In the 1990s, an elite rower had a recorded venous pH of 6.74 after a 2,000-meter race. Rather than collapsing or requiring hospitalization, he caught his breath, chatted with the crowd, and drove home.

Elite sprinters routinely hit pHs of 7.10 with lactate levels exceeding 10 mmol/L. Even weekend-warrior marathoners can easily enter the pH range of 7.20 to 7.30 near the end of a race.

So who’s afraid of a few extra stray hydrogen ions?

Well, the etiology matters: exertional metabolic acidosis is brief, self-limited, and occurring in healthy people with intact cardiopulmonary-renal function. By contrast, critically ill patients have ongoing acid generation from organ dysfunction with diminishing physiologic reserves. Acidemia can depress myocardial function and blunt catecholamine responsiveness, although the thresholds at which this occurs have been debated.

But many critically ill patients with acidemia have something distantly in common with those elite athletes: their acidemia is an adaptive response to stress, not necessarily harmful, and will resolve without consequence after the underlying stressors are relieved.

Others have severe acidemia from an unrelenting metabolic decompensation near death. For them, buffering out some H+ may make their labs look marginally better for a short time, but will have zero impact on the ultimate outcome.

Is there a third group, the Goldilockses of the ICU, who have acidemia that is both maladaptive (i.e., worsening an already bad situation) and remediable?

A series of randomized trials have sought to find this elusive group: BICARICU-1 and BICARICU-2, which we reviewed here —

— and most recently the SODa-BIC trial in NEJM.

The SODa-BIC Trial

In 55 ICUs in seven countries (in Australasia, Brazil, Japan, Canada, and Ireland), 500 critically ill adults with vasopressor-dependent shock and metabolic acidosis (pH < 7.30) were enrolled.

It was a sick group. About 40% had septic shock; ~25% had metabolic acidosis attributed to cardiogenic shock or cardiac arrest. Most had mild acute kidney injury (Cr ~1.5 mg/dL). About 80% were mechanically ventilated.

They were randomized to receive either 5% sodium bicarbonate (600 mmol/L) or placebo (5% dextrose) infusions for ≤5 hours to achieve pH ≥7.30, and followed for 30 days to assess the incidence of a “major adverse kidney event”: either death, need for dialysis or CRRT, or a persistent doubling (or greater) of baseline creatinine. (This composite endpoint, reused across trials, is referred to as MAKE30.)

SODa-BIC Results:

There was separation between groups, and bicarbonate improved acidosis. During the first 24 hours, the sodium bicarbonate group had pH values ~0.05 higher with serum bicarbonate levels than placebo.

But after 30 days, the rate of the MAKE30 composite outcome was no different between arms — 40.2% in the NaHCO3 group and 39.4% in the placebo group.

In-hospital mortality was near-identical (25.4% for NaHCO3 vs 24.0% with placebo).

However, sodium bicarbonate-treated patients had persistent renal dysfunction less often, numerically (14% vs 18% with placebo), required dialysis/CRRT 4% less often overall (16.8% vs 20.9%) and were less often dialysis-dependent at 30 days (5.7% vs 9.1%)—none of which were statistically significant.

There were no significant safety events, but four patients in the sodium bicarb group had significant hypokalemia (<3.0 mmol/L) requiring replacement, vs. none in the placebo arm.

Headlines announced “Sodium Bicarbonate Failed to Improve Kidney Outcomes in ICU” — which is technically true, but there’s more to consider.

Read more

Sepsis resuscitation decisions exist in a tension between two competing instincts: give fluids aggressively to restore perfusion, or limit initial fluid boluses in favor of early vasopressors to avoid potentially harmful fluid overload.

In the 1999 EGDT trial (Rivers NEJM), by repeatedly bolusing patients to a central venous pressure goal, the intervention arm received five liters of crystalloid (and the controls three-and-a-half) in the first six hours, with both groups exceeding +13 liters infused in the first 72 hours. Perhaps unsurprisingly, more than 60% required mechanical ventilation.

With its likely spurious signal of a mortality benefit in the goal-directed arm, this trial encouraged a generation of intensivists to continue the longstanding practice (which predated Rivers) of infusing unnecessarily large volumes of fluids into septic patients in the hope of restoring and maintaining perfusion.

Fifteen years later, the Process, Arise, and Promise trials (c. 2014) refuted the Rivers “goal-directed” package of dobutamine, transfusion, and central venous catheterization, without challenging the dominant fluid-forward paradigm of sepsis care. Patients in those trials received ~2 liters before randomization and another 2-3 liters in the first six hours (~3.5 to 5.5 liters total in 6 hours, generally).

Yet those three trials also showed clinical practice moderating toward less-aggressive fluids, with (e.g.) patients in Process and Arise “only” receiving an additional 4.5 to 5 liters from 6 to 72 hours (compared to ~8 liters in Rivers). There was no difference in mortality between arms in any of the three trials.

