Give me the current evidence for early goal-directed therapy in septic shock — and be specific about where the data weakened after ProCESS, ARISE, and ProMISe. What do we actually still follow from the Rivers trial?
Reading File
Finding Sources
Searching PubMed
"septic shock" AND "goal-directed therapy" AND "ProCESS OR ARISE OR ProMISe"
Searching the Web
EGDT septic shock ProCESS ARISE ProMISe what remains from Rivers trial 2024 2025 guidelines
Reading Article
Reading Article
Finding Sources
Searching PubMed
"septic shock" AND resuscitation AND guidelines
Reading File
Reading Article
Searching PubMed
"surviving sepsis campaign" guidelines 2021
I now have enough high-quality data to write a thorough, specific answer. Let me compile this.
Early Goal-Directed Therapy in Septic Shock: What the Evidence Actually Shows
The Rivers Trial (2001) - The Original Claim
Rivers et al. published their single-center RCT at Henry Ford Hospital (Detroit, n=263) in NEJM 2001. They enrolled patients with severe sepsis/septic shock - defined by at least 2 SIRS criteria plus either refractory hypotension (SBP <90 after 20-30 mL/kg fluid) or lactate ≥4 mmol/L. The EGDT protocol ran for the first 6 hours and targeted:
- CVP 8-12 mmHg (fluid boluses)
- MAP ≥65 mmHg (vasopressors)
- ScvO2 ≥70% - if not achieved after CVP/MAP targets: transfuse to Hct ≥30% first, then dobutamine up to 20 mcg/kg/min
- Urine output ≥0.5 mL/kg/hr
Result: 28-day mortality 30.5% (EGDT) vs 46.5% (usual care) - an absolute reduction of 16%, NNT ~6. This was a massive effect size that launched 15 years of protocol adoption and the Surviving Sepsis Campaign.
Context that matters: The control group was poorly resuscitated by modern standards. Mean ScvO2 at baseline was ~49%, lactate was 7.7 mmol/L (control) and 6.9 mmol/L (EGDT). Control-group mortality of 46.5% was exceptionally high - far above what even the sickest patients now experience in high-resource ICUs.
Where the Data Weakened: ProCESS, ARISE, ProMISe
All three trials ran roughly simultaneously, enrolled 2012-2014, and published in NEJM 2014-2015. Their combined patient-level data were then pooled in the PRISM meta-analysis (NEJM 2017, PMID 28320242, n=3723, 138 hospitals, 7 countries) - the definitive synthesis.
Key findings:
| Rivers (2001) | ProCESS (2014) | ARISE (2014) | ProMISe (2015) | PRISM Meta-analysis (2017) | |
|---|---|---|---|---|---|
| n | 263 | 1,351 | 1,600 | 1,260 | 3,723 |
| Control mortality | 46.5% | 18.9% (60d) | 18.8% (90d) | ~29% (90d) | 25.4% (90d) |
| EGDT mortality | 30.5% | 21.0% | 18.6% | ~29.5% | 24.9% (90d) |
| Difference | -16% | +2.1% (NS) | -0.2% (NS) | ~+0.5% (NS) | AOR 0.97 (95% CI 0.82-1.14), p=0.68 |
Why the effect vanished - the key mechanistic explanations:
1. Secular improvement in usual care ("control group contamination")
The control groups in the three trials had already absorbed the lessons of Rivers. They received fluid boluses before randomization (>1000 mL in ARISE/ProMISe, ~20-30 mL/kg in ProCESS), antibiotics were given early, and ICU care was far better than a 2001 Detroit ED. The Washington Manual notes this explicitly: "these studies were limited by practice changes in control group." A control-group mortality of 18.9% (ProCESS) vs 46.5% (Rivers) tells you the two populations are not comparable.
2. Pre-randomization resuscitation had already occurred
In ProCESS and ARISE, patients received substantial fluid and sometimes vasopressors before randomization. The mean CVP in control groups at 6 hours was 11-12 mmHg in all three trials - nearly identical to EGDT groups. The hemodynamic targets of EGDT were being hit in usual care patients anyway.
3. The ScvO2 and CVP targets were already being met without protocol
In Rivers, baseline ScvO2 was ~49%. By the time ProCESS/ARISE enrolled (2012-2014), usual care clinicians were already recognizing and treating overt shock earlier. At 6 hours, ScvO2 was ~74-76% in EGDT groups vs ~66-72% in usual care groups - a much smaller gap than Rivers showed.
4. The specific Rivers interventions were actively harmful in context
- Red cell transfusion to Hct ≥30%: EGDT patients received roughly twice as many transfusions as usual care patients. The TRISS trial (RCT, n=998) separately showed a Hgb threshold of 7 g/dL was non-inferior to 9 g/dL in sepsis. More transfusions = no benefit and potential harm.
- Dobutamine for ScvO2 <70%: Given to far more EGDT patients. There is no independent evidence this improves outcomes in this setting; it may increase myocardial oxygen demand.
- Central venous catheter requirement: EGDT mandated a specialized ScvO2-capable CVC (e.g. Edwards Lifesciences catheter). This added cost and procedural risk with no demonstrated benefit.
5. PRISM subgroup analyses found NO subset that benefited
The PRISM meta-analysis prospectively tested 16 patient characteristics and 6 care-delivery characteristics. Even patients with worse shock (higher lactate, combined hypotension + hyperlactatemia, higher predicted risk of death) showed no benefit from EGDT. Hospitals with lower baseline propensity to use vasopressors or fluids also showed no benefit.
