QuantiFERON TB gold testing for tuberculosis screening in an inflammatory bowel disease cohort in the United States

Although the anti-tumor necrosis factor-alpha (anti-TNF-α) agents (infliximab, adalimumab, certolizumab) have proven to be efficacious in the treatment of inflammatory bowel disease (IBD; Crohn's disease [CD] and ulcerative colitis [UC]),^1,–5 these agents have been associated with rare, but potentially devastating infectious complications.⁶ One of the most important complications of these drugs is the suppression of cell-mediated immunity leading to reactivation of Mycobacterium tuberculosis infection in previously exposed individuals. Studies have shown that these agents can cause up to a 5-fold increase in the reactivation of latent tuberculosis infection (LTBI) within the first 52 weeks after initiation of treatment.⁶ Prior to the initiation of anti-TNF-α therapy, it is recommended that patients undergo screening for LTBI.⁷, ⁸, ⁹ Patients with LTBI are then treated with antituberculous therapy (ATT) to decrease the chances of reactivation of tuberculosis following the initiation of long-term immunosuppression. These patients may also need surveillance of their LTBI on a regular basis after initiation of anti-TNF-α therapy.

In the recent past the only test available for LTBI was the tuberculin skin test (TST). However, TST has many disadvantages, including false-positive results in Bacillus Calmette–Guerin (BCG)-vaccinated patients¹⁰ and in patients exposed to nontuberculous mycobacterium,¹⁰ subjectivity in interpretation of results, and the need for two clinical encounters to administer and interpret the results.¹⁰ In May of 2005 the FDA approved a new test, the QuantiFERON TB Gold test (QFT-G, Cellestis, Carnegie, Australia). This is an interferon-gamma (IFN-γ) release assay (IGRA), where an in vitro white blood sample is stimulated by tuberculosis antigens (early secretory antigenic target 6 [ESAT-6] and culture filtrate protein 10 [CFP-10]), and the corresponding burst in cytokines will help determine whether the patient has been previously exposed. Due to use of antigens not present in M. bovis or other nontuberculous mycobacteria, this test offers higher specificity than TST.¹⁰, ¹¹

The positivity of QFT-G compared to TST has been recently examined in a European IBD cohort with a high rate of BCG vaccination.²⁷ However, data on the utility of these tests from a US population are limited. In the present study we report on the rate of positivity and performance of QFT-G in a US cohort of IBD patients on anti-TNF therapy.

Materials and Methods

Study Population and Variables

We conducted a retrospective, observational, descriptive study at the Inflammatory Bowel Disease Center at Froedtert Memorial Lutheran Hospital, the major teaching hospital of the Medical College of Wisconsin (Milwaukee, WI). We used our previously described IBD database²⁶ to identify all patients who have received or are receiving anti-TNF therapy and are being actively followed by our IBD center (with at least one clinic visit in the past year). For each of the above-identified patients, information on QFT-G and results were obtained from the electronic medical record. In addition to QFT-G results, the records were then searched for TST results and those were documented as well. Information on the use of immunosuppressive therapy use at the time of QFT-G testing was also extracted. Demographic information including age, gender, IBD type (CD or UC), and anti-TNF-α agent used were extracted from the database and medical records.

Tuberculosis Screening Modalities

QuantiFERON-TB Gold Test

All tests were performed at Dynacare Laboratories (Milwaukee, WI). The Centers for Disease Control (CDC) guidelines of December 2005 were used to identify positive, negative, and indeterminate results.⁶ In brief, the concentration of IFN-γ was measured from three wells (negative control or nil well, antigen well containing CFP-10 and ESAT-6, and positive control or mitogen well). The test was read as positive if the concentration of IFN-γ in the antigen well minus that of the nil well was ≥0.35 IU/mL. The results were read as negative if the concentration of IFN-γ in the antigen well minus that of the nil well was <0.35 IU/mL and the concentration in the mitogen well was ≥0.5 IU/mL. The results were read as indeterminate if the concentration of IFN-γ in the antigen well minus that of the nil well was <0.35 IU/mL but the concentration in the mitogen well was <0.5 IU/mL, or if the concentration in the nil was >0.7 IU/mL and the concentration in the antigen well was ≤50% above nil.¹¹

Tuberculin Skin Test

Based on the CDC guidelines,¹² 0.1 mL of tuberculin PPD (Parkdale Pharmaceuticals, Rochester, MI), equivalent of 5 TU, were injected intradermally. More than 10 mm of induration was used to denote a positive result in immunocompetent patients while induration greater than 5 mm was used to identify a positive reaction in patients on immunosuppressive therapy at the time of testing. Immunosuppressive therapy was defined to be the use of any of the following medications: systemic corticosteroids including budesonide, azathioprine (AZA), 6-mercaptopurine (6-MP), or methotrexate.¹²

Study Outcomes

Our primary outcome was a clinical or radiologic recurrence of tuberculosis infection. Symptoms suggesting possible reactivation (including cough, fevers, chills, hemoptysis, or weight loss) were inquired about at each clinic encounter and documented in the medical chart. Radiologic testing was utilized based on the presence of clinical symptoms.

