Hernandez G, Hunsicker O, de Backer D, et al. Twenty-five years of septic shock hemodynamic resuscitation trials: a conceptual perspective.
Six landmark trials, 25 years, one throughline: single fixed targets keep failing to reproduce, and the field keeps learning why.
This is a perspective piece, not a new trial — 21 co-authors, including several of the actual ANDROMEDA-SHOCK and CLASSIC trial investigators, tracing the conceptual arc from Rivers' 2001 EGDT trial through the just-published ANDROMEDA-SHOCK-2 (2025/2026). Its organizing claim: progress in septic shock resuscitation over 25 years came less from new drugs or devices, and more from a deepening understanding of how to use the tools already available — fluids, vasopressors, inotropes — more intelligently.
A single-center success that three later multicenter trials couldn't reproduce — and the paper is specific about why.
Rivers' EGDT trial targeted ScvO2 via a protocolized bundle and showed a significant survival benefit in a single-center ED, largely in patients with protracted shock or profound hypovolemia. It got folded into early Surviving Sepsis Campaign guidelines. Then three larger multicenter RCTs — ProCESS, ARISE, and ProMISe — all found neutral results. The paper attributes the failure to reproduce to three specific, verifiable factors: later cohorts never showed the very low ScvO2 values Rivers reported; ScvO2 turned out to be non-dichotomous (both low and supranormal values associate with worse mortality); and the algorithm itself rested on questionable physiology — an abnormally elevated CVP target and liberal transfusion/fluid thresholds that later fluid-overload research would directly contradict. ScvO2 was formally dropped as a resuscitation target in current SSC guidance, though it's retained as one input among several in multimodal perfusion monitoring.
This is the nuance most summaries skip — and it's the strongest evidence in the whole paper against treating lactate as a direct treatment target.
The LACTATE trial targeted a ≥20% lactate decrease every 2 hours over an 8-hour window and did show lower hospital mortality and earlier SOFA improvement as secondary outcomes. But the paper flags a specific internal contradiction: the intervention arm received more fluids and vasodilator support, yet lactate kinetics were essentially unchanged between arms, and the prespecified lactate-reduction target was achieved equally often in both groups. In other words, the protocol didn't actually "treat" lactate any more effectively than usual care — the benefit, if real, came from somewhere else in the bundle.
The paper's deeper point: lactate has been misread as a direct measure of oxygen debt. Its kinetics are biphasic — flow-dependent and genuinely informative in the first 6–12 hours, but increasingly driven by heterogeneous, non-hypoperfusion mechanisms afterward: adrenergic-driven aerobic glycolysis, impaired hepatic clearance, inflammatory metabolic reprogramming, mitochondrial dysfunction. Persistent hyperlactatemia beyond the early window shouldn't be assumed to mean ongoing tissue hypoxia, especially once other perfusion markers have normalized. This is why the paper doesn't call for abandoning lactate — it still flags high-risk trajectories and now anchors part of the SCCM/ESICM refractory-shock definition — but argues against using it as a late-phase resuscitation dial to keep turning.
A detail worth knowing before you cite this trial: the "standard" arm didn't actually run at the target it was supposed to.
SEPSISPAM compared a higher MAP target (80–85 mmHg) against standard (65–70 mmHg). The paper notes the standard-target group's actual achieved MAP was 75 mmHg — well above its intended 65 mmHg ceiling — which complicates any clean reading of "65 vs. 85." No overall mortality or renal-function difference, but significantly more new-onset atrial fibrillation across the whole population at the higher target. The one subgroup signal: chronic hypertensive patients assigned to the higher target needed less renal replacement therapy, without the added AF risk seen in the general population — an early hint that a single MAP number was never going to fit every patient.
| Trial | What it actually tested | Key caveat the paper flags |
|---|---|---|
| ANDROMEDA-SHOCK | CRT-targeted vs. lactate-targeted resuscitation, 8h window | Strong trend toward lower 28-day mortality and significant organ-dysfunction benefit with CRT; a later Bayesian reanalysis supported CRT's superiority more definitively than the frequentist primary analysis did. Systematic fluid-responsiveness testing before any bolus avoided fluids entirely in 25% of patients — a process benefit easy to overlook. |
| CLASSIC | Restrictive vs. "usual" IV fluid strategy | By the time this trial ran, usual care had already drifted restrictive — so CLASSIC is best read as testing extra restriction on top of already-conservative practice, not classic liberal-vs-restrictive. Fluid responsiveness wasn't systematically assessed in the trial itself, and venous congestion/fluid tolerance weren't incorporated into treatment decisions — both now considered important omissions. |
Not "CRT wins." A multi-component, phenotype-driven bundle wins — and the paper is explicit that no single ingredient can be credited.
The newest trial (JAMA 2025/2026) tested CRT-targeted personalized hemodynamic resuscitation (CRT-PHR): a sequential bedside assessment combining CRT, pulse pressure, diastolic arterial pressure, fluid-responsiveness testing, and early echocardiography to sort patients into hypovolemic, vasoplegic, or cardiac-dysfunction phenotypes, each triggering a different intervention.
86% of the intervention arm achieved CRT normalization vs. 62% with usual care. The trial hit its hierarchical composite primary outcome with a stratified win ratio of 1.16 (95% CI 1.02–1.33, P=.04) — statistically significant, but driven mainly by shorter duration of vital support, not by a 28-day mortality difference (mortality was similar between arms, and lower than expected in both, around 26.6%). Because the intervention was a bundled, multi-component protocol, the trial cannot isolate which specific piece — CRT targeting itself, the fluid-responsiveness gating, the DAP-triggered norepinephrine titration, the MAP/dobutamine tests, or the stopping rule — actually produced the benefit.
The authors are explicit that this trial is "the beginning of this process rather than its endpoint" — not a final answer that CRT-guided personalized resuscitation is proven, but the most complete operationalization of the idea tested so far.
The caption's question — lactate or capillary refill? — is a fair entry point, but the paper's actual argument is bigger: no single fixed physiologic target, tested in isolation, has reproducibly improved patient-centered outcomes across 25 years of trying, from ScvO2 to lactate to MAP to CRT alone. What has started to work, in ANDROMEDA-SHOCK-2, is treating resuscitation as short, reversible tests within iterative reassessment cycles, guided by whichever rapidly-responsive signal fits the patient's phenotype at that moment — not a permanent switch from one number to another. Lactate remains genuinely useful as an early-window signal and a refractory-shock criterion; it was never actually "abandoned" in this paper, just correctly resized.