Identification of subclinical tuberculosis in household contacts. The relevance of tuberculosis exposure scores and contact investigations.

Background prioritizes diagnosis and treatment of incipient and subclinical TB, recently defined by key stakeholders as an “asymptomatic, early pre-clinical disease during which pathology evolves”. Regarded as indicative of a high risk of TB progression, considerable efforts have been made to identify these cases through exploration of biomarkers. The present study aimed to evaluate simple scoring systems for TB exposure as screening tools for subclinical TB, the only identifiable of the incipient and subclinical disease states, in a contact investigation (CI) setting of low HIV-prevalence. Methods Nested within a large prospective study in household contacts (HHCs) of smear positive pulmonary TB cases in South-India conducted 2010-2012, we assessed 1) the association between the Tuberculosis Contact Score (TCS) and the Infectivity Score with established tools for Mycobacterium tuberculosis (Mtb) infection corrected for established TB risk factors and 2) the capability of the TB exposure scores to identify subclinical TB defined by Mtb-culture positivity in sputum or gastric aspirate (subjects <5 years) specimen. Results Of 525 HHCs, 29 were Mtb-culture positive and 96.6% of these asymptomatic. The TCS and the Infectivity Score associated with positive Tuberculin Skin Test and QuantiFeron TB-Gold In-tube assay (QFT) results in multivariate analyses (TCS: ORTST 1.16, 95%CI: 1.01, 1.33; ORQFT 1.33 95%CI: 1.16, 1.51. Infectivity Score: ORTST 1.39, 95%CI: 1.10, 1.76; ORQFT 1.41 95%CI: 1.16, 1.71). The Infectivity Score showed a moderate capability to identify subclinical TB (AUC of 0.61, 95%CI: 0.52, 0.70).

Conclusion Although our results did not identify an easily applicable screening tool for subclinical TB, the present study indicates that focusing on TB-related symptoms in CI settings may be of limited value for early identification of HHCs with high risk for TB progression, one of the goals of the End TB Strategy.

Background
Globally, tuberculosis (TB) caused by Mycobacterium tuberculosis ( Mtb) is the largest killer among infectious diseases, causing an estimated 1.2 million deaths in 2018 (1). Despite declining TB incidence achieved through considerable joint efforts since the Stop TB Initiative (2), the TB epidemic will represent a great challenge for years to come: the estimated number of new TB cases was 10.0 million in 2018 (1).
Traditionally, treatment of cases has been the cornerstone of epidemiological control (2,3). In most high-endemic countries, case detection relies on patients seeking health-care because of symptoms, but this strategy leaves 40% of estimated TB cases undiagnosed (4). In order to reduce the TB prevalence further, The End TB Strategy included in the Sustainable developmental goals (SDG), emphasizes early diagnosis of cases (5) including incipient TB defined as "asymptomatic, early pre-clinical disease during which pathology evolves, such as mycobacterial replication or the inflammatory response. Radiological abnormalities or positive microbiological tests may or may not be present. This state may either evolve and lead to symptomatic clinical TB or regress and remain asymptomatic" (6). Meeting this ambitious goal will require systematic screening of contacts and high risk groups for TB disease and subsequent preventive or full-length TB treatment where adequate (5). Being recently exposed, Mtb-infected household contacts (HHCs) have increased risk of TB progression and are therefore defined a target population for active casefinding as part of contact investigations (CIs) (7). The World Health Organization (WHO) strongly recommends CIs by health staff visiting households following diagnosis of all smear positive pulmonary TB cases. In low-middle income countries (LMICs) with TB incidence ≥100 per 100 000 population, CIs include evaluation for active TB in persons of all ages with symptoms. If active TB is excluded, preventive treatment should be given to children aged <5 years and persons living with HIV (PLHIV) (8). Although suggested as a possible strategy to eliminate TB (9-11), giving preventive treatment to all Mtb-infected subjects is not practical in moderate-high TB incidence LMICs as health systems are already overwhelmed. Although the Tuberculin Skin Test (TST) and/or interferon-gamma release assays (IGRAs) give evidence of Mtb-infection, these tests have poor predictive value for future TB (12,13) reducing their relevance in the identification of incipient and subclinical TB (14). These disease states represent early TB disease with high risk of TB progression and are likely to contribute to continued transmission (6,14). Therefore, the World Health Organization (WHO) and other stakeholders are strongly encouraging and facilitating the development of novel tests capable of identifying incipient and subclinical TB (6,15). CIs represents a reasonable well-established framework for case identification, and accurate identification of these TB cases would have a large impact on the TB epidemic (9,16) and subsequently increase national incentives to adhere to programmatic CIs (8). Although the scientific community are making progress in identifying host biomarker-based risk signatures for TB progression (17)(18)(19), validation and approval according to defined target product profiles (6, 15) will still take time. Notably, risk factors for incipient/subclinical and active TB can be assumed to be similar (14,(20)(21)(22). The risk of Mtb-infection and disease progression is generally accepted to be strongly affected by the degree of exposure (23, 24). Therefore, a scoring system has been developed to quantify TB-exposure in HHCs when evaluating the performance of the TST and IGRAs in pediatric and adult populations (24-27). The Tuberculosis Contact Score (TCS) contains multiple subscores (Infectivity Score, Duration of Contact Score, Type of Exposure Score and Duration of Symptoms Score) and, as opposed to TST and IGRAs, does not require repeated visits and laboratory facilities. Sputum smear grade alone has also been found to associate with TST (28,29) and QFT positivity (30,31 (14), we considered laboratory confirmation a more robust evidence of ongoing Mtb-replication than chest X-ray changes. We therefore defined subclinical TB by growth of Mtb in respiratory (or gastric aspirate for children aged >5 years) specimens from recently exposed HHCs and argue that we hereby aim to identify the subjects with the highest bacterial burden suggestive of risk of transmission and progression to overt disease. Patient-reported symptoms are highly variable in previous Asian studies even in definite TB cases (32) and were therefore not included in our definition.
Thus, in the context of a prospective study of HHCs of adult pulmonary TB index cases in South India, our primary aim was to evaluate the potential of the TB exposure scores TCS and the Infectivity Score (a TCS subscore) as screening tools for subclinical TB. We first assessed the association between the TB exposure scores for identifying Mtb-infection (TST and Quantiferon) and subclinical TB while correcting for established TB risk factors. We then evaluated the capability of the TB exposure scores to identify individual contacts and/or families with ≥1 subclinical TB case.

