Clinical Criteria for Persistent Inflammation, Immunosuppression, and Catabolism Syndrome: An Exploratory Analysis of Optimal Cut-Off Values for Biomarkers.

Background: While clinical criteria have been proposed for persistent inflammation, immunosuppression, and catabolism syndrome (PICS) using C-reactive protein (CRP), albumin, and lymphocyte count, there is no substantial basis for their optimal cut-off values. We herein aimed to develop and externally validate clinical criteria for PICS by investigating the optimal cut-off values for these biomarkers using machine-learning approaches and confirmed it with external validation. Methods: To develop criteria, we included ICU patients treated at a tertiary care hospital in Japan between 2018 and 2021 (derivation cohort). We introduced CRP, albumin and lymphocyte counts at around day 14 into six machine-learning models to predict PICS, defined as the compound outcome of the Barthel index (BI) < 70 at hospital discharge and in-hospital death. We incorporated the results of these models to assess the optimal cut-off values for biomarkers. We then developed and externally validated criteria for PICS using a nationwide claims database in Japan (validation cohort). Results: In the derivation cohort, 291 out of 441 patients had BI < 70 or in-hospital death. Based on machine-learning models, the optimal cut-off values for biomarkers to predict them were a CRP of 2.0 mg/dL, albumin of 3.0 g/dL, and a lymphocyte count of 800/μL, with an AUROC of 0.67. In the external validation cohort, 4492 out of 15,302 patients had BI < 70 or in-hospital death. The AUROC of the criteria was 0.71, with sensitivity of 0.71 and specificity of 0.68 to predict PICS. Conclusions: We herein provide a fundamental basis for PICS clinical criteria with CRP >2.0 mg/dL, albumin <3.0 g/dL, and a lymphocyte count <800/μL on day 14. The criteria developed will identify patients with PICS whose long-term mortality and activity of daily living may be poor.