Objective In North America, tuberculosis and nontuberculous mycobacterial (NTM) disease rates associated with antitumour necrosis factor α (anti-TNFα) therapy are unknown.
Methods At Kaiser Permanente Northern California, the authors searched automated pharmacy records to identify inflammatory disease patients who received anti-TNF therapy during 2000–2008 and used validated electronic search algorithms to identify NTM and tuberculosis cases occurring during anti-TNF drug exposure.
Results Of 8418 anti-TNF users identified, 60% had rheumatoid arthritis (RA). Among anti-TNF users, 18 developed NTM and 16 tuberculosis after drug start. Anti-TNF associated rates of NTM and tuberculosis were 74 (95% CI: 37 to 111) and 49 (95% CI: 18 to 79) per 100 000 person-years, respectively. Rates (per 100, 000 person-years) for NTM and tuberculosis respectively for etanercept were 35 (95% CI: 1 to 69) and 17 (95% CI: 0 to 41); infliximab, 116 (95% CI: 30 to 203) and 83 (95% CI: 10 to 156); and adalimumab, 122 (95% CI: 3 to 241) and 91 (95% CI: 19 to 267). Background rates for NTM and tuberculosis in unexposed RA-patients were 19.2 (14.2 to 25.0) and 8.7 (5.3 to 13.2), and in the general population were 4.1 (95% CI 3.9 to 4.4) and 2.8 (95% CI 2.6 to 3.0) per 100, 000 person-years. Among anti-TNF users, compared with uninfected individuals, NTM case-patients were older (median age 68 vs 50 years, p<0.01) and more likely to have RA (100% vs 60%, p<0.01); whereas, tuberculosis case-patients were more likely to have diabetes (37% vs 16%, p=0.02) or chronic renal disease (25% vs 6%, p=0.02).
Conclusions Among anti-TNF users in USA, mycobacterial disease rates are elevated, and NTM is associated with RA.
Statistics from Altmetric.com
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.
Over two million persons in USA suffer from rheumatoid arthritis (RA),1 and in the last decade, a number of biologic, immunosuppressive therapies have been widely used to treat this and other autoimmune inflammatory diseases.2 ,3 These therapies include those that inhibit tumour necrosis factor α (TNFα), a cytokine integral to host defense against intracellular pathogens such as mycobacteria, histoplasmosis and others.4,–,7 Although mycobacterial infections are known to be important in the anti-TNF setting,8 ,9 surprisingly little population-based data regarding their incidence and associated risk factors in the anti-TNF setting has been published, particularly in North America. In Europe, the risk of anti-TNF associated tuberculosis has been recently documented10 ,11 but in the USA, the incidence of tuberculosis in patients using anti-TNF therapy remains unknown. Further, in regions of low tuberculosis prevalence, it is possible that nontuberculous mycobacteria (NTM) are a more important complication of anti-TNF therapy.12 These environmental organisms are ubiquitously distributed in soil and municipal water systems, and are capable of causing both pulmonary and extrapulmonary disease.13 ,14 Contrary to tuberculosis, NTM incidence is increasing within USA and Europe, particularly among older women.15
Accordingly, we undertook the present study to describe the incidence, clinical characteristics and associated risk factors for anti-TNF associated tuberculosis and NTM disease within a defined population of patients using anti-TNF therapy. We conducted this study within Kaiser Permanente Northern California (KPNC), a large health maintenance organisation with over 3 million members with a background tuberculosis rate similar to that of the US (unpublished data, California Department of Health Services TB Control Branch).
Anti-TNF user cohort identification and characteristics
By searching electronic medical records, we identified all KPNC patients with ≥1 clinical visit and ≥1 outpatient prescription for etanercept or adalimumab, or ≥1 infusion of infliximab during 1 January 2000 to 31 December 2008. For all identified anti-TNF users, we searched their electronic medical records for inpatient or outpatient international classification of disease (ICD-9) codes during the study period for the following indications for anti-TNF therapy: RA (714.xx), psoriasis (696.1), psoriatic arthritis (696.0), Crohn's disease (555.xx), ulcerative colitis (556.xx) and ankylosing spondylitis (720.0).
