Latent Mycobacterium tuberculosis Infection

Introduction Despite the availability of effective therapy, tuberculosis (TB) continues to infect an estimated one-third of the world's population, to cause disease in 8.8 million people per year, and to kill 1.6 million of those afflicted [1]. Current TB control measures focus on the prompt detection and treatment of those with infectious forms of the disease to prevent further transmission of the organism. Despite the enormous success of this strategy in TB control, the persistence of TB in many parts of the world suggests the need to expand control efforts to identify and address the individual and social determinants of the disease. Since the early part of the 20th century, clinicians have observed an association between diabetes mellitus (DM) and TB, although they were often unable to determine whether DM caused TB or whether TB led to the clinical manifestations of DM [2–6]. Furthermore, these reports did not address the issues of confounding and selection bias. More recently, multiple rigorous epidemiological studies investigating the relationship have demonstrated that DM is indeed positively associated with TB [7–11]. While the investigators suggested that the association reflects the effect of DM on TB, some controversy over the directionality of the association remains due to observations that TB disease induces temporary hyperglycemia, which resolves with treatment [12,13]. A causal link between DM and TB does not bode well for the future, as the global burden of DM is expected to rise from an estimated 180 million prevalent cases currently to a predicted 366 million by 2030 [14]. Experts have raised concerns about the merging epidemics of DM and TB [15–17], especially in low- to middle-income countries, such as India and China, that are experiencing the fastest increase in DM prevalence [18] and the highest burden of TB in the world [19]. Given the public health implications of a causal link between DM and TB, there is a clear need for a systematic assessment of the association in the medical literature. We undertook a systematic review to qualitatively and quantitatively summarize the existing evidence for the association between DM and TB, to examine the heterogeneity underlying the different studies, and to evaluate the methodological quality of the studies. As our aim was to summarize the effect of DM on TB, we did not include studies that investigated the reverse association. Methods We conducted our systematic review according to the guidelines set forth by the Meta-analysis of Observational Studies in Epidemiology (MOOSE) group for reporting of systematic reviews of observational studies (see Text S2 for the MOOSE Checklist) [20]. Search Strategy and Selection Criteria We searched the PubMed database from 1965 to March 2007 and the EMBASE database from 1974 to March 2007 for studies of the association between DM and TB disease; our search strategy is detailed in Box 1. We also hand-searched bibliographies of retrieved papers for additional references and contacted experts in the field for any unpublished studies. Since we speculated that studies that examined the association between DM and TB may not have referred to the term “diabetes” in the title or abstract, we also searched for studies that examined any risk factors for active TB. We restricted our analysis to human studies, and placed no restrictions on language. We included studies if they were peer-reviewed reports of cohort, case-control, or cross-sectional studies that either presented or allowed computation of a quantitative effect estimate of the relationship between DM and active TB and that controlled for possible confounding by age or age groups. We also included studies that compared prevalence or incidence of DM or TB of an observed population to a general population as long as they had performed stratification or standardization by age groups. We excluded studies if they were any of the following: case studies and reviews; studies among children; studies that did not provide effect estimates in odds ratios, rate ratios, or risk ratios, or did not allow the computation of such; studies that did not adjust for age; studies that employed different methods for assessing TB among individuals with and without DM or for assessing DM among TB patients and controls; studies that investigated the reverse association of the impact of TB disease or TB treatment on DM; anonymous reports; and duplicate reports on previously published studies. Box 1. Search Strategy to Identify Observational Studies on the Association of Diabetes and Active Tuberculosis PubMed MeSH terms  1. “tuberculosis”  2. “diabetes mellitus”  3. “cohort studies” OR “case-control studies” OR “cross-sectional studies” OR “epidemiologic studies” OR “follow-up studies” OR “longitudinal studies” OR “prospective studies” OR “retrospective studies” Text terms  1. “tuberculosis”  2. “diabetes” OR “glucose intolerance” OR “glucose tolerance” OR “insulin resistance”  3. “chronic disease(s)” OR “non-communicable disease(s)”  4. “risk factor(s)” Search strings (all inclusive):  1. 1 AND 2  2. 1 AND 3 AND 5  3. 1 AND 3 AND 6  4. 1 AND 3 AND 7  5. 4 AND 5 (for the year preceding 3/2/07 for articles that may not have been assigned MeSH terms) EMBASE Text terms  1. “tuberculosis”, major subject  2. “diabetes mellitus”  3. “risk factor(s)”  4. “observational study” OR “longitudinal study” OR “prospective study” OR “case-control study” OR “cross-sectional study” Search strings (all inclusive):  1. 1 AND 2  2. 1 AND 3  3. 1 AND 4 Data Extraction The two investigators (CJ, MM) independently read the papers and extracted information on the year and country of the study, background TB incidence, study population, study design, number of exposed/unexposed people or cases/controls, definitions and assessment of DM and TB, statistical methods, effect estimates and their standard errors, adjustment and stratification factors, response rates, the timing of diagnosis of DM relative to that of TB, and the potential duplication of data on the same individuals. Differences were resolved by consensus. For the studies that did not directly report the background TB incidence, we obtained data for the closest matching year and state (or country) made available by public databases (WHO global tuberculosis database, http://www.who.int/globalatlas/dataQuery/; CDC Wonder, http://wonder.cdc.gov/TB-v2005.html). Data Analysis We separated the studies by study design and assessed heterogeneity of effect estimates within each group of studies using the Cochrane Q test for heterogeneity [21] and the I2 statistic described by Higgins et al. [22]. We determined the 95% confidence intervals (CIs) for the I2 values using the test-based methods [22]. We performed meta-analysis for computation of a summary estimate only for the study design (i.e., cohort) that did not show significant heterogeneity. Effect estimates of other study designs were not summarized due to significant heterogeneity. For those studies that reported age, sex, race, or region stratum-specific effects, we calculated an overall adjusted effect estimate for the study using the inverse-variance weighting method, then included this summary estimate in the meta-analyses and sensitivity analyses. We decided a priori to use the Dersimonian and Laird random effects method to pool the effect estimates across studies for the meta-analyses, because the underlying true effect of DM would be expected to vary with regard to underlying TB susceptibility and the severity of DM, and because it would yield conservative 95% confidence intervals [23]. In order to identify possible sources of heterogeneity and to assess the effect of study quality on the reported effect estimates, we performed sensitivity analyses in which we compared pooled effect estimates for subgroups categorized by background TB incidence, geographical region, underlying medical conditions of the population under study, and the following quality-associated variables: time of assessment of DM in relation to TB diagnosis, method of DM assessment (self-report or medical records versus laboratory tests), method of TB assessment (microbiologically confirmed versus other), adjustment for important potential confounders, and the potential duplication of data on the same individuals. To determine whether the effect estimates varied significantly by the above-mentioned factors, we performed univariate meta-regressions, in which we regressed the study-specific log-transformed relative risks (RRs) by the variables representing the study characteristics, weighting the studies by the inverse of the sum of within-study and between-study variance for all studies within the comparison. For background TB incidence, we created an ordinal variable, 1 representing < 10/100,000 person-years to 3 representing ≥ 100/100,000 person-years. Coefficients of meta-regression represent differences in log-transformed RRs between the subgroups; we tested the significance of these coefficients by Student t-test, and significance was set at p < 0.10. We considered studies to be of higher quality if they specified that DM be diagnosed prior to the time of TB diagnosis; used blood glucose tests for diagnosis of DM; used a microbiological definition of TB; adjusted for at least age and sex; were cohort, nested case-control, or population-based case-control studies; or did not have the potential for duplication of data. As the average background incidence rate of TB did not exceed 2 per 100 person-years in any of the of the case-control studies that had not employed incidence density sampling, we assumed TB to be sufficiently rare that the odds ratios would estimate the risk ratios [24], and that it would therefore be valid to compute summary RR in the sensitivity analyses regardless of the measure of association and design of the study. We explored possible effect modification by age by examining the three studies that reported results by age groups [7,9,25]. For this analysis, we graphed the stratum-specific estimates in a forest plot, and tested for heterogeneity of the effects within each study by the Q-test and I2 value. We also performed meta-regression within each study in which we regressed the log-transformed RRs by the mid-points of the age-bands. For the unbound age group, ≥ 60 y, we added half the range of the neighboring age-band, or 5 y, to the cutoff. We computed the factor reduction in RR with 10 y increases in age, and reported the p-value for significance of trend. We assessed publication bias using the Begg test and Egger test [26,27]. Statistical procedures were carried out using R version 2.5.1 [28]. 