Patients and methods

Patients

Successive patients attending the Department of Rheumatology, St Vincent’s University Hospital, Dublin, who were being considered for anti-TNFα therapy were subjected to a predefined screening protocol. The protocol was approved by the institutional ethics and research committee. A detailed clinical history and physical examination was performed on all patients by one of two investigators (JM or AG). The diagnosis of an inflammatory arthritis was based on standard criteria.43 44 45

LTBI screening

The clinical history format was designed to acquire information about possible risk factors associated with previous exposure to TB infection.10 11 Patients were considered to be at moderate risk of LTBI if they gave a history of occupational exposure to TB infection, birth or extended living in an area with a high prevalence of TB infection, or a history of incarceration, homelessness or intravenous drug use. Patients were considered to be at high risk of LTBI if they had previous close contact with TB infection, or a previous diagnosis of TB or LTBI which had been inadequately treated.

A TST was performed on all patients as previously recommended.13 Tuberculin PPD 2TU/0.1 ml was injected intracutaneously into the volar aspect of the forearm. The degree of induration was measured 48–72 h after the injection and the diameter was recorded. In this cohort of patients with active inflammatory arthritis, a diameter >5 mm induration was regarded as a positive test.13

All patients underwent a posteroanterior CXR. Every CXR was reviewed by a staff radiologist who was aware of the rheumatological diagnosis and that anti-TNFα therapy was being considered. Features that were regarded as suggestive of previous TB infection included apical pleural thickening, pulmonary nodules, upper lobe bronchiectasis, interstitial granulomatous calcification, cavitation and lymph node or pericardial calcification.13 The presence of two or more of the above features, or the presence of coarse upper zone fibrosis, represented probable LTBI.

Interferon γ release assays

Assays were performed according to the manufacturers’ instructions (see supplementary online data).

Statistical analysis

Data were recorded in Excel and coded according to conventional coding systems. Patients were divided into two groups based on their response to the T-SPOT.TB test as positive or negative responders. Associations between study variables and T-SPOT.TB were assessed as follows: for continuous variables (eg, age, duration of disease, etc) the statistical tests used were the two-sample t test or Wilcoxon rank sum test, as appropriate. The relationship of IGRA test outcome with categorical variables was analysed using the χ² test or Fisher exact test, as appropriate. Multivariate analysis was carried out using logistic regression with T-SPOT.TB as the dependent variable. Multivariate analysis of the QuantiFERON-TB Gold results was not undertaken owing to the relatively small sample sizes. Percentage agreement between T-SPOT.TB and conventional screening methods for detection of LTBI, and also between the two IGRAs, was calculated and the κ statistic generated. Statistical significance was deemed to be at p<0.05.

Results

Patients

A total 150 successive patients with chronic inflammatory arthritis being considered for anti-TNFα therapy were included in the study. Ninety-one (60.7%) were female. The mean age (range) was 50.1 (17–88) years. The mean disease duration (range) was 11.6 (0.2–60) years. The majority (80.5%), including all the patients with RA, had previously received at least one disease-modifying antirheumatic drug, mostly methotrexate. Eighty-seven patients (58%) had RA, 37 (24.7%) had psoriatic arthritis, 22 (14.7%) had ankylosing spondylitis, two (1.3%) had juvenile idiopathic arthritis and two (1.3%) had other diagnoses. No patient had previously received an anti-TNFα agent or other biological treatment. At least 82% of patients had been inoculated with BCG, 8.3% had not and 9.7% were uncertain. No patient had a prior history of active TB or LTBI.

Conventional LTBI screening

Table 1 gives the results of conventional LTBI screening. The CXR appearances that were suggestive of LTBI in 25 patients included 20 with calcified interstitial granulomata, 10 with apical pleural thickening, five with lymph node calcification, and one with cavitation. The appearances that represented likely LTBI in four patients included three with two or more of the above features and one with coarse upper zone fibrosis. The baseline CXR could not be retrieved in one patient.

