Influence of setting and diagnostic algorithm on disease severity among people diagnosed with symptomatic and asymptomatic tuberculosis in South Africa

Problem Statement
Traditional tuberculosis (TB) control in low- and middle-income settings has relied on passive case finding, where individuals self-present to clinics upon recognizing symptoms. However, with global TB incidence declining slowly, there is a growing shift toward active case finding (ACF) and community screening, which increasingly detects asymptomatic TB (aTB). This evolving landscape raises critical epidemiological questions: How does the setting of diagnosis (clinic vs. community) and the specific screening algorithm used influence the spectrum of disease severity and infectiousness? Specifically, it remains unclear whether reported symptoms reliably indicate disease severity or if the context of investigation (self-presentation vs. active screening) is a more informative determinant of the disease profile.

Methodology
This descriptive epidemiological study utilized data from two cross-sectional studies conducted in the same South African community (uMkhanyakude district, KwaZulu-Natal):

1. Community Survey (Vukuzazi, 2018–2019): A multimorbidity study involving 18,041 adults. Participants underwent symptom screening and chest radiography (CXR) read by computer-aided detection (CAD4TB v5). Sputum testing (Xpert MTB/RIF Ultra and culture) was performed based on symptoms, CAD scores (thresholds varied from 60 to 25), or radiologist findings.
2. Clinic-Based Study (2022–2023): Recruitment of adults starting treatment for bacteriologically confirmed pulmonary TB in local clinics. Participants provided CXRs read by CAD4TB v5 and symptom data.

Definitions and Metrics:

• TB Case Definition: Positive Xpert Ultra result (including "trace") and/or culture.
• Symptomatic TB (sTB) vs. Asymptomatic TB (aTB): Defined based on the WHO 4-symptom screen (cough, fever, night sweats, weight loss).
• Severity/Infectiousness Indicators: CAD scores derived from CXRs and the proportion of Xpert results exceeding "trace."

The authors compared these indicators between clinic-diagnosed and community-diagnosed sTB and aTB, stratified by HIV and antiretroviral therapy (ART) status. Furthermore, they modeled hypothetical screening scenarios (varying CAD thresholds and symptom criteria) to estimate how algorithmic choices alter the severity profile of detected cases. They also used bootstrapping to estimate the prevalence and severity of TB in the untested population (CAD <25 and no symptoms).

Key Results

• Setting Differences: People diagnosed with sTB in clinics exhibited significantly higher CAD scores and a higher proportion of Xpert results exceeding "trace" compared to those diagnosed with sTB in the community. Conversely, community-diagnosed aTB had higher CAD scores than clinic-diagnosed aTB.
• Symptom Reliability in Community: Within the community survey, there was minimal difference in CAD scores or Xpert results between sTB and aTB. This suggests that among those actively screened, reported symptoms do not reliably distinguish disease severity.
• Algorithmic Influence:
  • Under a hypothetical "universal testing" scenario (no screening step), sTB was estimated to have higher severity than aTB, though differences were not statistically significant in this dataset.
  • Under stricter screening algorithms (e.g., requiring CAD scores ≥50 and/or symptoms), the median CAD scores for detected aTB were higher than those for sTB. This occurs because strict thresholds filter out low-severity symptomatic cases while capturing higher-severity asymptomatic cases that meet the radiological threshold.
• HIV Status Stratification: The disparity in severity between clinic and community sTB was most pronounced among HIV-negative individuals. Among people living with HIV on ART, the severity profiles were more comparable between settings, likely due to routine active screening in ART clinics.

Key Contributions

• Re-evaluation of Symptom-Based Stratification: The study provides evidence that symptom status alone is not a reliable indicator of TB disease severity, particularly in community screening contexts where the spectrum of disease is broad.
• Contextual Determinants: The paper demonstrates that the setting (clinic vs. community) and the diagnostic algorithm (thresholds used) are primary determinants of the observed disease spectrum. Findings from clinic-based sTB cannot be generalized to the full spectrum of sTB in the community.
• Algorithm Sensitivity: The research highlights how screening design (e.g., raising CAD thresholds) can inadvertently select for more severe aTB than sTB, challenging the assumption that aTB is inherently less severe.

Significance and Claims
The authors conclude that the traditional dichotomy between symptomatic and asymptomatic TB is insufficient for characterizing disease severity. Instead, the "context of investigation" and "diagnostic algorithm design" are key factors shaping the spectrum of identified disease.

The paper claims that:

1. Generalizability is Limited: Data on diagnostic performance or treatment needs derived from clinic-based sTB patients may not apply to community-diagnosed sTB or aTB due to significant heterogeneity in disease severity.
2. Need for Nuanced Frameworks: Future research and policy should move beyond simple symptom-based classifications toward frameworks that integrate diagnostic context and objective severity measures (like CAD scores).
3. Cautious Interpretation: Comparisons of infectiousness or severity between aTB and sTB across different studies must account for the specific screening algorithms and settings used, as these factors heavily influence the results.

The authors remain modest regarding their claims, acknowledging limitations such as the extrapolation of universal testing results, potential biases in symptom reporting timing, and the subjective nature of symptom recall. They emphasize that their findings highlight the need for clear descriptions of identification methods in TB studies rather than proposing immediate clinical practice changes.