Introduction

Tuberculosis (TB) remains a major public health problem particularly in low/middle-income countries (LMIC). More than 80% of the global TB burden and TB-related deaths were reported from 22 high-burden LMIC in 2011 [1]. With the current modest annual decline in TB incidence (2%), many countries will be unable to achieve the Stop TB Partnership goal of halving TB incidence by 2015 [2]. Ambitious targets were recently set by the Partnership for the post-2015 era [3]. The spread of multidrug resistant TB (MDR-TB), the detrimental convergence with HIV infection and the unavailability of rapid diagnostic tools have contributed to the failure of global TB control [4–6]. Moreover, through altering the clinical and bacteriological presentations, HIV infection has contributed to the delay in diagnosing active TB disease, leading to increased morbidity and mortality and enhancement of transmission within communities [7–9]. These problems are further magnified in criminal justice settings where TB is a major contributing factor to the disproportionately high morbidity and mortality among prisoners [10,11]. Globally, poor prison living conditions, overcrowding, the concentration of people at high TB risk and the limited access to health services inside these congregate settings convert prisons into reservoirs fuelling the TB and MDR-TB epidemics into surrounding communities [12]. A systematic review confirmed the contribution of exposure in prisons to the population’s TB burden and showed that attributed fractions were 8.5% and 6.3% of overall TB cases in high- and low/middle-income countries, respectively [13]. Delay in TB diagnosis, due to limited screening procedures, inaccuracy of diagnostic algorithms and lack of adequate laboratory facilities, is recognized as a major limitation for TB control efforts in correctional settings [8,14–16]. This contributes not only to increased unfavorable disease outcomes but also to onward disease transmission to other inmates [9,16–20]. The current World Health Organization (WHO) symptom-based screening algorithm for prisons has proven to be diagnostically inadequate [21], therefore, in order to achieve effective TB control, the currently-recommended passive case detection needs to be coupled with an active case finding program, particularly in congregate settings and especially among HIV-infected individuals [7,22].

GeneXpert MTB/RIF (Xpert), a new nucleic acid amplification technology, offers rapid and accurate diagnostic results from biological specimens with minimal staff training requirements [20,23,24]. The analysis of this new technology as a point-of-treatment diagnostic tool in five distinct LMIC sites showed high sensitivity and specificity for the diagnosis of active TB (97% and 99.2%, respectively) and the detection of rifampicin (RIF) resistance (97.6% and 98.1%, respectively), the cornerstone of TB treatment regimens [20]. A recent review of published evaluation reports confirmed that the use of Xpert as an initial test replacing smear microscopy was highly accurate [23]. In 2010 and with the encouraging initial evaluation reports, the WHO endorsed Xpert to be used as the initial TB diagnostic test for individuals with HIV-infection or those suspected of having MDR-TB [25].

Though a dynamic transmission model predicted that the deployment of Xpert in annual TB screening surveys effectively reduces overall TB and MDR-TB prevalence in prisons of Former Soviet Union (FSU) countries, the utilization of this new technology in intensified case finding and its impact on TB burden in prisons of LMIC and where high prevalence of HIV exists has not yet been examined [26,27]. We aimed to estimate the diagnostic accuracy of a single Xpert assay and to assess correlates of active TB disease among HIV-infected inmates of a Malaysian prison.

Methods

Results

As of October 2012, there were 149 prisoners housed in the prison’s dedicated HIV units. Of these, 125 participants had complete data for analysis. Reasons for exclusion were (Figure 1): release within 24 hours (N=4); active TB cases currently taking anti-TB medications (N=9); considered too dangerous by prison authorities to transport (N=2); death (N=3); and negative HIV confirmatory test (N=6). Characteristics of the final sample (N=125) are summarized in Table S1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. The recruitment and findings flowchart.

https://doi.org/10.1371/journal.pone.0073717.g001

Despite having a high risk for TB, only one-third (34.4%) of subjects reported ever being screened for TB before while 36 (28.8%) subjects reported having previously been treated for active TB, mostly (75%) within the past five years. More than half (53.6%) of those surveyed were eligible (CD4 count ≤350 cells/µL) for antiretroviral therapy (ART), but only a total of 19 (15.2%) were prescribed ART at the time of the survey. The median CD4 lymphocyte count was 337 cells/µL (IQR 149-492) and the median HIV viral load was 28,042 copies/mL (IQR 1,138-73,566.50).