This set the stage for two large randomized trials testing a more restrictive (i.e., vasopressors-forward) approach to fluid resuscitation in sepsis after an initial bolus of ~2-3 liters. Neither showed a significant difference between the two fluid strategies:

• CLASSIC (NEJM 2022, n=1,554, Europe): identical 90-day mortality (very high at ~42%), with only 500 mL (restrictive) vs 1,300 mL (liberal) infused on the first day after randomization (plus ~3 liters in each group before enrollment);

• CLOVERS (NEJM 2023, n=1,563, U.S.): 90-day mortality 14% (restrictive) vs 14.9% (liberal), patients received ~2 liters before randomization, then 500 mL (restrictive) vs. 2,300 mL (liberal) in the first 6 hours; only 37% of liberal-arm patients required vasopressors vs 59% (restrictive) on the first day

Not to be outdone, the Australasian critical care consortium ANZICS recently published results from the ARISE FLUIDS trial, which will not extinguish the controversy over this deceptively simple-seeming clinical question.

ARISE FLUIDS Trial

At 51 sites in Australasia, 1,000 adults presenting to the ED receiving a diagnosis of sepsis, high lactate and hypotension after ≤1,000 mL I.V. fluid (but ≤2,000 mL) were randomized to six hours of either a restrictive (vasopressor-forward) or liberal fluid strategy (+1,000 mL then 500 mL boluses, 30 mL/kg in 3 hours advised, plus vasopressors after that or when deemed fluid-unresponsive).

There was meaningful separation: after the ~1,500 mL pre-randomization, restricted patients received only an additional ~500 mL in the first 6 hours vs. ~1,500 mL in the liberal arm (~2 L vs. ~3 L), with a ~1.1 L difference at 24 hours (~2,640 mL vs ~3,740 mL).

Fluid restriction led to higher vasopressor use in the first 24 hours (87% vs 68%)—about an additional one in five patients requiring vasopressors in the restricted arm.

No difference in the primary outcome: both arms had an identical number of 76 days alive and out of the hospital through day 90.

Mortality at 28 days (12.9% vs 10%) and 90 days (16.4% vs 14.4%) was nominally higher in the vasopressor group (non-significant).

There was eightfold more pulmonary edema in the liberal group: 5% vs. 0.6%, statistically significant. Pulmonary edema was identified by clinician opinion, based on hypoxemic respiratory failure, chest radiography, or other supporting findings. Clinicians were unblinded to treatment assignment, but were usually caring for study patients well past the six-hour mark, and thus not involved directly in the trial.

Despite this increase in clinician-identified pulmonary edema, the rate of mechanical ventilation was equal in both groups (15%).

Vasopressor-forward patients required ICU admission more often (77% vs 68%). Understandable, as many hospitals require vasopressors to be managed in ICUs.

Patients in the liberal group also required about one additional day of mechanical ventilation and renal replacement therapy—although these were not statistically significant.

Central lines were largely unnecessary. Most patients received vasopressors through a peripheral IV, without any major safety signal reported in the trial.

What to Take From Arise-Fluids?

Arise-Fluids “settles” the controversy of fluids-versus-vasopressors for septic shock by again demonstrating their general equivalence when all patients’ outcomes are combined.

But partisans can find favor in its data for their preferred tilt in either direction:

1. Restrict Fluids! After an initial fluid bolus of one or two liters for patients without volume depletion, restricting fluids and relying on vasopressors to maintain perfusion is reasonable. 30 cc/kg was always an arbitrary figure; over-resuscitation causes pulmonary edema and possibly adds time on the ventilator or dialysis circuit. Iatrogenic volume overload is an under-recognized cause of disability and incomplete recovery from critical illness, which Arise-Fluids did not measure.

2. Bolus Away! Liberal fluid resuscitation is generally superior from a practical clinical standpoint. A fluid-forward approach reduces the need for vasopressors, which in turn can reduce the need for ICU admission, conserving resources and improving care delivery system-wide. Pulmonary edema is not a hard clinical outcome and did not increase the need for mechanical ventilation either in Arise-Fluids or in Clovers, the most comparable major trial. The trend toward faster weaning from mechanical ventilation in the restrictive group in Arise-Fluids was not significant. Further, the numerically higher mortality in the restrictive arm is a potential warning of the limits of safe fluid restriction.

Fluids vs Vasopressors: A False Dichotomy

U.S. clinicians have appropriately chafed at policymakers’ attempts at national standardization of sepsis care via the governmental “quality” measure called SEP-1; this resistance has been in large part because of its mandate for providing three, four, or five-liter boluses to severely obese patients with sepsis who may also have concomitant cardio-pulmonary-renal failure.

Subsequent iterations of the rule permit clinicians to deliver smaller fluid boluses they consider beneficial, with documentation of their clinical reasoning.

The Arise-Fluids and Clovers trials together support the safety of just such an individualized, patient-centered approach to resuscitation and hemodynamic management. Deferring vasopressors for aggressive fluids for tolerant or volume-depleted patients; limiting fluid boluses and favoring vasopressors in patients with apparent impending respiratory failure or significant cardio-renal disease: the complexity of sepsis and each patient’s physiology necessitate thoughtful tailoring of their plan, rather than blindly following a trial result.

Rather than picking a winner in the fluids-vs-vasopressors debate, the Arise-Fluids trial reaffirms the truth that the best approach to sepsis resuscitation is the one tailored to the patient in front of you.

