目的·探讨病原宏基因组二代测序（metagenomics next-generation sequencing，mNGS）在肺部感染所致脓毒症患者中协助病原体早期诊断、指导临床抗感染方案调整、改善患者近期预后的价值。方法·该研究由一项多中心、前瞻性、非随机对照试验及一项诊断试验构成。纳入2020年3月—2021年10月在4家医院住院治疗的肺部感染所致脓毒症患者，所有患者均符合美国重症医学会与欧洲重症医学会发布的脓毒症3.0标准和肺部感染的临床诊断标准。入组患者根据自身意愿选择感染部位标本的病原学检测方法，包括仅送检传统病原学检测（传统病原学检测组），或在送检传统病原学检测的基础上同步送检mNGS（联合mNGS检测组）。患者预后的主要评价指标为7 d全因死亡率，次要评价指标包括7 d序贯器官衰竭评分（Sequential Organ Failure Assessment，SOFA）变化、7 d急性生理与慢性健康状况评分Ⅱ（Acute Physiology and Chronic Health Evaluation Ⅱ，APACHEⅡ）变化、28 d全因死亡率、28 d机械通气或死亡的复合终点发生率、28 d无呼吸机天数、28 d非住院天数和住院期间日均花费。采用倾向性评分匹配均衡2组的协变量分布，采用Kaplan-Meier生存曲线和Cox比例风险模型比较2组死亡率。使用联合mNGS检测组患者感染部位标本的病原学检测结果进行诊断试验。以临床综合判定的责任病原体为参考标准，分别将传统病原学检测和mNGS检测结果与参考标准相比较，计算2种方法和参考标准之间的阳性百分比一致性、阴性百分比一致性、阳性预测值和阴性预测值，并采用McNemar配对χ²检验比较2种检测方法的责任病原体检出能力。结果·共入组患者533例，其中311例选择接受额外的mNGS检测，222例仅接受传统病原学检测。非随机对照试验中，经倾向性评分匹配平衡协变量后，联合mNGS检测组的7 d全因死亡率低于传统病原学检测组［4.8% vs 8.6%，HR 0.37（95%CI 0.15~0.91），P=0.031］，联合mNGS检测组的28 d无呼吸机天数多于传统病原学检测组（19.9 d vs 18.4 d，P=0.041）。2组在28 d全因死亡率和日均住院花费上差异无统计学意义。诊断试验提示，mNGS检测结果与临床综合判定的责任病原体的阳性百分比一致性高于传统病原学检测［91.9%（95%CI 87.7%~95.0%） vs 56.1%（95%CI 49.7%~62.4%），P<0.001］，阴性百分比一致性低于传统病原学检测［29.2%（95%CI 18.6%~41.8%） vs 69.2%（95%CI 56.6%~80.1%），P<0.001］，阴性预测值高于传统病原学检测［48.7%（95%CI 32.4%~65.2%）vs 29.4%（95%CI 22.3%~37.3%），P=0.001］。结论·与传统病原学检测相比，肺部感染所致脓毒症患者感染部位标本行mNGS可提高责任病原体的检出率。相比于仅行传统病原学检测，联合mNGS检测的患者7 d全因死亡率更低，提示mNGS在肺部感染所致脓毒症患者中具有临床价值及应用前景。

Objective ·To explore the diagnostic, therapeutic, and prognostic value of metagenomics next-generation sequencing (mNGS) in patients with pneumonia-induced sepsis. Methods ·This study consisted of a multicenter, prospective, non-randomized controlled trial and a diagnostic test. Patients with pneumonia-induced sepsis who were hospitalized in four hospitals across China were enrolled between March 2020 and October 2021. All patients met the Sepsis-3 criteria issued by the Society of Critical Care Medicine and the European Society of Intensive Care Medicine, as well as the clinical diagnostic standard of pneumonia. Enrolled patients were assigned based on their preference to either the conventional test-only group [receiving only conventional test (CMT)] or the combined mNGS test group (receiving CMT and mNGS concurrently). The primary outcome was the 7-day all-cause mortality rate, and secondary outcomes included the changes in SOFA and APACHE Ⅱ scores from baseline to day 7, 28-day all-cause mortality rate, the composite endpoint of mechanical ventilation or death within 28 d, 28 d ventilation-free days, 28 d hospital-free days, and the average daily hospitalization cost. Propensity score matching was used to balance covariates between the two groups. Kaplan-Meier curves were plotted and Cox proportional hazards models were built to compare the risk of death between the two groups. Pathogen detection results from infection site samples in the combined mNGS test group were used for the diagnostic test. The clinically-adjudicated causative pathogens was used as the reference standard. The results of traditional pathogen detection and mNGS detection were compared respectively with the reference standard. The positive percent agreement, negative percent agreement, positive predictive value, and negative predictive value between the two methods and the reference standard were calculated. McNemar's χ² test was used to evaluate the causative pathogen detection capabilities of the two methods. Results ·A total of 533 patients were enrolled, of whom 311 opted for additional mNGS testing, while 222 received only conventional pathogenetic testing. In the non-randomized controlled trial, after propensity score matching to balance covariates, the 7-day all-cause mortality was lower in the combined mNGS test group compared to the conventional test-only group [4.8% vs 8.6%, HR 0.37 (95%CI 0.15‒0.91), P=0.031]. Additionally, the 28-day ventilation-free days were increased in the combined mNGS test group (19.9 d vs 18.4 d, P=0.041). No significant difference was observed between the two groups in terms of 28-day all-cause mortality or the average daily hospitalization costs. In the diagnostic test, compared to the reference standard, the positive percent agreement of mNGS with the clinical composite judgment for causative pathogens was higher than that of CMT [91.9% (95%CI 87.7%‒95.0%) vs 56.1% (95%CI 49.7%‒62.4%), P<0.001]. Conversely, the negative percent agreement of mNGS was lower than that of CMT [29.2% (95%CI 18.6%‒41.8%) vs 69.2% 95%CI 56.6%‒80.1%), P<0.001]. The negative predictive value of nNGS was higher than that of CMT [48.7%(95%CI 32.4%‒65.2%) vs 29.4% (95%CI 22.3%‒37.3%), P=0.001]. Conclusion ·In patients with pneumonia-induced sepsis, mNGS of infection site samples demonstrated a higher detection rate of causative pathogen compared to CMT. Furthermore, the combination of mNGS with CMT may help reduce the 7-day all-cause mortality, suggesting that mNGS has clinical value and potential for application in the management of sepsis caused by pulmonary infections.