Title: Earlier Recognition of Sepsis and Septic Shock With Sepsis-3 Criteria— It's Still Early Days!
Abstract: To the Editor: Evidence of earlier detection of sepsis would greatly support adoption of the sepsis-3 definition (1). We therefore read with great interest the recent study by Scheer et al. (2), comparatively evaluating sepsis onset based on sepsis-3 criteria in patients identified with the previous SIRS-based approach (sepsis-1). We believe the choice of study population does not permit a fair comparison of sepsis criteria and the study leaves important methodological questions unanswered, motivating us to caution against the rather strong conclusion that sepsis-3 criteria facilitate an earlier and more predictive sepsis diagnosis than sepsis-1. Notably, the study included only patients with severe sepsis and septic shock imposing two restrictions on the study population: it leaves out patients with simple sepsis according to sepsis-1 and patients detected with sepsis-3 not detected by sepsis-1. The first leads to a narrowed patient spectrum that shrinks the denominator of sepsis-1 cases that had to be successfully identified by sepsis-3 to show performance superior to sepsis-1. With the impact of simple sepsis cases unknown, patient selection may have favored sepsis-3, making the comparison unbalanced. Patients who developed severe sepsis or septic shock from simple sepsis would only be detected by this selection procedure once they developed organ dysfunction, although their septic status according to sepsis-1 may have been known before this point in time. In our opinion, evidence must first be provided that simple sepsis in the intensive care unit (ICU) should rather be classified as uncomplicated infection as per sepsis-3 and can, therefore, be deliberately excluded when comparing sepsis criteria. The second restriction disadvantages sepsis-3 as patients detected by sepsis-3 but not sepsis-1 would all formally have been detected earlier by sepsis-3. It also precludes determination of diagnostic specificity. The net effect of both restrictions on the comparison of sepsis criteria for early sepsis detection is unclear. The prospective identification of severe sepsis and septic shock and the retrospective assignment of sepsis-3 represent disparate approaches, and add uncertainty to the validity of the results. This could have been addressed by reporting how suspicion of infection was retrospectively identified, how the absence of SIRS was determined in those detected earlier with sepsis-3, if mechanical ventilation and pharmacological circulatory support masking SIRS (3, 4) were taken into account, and how many of the patients with non-ICU acquired sepsis detected earlier with sepsis-3 were SIRS negative. Unfortunately, we failed to reproduce the reported P value of 0.011 (Fig. 3A in (2)) used as evidence of superior early detection capability of sepsis-3 in ICU-acquired sepsis that, according to the Methods section, appears to come from a one-sided binomial test. Instead, when we compared 93% with 100% with a two-sided chi-squared test with Yates correction, we obtained a P value of 0.0118, which however underscores that a significant proportion went undetected by sepsis-3. Timely recognition was achieved in 86%, which means that in 14% it was delayed on day 0. Instead, the result of 93% on day +1 is emphasized in the study; the reason for granting what appears to be a grace period of 1 day to sepsis-3 remains unclear with the methodology presented. Finally, we wonder what proportion of undetected patients is considered acceptable as a tradeoff for potentially earlier recognition of patients with organ dysfunction. Evidence for global superiority of either set of sepsis criteria, in our view, requires studies with unselected patients and transparent methodology. Sincerely yours,