Two-thirds of participants (68%) reported having at least one of the four symptoms recommended by the WHO for clinical screening algorithm among HIV-infected individuals and 10 (8%) had ≥5 points using the WHO clinical scoring system [22,33]. Neither of the two clinical screening algorithms was associated with active TB disease in this sample (OR 1.34, p=0.64 and OR 0.80, p=0.84, respectively). Median BMI was 22.3 kg/m² (SD 3.25) and being underweight (BMI <18.5 Kg/m²) was not associated with having active TB disease (OR 1.96, p=0.43)

The study revealed a very high prevalence of undiagnosed active pulmonary TB (PTB) disease (12%, 15 out of 125 screened subjects) and overall TB prevalence when current active TB cases are added (16.7%, 24 out of 143 total HIV-infected prisoners). With the use of conventional fluorescence microscopy at the reference laboratory, all but one of the active TB cases had negative smear examination from the two sputum specimens. The single Xpert assay examination detected Mycobacterium tuberculosis accurately in sputa in eight (53.3%) of the culture-confirmed active TB cases. The assay however missed the diagnosis of seven active TB cases confirmed later by the liquid culture giving a sensitivity, specificity, PPV and NPV of 53.3% (95% CI 30.1-75.2%), 100% (95% CI 96.6-100%), 100% (95% CI 67.6-100%) and 94.0% (95% CI 88.2-97.1%), respectively (Table 1). Xpert was negative in a single case of Mycobacterium scrofulaceum detected subsequently by culture. Overall, the active case finding using Xpert increased TB case detection by one-fold over the current, passive, case detection (eight and nine active TB cases, respectively). Xpert-negative/ culture-positive subjects were more likely to have CD4 lymphocyte count more than 100 cells/µL (p=0.01) compared to Xpert-positive/culture-positive cases (Table 2). No rifampicin resistance was reported by Xpert in this sample, which was confirmed by the reference-laboratory mycobacteriology DST, giving 100% specificity for rifampicin resistance detection.

On univariate analysis (see Table S1), a number of unadjusted covariates significant at p<0.2 including age (OR 1.12, 95% CI 1.02-1.22, p=0.02), living in a crowded house before incarceration (OR 2.23, 95% CI 0.74-6.67, p=0.15), drug use duration (OR 1.14, 95% CI 1.03-1.26, p=0.01), recent alcohol use (OR 2.56, 95% CI 0.68-9.62, p=0.16), years of smoking cigarettes (OR 1.08, 95% CI 1.00-1.17, p=0.04), CD4 lymphocyte count ≤100 cells/µL (OR 3.17, 95% CI 0.95-10.55, p=0.06) and HIV-1 RNA of more than 100,000 (OR 3.04, 95% CI 0.91-10.20, p=0.07). Given that age, duration of drug use and years of cigarette smoking were identified as being collinear, separate multivariate logistic regression models using goodness-of-fit showed that the best model associated with active pulmonary TB was duration of drug use (adjusted OR 1.14, 95% CI 1.03-1.26, p=0.01 for every year of drug use). The other covariates no longer remained significant after controlling for these variables (Table S1).

Discussion

In this intensified TB case-finding survey, we report for the first time, the diagnostic performance of a single GeneXpert MTB/RIF assay in improving pulmonary TB case detection among HIV-infected inmates of a middle-income country prison. Overall, undiagnosed active TB disease was remarkably high (12%) among this HIV-infected inmate population housed in dedicated HIV units. With a very high sputum AFB smear-negativity, a single Xpert had a limited sensitivity for the diagnosis of active TB disease (53.3%) but high specificity for active TB disease (100%) and rifampicin resistance (100%) detection compared to the liquid culture. The new convenient assay markedly increased the number of active TB cases detected in this population, potentially at much earlier TB disease burden, where profoundly symptomatic disease is common at the time of clinical presentation.

Importantly in this study, the single Xpert impressively outperformed the fluorescence smear microscopy examination. This confirms findings from earlier evaluation studies conducted in geographically distinct settings and corroborates the WHO guidance on replacing smear microscopy with Xpert as the primary diagnostic tool for active TB among HIV infected individuals [20,25,36]. Similar utilization of a single Xpert assay in an active case finding among household contacts with unknown HIV status in Tanzania had 40% higher yield than smear microscopy examination [37].

