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. 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. 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 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 (=23·5%), high for studies investigating procalcitonin for serious bacterial infection (=75·5%), and moderate for studies investigating C-reactive protein for invasive bacterial infection (=49·5%) and serious bacterial infection (=28·3%). The absence of a single definition of serious bacterial infection across studies was the greatest source of interstudy variability and potential bias. 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. None.

中文翻译：

---

降钙素原和 C 反应蛋白诊断测试的准确性用于预测低龄发热婴儿侵袭性和严重细菌感染：系统评价和荟萃分析

出生后 90 天内出现发热的婴儿比年龄较大的儿童遭受侵袭性和严重细菌感染的风险更高。现代临床实践指南主要使用降钙素原作为诊断生物标志物，可以识别低风险婴儿，因此适合进行量身定制的管理。相比之下，C 反应蛋白广泛使用，但 C 反应蛋白和降钙素原是否具有相似的诊断准确性尚不清楚。我们的目的是比较降钙素原和 C 反应蛋白在预测发热婴儿侵袭性或严重细菌感染方面的测试准确性。为了进行这项系统评价和荟萃分析，我们使用 MeSH 术语“降钙素原”和“细菌感染”或“发烧”检索了 MEDLINE、EMBASE、Web of Science 和 Cochrane 图书馆截至 2023 年 6 月 19 日的诊断测试准确性研究”以及关键词“侵入性细菌感染*”和“严重细菌感染*”，没有语言或日期限制。研究由独立作者根据资格标准进行选择。符合资格的研究包括年龄不超过 90 天、因发烧（≥38°C）或在之前 48 小时内有发烧史而到医院就诊的参​​与者。主要指标检测为降钙素原，次要指标检测为C反应蛋白。测试套件必须是市售的，测试样本必须在送到医院时收集。侵袭性细菌感染被定义为通过培养或定量PCR检测到血液或脑脊液中存在细菌病原体；使用了作者对严重细菌感染的定义。从选定的研究中提取数据，并使用每种生物标志物的两个模型来分析侵入性或严重细菌感染的检测。根据国际公认的临界值（降钙素原为 0·5 ng/mL，C 反应蛋白为 20 mg/L）确定诊断准确性，并汇总计算每个生物标志物的部分曲线下面积 (pAUC) 值。确定了每个生物标志物的最佳截止值。本研究已在 PROSPERO 注册，CRD42022293284。在文献检索得出的 734 项研究中，有 14 项研究 (n=7755) 被纳入荟萃分析。对于侵袭性细菌感染的检测，降钙素原的 pAUC 值（0·72，95% CI 0·56–0·79）高于 C 反应蛋白（0·28，0·17–0·61；p= 0·016）。检测侵袭性细菌感染的最佳临界值是降钙素原为 0·49 ng/mL，C 反应蛋白为 13·12 mg/L。对于严重细菌感染的检测，降钙素原和 C 反应蛋白具有相似的 pAUC 值（0·55、0·44–0·69 0·54、0·40–0·61；p=0·92）。对于严重细菌感染，降钙素原和C反应蛋白的最佳临界值分别为0·17 ng/mL和16·18 mg/L。调查降钙素原检测侵袭性细菌感染的测试准确性的研究异质性较低 (=23·5%)，调查降钙素原治疗严重细菌感染的研究为高 (=75·5%)，调查 C 反应蛋白治疗侵袭性细菌感染 (=49·5%) 和严重细菌感染 (=28·3%) 的研究为中等。各研究中缺乏对严重细菌感染的单一定义是研究间变异和潜在偏差的最大来源。在一大群发热婴儿中，对于识别侵袭性细菌感染，降钙素原临界值 0·5 ng/mL 的 pAUC 值优于 C 反应蛋白临界值 20 mg/L。因此，在无法获得降钙素原的环境中，应谨慎使用 C 反应蛋白来识别侵袭性细菌感染，并应考虑低于 20 mg/L 的临界值。 C反应蛋白和降钙素原在鉴定严重细菌感染方面表现出相似的测试准确性，具有国际公认的临界值。这可能反映了确认严重细菌感染所面临的挑战以及缺乏普遍接受的严重细菌感染定义。没有任何。