Evaluation of Biomarkers in Critical Care and Perioperative Medicine: A Clinician’s Overview of Traditional Statistical Methods and Machine Learning Algorithms

Clinical phenotypes of cardiogenic shock survivors: insights into late host responses and long-term outcomes

AIMS

An elevated risk of adverse events persists for years in cardiogenic shock (CS) survivors with high mortality rate and physical/mental disability. This study aims to link clinical CS-survivor phenotypes with distinct late host-response patterns at intensive care unit (ICU) discharge and long-term outcomes using model-based clustering.

METHODS AND RESULTS

In the original prospective, observational, international French and European Outcome Registry in Intensive Care Units (FROG-ICU) study, ICU patients with CS on admission were identified (N = 228). Among them, 173 were discharged alive from the ICU and included in the current study. Latent class analysis was applied to identify distinct CS-survivor phenotypes at ICU discharge using 15 readily available clinical and laboratory variables. The primary endpoint was 1 year of mortality after ICU discharge. Secondary endpoints were readmission and physical/mental disability [short form-36 questionnaire (SF-36) score] within 1 year after ICU discharge. Two distinct phenotypes at ICU discharge were identified (A and B). Patients in Phenotype B (38%) were more anaemic and had higher circulating levels of lactate, sustained kidney injury, and persistent elevation in plasma markers of inflammation, myocardial fibrosis, and endothelial dysfunction compared with Phenotype A. They had also a higher rate of non-ischaemic origin of CS and right ventricular dysfunction on admission. CS survivors in Phenotype B had higher 1 year of mortality compared with Phenotype A (P = 0.045, Kaplan-Meier analysis). When adjusted for traditional risk factors (i.e. age, severity of illness, and duration of ICU stay), Phenotype B was independently associated with 1 year of mortality [adjusted hazard ratio = 2.83 (95% confidence interval 1.21-6.60); P = 0.016]. There was a significantly lower physical quality of life in Phenotype B patients at 3 months (i.e. SF-36 physical component score).

CONCLUSIONS

A phenotype with sustained inflammation, myocardial fibrosis, and endothelial dysfunction at ICU discharge was identified from readily available data and was independently associated with poor long-term outcomes in CS survivors.

Tracking longitudinal biomarkers in burn patients with sepsis and acute kidney injury: an unsupervised clustering approach

BACKGROUND

Sepsis is a grave medical disorder characterized by a systemic inflammatory response to infection. Furthermore, it is a leading cause of morbidity and mortality, especially in hospitalized patients. Acute kidney injury (AKI) is a common complication of sepsis and is associated with increased morbidity and mortality. Patients with burns are particularly vulnerable to developing sepsis and AKI due to the extensive tissue damage and immune suppression resulting from burn injury. In this study, unsupervised clustering algorithms were used to track longitudinal biomarkers in patients with burns and assess their impact on mortality.

METHODS

This retrospective study included adult patients with burns aged ≥ 18 years, who were admitted to the burn intensive care unit of Hallym University and Hangang Sacred Heart Hospital between July 2010 and December 2021. The patients were divided into two subgroups: those with sepsis (538 patients) and those without sepsis (826 patients). The longitudinal biomarkers were grouped into three clusters using the k-means clustering algorithm. Each cluster was assigned a letter from A to C according to its mortality rate.

RESULTS

The odds ratio (OR) of pH was 9.992 in the positive group and 31.745 in the negative group in cluster C. The OR for lactate dehydrogenase (LD) was 3.704 in the positive group and 6.631 in the negative group in cluster C. The OR for creatinine was 2.784 in the positive group and 8.796 in the negative group in cluster C. The OR for blood urea nitrogen (BUN) in the negative group was 0.348, indicating a negative predictor of mortality. Regarding the application of Continuous Renal Replacement Therapy (CRRT) and ventilation, ventilation was significant in both groups. In contrast, CRRT application was not significant in the sepsis-positive group. Furthermore, it was not selected as a variable in the negative group.

CONCLUSIONS

The pH, LD, and creatinine were significant in both groups, while lactate and platelets were significant in the sepsis-positive group. In addition, albumin, glucose, and BUN were significant in the sepsis-negative group. Continuous renal replacement therapy was not significant in either group. However, the use of a ventilator was associated with poor prognosis.

Distinct host-response signatures in circulatory shock: a narrative review

Circulatory shock is defined syndromically as hypotension associated with tissue hypoperfusion and often subcategorized according to hemodynamic profile (e.g., distributive, cardiogenic, hypovolemic) and etiology (e.g., infection, myocardial infarction, trauma, among others). These shock subgroups are generally considered homogeneous entities in research and clinical practice. This current definition fails to consider the complex pathophysiology of shock and the influence of patient heterogeneity. Recent translational evidence highlights previously under-appreciated heterogeneity regarding the underlying pathways with distinct host-response patterns in circulatory shock syndromes. This heterogeneity may confound the interpretation of trial results as a given treatment may preferentially impact distinct subgroups. Re-analyzing results of major 'neutral' treatment trials from the perspective of biological mechanisms (i.e., host-response signatures) may reveal treatment effects in subgroups of patients that share treatable traits (i.e., specific biological signatures that portend a predictable response to a given treatment). In this review, we discuss the emerging literature suggesting the existence of distinct biomarker-based host-response patterns of circulatory shock syndrome independent of etiology or hemodynamic profile. We further review responses to newly prescribed treatments in the intensive care unit designed to personalize treatments (biomarker-driven or endotype-driven patient selection in support of future clinical trials).