Mycobacterial case finding
Within the anti-TNF cohort, we used validated electronic search algorithms to identify pulmonary and extrapulmonary cases of tuberculosis and NTM disease.16 To identify culture-negative cases of tuberculosis, we relied upon validated algorithms combining ICD-9 codes11,–,18 and antimycobacterial therapy receipt.16 All tuberculosis and NTM cases were adjudicated with medical record review by coauthors with clinical expertise in mycobacterial disease (KLW, RB) using accepted case definitions for disease.16
Descriptive epidemiology of mycobacterial cases among the anti-TNF cohort
We assigned a diagnosis date to each case of mycobacterial disease (defined as first positive culture, or in the case of culture negative disease, the date of first ICD-9 code). We collected the following information from each case for the 2000-2008 study time-period: age at diagnosis date, sex, race/ethnicity and evidence of death. We examined important clinical comorbidities at two time-points during the study time-period:1 prior to anti-TNF start and2 anytime during the study time-period, as evidenced by one or more inpatient or outpatient ICD9 code given during these time-frames for diabetes mellitus (250), chronic kidney disease (585), neoplasm (200–208, 195, 196, 162), chronic lung disease including chronic bronchitis, asthma, bronchiectasis, silicosis (491, 492, 493, 494, 495, 496, 502), gastro-esophageal reflux disease (530.81, 530.11). Unfortunately, results of tuberculin skin testing were not reliably available for the study population.
Non-cases among the anti-TNF cohort
From the anti-TNF users that never developed tuberculosis or NTM during the study time-period, we collected background information regarding sex and clinical comorbidity information using ICD9 codes given at any time during the study time-period.
Comparison of tuberculosis and NTM cases with uninfected anti-TNF users
Data were entered into an Access Database (Microsoft, Redmond, Washington, USA) and analysed in SAS V. 9.1 (SAS Institute, Cary, North Carolina, USA). Univariate comparisons were made to evaluate epidemiological and clinical comorbidity differences between case-patients and uninfected anti-TNF users. We used χ2 and Fisher's exact tests to evaluate observed differences. For continuous variables, we used logistic regression to assess their association with tuberculosis and NTM. Multivariate analysis was precluded by the low number of mycobacterial cases identified.
Immunosuppressive exposure and disease rate calculations
We calculated patient exposure time to anti-TNF therapy as on-drug time (ie, date of first dispensing until date of first missed dose) plus 90 days after end of supply. During this interval, patients were considered ‘currently exposed.’ Patients having no record of resuming supply after a 90-day interruption were then considered unexposed until supply resumed. For our primary analysis, we attributed tuberculosis and NTM cases to anti-TNF exposure if the patient had evidence of infliximab infusion receipt, or in the case of adalimumab or etanercept, evidence of drug supply end-date within 90 days of tuberculosis or NTM diagnosis date. We then calculated disease incidence rates using only these patients in the numerator, and the denominator included all patient anti-TNF exposure time for the study time-period (with exposure time for case-patients censored at time of infection diagnosis). For sensitivity analysis, we also calculated rates considering cases currently exposed to anti-TNF therapy if they developed disease1 anytime after anti-TNF start or2 within 6 months of last exposure to anti-TNF therapy. These calculations produced similar estimates as our primary analysis, and for simplicity sake, were not presented here.
For non-biologic therapies (eg, prednisone, methotrexate, leflunomide, others), case-patients were considered exposed if they had evidence of pharmacy dispensings 30 days prior to mycobacterial diagnosis.