A total 57 patients (38%) had at least one abnormal screening test. Based on the treating doctor’s interpretation of the combined results, patients were advised to proceed with or to defer anti-TNFα therapy, or to start a prophylactic course of isoniazid. A total of 136 patients proceeded with anti-TNFα therapy, 27 after starting isoniazid, while anti-TNFα therapy was withheld in 14. No patient developed TB a median 38 months after starting anti-TNFα therapy.

Interferon γ release assays

For details of interferon γ release assays, see online supplementary file.

Conventional LTBI screening and IGRAs

An analysis of the associations between the IGRA results and selected clinical features, concomitant methotrexate and corticosteroid therapy and conventional LTBI screening was undertaken (tables 2 and 3). Patients were divided into two diagnostic categories: those with RA and those with non-RA inflammatory arthritis. In addition, a moderate- or high-risk history, a TST >5 mm and a CXR that was reported as suggestive or likely LTBI were regarded as a positive risk for LTBI by conventional screening. The mean ages (SD) of patients with positive and negative T-SPOT.TB tests were 58.7 (12.2) and 48.5 (13.8) years, respectively (p = 0.020 by univariate analysis, p = 0.026 by multivariate analysis) (table 2). In the multivariate model, the adjusted odds ratio (OR) and 95% confidence intervals (95% CI) were 1.1 (1.0 to 1.2). There were no statistically significant associations between the T-SPOT.TB result and disease duration, gender, diagnosis or concomitant methotrexate or corticosteroid treatment. Significant associations between T-SPOT.TB results and conventional screening tests were observed using univariate (TST and risk history) and multivariate (risk history) analysis. The OR (95% CI) values were 3.2 (0.5 to 21.0) for TST and 6.6 (1.1 to 40.0) for risk history. Conversely, there were no significant associations between the T-SPOT.TB results and CXR findings.

The mean ages (SD) of patients with positive and negative QuantiFERON-TB Gold tests were 58.7 (12.2) and 48.5 (13.8) years, respectively (p = 0.002) (table 3). After univariate analysis, no statistically significant associations between the QuantiFERON-TB result and disease duration, gender, diagnosis or concomitant methotrexate or corticosteroid treatment were seen. Statistically significant associations between the QuantiFERON-TB Gold results and both TST (p = 0.002) and risk history (p = 0.02), but not CXR, were confirmed. Multivariate analysis was not performed owing to the small sample sizes.

Further evaluation of associations between individual conventional screening modalities and T-SPOT.TB results was undertaken and the results are given in table 3 in the online supplementary file.

Discussion

This study evaluated two commercial IGRAs in patients with chronic inflammatory arthritis who were undergoing conventional screening for LTBI before starting anti-TNFα therapy. Less than 10% of the patients in this study had a positive IGRA result, which is consistent with previous reports.46 47 In contrast, a positive TST >5 mm was seen in 18% of patients, 20% had a positive risk history, and 20% had a CXR that was consistent with previous TB infection. In all, 38% of patients had at least one positive conventional risk factor for LTBI. Statistically significant associations were seen between each IGRA and both TST and risk history. Interestingly, in multivariate analysis the association between the T-SPOT.TB assay and risk history was shown to be independent of the other study variables. Owing to the relatively small sample sizes, multivariate analysis of the QuantiFERON-TB Gold results was not appropriate. This study suggested that a positive IGRA result was significantly associated with increased age.