References

ARISE-FLUIDS trial. Vasopressors or Fluids in Early Septic Shock. New England Journal of Medicine. Published online June 11, 2026. doi:https://doi.org/10.1056/nejmoa2516225

Meyhoff TS, Hjortrup PB, Wetterslev J, et al. Restriction of Intravenous Fluid in ICU Patients with Septic Shock. New England Journal of Medicine. 2022;386(26):2459-2470. doi:https://doi.org/10.1056/nejmoa2202707

‌‌The National Heart, Lung, and Blood Institute Prevention and Early Treatment of Acute Lung Injury Clinical Trials Network. Early Restrictive or Liberal Fluid Management for Sepsis-Induced Hypotension. New England Journal of Medicine. 2023;388(6). doi:https://doi.org/10.1056/nejmoa2212663

‌

Attending to the complex needs of patients after their ICU stays—physical, social, and psychological—seems intuitively certain to improve their long-term recovery.

Clinical evidence has not clearly shown that to be the case, but the overall quality of research in this underfunded area has been limited.

At first glance, the latest data from a rare multicenter randomized trial may seem uninspiring, but a closer look provides some encouragement that focused post-ICU interventions might yield benefits to some patients.

The iRehab Trial

At 52 hospitals in the U.K.’s National Health System, 429 ICU survivors of critical illness with ≥48 hours of mechanical ventilation were recruited within 12 weeks of their hospital discharge.

They were randomized to receive either usual care, or a lot of support for six weeks. This included a remotely administered individualized rehabilitation plan that focused on symptom management, exercise targeted at their weak areas, psychological support, and connections to a peer support group.

After eight weeks, there was no improvement in health-related quality of life (0.69 vs. 0.67 on the EuroQoL 5-dimension 5-level questionnaire).

However, the prespecified subgroup who were intubated for 7 days or less (about half in each arm) did experience a benefit (0.74 vs. 0.66 on the QOL metric, statistically significant).

Leg strength and exercise capacity were improved in the intervention group: they could perform 10.6 sit-to-stand repetitions in 30 seconds, vs 9.9 in the controls. They reported less anxiety and fatigue on validated scales, as well.

A Highly Selected Group: Relatively Young and Strong

The patients enrolled in iRehab were highly selected: out of 3,705 screened, 429 were randomized.

But all 3,705 had already been screened with a much more difficult test: surviving an episode of critical illness with mechanical ventilation (>7 days for half of them).

In most trials of early mobilization, the enrolled patients haven’t yet passed this test, which is a major signal of robustness, reserve, and resilience.

The most telling proof of this fact? After six months, only three patients had died (<1%). Their youth (mean age of 55) also aligns with this. They were already relatively strong: able to sit, stand, sit, etc every 3 seconds, 10 times in 30 seconds.

They were also motivated: a third of the excluded patients (~1,100) declined to participate, further distilling the group to those most dedicated to enhancing their quality of life.

Hope for Post-ICU Syndrome

Today, about 35 clinics in the U.S. are dedicated to helping people recover or manage their symptoms after ICU discharge. Almost all are at academic medical centers or large integrated healthcare systems (e.g., Brigham, Pittsburgh, UCLA, Mount Sinai, U of M, Geisinger, and Intermountain).

Until the iRehab trial, evidence has been lacking to support their resource-intensive, multidisciplinary, high-touch approach (e.g., three-hour intake visits including speech therapy, PT, OT, a pharmacist, nutritionist, and multiple physicians and nurses, plus peer support groups and additional services outside the clinic).

A recent JAMA Insights piece (a mini-review article) leans heavily on caveats like “difficult to ascertain … data are sparse … may promote … may decrease … may prevent … may enhance” … Clinicians have always agreed that patients with specific impairments like dysphagia, weakness, malnutrition, or anxiety should receive targeted therapies and support. But the limited research on PICS hadn’t yet established a benefit from a more holistic approach to the syndrome, nor the clinics that coordinate care.

iRehab provides the first convincingly encouraging signal that these efforts can yield benefits.

But not for everyone. iRehab enrolled younger patients who had passed what for most of them would be the defining stress test of their lives: surviving days or weeks of mechanical ventilation.

Intensive multimodal rehabilitation helped some of them recover more strength than patients who did not receive the intervention. And the subgroup analysis suggests that those with a shorter duration of ventilation may have seen small but meaningful gains in quality of life.

Post-ICU clinics prescribe hope, not just clinical interventions. The iRehab trial adds evidence that at least some ICU survivors can benefit from this focused additional care after they make it back home.

Whether the frail or elderly can also benefit would require another trial more representative of the standard population of today’s ICUs.

References

O’Neill et al. Remote Multicomponent Rehabilitation in Intensive Care Unit Survivors: A Randomized Clinical Trial: JAMA 2026

Post–Intensive Care Syndrome. JAMA. Published online January 15, 2026. doi:https://doi.org/10.1001/jama.2025.23666. Butcher BW.

Outcomes after Critical Illness. The New England Journal of Medicine. 2023. Herridge MS, Azoulay É.

Prevalence of Post-Intensive Care Syndrome Among Intensive Care Unit-Survivors and Its Association With Intensive Care Unit Length of Stay: Systematic Review and Meta-Analysis. PloS One. 2024. Ayenew T, Gete M, Gedfew M, et al.

Post-Intensive Care Syndrome as a Burden for Patients and Their Caregivers: A Narrative Review. Journal of Clinical Medicine. 2024. Schembari G, Santonocito C, Messina S, et al.

