Factors Affecting the Tuberculosis Risk in Patients Receiving Anti-Tumor Necrosis Factor-α Treatment

Tuberculosis (TB) is a major global public health problem, with one third of the world's population estimated to be infected with Mycobacterium tuberculosis. Although the primary site of involvement is the lungs (pulmonary TB; P-TB) in most cases, other organ systems may also be affected (extrapulmonary TB; EP-TB). Annually, 8.6 million people are infected, representing the second leading cause of death from an infectious disease worldwide (after human immunodeficiency virus) [1]. Certain biological drugs that have been introduced recently, such as tumor necrosis factor (TNF)-α inhibitors, are known to increase the risk of TB [2,3,4,5]. TNF-α inhibitors are increasingly being used for the treatment of a number of immune-related diseases in clinical practice. These agents inhibit TNF-α, which is a key cytokine with an important role in the host defense against infections and in particular against Mycobacterium infections through a number of mechanisms. These mechanisms include: formation of granulomas; stimulation of dendritic cell maturation and antigen-specific T cell proliferation; increasing neutrophil adherence, phagocytic activity, and degranulation, and inhibition of the reactivation of dormant bacilli [6,7,8,9,10]. Several guidelines recommend 6 or 9 months of isoniazid prophylaxis (INH-Px) for the treatment of latent TB infections, based on studies which have reported up to a 10-fold increase in the risk of reactivation of latent TB in patients undergoing anti-TNF-α treatment (aTNF-α-T) and a decrease with appropriate chemoprophylaxis in the same group of individuals [6,11,12,13,14,15,16].

One of every 4 individuals is estimated to be infected with M. tuberculosis in Turkey, with a TB incidence of 22/100,000 [17,18]. On the other hand, limited data from this country regarding the incidence of TB in patients undergoing aTNF-α-T suggests an 11-40 times higher risk of TB compared to the general population [5,19,20,21,22].

We aimed to determine the risk factors for TB in patients receiving aTNF-α-T in this study.

Pharmacovigilance regulations in Turkey mandate that all patients be screened by chest physicians for TB risk before initiating aTNF-α-T. Between October 2003 and October 2013, a total 3,094 patients were screened for TB risk before initiation of aTNF-α-T for various autoimmune diseases at the Outpatient Unit, Department of Pulmonary Diseases, Cerrahpasa Faculty of Medicine, Istanbul University. Among these, 1,964 patients on regular aTNF-α-T with a follow-up time longer than 6 months were identified and included in this study. The screening was carried out in accordance with the consensus report on using TNF-α inhibitors endorsed by the Rheumatology Society of Turkey [23]. A tuberculin skin test (TST) was performed using the Mantoux method. A standard dose of 5 tuberculin units (0.1 ml) is injected intradermally and evaluated after 48-72 h. The TST result was reported in millimeters as the measured value. If the initial reaction was between 0 and 4 mm in diameter, the TST was repeated 1 or 2 weeks later to evaluate the booster phenomenon. In case of suspicion of active TB, a comprehensive work-up was performed that included a sputum smear and culture, a thorax CT, and a bronchoscopy; if needed, further tests were carried out. INH-Px (300 mg/day) was given for 9 months if the screened patient had any of the following: a history of close contact with infectious cases within the past year, fibrotic changes consistent with TB on a chest X-ray (CXR) with no adequate history or documentation of a previous anti-TB treatment or a TST wheal ≥5 mm in diameter. In addition, INH-Px could be initiated at the discretion of the treating physician based on his/her clinical judgement. The aTNF-α-T was usually initiated 1 month after the start of INH-Px. Patients were followed up via regular visits performed every 3 months during the aTNF-α-T.

Since one of the primary objectives of this study was to evaluate the effect of INH-Px on the risk of TB, only those patients who had a follow-up time longer than 6 months under aTNF-α-T (n = 1964) were included.

Medical records of the patients were used to collect information on the following potential risk factors for TB: age, gender, smoking history, amount of smoking (<10 pack-years vs. ≥10 pack-years), primary disease, TST findings (anergic, wheal diameter <5 mm, wheal diameter 5-9 mm, and wheal diameter ≥10 mm), use of INH-Px, type of anti-TNF-α agent used, and use of other immunosuppressive agents such as methotrexate, azathioprine, and corticosteroids. Statistical analyses were performed using SPSS version 21.0 for Windows. A χ² test was used for comparison of categorical data. Continuous variables are reported as means ± SD or medians, and all categorical variables are expressed as numbers and percents of cases. The associated risk factors for TB were analyzed with multivariate logistic regression in order to define the most relevant predictive variables. The study protocol was approved by the Ethics Board of Cerrahpasa Medical School, Istanbul University (approval No. 83045809-3708).

