To review evidence about targeted screening for and treatment of latent tuberculosis infection (LTBI) among adults in primary care settings.
MEDLINE, the Cochrane Library, and trial registries through August 3, 2015; bibliographies from retrieved articles, outside experts, and reviewers, with surveillance of the literature through May 31, 2016.
Two investigators independently selected studies using a priori inclusion and exclusion criteria. We selected studies that evaluated the tuberculin skin test (TST) using the Mantoux method or tests evaluating commercial interferon-gamma release assays (IGRAs). We selected trials of treatment that evaluated pharmacotherapy regimens that are currently recommended by the Centers for Disease Control and Prevention for the treatment of LTBI for synthesis of benefits and harms. We excluded studies of persons with underlying immunosuppression and for whom LTBI screening and treatment would be part of standard disease management by specialty care providers (e.g., persons with HIV, history of or planned organ transplant, or planned or active use of tumor necrosis factor-alpha inhibitors). We excluded poor-quality studies, studies assessing specificity in countries with a high tuberculosis (TB) burden, and studies assessing harms and benefits in developing countries.
One investigator extracted data and a second checked accuracy. Two reviewers independently rated quality for all included studies using predefined criteria.
We did not identify any studies that compared screening with no screening. We included 72 studies of fair to good quality; 67 assessed test accuracy or reliability and five assessed benefits and harms of treatment. Pooled estimates for sensitivity of TST at both the 5-mm and 10-mm induration thresholds for positivity were 0.79; the pooled estimate at the 15-mm threshold was 0.52. Pooled estimates for sensitivity of IGRA tests ranged from 0.77 to 0.90. Estimates for specificity of TST at the 5-mm threshold varied considerably by TB burden of the study setting (0.94 to 0.97 in low-burden countries and 0.30 in an intermediate-burden country). Pooled estimates for specificity at the 10-mm and 15-mm thresholds were 0.97 and 0.99, respectively. Pooled estimates for specificity of IGRA tests ranged from 0.95 to 0.98. We found evidence for at least moderate interrater reliability for both TST and IGRA tests.
The best evidence on effectiveness of treatment of LTBI was from the International Union Against Tuberculosis (IUAT) trial, a large (N=27,830) good-quality randomized, controlled trial (RCT) that evaluated multiple treatment durations for daily isoniazid. It found a relative risk (RR) for progression to active TB at 5 years of 0.35 (95% confidence interval [CI], 0.24 to 0.52) for 24 weeks of isoniazid compared with placebo (N=13,955; number needed to treat, 112). Our sensitivity analyses adding four RCTs that did not meet all of our eligibility criteria (e.g., compared isoniazid with placebo using a longer duration of treatment or used different doses than currently recommended) found an RR of 0.31 (95% CI, 0.24 to 0.41; I2=0%; 5 RCTs, N=36,823). A head-to-head, open-label, noninferiority RCT that compared a combination of once-weekly rifapentine plus isoniazid for 3 months with daily isoniazid for 9 months found the combination therapy to be noninferior to isoniazid alone for preventing the development of active TB.
For harms, the IUAT trial reported an RR for hepatotoxicity of 4.59 (95% CI, 2.03 to 10.39; number needed to harm [NNH], 279) for 24 weeks of isoniazid compared with placebo. Sensitivity analyses pooling the IUAT with three RCTs that used a longer duration of isoniazid yielded a similar result (pooled RR, 5.04 [95% CI, 2.50 to 10.15]; I2=0%; 4 RCTs, N=35,161). The RR of treatment discontinuation because of adverse effects across all treatment duration arms in IUAT was 1.50 (95% CI, 1.18 to 1.89; N=27,830; NNH, 167). For isoniazid compared with rifampin, the pooled RR for hepatotoxicity was 3.29 (95% CI, 1.72 to 6.28; I2=0%; 3 RCTs, N=1,327) and the pooled RR for treatment discontinuation because of adverse events was 1.61 (95% CI, 0.57 to 4.57; I2=40.0%; 3 RCTs, N=1,327).
No test for the direct diagnosis of LTBI exists; thus, studies of test accuracy use subjects with confirmed active TB to establish sensitivity and healthy, low-risk subjects to establish specificity. Thus, applicability to other populations is uncertain. The single trial meeting all eligibility criteria that established the benefits of a currently recommended treatment (isoniazid 300 mg daily for 24 weeks) for preventing active TB was published more than 30 years ago and was conducted among subjects with pulmonary fibrotic lesions; whether it may overestimate the benefits of treatment for populations with lower risk for progression is not clear. No trials evaluated the effectiveness (compared with placebo) of regimens other than isoniazid. Contemporary treatment studies have not included placebo arms; when available, information on benefits and harms of newer treatments are derived from comparative studies (vs. isoniazid). The evidence on harms is limited by heterogeneous specification of outcomes across studies. This review is not applicable to persons with the highest risk, for whom testing and treatment is considered part of disease management or public health surveillance.
We did not find any studies evaluating the direct benefits and harms of screening for LTBI in the adult populations and settings included in this review. Both types of currently available tests (TST and IGRA) are moderately sensitive and, within countries with a low TB burden, highly specific. Isoniazid treatment reduces the risk of progression to active TB in persons with LTBI and pulmonary fibrotic lesions. The evidence is limited or not available for other regimens and outcomes (e.g., deaths due to TB, all-cause mortality) among the populations included in this review. Isoniazid is associated with higher rates of hepatotoxicity than placebo and rifampin regimens.