95% CI of the I2 value was computed using the “heterogi” module in STATA version 10 [29]. Results We identified and screened 3,701 papers by titles and abstracts; of these, 3,378 were excluded because they did not study risk factors for TB, were studies among children, were case reports, reviews, or studies of TB treatment outcome (Figure 1). Of the remaining 323 articles, 232 studies were excluded because they did not report on the association between DM and TB, and 56 studies were excluded because they were review articles (12) or ecological studies (2); studied the clinical manifestations of TB in people with diabetes (11); studied the association of DM and TB treatment outcome (6); assessed latent, relapsed, clustered, or drug-resistant TB as the outcome (6); studied the reverse association of the effect of TB on DM (5); had no comparison group (5); were case reports (3); did not give a quantitative effect estimate (3); had collapsed DM and other chronic diseases into a single covariate (2); or was a study that had been reported elsewhere (1). We contacted the authors of four papers that reported including DM in a multivariate analysis but that did not provide the adjusted effect estimate for DM; we included the papers of the two authors who responded and provided these adjusted estimates [30,31]. Further exclusion of studies that did not adjust for age (11), studies that used a general population as the comparison group for TB incidence or DM prevalence without standardization by age (9), and studies that used different methods for ascertaining TB in the people with diabetes and control group (2), left 13 eligible studies. These included three prospective cohort studies [7,30,32], eight case-control studies [8,11,31–37], and two studies for which study design could not be classified as either cohort or case control, as TB case accrual occurred prospectively while the distribution of diabetes in the population was assessed during a different time period after baseline [9,25]. The studies were set in Canada (1), India (1), Mexico (1), Russia (1), South Korea (1), Taiwan (1), the UK (1), and the US (6), and were all reported in English and conducted in the last 15 y. Two of the cohort studies were among renal transplant patients [30,32], and three of the case-control studies were hospital-based or based on discharge records [8,11,35]. The studies are summarized in Table 1. Figure 1 Flow Chart of Literature Search for Studies on the Association between Diabetes Mellitus and Active Tuberculosis Table 1 Summary of the 13 Observational Studies of Association between Diabetes and Active Tuberculosis Included in the Meta-analysis Table 1 Extended. Figure 2 summarizes the adjusted effect estimates of the 13 studies categorized by the study design. We found substantial heterogeneity of effect estimates from studies within each study design; between-study variance accounted for 39% of the total variance among cohort studies, 68% of the total variance among case-control studies, and 99% of the total variance in the remaining two studies. Despite this heterogeneity, the forest plot shows that DM is positively associated with TB regardless of study design, with the exception of the study by Dyck et al. [25]. DM was associated with a 3.11-fold (95% CI 2.27–4.26) increased risk of TB in the cohort studies. Of note, the study conducted within a nontransplant population provided greater weight (63%) to the summary estimate than the other two cohort studies combined. The effect estimates in the remaining studies were heterogeneous and varied from a RR of 0.99 to 7.83. Figure 2 Forest Plot of the 13 Studies That Quantitatively Assessed the Association between Diabetes and Active Tuberculosis by Study Designs Size of the square is proportional to the precision of the study-specific effect estimates, and the bars indicate the corresponding 95% CIs. Arrows indicate that the bars are truncated to fit the plot. The diamond is centered on the summary RR of the cohorts studies, and the width indicates the corresponding 95% CI. *Other: The studies by Ponce-de-Leon et al. [7] and Dyck et al. [25] were not specified as prospective cohort or case-control. TB case accrual occurred prospectively, while the underlying distribution of diabetes was determined during a different time period after baseline. Table 2 shows that there is an increased risk of active TB among people with diabetes regardless of background incidence, study region, or underlying medical conditions in the cohort. In the sensitivity analyses, we noticed that the strength of association increased from a RR of 1.87 to a RR of 3.32 as background TB incidence of the study population increased from < 10/100,000 person-years to ≥ 100/100,000 person-years, but the trend was not significant (trend p = 0.229). Effect estimates were heterogeneous within each category of background TB incidence (I2 = 60%, 98%, and 76% from highest to lowest background TB incidence category). Table 2 Results of Sensitivity Analyses to Identify Sources of Heterogeneity in the Magnitudes of the Association between Diabetes and Active Tuberculosis We also found that the associations of DM and TB in the study populations from Central America [9], Europe [33,37], and Asia [7,30,32] (RRCentralAm = 6.00, RREurope=4.40, RRAsia = 3.11) were higher than those of North American studies [8,11,33,34–36] (RRNA = 1.46) (meta-regression p CentralAm = 0.006, p Europe = 0.004, p Asia = 0.03). Among North American studies, the pooled estimate of the relative risks for Hispanics from two studies [8,11] was higher (RR = 2.69) than that of non-Hispanics from the same study [8] and other North American studies (RR = 1.23) (meta-regression p = 0.060) (Table 2). In general, stratification of the studies by quality-associated variables did not reduce the heterogeneity of effect estimates. Nonetheless, DM remained positively associated with TB in all strata. Studies that explicitly reported that DM was diagnosed prior to TB showed stronger associations (RR = 2.73) [7,31–34] than those that did not establish the temporal order of DM and TB diagnosis (RR = 2.10) [8,9,11,25,30,35–37], although the difference was not significant (meta-regression p = 0.483). Associations were stronger in studies that classified DM exposure through empirical testing (RR = 3.89) [7,9,32,34] rather than medical records (RR = 1.61) (meta-regression p = 0.051) [8,11,25,30,31,33]; and in those that confirmed TB status using microbiological diagnosis (RR = 4.91) [7,9,35,37] than in the studies that did not confirm by microbiological tests (RR = 1.66) (meta-regression p = 0.015) [8,11,25,30–34,36]. Among case-control studies, those that were nested in a clearly identifiable population or were population-based also reported stronger associations (RR = 3.36) [31,33,34,37] than those that used hospital based controls (RR = 1.62) [8,11,37], but the difference was not significant (meta-regression p = 0.321). Studies that had adjusted for smoking showed stronger associations (RR = 4.40) [33,37], while studies in which an individual may have contributed more than one observation to the data revealed weaker associations (RR = 1.62) [8,11]. Although these results suggest that higher-quality studies gave stronger estimates of association, we also found that the association was weaker in studies that adjusted for socioeconomic status (RR = 1.66) (Table 2) [8,11,37]. Figure 3 presents the summary measures of the association between DM and TB by age group based on the data from the three studies that presented age-stratified RRs. The plots from Kim et al. [7] and Ponce-de-Leon et al. [9] demonstrate stronger associations of DM and TB under the age of 40 y and declining RR with increasing age in age groups over 40 y (trend p Kim = 0.014, p Ponce-de-Leon = 0.184). Each 10 y increase in age was associated with a 0.6-fold reduction in magnitude of association in the study by Kim et al. [7]. This trend was not apparent in the study by Dyck et al. (Figure 3) [25]. Figure 3 Forest Plot of Age-Specific Association between Diabetes and Active Tuberculosis from Kim et al. [7], Ponce-de-Leon et al. [9], and Dyck et al. [25] Size of the square is proportional to the precision of the study-specific effect estimates, and the bars indicate 95% CI of the effect estimates. Arrows indicate that the bars are truncated to fit the plot. *Meta-regression: Factor reduction in RR with 10 y increase in age; p-values are given for test of linear trend. HR, hazard ratio. Both the Egger test and Begg test for publication bias were insignificant (p = 