A meta-analysis1 included 18 studies with a total of 2 097 subjects. The diagnostic accuracy of PCT in sepsis diagnosis in critically ill patients was estimated. The enrolled studies included cardiac, pulmonary, neurological, gastrointestinal, renal, trauma, and surgical illnesses. The prevalence of sepsis across the studies ranged from 31% to 88%.
Three were phase 3 studies (n=4, 495 patients), i.e. those that assessed the test's real-life performance in patients suspected to have the disorder. Their diagnostic performance was upwardly biased in smaller studies, but moving toward a null effect in larger studies. The rest of the studies were phase 2 studies (n=14, 1 602 patients), i.e. those that examined how the index test discriminated between patients with or without the target disorder. These were found to have a low pooled diagnostic OR of 7.79 (95% CI 5.86 to 10.35). As a general rule, a diagnostic OR of greater that 100 indicates high accuracy, 25-100 indicates moderate accuracy, and less than 25 indicates an unhelpful test. Overall the diagnostic performance of PCT was found to be low, with mean values of both sensitivity and specificity being 71% (95% CI 67 to 76) and an area under the summary receiver operator characteristic curve of 0.78 (95% CI 0.73 to 0.83).
Comment: The quality of evidence is downgraded by the inconsistency of the diagnostic criteria of the enrolled patients, the small number of phase 3 studies, and the heterogeneity of the phase 3 studies results. Low diagnostic performance was supported by phase 2 tests.
Primary/Secondary Keywords