Sepsis is a heterogeneous disease and identifying sepsis subphenotypes can help optimize sepsis management.

To identify sepsis subphenotypes and risk stratification using procalcitonin trajectories.

Retrospective analysis of 800 cases admitted to the General Hospital of Ningxia Medical University and 202 adult patients with sepsis (age >18 years) in Gansu Provincial Hospital from January 1, 2021 to August 1, 2023 was performed. 597 patients from the General Hospital of Ningxia Medical University were randomized into the development cohort (60%) , and another 202 and 203 from Gansu Provincial Hospital, totaling 405 patients, were included in the validation cohort (40%) . Firstly, the development cohort was divided into survival and death groups to analyze the prognostic value of procalcitonin measurements for sepsis at different times, and ROC curves were plotted to assess predictive efficacy. Then, Group-based trajectory modeling (GBTM) based on repeated measurements of procalcitonin was performed to identify sepsis subphenotypes, which were characterized based on trends in procalcitonin changes and clinical features, and survival analysis and risk stratification were performed, and, Finally, the predictive model was validated.

In the development cohort, 512 patients survived and 85 died, and the overall 28-day mortality was 14.2%. In the validation cohort, 341 patients survived and 64 died, with an overall 28-day mortality of 16.3%. The death group had significantly higher PCT d3, PCT d5, and PCT d7 than the survival group (P<0.01) , and PCT d7 had the highest predictive efficacy with an area under the ROC curve of 0.833. The "Middle Start Rapid Rise" was characterized by respiratory dysfunction; the "Low Start Slow Decline" had the lowest comorbidity and critical care scores and was considered to be the baseline group; the "High Start Rapid Decline" was characterized by higher comorbidity and critical care scores; and the "High Start Slow Decline" was characterized by multiple organ dysfunction and had the highest value of critical care scores and was considered to be the most severe group on admission. Survival analyses of the four subphenotypes showed that "Middle Start Rapid Rise" had the highest mortality rate and was defined as the high-risk group, followed by "High Start Slow Decline" and was defined as the intermediate-risk group, "Low Start Slow Decline" and "High Start Rapid Decline" had the lowest mortality rate and were defined as the low-risk group. The relative distributions of calcitonin trajectories and comorbidities in the validation and development cohorts were generally consistent.

Procalcitonin trajectories can be used to identify sepsis subphenotypes, and the combination of procalcitonin values and trajectories can be used to achieve risk stratification for sepsis, providing a theoretical basis for clinicians to assess the prognosis of patients using procalcitonin trajectories.

脓毒症是一种异质性疾病，识别脓毒症亚表型有助于优化脓毒症管理。

利用降钙素原（PCT）轨迹识别脓毒症亚表型并进行风险分层。

回顾性分析2021-01-01至2023-08-01宁夏医科大学总医院收住的800例及甘肃省人民医院收住的202例成年脓毒症患者（年龄>18岁），随机将宁夏医科大学总医院其中597例患者纳入开发队列（60%），另外203例及甘肃省人民医院202例共405例患者纳入验证队列（40%）。首先根据患者28 d生存情况，将开发队列分为存活组与死亡组，分析不同时间PCT测量值（PCT第1天、第3天、第5天、第7天分别标记为PCT d1、PCT d3、PCT d5、PCT d7）对脓毒症的预后价值，并绘制受试者工作特征（ROC）曲线评估预测效能；然后，基于PCT重复测量进行组基轨迹建模以识别脓毒症亚表型，根据PCT变化趋势及临床特征对亚表型进行特征分析，并进行生存分析与风险分层；最后，对预测模型进行验证。

开发队列中512例存活、85例死亡，总体28 d死亡率为14.2%；验证队列中341例存活、64例死亡，总体28 d死亡率为16.3%。开发队列中死亡组PCT d3、PCT d5、PCT d7高于存活组（P<0.01）；ROC曲线结果显示，PCT d7预测效能较高，ROC曲线下面积为0.833。开发队列组基轨迹建模确定了4种脓毒症亚表型："中起点快速上升型"的特征是呼吸功能障碍；"低起点缓慢下降型"的特征是并发症及危重症评分均较低，视为基线组；"高起点快速下降型"的特征是合并症及危重症评分均较高；"高起点缓慢下降型"的特征是多器官功能障碍，危重症评分较高，视为入院时最严重组。通过对4种亚表型进行生存分析，结果显示"中起点快速上升型"死亡率最高，定义为高危组；其次为"高起点缓慢下降型"，定义为中危组；"低起点缓慢下降型"和"高起点快速下降型"死亡率较低，定义为低危组。验证队列与开发队列PCT轨迹和合并症的相对分布基本一致。

利用PCT轨迹可以识别脓毒症亚表型，结合PCT数值与变化轨迹可以实现对脓毒症的风险分层，为临床医师利用PCT变化轨迹评估患者预后提供理论依据。