Introduction

Biologic agents antagonising tumour necrosis factor alpha (TNF-α) are highly effective in treating patients with rheumatoid arthritis (RA), ankylosing spondylitis (AS) or psoriatic arthropathy (PsA), reducing symptoms and in most cases also inhibiting the radiological progression of the disease [1,2,3,4]. However, this treatment has been shown to be associated with various side effects, among which serious infections are one of the major concerns [5, 6]. TNF-α plays a major role in the defence against infection and in the formation and maintenance of granulomas; therefore, treatment with TNF-α inhibitors (TNFi) is recognised as a risk factor for tuberculosis (TB) [7, 8].

TNFi have been available in Romania for the treatment of patients with severe RA since 2002. The first available TNFi has been infliximab, followed by etanercept since 2005 and adalimumab since 2007. In patients with severe AS or PsA, TNFi have been used since 2008. In order to prevent TB in patients treated with TNFi, even since the beginning of their use, screening for both active TB and latent TB infection (LTBI) has been mandatory prior to the initiation of the TNFi. However, re-testing for LTBI in patients already on biologic therapy is still under debate [9] and there is no consensus for a specific protocol.

While active TB can be reasonably ruled out by history, clinical examination and chest radiography, there is no ‘gold standard’ for the diagnosis of LTBI. Screening for LTBI includes history (personal or family history of TB, history of TB exposure, ethnicity), chest radiography (in search of inactive TB lesions) and immunodiagnostic testing for latent TB. The latter has varied in time in the Romanian protocols for TNFi treatment. The first method used was the tuberculin skin test (TST), also known as Mantoux test [10] using Tuberculin PPD I.C. 65 [11]; the threshold for a positive TST varies with the underlying risk of developing active tuberculosis [12]. We have used a threshold of ≥ 10 mm from 2002 to 2006 and of ≥ 5 mm from 2007 onwards, taking into account the known increased risk of active TB in this patient population. Patients having a positive test were treated for 9 months with isoniazid 5 mg/kg/day (maximum 300 mg/day).

Limitations of TST diagnostic value for LTBI include false-positive results secondary to previous bacillus Calmette-Guerin (BCG) vaccination or exposure to other mycobacteria, and false negative results in immunocompromised patients, most notably in HIV-infected individuals [13]. Interferon-γ-releasing assays (IGRAs) are in vitro tests that measure the interferon-γ release of peripheral T cells in response to Mycobacterium tuberculosis-specific antigens [13]. IGRAs are not influenced by previous BCG vaccination and by exposure to several mycobacteria and might be less influenced by immunosuppression. Based on this evidence, the QuantiFERON-TB Gold In-Tube (QFT) test [14] was used starting with September 2010 as a mandatory immunodiagostic test for LTBI screening in arthritis patients before starting therapy with TNFi. The cutoff for a positive test was the one recommended by the manufacturer, including criteria for a valid test.

As there is little information assessing the impact of different LTBI screening protocols in patients requiring treatment with TNFi, we conducted this retrospective, single-centre study aiming to calculate the incidence rate of TB according to the initial screening method used and to evaluate the efficacy of each of the above-mentioned LTBI screening methods on preventing active TB development. Another objective of this study was to identify risk factors for developing active TB while receiving TNFi.

Patients and methods

Patients

We retrospectively reviewed the charts of all patients diagnosed with RA, AS or PsA, who were evaluated in our department and were initiated on treatment with TNFi, taken for at least 1 month, following the Romanian national guidelines, between January 1, 2002, and June 1, 2014. Only patients who were naïve to TNFi before their first visit to our clinic and who have been evaluated longitudinally for at least 12 months were included. The data were collected from the hospital electronic database and follow-up was documented until June 1, 2015. We collected demographic data, diagnosis, disease duration and a detailed history of all biologic treatments. We recorded for all patients the type and the result of LTBI screening immunodiagnostic test used before the TNFi initiation, LTBI treatment, starting and ending date of the TNFi treatment, type of TNFi used and any diagnosis of active TB after TNFi initiation. All patients receiving TNFi in our centre gave written informed consent enabling the centre to use their de-identified medical data in research projects, and this study was approved by the local ethics committee.