Study population
The present study was a cross-sectional study nested within a large prospective years except for the muslim minority (81.4%) (33). The HIV prevalence in tested pregnant women in the aera is 1-2% (34). Index cases were recruited through the RNTCP (Revised National TB Control Program) when diagnosed with smear positive pulmonary TB (PTB). Both index cases and household contacts were enrolled in the HHC study: eligible PTB cases were smear positive subjects aged >18 years. Eligible contacts were persons living ≥75% of the time in the same household as the index case and sharing the same kitchen (35). Contacts with previous PTB or already on TB treatment/prophylaxis were excluded. Written informed consent was given by all adults. Children aged 7-18 years gave their assents followed by parental consent whereas parental consent alone was given for children aged <7 years.

Tuberculosis contact investigation
According to the study protocol, all participating contacts were offered a comprehensive evaluation at baseline. This included an interview on socioeconomic conditions, medical history, TB-symptoms and clinical examinations including weight, TST, the QuantiFeron TB-Gold In-tube assay (QFT), two sputa (gastric aspirates for children ≤5 years), harvested on two consecutive days, for smear and culture, and chest X-rays (anterior view for all, lateral view added on selected children <5 years). First, a blood sample was drawn for QFT, performed and interpreted according to the manufacturer's instructions (a positive test was defined as ≥0.35 IU/ml). Then a TST was performed by trained staff injecting two TU of PPD (SPAN Diagnostics Ltd, Surat, India) intra-dermally on the volar part of the left arm and read between 48-96h (~80% evaluated within 72h, the remaining within 96h).
An induration ≥10 mm was defined as positive. HHCs with an induration <5 mm (n=54) had a repeated test after 1-4 weeks (~96.3% evaluated within 72h, two subjects within 96h), and baseline TST results defined as the larger of the two tests.
The chest X-rays were interpreted as either "normal", "abnormal not TB" or "abnormal TB", first by a medical officer in the field, later by a radiologist whose interpretation was preferred if discrepancy. Sputum/gastric samples were evaluated by smear microscopy for acid-fast bacilli (AFB) and cultured on both liquid (BACTEC MGIT 960™ [Becton and Dickinson, USA]) and solid (Lowenstein-Jensen) media.
Identification of Mtb was done using the GenoType MTBC test kit (HAIN kit). All HHC participants were offered HIV-testing and pre-test counselling at baseline.

