Accuracy of C-Reactive Protein for Tuberculosis Detection in General-Population Screening and Ambulatory-Care Triage in Uganda.

C-reactive protein (CRP) has demonstrated utility as a point-of-care triage test for tuberculosis (TB) in clinical settings, particularly among people with human immunodeficiency virus (HIV), but its performance for general-population TB screening is not well characterized. To assess the accuracy of CRP for detecting pulmonary TB disease among individuals undergoing community-based screening or presenting for evaluation of TB symptoms in Kampala, Uganda. We pooled data from two case-control studies conducted between May 2018 and December 2022 among adolescents and adults (⩾15 yr) in Kampala, Uganda. We conducted community-based screening for TB, regardless of symptoms. We enrolled people with Xpert MTB/RIF Ultra-positive (including trace) sputum results and a sample of people with Ultra-negative results. We also enrolled symptomatic patients diagnosed with TB and controls with negative TB evaluations from ambulatory care settings. Participants underwent further evaluation, including sputum culture, CRP, and HIV testing. We assessed the accuracy of CRP alone or with symptom screening against a bacteriologic reference standard. Our primary analysis evaluated the sensitivity and specificity of CRP at a cutoff of 5 mg/L. Diagnostic performance was summarized by calculating the area under the receiver operating curve (AUC). In the community setting ( = 544), CRP ⩾ 5 mg/L had a sensitivity of 55.3% (95% confidence interval, 47.0-63.4%) and specificity of 84.7% (79.7-88.8%) for confirmed TB; AUC was 0.75 (0.70-0.79). Screening for CRP ⩾ 5 mg/L or positive symptoms increased sensitivity to 92.0% (86.4-95.8%) at the expense of specificity (57.1% [50.8-63.2%]). In the ambulatory care setting ( = 944), sensitivity of CRP ⩾ 5 mg/L was 86.7% (81.8-90.7%), specificity was 68.6% (64.8-72.2%), and AUC (0.84 [0.81-0.87]) did not differ significantly by HIV status. CRP ⩾ 5 mg/L was >90% sensitive among individuals with a medium or high semiquantitative Xpert result in both settings. Although CRP did not meet World Health Organization (WHO) TB screening benchmarks in the community, it demonstrated high specificity, and sensitivity was high among individuals with high sputum bacillary burden who are likely to be most infectious. In ambulatory care, estimated sensitivity and specificity were each within 4 percentage points of WHO benchmarks, with no meaningful difference in performance by HIV status.