A score to predict and stratify risk of tuberculosis in adult contacts of tuberculosis index cases: a prospective derivation and external validation cohort study

Abstract

Background: Contacts of tuberculosis index cases are at increased risk of developing tuberculosis. Screening, preventive therapy, and surveillance for tuberculosis are underused interventions in contacts, particularly adults. We developed a score to predict risk of tuberculosis in adult contacts of tuberculosis index cases.

Methods: In 2002-06, we recruited contacts aged 15 years or older of index cases with pulmonary tuberculosis who lived in desert shanty towns in Ventanilla, Peru. We followed up contacts for tuberculosis until February, 2016. We used a Cox proportional hazards model to identify index case, contact, and household risk factors for tuberculosis from which to derive a score and classify contacts as low, medium, or high risk. We validated the score in an urban community recruited in Callao, Peru, in 2014-15.

Findings: In the derivation cohort, we identified 2017 contacts of 715 index cases, and median follow-up was 10·7 years (IQR 9·5-11·8). 178 (9%) of 2017 contacts developed tuberculosis during 19 147 person-years of follow-up (incidence 0·93 per 100 person-years, 95% CI 0·80-1·08). Risk factors for tuberculosis were body-mass index, previous tuberculosis, age, sustained exposure to the index case, the index case being in a male patient, lower community household socioeconomic position, indoor air pollution, previous tuberculosis among household members, and living in a household with a low number of windows per room. The 10-year risks of tuberculosis in the low-risk, medium-risk, and high-risk groups were, respectively, 2·8% (95% CI 1·7-4·4), 6·2% (4·8-8·1), and 20·6% (17·3-24·4). The 535 (27%) contacts classified as high risk accounted for 60% of the tuberculosis identified during follow-up. The score predicted tuberculosis independently of tuberculin skin test and index-case drug sensitivity results. In the external validation cohort, 65 (3%) of 1910 contacts developed tuberculosis during 3771 person-years of follow-up (incidence 1·7 per 100 person-years, 95% CI 1·4-2·2). The 2·5-year risks of tuberculosis in the low-risk, medium-risk, and high-risk groups were, respectively, 1·4% (95% CI 0·7-2·8), 3·9% (2·5-5·9), and 8·6%· (5·9-12·6).

Interpretation: Our externally validated risk score could predict and stratify 10-year risk of developing tuberculosis in adult contacts, and could be used to prioritise tuberculosis control interventions for people most likely to benefit.

Funding: Wellcome Trust, Department for International Development Civil Society Challenge Fund, Joint Global Health Trials consortium, Bill & Melinda Gates Foundation, Imperial College National Institutes of Health Research Biomedical Research Centre, Foundation for Innovative New Diagnostics, Sir Halley Stewart Trust, WHO, TB REACH, and Innovation for Health and Development.

Figure 1. Characteristics of populations and study areas for the derivation and internal validation cohort and the external validation cohort

Statistics are reported by the Peruvian Instituto Nacional de Estadistica e Informatica unless indicated otherwise. *Collected collaboratively from government-run health posts.

Figure 2. Study profile

Figure 3. A score to predict risk of tuberculosis in adult contacts of index cases

(A) An example risk score for field use. (B) Predicted 10-year risk of tuberculosis plotted against risk scores. (C) Numbers needed to treat with preventive therapy to prevent one case of tuberculosis among contacts, by risk group.

Figure 4. Cumulative observed risk of tuberculosis among contacts, stratified by risk group

(A) Ventanilla derivation cohort (n=2017). (B) Callao validation cohort (n=1910). Data are derived from Kaplan-Meier functions. Community risk was defined by the average tuberculosis case notification rate during corresponding years, corrected by 20% to assume under-reporting of cases treated outside the public system, as is the local practice.

Figure 5. TST results and 10-year observed risk of tuberculosis in the Ventanilla derivation cohort

(A) TST results among contacts, stratified by risk group. (B) Observed 10-year risk of tuberculosis stratified by TST results and risk group. Tuberculosis risk was not significantly different within risk groups when stratified by TST result (negative, unknown, or positive), but was significantly different between risk groups for each TST result. Error bars represent 95% CIs. Data are derived from Kaplan-Meier functions. The p values represent log-rank tests for equality of survival functions. TST=tuberculin skin test.