Score.
The Tuberculosis Contact Score (TCS) was based on previously published scores when assessing children (24) and adults (26), and modified to include HHCs of all ages. Relationship score was excluded as data were judged unreliable (Table 1). The TCS was based on interview (three questions) and sputum evaluation of the index cases assuming the gradient of exposure to be a composite function of: 1) the infectivity of the index case represented by the sputum smear grade, graded 0-6 (Infectivity Score), 2) closeness to the index case during sleep, graded 0-4 (Type of Exposure Score), 3) the time (hours per day) spent with the index case, graded 0-4 (Duration of Contact Score), and 4) duration of the index case's symptoms before first visit to the doctor, graded 0-4 (Duration of Symptoms Score). The TCS is a sum of all subscores (max score 18), the higher the score, the greater the exposure and risk of Mtb-infection and disease.

Mtb-infection status and subclinical TB
For analysis, HHCs were categorized according to their Mtb-infection status at baseline defined by 1) TST and QFT results (both tests required: TST and QFT negative; TST or QFT positive; TST and QFT positive) or 2) subclinical TB defined by positive Mtb-culture in sputum or gastric aspirate (subjects <5 years) specimen.
Although being asymptomatic is the key in the suggested definition for subclinical TB, broad reports from similar CI settings in Asia suggest that 40-79% of active TB cases do not report symptoms (32) highlighting the disease continuum and difficulty in drawing a strict line between subclinical, subclinical and active TB. After all, finding all these cases must be the goal of CIs. In accordance with the suggested subclinical TB definition (6) and the study aim of finding a screening tool applicable in the field, categorization of subclinical TB was done irrespective of chest X-ray results.

Statistical analysis
Categorical variables were reported as percentages. Continuous variables were summarized by means of mean and standard deviation or median and interquartile range, as appropriate. TST (mm) and QFT (IU/ml) results were analyzed both as continuous variables using median and range and as categorical variables dichotomized around their cut-offs (TST ≥10 mm) and (QFT ≥0.35 IU/ml).

Distribution of clinical characteristics between HHCs categorized according to Mtb-
infection status was assessed with Chi-square test and one-way ANOVA (Tukey "Honest Significant Difference" method for post-hoc comparisons), where appropriate. Associations between the independent variables TCS and established risk factors for TB (age, gender, BCG-scar, diabetes, smoking, indoor pollution, crowding) and the dependent variables 1) TST and QFT results were evaluated by univariate logistic regression, and multivariate logistic regression adjusted for previously listed risk factors for TB. Since assessment of body weight is different in subjects aged ≥15 years (Body Mass Index, BMI) and <15 years (Body Mass Index Zscore, BMIZ), BMI and BMIZ were only entered in the models evaluating this variable, which were then stratified for age. All other listed TB risk factors were entered in the BMI/BMIZ models. Similar models were fitted to the Infectivity Score.

Characterization of the study population
A total of 525 household contacts (HHCs) were recruited from 176 index cases in the study area accounting for 94.3 % of all eligible households. Index cases reported a median of 20 days (IQR: [15,30] days) from onset of symptoms to first doctor visit, and median of 6 days (IQR: [3,20] days) from doctor visit to diagnosis. At inclusion, 93 index cases (57%) had started treatment. In these subjects, the median time from diagnosis to initiation of treatment was 2 days (IQR: [2,4] days), and a median 2 days of treatment was given at the time of CI (IQR: [1,2] days). Baseline investigations were completed in 490 (93%) HHCs, in a median of 4 days (IQR: [3,5] days) in the 436 (83%) HHCs with only one TST needed and a median of 17 Among the 29 subclinical TB cases, one (5.5%) had symptoms, one had a positive smear and "abnormal TB" chest X-rays, one had "abnormal TB" and one "abnormal, not TB" chest X-ray.

Association between TB exposure scores and Mtb-infection in HHCs.
Although the association between TCS and TST/QFT is established in other cohorts (24-27), we first investigated the strength of this association in the present Indian cohort corrected for established risk factors for TB. The TCS and the Infectivity Score was associated with positive TST and QFT both in univariate and in multivariate analysis (Table 3). Table 3. See at the end of the manuscript.
In the multivariate models, the likelihood for a positive TST increased by 16%

Association between TB exposure scores and subclinical TB in HHCs.
There was no association between the TCS or the Infectivity Score for subclinical TB (Table 4).  (36). Notably, crowding was not significant in the BMI model, suggesting interaction between BMI and crowding. To explore this, the interaction BMI/crowding was entered in the model resulting in no significant association for crowding alone.

