Among 9,166 adult household contacts, 4,506 (49%) were TST-positive at baseline; of those, 4,503 (99%) underwent symptom screening. Of the 3,610 persons with TB symptoms and 893 without symptoms at baseline, 1,489 (41%) persons with symptoms and 358 (40%) without symptoms underwent a chest radiograph at baseline. Those referred for chest radiograph were less likely to be male (37% vs. 45%; p = 0.001), HIV-positive (0.4% vs. 0.8%; p = 0.02), or have a previous history of TB disease (17% vs. 26%; p<0.001) (Appendix Table 1). The presence of TB symptoms was similar among those referred for radiograph and those not referred (19% vs. 20%; p = 0.54). Of the 1,847 subjects with a baseline radiograph, TB disease developed in 100 (5%) persons within 30 days after enrollment; we considered those to be coprevalent cases and excluded them from analysis (Figure 1).

In this study, we found that among persons >15 years of age who were exposed to TB at home, those who were initially deemed TB-free but had an abnormal baseline chest radiograph had a 15-fold higher risk for incident TB developing during 12 months of follow-up than did persons with an unremarkable radiograph. That finding was more pronounced in household contacts with no TB history and in younger participants, who were less likely to have residual abnormalities on radiograph from past TB or other comorbidities. Our findings remained unaltered in multivariate models when we further adjusted for demographic and clinical variables, as well as history of comorbidity. Among young persons with abnormal radiograph who were initially deemed TB-free at baseline, incident TB developed in 50% during follow-up. The severity grade of radiograph abnormalities was a strong indicator of an early-state TB trajectory in young persons.

Although numerous investigators have examined the use of chest radiography in screening for clinically apparent TB, few have evaluated whether abnormal radiographs could help identify early TB. The idea of using chest radiograph examination for incipient pulmonary TB was first proposed by Dr. Francis Williams, the father of chest radiology in the United States, in 1899 (23). Chest radiography for mass screening and active case-finding for pulmonary TB was recommended by the New York City Health Department in the United States as early as the 1920s–1930s (24,25) and was widely implemented in industrialized countries in the 1940s–1960s, contributing to the success of TB control in developed countries (3,26). However, by the late 1960s, WHO no longer recommended active case finding through mass radiograph screening for developing countries because of financial constraints (27). Not until 2015 was chest radiograph screening re-recommended by WHO as part of the end TB strategy (1,27). Even today, radiograph screening remains inaccessible in resource-limited areas in many low- and middle-income countries. Screening criteria typically focus on previously underdiagnosed TB disease, which is known as the late stage of TB disease (26). As a consequence, the very early stage of TB, known as subclinical TB, which is characterized by minimal or atypical presentations on chest radiograph, could still be underrecognized.

Nonetheless, our findings are consistent with several previous studies. In a study conducted in Czechoslovakia among a district-level cohort of 7,800 adults followed during 1961–1964 (28), 44% of persons in whom incident TB developed during follow-up had a previous abnormal chest radiograph indicating inactive TB or a fibrotic lesion. In a cohort study conducted during 1950–1952 in Denmark, persons with fibrotic lesions and shadows suspicious of TB on chest radiograph had a >10-fold higher annual rate of subsequent TB during the follow-up period (29). Similarly, in cohort studies conducted in 1988 among Southeast Asian refugees in Seattle (30) and in the early 1980s among recent Asian immigrants to northern Canada (31), persons with radiograph abnormalities indicative of or interpreted as inactive TB had 6-fold higher and 19-fold higher risk for subsequent TB. Those studies demonstrate that different types of lesions on baseline radiograph are associated with varying risks of subsequent TB in different mass screening settings. In a 1981 study from Hong Kong, 176 smear- and culture-negative adults with TB-compatible lesions were monitored for up to 30 months (32). Of those, TB was eventually diagnosed in 30%; most cases occurred within the first 12 months of follow-up. Similarly, a 1985 study conducted in South Africa found that among 152 TST-positive gold miners with evolving apical lung lesions on baseline radiograph who were smear- and culture-negative, bacteriologically confirmed pulmonary TB developed in 58% during a 58-month follow-up period (33). In a recent systematic review and meta-analysis of studies of 34 cohorts observed in the prechemotherapy era of persons with untreated TB who underwent follow-up, progression from microbiologically negative to positive disease was observed at an annualized rate of 10% in those with initial radiographs suggestive of active TB and at a rate of 1% in those initially suggestive of inactive TB (12). Although the incidence rate in this systemic review is lower than in our study, only 1 of the studies cited previously focused on a cohort of household contacts with known exposure to TB. The higher incidence in our study might be because we did not require microbiological confirmation to classify incident disease.