Post-Intensive Care Syndrome: A Concept Analysis. International Journal of Nursing Studies. 2021. Yuan C, Timmins F, Thompson DR.

Post-Intensive Care Syndrome and Its New Challenges in Coronavirus Disease 2019 (COVID-19) Pandemic: A Review of Recent Advances and Perspectives. Journal of Clinical Medicine. 2021. Nakanishi N, Liu K, Kawakami D, et al.

Communicating to Patients and Families About Post-Intensive Care Syndrome. Chest. 2025. Rolfsen ML, Wilcox ME, Mart MF, et al.

Delirium-Related Psychiatric and Neurocognitive Impairment and the Association With Post-Intensive Care Syndrome-a Narrative Review. Acta Psychiatrica Scandinavica. 2023. Ramnarain D, Pouwels S, Fernández-Gonzalo S, Navarra-Ventura G, Balanzá-Martínez V.

Prevalence and Incidence of Post-Intensive Care Syndrome Among Intensive Care Unit Survivors: A Systematic Review and Meta-Analysis. Annals of Medicine. 2026. Zare-Kaseb A, Sanaie N, Sarmadi S.

Resource Document on the Neuropsychiatric Symptoms of Subacute and Chronic Long COVID. American Psychiatric Association (2024). 2024. Aisha Gillan MD, Melissa Peace MD, Davin Quinn MD, Jon Levenson MD, Thida Thant MD

Post Intensive Care Syndrome (PICS): An Overview of the Definition, Etiology, Risk Factors, and Possible Counseling and Treatment Strategies. Expert Review of Neurotherapeutics. 2021. Ramnarain D, Aupers E, den Oudsten B, et al.

Post-Intensive Care Syndrome After Critical Illness: Incidence and Predictors in a Nationwide Cohort. Anaesthesia. 2026. Oh, Song IA.

Guideline on Multimodal Rehabilitation for Patients With Post-Intensive Care Syndrome Critical Care. 2023. Renner C, Jeitziner MM, Albert M, et al.

Needham DM, Davidson J, Cohen H, et al. Improving long-term outcomes after discharge from intensive care unit: report from a stakeholders’ conference. Critical Care Medicine. 2012.

A 79-year-old woman nursing home resident with chronic kidney disease and chronic back pain on long-term oxycodone is admitted to the ICU with septic shock from healthcare-associated pneumonia. She is intubated for hypoxemic respiratory failure and provided cefepime, vancomycin, hydrocortisone, and a fentanyl infusion.

On hospital day 3, fentanyl is discontinued for somnolence (RASS -3). Over the next 24 hours, she awakens but remains inattentive and does not follow commands. She has prolonged episodes of staring. An EEG during an episode shows epileptiform discharges >2.5/sec. She has no past seizure history. Lorazepam and levetiracetam are provided, and neurology consultation is requested.

Laboratory studies (including sodium, magnesium, calcium, pH, and glucose) are unremarkable except for creatinine 3.1 mg/dL, increased from 1.4 mg/dL at admission, with BUN 82 mg/dL. Urine output is 0.7 mL/kg/hr. CT head shows no acute abnormality.

Read more

Most critically ill patients either arrive anemic or become so during their ICU stays. Many years ago, they were routinely transfused to near-normal hemoglobin targets until the TRICC trial (NEJM 1999) strongly suggested that withholding red cell transfusion until hemoglobin fell to ≤7 g/dL was safe (but not superior, from a mortality or secondary-outcome standpoint).

In a rare example of replicability, dozens of subsequent randomized trials and meta-analyses generally confirmed the safety of restricting transfusion to Hb ≥ 7 or 8 g/dL for most hospitalized patients, including the critically ill and those with active GI bleeding. Restrictive strategies can reduce the total volume of transfused blood by ~50%, in aggregate, without apparent harm (or benefit).

Guidelines from multiple specialty societies have since advised a restrictive approach to transfusion for hospitalized patients generally.

Patients with ischemic heart disease have been a lingering question mark. Those with significant cardiac morbidity, and especially active cardiac ischemia, were either excluded from these trials or were treated off-protocol with liberal transfusions.

The MINT trial is the only randomized study of any significant size to test transfusion targets in patients with acute myocardial infarction. Its few exclusion criteria produced a very ill, real-world population.

In MINT (n=3,500), liberally transfused patients received more than three times as much blood (~2.5 units per patient vs 0.7 per patient). Blood was spared, but the results were not reassuring, safety-wise:

• Recurrent myocardial infarction or death, the composite primary outcome (~17% in restrictive, 14.5% in liberal)

• Death (~10% in restrictive, ~8% in liberal)

• Recurrent MI (8.5% in restrictive, ~7% in liberal)

• Cardiac death (5.5% in restrictive, ~3% in liberal)

Confidence intervals included 1.0, but just barely, suggesting occasional harm.

Guidelines Tilt Toward Liberal Transfusion for Acute MI

Swayed by the MINT data, major professional societies gently nudged clinicians toward more liberal transfusion of patients with acute coronary syndromes:

• AABB recommended transfusing to Hb ≥10 g/dL for hospitalized patients with acute myocardial infarction, in 2025.

• ESICM weakly recommended a target hemoglobin of 9 to 10 g/dL, citing low-certainty evidence.