Statistical Analysis

Simple descriptive statistics were used to analyze the data given the exploratory design of the study. Continuous variables are summarized using the mean and/or median where appropriate depending on the distribution of data (normal or skewed). Cohen's kappa was used to calculate correlation between dichotomous variables while weighted kappa was used for tests with more than two variables. Fisher's exact test (2-tailed) was used to analyze associations of categorical data in two independent groups with a P-value <0.05 considered statistically significant. Data were analyzed using StatXact v. 3.0.2 (Cytel Software, Cambridge, MA) statistical software.

This study was approved by the Medical College of Wisconsin's Human Research Review Committee.

Results

Of 542 patients at our center who were considered for or started on anti-TNF therapy, 340 patients were ever tested with QFT-G test. This group formed our study population. Those 340 patients had 518 QFT-G tests performed on them during the review period. The demographic characteristics for the study group are summarized in Table 1. The majority of patients were on concomitant immunosuppressive at the time of testing. Forty-two percent (n = 142) of patients were tested with QFT-G before initiation of anti-TNF therapy, while 58% (n = 198) were tested while on such therapy. Three patients were tested within 2 months of stopping one anti-TNF agent. These patients were considered to be on therapy.

Of the 340 patients who were tested with QFT-G test, five patients (1.5%) had a positive QFT-G test, nine patients (2.7%) had an indeterminate result, and the remaining 326 patients (95.8%) had a negative result. Since their first QFT-G test, these patients were followed up for a mean of 17 (±10) months with one case of TB reactivation for a rate of 0.3%.

In the analysis of QFT-G results by test for the 518 tests submitted, the rate of positivity was 1.2% (n = 6). In 2.5% (n = 13) the test was indeterminate, and the remaining 96.3% (n = 499) were negative.

Effect of Immunosuppressive Therapy on QFT-G

Forty percent (n = 136) of all patients were tested with QFT-G while on immunosuppressive therapy (methotrexate, n = 41; azathioprine, n = 60; mercaptopurine, n = 29; thioguanine, n = 1; mycophenolate mofetil, n = 4; or lenalidomide, n = 1). Of those patients 3% (n = 4) had an indeterminate QFT-G, 97% (n = 132) had a negative QFT-G, and none had a positive QFT-G. None of these patients had evidence of TB reactivation during the follow-up period. Of the remaining 60% (n = 204) of patients who were tested without being on any immunosuppressive therapy, 2.5% (n = 5) had indeterminate QFT-G, 2.5% (n = 5) had positive QFT-G, and 95% (n = 194) had a negative QFT-G. The one patient who had TB reactivation was on immunosuppressive therapy before testing and was on infliximab when he had an indeterminate QFT-G test. There was no significant difference between the two groups in the rates of positive QFT-G (0% versus 2.5%, P = 0.1617) or in the rate of indeterminate QFT-G results (3% versus 2.5%, P = 1).

Effect of Anti-TNF Therapy on QFT-G Testing

Fifty-eight percent (n = 198) of all patients were tested with QFT-G while on anti-TNF therapy. Most of them were tested as part of annual screening for LTBI while on maintenance anti-TNF agent. The remaining 42% (n = 142) of patients were tested before initiation of anti-TNF therapy. The rate of positive, negative, and indeterminate result in each group is summarized in Table 2.

There was no significant difference between the two groups in the rates of positive QFT-G (1.5% versus 1.4%, P = 0.653) or in the rate of indeterminate QFT-G results (1.5% versus 4.2%, P = 0.1723).

Reproducibility

Thirty-nine percent (n = 131) of all patients had more than one QFT-G done (range, 2–6 tests). All of them took place in patients who either had negative QFT-G results and were later tested as part of annual screening, or in patients who had indeterminate results. Among 309 repeated tests the test–retest results were identical in 97.4% (n = 301), while disagreement occurred in 2.6% (n = 8) of the tests. Test reproducibility for the initial two QFT-G testes was calculated for those patients (Table 3) using Cohen's kappa, which was κ = 0.72, P < 0.0001. This indicates a substantial agreement between the two tests. Disagreement took place in four patients. This involved patients who had an indeterminate test that was later found to be negative or vice versa.

Concordance Between TST and QFT-G

Among the 340 patients in the study group, 25% (n = 85) had documentation of TST testing in our medical records. Six patients (7%) tested positive, one patient (1.2%) had an allergic reaction (immediate induration after injection), and the remaining 78 patients (91.8%) tested negative (Table 4).