Diagnosis of LTBI

All patients were screened for active TB and LTBI prior to TNFi therapy. This included medical history (regarding personal or family history of TB, or exposure to TB), clinical examination, chest radiography and immunodiagostic test for LTBI (TST or QFT). The TST was performed as previously described, with a cutoff of 10 mm until December 2006, and of 5 mm between January 2007 and August 2010. Beginning with September 2010, QFT was performed, according to the manufacturer’s recommendations. All patients with a positive immunodiagnostic test or with other criteria for a high index of suspicion to have LTBI (such as presence of fibrotic or calcified sequelae of TB on the chest radiography, or low socio-economic status) were prescribed LTBI treatment with isoniazid 5 mg/kg/day (maximum 300 mg per day) for 9 months and started the TNFi after at least 1 month of isoniazid treatment. The diagnosis of active TB was based on expert opinion, sustained whenever possible by smear- or culture-positive sputum for M. tuberculosis and/or highly suggestive histologic findings. Active TB was conventionally defined as ‘early TB’ if occurring within the first 12 months of TNFi therapy or ‘late TB’ if occurring after more than 12 months of TNFi therapy. For each group, the exposure to TNFi was computed in patient-years (PY), considered from the first administration of a TNFi to the date of active TB diagnosis or last documented TNFi administration whichever was earlier. The incidence of active TB was reported as cases per 105 PY. Data on TB incidence in Romanian general population were retrieved from the National Program for Prevention, Surveillance and Control of Tuberculosis (PNPSCT) reports [15].

Statistical analysis

We divided the study population, according to the baseline immunodiagnostic LTBI screening method, in 3 groups: TST1 (TST positive if ≥ 10 mm), TST2 (TST positive if ≥ 5 mm) and QFT. The statistical analysis was done using the IBM SPSS 20.0 software [16] and the Open Source Epidemiologic Statistics for Public Health application [17]. For categorical variables, comparisons between groups were performed with the χ 2 test or with Fisher’s exact test in case that more than 20% of the cells of a contingency table had less than 5 cases. For numeric variables, comparisons were performed with the independent samples t test (for normally distributed variables) or with the Mann-Whitney U test (for non-parametrically distributed variables). Incidence rates and 95% confidence intervals for rates were compared using Fisher’s exact test adapted for the comparison of rates as the number of events was small. Further, to assess the baseline risk factors for LTBI reactivation and for any active TB while receiving TNFi, we used multivariable Cox proportional hazard regression, in order to take into account the variable time of exposure to TNFi of each patient. Each potential predictor of active TB was analysed first as a separate categorical covariate, only adjusting for age and disease duration (as continuous covariates), thereafter included in a multivariable model with all potential predictors as categorical covariates, and also age and disease duration as continuous covariates. A p value < 0.05 was considered as statistically significant.

Results

Demographic characteristics

A total of 550 patients were included in the analysis: 305 with RA, 42 with PsA and 203 with AS. Each patient belonged, per the baseline LTBI screening method, to one of the 3 groups (TST1, TST2 and QFT) as already described.

Total TNFi exposure time was 2192.8 PY with the mean duration on TNFi therapy per patient of 4.0 ± 2.4 years, and a median duration of 3.8 years. The baseline characteristics of the patients are presented in Table 1. We had follow-up data from TNFi initiation until June 1, 2015 for 532 patients, while the remaining 18 (3.4%) were considered as lost to follow-up. In the latter, we had longitudinal data for 4.6 ± 3.1 years (range 1.17–10.53).

Table 1 Demographic and clinical data of the 550 TNFi-treated patients
Full size table

TB screening

All 550 patients underwent screening for active TB and LTBI before starting the first TNFi, and 88 patients (16.0%) were positive at LTBI screening, with fewer positive screening cases in TST1 group than in the TST2 and QFT groups (6.2 vs 17.9% and 18.3%, p < 0.001 by χ 2 test), suggesting a lower sensitivity of TST1 to detect LTBI. There were no patients diagnosed with active TB at this moment (Table 2). All patients screened positive for LTBI were referred to a pulmonologist and received LTBI treatment with isoniazid 5 mg/kg/day (maximum 300 mg/day), for 9 months.