• A U.S. critical care society recommended against using a restrictive strategy (Hb 7 to 8 g/dL) based on trials using 10 g/dL as the permissive comparator.

A new randomized trial recently added to the body of data on red cell transfusions for patients at high risk for perioperative MIs.

The TOP Trial

TOP enrolled adults at 16 Veterans Affairs medical centers in the United States from 2018 through 2023 who had recently undergone major vascular or general surgery, were anemic (Hb <10g/dL), and at elevated cardiac risk (by RCRI).

More than 1,400 patients were randomized to either transfusion and maintenance of ≥10 g/dL until discharge, or transfusion only when hemoglobin was <7 g/dL.

More than 90% underwent vascular surgery, half had diabetes, and peripheral arterial disease and coronary artery disease were common (i.e., at high risk for periop MIs and strokes).

Transfusion rates (2 units vs 0 median) and hemoglobin separation (~2 g/dL divergence at day 5) were meaningful.

But at 90 days, the composite of all-cause death, myocardial infarction, coronary revascularization, acute kidney failure, or ischemic stroke had occurred almost equally often in 9% of patients in the liberal group vs. 10% in the restrictive group, a nonsignificant difference.*

*To build their power calculations, the authors over-pessimistically assumed the primary outcome would occur in 30% of the restricted patients, and over-optimistically assumed that liberal transfusion would reduce this occurrence by 25% (to 22.5%). This decision had the salutary effect (for the trialists) of drastically reducing the required enrollment to an attainable level, but reduced the TOP trial’s power to detect smaller effects.

Fewer Non-MI Cardiac Complications in the Liberally-Transfused

The liberally transfused patients had a lower rate of cardiac complications other than MI (arrhythmias, heart failure, and nonfatal cardiac arrest): 5.9% with liberal transfusion, 9.9% with restriction. This was a prespecified endpoint and was statistically significant.

Rates of the individual components of the primary outcome (mortality, MI, AKI, stroke) occurred infrequently and did not differ statistically between groups.

Does TOP “Refute” MINT?

TOP was not an acute MI trial, instead enrolling postoperative patients at high cardiac risk, almost none of whom had an active myocardial infarction at enrollment. MINT enrolled patients already experiencing MIs; TOP sought to prevent them in the perioperative period.

TOP’s topline negative finding suggests that high cardiac risk alone isn’t reason enough to routinely transfuse every postoperative patient to hemoglobin ≥10 g/dL.

But nor does it prove that hemoglobin 7 g/dL is always safe in cardiovascularly high-risk surgical populations. The confidence interval leaves room for benefit or harm, and the secondary cardiac-complication signal aligns with MINT.

TOP was negative, but it was also designed with unreasonably high event rate assumptions and was underpowered to detect small but potentially clinically important differences. (It was also stopped slightly early for funding concerns, but even at full enrollment would have remained underpowered.)

Other Trials Didn’t Show Benefits of Prophylactic Transfusions Either

More support for the thesis that every high-risk patient need not be liberally transfused perioperatively comes from the FOCUS trial (Carson et al NEJM 2011). In more than 2,000 older patients undergoing hip fracture surgery who had cardiovascular disease or cardiovascular risk factors, liberal and restrictive transfusion strategies had similar outcomes (although MIs were, again, about an absolute 1% lower in the liberally transfused).

The TRICS-III trial (Mazer et al NEJM 2017) found no significant difference in its composite outcome in 5,243 cardiac bypass surgery patients transfused to either Hb ≥7.5 or ≥10 g/dL (numerically favoring the restricted-transfusion arm by 1%).

Conclusions

For stable postoperative patients with cardiac risk factors but no active ischemia, TOP, FOCUS and TRICS-III support continuing a restrictive approach rather than reflexively transfusing to 10 g/dL.

For a patient with acute MI, MINT and subsequent professional society guidance point toward a higher threshold, perhaps hemoglobin 9 to 10 g/dL, though the evidence remains low certainty.

For the patient in between (postoperative anemia plus severe coronary disease, tachyarrhythmias, heart failure, dynamic troponin elevation, ischemic symptoms, or tenuous hemodynamics), careful individualized decision-making is required. TOP offered something for both the restrictive (virtually identical mortality in both groups) and the liberal clinician (fewer non-MI complications as a secondary outcome).

There is no ongoing trial comparable to MINT testing transfusion in active MI. The TRICS-IV trial is further testing restrictive vs. liberal transfusion strategies in ~1,400 younger adults undergoing cardiac bypass surgeries. Results are expected in 2026.

References

Kougias P, ; TOP Trial Investigators. Liberal or Restrictive Postoperative Transfusion in Patients at High Cardiac Risk: The TOP Randomized Clinical Trial. JAMA. 2025 Dec 23;334(24):2197-2207. doi: 10.1001/jama.2025.20841. PMID: 41205227; PMCID: PMC12596742.

Carson et al. Restrictive versus Liberal Transfusion in Myocardial Infarction — A Patient-Level Meta-Analysis. NEJM Evid 2024

Carson JL, Brooks MM, Abbott JD, Chaitman B, Kelsey SF, Triulzi DJ, Srinivas V, Menegus MA, Marroquin OC, Rao SV, Noveck H, Passano E, Hardison RM, Smitherman T, Vagaonescu T, Wimmer NJ, Williams DO. Liberal versus restrictive transfusion thresholds for patients with symptomatic coronary artery disease. Am Heart J. 2013 Jun;165(6):964-971.e1. doi: 10.1016/j.ahj.2013.03.001. Epub 2013 Apr 8. PMID: 23708168; PMCID: PMC3664840.

