What question did this study set out to answer?

April 20, 2026Open Access

Dynamic biomarker profiling and phenotyping in burn sepsis: a retrospective cohort study using growth mixture modeling

PXPei XuNingbo No. 2 Hospital JLJiaqi LouNingbo No. 2 Hospital HKHong KongNingbo College of Health Sciences

Key Points

To evaluate the prognostic value of multicategory biomarkers in burn sepsis and identify distinct patient phenotypes.
Utilized a retrospective cohort design with 712 adult burn sepsis patients in a regional burn ICU.
Measured 17 biomarkers at five time points (days 1, 3, 7, 14, and 21).
Compared survivor/non-survivor biomarker trajectories using linear mixed-effects models.
Analyzed baseline biomarker associations with 21-day mortality using Cox regression.
Employed growth mixture modeling to identify distinct patient phenotypes based on biomarker trajectories.
21-day mortality rate was 17.9% (81 deaths).
Identified two patient phenotypes: high-risk (mortality 15.7%) and low-risk (mortality 8.8%) with significant biomarker differences (P = 0.005).
Baseline markers like ALB and IL-6 were associated with mortality but did not retain independent significance in multivariable analysis.
C-indices for baseline ALB, PA, and IgG were modest (0.604, 0.583, and 0.585).

Abstract

Objective To assess the dynamic prognostic value of multicategory biomarkers in burn sepsis and identify patient phenotypes based on their longitudinal trajectories. Methods This retrospective cohort study included 712 adult patients with burn sepsis (Sepsis−3) admitted to a regional burn ICU (2022–2025). Seventeen biomarkers covering nutrition (albumin, ALB; prealbumin, PA; transferrin, TRF; nitrogen balance), immunity (immunoglobulins A/G/M; CD3 + /CD4 + /CD8 + T cells, CD4 + /CD8 + ratio; natural killer, NK cells), and inflammation (interleukin−6, IL−6; C−reactive protein, CRP; procalcitonin, PCT; platelet count, lactate) were measured at five time points (days 1, 3, 7, 14, and 21). We compared survivor/non-survivor trajectories using linear mixed-effects models, assessed baseline biomarker associations with 21−day mortality via Cox regression, and evaluated predictive performance with Harrell’s C-index. Growth mixture modeling (GMM) identified phenotypes from integrated ALB, IL−6, and immunoglobulin G (IgG) trajectories. Results The 21−day mortality was 17.9% (81 deaths). Survivor and non-survivor trajectories differed significantly for multiple biomarkers ( P 0.05). Growth mixture modeling identified two distinct patient phenotypes: a high-risk phenotype ( n = 267, mortality 15.7%) characterized by persistently lower ALB and IgG and sustained IL−6 elevation over 21 days and a low-risk phenotype ( n = 445, mortality 8.8%) with favorable biomarker trajectories ( P = 0.005). Univariable analysis associated several baseline markers with mortality (e.g., ALB: HR = 0.97, P = 0.013; IL−6: HR = 1.004, P = 0.004). However, no biomarker retained independent significance in multivariable analysis, likely due to multicollinearity among nutritional markers variance inflation factor (VIF) up to 8.4. Harrell’s C-indices for baseline ALB, PA, and IgG were modest (0.604, 0.583, and 0.585, respectively). Conclusions Longitudinal multicategory biomarker trajectories predict 21−day survival in burn sepsis. Trajectory-based phenotyping identifies patient subgroups with markedly different outcomes, offering superior prognostic stratification over static measurements. The integrated phenotype, reflecting the dynamic interplay of catabolism, immune paralysis, and inflammation, emerges as a robust prognostic marker, supporting personalized management approaches.

Ask AI

Helpful

Bookmark

View Full Paper