This week in Critical Care
Nov 18, 2026
Optimizing Sepsis Time Zero Definitions Through Associations between Antibiotic Timing and Mortality.
Across 145 United States hospitals, antibiotic delay beyond one hour was linked to higher mortality only when sepsis time zero was defined by hypotension or lactate of at least 4 mmol/L.
American Journal of Respiratory and Critical Care Medicine · 2026 · PubMed
This week’s highlights
Read the weekly briefing
This AudioScholar briefing is generated by artificial intelligence from PubMed abstracts for healthcare professionals and is not medical advice.
This week: sepsis antibiotic urgency depends on which organ dysfunction starts the clock, a lower carbon dioxide target in ARDS shows no survival gain, and peripheral perfusion targets in septic shock trim early fluids without a proven mortality effect.
Welcome to This Week in Critical Care. We're covering 10 papers this week: 3 in depth, 4 in brief, 3 quick signals.
In the American Journal of Respiratory and Critical Care Medicine, Rhee and colleagues asked which of the SEP-1 organ dysfunction criteria actually identify patients harmed by antibiotic delay [1]. This was a retrospective cohort built from electronic health record data at 145 United States hospitals between 2018 and 2023. Of nearly 1.85 million encounters with suspected infection, about 624,000 had at least one organ dysfunction and received antibiotics within six hours of time zero. Using inverse probability weighting for baseline and time-varying confounders, antibiotics given at one to three hours rather than within the first hour were associated with about seven percent higher relative mortality when time zero was defined by hypotension, and about ten percent higher when defined by a lactate of 4 mmol per litre or more. No such association appeared for the other dysfunctions, alone or in combination. Among patients without hypotension or high lactate, giving antibiotics at three to six hours rather than within three hours was linked to higher mortality only in acute respiratory failure. The main limitation is the observational design, so residual confounding by indication remains possible. What it adds is a large, criterion-by-criterion map of where antibiotic timing appears to matter, which the authors say could inform refinements to sepsis time zero definitions and performance measurement.
Still on sepsis, Bispo and colleagues in the Journal of Critical Care pooled randomized trials of resuscitation strategies that incorporate peripheral perfusion targets in septic shock [2]. This systematic review and meta-analysis included five trials with 2,180 patients, assessed with standard risk of bias and certainty tools. Compared with conventional strategies, these approaches were not associated with a significant difference in 28 or 30-day mortality, with moderate certainty, or in 90-day mortality, with low certainty. The one consistent signal was less resuscitation fluid in the first six to eight hours, a reduction of roughly 265 millilitres, rated low certainty. There were no significant differences in length of stay, organ dysfunction, vasopressor-free days, ventilator-free days or renal replacement therapy-free days. The key limitation, which the authors state themselves, is that the intervention algorithms and comparators differed across trials, so the findings cannot be attributed to peripheral perfusion monitoring alone. The authors also note that clinically important benefit or harm cannot be excluded. As evidence, this is the most rigorous synthesis this week, and it places perfusion-targeted resuscitation as a fluid-sparing strategy without demonstrated mortality benefit.
Turning to ventilation, Giannakoulis and colleagues in Critical Care Medicine emulated a target trial comparing normocapnia with mild hypercapnia during lung protective ventilation in ARDS [3]. They drew on patients enrolled in seven ARDS Network and PETAL Network randomized trials, and from 5,367 potentially eligible patients assigned 292 to a carbon dioxide target of 35 to 45 millimetres of mercury and 292 to a target of 50 to 55, for four days after starting low tidal volume ventilation. Separation in carbon dioxide held, and normocapnia was achieved through a higher respiratory rate, a median of 33 versus 28 breaths per minute, while driving pressure, tidal volume and dead space were similar. Ninety-day in-hospital mortality was identical, with 84 deaths in each group, just under 29 percent. The null result held in older and obese subgroups. The authors caution that as an emulation, residual bias may be substantial, and the analysed sample is a small fraction of the source cohort. What it adds is evidence that pushing respiratory rate to normalise carbon dioxide was not associated with a survival advantage over tolerating mild hypercapnia.
Now, four in brief.