0.37, p = 0.14). Discussion Summary of Findings Our meta-analysis shows that DM increases the risk of TB, regardless of different study designs, background TB incidence, or geographic region of the study. The cohort studies reveal that compared with people who do not have diabetes, people with diabetes have an approximately 3-fold risk of developing active TB. Higher increases in risk were seen among younger people, in populations with high background TB incidence, and in non-North American populations. Heterogeneity of strengths of association may reflect true geographic/ethnic differences in severity of DM, transmission dynamics of TB, and the distribution of effect modifiers such as age, or it may be due to differences in study methodology or rigor. Given this heterogeneity of the RR estimates and the fact that all the cohort studies were conducted in Asia, we note that the summary estimate may not be applicable to other populations and study types. While the included studies covered a relatively broad range of geographic areas, there were none from Africa, where TB incidence is high. Nonetheless, a positive association of DM and TB was noted in two African studies [38,39] and several other studies that we excluded from the meta-analysis [10,40–42], as well as in a previous narrative review [43] of the association of DM and TB. Unlike the previous review, our systematic review identified five additional studies that had examined the association of DM and TB, computed a pooled summary estimate among the cohort studies, and determined important sources of heterogeneity through rigorous sensitivity analyses. Public Health Implication With an estimated 180 million people who have diabetes, a figure expected to double by year 2030, it is clear that DM constitutes a substantial contributor to the current and future global burdens of TB. For example, if we assume a RR of 3 and a prevalence of DM in Mexico of 6%, we can conclude that DM accounts for 67% of active TB cases among people with diabetes, and 11% of cases among the entire Mexican population (see Text S1 for the calculation) [44]. The contribution of DM to the burden of TB may be even higher in countries such as India and China where the incidence TB is greater and mean age is lower. In fact, a recent study by Stevenson et al. determined that DM accounts for 80.5% of incident pulmonary TB among people with diabetes, and 14.8% of incident TB in the total population in India [16]. The population-attributable risk for diabetes is comparable to that of HIV/AIDS; while HIV/AIDS is strong risk factor for TB (RRHIV = 6.5–26 [45], approximately 2–9 times greater than the RRDM estimated in this study), it is a less prevalent medical condition (33 million people infected in 2007 [46], approximately 5–6 times less prevalent than DM). Given these figures it may be puzzling to observe a decrease in TB in those areas that have experienced a growing burden of DM. We attribute this observation to negative confounding by factors such as improved nutrition and TB control measures in the areas of increasing DM such as India and China. Were these other factors to remain the same, we would expect to see a TB incidence trend reflecting that of DM in accordance with the positive association. Biological Plausibility Numerous studies have presented convincing biological evidence in support of the causal relationship between DM and impaired host immunity to TB. Studies in animal models have demonstrated that diabetic mice experimentally infected with M. tuberculosis have higher bacterial loads compared to euglycemic mice, regardless of the route of inoculation of M. tuberculosis [47,48]. Compared to euglycemic mice, chronically diabetic mice also had significantly lower production of interferon-γ (IFN-γ) and interleukin-12 (IL-12) and fewer M. tuberculosis antigen (ESAT-6)-responsive T cells early in the course of M. tuberculosis infection, marking a diminished T helper 1 (Th1) adaptive immunity, which plays a crucial role in controlling TB infection [48]. In experimental studies of human plasma cells, high levels of insulin have been shown to promote a decrease in Th1 immunity through a reduction in the Th1 cell to Th2 cell ratio and IFN-γ to IL-4 ratio [49]. Additionally, an ex vivo comparison study of production of Th1 cytokines showed that nonspecific IFN-γ levels were significantly reduced in people with diabetes compared to controls without diabetes [50]. Another study indicated a dose–response relationship; levels of IFN-γ were negatively correlated with levels of HbA1c (a measure of serum glucose levels over time in humans) [51]. Furthermore, neutrophils from people with diabetes had reduced chemotaxis and oxidative killing potential than those of nondiabetic controls [52], and leukocyte bactericidal activity was reduced in people with diabetes, especially those with poor glucose control [53]. Taken together, these studies strongly support the hypothesis that DM directly impairs the innate and adaptive immune responses necessary to counter the proliferation of TB. Limitations There are several potential limitations to this study. Our analysis was based on estimates derived from observational studies that are vulnerable to confounding by variables associated with both DM and TB. To address the issue of potential confounding, we performed a sensitivity analysis in which we reported separate summary estimates for the studies that adjusted for important potential confounders and those that did not. Studies that controlled for socioeconomic status in a multivariable model found that the adjusted effect of DM was reduced, but not eliminated. Crude effect estimates were not provided in two of the larger studies that adjusted for socioeconomic status, thus the direction of bias cannot be determined. The three studies that did report both crude and the adjusted estimates [33,34,37] found that the adjusted RRs for DM were higher. Although we could not exclude the possibility of residual confounding by unmeasured confounders in these observational studies, such as other chronic diseases that often coexist with diabetes, we found that the effect of DM on TB risk persisted even after adjustment for multiple potential confounders that are likely to be correlated with unmeasured factors. Eight of the studies included in this meta-analysis were case-control studies. Control selection strategies included sampling from hospitals, discharge records, department of health records, the general population, and the cohort in which the study was nested. Sampling controls from hospital or discharge records may have introduced a Berkson bias—a selection bias that can occur when both the exposure and the outcome are associated with attendance at a health-care facility from which cases and controls are recruited [54]. Since DM can lead to multiple health problems, the prevalence of DM is likely to be higher among persons attending clinics or being admitted to hospitals than it is in the general population. This bias would be expected to result in an underestimation of the effect of DM on TB, an expectation that was consistent with our finding that studies using hospital-based controls reported lower effect estimates [54]. Other sources of potential bias include misclassification of either exposure or outcome, such as may have occurred in studies that did not employ laboratory tests to diagnose DM or TB. When we restricted our analysis to studies that used glucose tests to determine DM status, we found that effect estimates were higher than in the studies that relied on less-rigorous methods, consistent with our expectation of a bias toward the null among studies that nondifferentially misclassify the exposure. Studies that utilized glucose tests to classify the exposure may also have reported higher RRs of TB among people with diabetes, since they may have identified undiagnosed people with diabetes who remained untreated and therefore may have had higher glycemic levels that those who self-reported their status. Those studies that confirmed TB through microbiological diagnosis also reported stronger associations, suggesting that diabetes may have a stronger impact on smear-positive and thus transmissible forms of TB. Our result underscores the conclusion by Stevenson et al. that DM accounts for a greater proportion of smear-positive TB than of other forms [17]. In short, we found that magnitudes of association varied by the quality of the studies; at the same time, variations may have been influenced by differences in population characteristics that are correlated with quality-associated variables. Another important limitation of our systematic review is that most of the studies we included failed to examine age as an effect modifier of the relationship between DM and TB. The studies by Kim et al. [7] and Ponce-de-Leon et al. [9] found that estimates varied markedly by age, with substantially higher estimates among younger people. This finding may be explained by heterogeneity of the individuals without diabetes between the age groups. Because baseline glucose tolerance is lower in older persons without diabetes, elderly controls may have had an elevated risk of TB compared to younger ones [55], thus reducing the apparent effect of DM. It is possible that younger people with diabetes might have had type I diabetes, a more severe form of diabetes with a juvenile onset; however, because most studies did not distinguish between type I and type II diabetes we cannot conclude whether the effect modification by age would have been due to differences in types of diabetes. Notably, the study by Dyck et