Peripheral blood samples were taken to detect LTBI using QuantiFERON-TB Gold In-Tube assay (QFT) (Celestis, Australia), which was performed according to the manufacturer's instructions [22]. Interferon-gamma level of the post-reaction supernatant was then measured by enzyme-linked immuno-sorbent assay (ELISA) and results were interpreted as positive, negative, or indeterminate accordingly [23], [24]. In this study, LTBI was defined as a positive QFT result.

Compared to HD patients, PD patients were significantly younger, predominantly female, more symptomatic with chronic cough, had shorter length of dialysis duration, had lower serum albumin level, and were less likely to have pulmonary lesions on chest radiograph (18% vs. 37%; p<0.001). The proportion of QFT-positive patients was similar in the two groups (19% vs. 22%, p=0.109).

The QFT result was positive in 91 (21.3%) patients, negative in 316 (74.0%), and indeterminate in the remaining 20 (4.7%) (Table 2). Among the 20 QFT-indeterminate results, 19 were due to weak response to mitogen and one due to strong response of negative control. The QFT tests were repeated in 14 of the 20 patients and were negative in 9, positive in 1, and indeterminate in the remaining 4.

Compared to combined group of QFT- positive and QFT-negative patients, the 20 QFT-indeterminate patients were older and less likely to be ex-smokers, received HD for a longer period, and had more dyspnea and constitutional symptoms (Table 2). They also had lower serum albumin level and a higher proportion of patients with pulmonary lesions. Multivariate logistic regression analysis revealed that dialysis duration (Odds Ratio [O.R.] 1.113, 95% C.I. 1.015–1.221 per year increment), HD (O.R. 10.535, 95% C.I. 1.336–83.093), anemia (O.R. 8.760, 95% C.I. 1.014–75.651), and serum albumin level (O.R. 0.244, 95% C.I. 0.086–0.693 per unit increment) were significant predictors of QFT-indeterminate results (Table 3). If 3.5 g/dl was used as cut-off value for serum albumin level to predict an indeterminate QFT result, sensitivity and specificity were 80% and 93%, respectively.

Patients with LTBI (QFT-positive) were older and more likely had a prior TB history compared to QFT-negative patients, by univariate analysis (Table 2). In multivariate logistic regression analysis, independent predictors of LTBI included age (O.R. 1.034, 95% C.I. 1.013–1.056, per year increment), prior TB history (O.R. 6.467, 95% C.I. 1.985–21.066), and current smoker (O.R. 6.467, 95% C.I. 1.985–21.066) (Table 4; Figure 1). The presence of underlying co-morbidity was not an independent predictor. In patients with prior TB history and those who were current smokers, the prevalence of LTBI was 57.1% and 29.6%, respectively. This prevalence increased by <15% in patients younger than 50 years old to almost 30% in those older than 80 years old.

Though IGRA-positive is not 100% equivalent of LTBI, it has several advantages over tuberculin skin test (TST) in terms of convenience and accuracy [10]–[12]. The TST has a significant limitation in Taiwan due to BCG vaccination [29] and the high prevalence of NTM disease [30]. In QFT, by incorporating positive control (mitogen) and negative control (no antigen) tubes, the true immune reaction against Mycobacterium tuberculosis-specific antigens can be differentiated from false-positive result due to non-specific activation and false-negative result due to immuno-suppression. Thus, IGRA is a better screening test for LTBI than TST while implementing public health policy, especially for an immune-compromised host.

Previous studies using IGRA report an LTBI prevalence of 21–40% in HD patients [13]–[15]. The QFT-positive rate in the present study is within this range and lower than 40%, as reported in a study conducted in south Taiwan [14]. This is probably because the incidence of TB has been decreasing in Taiwan [31]. However, the prevalence in the present study (21.3%) is similar to that reported in household contacts in large-scale studies (11~30%) [12], [32] and much higher than the results of new health-care staff in the study institute at the same period (13 [5.7%] QFT-positive in 229 [unpublished data]). As such, it can be posited that dialysis patients in Taiwan have a much higher prevalence of LTBI than the general population and should be a priority group for targeted screening for active TB disease, especially IGRA-positive patients. If IGRA is unavailable, focus should be on older patients, current smokers, and those with prior TB history. Quite interestingly, the three predictors for LTBI identified in the present study are also risk factors of active TB disease [14], [33], [34]. Different combinations of these predictors may be useful to select the target population for preventive therapy for LTBI. However, the cost and benefit of preventive therapy in this special population should be further evaluated.

The population of chronic renal failure patients receiving long-term dialysis is increasing worldwide and TB is a commonly associated infectious disease [18], [19]. It has been previously assumed that because HD patients frequently visit the HD room, they are more likely to acquire Mycobacterium tuberculosis infection than PD patients via airborne transmission. Only a report of PD patients in Spain has shown a comparable LTBI prevalence of 18% [35]. By simultaneously enrolling HD and PD patients, this is the first study to demonstrate similar LTBI prevalence in the two patient groups, thereby challenging the hypothesis of occult transmission in the HD room. Although the two dialysis groups are different in more ways than just dialysis place and duration, this observation suggests that transmission of TB to HD patients within crowded dialysis facilities may be similar to PD patients at home [36]. This suggests that the study institute has an effective TB infection control policy on early detection, prompt treatment, and rapid isolation.

Although 71.4% of QFT-indeterminate patients have definite results after repeat testing, 4.7% of the initial QFT tests with an indeterminate result are associated with hypoalbuminemia, anemia, HD, and longer dialysis duration. The association between anemia and indeterminate status has been shown before [37]. Along with hypoalbuminemia, these predictors suggest that malnutrition attenuates immune response and compromises the performance of IGRA [38]. The current finding that HD, but not PD, is associated with QFT-indeterminate is interesting and worth discussing further. A previous study reveals that while both HD and PD patients have lower but insignificant HLD-DR expression on peripheral blood monocytes compared to healthy controls, HLA-DR expression is significantly higher in PD than in HD patients [39]. This implies that continuous dialysis like PD can attenuate immune dysfunction compared to intermittent modes like HD. Moreover, longer duration of dialysis in dialysis patients has been correlated with worse cellular immunity [40]. This may explain how a much higher percentage (95%) of QFT-indeterminate results in the present study may come from low mitogen responses, compared to 51% in a public health clinic setting [24]. For diagnosing LTBI in such patients, IGRA should be meticulously applied. Repeating IGRA or using alternative tests may be necessary.

In contrast to a previous report [24], female gender is not an independent factor of QFT-indeterminate results in the present study. This may be due to different patient characteristics, such as age, race, and prevalence of HIV infection, between studies. Further large-scale investigations are necessary to confirm this finding and investigate possible reasons.

The present study has several limitations. First, this study was conducted in a tertiary referral center and its branch, so patients had more underlying co-morbidities and the LTBI prevalence might be higher. Second, without detailed contact investigation, the epidemiologic link and biological implication of QFT-positivity cannot be confirmed. Lastly, this is a cross-sectional study. Further prospective studies with long-term follow-up on the development of active TB are needed.

In conclusion, patients receiving PD have a similar prevalence of LTBI as those receiving HD (19% and 22%, respectively). The prevalence of LTBI in long-term dialysis patients is even higher in the elderly, current smokers, and those with prior TB history. These risk factors can be used to select a target group for cost-effective LTBI screening. Patients receiving HD or long duration of dialysis, and those with anemia or lower albumin level are likely to have a QFT-indeterminate result. For such patients, repeat IGRA or alternative test may be necessary to detect LTBI.