6. EGDT was more expensive and resource-intensive
PRISM: EGDT patients spent significantly more days on cardiovascular support (1.9 vs 1.6 days, p=0.01) and in the ICU (5.3 vs 4.9 days, p=0.04). Average costs were higher in EGDT arms, with no mortality benefit.
The 2025 Heterogeneity Data (PMID 39440873, Crit Care Med 2025)
This secondary analysis of ProCESS + ARISE (n=~2,900, 81 sites) used machine learning (local-linear random forest) to identify heterogeneity of treatment effect. It found that treatment response varied substantially across patients:
- Quintile 1 (lowest predicted benefit): ARD -8% (95% CI -19% to +4%) - suggesting EGDT may have harmed these patients
- Quintile 5 (highest predicted benefit): ARD +12% (95% CI +1% to +23%) - suggesting real benefit
The strongest predictor of heterogeneity was pre-intervention albumin level. This suggests a phenotype-based approach may eventually identify who benefits - but the tools to do this at the bedside remain unvalidated.
What We Still Follow From Rivers
The Rivers trial did not prove EGDT was uniquely effective - it proved that doing something systematic and early for septic shock saves lives compared to doing it late and haphazardly. The components that survived are:
1. Early recognition and time-sensitive treatment
The "1-hour bundle" concept. The Surviving Sepsis Campaign (SSC) 2021 guidelines strongly recommend:
- Blood cultures before antibiotics
- Broad-spectrum antibiotics within 1 hour of recognition
- 30 mL/kg IV crystalloid for hypotension or lactate ≥4 mmol/L
- Vasopressors if MAP <65 persists during/after fluids
2. Lactate measurement and clearance as a resuscitation endpoint
Rivers pioneered the concept of tissue hypoperfusion as the target. Lactate remains a cornerstone. SSC 2021 recommends:
- Measure lactate; if ≥2 mmol/L, remeasure at 2 hours
- Target lactate normalization rather than a fixed ScvO2 number
- The Washington Manual (Surviving Sepsis source text): "Lactate clearance is associated with improved mortality in septic patients... targeting resuscitation to normalize lactate in patients with elevated lactate levels"
3. MAP ≥65 mmHg as the vasopressor target
This specific numeric threshold from Rivers is retained in SSC 2021. Norepinephrine remains first-line (the Rivers protocol used vasopressors to target MAP ≥65 - this principle holds).
4. Reassessment and dynamic monitoring
The concept of assessing fluid responsiveness rather than just giving fluid. The ESICM 2025 circulatory shock guidelines (PMID 41236566) now explicitly recommend dynamic variables over static CVP for assessing preload - a direct repudiation of the CVP target from Rivers, while preserving the principle of goal-directed hemodynamic assessment.
5. Source control
Always was, still is, critical. Rivers' framework for structured source control is retained.
What Was Abandoned
| Rivers Element | Current Status |
|---|---|
| CVP 8-12 mmHg target | Abandoned - CVP is a poor predictor of fluid responsiveness (SSC 2021, ESICM 2025) |
| ScvO2 ≥70% target | No longer a mandatory endpoint - lactate clearance substituted |
| Transfuse to Hct ≥30% | Abandoned - liberal transfusion harms; Hgb threshold now 7 g/dL (SSC 2021, TRISS) |
| Dobutamine for ScvO2 | Not routinely recommended; reserved for clinical evidence of low CO |
| Protocolized 6-hour bundle (rigid) | Replaced by individualized resuscitation with dynamic reassessment |
| Specialized ScvO2 CVC requirement | Not required |
Bottom Line
The three trials (ProCESS, ARISE, ProMISe) and the PRISM patient-level meta-analysis are unambiguous: protocolized EGDT as Rivers described it does not reduce mortality compared to good usual care, and is more expensive. The effect in Rivers was almost certainly driven by how bad usual care was in 2001, not by anything unique to the CVP/ScvO2 protocol.
What Rivers got right - and what we still practice - is the core principle: identify septic shock early, give fluids and antibiotics fast, start vasopressors early for MAP <65, and use a measurable physiologic endpoint (now lactate, not ScvO2) to guide resuscitation. The ESICM 2025 guidelines further emphasize dynamic preload assessment and echo-based phenotyping over static filling-pressure targets.
The 2025 heterogeneity data (Shah et al., Crit Care Med) add an important caveat: EGDT may still benefit a phenotypically distinct subset - possibly hypoalbuminemic patients with profound inflammatory activation. This is a hypothesis-generating finding that has not changed guidelines, but it explains why some clinicians in specific contexts may still use ScvO2-guided resuscitation.
Key citations:
- Rivers et al., NEJM 2001 (original EGDT trial)
- ProCESS Investigators, NEJM 2014
- ARISE Investigators, NEJM 2014
- Mouncey et al. (ProMISe), NEJM 2015
- PRISM Investigators/Rowan & Angus, NEJM 2017 [PMID 28320242] - patient-level meta-analysis
- Shah et al., Crit Care Med 2025 [PMID 39440873] - heterogeneity of treatment effect
- Holst et al. (TRISS), NEJM 2014 - transfusion thresholds
- SSC Guidelines 2021 (Evans et al., Crit Care Med)
- ESICM Circulatory Shock Guidelines 2025 [PMID 41236566]
- Washington Manual of Medical Therapeutics (septic shock chapter)
This is a shared conversation. Sign in to Orris to start your own chat.