Twenty-two studies met the inclusion criteria (2,836 newborns, range 55 to 286). The major methodological limitations of the included studies were identified as a lack of blinding of interpreters of the index test and reference standard, explanation of withdrawals, reporting of uninterpretable results and disease progression bias. Studies performed better with regards avoided partial and differential verification bias and incorporation bias and recruiting a representative patients sample, but the graphic showed that several studies still failed these criteria.
Early onset sepsis: When using umbilical blood (three studies), procalcitonin test sensitivity was 78% (95% CI 68% to 86%) and specificity was 83% (95% CI 80% to 86%). For blood samples taken at zero to 12 hours (eight studies) sensitivity was 77% (95% CI 72% to 81%) and specificity was 87% (95% CI 84% to 90%). At 12 to 24 hours (four studies) sensitivity was 77% (95% CI 68% to 84%) and specificity was 89% (95% CI 85% to 91%). At 24 to 48 hours (six studies) sensitivity was 70% (95% CI 64% to 75%) and specificity was 88% (95% CI 85% to 91%).
Late-onset sepsis (>72 hours): Where sepsis was proven or suspected (seven studies), sensitivity was 62% (95% CI 57% to 67%) and specificity was 92% (95% CI 89% to 95%). Where sepsis was proven (eight studies), sensitivity was 82% (95% CI 77% to 86%) and specificity was 77% (95% CI 73% to 80%).
Considerable heterogeneity was observed for most analyses; where heterogeneity was considered to be absent, this was using p<0.05. Results were reported that compared the diagnostic outcomes of the procalcitonin test to those of the C-reactive protein test; sensitivity was higher for the procalcitonin test, but the results for specificity differed depending on whether the comparisons were conducted in the same patients or not. The diagnostic odds ratio and results of further subgroup and sensitivity analyses were reported.