Indeterminate results of a tuberculosis-specific interferon-γ release assay in immunocompromised patients

To the Editors:

Immunocompromised patients with various causes and degrees of immunodeficiencies, such as stem cell and solid organ transplant recipients, patients with autoimmune diseases, patients with chronic renal failure or HIV-positive patients, are at increased risk of progression from latent Mycobacterium tuberculosis infection to active disease. Therefore, screening for latent tuberculosis and preventive treatment is recommended in this patient population. Tuberculosis-specific interferon (IFN)-γ release assays (TIGRAs) lacking cross-reactivity with Mycobacterium bovis bacille Calmette-Guérin have been introduced into the routine diagnosis of latent tuberculosis infection (LTBI) in the last few years as a more specific alternative to tuberculin skin test (TST). More recent results implicate that QuantiFERON®-TB Gold in tube (QFT-G-IT; Cellestis, Carnegie, Australia) may better predict progression from latent to active disease compared with TST 1, 2.

TIGRAs using stimulation of T-cells with phytohemagglutinin (PHA) as a positive control for identification of false negatives and the classification of indeterminate results may be a better alternative for the prediction of LTBI in immunocompromised patients where IFN-γ release may be affected by immunosuppression. The growing list of data existing on reliability of TIGRAs in immunocompromised patients show that the prevalence of indeterminate results may vary depending on the degree of immunosuppression and the TIGRA test used 3. In addition, PHA and recall antigens use different IFN-γ secretion pathways 4, which may be differentially affected by immunosuppressive conditions. Thus, an in-depth analysis of factors influencing PHA-associated IFN-γ secretion is important for the assessment of TIGRAs in immunocompromised patients. Since disease group dependent and independent risk factors of indeterminate results in immunocompromised patients have not been evaluated prospectively using the third generation QFT-G-IT test, we tested QFT-G-IT in patients with diverse conditions of immunosuppression to determine the rate of and identify risk factors for indeterminate results.

After approval of the ethical committee of the University of Freiburg (Freiburg, Germany), outpatient or hospitalised immunocompromised patients and 104 healthy students and healthcare workers (controls) were included prospectively (October 2006 to October 2007) in a single centre in Germany. Patients (aged >18 yrs, mean age 64.5 yrs) had either undergone organ transplantation or stem cell transplantation (39 allogeneous, four autologous), had an autoimmune disease (39% rheumatoid arthritis, 12% systemic lupus erythematosus (SLE), 5% chronic inflammatory bowel disease, 44% other autoimmune diseases), had been receiving immunosuppressive therapy for at least 6 months (glucocorticoids equivalent to ≥10 mg prednisolone, cyclosporin, tacrolimus, methotrexate or mycophenolate), and had primary immunodeficiencies (82% with chronic variable immunodeficiency syndrome) or HIV-infection (mean CD4 count: 435 cells·µL⁻¹; 42 patients with a CD4 count >200 cells·µL⁻¹; 33 with HAART therapy; 33 were male). Of 456 screened patients, nine were excluded for not meeting inclusion criteria or not enough data to define length or dosage of immunosuppressive therapy, two patients (one stem cell transplant and one with autoimmune disease) had active tuberculosis (table 1⇓). Prospectively planned data collection included demographic data, medical history, medication and laboratory values.

QFT-G-IT was performed in one laboratory following manufacturers’ instructions after in-house transport at room temperature. After vigorous resuspension of the antigen, test tubes were incubated at 37°C 2–6 h after collection and ELISA was performed after 16–20 h of incubation. Indeterminate results were defined as PHA-associated IFN-γ secretion of <0.5 IU·mL⁻¹ or negative control values >8 IU·mL⁻¹. Lymphocyte counts and haemoglobin levels were performed in 68% and 95% of patients, respectively, with a similar percentage in different patient groups. The Chi-square based measure Cramer V was used to determine relationships of categorical variables. For multivariate analysis a backward stepwise logistic regression was performed for variables with more than 25 cases per category testing disease group, sex, age, immunosuppressive medication, blood values (haemoglobin, erythrocyte count and lymphocyte count), and risk factors for immunosuppression (diabetes, smoking, current dialysis). To compare means of numeric variables we used an unpaired t-test.

In our cohort of immunocompromised patients positive QFT-G-IT results were more common than in controls (8.3% (95% CI 5.7–10.7%) versus 1.9% (95% CI 2.3–6.8%); p = 0.060) and more common in males than in females (10.5% (95% CI 7.2–15.4%) versus 5.3% (95% CI 2.7–10%); p = 0.056). In a logistic backward regression, a positive QF-GT-IT result was significantly influenced by immigration status and old age (p<0.05) reflecting known risk factors of LTBI in Germany.