The capacity of TB exposure scores to identify subclinical TB.
Finally, we assessed the capability of the TB exposure scores to identify individual contacts and/or families with ≥1 contact with subclinical TB. For comparison, the same analyses were performed for TST and QFT at baseline CI. As our aim was to

Discussion
To our knowledge, the present study is the first to assess the Tuberculosis Contact Score (TCS) and the Infectivity Score as screening tools for subclinical TB in a CI setting. Currently, no tools are available for this purpose. Even with promising host biomarker-based risk signatures for TB progression (17)(18)(19), validation in various populations as well as translation to a point-of-care test format will take time (6,15). Therefore, the present study was motivated by our hope that simple scoring systems, previously proven to reflect TB exposure, a well-established risk factor for TB, could fill some of this gap. Being independent of laboratory and X-ray facilities, the TCS or the Infectivity Score can easily be applied as a screening tool resulting in more targeted CIs either by sorting out low-risk families/subjects where no CI could be justified, or identify high-risk families/subjects warranting referral for further TB investigations. Regrettably, the TCS could not reliably identify individual subjects or families with subclinical TB. The Infectivity Score performed better with a significant capacity to identify subjects and families with subclinical TB with an AUC of 0.61 and 0.58 respectively. This is however, not good enough for a screening test (37) to identify low-risk ("rule-out") or high-risk ("rule-in") families/subjects, underlining the need for continued search for host biomarkers for this screening purpose. in large-scale studies between smoking and TB disease due to passive smoking. The size of the households in the present study (median of 5) was similar to another HHC study in rural Indian (53). Semi-urban and urban Indian families tend to be smaller (41, 54) but congested living is common in both rural and urban areas (41) with 96.8 % of families in the present study living in 1-2 rooms. Interestingly, multivariate analyses revealed a possible interaction between BMI and crowding that could confound the results. Notably, children aged <15 could not be included in multivariate analyses assessing BMI as BMI is an invalid measurement in this age group. For HHCs aged ≥15 years, HHCs with and without subclinical TB had mean BMI of 18.7 (St.dev 2.13) and 19.6 (St.dev 3.25) respectively. This was considerably lower than the BMI in two large Peruvian HHC Study cohorts (mean BMI 25.2 and 25.6). Interestingly, the dose-response log-linear relationship between BMI and TB incidence reported in a meta-analysis, was less certain at BMI <18.5 and >30 (21).
The strengths of the present study are inclusion of HHCs of all ages, detailed sociodemographic data, a thorough work-up of most contacts with TST, QFT, sputum or gastric aspirate samples for both smear and cultures, combined with data on the TCS and Infectivity Score for all HHCs. A major weakness is the lack of 18-24 months follow-up data in accordance with current consensus definition for incipient TB (6).
Another limitation is due to the 18% missing data in the study. The missing data may be the reason that some associations may not have been detected. Moreover, as the missingness pattern is at least partly attributable to missing not at random mechanisms (e.g., different types of non-responders) there is a risk of bias in the reported associations.

Conclusions
Although our results were disappointing with regard to our hope of identifying an easily applicable screening tool for subclinical TB, the present study provides relevant information to CIs as framework for identification and early treatment of subclinical TB required to achieve the ambitious goal of the End TB Strategy (5). Although other studies support more targeted identification of subjects with high TB risk based on other TB risk assessment based framework (21, 22, 36), it is questionable whether exploring and validating eventual new TB risk scores offers a viable alternative to host biomarker based screening tools for subclinical TB, which are strongly needed to reach the ambitious goal of TB elimination.

Consent for publication
Not applicable.

Availability of data and materials
The datasets generated and/or analysed during the current study are not publicly available due to ongoing work on immune readouts in the same cohort but are available from the corresponding author on reasonable request.

Competing interests
The authors declare that they have no competing interests.     ¹Tuberculin Skin Test (TST) was dichotomized around the cutoff 10mm; ≥10 mm corresponding to TST+ and <10mm corresponding to TST-. Quantiferon (QFT) was dichotomized around the cutoff 0.35IU/mL; ≥0.35IU/mL corresponding to QFT+ and <0.35IU/mL corresponding to QFT-(Subjects with indeterminate QFT is excluded from analysis). ²Analysis including the TCS. ³Analysis including the Infectivity Score. ⁴WHO based age categories. ⁵Since assessments of body weight is different in ⁵ᵅsubjects aged ≥15 years (Body Mass Index, BMI) and ⁵ᵇ<15 years (Body Mass Index Z-score, BMIZ), BMI and BMIZ were only entered in the models evaluating this variable, which were then done stratified for age. ⁶Number of household members divided on rooms in the house. Figure 1 The capability of the Infectivity Score to identify individual household contacts (HHCs) with g