al. [25] did not demonstrate this trend in the age-specificity of the effect of DM on TB and in fact showed a negative association among the elderly. The authors of the study note that results may have been biased by differential mortality in the elderly since individuals with diabetes who would have been most at risk for TB would have already died. Moreover, this study also differed from the others in that it relied on medical records rather than laboratory tests to determine DM status, and it had not included DM occurring in the last six of the 16 y during which TB case accrual occurred. Conclusions In summary, we found consistent evidence for an increased risk of TB among people with diabetes despite heterogeneity in study design, geographic area, underlying burden of TB, assessment of exposure and outcome, and control of potential confounders. Data from these human studies are consistent with emerging information on the biological mechanisms by which hyperglycemia may affect the host immune response to TB. Our findings suggest that TB controls programs should consider targeting patients with diabetes for interventions such as active case finding and the treatment of latent TB and, conversely, that efforts to diagnose, detect, and treat DM may have a beneficial impact on TB control. We also recommend further studies investigating how TB risk varies by type, duration, and severity of DM, for a more thorough understanding of the association that could be translated to a clear public health message. Supporting Information Text S1 Calculation of Attributable Risk Fraction of TB among Patients with Diabetes and Population-Attributable Risk Fraction of TB Due to Diabetes (17 KB DOC) Click here for additional data file. Text S2 MOOSE Checklist (61 KB DOC) Click here for additional data file.

Introduction By the end of 2009, an estimated 33 million people were living with HIV, the vast majority in sub-Saharan Africa and Asia. Tuberculosis (TB) remains the most common cause of death in people living with HIV. Compared to people without HIV, people living with HIV have a more than 20-fold increased risk of developing TB [1]. TB disease may occur at any stage of HIV disease and is frequently the first recognized presentation of underlying HIV infection [2],[3]. Without antiretroviral treatment (ART), up to 50% of people living with HIV who are diagnosed with TB die during the 6–8 mo of TB treatment [4]–[6]. This risk increases to 72%–98% among those with multi-drug (MDR) or extensively drug-resistant (XDR) TB [7],[8]. Although ART can reduce the incidence of TB both at individual [9] and population [10] levels, people living with HIV on ART still have higher TB incidence rates and a higher risk of dying from TB [11]. Routine TB screening offers the opportunity to diagnose and promptly treat TB disease, and to identify those without TB disease who may be eligible for TB preventive therapy [12]. The use of TB preventive therapy can reduce TB incidence and is therefore of considerable benefit to patients [13]. For these reasons, the World Health Organization (WHO) recommends regular screening for active TB disease of all people living with HIV and providing either treatment for active disease or isoniazid preventive therapy (IPT) to mitigate TB morbidity, mortality, and transmission [14]. However, in 2009, of the estimated 33 million people living with HIV, only 1.7 million (5%) were screened for TB, and about 85,000 (0.2%) were offered IPT [15]. Currently there is no internationally accepted evidence-based tool to screen for TB in people living with HIV. Several studies have shown that the presenting signs and symptoms of TB in people living with HIV are different from those in people without HIV to diagnose TB; for example, many people living with HIV who have culture-confirmed TB do not report having a prolonged cough, which is one of the standard TB screening questions used by national TB control programs globally [16]. Moreover, the most widely available TB diagnostic tests such as smear microscopy and chest radiography perform poorly among people living with HIV; because most people living with HIV and TB have either sputum acid-fast bacilli (AFB) smear negative pulmonary or extrapulmonary TB [17]. We conducted an individual participant data meta-analysis of published and unpublished studies to develop a simple, standardized TB screening rule for resource-constrained settings that will adequately separate patients into two groups: (1) those for whom TB is reliably excluded, and IPT and ART, if indicated, can be initiated; and (2) those who require further investigation for TB disease. We describe the results of this meta-analysis and propose an algorithm for TB screening among people living with HIV in resource-constrained settings. Methods We proceeded through several steps. First, we prospectively enumerated criteria for studies to be included in our meta-analysis. Second, we searched for and selected studies that met these criteria. Third, we sought primary data from investigators and mapped individual-level data to common symptoms. Fourth, we identified five symptoms available in most studies and, within each study, computed the sensitivity and specificity of 23 screening rules derived from these five symptoms. Finally, we used meta-analysis methods to estimate the performance of all 23 rules, as well as the association of study-level and individual-level correlates with performance. Inclusion of Studies We defined studies as being eligible for inclusion in this analysis if they met the following criteria: (1) collected sputum specimens from people living with HIV regardless of signs or symptoms; (2) used mycobacterial culture of at least one specimen to diagnose TB; and (3) collected data about signs and symptoms. Search Strategy and Selection of Studies To identify studies eligible for the meta-analysis, we conducted a systematic literature review of studies related to TB screening among people living with HIV in June 2008 using PubMed and various combinations of the following keywords: “HIV,” “Tuberculosis,” “TB screening,” “Smear-negative TB,” “Sputum negative TB,” “TB case finding,” “Intensified TB case finding,” “Isoniazid prevention treatment, trial or therapy.” We also searched for abstracts presented at conferences organized by the International Union Against TB and Lung Diseases and the International AIDS Society between 2000–2008. No language restriction was placed on the search. We reviewed all retrieved titles and abstracts for relevance to the topic. The reference lists of retrieved studies were also reviewed to identify further studies that meet the eligibility criteria. In addition, recognized experts in the field were contacted to identify studies that were not available (e.g., unpublished) in the initial electronic search. Studies that involve concomitant HIV testing and mycobacterial culture on all patients are resource intensive and challenging to implement in countries with a high burden of TB and HIV. Therefore, we believe it is unlikely that eligible studies would have been completed but missed by our search strategy. We found 2,119 publications and reviewed all their abstracts. Using the criteria above, we selected 53 for review of the full text. Twenty-one articles included information on signs and symptoms for TB screening in people living with HIV. A total of 14 studies (six published and eight unpublished at the time of the search) met the inclusion criteria of our meta-analysis (Figure 1). The corresponding authors or principal investigators were contacted for all 14 studies to confirm that their studies met all the eligibility criteria. One unpublished dataset was excluded for not meeting the inclusion criteria after verification with the principal investigator, and another one was excluded because the investigators could not submit the data within the agreed timeframe. A total of 12 investigators (representing six published and six unpublished studies at the time of the search) provided de-identified individual patient data for inclusion in the primary meta-analysis within an agreed time framework (Table 1) [18]–[29]. In November 2010, immediately preceding manuscript publication, we re-ran the search strategy again to look for additional studies that were reported since the initial search and should have been included in the meta-analysis. The search found seven eligible studies, of which all except one [30] were included in our meta-analysis as unpublished datasets. 10.1371/journal.pmed.1000391.g001 Figure 1 Search strategy and studies included in the meta-analysis (PRISMA flow diagram). 10.1371/journal.pmed.1000391.t001 Table 1 Summary of studies included in the meta-analysis. Reference Study Population Characteristics (Setting, Continent, n Samples, Culture Method Used) Sample Size PLTB/PLHIVa (%) PLTB/PLHIV (%) with Data on the Five Symptoms Ayles et al. 2009 Adults of more than 15 y of age sampled from one rural and one urban communities in Zambia (community, sub-Saharan Africa, 1 LJ and MGIT). 8,044 43/2,253 (1.9) 41/2,145 (1.9) Corbett et al. 2010b Random sample of adults in 46 previously enumerated neighbourhoods in the high density suburbs of Harare, Zimbabwe (community, sub-Saharan Africa, 3 LJ). 10,079 31/1,841 (1.7) 31/1,834 (1.7) Cain et al. 2010 PLHIV from 8 outpatient clinics in Cambodia, Thailand, and Vietnam who were enrolled regardless of signs or symptoms suggestive of TB (clinical, Southeast Asia, 3 MGIT and LJ). 1,748 267/1,724 (15.5) 267/1,721 (15.5) Day et al. 2006 Employees of a gold mining company first attending the TB preventive therapy clinic in South Africa (miners, sub-Saharan Africa, 2 LJ). 