Oct 312014

2014 has been a rough year for advocates of sepsis protocols.

First, the long-awaited ProCESS trial did not show any benefit from use of the original early goal-directed therapy (EGDT) protocol used in the single-center 2001 trial by Rivers et al that, despite criticism, became the standard of care for the following decade. Patients cared for in the 2 non-EGDT arms of ProCESS -- one, a "softer" protocol arm not including central venous oxygenation monitoring or related interventions, and a 3rd arm of "usual care" with no protocol at all -- together received less of everything (central lines, crystalloid, dobutamine, and blood transfusions) but had statistically identical outcomes to those on the EGDT protocol.

The ARISE trial's design was similar, with half the 1,600 patients receiving classic protocol-driven EGDT and the other half cared for solely at the physician's discretion (usual care). ARISE went even further than ProCESS by challenging a core principle of early goal-directed therapy, central venous oxygenation (ScvO2) as a marker for end-organ perfusion. ARISE actually forbade measurement of ScvO2 in the usual care arm; physicians had to care for patients with no awareness of this supposedly critical variable for the first 6 hours. As in ProCESS, patients in the usual care arm of ARISE received less of everything else in the EGDT protocol or "sepsis bundle" (early crystalloid resuscitation, dobutamine, and red blood cell transfusions).

What happened? Those treated for septic shock with usual care with fewer interventions survived at precisely the same rate as those treated with guideline-compliant early goal-directed therapy, in both ARISE and ProCESS.

My cheeky headline "Sepsis Protocols Pronounced Dead" after ARISE echoed ProCESS's findings was considered irresponsible by some observers. They argued that abandoning sepsis protocols could lead to harm as physicians and team members lose the vigilance enforced by a written checklist. These studies' investigators raised similar concerns. I worried they might be right. I revised the post and the headline, to emphasize the importance of early aggressive care for septic shock, whether protocolized or not.

But after thinking harder about it, I've been wondering ... what care are they talking about? What exactly might we lose -- and where would the risk originate -- by disregarding most aspects of sepsis protocols tomorrow? Let's unpack the time-honored components of the sepsis bundle as viewed through the lens of the ARISE and ProCESS trials -- the largest and most current body of randomized trial data in sepsis care:

  • Central venous catheterization: In ARISE and ProCESS, only about 60% of patients in the non-EGDT arms received central lines, i.e., ~40% were managed with peripheral IVs. Those in ProCESS got their central lines late -- usually, more than 3 hours after sepsis was identified. Result: no detectable harm.
  • Central venous oxygen saturation: In ARISE, this variable was not checked in any patient in the usual care arm. In ProCESS, ≤4% of patients in the non-EGDT arms had serial monitoring of ScvO2. Result: no detectable harm.
  • Red blood cell transfusion: In ARISE, 14% of EGDT-treated patients got them, vs. 7% in usual care; in ProCESS, it was 14% with EGDT vs. ~8% in the non-EGDT arms. Most RBC transfusions in non-EGDT arms were for hemoglobin < 7 g/dL. Result: no detectable harm.
  • Dobutamine infusion: ARISE, 15% vs. 2.6%; ProCESS, 8% vs. ~1% -- i.e., there was almost no use of dobutamine outside the EGDT protocol arms. Result: no detectable harm.

(As to crystalloid, vasopressors, and antibiotics -- I think we can leave these out of the debate for now.)

These were not small divergences in treatment. The doctors in the usual-care arms of ARISE and ProCESS weren't treating patients with de facto goal-directed therapy using unconscious muscle memory from years of practice, blurring differences between arms. Quite the contrary: they explicitly abandoned virtually the entire algorithmic protocol for sepsis treatment and the Surviving Sepsis Guidelines, with no untoward effects, without the tiniest hint of a signal of ill effect. Without measuring central venous oxygenation, what is a sepsis protocol, exactly? And if one wants to defend or promote sepsis protocols in light of this, what is one left defending or promoting, practically speaking?

Arguments Against Believing ARISE and ProCESS

1. These patients weren't sick enough (APACHE ~15 in ARISE, ~20 in ProCESS).  There were a lot of exclusion criteria (really not that many in ARISE; ProCESS did have a lot, including MIs, decompensated CHF and GI bleeds). They don't represent the real-world sepsis we see out there.