Biomarkers: Promising and valuable tools towards diagnosis, prognosis and treatment of Covid-19 and other diseases

The use of biomarkers as early warning systems in the evaluation of disease risk has increased markedly in the last decade. Biomarkers are indicators of typical biological processes, pathogenic processes, or pharmacological reactions to therapy. The application and identification of biomarkers in the medical and clinical fields have an enormous impact on society. In this review, we discuss the history, various definitions, classifications, characteristics, and discovery of biomarkers. Furthermore, the potential application of biomarkers in the diagnosis, prognosis, and treatment of various diseases over the last decade are reviewed. The present review aims to inspire readers to explore new avenues in biomarker research and development.

Machine Learning Approaches for Phenotyping in Cardiogenic Shock and Critical Illness: Part 2 of 2

Progress in improving cardiogenic shock (CS) outcomes may have been limited by failure to embrace the heterogeneity of pathophysiologic processes driving the underlying syndrome. To better understand the variability inherent to CS populations, recent algorithms for describing underlying CS disease subphenotypes have been described and validated. These strategies hope to identify specific patient subgroups with more favorable responses to standard therapies, as well as those who require novel treatment approaches. This paper is part 2 of a 2-part state-of-the-art review. In this second article, we present machine learning-based statistical approaches to identifying subphenotypes and discuss their strengths and limitations, as well as evidence from other critical illness syndromes and emerging applications in CS. We then discuss how staging and stratification may be considered in CS clinical trials and finally consider future directions for this emerging area of research.

Machine learning prediction of postoperative major adverse cardiovascular events in geriatric patients: a prospective cohort study

Background

Postoperative major adverse cardiovascular events (MACEs) account for more than one-third of perioperative deaths. Geriatric patients are more vulnerable to postoperative MACEs than younger patients. Identifying high-risk patients in advance can help with clinical decision making and improve prognosis. This study aimed to develop a machine learning model for the preoperative prediction of postoperative MACEs in geriatric patients.

Methods

We collected patients’ clinical data and laboratory tests prospectively. All patients over 65 years who underwent surgeries in West China Hospital of Sichuan University from June 25, 2019 to June 29, 2020 were included. Models based on extreme gradient boosting (XGB), gradient boosting machine, random forest, support vector machine, and Elastic Net logistic regression were trained. The models’ performance was compared according to area under the precision-recall curve (AUPRC), area under the receiver operating characteristic curve (AUROC) and Brier score. To minimize the influence of clinical intervention, we trained the model based on undersampling set. Variables with little contribution were excluded to simplify the model for ensuring the ease of use in clinical settings.

Results

We enrolled 5705 geriatric patients into the final dataset. Of those patients, 171 (3.0%) developed postoperative MACEs within 30 days after surgery. The XGB model outperformed other machine learning models with AUPRC of 0.404(95% confidence interval [CI]: 0.219–0.589), AUROC of 0.870(95%CI: 0.786–0.938) and Brier score of 0.024(95% CI: 0.016–0.032). Model trained on undersampling set showed improved performance with AUPRC of 0.511(95% CI: 0.344–0.667, p < 0.001), AUROC of 0.912(95% CI: 0.847–0.962, p < 0.001) and Brier score of 0.020 (95% CI: 0.013–0.028, p < 0.001). After removing variables with little contribution, the undersampling model showed comparable predictive accuracy with AUPRC of 0.507(95% CI: 0.338–0.669, p = 0.36), AUROC of 0.896(95%CI: 0.826–0.953, p < 0.001) and Brier score of 0.020(95% CI: 0.013–0.028, p = 0.20).

Conclusions

In this prospective study, we developed machine learning models for preoperative prediction of postoperative MACEs in geriatric patients. The XGB model showed the best performance. Undersampling method achieved further improvement of model performance.

Trial registration

The protocol of this study was registered at www.chictr.org.cn (15/08/2019, ChiCTR1900025160)

Supplementary Information

The online version contains supplementary material available at 10.1186/s12871-022-01827-x.

Identifying clinical subtypes in sepsis-survivors with different one-year outcomes: a secondary latent class analysis of the FROG-ICU cohort

BACKGROUND

Late mortality risk in sepsis-survivors persists for years with high readmission rates and low quality of life. The present study seeks to link the clinical sepsis-survivors heterogeneity with distinct biological profiles at ICU discharge and late adverse events using an unsupervised analysis.

METHODS

In the original FROG-ICU prospective, observational, multicenter study, intensive care unit (ICU) patients with sepsis on admission (Sepsis-3) were identified (N = 655). Among them, 467 were discharged alive from the ICU and included in the current study. Latent class analysis was applied to identify distinct sepsis-survivors clinical classes using readily available data at ICU discharge. The primary endpoint was one-year mortality after ICU discharge.

RESULTS

At ICU discharge, two distinct subtypes were identified (A and B) using 15 readily available clinical and biological variables. Patients assigned to subtype B (48% of the studied population) had more impaired cardiovascular and kidney functions, hematological disorders and inflammation at ICU discharge than subtype A. Sepsis-survivors in subtype B had significantly higher one-year mortality compared to subtype A (respectively, 34% vs 16%, p < 0.001). When adjusted for standard long-term risk factors (e.g., age, comorbidities, severity of illness, renal function and duration of ICU stay), subtype B was independently associated with increased one-year mortality (adjusted hazard ratio (HR) = 1.74 (95% CI 1.16-2.60); p = 0.006).

CONCLUSIONS

A subtype with sustained organ failure and inflammation at ICU discharge can be identified from routine clinical and laboratory data and is independently associated with poor long-term outcome in sepsis-survivors. Trial registration NCT01367093; https://clinicaltrials.gov/ct2/show/NCT01367093 .