For comparison, we calculated background tuberculosis and NTM disease incidence among patients without evidence of anti-TNF therapy during the study time-period for two groups:1 those with RA, and2 the background KPNC health plan population. For tuberculosis, we used microbiological data to ascertain culture-confirmed TB cases (we did not ascertain culture-negative cases for these background calculations), and for NTM we used validated microbiological criteria of the ATS (American Thoracic Society)/IDSA (Infectious Diseases Society of America) case definition for pulmonary NTM disease, and extrapulmonary isolates were considered as disease if they were isolated from a sterile site.16 We calculated denominator person-time for RA patients and the general background population who were unexposed to anti-TNF therapy during the entire study time-period. Person-year time for RA patients lacking anti-TNF use began with their first diagnostic code for RA (they had to have at least two 714.xx codes to meet RA case definition) and was censored at the time of TB or NTM development, death, health plan disenrollment, or study time-period end, whichever came first. For the unexposed general KPNC population, follow-up time began at date of health plan enrolment (or start of study time-period for those already enrolled) with similar methods used for right censoring as above. All incidence rate calculations used Poisson 95% CIs.
Anti-TNF cohort characteristics
We identified 8418 anti-TNF users during the study period, with a total of 20 330 person-years of anti-TNF exposure. Most (61%) were patients carrying diagnostic codes for RA, 64% were women and 61% were white, non-Hispanic. Among this group, we identified 16 tuberculosis and 18 NTM cases that developed disease after starting anti-TNF therapy. An additional six tuberculosis and four NTM cases had developed disease before initiating anti-TNF therapy; these cases were excluded from the study.
Characteristics of tuberculosis and NTM cases
Of the 16 total tuberculosis cases identified, 11 (69%) were pulmonary, four (25%) extrapulmonary and one unknown. Twelve (75%) had RA, one (6.3%) had Crohn's disease, one (6.3%) had ulcerative colitis and three (18.8%) had psoriasis. Case-patients were median aged 57.5 years (range, 31–77 years), mostly women (n=13.81%), with median time to onset after drug start of 670 days (range, 1–3181 days). Three (19%) case-patients died during the study period. The median time between infection and death was 74 days (range 72–1946 days). Nine (44%) were taking prednisone and four (25%) were taking methotrexate during the 30 days preceding their diagnosis date. Ten (63%) were ‘currently exposed’ to anti-TNF therapy at time of diagnosis (figure 1).
Of 18 NTM cases identified, 12 (67%) were pulmonary, four (22%) extrapulmonary, two unknown; median onset between anti-TNF start and diagnosis was 1027 days (range, 77–2832 days). Case-patients were median aged 68 years (range 49–81 years), mostly women (n=13, 72%), and 17 (94%) were white and non-Hispanic. Eighteen (100%) had RA, and three (17%) and two (11%) carried codes for psoriasis and ankylosing spondylitis, respectively. Seven (39%) case-patients died with a median time between infection and death of 569 days (range, 21–2127 days). At time of NTM diagnosis, nine (50%) and two (11%) were taking prednisone and methotrexate respectively, and 15 (83%) were ‘currently exposed’ to anti-TNF therapy. (figure 2).
Incidence rate of mycobacterial disease among anti-TNF users
NTM and tuberculosis incidence rates were calculated using only cases who developed disease while currently exposed to anti-TNF therapy (NTM n=15, tuberculosis n=10). Anti-TNF associated NTM rates were higher than tuberculosis rates, and rates of both tuberculosis and NTM were higher among patients using monoclonal antibodies; however, 95% CI overlapped between all comparisons (table 1). Anti-TNF associated tuberculosis and NTM incidence rates were significantly higher than rates observed in unexposed RA and background KPNC populations (table 2). Among unexposed RA patients, 19 tuberculosis and 42 NTM cases were detected in 219 295 patient-years of follow-up. Within the unexposed background KPNC population, 774 tuberculosis and 1145 NTM cases were detected in 27 668 456 years of follow-up.
Anti-TNF exposure after mycobacterial diagnosis
Of the 45 total mycobacterial cases (including those who developed disease prior to anti-TNF start), 24 (53%) had evidence of taking anti-TNF therapy after their diagnosis date; 15 (68%) NTM cases and nine (39%) tuberculosis cases. In this group, there were five deaths, all in NTM patients (figures 1 and 2).