Complementing the perfusion meta-analysis, Rietdijk and colleagues in the Journal of Critical Care ran a post hoc analysis of the ANDROMEDA-SHOCK 1 and 2 trials, covering 1,885 patients with septic shock [4]. About 37 percent of patients had a normal baseline capillary refill time of three seconds or less. These patients had lower 28-day mortality, about 21 percent against 34 percent, with a similar gap at 90 days. Yet many still looked unwell: close to half of these patients were tachycardic, and almost all had a lactate above 2. As a post hoc association, this is prognostic information rather than evidence about treatment, and the authors call for research on how to improve care in this group.
Also a target trial emulation, Reep and colleagues in Critical Care used the MIMIC-IV database to study when to switch from controlled to assisted ventilation in 1,635 patients [5]. Delaying the switch until the oxygenation ratio exceeded 250 was associated with a longer time to successful extubation, by roughly two days on the restricted mean time measure, compared with switching above 150 or 200. Exploratory analyses suggested a slightly higher threshold of 200 for patients on higher PEEP. No mortality differences appeared across strategies. This is single-database observational evidence, and the authors state that prospective validation is needed.
Also in Critical Care, Foucrier and colleagues reported a retrospective, propensity-weighted cohort of 1,321 ventilated trauma patients from 15 French trauma centres [6]. Early-onset ventilator-associated pneumonia occurred in 32 percent of patients given antibiotic prophylaxis, mostly amoxicillin-clavulanic acid, and in 34 percent of those without it. After weighting, prophylaxis was not associated with less early pneumonia, and a competing-risk model agreed. Secondary outcomes, including mortality and ventilation duration, did not improve. The design is observational, but the null finding is consistent across analyses and argues against extending brain-injury data to trauma broadly.
In Critical Care Medicine, Zaaqoq and colleagues analysed the Extracorporeal Life Support Organization registry, with 115,849 adults across 495 hospitals from 2012 to 2023 [7]. Each additional ten annual cases was associated with about two percent lower odds of in-hospital death. Centres delivering ECMO across four or more ICUs had roughly 18 percent higher adjusted odds of death than single-ICU centres, which the authors suggest reflects patient complexity offset by higher volume. Registry data cannot exclude residual confounding, and the effect sizes are small.
Three quick signals.
In Critical Care Medicine, a Bayesian meta-analysis by Tan and colleagues of 54 studies found new-onset atrial fibrillation in about one in ten critically ill patients, with about seven percent having thromboembolism after discharge and roughly twice the relative risk [8]. Only about a third received anticoagulation. Follow-up and anticoagulation exposure varied widely.
In Annals of Intensive Care, Benghanem and colleagues reviewed how the prognostic value of examination, EEG, evoked potentials, biomarkers and imaging changes over time after cardiac arrest, arguing for timing each marker to its optimal window [9]. This is a narrative review, with no new patient data.
In Chest, Sridhar and colleagues offer a case-based, physiology-driven guide to acute right ventricular failure, covering point-of-care ultrasound, preload, afterload, contractility, airway management and escalation to right ventricular assist devices [10]. It is expert practice guidance rather than trial evidence.
Stepping back across this week's papers, much of the evidence came from observational designs dressed in trial-like structure: target trial emulations, propensity weighting and post hoc analyses of randomized data. These methods sharpen the questions but cannot remove residual confounding, and the one randomized synthesis this week found that a change in a process measure did not carry through to survival. Several of the most useful results were nulls, which narrow the space of plausible benefit.
If you only have time for one paper this week, make it the sepsis time zero analysis from Rhee and colleagues [1]. It reopens the question of whether a single antibiotic clock fits every organ dysfunction in SEP-1, with data large enough to shape how sepsis performance measures are defined.
That's your roundup for This Week in Critical Care. The full transcript and references are available on the episode page. See you next week.
If this weekly briefing is useful, follow the show in your podcast app so new episodes arrive automatically. And think of one colleague, in any specialty, who never has time to keep up with the literature. Tell them about AudioScholar: a ten-minute weekly for every specialty, to listen to in any podcast app, or to read at audioscholar dot C C.
This is an automated summary generated by artificial intelligence, which can make mistakes. Use it to choose which papers to read in full.
Get Critical Care every week
By email, Spotify, Apple or any podcast app.
Spot something worth flagging?