Ducrocq G, Gonzalez-Juanatey JR, Puymirat E, Lemesle G, Cachanado M, Durand-Zaleski I, Arnaiz JA, Martínez-Sellés M, Silvain J, Ariza-Solé A, Ferrari E, Calvo G, Danchin N, Avendaño-Solá C, Frenkiel J, Rousseau A, Vicaut E, Simon T, Steg PG; REALITY Investigators. Effect of a Restrictive vs Liberal Blood Transfusion Strategy on Major Cardiovascular Events Among Patients With Acute Myocardial Infarction and Anemia: The REALITY Randomized Clinical Trial. JAMA. 2021 Feb 9;325(6):552-560. doi: 10.1001/jama.2021.0135. PMID: 33560322; PMCID: PMC7873781.

Carson et al. Restrictive or Liberal Transfusion Strategy in Myocardial Infarction and Anemia. MINT Trial. NEJM 2024

FOCUS trial (Carson et al NEJM 2011)

TRICS-III trial (Mazer et al NEJM 2017)

For patients with severe hemorrhage, transfusion with whole blood has theoretical advantages over component blood products: whole blood is a complete, balanced resuscitation fluid that can also reverse coagulopathy with native plasma proteins.

Observational studies both in military and civilian settings have suggested whole blood may be superior to component blood transfusion for hemorrhage.

In forward-deployed military units, whole blood has the advantage of not always requiring refrigeration—it can be drawn from active-duty personnel and transfused warm within hours.

Such a “just-in-time,” on-demand approach to transfusion creates its own logistical challenges, however, and so in U.S. military forward deployments today, warm blood transfusion from matched active-duty donors tends to be paired with stored low-titer group O whole blood, which requires refrigeration but still retains logistical simplicity compared to a component strategy.

The relative simplicity and theoretical physiologic advantages of a whole-blood approach to transfusion led to its expansion into the civilian trauma community in the 2010s.

Prehospital whole blood transfusion has been tested in two randomized trials to date.

SWIFT Trial

The SWIFT trial (Smith et al NEJM 2026) did not demonstrate benefit from transfusing up to two units of whole blood in hemorrhaging civilian trauma victims in the U.K. en route by air ambulance to trauma centers, as compared to transfusion with standard blood components.

Among 616 patients transfused prehospital, those who were randomized to whole blood had numerically equal or higher mortality at all time points measured, from 6 hours up to 90 days after randomization. They also had near-identical rates of massive transfusion as those receiving component blood. Death and massive transfusion were combined into a composite outcome, which did not differ between groups.

TOWAR Trial

In the TOWAR trial, 1,020 traumatically hemorrhaging casualties were transfused with either up to two units of low-titer group O whole blood or component blood products, in 44 air medical bases that initiated the intervention in a cluster-randomized design.

At 30 days, all-cause mortality was numerically higher in the whole-blood group (25.9% vs 20.5%; non-significant, confidence interval 0.87 to 1.76; P=0.24). There was no detectable difference in adverse events between the two groups.

In an observational substudy, the age of the blood (at 1 to 14 days old, vs. 15 to 21 days old) had no association with mortality (an odds ratio of 0.99).

TOWAR was a pragmatic trial and was consequently “messy” to the expected extent, with numerous crossovers in both directions; many patients were transfused before enrollment; coagulation parameters were not obtained.

Is Two Units of Whole Blood Enough? TROOP Trialists Intend to Find Out

In both the SWIFT and TOWAR trials, a maximum of two units of whole blood were transfused prehospital, raising the question of whether this was an adequate dose to produce a benefit.

Total volumes transfused were likewise low. In both SWIFT and TOWAR, 75% of patients were transfused a total of four units of blood or fewer in the first 24 hours.

Importantly, neither trial tested a whole-blood–predominant massive transfusion strategy. Most patients subsequently received conventional component-based resuscitation after hospital arrival.

Resuscitation from traumatic hemorrhage with whole blood is currently under more definitive study in the TROOP trial.

In TROOP, around 1,100 severely injured adults at 15 major U.S. trauma centers will be randomized to resuscitation with either low-titer type O whole blood or component blood products, with a primary outcome of 6-hour mortality and numerous secondary outcomes.

TROOP began enrolling patients in 2023 and is expected to be complete by mid-2027.

The Verdict

Whole blood transfusion for severe hemorrhage doesn’t clearly improve outcomes compared to component transfusion … when only two units of blood are transfused.

Whole blood provides logistical advantages in the military setting (and possibly civilian as well), but any health benefits have yet to be established in a randomized trial.

The TROOP trial will help to establish whether whole blood provides any real physiologic or survival benefits over component blood transfusion.