1,093 32/991 (3.2) 0/0 (–) Corbett et al. 2007b Employees of 22 small and medium-sized enterprises in Zimbabwe (community, sub-Saharan Africa, 3 LJ). 4,668 3/797 (0.4) 3/797 (0.4) Lewis et al. 2009b All consenting employees of a gold mining indusry who undergo annual medical examinations in an occupational health centre in South Africa (miners, sub-Saharan Africa, 2 LJ). 1,955 18/560 (3.2) 18/560 (3.2) Shah et al. 2009 All newly diagnosed HIV-positive clients of at least 18 y old from the VCT Clinic in a large referral hospital in Addis Ababa, Ethiopia (clinical, sub-Saharan Africa, 1 LJ). 453 27/427 (6.3) 22/357 (6.2) Kimerling et al. 2002 PLHIV of at least 15 y of age and enrolled in an HIV home-based care service in Phnom Penh, Cambodia (community, Southeast Asia, 1 LJ). 441 36/393 (9.2) 36/393 (9.2) Lawn et al. 2009 PLHIV with more than 18 y of age who were referred to a community-based ART service in South Africa (clinical, sub-Saharan Africa, 2–4 MGIT). 235 58/226 (25.7) 57/218 (26.1) Wood et al. 2007 Randomly sampled and consenting adults living in shacks in a high-density residential area in South Africa (community, Sub-Saharan Africa, 4 MGIT). 174 12/163 (7.4) 0/0 (–) Mohammed et al. 2004 PLHIV with WHO clinical stage 3 or 4 disease referred for possible participation in a trial of TB-preventive therapy in 3 hospital-based adult HIV clinics in South Africa (clinical, sub-Saharan Africa, 1 BACTEC). 129 10/128 (7.8) 0/0 (–) Chheng et al. 2008 All consenting participants of more than 19 y old who were tested for HIV in a Voluntary Counseling Center and referred for TB screening in Cambodia (clinical, Southeast Asia, 3 LJ). 504 20/123 (16.3) 20/123 (16.3) Total 29,523 557/9,626 (5.8) 495/8,148 (6.1) a The following patients were excluded: (1) patients who were on TB treatment or prophylaxis at enrolment; (2) patients who were smear positive, but culture grew non- tuberculosis mycobacterium (NTM); and (3) patients who were smear positive, but culture negative. b Patients were previously screened for TB before enrolment into the study. LJ, Lowenstein-Jensen culture medium; MGIT, Mycobacterial Growth Indicator Tube culture system; PLHIV, persons living with human immunodeficiency virus; PLTB, persons with tuberculosis disease; VCT, voluntary counselling and testing for HIV. Investigators for all included studies signed a data sharing and confidentiality agreement, and agreed to a data management, analysis, and publication plan. During design and analysis phases of the meta-analysis, the investigators of the studies and data managers of the included studies held multiple discussions by email, by teleconference, and in person in Geneva, Switzerland and Atlanta, Georgia, United States. Data Abstraction and Management The list of variables from the most comprehensive dataset [20] was used to construct an initial master list of variables. All the variables from each study included in the meta-analysis were mapped to this master list. Principal investigators and data managers for the 12 studies worked with the meta-analysis investigators to ensure accurate mapping of data from the primary studies to the master variable list. In the end, the final dataset of the meta-analysis included 159 variables that appeared in at least two of the studies. We identified five symptoms common to most studies and limited the meta-analysis to the nine studies with substantially complete information for those five symptoms. Case Definitions We defined a TB patient as any person living with HIV and at least one specimen culture positive for Mycobacterium tuberculosis (MTB). We defined participants as having no TB if cultures were negative for MTB and participants were judged not to have TB on the basis of the original study definition of the investigators. We excluded from the analysis: (1) patients who were receiving treatment for TB (infection or disease) at enrolment; (2) patients who were AFB smear positive, but whose culture grew non-tuberculous mycobacteria; and (3) patients who were AFB smear positive or scanty, but sputum culture negative. Sources of Study Heterogeneity All studies were reviewed to identify study-level characteristics that could substantially impact the findings of the meta-analysis. Two studies were conducted exclusively among gold miners living in South Africa [21],[23], a population that may not be broadly generalizable, because of its demographics, its high prevalence of non-TB illnesses (e.g., silicosis) that can produce cough, and the practice of annual TB radiological screening. Five studies [18],[19],[22],[25],[27] were conducted among individuals drawn from a community setting through prevalence surveys, which may lead to enrolment of patients with a different spectrum of TB and HIV disease than would be found among patients seeking care. Three studies [19],[22],[23] involved participants who were previously screened for TB or who had had access to routine TB screening before being enrolled into the studies. Finally, three studies exclusively used liquid media to culture specimens [26]–[28], two studies used both solid and liquid media [18],[20], and seven studies exclusively used solid media (Table 1). Liquid media have substantially increased sensitivity for growing MTB, particularly in patients with low levels of TB bacilli, as would be expected in a population of people living with HIV being screened for TB [31]. Studies that used liquid media, therefore, would have improved ability to classify patients correctly into those who have TB and those who do not. We explored the impact of these factors on the performance of the screening algorithms and analyzed subsets of the final dataset grouped according to these characteristics. Data Analysis We compared characteristics of patients with TB to those of patients without TB to derive a standardized rule for TB screening among people living with HIV. The goal of TB screening is to divide the population of people living with HIV into two groups: (1) those who do not have TB; and (2) those who need further evaluation for the diagnosis of TB (i.e., TB suspects). We restricted our analysis to clinical symptoms that could be readily assessed at any level of the health system and that were asked about in all studies: current cough (C), haemoptysis (H), fever (F), night sweats (S), and weight loss (W). Using the four studies that included chest radiography data [20],[23],[24],[26], we also evaluated the impact of adding abnormal chest radiography findings to the TB screening rule. Only observations with no missing data for the symptoms of interest were included in the analysis. We considered “1-of-n” rules as candidates for screening for TB that could best classify patients into two groups (not TB and suspected TB) with high sensitivity [20]. The “1” represents the minimum number of symptoms that must be present in an individual to be classified as a suspected TB patient and the “n” represents the number of symptom(s) in a given rule. For example, a “1-of-3” rule could be a set of symptoms such as current cough, fever, and weight loss, abbreviated here as CFW. An individual with at least one symptom specified in this particular rule would be classified as a TB suspect; an individual without any of these symptoms would be classified as not having TB. We considered all combinations of the five candidate symptoms except for combinations that include both current cough and haemoptysis, yielding a total of 23 candidate rules: two 1-of-4 rules (CFSW, HFSW), seven 1-of-3 rules (CFS, CFW, CSW, HFS, HFW, HSW, FSW), nine 1-of-2 rules (CF, CS, CW, HF, HS, HW, FS, FW, SW), and five 1-of-1 rules (C, H, F, S, W) (see also Table 3.) The analysis with abnormal chest radiographic findings (X) considered these 23 rules, each augmented with this additional sign (e.g., CFSWX). Other analyses have considered m-of-n rules with m>1 [20]. These rules cannot exceed the sensitivity of 1-of-n rules. Suppose that a positive screen requires the presence of at least two symptoms out of current cough, fever, and night sweats, the number of true positives for this 2-of-3 rule will not be greater than the number of true positives from the corresponding 1-of-3 rules. Because our aim in this analysis was to identify the most sensitive rule, we did not include rules of this kind in our analysis. We applied two closely related methods for cross-study analysis of sensitivity and specificity of these 23 candidate screening rules: bivariate random-effects meta-analysis (BREMA) and the hierarchical summary relative operating characteristic (HSROC) curve [32],[33]. BREMA jointly models sensitivity and specificity while accommodating between-study heterogeneity, and HSROC models tradeoffs between sensitivity and specificity across study populations. Both methods can be unified in the same model. In addition to sensitivity and specificity, we calculated predictive value negative and likelihood ratio negative of each candidate rule [34]. Our goal was to identify a combination of symptoms that achieved the highest possible sensitivity and the lowest possible negative likelihood ratio for ruling out TB disease (without any predetermined cut-off points). To further understand between-study heterogeneity and other factors associated with the diagnostic performance of the most sensitive rule, we analyzed several study-level predictors (setting, prior screening of study participants, culture medium used, and geographic region) and participant-level predictors (age, gender, CD4 T-lymphocyte cell count [CD4], and abnormal chest radiographic findings). Our analytic