Source Control in Sepsis Urgent or Not So Fast?*
The investigators find that patients requiring source control (n = 1,173 [32%]) were significantly older (66.7 vs 62.8 yr; p < 0.001) and more likely to be in shock (73.9 vs 65.5; p < 0.001) with bacteremia, elevated lactate levels, and other standard measures of disease severity and multiple organ dysfunction than the 2,490 contemporary septic patients not requiring source control (1). Despite features that would predict a worse outcome for patients requiring source control, these patients fared better overall with a crude ICU mortality rate of only 21.2% versus the no source control, comparator population (25.1%; p = 0.01), even after adjusting for potential confounders. Importantly, the survival advantage in the source control group persisted even though overall compliance with the sepsis resuscitation bundles was significantly inferior to the nonsource control patients. However, the source control group ultimately ended up with a higher percentage of patients receiving appropriate antimicrobial use than the no need for source control group (57.1 vs 49.4%; p < 0.001).
In contrast to the conventional wisdom and a number of other observational studies (4–10), no differences in outcome could be detected that were attributable to the timing of the source control procedures whether source control was performed percutaneously by interventional radiologists or by surgeons in the operating room (1). A very similar, multicenter, observational study, conducted in Germany during the same time period as the current study, reported significantly worse outcomes in patients whose source control interventions were performed more than 6 hours after onset when compared with earlier intervention times (7). Other studies have found that a 12-hour cutoff is a better indicator between early and late intervention times with better outcomes observed in the early intervention groups (9, 10).
How do we reconcile these observational findings and differences in determining the need and the optimal timing for source control in septic patients? These observations need to be viewed in a clinical context with contemporary critical care practice. We must acknowledge that large observational studies from this Spanish cooperative multicenter database, and similar collaborative critical care study groups, provide extremely valuable, clinically relevant, information when addressing questions such as the value of source control in sepsis. Despite their considerable value and intrinsic utility, retrospective observational studies, looking back at outcomes from nonrandomized subgroups, are susceptible to systematic error from unintended bias (11).
The major statistical issue is allocation bias where one assumes that the study group (e.g., early intervention source control subgroup) and the comparator subgroup (e.g., late intervention group) are drawn from the same subpopulation (patients needing source control to manage their septic event). This assumption in retrospective studies, no matter how well intended, is often subject to selection bias (11). First, source control versus nonsource control septic patients are clearly different patient populations when considering the source of infection leading to sepsis.