A recently developed RD-1 gene-based assay for diagnosing tuberculous peritonitis (TBP) has given promising results. We therefore created a clinical algorithm for differentiating TBP from other diagnoses using peripheral blood and peritoneal fluid mononuclear cells (PBMC/PF-MC) along with conventional tests.

All adult patients with suspected TBP in whom enzyme-linked immunosorbent spot (ELISPOT) assays were performed both on PBMC and PF-MC were prospectively enrolled over a 6-year period. Confirmed TBP with positive cultures or Mycobacterium tuberculosis PCR, probable TBP with PF changes consistent with TBP, caseating granuloma, and a successful response to anti-TB therapy, as well as possible TBP without exclusion of TBP, were each defined.

A total of 74 patients were enrolled. Of these, 45 (61%) (19 confirmed, 16 probable, and 10 possible) were classified as TBP. The other 29 (39%) patients were classified as not TB. The sensitivity and specificity, respectively, of the tested methods for diagnosing TBP were as follows: PBMC ELISPOT (≥6 spots), 84% and 59%; PF-MC ELISPOT (≥6 spots), 87% and 86%; PF-MC/PBMC ratio (≥3), 69% and 97%; and PF-ADA level (≥21 U/L), 82% and 79%. The areas under the ROC curves were as follows: PF-MC ELISPOT, 0.90; PF-MC/PBMC ratio, 0.82; PBMC ELISPOT, 0.80; and PF-ADA, 0.80, respectively. When a 2-step algorithm (‘PBMC ELISPOT ≥6 spots or PF-ADA ≥21 U/L’ as a rule-out test and ‘PF-MC/PBMC ratio ≥3’ as a rule-in test) was applied, 67% (30/45) of the patients with TBP were accurately classified without undergoing invasive procedures.

A 2-step algorithm using the PBMC/PF-MC ELISPOT assays and PF-ADA appears to be a promising rapid and non-invasive approach for diagnosing TBP.