methods produced odds ratios that reflect the magnitude of association between each factor and the probability of correctly identifying persons with TB (sensitivity) or without TB (specificity). For a range of TB prevalence values, we calculated the negative predictive values at levels of each covariate. We calculated the ratio of the number of patients that screen positive but who actually have no TB (false positives) and hence unnecessarily require additional TB diagnostic evaluation (e.g., culture) to the number of patients that screen positive and actually have TB (true positives), which is referred to as the number needed to screen. We calculated this ratio for different rules using a theoretical population of 1,000 people living with HIV with different levels of TB prevalence. This ratio provides proxy information similar to a marginal cost-effectiveness analysis for different screening rules and it helps quantify how much a health program would need to invest (as measured by additional diagnostic tests) for every patient with TB identified through the screening rule [35]. Each observation with a missing covariate value was omitted from analysis of that covariate. BREMA models were fitted using SAS procedure Glimmix (SAS 9.22, SAS Institute), and further calculations were performed in R (R 2.10.1, R Development Core Team). Ethical Review All data collection included in the meta-analysis was approved by institutional ethical review boards at the respective institutions during the original study; if necessary, principal investigators requested additional approval from institutional review boards for the inclusion of the primary dataset in the meta-analysis. All data for the meta-analysis were provided completely de-identified. In the meta-analysis dataset, investigators were not able to link case records to individuals. Results Investigators provided data about 29,523 participants, of whom 10,057 were people living with HIV. The dataset included 9,626 people living with HIV who had TB screening and sputum culture performed, of whom 8,148 could be evaluated on the five symptoms of interest from nine of 12 studies (Figure 2). 10.1371/journal.pmed.1000391.g002 Figure 2 Flow chart of study participants included in the individual patient data meta-analysis. Most patients (77% [7,386/9,626]) were from sub-Saharan Africa; the rest were from Southeast Asian countries. The median age was 34 y (interquartile range [IQR], 27–41 y). Of the 9,626 patients with HIV in the 12 studies, CD4 cell count information was available for 3,489 (36%) and chest radiography information for 3,903 (41%). The median CD4 count was 248 cells/µl (IQR, 107–409). The overall prevalence of TB disease was 5.8% (557/9,626), ranging across studies from 0.4% to 25.7% (Table 1). More than half of TB patients (52% [288/557]) had sputum smear negative pulmonary TB, whereas 39% (218/557) had sputum smear positive pulmonary, and 5% (28/557) had exclusively extrapulmonary TB. The anatomic site of TB was not specified in 4% (23/557) of patients. Table 2 summarizes the distribution of common variables, and Table S1 summarizes how each question was actually asked in each study. Because duration of cough was included in many studies but was asked about in different ways, we were able to analyze data using three different cough variables: cough in the past 4 wk (information available for 39.3% of participants); cough lasting for 2 wk or more (information available for 47.1%); and cough present in the last 24 h, which is referred to as “current cough” (information available for 89.6%). 10.1371/journal.pmed.1000391.t002 Table 2 Characteristics of participants with and without TB for variables included in the analysis. Characteristic All PLHIV (n = 9,626) PLHIV with Data on the Five Symptoms (n = 8,148) No TB Disease (n = 9,069), n (%) TB Disease (n = 557), n (%) TB Disease (n = 7,653), n (%) TB Disease (n = 495) n (%) Origin of patient Sub-Saharan Africa 7,152 (78.9) 234 (42.0) 5,739 (75.0) 172 (34.8) Southeast Asia 1,917 (21.1) 323 (58.0) 1,914 (25.0) 323 (65.2) Setting Clinical 2,246 (24.8) 382 (68.5) 2,053 (26.8) 366 (73.9) Community 5,322 (58.7) 125 (22.4) 5,058 (66.1) 111 (22.4) Miners 1,501 (16.5) 50 (9.0) 542 (7.1) 18 (3.6) Sex Male 4,957 (54.7) 356 (63.9) 3,811 (49.8) 309 (62.4) Female 4,111 (45.3) 201 (36.1) 3,841 (50.2) 186 (37.6) Missing or not recorded 1 (0.0) 0 (0.0) 1 (0.0) 0 (0.0) Median age (IQR), n =  8,633 (7,286) 34 (27–41) 33 (28–40) 33 (27–40) 32 (28–39) Median CD4+ count (IQR), n =  3,489 (2,409) 268 (126–427) 106 (38–241) 229 (94–391) 94 (33–215) Cough in the past 4 wk Yes 1,439 (15.9) 303 (54.4) 1,270 (16.6) 278 (56.2) No 1,909 (21.0) 129 (23.2) 1,067 (13.9) 110 (22.2) Missing or not recorded 5,721 (63.1) 125 (22.4) 5,316 (69.5) 107 (21.6) Cough lasting for 2 wk or more Yes 957 (10.6) 197 (35.4) 848 (11.1) 177 (35.8) No 3,093 (34.1) 288 (51.7) 2,046 (26.7) 260 (52.5) Missing or not recorded 5,019 (55.3) 72 (12.9) 4,759 (62.2) 58 (11.7) Haemoptysis Yes 543 (6.0) 60 (10.8) 523 (6.8) 58 (11.7) No 8,509 (71.4) 495 (88.9) 7,130 (93.2) 437 (88.3) Missing or not recorded 17 (0.2) 2 (0.4) 0 (0.0) 0 (0.0) Current cough or cough in the past 24 h Yes 1,625 (17.9) 274 (49.2) 1,530 (20.0) 260 (52.5) No 6,474 (71.4) 250 (44.9) 6,123 (80.0) 235 (47.5) Missing or not recorded 970 (10.7) 33 (5.9) 0 (0.0) 0 (0.0) Current fever or fever in the past 4 wk Yes 1,801 (19.9) 294 (52.8) 1,669 (21.8) 280 (56.6) No 7,002 (77.2) 246 (44.2) 5,984 (78.2) 215 (43.4) Missing or not recorded 266 (2.9) 17 (3.0) 0 (0.0) 0 (0.0) Current night sweats or night sweats in the past 4 wk Yes 1,710 (18.9) 242 (43.4) 1,497 (19.6) 225 (45.4) No 7,329 (80.8) 313 (56.2) 6,156 (80.4) 270 (54.6) Missing or not recorded 30 (0.3) 2 (0.4) 0 (0.0) 0 (0.0) Current weight loss or weight loss in the past 4 wk Yes 2,434 (26.8) 333 (59.8) 2,258 (29.5) 308 (62.2) No 6,478 (71.4) 218 (39.1) 5,395 (70.5) 187 (37.8) Missing or not recorded 157 (1.7) 6 (1.1) 0 (0.0) 0 (0.0) Abnormal chest radiography Yes 581 (6.4) 271 (48.7) 294 (3.8) 239 (48.3) No 2,900 (32.0) 151 (27.1) 2,155 (28.1) 145 (29.3) Missing or not recorded 5,588 (61.6) 135 (24.2) 5,204 (68.0) 111 (22.4) Abnormal chest radiography consistent with TB Yes 377 (4.2) 227 (40.8) 261 (3.4) 209 (42.2) No 2,589 (28.5) 144 (25.8) 1,641 (21.4) 129 (26.1) Missing or not recorded 6,103 (67.3) 186 (33.4) 5,751 (75.2) 157 (31.7) Any one of current cough, fever, night sweats, or weight loss Yes 3,591 (39.6) 425 (76.3) 3,563 (46.6) 418 (84.4) No 4,090 (45.1) 77 (13.8) 4,090 (53.4) 77 (15.6) Not evaluable 1,388 (15.3) 55 (9.9) 0 (0.0) 0 (0.0) We analyzed the performance of individual and combinations of symptoms as screening rules using data from the 8,148 participants who could be evaluated based on the five candidate symptoms. Table 3 shows the diagnostic performance characteristics for the 23 candidate combinations of symptoms, sorted from highest sensitivity to lowest. The most sensitive rule was the presence of any one of the following symptoms: current cough, fever, night sweats, and weight loss (CFSW). The population-average sensitivity of this symptom combination was 78.9% (95% confidence interval [CI] 58.3%–90.9%) with the negative likelihood ratio of 0.426 (95% CI 0.349–0.520), which corresponds to a postscreening reduction in the probability of TB by 15%–20% [34]. 10.1371/journal.pmed.1000391.t003 Table 3 Diagnostic performance of 23 candidate 1-of-n rules. Rule Sensitivity (95% CI) Specificity (95% CI) LRN (95% CI) CFSW 78.9 (58.3–90.9) 49.6 (29.2–70.1) 0.426 (0.349–0.520) HFSW 75.7 (53.9–89.2)a 52.7 (31.8–72.7) 0.461 (0.391–0.544) CFW 74.0 (51.7–88.3)a 53.8 (32.8–73.6) 0.483 (0.416–0.561) CSW 73.4 (51.0–88.0) 53.8 (32.8–73.5) 0.494 (0.428–0.570) CFS 73.1 (50.6–87.9) 61.1 (39.7–79.0) 0.440 (0.382–0.506) HFW 70.6 (47.5–86.4) 57.5 (36.2–76.4) 0.511 (0.454–0.576) FSW 69.2 (45.9–85.6) 55.7 (34.5–75.0) 0.554 (0.497–0.617) HSW 68.1 (44.6–85.0) 58.7 (37.3–77.2) 0.544 (0.492–0.602) CW 65.3 (41.6–83.3) 60.3 (38.8–78.4) 0.576 (0.530–0.625) CF 65.0 (41.3–83.1) 68.6 (47.7–83.9) 0.510 (0.470–0.553) HFS 63.7 (39.9–82.3) 66.3 (45.2–82.4) 0.548 (0.509–0.589) FW 63.1 (39.3–81.9) 61.4 (40.0–79.1) 0.601 (0.560–0.644) SW 61.0 (37.2–80.5) 61.9 (40.5–79.5) 0.630 (0.594–0.669) CS 59.7 (35.9–79.6) 69.4 (48.7–84.4) 0.581 (0.551–0.613) HW 56.8 (33.3–77.6) 66.8 (45.7–82.8) 0.647 (0.620–0.675) FS 56.3 (32.8–77.3) 70.1 (49.6–84.9) 0.623 (0.598–0.649) HF 52.0 (29.2–74.1) 75.0 (55.6–87.7) 0.640 (0.620–0.660) W 49.3 (27.0–71.9) 71.1 (50.8–85.5) 0.712 (0.693–0.733) F 42.8 (22.2–66.3) 79.8 (62.4–90.4) 0.716 (0.695–0.738) HS 38.9 (19.5–62.6) 78.1 (59.9–89.5) 0.782 (0.753–0.813) C 38.5 (19.2–62.2) 81.8 (65.3–91.5) 0.753 (0.724–0.783) S 31.4 (14.8–54.6) 82.2 (65.9–91.7) 0.835 (0.780–0.893) H 5.9 (2.3–14.5) 94.4 (87.6–97.6) 0.996 (0.735–1.351) Sensitivity, specificity, and likelihood ratio negative (LRN) from the bivariate random-effects meta-analysis model. Rule is at least one of the indicated symptoms. C, current cough; H, haemoptysis; F, fever; S, sweats; W, weight loss. a p-Value >0.05 for the same sensitivity of the CFSW rule and the indicated rule. All specificities are significantly different from that of the CFSW rule. LRN, likelihood ratio negative. The nine included studies demonstrated significant between-study heterogeneity on both sensitivity (p<0.001) and specificity (p<0.001) of the rule CFSW (see also Figure 3). The bivariate graphic shows that six studies have study-level specificities below and three above the population average specificity. Furthermore, this rule has the highest-ranking sensitivity in eight of the nine included studies (Table S2). The