Shock-Stratified Prognostic Performance of Lactate and Procalcitonin in Pediatric Sepsis In-Hospital Mortality; A Retrospective Cohort Study.

Introduction: Admission lactate and procalcitonin (PCT) are routinely measured in pediatric sepsis; however, their prognostic performance may vary by clinical phenotype. This study aimed to evaluate the shock-stratified prognostic performance of admission lactate and PCT for predicting in-hospital mortality in children with sepsis. Methods: In this retrospective cohort study (July 2022 - July 2025), children aged 2 months to 15 years evaluated for suspected sepsis were screened; 193 met eligibility criteria and were included in analyses. Septic shock was recorded at admission, and organ dysfunction was recorded at admission or within the first 24 hours. The Phoenix Sepsis Score was calculated in 143 patients with complete lactate and other required data, whereas the Pediatric Sequential Organ Failure Assessment (pSOFA) score was calculated for all 193 patients using data recorded within the first 24 hours. Admission lactate and PCT were summarized by survival status within shock and non-shock phenotypes. Multivariable logistic regression models adjusted for age and malnutrition assessed biomarker-mortality associations and biomarker × shock interactions. Discrimination was evaluated using shock-stratified receiver operating characteristic (ROC) curves and area under the curve (AUC), with DeLong tests comparing AUCs between phenotypes. Results: Among 193 children, 93 (48.2%) had septic shock, 145 (75.1%) had organ dysfunction, and 79 (40.9%) died in hospital. The median Phoenix Sepsis Score was 2.0 (range, 0--12) among 143 patients, while the median pSOFA score was 4.0 (range: 0--21) among all 193 patients. Mortality was higher in shock than non-shock sepsis (66.7% vs 17.0%; odds ratio (OR): 9.77 (95% confidence interval (CI): 4.96-19.24)) and in those with organ dysfunction (53.1% vs 4.2%; OR = 26.03 (95% CI: 6.09-111.34)). Admission lactate was higher in non-survivors than survivors in both non-shock and shock groups (p<0.05). In adjusted interaction models, each 1 mmol/L increase in lactate (OR = 1.50, 95% CI: 1.11--2.01; p = 0.0075) and each one-unit increase in natural-log--transformed PCT (ln(PCT)) (OR = 1.34, 95% CI: 1.01-1.79; p=0.0458) were associated with mortality in non-shock sepsis. The lactate × shock interaction (OR = 0.76, 95% CI: 0.55--1.05; p = 0.092) and ln(PCT) × shock interaction (OR = 0.74, 95% CI: 0.52--1.07; p = 0.114) were not statistically significant. Shock-stratified ROC curve analysis showed better discrimination in non-shock than shock for lactate (AUC: 0.76 (95% CI: 0.67-0.85) vs 0.62 (95% CI: 0.40-0.83); DeLong p=0.233) and modest discrimination for natural-log-transformed PCT [ln(PCT)] in both phenotypes (AUC: 0.65 (95% CI: 0.55-0.75) vs 0.61 (95% CI: 0.41-0.80); DeLong p=0.695). Conclusion: Admission lactate showed moderate discrimination for in-hospital mortality in non-shock sepsis, whereas discrimination was lower in septic shock. PCT showed modest discrimination across phenotypes. Although AUCs were numerically higher in non-shock sepsis, between-phenotype differences were not statistically significant, supporting phenotype-based interpretation and the need for external validation.