<![CDATA[Background: Rapid molecular tests are central to pulmonary tuberculosis (TB) diagnosis, yet sputum scarcity, pre-analytic barriers, and operational delays sustain diagnostic uncertainty in referral hospitals. Urine lipoarabinomannan (LAM) and serum C-reactive protein (CRP) may provide adjunctive, point-of-care signals, but head-to-head evidence in routine care remains limited. Methods: We conducted a cross-sectional diagnostic accuracy study in a tertiary referral hospital in Indonesia and reported according to STARD principles. Adults with suspected pulmonary TB who had not started anti-tuberculosis treatment underwent urine LAM testing and serum CRP measurement (cut-off >5). The reference standard was bacteriological testing using a rapid molecular assay on respiratory specimens. We calculated sensitivity, specificity, predictive values, likelihood ratios, and receiver operating characteristic (ROC) area under the curve (AUC). Results: Fifty participants were included with complete index and reference test data. Urine LAM demonstrated sensitivity 78% and specificity 75%, with PPV 56.7% and NPV 75% (LR+ 3.12; LR− 0.29) and AUC 0.77. CRP showed sensitivity 75% and specificity 50%, with PPV 58.1% and NPV 68.4% (LR+ 1.50; LR− 0.50) and AUC 0.62. LAM and CRP distributions did not differ significantly between clinically diagnosed and bacteriologically confirmed TB groups (p=0.056 and p=0.072, respectively). Formal AUC comparison and subgroup analyses were not performed because of limited granularity and few HIV-positive participants. Conclusion: In this referral hospital cohort, urine LAM provided better overall discrimination and stronger rule-in support than CRP, while CRP appeared more suitable for triage-oriented use. Multicenter studies should validate these findings and test combined algorithms for implementation in practice.]]>
