# Serum growth differentiation factor 15 trajectory predicts 28-day mortality in critically ill patients: a multicenter cohort study

**Authors:** Qinxue Wang, Jiawei Wang, Yuhan Zhao, Yuanze Ma, Xiang Li, Xinyi Chang, Nan Zheng, Yong Ji, Yi Han

PMC · DOI: 10.7717/peerj.20317 · PeerJ · 2025-11-03

## TL;DR

Tracking changes in GDF15 levels in ICU patients can predict 28-day mortality, with higher levels linked to worse outcomes.

## Contribution

Introduces GDF15 trajectory and GDF15-load as novel, robust predictors of mortality in ICU patients.

## Key findings

- Four GDF15 trajectory subtypes were identified, with high-increase and high-persistent subtypes showing worst outcomes.
- GDF15-load was a strong predictor of 28-day mortality, comparable to APACHE II and SOFA scores.
- Initial GDF15 levels (GDF15-D1) strongly correlated with cumulative GDF15 burden and improved mortality prediction when combined with existing scores.

## Abstract

Growth differentiation factor 15 (GDF15) has been linked to critical illnesses, particularly cardiovascular and infectious diseases, but its dynamic patterns and prognostic value in critically ill patients remain unclear. This study investigates the predictive utility of serum GDF15 trajectories for 28-day mortality among patients in the intensive care unit (ICU).

In this multicenter, prospective cohort study, ICU patients were enrolled, and serum GDF15 trajectories during the first week were analyzed using group-based trajectory modeling (GBTM). The association between trajectory subtypes and 28-day mortality was assessed through hierarchically adjusted multivariable logistic regression. A cumulative index, “GDF15-load,” was introduced to quantify overall GDF15 exposure and compared with the Acute Physiology and Chronic Health Evaluation II (APACHE II) and Sequential Organ Failure Assessment (SOFA) scores. The correlation between initial GDF15 levels (GDF15-D1) and GDF15-load was evaluated using Spearman’s correlation test. Predictive performance was assessed via the area under the receiver operating characteristic curve (AUROC), and feature importance was interpreted using Shapley Additive Explanations (SHAP).

Among 1,973 patients, 493 comprised the cohort for development with full serum profiles on days 1, 3, and 7. Four GDF15 trajectory subtypes were identified: low-maintenance (LM), medium-maintenance (MM), high-increase (HI), and high-persistent (HP). Trajectory subtypes showed significant differences in inflammatory markers, organ dysfunction, and 28-day mortality, with the HI and HP groups having the worst outcomes. GDF15-load increased progressively from LM to HP and emerged as the most important predictor of 28-day mortality, not inferior to APACHE II and SOFA scores. GDF15-D1 was strongly correlated with GDF15-load (Spearman r = 0.778) and demonstrated robust predictive value, particularly in postoperative ICU patients, where its combination with APACHE II or SOFA further improved prognostic accuracy.

Serum GDF15 trajectory and GDF15-load are robust predictors of 28-day mortality in ICU patients. GDF15-D1 strongly reflects cumulative GDF15 burden and provides a rapid, practical tool for early risk stratification, especially in postoperative ICU patients. These findings support the use of GDF15 as both a dynamic and point-of-care biomarker in intensive care settings.

## Linked entities

- **Genes:** GDF15 (growth differentiation factor 15) [NCBI Gene 9518]

## Full-text entities

- **Genes:** GDF15 (growth differentiation factor 15) [NCBI Gene 9518] {aka GDF-15, HG, MIC-1, MIC1, NAG-1, PDF}
- **Diseases:** inflammatory (MESH:D007249), Failure (MESH:D051437), critical illnesses (MESH:D016638), organ dysfunction (MESH:D009102), cardiovascular and infectious diseases (MESH:D003141)
- **Species:** Homo sapiens (human, species) [taxon 9606]

## Full text

_Full body text omitted from this summary view._ Fetch the complete paper as Markdown: https://tomesphere.com/paper/PMC12591050/full.md

## Figures

6 figures with captions in the complete paper: https://tomesphere.com/paper/PMC12591050/full.md

## References

36 references — full list in the complete paper: https://tomesphere.com/paper/PMC12591050/full.md

---
Source: https://tomesphere.com/paper/PMC12591050