The mean follow-up time for the 1,964 participants was 27.9 ± 21.0 months (median 21). Of these, 1,009 (51%) were male and 955 (49%) were female, with a mean age of 39.7 ± 13.9 years. The primary diseases for aTNF-α-T were ankylosing spondylitis in 875 (44.6%), rheumatoid arthritis in 711 (36.2%), Behçet's disease in 83 (4.2%), and others in the remaining 295 (15%) patients. Fifty-seven patients (2.9%) had a previous history of TB; 886 patients [45.1%; 592 (66.8%) males and 294 (33.2%) females] were smokers (mean 14.8 ± 12.3 pack-years). The following anti-TNF-α agents were given to the patients: etanercept (n = 760; 38.7%), infliximab (n = 684; 34.8%), and adalimumab (n = 520; 26.5%). The demographic characteristics of the patients are shown in table 1.

Of 1,964 patients, 1,250 (63.6%) received INH-Px with the following indications: positive TST (n = 1,162; 93%), fibrotic changes consistent with TB on a chest CXR (n = 34; 2.7%), a history of close contact with infectious individuals within the past year (n = 9; 0.7%), and at the discretion of the physician (n = 45; 3.6%).

Sixteen patients [8 (50%) males and 8 (50%) females] developed TB (mean age 40.1 ± 14.6 years, range 25-74). Eight (50%) of them were in the INH-Px group. The annual incidence of TB in this patient group was 466/100,000.

Of the 16 TB cases identified, 5 (31.2%) had P-TB and 11 (68.8%) had EP-TB [military, n = 4 (25%); TB pleurisy, n = 3 (18.8%); intestinal TB, n = 2 (6.6%); TB of the ear, n = 1 (6.3%), and TB of the urogenital system, n = 1 (6.3%)]. All 16 patients were on aTNF-α-T at the time of the diagnosis of TB. The mean time between the initiation of aTNF-α-T and the TB diagnosis was 26.8 ± 17.1 months (range 1-60, median 26.5). TB developed in 3 (18.8%) patients during the first 12 months, in 3 (18.8%) patients between 12 and 18 months, and in 10 (62.4%) patients after 18 months. The characteristics of the 16 patients who developed TB during aTNF-α-T are shown in table 2.

A significantly lower proportion of patients with a TST wheal ≥10 mm in diameter (3/885) developed TB as compared to those with a TST wheal <10 mm in diameter (12/1,079; p = 0.027).

The primary condition necessitating the use of aTNF-α-T in these 16 patients with a diagnosis of TB was ankylosing spondylitis in 8 (50%), rheumatoid arthritis in 4 (25%), Behçet's disease in 3 (18.8%), and Takayasu arteritis in 1 (6.3%) patient. TB was diagnosed in 8 of 875 (0.9%) ankylosing spondylitis patients, in 4 of 711 (0.5%) rheumatoid arthritis patients, and in 3 of 83 (3.6%) Behçet disease patients. A significantly higher proportion of patients with Behçet's disease were diagnosed with TB compared to patients with other conditions (p = 0.028). Thirteen of the 16 TB patients (81.2%) were also receiving at least one other immunosuppressive agent in addition to TNF inhibitors as follows: azathioprine, n = 1 (7.6%); prednisolone, n = 6 (46.2%); azothioprine + prednisolone, n = 2 (15.4%), and methotrexate, n = 4 (30.8%). TB occurred significantly more frequently among patients receiving azathioprine (3/82) than among those receiving other immunosuppressives (13/1,882; p = 0.038); however, it should be pointed out that azathioprine was prescribed only to patients with Behçet's disease. No association was found between use of methotrexate or steroids and TB development (p > 0.05).

There was also no association between TB development and the age and gender of the patients (p > 0.05) or the type of anti-TNF-α agent (p > 0.05). However, the rate of TB in patients treated with infliximab was approximately twice that of patients who received adalimumab or etanercept (the patient-years for etanercept were 23/10,000 compared to 52/10,000 for infliximab and 27/10,000 for adalimumab). Six (37.5%) of the 16 patients with TB were nonsmokers, while 10 (62.5%) were current smokers (mean 12.1 ± 10.4 pack-years, range 4-40). Three (60%) of the 5 patients with pulmonary TB were current smokers, while the corresponding figure was 7 (63.6%) among 11 patients with EP-TB. There was no statistically significant difference between current smokers and nonsmokers in terms of TB risk (p > 0.05). In addition, the smoking history length was not associated with an increased TB risk (<10 vs. ≥10 pack-years, p > 0.05).