References

Sperry JL, Guyette FX, Cotton BA, et al. Prehospital Resuscitation with Type O Whole Blood for Trauma and Hemorrhage. New England Journal of Medicine. Published online May 18, 2026. doi:https://doi.org/10.1056/nejmoa2602167

‌Smith JE, Cardigan R, Sanderson E, et al. Prehospital Whole Blood in Traumatic Hemorrhage — a Randomized Controlled Trial. New England Journal of Medicine. Published online March 17, 2026. doi:https://doi.org/10.1056/nejmoa2516043

‌‌Guyette FX, Sperry JL. Prehospital low titer group O whole blood is feasible and safe: Results of a prospective randomized pilot trial. J Trauma Acute Care Surg. 2022 May 1;92(5):839-847. doi: 10.1097/TA.0000000000003551. Epub 2022 Jan 25. PMID: 35081595; PMCID: PMC9038638.

Hazelton JP, et al. Use of Cold-Stored Whole Blood is Associated With Improved Mortality in Hemostatic Resuscitation of Major Bleeding: A Multicenter Study. Ann Surg. 2022 Oct 1;276(4):579-588. doi: 10.1097/SLA.0000000000005603. Epub 2022 Jul 18. PMID: 35848743.

Brill JB, et al . Impact of Incorporating Whole Blood into Hemorrhagic Shock Resuscitation: Analysis of 1,377 Consecutive Trauma Patients Receiving Emergency-Release Uncrossmatched Blood Products. J Am Coll Surg. 2022 Apr 1;234(4):408-418. doi: 10.1097/XCS.0000000000000086. PMID: 35290259.

Dorken-Gallastegi A, et al . Whole Blood and Blood Component Resuscitation in Trauma: Interaction and Association With Mortality. Ann Surg. 2024 Dec 1;280(6):1014-1020. doi: 10.1097/SLA.0000000000006316. Epub 2024 May 6. PMID: 38708894; PMCID: PMC11538373.

Spinella PC, Perkins JG, Grathwohl KW, Beekley AC, Holcomb JB. Warm fresh whole blood is independently associated with improved survival for patients with combat-related traumatic injuries. J Trauma. 2009 Apr;66(4 Suppl):S69-76. doi: 10.1097/TA.0b013e31819d85fb. PMID: 19359973; PMCID: PMC3126655.

Gurney J, Staudt A, Cap A, Shackelford S, Mann-Salinas E, Le T, Nessen S, Spinella P. Improved survival in critically injured combat casualties treated with fresh whole blood by forward surgical teams in Afghanistan. Transfusion. 2020 Jun;60 Suppl 3:S180-S188. doi: 10.1111/trf.15767. Epub 2020 Jun 3. PMID: 32491216.

Nessen SC, et al . Fresh whole blood use by forward surgical teams in Afghanistan is associated with improved survival compared to component therapy without platelets. Transfusion. 2013 Jan;53 Suppl 1:107S-113S. doi: 10.1111/trf.12044. PMID: 23301962.

Torres CM, Kenzik KM, Saillant NN, Scantling DR, Sanchez SE, Brahmbhatt TS, Dechert TA, Sakran JV. Timing to First Whole Blood Transfusion and Survival Following Severe Hemorrhage in Trauma Patients. JAMA Surg. 2024 Apr 1;159(4):374-381. doi: 10.1001/jamasurg.2023.7178. Erratum in: JAMA Surg. 2024 Apr 1;159(4):470. doi: 10.1001/jamasurg.2024.0324. PMID: 38294820; PMCID: PMC10831629.

Hanna K, Bible L, Chehab M, Asmar S, Douglas M, Ditillo M, Castanon L, Tang A, Joseph B. Nationwide analysis of whole blood hemostatic resuscitation in civilian trauma. J Trauma Acute Care Surg. 2020 Aug;89(2):329-335. doi: 10.1097/TA.0000000000002753. PMID: 32744830.

Ibrahim W, Meza Monge K, Menzel J, et al. Whole-Blood vs Component Therapy in Adult Trauma: An Updated Systematic Review and Meta-Analysis. JAMA surgery. Published online November 2026:e260197. doi:https://doi.org/10.1001/jamasurg.2026.0197

Secondary outcomes include 24-h and 30-day or hospital mortality (whichever is earlier); prespecified complications; adjudicated cause of death; time to death; length of stay (ICU and hospital); and hospital-, ventilator- and ICU-free days; the incidence of major surgical procedures; time to hemostasis in those undergoing procedures with a hemostatic component; number and type of blood products used until hemostasis is achieved (and randomized products are discontinued), as well as after hemostasis has been achieved, to 24 h post-admission; discharge destination and functional status and quality of life at hospital discharge or 30 days, as measured by Glasgow Coma Scale (GCS) and EuroQol (EQ-5D) quality of life measurement.

is Trauma resuscitation with Low-Titer Group O Whole Blood Or Products: study protocol for a randomized clinical trial (the TROOP trial)

“MR. JOHNSON! IF YOU KEEP DOING THAT, WE’RE GOING TO HAVE TO TIE YOU DOWN!”

Restraining the wrists of mechanically ventilated patients with cloth wrist cuffs strapped to bedrails has become a standard practice in many, if not most ICUs in the U.S. This has been defended as necessary for patients (for their safety), nurses (overstressed by attending to fidgety patients), or the hospital (reducing liability claims).

None of these purported benefits has ever been demonstrated to be true. Until 2026, the use of physical restraints in critically ill patients had not been examined in any randomized trial.

