3 days popular7 days popular1 month popular3 months popular

Predictive Model Created For Mortality Risk In The ICU

A metabolic profile of intensive care unit () patients based on biomarkers of four can be used to accurately predict mortality, according to a new study.

“Existing models for predicting mortality in the ICU may not always be accurate and they can also be cumbersome to use,” said researcher Angela J. Rogers MD, MPH, Instructor in Medicine at Channing Laboratory, Brigham and Women’s Hospital, in Boston. “Levels of lactate, a metabolite from the carbohydrate pathway, are an established biomarker for ICU mortality. Using metabolomics, the scientific study of chemical processes involving metabolites [molecules that are involved or are the product of important cellular processes], we were able to identify several novel biomarkers for predicting mortality in ICU patients and create a network of these metabolites that was highly associated with mortality in two independent populations.”

The results of the study were presented at the ATS 2013 International Conference.

In the study, plasma samples from 90 ICU patients from the Brigham and Women’s Hospital Registry of Critical Illness, with systemic inflammatory response syndrome (SIRS), sepsis, or sepsis/acute respiratory distress syndrome (ARDS) were tested for the presence of 411 metabolites. These metabolites were analyzed for possible associations with 28-day mortality using a statistical model that adjusted for age, race, gender, malignancy status, and Acute Physiologic and Chronic Health Evaluation (APACHE) score.

Of 90 enrolled patients, 30 died by Day 28. Before statistical adjustments were made, a total of 59 metabolites were differentially expressed in survivors and patients who died. These were tested in an independent cohort of 150 patients with sepsis participating in the Community Acquired Pneumonia & Sepsis Outcome Diagnostics (CAPSOD) study. 34 of the individual metabolites were associated with death in that cohort as well.

A final predictive model was created using a statistical model, Bayesian Networks, and included five metabolites: sucrose, and mannose from the carbohydrate pathway and arginine, methionine, and beta-hydroxyisovalerate amino acids. The network was highly associated with mortality in both populations.

“We were impressed with the profound metabolic derangements that were associated with death in two separate cohorts of critically ill patients,” said Dr. Rogers. “Because these results replicate well in two populations, it suggests that measuring plasma metabolites may be an important tool for identifying patients at highest risk for death in the ICU.”


American Thoracic Society International Conference May 17-22, 2013 Philadelphia, Pennsylvania

* Please note that numbers in this release may differ slightly from those in the abstract. Many of these investigations are ongoing; the release represents the most up-to-date data available at press time.

Abstract 40370

A Metabolomic Signature For Death In The ICU

Type: Scientific Abstract

Category: 03.04 – Genomics/Proteomics/Bioinformatics (RCMB)

Authors: A.J. Rogers1, M. Michael1, L. Gazourian2, T. Dolinay2, G.M. Hunninghake2, L.E. Fredenburgh2, A. Massaro2, R.J. Langley3, B.A. Raby1, R.M. Baron2, A.M.K. Choi4; 1Channing Division of Network Medicine – Boston, MA/US, 2Brigham and Women’s Hospital – Boston, MA/US, 3Lovelace Respiratory Research Institute – Albuquerque, NM/US, 4Brigham and Women’s Hospital, Harvard Medical School – Boston, MA/US; BWH MICU Registry

Abstract Body:

Rationale: Metabolomic profiling can provide insight into cellular activity—compared to SNP and gene expression, metabolism products can be seen as the final step in the omics pathway from genotype to phenotype. While SNP and gene expression studies have been conducted in the ICU setting, a metabolomic examination of plasma at ICU admission has never been attempted. Lactate, a breakdown product of anaerobic metabolism is one of the most-useful biomarkers for ICU mortality and response to therapy. Comprehensive testing for novel metabolomic biomarkers is now feasible.

Methods: We studied 90 medical ICU subjects with SIRS, Sepsis, and Sepsis/ARDS enrolled in the BWH Registry of Critical Illness (RoCI). Plasma samples on Day 0 of enrollment were tested for 411 metabolites using lipid chromatography and gas chromatography mass spectroscopy at Metabolon, Inc. Samples were analyzed for association with 28-day mortality, using logistic regression in R, adjusting for age, race, gender, and malignancy status. We then used a conditional Gaussian Bayesian network to identify the network of metabolites and phenotypes most highly predictive of death.

Results: 30 of the 90 RoCI subjects died by day 28. We identified 50 metabolites that were present differentially among ICU survivors vs. those who died by day 28 with FDR p< .05. The top metabolites remain significant after adjustment for malignancy status and APACHE score. We used Bayesian networks to identify a metabolic signature for death in the ICU. The metabolic signature included 4 metabolites from 3 metabolic pathways: lactate and mannose from the carbohydrate pathway, gamma-glutamyltyrosine (a peptide), and stearidonate (18:4n3) (a long-chain fatty acid). We performed five-fold cross-validation using these 4 metabolites and race in our cohort, and achieved an accuracy of 82% AUC predicting death. This combined metabolomic signature outperforms individual biomarkers like lactate and the APACHE score (AUCs <67%). Conclusion: In this analysis, we have performed metabolomic profiling early in ICU admission, and identified biomarkers for 28-day mortality. We have identified several novel candidate biomarkers for ICU mortality, and using a conditional Gaussian Bayesian network, are able to predict death with much greater accuracy than the widely-used but cumbersome APACHE score. These results establish the study of metabolomics as a powerful new genomic technique for identifying novel ICU biomarkers.

Funding: Parker B. Francis foundation American Thoracic Society