Tuberculosis and NTM risk factors among patients who use anti-TNF therapy
Within the anti-TNF using cohort, tuberculosis patients were significantly less likely to be white, non-Hispanic and significantly more likely to have diabetes mellitus or chronic renal disease as compared with uninfected patients (table 3). NTM patients by contrast, were significantly older, more likely to be white, non-Hispanic, or have gastro-eesophageal reflux disease (GERD) or chronic lung disease as compared with uninfected anti-TNF users (table 3). All NTM cases had RA, compared with only 60% of uninfected anti-TNF users. Infliximab use at any time during the study period was also more likely in the NTM cases (67% vs 33.1%, OR=4.0, 95% CI 1.5 to 10.8) as compared with uninfected individuals. When stratified by infliximab use, RA continued to be a risk factor in those exposed to infliximab (p<0.01) and trended towards significance in those without infliximab exposure (p=0.09).
Since most (89%) NTM cases were over 50 years old, we completed a secondary analysis restricted to those ≥ 50 years old. Comparison of these NTM cases with uninfected anti-TNF users in the ≥ 50 years old population yielded similar results, including a significant association between NTM and RA (p<0.05). All seven of the observed NTM case deaths were in this group. NTM disease rates were substantially higher in this ≥50 years old subset (table 1).
We calculated the incidence of tuberculosis and NTM disease in a large cohort of anti-TNF users from an over 3 million member integrated healthcare organisation in Northern California. We found tuberculosis and NTM disease rates among those using anti-TNF therapy to be five to tenfold higher than that in the unexposed RA and general background populations respectively; rates were highest in patients using monoclonal antibodies. Among anti-TNF users, NTM cases occurred more frequently than tuberculosis and a greater proportion of NTM patients died. Tuberculosis patients were more likely to have renal disease and diabetes, whereas NTM patients were more likely to have chronic lung disease and GERD. Among the inflammatory diseases represented in this anti-TNF using population, RA was significantly associated with the development of NTM independent of age and type of anti-TNF therapy used.
In the USA, a region of low tuberculosis prevalence, our study highlights the importance of NTM in the setting of anti-TNF therapy and RA. While it is unclear from our work that anti-TNF therapy itself raises the risk of NTM disease, the rate of NTM disease in this group of patients is elevated over the general population, including the background unexposed RA population, and elevated relative to the rate of tuberculosis (although CI overlapped between estimates). Consistent with a prior survey of US infectious disease clinicians, our study suggests that NTM occurs as or more frequently in this patient population than does tuberculosis.12 This is perhaps not surprising, given the low prevalence of tuberculosis in USA, and given that tuberculosis screening guidelines have been well-publicised in the last 5 years making it likely that tuberculosis cases were averted in our study population. Further, recent epidemiological studies suggest NTM disease to be increasing in prevalence particularly within older, white women, a demographic similar to that of RA.15 ,17 ,18 Pulmonary NTM disease is difficult to treat and in many cases cure is not attainable,.14 And in our series, more anti-TNF treated cases with NTM died as compared with TB. It is currently unclear if patients with active NTM disease can safely receive anti-TNF therapy. The cases herein, and the reports of others, suggest patients with NTM disease should discontinue anti-TNF therapy.12 ,19,–,21
Chronic underlying lung disease is a known risk factor for NTM, and it is likely that the lung disease of RA predisposes to pulmonary NTM acquisition. Similar to findings from our review of anti-TNF associated NTM cases contained within the FDA MedWatch system, our current study suggests approximately 60% of anti-TNF associated NTM cases are pulmonary. Little is known regarding the pathogenesis of pulmonary or extrapulmonary NTM disease, although prior work has suggested that patients with pulmonary NTM do have lower levels of TNFα, suggesting a theoretical role of anti-TNF therapy in predisposing patients to pulmonary disease.22 Extrapulmonary NTM disease is more difficult to explain as a feature of RA, and the existence of such cases suggest these or other immunosuppressive therapies predispose patients to such infections.