Of the six patients who had a positive TST, two had a positive QFT-G, three had a negative QFT-G, and one had an indeterminate QFT-G result. One patient had an allergic reaction to TST and tested negative with QFT-G (Table 4).

Of the 78 who tested negative with TST, one had a positive QFT-G and five patients had an indeterminate QFT-G. Moderate concordance between the two tested was detected with κ = 0.4152, P = 0.0041.

Patients with Positive QFT-G

Five patients had a positive QFT-G test, two of whom never tested with TST. Of these two, one was treated for LTBI and had no evidence of TB reactivation at 4 months after initiation of anti-TNF therapy. The second patient tested positive on adalimumab therapy and was referred to an infectious diseases clinic. Of the remaining three patients who had a positive QFT-G and had a TST, two had a positive TST as well (Table 2). One had received treatment several years prior and had no TB reactivation after 1 year of anti-TNF therapy. The second patient was treated with INH and subsequently on certolizumab with no reactivation after 4 months of follow-up.

The remaining one patient with a positive QFT-G (for annual surveillance) had a negative TST (prior to initiation of anti-TNF therapy). She was at low risk for exposure to TB and had a normal chest x-ray. She was referred to ID. No LTBI treatment was started given that the “low degree of positivity on her QFT-G,” negative TST, and low risk for exposure to TB and normal chest x-ray. She continues to be on adalimumab and has been free of any pulmonary symptoms.

Positive TST and Negative QFT-G

Two patients had a positive TST and negative QFT-G. One had a history of being exposed to TB as a child and being treated at that time. QFT-G was found to be negative. The patient was started on adalimumab and has done well for 18 months since. The second patient had a positive TST and was treated with INH and started on adalimumab. Subsequent QFT-G tests have been negative and he had no reactivation during 4 years of follow-up.

Indeterminate QFT-G

As previously mentioned, 2.7% (n = 9) of patients had an indeterminate QFT-G result. Five of those patients had a negative TST and were not treated for LTBI and given low-risk and normal chest radiographs. All patients have been free of any symptoms to suggest TB reactivation so far.

TB Reactivation

As mentioned earlier, the rate of TB reactivation in our cohort over the follow-up period was only 0.3% (one patient). This patient had a history of imprisonment for about 4 years prior to being started on anti-TNF therapy. He had had several TSTs while in jail, which he reported were negative. He was on infliximab before QFT-G testing, which was indeterminate. Nine months after QFT-G testing he presented to the hospital and was found to have miliary TB.

Discussion

Reactivation of latent tuberculosis infection is a major concern in patients on long-term anti-TNF α therapy.⁷, ²⁴, ²⁵ Hence, initial screening as well as annual surveillance testing for LTBI is recommended for patients prior to starting biologic therapy.⁷, ²⁵ We report the first US experience with the use of QFT-G in patients with IBD.

QFT-G is gaining wider approval for use in detecting LTBI and active TB in several patient populations.^13,–16 However, its use in IBD patients who are considering or are currently on anti-TNF-α medications has not been adequately studied. To date, there are no studies about the frequency and the usefulness of QFT-G in IBD patients in the US. The only study regarding the use of IGRA in IBD patients comes from a recently published study by Schoepfer et al.²⁷ They prospectively compared the use of QFT-G in-tube (QFT-G-IT) to TST in cohort of 212 patients (168 with IBD and 44 controls) in Switzerland. They had a high BCG vaccination rate of 71%. Overall, the rate of positive test was 8% in their IBD group. Agreement between the two tests was found to be poor. The study concluded that QFT-G-IT is a better screening tool in immunocompromised IBD patients because of the increased rate of false-negative results in TST in this group.

In theory, if a patient is on immunosuppressive therapy at the time of QFT-G testing, there should be a higher likelihood of an indeterminate test (poor response to the internal control [mitogen well]). In our study, 40% of our patients were on immunosuppressive therapy at the time of their first QFT-G test, while 60% were not on such therapy. No difference was found in the rates of positive or indeterminate results between both groups, which is consistent with findings of similar studies.²⁷, ³¹ However, it was interesting to find that all five patients with positive QGT-G were not on any immunosuppressive therapy at the time they tested positive.

In addition, we reviewed the effect of being on anti-TNF therapy at the time of QFT-G testing, which took place in 58% of all patients. Some studies³¹, ³² have suggested that there is an increased tendency to obtain false-negative results in such patients. There was no statistically significant difference in the rates of positive and indeterminate results between patients who were receiving anti-TNF therapy at the time of testing and those who were tested before initiation of such therapy. However, the only patient who did develop TB was on infliximab prior to QFT-G testing and had an indeterminate result.

