Reply to Maitra and Bhattacharjee: Adherence to the Prevailing Sepsis Definition Is Quintessential to Subphenotype Identification.

From the Authors:

We thank Dr. Maitra and Dr. Bhattacharjee for their comments on our recent article on using temperature trajectories to identify sepsis subphenotypes (1). We agree that our study cohort was not restricted to patients who met the criteria for sepsis but instead included all hospitalized patients who had been admitted through the emergency department with suspected infection. We included all patients with suspected infection in this study for the following reasons: 1) dysregulated responses to infection occur on a spectrum, and the biological response to infection is unlikely to change abruptly as soon as a patient meets the current sepsis definition; 2) the definition of sepsis will likely continue to evolve, and we developed the temperature trajectory subphenotypes to be generalizable to past and future definitions; and 3) temperature trajectories derived only from patients with organ dysfunction on presentation would exclude patients with infection who later developed sepsis during hospitalization; because the development of organ dysfunction due to infection is likely in part related to the immune response, we did not want to exclude these patients from our analysis.

On the suggestion of Dr. Maitra and Dr. Bhattacharjee, we tested the association between temperature trajectory membership and mortality adjusting for the Sequential Organ Failure Assessment (SOFA) score instead of the quick SOFA score. In logistic regression, when we controlled for age, comorbidities, SOFA, and time to antibiotics, membership in the “hyperthermic, fast resolvers” group remained associated with decreased mortality risk (odds ratio, 0.55; 95% confidence interval, 0.42–0.72; P < 0.001) compared with the “normothermic” group. Membership in the “hypothermic” group was associated with increased mortality risk (odds ratio, 1.56, 95% confidence interval, 1.30–1.88; P < 0.001). These results are similar to those we obtained in the primary analyses presented in our paper.

Although the metric used to determine the accuracy of sepsis definitions is often risk of mortality, definitions developed based on that outcome may not capture the heterogeneity of the sepsis syndrome (2, 3). Developing a trajectory model based on body temperature (a biologically relevant clinical measurement) allowed us to establish subphenotypes that were disentangled from but still predictive of the outcome. Further studies are required to establish the precise biological significance of the temperature trajectory subphenotypes.