Prior studies evaluating the risks of tuberculosis associated with anti-TNF therapy have primarily come from Europe where the background rate of tuberculosis is generally higher than in USA. Remarkably, our risk estimates for tuberculosis among anti-TNF users are proportionally similar to those documented by the British Biologic Registry (BBR) last year.11 While the BBR documented rates approximately twice as high as ours, this was consistent with their background population TB rate being higher than that of KPNC (12 vs 2.8 per 100 000). Similarly, like the BBR study, we documented etanercerpt-associated tuberculosis rates to be approximately threefold higher than background rates, with observed rates for infliximab and adalimumab an additional three to fourfold higher than etanercept-related rates. A variety of explanations have been proposed to explain the difference in observed tuberculosis risk between these compounds.23
Our work highlights important risk factors for both tuberculosis and NTM in inflammatory disease patients. These include chronic renal disease and diabetes mellitus for tuberculosis, and chronic lung disease for NTM. These associations are important for physicians to recognise when considering anti-TNF therapy use.18 ,24 Tuberculosis screening prior to anti-TNF therapy use has been well-publicised, and has been associated with prevention of tuberculosis in at least one published experience from Spain.25 Unfortunately we were unable to evaluate the prevalence or success of tuberculosis screening efforts in our study population. For now, treating physicians should use local guidelines24 ,26,–,28 to screen and treat for latent tuberculosis prior to using biologic therapies. All guidelines cite the importance of taking a close history of tuberculosis risk factors followed by employment of one or more diagnostic tests for tuberculosis exposure (ie, interferon γ release assays, tuberculin skin test).28 We suspect some patients with underlying NTM are discovered during tuberculosis screening. During such screening, physicians should consider obtaining sputum for culture if patients have a chronic unexplained cough or a chest radiograph with infiltrate or bronchiectasis.18 ,29
The strengths of using the KPNC database included the ability to use culture data to better ascertain cases, and the ability to validate cases by record review. Further, the health plan has a large population (3.1 million) with demographic characteristics and a background tuberculosis rate similar to USA. However, in calculating the background rate of tuberculosis within the KPNC and unexposed RA populations, we were limited to using only culture data to find cases. Our rate of 2.8/100 000 underestimates the true rate of tuberculosis within this population during the study time-period, as approximately 15–20% of tuberculosis cases are not culture confirmed.30 On the other hand, the rates calculated within our anti-TNF cohort included all culture confirmed and unconfirmed cases as we were able to use validated search algorithms in combination with chart review to ascertain all cases.16 Accordingly, our calculations have slightly overestimated the magnitude by which anti-TNF associated tuberculosis rates differ from background. In addition, we were limited by our study methodology in that we cannot definitively conclude that anti-TNF drugs independently increase the risk of tuberculosis and NTM. While our findings and the findings of others support this notion,10 ,11 we did not directly compare exposed patients with unexposed patients, and did not control for underlying disease severity or the presence of comorbid diseases that could influence the risk of tuberculosis and NTM.
In summary, in population-based fashion, we have documented rates of tuberculosis and NTM disease among users of anti-TNF therapy to be five to tenfold higher than disease rates seen in the unexposed background RA and general populations. Importantly, work suggests NTM to be of particular importance to patients with RA within this setting. Clinicians should continue to screen for tuberculosis prior to use of anti-TNF therapy, and should remain vigilant for the presence or development of NTM disease particularly in patients who use these therapies.
Funding This work was funded by a grant from UCB pharmaceuticals. KL Winthrop's work on this manuscript was funded by an Agency for Healthcare Research and Quality (AHRQ) grant (1K08HS017552-01).
Competing interests KL Winthrop has received a grant from UCB pharmaceuticals and scientific advisory board fees from Amgen, Genentech, and Oxford Immunotech. JR Curtis has served as a consultant for Roche/Genentech, UCB, Centocor, CORRONA, Amgen, Pfizer, BMS, Crescendo and Abbott, and has received research support from Amgen, Roche/Genentech, Centocor and CORRONA. JW Baddley has served as a consultant for Abbott. EB Suhler has received research support from Abbott and Centocor. JT Rosenbaum consults for Amgen, Abbott, Centocor, and Genentech and has received grant support from Centocor, Genentech, and Abbott. LJ Herrinton has a research contract with Genentech and Proctor and Gamble
Provenance and peer review Not commissioned; externally peer reviewed.