We postulated that the low sensitivity reported in this sample might be related to several factors. Cumulative data from different settings confirmed the impact of smear-negative microscopy status on the Xpert accuracy, which can be improved by increasing the number of analyzed sputa [20,23]. This observation is probably related to the low bacillary load given that Xpert requires a minimal threshold of 131 bacilli/mL (compared to 10-100 bacilli/mL for culture) [38]. Lawn SD, et al reported a similarly low (43.4%) Xpert sensitivity from a TB screening before ART initiation with prevalent smear negativity in South Africa, which improved modestly when a second sample was analyzed (62.3%) [39]. In HIV settings with considerable smear-negative TB disease, a single Xpert analysis is often deployed due to resource constraints.

HIV infection, on the other hand, complicates the diagnosis of active TB through reducing the accuracy of available diagnostic tools [40]. A diagnostic evaluation from a predominantly HIV-infected community in Tanzania showed that the vast majority (88.9%) of smear-negative TB cases were among PLWHA [41]. While HIV infection limits the sensitivity of smear microscopy in diverse settings, irrespective of the immunological status [42–44], the impact of HIV infection on Xpert diagnostic accuracy is yet to be established. In 5 LMIC sites, the overall sensitivity of Xpert was modestly lower among HIV-infected than HIV-uninfected (93.9% vs 98.4%; p=0.02) individuals [20]. A similar reduced trend in the diagnostic utility of Xpert among 496 archived sputum samples from TB suspects was found in South Africa among HIV-infected and HIV-uninfected (69.6% vs 82.9%; p=0.09) patients with significantly lower NPV (78.3% vs 90.3%; p<0.01) for HIV-infected and non-infected patients, respectively [36]. On the other hand, two clinical evaluation studies among symptomatic TB suspects in Peru [45] and Tanzania [41] reported no differences in the diagnostic accuracy of Xpert among HIV-infected and non-infected patients. A recent review of related evaluation studies revealed heterogeneity in reported accuracies of Xpert among PLWHA (70-100%) but the pooled sensitivity of Xpert was generally lower among PLWHA (76%) than HIV-uninfected subjects (89%) [23]. This heterogeneity was partially explained by the high prevalence of smear negativity among PLWHA.

Finally, intensified case finding (ICF) programs support identification of earlier TB cases, especially with less advanced TB disease and low bacterial burden, but potentially insufficient to be detected using the Xpert assay. One third of culture-confirmed active TB cases from an active pre-ART screening in South Africa were false-negative using Xpert and this phenomenon was associated with smear microscopy negativity, prolonged duration to culture positivity, less likely to have chronic cough and with less extensive radiological abnormalities [46]. Moreover, Xpert-negative active TB cases had less advanced HIV infection (higher CD4 count, lower HIV-1 RNA levels and higher BMI). Similarly, Xpert-negative cases in our report were more likely to have CD4 cell count more than 100 cells/µL, but despite having longer median time to culture positive (28 days compared to 21 days of Xpert-positive cases), the difference was not statistically significant (p=0.28). To examine this issue further, an analysis of the impact and cost of alternative diagnostic algorithms in a high TB/HIV co-infection setting projected that the addition of a second Xpert test (Xpert/Xpert) performs similarly but less expensively than the addition of culture (Xpert/culture) [47], while chest radiography performed well as a ruling-out, rather than rule-in, tool among Xpert-negative TB cases [48]. Empiric data supporting these assertions, however, are lacking.

Moreover, the finding of the low reported NVP of Xpert in this sample conforms the earlier studies of using Xpert as a rule-in rather than rule-out tool in the diagnosis of active TB among PLWHA [36]. On the other hand, Xpert was highly specific for detecting M tuberculosis and rifampicin resistance in this sample with no drug resistant case reported.

The deployment of Xpert in an intensified case finding program may have an immediate and sustained impact on TB epidemiology, but the finding that Xpert performed less than ideal among HIV-infected inmates suggests the need to explore combined diagnostic strategies [48–50]. In Brazil, a similar socioeconomic country, symptom-based screening was unreliable and regular mass radiographic screening was concluded to have more pronounced impact on the prevalence of active TB among prisoners [51,52]. There is a growing evidence that, whenever resources allow, bacteriological examination (smear microscopy and culture) should be provided to all PLWHA, regardless of their symptom status [53–57]. In high TB burden settings, culture yielded almost double the number of TB cases detected by smear microscopy and required fewer number needed to diagnose (NND) compare to smear microscopy (8 compared to 19). This yield, however, was more pronounced among HIV-infected compared to HIV-uninfected individuals [58].