In contrast to this study, a low level of agreement between TST and the QuantiFERON-TB Gold assay was seen in a recent study of patients with inflammatory rheumatic diseases.42 That study differed in a number of important respects. First, 60% of patients were receiving a TNFα inhibitor at the time of testing, which significantly reduced the likelihood of a positive IGRA. Second, the TST was not performed in 19% of the study patients. Third, the TST and the IGRA were performed at different time points, so that matched test results in individual patients may have been differentially influenced by changing therapeutic combinations. Observations from other studies have been inconclusive, possibly owing to different prevalence rates of TB infection and to variations in study design and data analysis.46 47 48 49

There was a high level of agreement between the T-SPOT.TB and the QuantiFERON-TB Gold assays. This observation is consistent with Lee et al,31 who compared TST with the two commercial IGRAs in a cohort of patients with active TB infection and a group with a low risk for TB. They reported that T-SPOT.TB had significantly greater sensitivity than TST and QuantiFERON-TB Gold and that both T-SPOT.TB and QuantiFERON Gold demonstrated superior specificity to TST. An analysis of the sensitivity and specificity of the IGRAs in this study was not possible as none of the patients developed reactivation of TB infection. This is not surprising as the prevalence of tuberculosis in the Irish population is 11 per 100 000. The prevalence of LTBI is not known. Arend et al38 compared the utility of TST, T-SPOT.TB and an in-tube modification of QuantiFERON-TB Gold in large-scale contact tracing. Interassay agreement was 89.6%, and the sensitivity of the T-SPOT.TB assay was found to be better than the QuantiFERON-TB Gold test.

The TST was considered positive in this study cohort if a reaction of >5 mm was seen. In 22/27 patients (81.5%) with a positive TST, the reaction was >10 mm. The 5 mm cut-off point was considered appropriate in this population with chronic illness, many of whom were receiving immunosuppressive drugs at the time of testing.13 50 51 A large majority of patients had received BCG vaccination in childhood, which can complicate interpretation of the TST result as TST may have low specificity after vaccination with BCG.13 A statistically significant correlation between the IGRA and the TST was seen in this study, which is consistent with other reports.34 37 41 52 53 Nevertheless, several patients provided discordant results. Seventy-six per cent of patients with a positive TST had a negative IGRA, consistent with the suggestion that the IGRA has greater specificity for recently acquired infection.37 40 5 54

A similar proportion of discordant results were reported in a Dutch study of contact tracing in an unvaccinated population.38 In that study a positive TST, but not IGRA, was associated with older age and, arguably, remote exposure. In some instances, this discordance may be the result of HLA-associated inability to respond to CFP-10 and ESAT-6.55 In this study, 6.8% of patients with a negative TST had a positive IGRA, which is more difficult to explain. This pattern of discordance has also been previously reported and was associated with indicators of recent exposure.27 35 37 56 As no patient developed TB after starting anti-TNFα treatment, it is unlikely that recent exposure to M tuberculosis is the sole explanation for these discordant results, unless some patients receiving anti-TNFα treatment retain the capacity to suppress TB infection for an extended period despite TNFα inhibition.

CXR is recommended before anti-TNFα therapy primarily to exclude active TB.57 Most patients with LTBI have a normal CXR.58 The most common radiographic abnormalities that suggested possible LTBI in this study were calcified interstitial granulomata, lymph node calcification and apical pleural thickening. These features, which were almost invariably associated with a negative risk history and negative TST, were previously reported to carry a low risk of progressing to active TB.59 60 This conclusion is supported by the IGRA results reported here. In contrast, non-calcified nodular and fibrotic lesions suggestive of previous TB, which were uncommon in this study, should be given higher priority for prophylactic treatment of LTBI, particularly if associated with a positive TST61

This study provides compelling evidence to suggest that IGRAs may have utility when screening for LTBI before anti-TNFα therapy in patients with immune-mediated inflammatory diseases. The study was completed in an area of low TB incidence. Therefore, it may not be appropriate to apply the findings to areas of intermediate or high incidence where a greater number of positive conventional screening tests and positive IGRAs might be seen. The study did not permit a determination of the sensitivity and specificity of IGRAs. These critical analyses will require further large-scale prospective studies that include cases of proven TB. The observations reported here also highlight the inadequate performance of CXR as a marker of LTBI. CXR should be retained as a modality of pre-anti-TNFα screening solely for the purpose of excluding active TB.