We focused our analysis on the reliability and risk factors of disease group dependent and independent indeterminate results in immunocompromised patients (table 1⇑). As expected, the overall rate of indeterminate result in immunocompromised patients was significantly higher (13%, 95% CI 9.9–16.1%) than in the control group (0.9%, 95% CI 0–2.75%; p<0,001), and similar to previously published results. However, further comparability is limited due to the use of the second generation QFT-Gold test or due to differences in patient cohorts 3, 5.

Overall indeterminate results differed significantly depending on the disease group. Patients with autoimmune diseases had a significantly higher percentage of indeterminate results than organ transplant patients (p<0.05), which may be due to a previously described decreased IFN-γ response to different antigens in patients with rheumatoid arthritis 6. In a subgroup of patients with SLE, seven out of 12 patients had an indeterminate result (univariate p<0.001) which may be related to an impaired in vitro response to PHA in patients with active SLE 7. The overall indeterminate result rate was 9.6% if results of stem cell transplant patients were omitted. The high indeterminate result rate in stem cell transplant patients is not surprising as in vitro a prolonged impairment of IFN-γ production has been shown in patients with autologous stem cell transplantation 8, and most of our patients have been studied within their first year after transplantation. Thus, the use of TIGRA prior to immunosuppressive therapy is especially important in this patient group, not only because of the high percentage of indeterminate results, but since the donor immune system will be studied after stem cell transplantation.

Indeterminate result rates were higher in females than males (univariate p<0.05; not significant in a multivariate analysis comprising medication, haemoglobin and lymphocyte count), a finding that has not been reported in other studies evaluating TIGRAs in immunocompromised patients. The sex-difference was most pronounced in patients with autoimmune diseases (29% in females versus 0% in males; p<0.05), and it was also significant in multivariate analysis. A nonsignificant sex influence favouring indeterminate results in female over male patients was also found in other patient groups. Several in vitro results (in young females, females and female deer) showed a lower IFN-γ response towards PHA stimulation or towards a multi-test cellular-mediated immunity skin test 9. Therefore, it may well be that TSTs are more frequently false negative in females than in males and that data based on the skin test showing that females are less likely to have latent infection with M. tuberculosis must be interpreted more cautiously. Future contact tracing studies combining skin tests and TIGRAs should address this issue prospectively.

Low haemoglobin levels, as well as a low lymphocyte count, that were not correlated (p = 0.179) correlated significantly with low IFN-γ secretion in univariate analysis (p<0.001). In a backward logistic regression analysis a low haemoglobin level as well as a low lymphocyte count were significant factors for explaining variance in QFT-G-IT results (p<0.001, Nagelkerke R = 0.29). In a linear regression model with the same independent variables and the dependent variable “PHA-induced IFN-γ secretion”, both lymphocyte count and haemoglobin level explained IFN-γ secretion variance significantly (p<0.001, R² = 0.164) (table 1⇑). A correlation between lymphocyte count and IFN-γ response is expected and is similar to the results published by Kobashi et al. 5. The reason for the surprising and previously unreported finding of a correlation between low erythrocyte count or haemoglobin level and indeterminate results in QFT-G-IT is unclear and not well explained by the influence of red blood cells on T-cell function, or of erythropoietin on T-lymphocytes and pro-inflammatory cytokine response 10. If anaemia can be expected, QFT-G-IT tests should be used in immunocompromised patients before its development. Further studies are needed to evaluate the effect of low erythrocyte count/haemoglobin levels on the PHA-associated IFN-γ response in TIGRAs.

In a subgroup of 67 patients (HIV patients and patients with primary immunodeficiencies) where a differential lymphocyte count was available, a strong positive correlation of both CD3+ T-lymphocytes and CD3+CD8+ with the IFN-γ response was found. Similar to previous reports of indeterminate results in patients with HIV infections, in our study no indeterminate results occurred in HIV patients with a CD4-count >200 cells·μL⁻¹. However, indeterminate results did not correlate with CD3+CD4+-T-lymphocytes in this specific subgroup of patients possibly due to the low number of patients, which was a limitation of our study. In this selective patient sub-cohort a low CD8 count was the best predictor of low IFN-γ response after PHA stimulation.

In conclusion, different disease groups bear an independent risk of indeterminate results in the QFT-G-IT. It is prudent to perform TIGRAs for the diagnosis of latent tuberculosis as soon as the need for therapeutic immunosuppression is evident. Prospective studies may be able to confirm that low lymphocyte, CD8 T-cell and haemoglobin levels are better predictors of indeterminate QFT results than disease group or immunosuppressive medication and confirm the surprising finding of female sex being a risk factor for indeterminate results.

Statement of interest

A statement of interest for D. Wagner can be found at www.erj.ersjournals.com/misc/statements.dtl