Table 2 LTBI screening results and TB occurrence in the 550 TNFi-treated patients. All data on TB incidence are per 105 patient-years. CI 95% = 95% confidence intervals
Full size table

Incidence of active TB in TNFi-treated patients

Among the 550 patients, we recorded 20 cases (3.6%) of active TB, accounting for an incidence of 912.1 cases/105 PY, about nine times higher than the average TB incidence in Romania for the period 2002–2014, which, according to the WHO reports on TB, was of 104.6 cases/105PY. Only seven patients had pulmonary TB, all confirmed by positive sputum cultures. There were seven cases of disseminated TB, four with positive sputum cultures and/or histologic evidence of TB and three with a clinical diagnosis based on symptoms, chest radiography, absence of alternative diagnosis and significant improvement on TB chemotherapy. Among the remaining six patients, four had TB pleurisy and/or pericarditis (none confirmed by smears or cultures), one mediastinal lymph node TB and one isolated hepatic TB, both with highly suggestive histologic findings for TB.

Of all patients with active TB, three had been positive at the screening for LTBI prior to the beginning of the TNFi and had consecutively received izoniazid treatment. Among the 20 active TB cases, only five were classified as ‘early TB’ (one pulmonary and four extrapulmonary TB), while 15 were classified as ‘late TB’ (six pulmonary/nine extrapulmonary). None of the five patients with early TB had been positive at the LTBI screening, and all five were receiving infliximab. Among the 15 patients with late TB, seven were receiving infliximab, seven adalimumab and one had more than one TNFi (21 months of etanercept, replaced because of loss of efficacy with adalimumab, which had been administered until the diagnosis of TB for about 16 months). This case was analysed as exposed to adalimumab before the diagnosis of TB.

Time of exposure to TNFi (mean ± SD) in patients with early TB was 0.4 ± 0.4 years [range 2–12 months], while in patients with late TB was 3.0 ± 1.4 years [range 1.3–5.4 years]. The LTBI frequency at baseline and TB cases distribution during follow-up in each group, together with TB incidence values, are shown in Table 2. TB incidence was highest in TST1group and lowest in QFT group. Incidence of early TB was 7 times greater in TST1 group when compared with TST2 and QFT groups, although these differences failed to reach statistical significance. Late TB incidence was similar in all groups. Similarly, TB incidence in the first 2 years was 6 times greater in TST1 group than in TST2 and QFT groups (data not shown here).

Risk factors for active TB

Comparing patients who developed TB with those who did not, we found a large difference in TNFi treatment, when this was dichotomised into monoclonal antibodies (MAb)—in 423 patients and non-MAb—in 107 patients, with all 20 patients developing active TB receiving MAb TNFi as the last biologic agent before TB diagnosis.

Further, we performed multivariable Cox proportional hazards regression for detecting risk factors for early TB and for any active TB, first using in each model age, disease duration at baseline and one (other) potential risk factor, then using in a final model all potential risk factors including age and disease duration. Both the results for early TB (not shown here) and those for any active TB (displayed in Table 3) have been statistically not significant. In all models, TNFi treatment dichotomised into MAb and non-MAb was statistically not significant because one of the groups (patients on non-MAb TNFi) had zero events (active TB).

Table 3 Risk factors for active TB in patients with RA, AS or PsA receiving treatment with TNFi
Full size table

Discussion

In a country with a high TB burden, TB risk in patients treated with TNFi is a major concern. This is the first Romanian study to compare the efficacy of different protocols for LTBI screening at TNFi treatment initiation, in real-life patients with inflammatory arthritis. As expected in a high-TB burden country, the TB incidence of patients treated with TNFi was about 9 times higher than the general population.

Our results show that screening for LTBI before TNFi initiation, followed by 9 months treatment with isoniazid in all patients positive at this screening, was associated with a rather low rate of early TB: 0.9% (5/550) in our cohort. We considered the patients with early TB to have most probably developed LTBI reactivation. A definitive diagnosis of reactivation TB is difficult to make without previous diagnosis of LTBI and genotyping technique. As expected, early TB rate was higher than the one of 0.0–0.3% observed in patients enrolled in international, multicentre clinical trials of TNFi and their open-label extensions [18] and this is in line with the higher TB incidence in Romania, compared to countries where these studies were performed.

An interesting result is the similar proportion of patients screened positive for LTBI when immunodiagnostic testing is performed by TST 2 (17.9%) or by QFT (18.3%). As expected, TST1 identified fewer patients with LTBI (6.2%) than TST2. The frequency of a positive LTBI screening test was lower in our cohort than in other studies, performed in countries with a significant, but lower TB incidence than Romania: 29% by TST in Spanish patients with psoriasis, 28.9% by TST /22.2% by QFT, in patients with RA and SA from Poland and 47.15% by TST / 17.8% by QFT in French patients with inflammatory arthritis [19,20,21].

