Open AccessEvaluation of the diagnostic performance of laboratory-based c-reactive protein as a triage test for active pulmonary tuberculosis
Author(s) -
Thomas H. A. Samuels,
Romain Wyss,
Stefano Ongarello,
David Moore,
Samuel G Schumacher,
Claudia M. Denkinger
Publication year - 2021
Publication title -
plos one
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.99
H-Index - 332
ISSN - 1932-6203
DOI - 10.1371/journal.pone.0254002
Subject(s) - triage , medicine , receiver operating characteristic , tuberculosis , sputum , cutoff , c reactive protein , referral , human immunodeficiency virus (hiv) , immunology , emergency medicine , pathology , family medicine , physics , quantum mechanics , inflammation
A highly sensitive triage test that captures most symptomatic patients at increased likelihood of having pulmonary tuberculosis (PTB) would ‘rule-out’ lower-risk patients from expensive confirmatory testing. Although studies have assessed the diagnostic accuracy of a C-reactive protein (CRP) triage test for PTB in HIV+ patients, limited data are available from HIV- cohorts. Materials and methods In this retrospective case-control study, 765 serum samples were selected from FIND’s biobank. Each sample had been collected from an adult presenting with respiratory symptomatology to district hospitals in South Africa and referral hospitals in Cambodia, Peru, Georgia and Vietnam between 2007–2017. Serum CRP measurements were obtained using a laboratory-based assay. CRP cutoff-points of ≥8mg/L and ≥10mg/L were predefined as a positive triage test result. The PTB reference standard was two contemporaneously collected sputum liquid culture results. Results CRP demonstrated an overall sensitivity for PTB of 79.8% (95%CI 75.5–83.5) and 77.7% (95%CI 73.4–81.6) for cutoff-points of 8mg/L and 10mg/L respectively. Specificity was 62.8% (95%CI 57.8–67.6%) and 66.6% (95%CI 61.1–70.7) respectively. Area-under-the-curve using Receiver Operating Characteristic analysis was 0.77 (95%CI 0.74–0.81). Threshold analysis showed optimal CRP cutoff-points were higher in HIV+ than HIV- participants. An algorithm in which CRP triage was followed by confirmatory Xpert MTB/Rif testing achieved a sensitivity of 75.1% (95%CI 69.0–80.4%) whilst decreasing Xpert usage by 40.6%. Discussion CRP may not meet the challenge of a catch-all TB triage test. However, it shows promise in HIV+ individuals. Further research is required in a prospective study using point-of-care platforms to further evaluate its capabilities.