hierarchical summary relative operating characteristic curves (Figure S1) show slightly better overall diagnostic performance of the rules CFS and CF, but our application requires the highest sensitivity possible, allowing for some tradeoffs with lower specificity. Figure 3 shows that three studies are outliers, and they represent studies of patients who were previously screened for TB or studies in which much of the population likely had previous TB screening (e.g., miners); this can modify the performance characteristics of the screening rule. 10.1371/journal.pmed.1000391.g003 Figure 3 Diagnostic performance of CFSW rule in the included studies. BREMA, bivariate random-effects meta-analysis; HSROC, hierarchical summary relative operating characteristic. The CFSW rule has sensitivity of 90.1% (95% CI 76.3%–96.2%) and 67.1% (95% CI 41.7%–85.3%) among participants selected from clinical and community settings, respectively. Similarly the sensitivity of the rule among those who had not been previously screened for TB was higher at 88.0% (95% CI 76.1%–94.4%) compared to those who had been screened for TB at 40.5% (95% CI 16.6%–69.9%). At the 95% confidence level, the sensitivity of this rule could not be statistically distinguished from the sensitivity of the rule that substitutes haemoptysis for current cough (HFSW, 75.7% sensitive [95% CI 53.9–89.2%]) or the rule that drops night sweats (CFW, 74.0% sensitive [51.7–88.3%]). All other rules had lower sensitivity. Regression analysis of study-level predictors revealed that studies in which TB screening was performed in clinical settings had 4.5 times the odds for a true-positive screening result compared to studies in which TB screening was performed in a community setting (95% CI 1.0–19.5). Studies of participants who had not previously been screened for TB had 10.8 times the odds for a true-positive screen (95% CI 2.4–47.8) compared with studies in which participants had previously been screened for TB. Participants with CD4 cell count <200 cells/µl had 6.4 times the odds of a true-positive screen (95% CI 2.9–14.2). Statistically significant predictors of true-negative results include prescreening, geographic region, participant age ≥33 y, CD4 cell count <200 cells/ml, and abnormal result on chest radiograph (Table 4). 10.1371/journal.pmed.1000391.t004 Table 4 Association of study-level and individual-level predictors with the diagnostic performance of CFSW rule. Predictors Sensitivity (95% CI) Specificity (95% CI) Study level Setting Community 1.0 Clinical 4.45 (1.02, 19.46)a 0.25 (0.06–1.01) Miners 0.25 (0.02–2.51) 4.07 (0.44–37.68) Screening Prescreened for TB 1.0 Not screened for TB 10.82 (2.45–47.78)a 0.08 (0.06–0.12)a Culture medium Solid 1.0 Liquid 3.41 (0.57–20.30) 0.33 (0.06–1.97) Region Sub-Saharan Africa 1.0 Southeast Asia 4.03 (0.65–24.84) 0.20 (0.04–1.00)a Individual level Ageb <33 y 1.0 ≥33 y 1.43 (0.81–2.52) 0.74 (0.66–0.84)a Gender Female 1.0 Male 1.26 (0.71–2.24) 1.04 (0.93–1.16) CD4 cell countc ≥200 cells/µl 1.0 <200 cells/µl 6.38 (2.87–14.17)a 0.46 (0.38–0.57)a Abnormal chest radiographd No 1.0 Yes 1.36 (0.68–2.73) 0.41 (0.30–0.57)a Values in each cell indicate the odds ratio for sensitivity or specificity compared with a referent group. a p-value <0.05 for null hypothesis that odds ratio  = 1. b Excludes Shah et al. [24]. c Includes only studies Cain et al. [20], Shah et al. [24], Lawn et al. [26], and Chheng et al. [29]. d Includes only studies Cain et al. [20], Lewis et al. [23], Shah et al. [24], and Lawn et al. [26]. Table 5 shows the negative predictive value and the numbers needed to screen for the CFSW rule adjusted for individual- and study-level covariates. In a setting with 5% TB prevalence among people living with HIV, the rule has a negative predictive value of 98.3% (95% CI 97.5%–98.8%) for patients screened in a clinical setting and 97.3% (95% CI 96.9%–97.7%) for patients screened in a community setting. The numbers needed to screen at the same prevalence of TB are 15 and 11 for clinical and community setting, respectively. The negative predictive value was similar in those having high (≥200) and low (<200) CD4 count at 96.9% (95% CI 95.1%–98.0%) and 98.9% (95% CI 97.5%–99.5%), respectively (see also Table S3). 10.1371/journal.pmed.1000391.t005 Table 5 Negative predictive value (NPV) and number needed to screen (NNS) using rule CFSW in a hypothetical population of 1,000 people living with HIV stratified by study and individual level predictors. Participants 1% TB Prevalence 5% TB Prevalence 10% TB Prevalence 20% TB Prevalence NPV 95% CI NNS NPV 95% CI NNS NPV 95% CI NNS NPV 95% CI NNS All study participants 99.6 (99.5–99.6) 62 97.7 (97.4–98.0) 12 95.3 (94.6–95.9) 6 90.0 (88.6–91.3) 3 All study participants excluding miners 99.6 (99.5–99.7) 67 97.9 (97.5–98.2) 13 95.6 (94.8–96.3) 7 90.6 (89.0–92.1) 3 Setting Clinical 99.7 (99.5–99.8) 78 98.3 (97.5–98.8) 15 96.4 (94.8–97.5) 8 92.3 (89.0–94.6) 4 Community 99.5 (99.4–99.5) 55 97.3 (96.9–97.7) 11 94.5 (93.7–95.2) 5 88.5 (86.9–89.9) 3 Miners 99.2 (98.7–99.5) 38 96.1 (93.8–97.6) 8 92.2 (87.8–95.1) 4 84.0 (76.1–89.6) 2 Screening Nonscreened for TB 99.6 (99.5–99.7) 77 98.1 (97.5–98.5) 15 96.0 (94.9–96.9) 7 91.5 (89.2–93.3) 4 Prescreened for TB 99.3 (99.3–99.3) 36 96.5 (96.2–96.7) 7 92.8 (92.4–93.2) 4 85.1 (84.3–86.0) 2 Culture medium Liquid 99.6 (99.3–99.8) 75 98.2 (96.7–99.0) 15 96.2 (93.2–98.0) 7 91.9 (85.9–95.5) 3 Solid 99.5 (99.4–99.5) 57 97.3 (97.0–97.7) 11 94.6 (93.8–95.2) 6 88.5 (87.1–89.8) 3 Geography Southeast Asia 99.6 (99.2–99.8) 81 98,0 (95.9–99.0) 16 95,9 (91.6–98.0) 8 91.2 (83.0–95.6) 4 Sub-Saharan Africa 99.5 (99.4–99.6) 52 97.4 (97.1–97.8) 10 94.8 (94.0–95.4) 5 88.9 (87.5–90.2) 3 Age ≥33 y 99.6 (99.5–99.7) 63 97.8 (97.2–98.2) 12 95.4 (94.3–96.4) 6 90.3 (88.0–92.1) 3 <33 y 99.5 (99.4–99.6) 59 97.5 (97.0–97.9) 12 94.8 (94.0–95.6) 6 89.1 (87.4–90.6) 3 Gender Male 99.5 (99.4–99.6) 64 97.5 (97.2–97.9) 13 95.0 (94.2–95.6) 6 89.3 (87.8–90.6) 3 Female 99.6 (99.5–99.7) 60 98.0 (97.5–98.4) 12 95.8 (94.9,96.6) 6 91.0 (89.2–92.6) 3 CD4 cell count ≥200 cells/µl 99.4 (99.0–99.6) 80 96.9 (95.1–98.0) 16 93.6 (90.2–95.9) 8 86.7 (80.4–91.2) 4 <200 cells/µl 99.8 (99.5–99.9) 81 98.9 (97.5–99.5) 16 97.8 (94.8–99.1) 8 95.1 (89.1–97.9) 4 Abnormal chest radiograph Yes 99.4 (99.0–99.6) 83 97.0 (95.2–98.2) 16 93.9 (90.3–96.2) 8 87.2 (80.6–91.8) 4 No 99.5 (99.3–99.7) 61 97.7 (96.7–98.4) 12 95.2 (93.2–96.7) 6 89.9 (85.9–92.9) 3 Four studies [20],[23],[24],[26] consistently recorded information on chest radiograph, allowing screening rules with this sign to be evaluated using data from 2,805 participants The addition of abnormal chest radiographic findings into the CFSW rule increases the sensitivity to 90.6% (95% CI 66.7%–97.9%) with a specificity of 38.9% (95% CI 12.8%–73.3%), and a likelihood ratio negative of 0.242 (95% CI 0.102–0.571). Fifteen of the 23 rules included in our analysis outperform the symptom-based CFSW rule when abnormal chest radiographic findings are added (Table S4). On the basis of our meta-analysis findings and incorporating current WHO recommendations on provision of IPT, we developed a simple TB screening algorithm for public health programmes to screen people living with HIV, and, depending on the outcome of screening, to either provide IPT or evaluate patients further for TB or other diseases (Figure 4). 10.1371/journal.pmed.1000391.g004 Figure 4 Algorithm for TB screening in person living with HIV in HIV prevalent and resource-constrained settings. * Every person living with HIV needs to be evaluated for ART eligibility, and all settings providing care should reduce TB transmission through proper measures. ** Chest radiography is not required to classify patients into the TB and not-TB groups, but can be done, if available, to increase the sensitivity of screening. *** Assess for contraindications, including active hepatitis (acute or chronic), regular and heavy alcohol consumption, and symptoms of peripheral neuropathy, is required prior to initiating IPT. Past history of TB is not a contraindication for starting IPT. Tuberculin skin test may be performed as part of eligibility screening in some settings. **** Investigations for TB should be done in accordance with existing national guidelines. Discussion We found that the absence of all of current cough, fever, night sweats, and weight loss can identify a subset of people living with HIV who have low probability of having TB disease. This screening rule was superior over other candidate rules in eight of the nine studies included and had an overall favourable performance over competing rules in the hierarchical summary relative operating characteristic (HSROC) analysis. We also demonstrated that the negative predictive value of the rule was high across a range of TB disease prevalence estimates and across different population subsets, including those with low and high CD4 count, and those drawn from clinical and community settings and South African miners. We believe that these screening questions are likely to be acceptable to practitioners, because they are symptoms classically associated with TB disease. Underdiagnosis and delayed diagnosis of TB contribute to excess mortality among people living with HIV [17]. Similarly, concerns about the ability to reliably rule out active TB before initiating IPT have been a major barrier for wider use of this intervention. In the absence of a rapid and effective TB diagnostic tool available at the point-of-care, simple clinical algorithms must be used to screen people living with HIV for TB, dividing them into those in whom active TB is excluded and those who require