strength of our study is that the provincial coverage for TB medications is universally free and recorded in the pharmacy claims database. However, one limitation is that any information on RA medication not covered by the Quebec drug plan would not be included in the data. This is of particular concern, because anti–tumor necrosis factor agents were available prior to their introduction to the Quebec formulary in 2002 through private medication plans or clinical trials. If coadministered with nonbiologic DMARDs (which is often the case), these unmeasured exposures could potentially explain some of our estimated effects for DMARDs on TB risk.

In summary, the age- and sex-standardized incidence rates of TB in RA patients are 10 times that of the general population. At least some of this risk may be related to DMARD and corticosteroid therapies, especially in those age >65 years. Our findings are supportive of published guidelines that recommend TB screening before initiation of any immunosuppressive therapy (13, 24–28).

AUTHOR CONTRIBUTIONS

Dr. Brassard 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. Brassard, Suissa.

Acquisition of data. Suissa.

Analysis and interpretation of data. Brassard, Lowe, Bernatsky, Kezouh, Suissa.

Manuscript preparation. Brassard, Lowe, Bernatsky, Suissa.

Statistical analysis. Kezouh, Suissa.

Ancillary