To the Editor This letter is in response to the systematic review and meta-analysis by Achten et al¹ on the use of the neonatal early-onset sepsis (EOS) calculator. The meta-analysis clearly demonstrated that the use of the neonatal EOS calculator is associated with a reduction in the use of empirical antibiotics. However, the authors’ conclusion that “Available evidence regarding safety of the use of the EOS calculator is limited, but shows no indication of inferiority compared with conventional management strategies”¹ may not be true. Many of the cases of EOS in which the EOS calculator was applied were not included in the meta-analysis. The authors included 18 cases of EOS in which the management was guided by the use of the EOS calculator (Table 2) and 12 cases of EOS from retrospective studies for which the EOS calculator was hypothetically applied (Table 3).¹ Surprisingly, 39 cases of EOS from the landmark study by Kuzniewicz et al,² in which EOS management was assessed using the EOS calculator, are not included in the meta-analysis (eTable 1 in the Supplement). In this study, the EOS calculator was applied to 24 neonates during the baseline period and 15 infants during the learning period. Although management was not guided by the EOS calculator during these periods, the calculator would have failed to recommend empirical antibiotics in 29 of 39 infants (74%) with blood culture–positive EOS. Moreover, the calculator would not have recommended even a blood culture in 63% of infants (25 of 39) with blood culture–positive EOS. Similar concern was raised by others regarding missing cases of EOS during the baseline period.³ The authors¹ have reported that during the baseline period, 18 of 24 infants received empirical antibiotics before 24 hours by using a conventional method (Table 2). However, during this period, if the management was guided by the recommendations from the EOS calculator, only 6 of 24 infants would have received antibiotics before 24 hours with an additional 2 infants if their blood culture became positive within 24 hours (18 of 24 vs 8 of 24; P = .009). It is unclear why the authors have included these 12 cases of hypothetically managed infants from retrospective studies but excluded a large number of infants from such a seminal study. The number of infants with EOS (calculator applied) excluded in the meta-analysis was more than the combined infants with EOS included in the meta-analysis. Including these infants with EOS would likely change the conclusion of the meta-analysis.

给编辑的信这是对Achten等人¹对使用新生儿早发型脓毒症（EOS）计算器的系统评价和荟萃分析的回应。荟萃分析清楚地表明，新生儿EOS计算器的使用与经验性抗生素的使用减少有关。但是，作者的结论是“关于使用EOS计算器的安全性的可用证据有限，但与传统管理策略相比，并没有显示自卑感” ¹可能不是真的。荟萃分析未包括应用了EOS计算器的许多EOS案例。作者包括18例EOS，其中以EOS计算器为指导进行管理（表2）和12例回顾性研究中的EOS（假设使用EOS计算器）（表3）。¹令人惊讶的是，来自Kuzniewicz等人具有里程碑意义的研究的39例EOS，²Meta分析中不包括使用EOS计算器评估EOS管理的情况（补编中的e表1）。在这项研究中，EOS计算器在基线期应用于24名新生儿，在学习期应用于15名婴儿。尽管在这段时期内EOS计算器未指导管理，但该计算器未能为39名血培养阳性EOS婴儿中的29名（74％）推荐经验性抗生素。此外，计算器甚至不建议在63％的血液培养阳性EOS的婴儿中进行血液培养（39岁中的25岁）。其他人也对基准期内缺少EOS的情况提出了类似的关注。³作者¹据报道，在基线期，24名婴儿中有18名在24小时之前通过常规方法接受了经验性抗生素治疗（表2）。但是，在此期间，如果根据EOS计算器的建议进行管理，则24小时内只有6例婴儿会在24小时之前接受抗生素治疗，如果其血液培养在24小时内呈阳性，则另外2例婴儿会接受抗生素治疗（24中的18 vs 24之8；P = .009）。目前尚不清楚为什么作者从回顾性研究中纳入这12例假设治疗的婴儿，却从此类开创性研究中排除了许多婴儿。荟萃分析中排除的EOS（应用计算器）婴儿的数量比荟萃分析中包含EOS的合并婴儿多。包括这些EOS婴儿可能会改变荟萃分析的结论。