further evaluation. This meta-analysis synthesizes the best available evidence for how to do this by relying on individual patient data of culture-confirmed TB cases from people living with HIV in the two regions of the world with the most severe burden of the TB and HIV dual epidemic. The major change to existing practice would be the replacement of chronic cough with current cough as a screening question and the addition of other symptoms to standard screening. National TB programs have traditionally defined a TB “suspect” as someone with cough lasting greater than 2 or 3 wk, and designed case-finding activities to investigate up to ten TB suspects for every TB case detected [36]. However, studies included in this analysis have shown that chronic cough is highly insensitive for TB disease in people living with HIV; using this symptom as a screening rule would miss cases and contribute to diagnostic delays [16],[20]. Using the combination of symptoms that we propose, in a population of people living with HIV with a 5% TB prevalence (excluding miners), requires the investigation of 13 extra patients for every TB case detected, a ratio of TB suspects to a TB case not much different from what national TB control programmes target in the general population. There has been ongoing debate about the importance of chest radiography in screening people living with HIV for IPT eligibility [37],[38]. Our analysis showed that the addition of abnormal chest radiography findings into the screening rule of CFSW increases the sensitivity of the rule by 11.7% (90.6% versus 78.9%) with a reduction of specificity by 10.7% (49.6% versus 38.9%). However, for example at a 5% TB prevalence rate among people living with HIV, augmenting the CFSW rule with abnormal chest radiographic findings increases the negative predictive value by a margin of only 1% (98.7% versus 97.8%), albeit with the same number of cases needed to be screened. On the other hand, the addition of abnormal chest radiographic findings to the rule at TB prevalence of 20% among people living with HIV increases the negative predictive value by almost 4% (94.3% versus 90.4%) without additional cases needed to be screened. It is also worth noting that the CFSW screening rule has higher sensitivity among those who presented into a clinical setting (90%) and among those who have not been previously screened for TB (88%). Our findings show that the utility of the proposed symptom-based screening rule have excellent performance in most settings with TB and HIV burden. However, the negative predictive value will fall in those settings with higher TB prevalence when symptom screening alone is used, as it depends on prevalence of disease. In particular settings (e.g., antiretroviral clinics with a very high TB burden [39]), consideration must be given to use of an algorithm that contains chest radiography, or even adding additional sensitive investigations (e.g., culture) while screening people living with HIV for TB [30],[40]. People living with HIV and receiving IPT should also be regularly screened for TB during their visit to a health facility or contact with health care provider so as to promptly detect active TB, if it develops. Programme managers need to weigh the financial, technical, and logistic difficulties, and patient cost and inconvenience associated with performing chest radiography or other additional sensitive investigations on all people living with HIV as part of a screening program compared with an approach that relies only on symptomatic screening. When interpreting our results, one must bear in mind that only a few variables were common to all studies included in the meta-analysis. It is possible that the addition of one or more symptoms not included in our list of common symptoms could have improved the performance of our proposed screening rule. However, at least one study included in our meta-analysis explored over 80 million combinations of about 100 signs and symptoms and found a symptom combination (cough and fever of any duration and night sweats for 3 weeks or longer), which was similar to the one we propose as the best performing one [20]. Furthermore, questions were not asked in a uniform manner across all studies, and the reporting of symptoms can be highly dependent on factors such as the quality of the interview and interviewer, the circumstances under which questions are asked, and the social and cultural factors that shape individual perceptions of symptoms and disease [41]. We reviewed all questions carefully with principal investigators and data managers to ensure accurate mapping of differently phrased questions to common variables. Our study relied on patients drawn from multiple countries and multiple settings, and the variation in the performance of the proposed screening rule across these different settings suggests that variation in patient self-report of symptoms is unlikely to have major impact, at least at the population level. In some studies, only one sputum specimen was collected for culture, while multiple cultures are required to maximize sensitivity. Some patients with TB may have been incorrectly classified as not having TB. Extrapulmonary TB is an important cause of morbidity and mortality in people living with HIV, but most studies included in the meta-analysis focused on screening for pulmonary TB. Young children were not included in the studies. We did not specifically look into the role of tuberculin skin test in the proposed screening rule. Ideally, the utility of the algorithm we propose, based on the screening rule from our meta-analysis, should be studied prospectively using a standardized protocol in multiple diverse sites; this is particularly important as the studies included in our analysis came from only two geographical regions of the world. Similarly, because of the time required for the data aggregation, statistical analysis, manuscript preparation, and publication, there was one potentially eligible study that was not included in our analysis [30]. We believe that the exclusion of this single study from South Africa, a country from which we have included similar studies already, will not affect the interpretation of our data and conclusions. In the future, as more studies are reported, particularly from other regions, it will be important to repeat the meta-analysis. Greatly improving TB screening, diagnosis, and treatment in people living with HIV will require deployment of a rapid, accurate, point-of-care TB diagnostic test. In the absence of such a test, we believe that a standardized algorithm employing symptoms, as we propose here, can improve the diagnosis and treatment of TB for people living with HIV, and by doing so would save many lives. Reliable exclusion of TB disease will facilitate safer initiation of antiretroviral therapy and will allow for broader use of IPT, which can substantially reduce TB incidence. Earlier and accurate HIV and TB screening and treatment may also help identify infectious cases earlier, thereby reducing both HIV and TB transmission. Evidence-based and internationally recommended guidelines should be used to expedite the diagnosis and treatment of TB in people living with HIV [42]. Supporting Information Figure S1 Hierarchical summary relative operating characteristic (HSROC) curves for the 23 candidate 1-of-n rules. (0.08 MB DOC) Click here for additional data file. Table S1 Phrasing of questions that were used in all studies to ask about five common symptoms. (0.09 MB DOC) Click here for additional data file. Table S2 Study-specific values and rankings of the sensitivity of each candidate screening rule in the nine studies included. (0.14 MB DOC) Click here for additional data file. Table S3 Diagnostic performance of all 23 candidate rules and number needed to screen in a hypothetical population of 1,000 people living with HIV stratified by TB prevalence among people living with HIV. (0.15 MB DOC) Click here for additional data file. Table S4 Diagnostic performance of 23 candidate rules that include abnormal chest radiograph and number needed to screen in a hypothetical population of 1,000 people living with HIV stratified by TB prevalence among people living with HIV. (0.15 MB DOC) Click here for additional data file.

Tuberculosis control relies on the identification and preventive treatment of individuals who are latently infected with Mycobacterium tuberculosis. However, direct identification of latent tuberculosis infection is not possible. The diagnostic tests used to identify individuals latently infected with M. tuberculosis, the in vivo tuberculin skin test and the ex vivo interferon-gamma release assays (IGRAs), are designed to identify an adaptive immune response against, but not necessarily a latent infection with, M. tuberculosis. The proportion of individuals who truly remain infected with M. tuberculosis after tuberculin skin test or IGRA conversion is unknown. It is also uncertain how long adaptive immune responses towards mycobacterial antigens persist in the absence of live mycobacteria. Clinical management and public healthcare policies for preventive chemotherapy against tuberculosis could be improved, if we were to gain a better understanding on M. tuberculosis latency and reactivation. This statement by the TBNET summarises knowledge and limitations of the currently available tests used in adults and children for the diagnosis of latent tuberculosis infection. In summary, the main issue regarding testing is to restrict it to those who are known to be at higher risk of developing tuberculosis and who are willing to accept preventive chemotherapy.