In the present retrospective study, we found the rate of positivity for QFT-G in a US cohort of IBD patients to be low (1.3%) when compared to the European study. As a result of QFT-G testing, three out of the five patients who tested positive with QFT-G had treatment for LTBI (two were never tested with TST and one had TST to confirm the results). The rate of indeterminate results, however, was found to be 2.5%, which is close to the 3% found by Schoepfer et al.²⁷ In addition, we reported on the reproducibility of the QFT-G by comparing results from repeat testing. Such a measure has not yet been reported in this population and we found test–retest agreement to be good (κ = 0.72) but not perfect. The most test–retest disagreement was seen in patients with indeterminate results. This measure of reproducibility, however, should be considered with caution, as some studies³¹, ³² suggested increased rates of false-positive IRGAs in patients who are already on anti-TNF therapy.

In a subgroup of our patients (25%), we were able to compare the results from TST and QFT-G. As a blood test, the results of QFT-G testing are easily identified in our electronic records. TST, on the other hand, is harder to identify, as the test was sometimes performed or read by the patients' primary care providers who in many cases were not part of our IBD referral center. We only reported on the patients for which we had clear documentation of TST results in our records. The concordance between the two tests was found to be moderate (κ = 0.4125). When calculating concordance between the two tests, previous studies²⁷, ³⁰ used 2 × 2 tables that included the positive and negative IRGA versus positive and negative TST. This method, however, ignores the significant number of patients who have indeterminate results. When calculating the above concordance, we used a 3 × 3 table (Table 4) that included positive, negative, and indeterminate results for QFT-G and TST (i.e., the patient who had allergic response to TST). Although an indeterminate result in not helpful (you are not sure if the patient has or has not been exposed to TB), the rate of obtaining such a result seems consistent in this population and is a significant number that should be analyzed in the calculation of concordance. Alternatively, kappa was also calculated using the positive and negative results alone and was found to be κ = 0.4747 (P = 0.0097). This is in contrast to the poor concordance preciously reported by Schoepfer et al²⁷ and in agreement with concordance of the two tests in a different patient population in the US.³⁰ The reason for this higher concordance compared to the European study is unknown. We hypothesize that this could be due to the lower prevalence of TB in Wisconsin and the low rate of BCG vaccination in the US.

In a meta-analysis of 58 studies, Menzies et al¹⁰ concluded that the specificity of TST for LTBI was 66% compared to corresponding value of 97% for QFT. In our study there were three patients who had a history of positive TST for which they were treated in the past and had been started on anti-TNF therapy and followed for many years. In later testing, their QFT-G was found to be negative. These patients may represent false-positive TSTs since QFT-G does not diminish after treatment. Another case where QFT-G proved to be helpful was in the patient who had an allergic reaction to TST.

The rate of TB reactivation in our cohort was 0.3%, as there was one patient who did develop miliary TB 7 months testing negative with QFT-G. This patient was initiated on infliximab in 2006 at an outside institution and transferred care to us 2 years later. At that time he was on infliximab and had just finished a month-long prednisone taper for UC flare. QFT-G was negative. He had two follow-up visits in our clinic for the next 7 months and reported good response on the medication with no side effects noted. The patient was admitted to the hospital in January 2009 for cough and diagnosed to have miliary TB. This patient was at high risk for TB, given the history of prolonged incarceration, yet both TST and QFT-G were falsely negative. The potential reason for the negative TST and QFT-G in this patient at the most recent evaluation may have been due to his being on anti-TNF agents at the time of testing. This draws to attention the potential limitation of both these tests in an immunosuppressed population.

In considering the cost-benefit analysis of TST versus QFT-G, there are several variables to be considered. The direct monetary cost for running QFT-G at our laboratory is around $140, compared to $30 for TST. However, it is important to consider the additional, indirect resources involved and work hours lost in needing to take a second trip to a doctor's office to get the TST read. There are also potential monetary costs and health risks for treating false-positive TSTs that otherwise would not be treated if patients were tested with the more specific QFT-G. In addition, a single encounter test such as the QFT-G may have better compliance than tests that require two separate visits.

Our study had several limitations. First, Wisconsin has a low incidence (12/1,000,000 in 2008)²⁸, ²⁹ of active TB. The retrospective design and relatively small sample size of our study limit the power to examine associations as well as make it vulnerable to certain biases. In addition, the lack of a prospectively planned testing of all patients with both modalities limits our ability to compare the performance of the results.

In conclusion, we report on the use of QFT-G in a US IBD cohort on anti-TNF therapy. The rate of positivity was found to be low, at 1.5%, and the rate of indeterminate results was found to be 2.7%. Most patients who tested negative with QFT-G were able to start anti-TNF therapy with no evidence of reactivation. However, one case of TB reactivation took place in a high-risk patient with an indeterminate QFT-G while on infliximab, suggesting cautious interpretation of either TST or QFT-G in an immunosuppressed population. Larger, prospective studies are needed to answer the question of whether QFT-G should replace TST in TB screening for IBD patients before initiating anti-TNF therapy.

References

Author notes