We performed a retrospective cohort study of children <5 years of age referred to pediatric TB clinics in Houston, Texas, USA, during 2007–2017. Our clinic (Children’s Tuberculosis Clinic at Texas Children’s Hospital, Houston, Texas, USA) is the main referral source for children seen in 12 counties of the greater Houston metropolitan area (population ≈7 million), where TB incidence is twice the national average (6). We included all infants <6 months of age who had recent contact with an index TB patient, as well as all other children <5 years of age who had recent contact with an index patient and initial negative diagnostic test results (TST <5 mm or negative IGRA), no symptoms of TB disease, and an unremarkable 2-view chest radiograph and who were started on window prophylaxis. We excluded children whose contacts were later determined to not have TB or to have solely extrapulmonary TB.

Children are within the window period until 8–12 weeks after contact with the index patient has ceased, either by physical separation or effective treatment of the index patient (generally reflected by 3 consecutive negative results on acid-fast bacilli [AFB] sputum smear testing) (7). Health department workers performed a TST or obtained blood for IGRAs in the home of the child as soon as the child’s exposure was recognized and 8–12 weeks after contact was broken. In the interim, isoniazid window prophylaxis was offered to exposed children twice weekly (20–30 mg/kg/dose, maximum 900 mg/d) under directly observed preventive therapy (DOPT) in the child’s home by health department workers, as previously described (8). If the index case isolate was resistant to isoniazid or isoniazid was contraindicated, healthcare workers administered rifampin daily (10–20 mg/kg/dose, maximum dose 600 mg/d). Children were seen in the clinic at the initiation of therapy and every 1–2 months thereafter. We abstracted medical records for demographic variables, epidemiology and microbiology results for the index patient, medication regimens, and adverse events (AEs). We obtained institutional review board approval from Baylor College of Medicine (Houston, Texas, USA).

The primary outcome was the proportion of families reporting AEs for their children while they were on window prophylaxis. All children were assessed for signs and symptoms of hepatotoxicity at each DOPT medication visit. Children with abdominal pain, vomiting, anorexia, icterus, or weight loss had their medications withheld pending clinical and laboratory evaluation. Otherwise healthy asymptomatic children did not have baseline or serial liver testing performed. We used the National Cancer Institute guidelines to grade AE severity (9). The secondary outcomes were the proportion of infants with an initially negative test of infection in which TB infection or disease subsequently developed while on treatment and the epidemiologic factors associated with TST or IGRA conversion. We considered TST results of ≥5 mm of induration and QuantiFERON (QIAGEN, https://www.qiagen.com) results ≥0.35 international units/mL positive (10,11). We performed a priori analyses to evaluate test conversion in children <2 years and ≥2 years of age because children in the first 2 years of life are at the highest risk for disease progression with untreated TB infection (2). We compared proportions using odds ratios (ORs) with 95% CIs and expressed continuous variables as medians with interquartile ranges (IQRs). We used Stata 14 (StataCorp, https://www.stata.com) for analyses.

We found that most families accepted their young children being treated for TB exposure. Therapy was initiated and completed for most children, most of whom were treated with twice-weekly DOPT with isoniazid. This treatment was safe; no child developed hepatotoxicity. Progression from exposure to infection was uncommon and not associated with epidemiologic factors typically associated with transmission of M. tuberculosis. Health department workers predominantly used TSTs for testing and isoniazid for window prophylaxis and only infrequently used IGRA and rifamycin-based regimens.

Prior studies have shown that up to 50% of children do not begin or complete isoniazid therapy for TB infection when children or other family members administer the drug (5,12,13). In contrast, families of TB-exposed children felt comfortable directly observing the administration of medication to their young, asymptomatic children who had unremarkable examination findings and chest radiograph results and negative tests of infection. Their acceptance of this therapy might reflect them having known a symptomatic contact (14) but, we hope, was also supported by our counseling. Directly observed therapy (DOT), which removes barriers to medication acquisition and administration, is often used to optimize adherence (13) and enables close monitoring for drug toxicities. We had a low threshold to change drug regimens in the event of any AE to facilitate continuation of window prophylaxis.