TB developed in 8 of 1,250 patients who received INH-Px (the TB incidence was 367/100,000) and in 8 of 714 patients who did not receive INH-Px (the TB incidence was 635/100,000), with the difference being insignificant (p = 0.3).

When patients who received INH-Px were assigned to 2 groups based on purified protein derivative (PPD) results as those with a wheal diameter ≥10 mm and those with a wheal diameter between 5 and 9 mm, a borderline insignificant difference was observed between the 2 groups (3/885 for >10 mm and 4/277 for 5-9 mm; p = 0.05).

INH resistance was found in only 1 (12.5%) of the 8 patients who developed TB after INH-Px, but the subsequent treatment for TB in this patient was successful (case 7 in table 2).

In the multivariate regression analysis, Behçet's disease (p = 0.003) and having a TST wheal ≥10 mm in diameter (p = 0.027) were found to be associated with the development of TB infection. The effect of azathioprine use disappeared in the multivariate analysis (p > 0.05; table 3).

TNF-α plays key role in the pathogenesis of TB through its involvement in almost all immunological and pathological responses of the host against M. tuberculosis [24,25]. Therefore, an increased risk of developing P-TB, EP-TB, or disseminated forms of TB infection is expected in patients receiving aTNF-α-T [26,27,28,29]. Askling et al. [4] reported a 4-fold increased risk of TB in subjects with rheumatoid arthritis receiving aTNF-α-T from 1999 to 2004. In another report from Japan, TB was found to develop in 13 out of 4,000 patients receiving these agents [30]. In contrast, there is a relative scarcity of data on the prevalence of TB in patients receiving aTNF-α-T in Turkey [5,19,20,21,22]. Hanta et al. [20] reported 3 patients developing TB out of 192 who received aTNF-α-T. Elbek et al. [5] concluded that aTNF-α-T is associated with a nearly 40-fold increased risk of TB after observing active TB in 2 of 240 subjects during the course of aTNF-α-T. Recently, Cagatay et al. [22] reported that 20 of 1,794 patients developed TB during such treatment. In our study, 16 of 1,964 patients who received aTNF-α-T for more than 6 months developed TB during an average follow-up period of 21 months, corresponding to an annual incidence of 466/100,000, i.e. approximately 21 times the incidence of TB in Turkey (22/100,000) [18].

Several guidelines recommend patient assessment prior to the initiation of TNF inhibitors, with careful history taking regarding exposure to TB, TST, and CXR. These guidelines also recommend INH-Px (300 mg/day) for 6 or 9 months in patients diagnosed with latent TB [11,12,23]. Sichletidis et al. [31] reported the occurrence of active TB in 1 out of 3 patients despite INH-Px (patients with a TST wheal ≥10 mm in diameter received INH-Px for 6 months) during a 3-year follow-up. In contrast with the results reported by Sichletidis et al. [31], Manadan et al. [32] reported no active TB in patients who received INH-Px. Cagatay et al. [22] reported that 17 out of 20 TB cases developed TB in spite of INH-Px. Compared to all these studies, our study had a larger sample size. Among 16 patients who developed TB during 21 months of follow-up, TB developed despite INH-Px in 8 cases. Although no significant association between administration of chemoprophylaxis and TB development was found (8/1,250 among those who received INH-Px vs. 8/714 in those without INH-Px; p = 0.3), the TB incidence was lower among patients who received INH-Px (367/100,000) than among those who did not receive INH-Px (635/100,000). Since patients undergoing INH-Px represented a high-risk group, the absence of differences between INH users and INH nonusers in terms of the occurrence of TB, together with a lower TB prevalence among INH users, may be interpreted as an indication of the efficacy of INH-Px in reducing TB in high-risk groups. Also, in our study no patient developed TB during INH-Px, although no relationship was found between chemoprophylaxis and the risk of developing TB. This result is similar to the recently published study of Churchyard et al. [33], who examined workers in gold mines and found that, during the first 9-month period when INH was given, the incidence of TB was 58% lower in the INH cohort than in the control cohort but subsequently the TB incidence was similar in the 2 cohorts after INH-Px [33]. Clearly, more studies are warranted to shed more light on this issue. On the other hand, the 21 times higher TB incidence reported despite the use of follow-up protocols and INH-Px raises some doubts about the efficacy of the follow-up protocols and INH-Px implemented.

