Effectiveness of preventive treatment among different age groups and Mycobacterium tuberculosis infection status: a systematic review and individual-participant data meta-analysis of contact tracing studies

Abstract

Background: Tuberculosis is a preventable disease. However, there is debate regarding which individuals would benefit most from tuberculosis preventive treatment and whether these benefits vary in settings with a high burden and low burden of tuberculosis. We aimed to compare the effectiveness of tuberculosis preventive treatment in exposed individuals of differing ages and Mycobacterium tuberculosis infection status while considering tuberculosis burden of the settings.

Methods: In this systematic review and individual-participant meta-analysis, we investigated the development of incident tuberculosis in people closely exposed to individuals with tuberculosis. We searched for studies published between Jan 1, 1998, and April 6, 2018, in MEDLINE, Web of Science, BIOSIS, and Embase. We restricted our search to cohort studies; case-control studies and outbreak reports were excluded. Two reviewers evaluated titles, abstracts, and full text articles for eligibility. At each stage, two reviewers discussed discrepancies and re-evaluated articles until a consensus was reached. Individual-participant data and a pre-specified list of variables, including characteristics of the exposed contact, the index patient, and environmental characteristics, were requested from authors of all eligible studies; contacts exposed to a drug-resistant tuberculosis index patient were excluded. The primary study outcome was incident tuberculosis. We estimated adjusted hazard ratios (aHRs) for incident tuberculosis with mixed-effects Cox regression models with a study-level random effect. We estimated the number-needed-to-treat (NNT) to prevent one person developing tuberculosis. Propensity score matching procedures were used in all analyses. This study is registered with PROSPERO (CRD42018087022).

Findings: After screening 25 358 records for eligibility, 439 644 participants from 32 cohort studies were included in the individual-participant data meta-analysis. Participants were followed for 1 396 413 person-years (median of 2·7 years [IQR 1·3-4.4]), during which 2496 people were diagnosed with incident tuberculosis. Overall, effectiveness of preventive treatment was 49% (aHR 0·51 [95% CI 0·44-0·60]). Participants with a positive tuberculin-skin-test (TST) or IFNγ release assay (IGRA) result at baseline benefitted from greater protection, regardless of age (0·09 [0·05-0·17] in children younger than 5 years, 0·20 [0·15-0·28] in individuals aged 5-17 years, and 0·17 [0·13-0·22] in adults aged 18 years and older). The effectiveness of preventive treatment was greater in high-burden (0·31 [0·23-0·40]) versus low-burden (0·58 [0·47-0·72]) settings. The NNT ranged from 9 to 34 depending on age among participants with a positive TST or IGRA in both high-burden and low-burden settings; among all contacts (regardless of TST or IGRA test result), the NNT ranged from 29 to 43 in high-burden settings and 213 to 455 in low-burden settings.

Interpretation: Our findings suggest that a risk-targeted strategy prioritising contacts with evidence of M tuberculosis infection might be indicated in low-burden settings, and a broad approach including all contacts should be considered in high-burden settings. Preventive treatment was similarly effective among contacts of all ages.

Funding: None.

Figure 1:

Flowchart of the systematic search process

Figure 2:. The effect of preventive treatment on the incidence of tuberculosis, stratified by age and Mycobacterium tuberculosis infection status

Each individual row represents an independent mixed-effects model stratified by the given row characteristic. Each model uses a propensity score using a nearest neighbour matching algorithm with matching based on individual-level covariates of age, sex, study design, and previous tuberculosis. A preventive drug treatment regimen was defined as any preventive drug regimen given at the baseline visit of the contact tracing intervention. Preventive treatment was administered to contacts at the discretion of each study site, and we accepted each study’s decision to administer preventive treatment. Completion of preventive treatment was not reported for most studies. IGRA=IFNγ release assay. TST=tuberculin-skin-test.

Figure 3:. Effectiveness of preventive treatment in settings with a high-burden and low-burden of tuberculosis

We defined tuberculosis burden as high if the study was conducted in a country with a background tuberculosis incidence above 100 cases per 100 000 population at the time of the study, and low if the incidence was below 100 cases per 100 000 population. Each individual row represents an independent mixed-effects model stratified by the given row characteristic. Each model uses a propensity score with a nearest neighbour matching algorithm for matching based on individual-level covariates of age, sex, study design, and previous tuberculosis. A preventive drug treatment regimen was defined as any preventive drug regimen given at the baseline visit of the contact tracing intervention. Preventive treatment was administered to contacts at the discretion of each study site, and we accepted each study’s decision to administer preventive treatment. Completion of preventive treatment was not reported for most studies. IGRA=IFNγ release assay. TST=tuberculin-skin-test.

Figure 4:. Estimated number-needed-to-treat to prevent one person from developing tuberculosis among people with differing Mycobacterium tuberculosis infection status and background tuberculosis burden

We defined tuberculosis burden as high if the study was conducted in a country with a background tuberculosis incidence above 100 cases per 100 000 population at the time of the study, and low if the incidence was below 100 cases per 100 000 population. We calculated a number-needed-to-treat for young children (age <5 years), older children (age 5–17 years), adults (≥18 years), and all ages. We estimated the number-needed-to-treat to prevent one person from developing tuberculosis under distinct situations and groups of participants by calculating a pooled absolute risk difference through mixed-effects multivariable models and study-level random effects for each specific group of interest. The number-needed-to-treat was estimated as the reciprocal of the absolute risk differences between the incidence of groups that were prescribed and not prescribed preventive treatment after adjustment through multivariable modelling. We repeated these analyses among participants that tested IGRA or TST positive (right hand panel) or unrestricted by test results (left hand panel). We further stratified these analyses by the entire population (red), study results from high-burden countries (blue), and study results from low-burden countries (green). We did not include IGRA results alone in this figure due to limited sample size after stratification by age and background burden. IGRA=IFNγ release assay. TST=tuberculin-skin-test.