Identifying Cardiovascular Subphenotypes in Sepsis: Toward Personalized Hemodynamic Management

Highlight

• Four distinct cardiovascular subphenotypes were identified in ICU patients with sepsis using unsupervised clustering of echocardiographic and hemodynamic data.
• These subphenotypes correlate with varying mortality rates and reflect different mechanisms of circulatory failure.
• Simple three-variable models accurately classify patients into these subphenotypes in both derivation and validation cohorts.
• Subphenotype-specific responses to vasoactive agents and fluid therapies suggest a potential for personalized hemodynamic treatment in septic shock.

Study Background

Sepsis remains a major cause of intensive care unit (ICU) admissions worldwide, associated with high morbidity and mortality. Cardiovascular dysfunction, including circulatory failure and shock, is a key driver of poor outcomes in sepsis. However, sepsis-induced cardiovascular abnormalities are heterogeneous, ranging from preserved cardiac function with systemic vasodilation to depressed ventricular contractility and low cardiac output states. This clinical heterogeneity complicates hemodynamic management decisions, which often rely on generalized protocols rather than individual patient physiology. Recent advances suggest that phenotyping or subphenotyping patients based on cardiovascular profiles could identify treatable subgroups and improve outcomes by tailoring therapy. Unsupervised clustering methods that integrate hemodynamic variables and transthoracic echocardiography (TTE) parameters represent a promising approach to dissect this phenotypic diversity in septic patients. This study aimed to identify distinct cardiovascular subphenotypes in patients with sepsis and to evaluate their associations with mortality and responses to hemodynamic therapies.

Study Design

This retrospective cohort study was conducted in the University Hospital ICU, Birmingham, UK. ICU patients diagnosed with sepsis who underwent TTE within 7 days of sepsis onset were enrolled into two temporal cohorts: a derivation cohort (April 2016 to December 2019) and a validation cohort (January 2020 to December 2021). No interventional treatments were administered as part of this study. Demographic data, clinical variables, TTE parameters assessing left and right ventricular function, and hemodynamic indices such as cardiac index and ejection fraction were extracted from electronic records. Unsupervised clustering algorithms applied to these data identified distinct cardiovascular subphenotypes. Primary endpoints included 90-day mortality. Secondary analyses examined interactions between subphenotypes and different vasoactive agent or fluid administration strategies.

Key Findings

A total of 995 patients formed the derivation cohort, and 804 patients comprised the validation cohort. A four-class cardiovascular subphenotype model optimally fit the data in both cohorts. The subphenotypes were characterized as follows:

• Class 1 (Normal Function): This majority group (51% derivation, 56% validation) had predominantly normal left and right ventricular systolic function and cardiac index.
• Class 2 (Hyperdynamic): Represented 30% in derivation and 22% in validation, characterized by elevated cardiac index and hyperdynamic left ventricular ejection fraction.
• Class 3 (RV Dysfunction): Comprised 10-12% of patients, exhibiting dilated right ventricles with impaired systolic function.
• Class 4 (Low Cardiac Output): Made up ~9-10% and featured low cardiac output with depressed left ventricular ejection fraction.

Mortality analyses revealed the following 90-day mortality rates: class 1 had the lowest mortality (18-20%), whereas classes 2, 3, and 4 exhibited significantly higher mortality rates ranging from 41% to 58% in both cohorts (p < 0.0001). Multivariable logistic regression confirmed classes 2-4 as independent predictors of death. Notably, relatively simple models consisting of three variables yielded high diagnostic accuracy for classifying subphenotypes across both cohorts, indicating potential for rapid bedside identification.

Importantly, the relationship between cardiovascular subphenotype and mortality was modulated by hemodynamic management strategies. Different vasoactive agents and fluid resuscitation volumes showed variable associations with outcome depending on subphenotype, underscoring the heterogeneity of treatment responses.

Expert Commentary

This study elegantly demonstrates that sepsis-related cardiovascular dysfunction is not a monolithic entity but comprises discrete physiologic states with prognostic and therapeutic implications. The identification of hyperdynamic, right ventricular failure, and low-output subphenotypes reflects well-established hemodynamic disturbances in septic shock that have traditionally been difficult to classify in ICU practice. The prognostic differentiation between these classes affirms the clinical relevance of phenotyping.

Previous literature has suggested that indiscriminate use of fluids and vasoactive drugs may contribute to poorer outcomes in subsets of septic shock patients. The current findings offer a pathway toward precision medicine by enabling tailored interventions targeted to specific cardiovascular dysfunction patterns. For instance, patients with right ventricular failure may benefit from different vasoactive support or fluid strategies than those with hyperdynamic circulation.

Limitations include the single-center retrospective design and observational nature, which restrict causality assessment and generalizability. Echocardiographic assessment is operator-dependent, though use of objective clustering may mitigate bias. Further validation in multicenter prospective studies and integration with biomarkers could refine subphenotyping.

Conclusion

This study identifies and validates four cardiovascular subphenotypes in sepsis characterized by distinct hemodynamic and echocardiographic profiles. These subphenotypes are associated with markedly different mortality risks and demonstrate variable responses to vasoactive and fluid therapies. The findings support a paradigm shift toward personalized shock management guided by straightforward echocardiographic and hemodynamic measurements. Prospective studies are warranted to test whether individualized treatment protocols based on cardiovascular subphenotypes can improve outcomes in sepsis-induced shock.

Funding and Clinical Trials

No specific funding was reported for this study. The study was a retrospective analysis and was not registered as a clinical trial.

References

1. Chotalia M, Ali M, Chotalia R, et al. Cardiovascular Subphenotypes in Sepsis. Crit Care Med. 2026;54(7):1622-1634. doi:10.1097/CCM.0000000000005993.
2. Singer M, Deutschman CS, Seymour CW, et al. The Third International Consensus Definitions for Sepsis and Septic Shock (Sepsis-3). JAMA. 2016;315(8):801–810.
3. Levi M, van der Poll T. Coagulation and sepsis. Thromb Res. 2017;149:38-44.
4. Vieillard-Baron A, Caille V, Charron C, Belliard G, Page B, Jardin F. Actual incidence of global left ventricular hypokinesia in adult septic shock. Crit Care Med. 2008;36(6):1701-1706.
5. Pulido JN, Afessa B, Masaki M, et al. Clinical Spectrum, Risk Stratification, and Outcomes of Right Ventricular Failure in Sepsis and Septic Shock: A Prospective Cohort Study. Crit Care Med. 2015;43(5):1043-1051.