An unexpected result of our study was the significantly lower occurrence of TB in patients with a TST wheal ≥10 mm in diameter (3/885) compared to those with a TST wheal <10 mm in diameter (13/1,079; p = 0.027). This finding may be explained by the fact that all of these patients had chronic immune-related diseases and most of them also received immunomodulator drugs. A strong reaction on a TST despite the presence of an immune-related disease and chronic illness may be considered a sign of a relatively better immune status, conferring protection for the patient. In these patients, generally an inadequate response to a TST is expected. When patients on INH-Px were categorized on the basis of PPD results as those with a skin reaction ≥10 mm in diameter and those with a reaction between 5 and 9 mm in diameter, no significant difference in terms of TB risk was found, suggesting that the low risk of TB in those with a PPD result ≥10 mm was not associated with INH-Px.

Of the 16 patients with an established diagnosis of TB, 5 (31.2%) had P-TB, while 11 (68.8%) had EP-TB. Similar to our results, Sichletidis et al. [31] reported that, of the 11 TB cases they identified despite chemoprophylaxis among patients receiving an anti-TNF agent, 6 (54.5%) had P-TB and 5 (45.5%) had EP-TB (nasopharygeal TB, n = 1; oropharyngeal TB, n = 1; lymph node TB, n = 1; TB pleurisy, n = 1, and spleen TB, n = 1) [31]. This high occurrence of EP-TB indicates that, in patients receiving aTNF-α-T, a high index of suspicion should be maintained for P-TB and EP-TB when signs of infection are observed.

All 16 (100%) patients with an established diagnosis of TB were on aTNF-α-T at the time of diagnosis. The mean time between the initiation of aTNF-α-T and the TB diagnosis was 26.8 ± 17.1 months. TB was diagnosed during the first 12 months in 3 patients, between 12 and 18 months in 3 patients, and after 18 months in 10 patients. These data suggest an increased risk of TB during the late phase of the aTNF-α-T and give rise to a number of intriguing questions. For instance, can there be excessive immunosuppression in the long term in patients on aTNF-α-T, or could late TB infections occurring despite INH-Px be due to reinfections rather than latent infections, or should INH-Px be continued for more than 9 months? Certainly, further research is warranted to answer these questions.

Rheumatoid arthritis patients are known to have a higher risk of TB than the general population. Also agents used for the treatment of rheumatoid arthritis may further increase the risk of TB [34,35,36]. Recently, Kim et al. [37] reported that the TB incidence in ankylosing spondylitis and rheumatoid arthritis patients receiving aTNF-α-T is similar, and EP-TB was prevalent in both conditions. On the other hand, a higher risk of TB was observed in patients diagnosed with Behçet's disease in our study (p = 0.003), consistent with some previous reports suggesting an association between an elevated TB risk and Behçet's disease [38,39]. This maybe related to the fact that Behçet's disease generally involves large and small blood vessels, facilitating TB infection through a ventilation-perfusion mismatch subsequent to microclotting.

A differential risk of TB was reported for different anti-TNF-α agents, with more cases reported for monoclonal antibodies compared to receptor proteins (etanercept, 35/100,000; infliximab, 144/100,000, and adalimumab, 240/100,000) [38]. Wallis et al. [16] reported a 3-fold increased risk of TB with infliximab treatment compared to etanercept. However, in our study the TB incidence was comparable between the 3 anti-TNF-α agents (the patient-years for etanercept were 23/10,000 as compared to 52/10,000 for infliximab and 27/10,000 for adalimumab; p > 0.05). Also, there was no relationship between age, gender, and the risk of TB (p > 0.05)

The potential strengths of our study include the fact that it represents one of the largest and longest studies investigating TB and related risk factors in patients receiving aTNF-α-T.

However, its retrospective design with the absence of an appropriate control group, the absence of interferon-γ release assay testing for TB in addition to TST, and the absence of data on compliance with INH treatment are among its limitations.

In conclusion, our results suggest that aTNF-α-T increases the risk of P-TB and EP-TB despite the implementation of follow-up protocols and INH-Px, and that TB often develops during the late phase of such treatment. Although INH-Px appears to reduce the TB risk in high-risk subjects, TB development despite Px in some patients raises questions about the duration of the Px in high-risk groups in populations with a high prevalence of TB infection. The risk of TB was higher in patients with Behçet's disease and lower in patients who had a strong reaction to the TST. Patients receiving aTNF-α-T should be monitored closely for TB even at the late phases of the therapy.