Meta-analysis confirms EGDT for sepsis is unhelpful and wasteful (PRISM)
Three large, well-conducted randomized trials around the world (ProCESS, ARISE, and ProMISe) all agreed: use of early goal-directed therapy (EGDT) for sepsis does not improve mortality or any other important clinical outcome. The Big Three sepsis trials were a death knell for the formerly ubiquitous "sepsis bundles," protocols based on the single-center 2001 trial of EGDT for sepsis. Evidence this convincing is rare in critical care, but experts at some centers noted the relatively low mortality (~25%) in the Big Three trials and continued advocating EGDT, believing it might help the sickest patients with sepsis. Each of the three trials individually could have been underpowered to find this benefit, went the argument. As it turns out, the clever designers of ProCESS, ARISE, and ProMISe thought of this ahead of time. They synchronized entry criteria, treatment protocols, outcomes, and data collection -- so they could pool their data into a meta-analysis, gaining increased power for their conclusions. That meta-analysis, called PRISM and published in the New England Journal of Medicine, concludes that EGDT did not improve survival from sepsis, even in the sickest patients, but did increase use of intensive care, vasoactive infusions, and overall costs.
EGDT for Sepsis, Seen Through PRISM
The PRISM analysis included a total of 3,723 patients at 138 hospitals in seven countries, with a subgroup analysis of the sickest patients included in the three trials. Investigators found:
90-day mortality was statistically identical for EGDT (24.9%) and usual care (25.4%)
EGDT was associated with more ICU days (mean 5.3 days vs. 4.9 days) and vasoactive drug use (1.9 days vs. 1.6 days)
There were higher costs with EGDT
No benefit from EGDT for patients with worse septic shock (higher lactate levels, predicted mortality, or with combined hypotension and hyperlactatemia)
PRISM's sickest one-third of patients by risk scoring had a 45% mortality in both the EGDT and usual care arms, and was 4x the size of the entire 2001 EGDT trial.
"We found no evidence of treatment benefit with EGDT in patients with greater severity of illness, despite using several approaches to identify subgroups of very sick patients that were considerably larger than the entire population in the trial by Rivers et al.," the researchers wrote. University of Pittsburgh's Dr. Derek Angus told MedPage Today, "It's time to roll things back and focus on the basics. It is definitely important to have a high index of suspicion, to be ready with early antibiotics and fluids, and to confirm whether patients are in shock. If they are in shock that's when we need to be aggressive." What aggressive means exactly, authors argue, will require further research on "unresolved questions regarding the most effective fluid and vasopressor regimens, the role of hemodynamic monitoring, and appropriate targets in the resuscitation of patients with sepsis and septic shock." But we already know that higher MAP targets in sepsis probably don't help, and available hemodynamic monitoring tools and techniques are impractical, imprecise or both. It's hard to see interventions based on these variables having an effect. Overuse of IV fluid resuscitation may be harmful in critical illness; perhaps a vasopressors >> fluids resuscitation strategy (a la Paul Marik) could be tried?
PRISM Pops Physician-Physiologists' Pride
EGDT seduced physicians so well because it cast us in an irresistibly heroic role: a cross between a pilot ("this patient is crashing"), an embattled general (commanding our team to "hit 'em with some big guns"), and a minor deity, benevolently manipulating our patients' physiology in real time. We were charging into new frontiers of medicine at the helm of the Starship Enterprise. Unfortunately, it turned out the shiny console buttons we were pushing weren't connected to anything. In this sense, PRISM is a bummer, a humbling reminder of the current limits on our ability to understand or control the physiologic response to disease, our tendency to exaggerate our powers, and the pitfalls that abound when we do. On the positive side, physicians developed a healthy skepticism to single-center studies with large treatment effects published in major journals.
Sepsis Care: A Brief History
Summing up: Sepsis care, pre-2001: 1. recognize it early. 2. give antibiotics, fluids, and vasopressors if needed. Monitor attentively. Sepsis care, 2017 - : See above. (Do it better.)
EGDT: Last One To Leave, Turn Out the Lights
Did EGDT's "benefit" come from a simple type I error, or was there something else afoot, as a Wall Street Journal piece wondered? There is a document that looks like an unsigned, unpublished letter to the WSJ in response to that piece, which you can (at this writing) read on Scott Weingart's EMCrit blog. It is signed only "The Department of Emergency Medicine - Henry Ford Hospital". Scott's podcast describes it as being authored by both Henry Ford Hospital and the lead author of the 2001 EGDT trial. I have not verified its authorship or authenticity. The author(s) of that document respond to the WSJ reporter's assertions of potential mishandling of data in the 2001 EGDT trial, which (the WSJ piece asserts) could have led to a false rejection of the null hypothesis (i.e., false positive result of benefit for EGDT): https://emcrit.org/wp-content/uploads/Henry-Ford-Hospital-Reply-to-WSJ-10.27.2008.pdf The internal hospital inquiry into the 25 patients who were randomized but excluded from the intention-to-treat analysis* is ostensibly described on page 4:
[A] chi-square analysis weighted for the differences in illness severity [emphasis added] between control and treatment groups ... demonstrated a statistically significant retained mortality benefit at all time points for the EGDT treatment group equal to or better than the original reported results of 263 patients.
But it's unclear why would one weight for the differences in illness severity between control and treatment groups in a randomized trial (or post hoc "hypothetical randomization of all patients" as the document puts it). Isn't that randomization's job: to ensure balance between groups? Shouldn't the data from the 25 patients (previously randomized) thus simply have been included as is, and the chi-square analysis repeated? I'm not a statistician or a researcher. Maybe someone who is can answer that question. *According to the WSJ, the 2001 trial's lead author stated in an interview " 'all 288 are randomized' -- that is, placed into either the standard-therapy group or the other group -- 'and then some are excluded because they're not considered severe septic shock.' " He contradicted this in an email, stating "there were absolutely no patients pulled out before the analysis." These statements are not clarified in the above linked response to the WSJ; it only says that the hospital's review concluded the design was appropriate and "the patients of concern, when evaluated, did not affect the results of the study. " According to the reporter, the hospital declined to share their inquiry's report. The end of this post has been edited. Read more: Early, Goal-Directed Therapy for Septic Shock — A Patient-Level Meta-Analysis. N Engl J Med. 2017 Mar 21.