The finding that chest radiography might help identify pathological changes before the onset of TB symptoms aligns with recent evidence from South Africa. Esmail et al. (34) followed 250 HIV-negative adults exposed to TB at home for >4.7 years. Among 14 persons in whom active disease developed during follow-up, 6 exhibited baseline abnormalities on computed tomography/positron emission tomography (34). The TB research community now widely acknowledges that TB follows a dynamic continuum, encompassing TB infection, asymptomatic incipient or subclinical TB, and symptomatic clinically manifested disease (7–10). A 1982 study by the International Union Against Tuberculosis Committee on Prophylaxis revealed that among 28,000 adults with radiograph-identified fibrotic pulmonary lesions compatible with TB, a 24-week regimen of isoniazid preventive treatment decreased tuberculosis incidence by 65% compared with the placebo group over a 5-year follow-up period (35). Those findings suggest that lung changes can be observed in the asymptomatic phase of TB. Given that the cost of computed tomography/positron emission tomography makes it impractical for many screening programs, radiography could play a pivotal role in early TB detection, enabling early interventions such as close follow-up, preventive therapy, or full-dose combination therapy in populations at risk.

Our study revealed that young persons 16–24 years of age with abnormal baseline chest radiograph had a 3-fold higher risk for TB than older adults with abnormal radiograph. The observed association between the severity of baseline radiograph abnormalities and the trajectory of early TB disease within this young age group further supports the notion that an abnormal radiograph might serve as an indicator of undiagnosed or subclinical TB in this specific population, whereas older adults are more likely to have abnormal radiographs caused by non-TB lung diseases. That notion is supported by 2 cross-sectional studies conducted in Malawi during 2018–2020 (36) and the 2016 Kenya national TB prevalence survey (37). Both those studies reported a high prevalence of non-TB abnormalities on chest radiographs among older adults, with the proportion increasing with age.

The first limitation of this study is that only 40% of TST-positive adults received a baseline radiograph independent of symptoms; the decision to request a radiograph was made by local health center clinicians. Second, we lacked lateral positioning radiographs, which might have led to misclassifications of abnormalities. Third, readings were initially blinded for the first reader, and only abnormal radiographs were subsequently validated by a second reader who was not blinded. Therefore, misclassification of findings could have occurred. The fact that the second reader only read the films initially deemed to be abnormal and was not blinded to this classification might have resulted in overidentifying abnormalities in keeping with the first reader’s interpretations, leading to inflated measures of inter-rater agreement and consequently overestimating the effect size of the association between abnormalities and subsequent TB disease. Because the radiographs initially read as unremarkable were not read by a second reader, some of those might have been classified as abnormal if all radiographs had been read. Fourth, the relatively high prevalence of previous TB history among household contacts raises the possibility that some abnormal baseline radiographs were a result of inadequately treated previous TB or residual, minimally active TB. We did not have previous treatment or drug sensitivity information for those contacts and so cannot rule out the possibility that a history of TB might contribute to the elevated risk for incident TB among those persons. However, we also observed that persons with TB history were less likely to undergo baseline examinations. Moreover, when we excluded those with TB history from the 1,747 contacts who underwent radiograph screening, we found that the effect size increased by 30%. Finally, health center clinicians were expected to follow the clinical diagnosis guidelines outlined by Peru’s National TB Program, but different clinicians might have applied varying criteria for TB diagnosis. The absence of a standardized set of criteria for TB diagnosis might have resulted in nondifferential misclassification, potentially influencing the results toward a null effect.

Overall, our study revealed a significant association between abnormal baseline chest radiograph results and subsequent diagnosis of TB among asymptomatic adults with recent TB exposure. The severity of baseline radiograph results also reflected the early TB progression trajectory. Our findings strongly support the wide use of chest radiography with computer-aided detection as a screening tool in exposed adults and prompt the development of accurate criteria and algorithms for human and artificial intelligence readers to effectively identify early TB states beyond active TB disease, leading to timely intervention and proper management.