Assume this may be true. How should we treat the sicker patients differently? Since most patients who get really sick do so over a few hours, is it proposed to initiate the standard EGDT sepsis protocol early on all patients -- which is clearly inappropriate and unnecessary in most -- or to treat with usual care as a rule and wait until the really sick patients "declare themselves" -- at which time many/most will already be leaving the supposed 6-hour window of prime opportunity to avert disaster? Or do we need a randomized trial to test EGDT protocols in the sickest patients with septic shock -- and does anyone really believe this will show a benefit, after an n of ~3,000 in ProCESS and ARISE (a large portion of whom did have the full-on-badness of "wild type" septic shock)?

2.  This doesn't disprove any specific element of the sepsis bundle, because they weren't tested individually.

You can't have your protocol and cheat it too. Any protocol must be adopted or rejected en bloc according to the best empiric evidence for its utility as a whole. The original Rivers trial protocol was adopted in its entirety without knowledge of which elements were responsible for the reported benefits -- and by the same standard, should now be rejected. The current evidence base is of vastly higher quality (much higher power, multiple centers, nations, health systems, and without an overt financial conflict of interest by its principal investigator and study institution).

3. Abandoning sepsis protocols will lead to worse care as physicians let down their guard.

What intervention is it suggested that non-vigilant physicians will overlook? Once crystalloid, antibiotics, and vasopressors are begun (hopefully all within an hour or 2), what more from the protocol can it still be argued should be done?

4. If sepsis protocols aren't helping, how do you explain the improvement in outcomes from sepsis care over the past decade?

Earlier recognition of sepsis and earlier provision of antibiotics, crystalloid, and vasopressors.

Summing Up

Physicians will always have the option for testing and treating patients with septic shock according to their own training, intuition, and beliefs about physiology. Placing central lines, analyzing a constant stream of data (which surely must be telling us a story of the patient's trajectory -- or why was it on our boards?), and the comfort of following algorithms created by "expert consensus" all reassure us of our shared intellectual power, our potency as healers, and science's power to tame natural processes.(The same phenomenon is present when doctors pore over Swan-Ganz numbers.)

But in the treatment of sepsis, the evidence shows those feelings of potency and power conferred by the protocol are an illusion, with little or no connection to the reality or direction of the patient's condition. In truth, the carefully considered actions we take in response to the data have no net effect on our patients' outcomes.

That is, as long as we give antibiotics, crystalloid, and vasopressors.

As to what's in my heart of hearts, I believe all aspects of critical care (and medical care generally) that can be protocolized or made into checklists, should. There is too much information, too much distraction, too much cognitive load in the health care environment, and protocols and checklists can help "chunk" tasks and data, encourage teamwork, and improve outcomes. The Checklist Manifesto is a great book. I have been an ardent advocate of sepsis protocols, minus the transfusion piece, since internal medicine residency. That changed with ProCESS and ARISE. The same empiricism that led me to abandon transfusion for hemoglobin > 7 g/dL (after TRICC) now compels me to abandon nearly all the other aspects of the early goal-directed therapy protocol.

Except for promptly giving antibiotics, crystalloid, and vasopressors.

Preferably in the emergency department, within 1 hour of presentation, in response to a new, universal sepsis screening tool that should be given a status co-equal to that of ECGs and troponins to quickly identify and treat myocardial infarctions.

We should all follow protocols, but only those that are actually shown to work. Strong evidence shows that sepsis protocols are not necessary for high-quality care of severe sepsis and septic shock, as long as antibiotics, crystalloid, and vasopressors are given promptly and appropriately. The burden now lies on those who believe otherwise to explain why additional interventions from a disproven, resource-intensive strategy should continue to be recommended as standard treatment.

Now It's Your Turn to Talk.

Get our weekly email update, and explore our library of practice updates and review articles.

PulmCCM is an independent publication not affiliated with or endorsed by any organization, society or journal referenced on the website. (Terms of Use | Privacy Policy)


Life after sepsis protocols: What now? (You decide.)