Sepsis-3: Of Love and Bliss
The author’s choice of a title for their viewpoint is interesting—“To systemic inflammatory response syndrome with Love,” a play on the classic 1967 movie starting the great Sidney Poitier. Because, it would seem, the authors professed love—something ephemeral, recognizable but defying easy definition—is not unlike sepsis (or, for that matter, obscenity, per the U.S. Supreme Court Justice Potter Stewart in the 1964 case “Jacobellis versus Ohio,” where he stated “I can’t define it but I know it when I see it”). Truly defining sepsis presented the Sepsis-3 task force with a number of problems. Indeed, a large part of our discussion involved the nature of what constitutes a “definition” (the definition of “definition”?). Per the Merriam-Webster dictionary, a definition is “the essence of a thing” or “what something is.” Using the most current understanding of sepsis pathobiology, the task force determined that “life-threatening organ dysfunction caused by a dysregulated host response to infection” best reflects “what sepsis is” (2). However, we recognized that clinicians require something more tangible to aid in the identification of those patients most likely to have sepsis. In this regard, we felt an obligation to assure that advocacy for any clinical criteria be based upon data-driven derivation and validation. Therefore, Seymour et al (3) began an analysis to identify patients likely to have sepsis by limiting the exercise to only those with suspected, presumed, or documented infection. The authors (and I was one) went on to use receiver operator characteristic curve analysis to examine the ability of the systemic inflammatory response syndrome (SIRS), the Sequential Organ Failure Assessment (SOFA) score (4), the Logistic Organ Dysfunction Score (5), and other approaches currently in use, to identify patients whose outcomes were consistent with a diagnosis of sepsis. We further used logistic regression to derive a new entity “quick SOFA” (qSOFA). These composites represent “clinical criteria” that either have been previously touted or, in the case of qSOFA, were derived a priori from a large database of patients in whom infection was suspected. Seymour et al (3) determined that the predictive validity of SOFA greater than or equal to 2 exceeded that of SIRS greater than or equal to 2; they further determined that, outside the ICU, the predictive validity of qSOFA greater than or equal to 2 exceeded that of SIRS greater than or equal to 2. We pointed out that their analysis and conclusions carried a number of limitations, including but not restricted to the study’s retrospective design, the U.S.-centric nature of the datasets examined and need to extend the work to region with less fully developed healthcare systems. But our response to Sprung et al (1) begins and ends with a reminder that entry into the analysis performed by Seymour et al (3) and for septic shock by Shankar-Hari et al (6) required that infection be at least suspected.
Thus, the Sepsis-3 recommendations are, by design, completely agnostic to the value of SIRS in identifying organ dysfunction and/or an aberrant host response in patients in whom infection was not suspected—because we chose to limit the analysis to infected patients and infected patients alone. We specifically did not seek “a descriptor to differentiate patients with infection from those with similar characteristics who are not infected.” All of the recommendation contained in the articles by Singer et al (2), Seymour et al (3), and Shankar-Hari et al (6) was data-driven, tested initially or derived from a large dataset and validated in four additional databases of patients with suspected, presumed, or documented infection.