Our survey revealed a very high prevalence (12%) of undiagnosed active TB in this prison’s HIV-infected population. Negative consequences of missing active TB cases in congregate settings are more pronounced compared to community settings [59,60]. Reports from geographically-distinct settings showed that several TB outbreaks, particularly of MDR-TB in the community were linked to exposure in prisons [61–63]. Additionally, missed infectious TB cases may undermine TB control in such high-transmission closed settings. Despite intensive TB control efforts, a single strain of M tuberculosis persisted in a prison in the United States and contributed to 50% of diagnosed TB cases over nine years. This was attributed to infection/re-infection and reactivation of latent TB infection (LTBI) among prisoners in contact with index cases before being diagnosed [64]. Finding that more than one third (38%) of pulmonary TB cases were missed in spite of a regular screening program in a prison in the Cameroon, Noeske et al questioned the possibility of ever achieving TB control under any conditions of confinement [60]. Moreover, missing pulmonary TB cases in prisons complicates the scaling up of two major TB control interventions in high HIV prevalence settings, namely isoniazid preventive therapy (IPT) and ART [12,17,57,65–67]. An active case finding among ambulatory PLWHA with good immunological status (CD4 count >200 cells/µL) showed that 8 of 10 PLWHA with no TB symptoms were prescribed IPT for more than 4 weeks before the bacteriological confirmation of their active TB status [10] and may potentially contribute to isoniazid resistance since a single drug was offered to active TB cases.

The overall TB prevalence (newly identified and previously diagnosed TB cases) of 16.7% is one of the highest reported prevalence of active TB from a prison setting worldwide and is 165 times higher than that in the Malaysian general population [1,22]. It is well documented that prisoners come from communities at elevated TB risk, and this may have contributed to such high TB prevalence, among others. All, but one of the study participants used illicit drugs before being incarcerated and active TB disease was independently associated with longer duration of drug use in this sample. Given this variable’s co-linearity with age, it may be that longer duration of drug use was associated with longer-standing HIV infection and increased vulnerability to TB infection. Substance misuse, particularly injection drug use, is a recognized independent risk factor for LTBI and active TB disease [68]. The high prevalence of tuberculin skin test positivity among PWID in a substance abuse treatment center (86.7%) [69] and in another Malaysian prison (87.6%) [70] confirmed the excess risk of LTBI among this population in Malaysia. Moreover, TB remains a leading cause of death among HIV infected PWID [71]. To better control co-morbid illnesses among PWID, the WHO strongly recommends the co-location of treatment for TB, HIV, and opioid dependence using a “one-stop shopping” model of care, but many countries facing the detrimental convergence of these epidemics run these programs vertically [71–73]. The availability and impact of utilizing new rapid diagnostic tools, like the one used in this study, improve the likelihood of better integration of services to treat multiple co-morbidities, yet needs further empiric investigation.

Prisons have been identified as reservoirs for TB, ultimately spreading the disease to the general population through prison staff and released inmates. Increasing incarceration rates alone was projected to increase TB and MDR-TB incidence in civilian communities of the FSU [22,74]. Finding alternatives to incarceration through penal reforms, particularly for PWID, in an attempt to reduce overcrowding would likely reduce TB burden in similar communities [74]. Proposed alternatives include drug courts with alternatives to incarceration, noncustodial measures, enhanced bail procedures and adapted sentence policies [27]. In a similar setting, the diversion of PWID from prisons into community-based treatment facilities was a potential contributory factor leading to the reported reduction in TB prevalence in 27 Thai prisons over a six-year period [75].

Conclusions

In this high HIV and TB prevalence setting, active case finding using a single Xpert analysis has increased active pulmonary TB case detection among HIV-infected Malaysian prisoners. The new diagnostic tool is markedly superior to sputum smear microscopy, however, missed almost half of previously undiagnosed pulmonary TB cases (sensitivity of 53.3%) in the predominantly smear-negative sample. The extraordinarily high TB prevalence reported here affirms the importance of intensified TB case finding in congregate settings and warrants further exploration of multiple diagnostic approaches using the Xpert technology, potentially combined with other diagnostic screening modalities in correctional institutions with high HIV prevalence.

Supporting Information

Acknowledgments

The authors would like to especially thank all participants for their time and patience as well as to the many correctional officers at Penjara Kajang who helped facilitate the conduct of the study.