In this study, we did not find an association between TST conversion and AFB sputum smear status or cohabitation with the index patient, both found to be risk factors for acquisition of TB infection in other studies (15,16). Several factors could help explain this discrepancy. First, baseline smear positivity is not an accurate proxy of the degree of infectiousness. Other factors, such as intensity of exposure, room air exchanges, radiographic findings, sputum viscosity, and aerosolization of viable bacteria, also affect transmission (17,18). In fact, some data suggest that adults with high concentrations of bacteria in their smears might not transmit TB efficiently because they are too weak to produce a vigorous cough (19). Second, infected persons are not equally contagious throughout the natural history of infection, and discerning when during the index patient’s course of infection the exposure to the child occurred is often difficult. Third, children were defined as household contacts if they resided in the same home as the index patient. However, some children spent most of their waking hours in another home cared for by a nonparent index patient, in which case, the child was defined as a nonhousehold contact. A rigid definition of what constitutes a household contact lacks epidemiologic meaning (20). Another finding from our study was that epidemiologic and microbiological factors were inadequate to risk stratify the children at higher risk of progressing from TB exposure to infection. As such, we feel that all exposed young children need to be evaluated with a physical examination, chest radiograph, and a test of infection and that all families should be offered window prophylaxis for their children. Of note, a TB exposure risk scoring system has been validated for children in settings of TB hyperendemicity (21). Although this system was not validated in low-incidence settings, such as the United States, discussing some of the scoring system variables associated with increased risk for exposure (maternal TB, sleep proximity, and duration of exposure) with families reluctant to initiate therapy might be useful.

Our primary diagnostic test for exposed children was TSTs, and treatment was primarily with isoniazid. The decision to use this testing and treatment strategy was driven by several considerations: a paucity of data on IGRAs and other treatment regimens in young children, ease of administration, and cost. The ability to place and read a TST in the home, rather than referring a child to a laboratory for an IGRA, has cost and convenience advantages for families. Given the large number of children evaluated for TB annually, using lower cost tests of infection and medications enabled health departments to provide services to more patients. Video DOT (i.e., the secure uploading and transmission of videos to the health department showing the child taking medication) enables remote monitoring of adherence and toxicity and is less costly than standard DOT (22). Rifampin is considerably more expensive than isoniazid, and data do not support giving rifampin less frequently than daily, making DOPT with this drug difficult and expensive. The safety and efficacy of giving window prophylaxis with either medication by DOPT or by family member administration has not been compared in any studies.

Routine use of rifampin for window prophylaxis is reasonable in certain communities where isoniazid-resistant M. tuberculosis isolates are highly prevalent. In the United States, ≈8% of isolates are isoniazid resistant and rifampin susceptible (6). In our clinic, efficient communication with our health department partners facilitated prompt changing of regimens on the basis of source case drug susceptibility results. Another possible use for rifampin may be for exposed young infants, where concern about TST anergy can result in children receiving therapy for much longer than 8–12 weeks. For these children, 4 months of rifampin might be provided before physicians feel comfortable relying on the definitive TST.

For the children whose test of infection converted to positive, indicating TB infection, we chose to continue twice-weekly isoniazid and complete the 9-month regimen. Considering the data on safety, effectiveness, and adherence to 4-month regimens of daily rifampin (among children) and 3-month regimens of weekly isoniazid and rifapentine (for children ≥2 years of age) (4,23), changing to 1 of these shorter regimens seems reasonable. Whether the duration of either of these treatments could be shortened in the event of previous isoniazid window prophylaxis is unclear, but giving the infected child the full course of either regimen is most reasonable.

This study had some limitations. Not all variables were documented for all children. Children could have moved outside the area and TB disease could have developed thereafter. Our study was not powered to identify exposure risk factors for TST conversion; with only a 4.9% conversion rate, we might not have had enough cases to adequately assess the influence of certain putative risk factors. Almost all children had serial TSTs and not IGRAs; thus, we cannot rule out the possibility of boosting in the children who had TST conversions. Radiographic findings for the index patient were not uniformly available from local health departments; as such, we could not correlate cavitary disease or other forms of multibacillary disease with the risk for infection in children. Quantifying the duration of exposure was impossible; instead, we used living within the same or different household as a proxy. Timing of when contact was broken for children who were no longer around the contagious person was subject to recall bias. The study was not powered to evaluate efficacy of treatment. However, given the age of the children who had TST conversions (2), we estimate that 4–5 children would have progressed to TB disease if untreated. Last, because window prophylaxis is a recommended intervention, performing a randomized controlled trial comparing treatment and no treatment would have been unethical. The administration of window prophylaxis potentially could have prevented the establishment of TB infection in some children; thus, our study potentially underestimates the effect of window prophylaxis in the prevention of infection and disease.

In conclusion, ≈5% of TB-exposed children given prophylaxis progressed to TB infection, and available epidemiologic or microbiological data could not accurately identify which children might receive more benefit from window prophylaxis. However, given the safety of the medications and the acceptability of treatment by families, the benefits of therapy greatly outweigh the risks in this vulnerable patient population.