We found that TB incidence in the first year of TNFi treatment was 7 times higher in TST1 group compared to TST2 and QFT groups, while TB incidence after the first year was similar in the three groups. These findings suggest that TST2 and QFT are more efficient than TST1 in preventing early TB (likely due to reactivation), but are similar in regard to prevention of late TB (presumably due to a new infection).

We would have expected that the TST2 protocol, due to its expectedly higher rate of positive results, would lead to more patients receiving unnecessary LTBI treatment when compared to the QFT protocol—but this concern is not sustained by our observations. Actually, our data fail to show any significant difference in the performance of the TST2 and QFT screening protocols; consecutively, replacing TST2 with QFT seems not to have offered much benefit to our patients in terms of reduction of unnecessary LTBI treatment or prevention of active TB.

Not unexpectedly, our results showed that the incidence of late TB exceeded that of early TB, suggesting that baseline LTBI screening has a limited effect in reducing the long-term risk of active TB in patients treated with TNFi. To address this issue, annual re-testing for LTBI in patients who were negative before TNFi initiation might be a solution. Taking into consideration the costs of QFT re-testing, the frequency and method of re-screening is still a matter of debate. QFT was proven to be cost effective as a baseline screening tool when compared to TST, in a study on patients with inflammatory bowel disease (IBD) requiring TNFi, but cost-effectiveness of re-testing has not been evaluated. In the above-mentioned study, for every 1000 patients with IBD screened by choosing the QFT over TST, 30 false negative results, 4.92 TB reactivations and 1.4 deaths were avoided [22].

Our study also aimed to identify risk factors for active TB during treatment with TNFi. The most striking result is that we did not find any TB cases in patients treated with non-MAb TNFi, who represented about 20% (107/550) of the cohort. We attempted to identify further potential risk factors for TB, such as disease (RA, AS or PsA), disease duration, TNFi treatment, baseline status at screening for LTBI and baseline method of screening, using univariable and multivariable Cox proportional hazards regression; however, none of these analyses led to statistically significant results. The lower risk of TB in patients treated with etanercept, in comparison to those receiving MAb TNFi, has already been demonstrated in larger cohorts [23,24,25]. This suggests that besides careful screening for LTBI before initiating treatment with TNFi, patients considered to have a higher risk of TB (such as health professionals or persons travelling frequently to areas of even higher TB incidence than our country) would benefit if given a non-MAb TNFi.

Prior studies have shown a slight majority of extrapulmonary TB cases in patients treated with TNFi. Our study confirms these results, showing a high percentage of extrapulmonary TB cases (57.2%), with the predominant type being disseminated TB (50% of the extrapulmonary cases and 28.6% of the total number of TB cases) [25,26,27].

Coming from a single-centre cohort, our data had high accuracy and reliability, with a very small loss to follow up and a good documentation of all TB cases, which are all strong points of this study. Limitations come from the relatively small cohort and from the lack of bacteriologic confirmation in 7 of the 20 TB cases. Finally, we were confronted with several statistical issues. The very small number of cases of TB reactivation (3% in TST1, and 1% both in TST2 and QFT) made the use of the Pearson chi-square test improper. For the same reason, the different incidence of active TB in patients treated with MAb TNFi (20/423) vs. those receiving non-MAb TNFi (0/107) could not be tested for statistical significance because there were fewer than 5 cases in the non-MAb group, in a cohort of more than 300 subjects. Limitations of the statistical methods occurred also in the Cox proportional hazards analysis, where MAb TNFi treatment was not identified as a risk factor for TB because of zero TB cases in the reference (non-MAb TNFi) group, which led to infinite 95% CIs in the Cox models and thus to apparent lack of statistical significance. These limitations may be overcome if the study would be extended to a larger cohort of patients, such as the National Registry for treatment of arthritis with biologic agents.

In conclusion, our study showed that even in a country with high TB incidence, mandatory screening for LTBI with the current protocols and LTBI treatment with isoniazid leads to a low risk of TB in the first 12 months (probably mostly by preventing LTBI reactivation) after TNFi initiation, with no significant difference between TST with a cut-off of 10 mm and QFT. However, it is necessary to further diminish the risk of TB in patients exposed to TNFi for more than 1 year. This suggests that patients initially negative at TST or QFT might benefit from periodical re-screening with one of these tests, in order to detect new TB while it is not yet clinically manifested.