Diagnostic test accuracy of procalcitonin and C-reactive protein for predicting invasive and serious bacterial infections in young febrile infants: a systematic review and meta-analysis
Hannah Norman-Bruce,Etimbuk Umana,Clare Mills,Hannah Mitchell,Lisa McFetridge,David McCleary,Thomas Waterfield
DOI: https://doi.org/10.1016/S2352-4642(24)00021-X
Abstract:Background: Febrile infants presenting in the first 90 days of life are at higher risk of invasive and serious bacterial infections than older children. Modern clinical practice guidelines, mostly using procalcitonin as a diagnostic biomarker, can identify infants who are at low risk and therefore suitable for tailored management. C-reactive protein, by comparison, is widely available, but whether C-reactive protein and procalcitonin have similar diagnostic accuracy is unclear. We aimed to compare the test accuracy of procalcitonin and C-reactive protein in the prediction of invasive or serious bacterial infections in febrile infants. Methods: For this systematic review and meta-analysis, we searched MEDLINE, EMBASE, Web of Science, and The Cochrane Library for diagnostic test accuracy studies up to June 19, 2023, using MeSH terms "procalcitonin", and "bacterial infection" or "fever" and keywords "invasive bacterial infection*" and "serious bacterial infection*", without language or date restrictions. Studies were selected by independent authors against eligibility criteria. Eligible studies included participants aged 90 days or younger presenting to hospital with a fever (≥38°C) or history of fever within the preceding 48 h. The primary index test was procalcitonin, and the secondary index test was C-reactive protein. Test kits had to be commercially available, and test samples had to be collected upon presentation to hospital. Invasive bacterial infection was defined as the presence of a bacterial pathogen in blood or cerebrospinal fluid, as detected by culture or quantitative PCR; authors' definitions of serious bacterial infection were used. Data were extracted from selected studies, and the detection of invasive or serious bacterial infections was analysed with two models for each biomarker. Diagnostic accuracy was determined against internationally recognised cutoff values (0·5 ng/mL for procalcitonin, 20 mg/L for C-reactive protein) and pooled to calculate partial area under the curve (pAUC) values for each biomarker. Optimum cutoff values were identified for each biomarker. This study is registered with PROSPERO, CRD42022293284. Findings: Of 734 studies derived from the literature search, 14 studies (n=7755) were included in the meta-analysis. For the detection of invasive bacterial infections, pAUC values were greater for procalcitonin (0·72, 95% CI 0·56-0·79) than C-reactive protein (0·28, 0·17-0·61; p=0·016). Optimal cutoffs for detecting invasive bacterial infections were 0·49 ng/mL for procalcitonin and 13·12 mg/L for C-reactive protein. For the detection of serious bacterial infections, procalcitonin and C-reactive protein had similar pAUC values (0·55, 0·44-0·69 vs 0·54, 0·40-0·61; p=0·92). For serious bacterial infections, the optimal cutoffs for procalcitonin and C-reactive protein were 0·17 ng/mL and 16·18 mg/L, respectively. Heterogeneity was low for studies investigating the test accuracy of procalcitonin in detecting invasive bacterial infection (I2=23·5%), high for studies investigating procalcitonin for serious bacterial infection (I2=75·5%), and moderate for studies investigating C-reactive protein for invasive bacterial infection (I2=49·5%) and serious bacterial infection (I2=28·3%). The absence of a single definition of serious bacterial infection across studies was the greatest source of interstudy variability and potential bias. Interpretation: Within a large cohort of febrile infants, a procalcitonin cutoff of 0·5 ng/mL had a superior pAUC value to a C-reactive protein cutoff of 20 mg/L for identifying invasive bacterial infections. In settings without access to procalcitonin, C-reactive protein should therefore be used cautiously for the identification of invasive bacterial infections, and a cutoff value below 20 mg/L should be considered. C-reactive protein and procalcitonin showed similar test accuracy for the identification of serious bacterial infection with internationally recognised cutoff values. This might reflect the challenges involved in confirming serious bacterial infection and the absence of a universally accepted definition of serious bacterial infection. Funding: None.