In small observational studies, most patients who self-extubated were restrained. (This confounded data point doesn’t prove restraints don’t prevent self-extubation, because caregivers may have been more likely to restrain agitated patients they thought might self-extubate, who then did. It doesn’t tell us if or how often restraints prevented self-extubation in the many others who did not succeed at the task.)

Restraints have also been associated with higher sedative use, longer lengths of stay, and delirium or disorientation—all of which likewise have obvious potential for confounding with the need for restraints. A 2021 meta-analysis of low-quality studies suggested an association between restraint use and PTSD.

A mere six decades after mechanical ventilation and physical restraints entered wide use in developed nations’ ICUs, a randomized trial was performed to test whether the practice provided any benefit.

The R2D2 Trial

Among 405 mechanically ventilated adults in 10 French ICUs, patients were randomized to a restrictive strategy (wrist restraints avoided unless severe agitation occurred, defined as RASS ≥3) or a liberal strategy (in which wrist restraints were applied by default and reassessed daily).

In the “low-use” (unrestrained) group, wrist-straps were used on at least one day in 36% of patients (i.e., 64% of vented patients went completely restraint-free throughout their ICU stay). Most of these were from caregiver noncompliance with the protocol — only 3% of patients were recorded as requiring restraints for severe agitation.

There was no difference in the primary endpoint of “days alive without coma or delirium at 14 days”. This is a muddy endpoint to parse in this unblinded, dynamic trial, due to the complex interactions between restraints, agitation, delirium, and sedation levels.

What everyone wants to know instead is, did the unrestrained patients self-extubate more often? They did not: 18 patients self-extubated in the unrestrained group (9.2%) vs 17 in the restrained (8.5%).

Nor did patients remove devices like I.V.s or urinary catheters more often (two devices in the unrestrained group vs. one in the restrained group). They did not have more agitation, either.

The second most salient question: did the restrained patients receive more sedation (to “snow” them pre-emptively or reactively, to reduce the perceived risk for self-extubation or other self-harm)? They did not. (See eFigures 3-7 in the supplementary appendix for the closely comparable dosing of multiple sedatives.)

The unrestrained patients got up and moved ever so slightly more in the first 14 days, but at two weeks, mobilization was nearly identical between groups.

Mortality and 90-day functional, cognitive, and psychological outcomes were also similar. About 4% (absolute) more patients in the restrained group developed PTSD at 90 days, but this was not statistically significant.

Pressure ulcers were reported in 30 unrestrained patients (15.3%) and 34 restrained patients (17.0%), nonsignificant.

But Let’s Also Show Some Restraint

It would be premature to draw broad conclusions about the non-utility (or disutility) of restraints outside the scope of this trial performed exclusively in French ICUs.

Many in the U.S. will raise a skeptical eyebrow at the report that fewer than 1% (only three patients out of ~400) removed their I.V.s, nasogastric tubes, urinary catheters, or other devices. Anecdotally speaking, critically ill patients in many U.S. ICUs seem to be significantly more unruly than this.

The caregivers participating in the trial were doubtless influenced by the knowledge that their behavior was being observed and the outcomes recorded. This may have resulted in extra vigilance toward reduction of adverse events in the intervention (low-restraint) arm.

The accepted way to account for this and other potential secular (over time) changes in clinician behavior is to randomly start the intervention in each of the ICUs over several months or a year, in a cluster-randomized design. That wasn’t done here.

Should Restraints Be So Routine in the ICU?

Per the CMS regulatory code, restraints may only be “imposed to ensure the immediate physical safety of the patient, a staff member, or others,” and never “as a means of coercion, discipline, convenience, or retaliation by staff. [They] must be discontinued at the earliest possible time.”

One hopes that restraints are virtually always imposed with some motivation to ensure ventilated patients’ safety. But there exists an uncomfortably large overlap between a concern for patient safety and a desire to maximize care teams’ convenience.

The R2D2 trial raises real doubts about whether restraints are required to maintain the safety of most mechanically ventilated patients.

Hundreds of thousands of people are physically restrained during their ICU stays every year around the world, without any good evidence to support the practice. A meta-analysis (albeit of low-quality studies) suggested that restraint use may be associated with the development of PTSD.

Now that the benefits for restraints have been called into question, conducting a properly powered cluster-randomized trial (in thousands of patients) would seem to be the bare minimum we should do to ethically justify this common but often disturbing and conflicted act of care.

References

Restrictive vs Liberal Physical Restraint Strategies in Critically Ill Patients. The Journal of the American Medical Association. 2026. Sonneville R, Couffignal C, Sigaud F, et al.

Clinical Practice Guidelines for the Prevention and Management of Pain, Agitation/Sedation, Delirium, Immobility, and Sleep Disruption in Adult Patients in the ICU. Critical Care Medicine. 2018. Devlin JW, Skrobik Y, Gélinas C, et al.

Physical Restraints and Post-Traumatic Stress Disorder in Survivors of Critical Illness. A Systematic Review and Meta-Analysis. Annals of the American Thoracic Society. 2021. Franks ZM, Alcock JA, Lam T, et al.

Unplanned Endotracheal Extubations in the Intensive Care Unit: Systematic Review, Critical Appraisal, and Evidence-Based Recommendations. Anesthesia and Analgesia. 2012. da Silva PS, Fonseca MC.

Non-Pharmacological Interventions for Minimizing Physical Restraints Use in Intensive Care Units: An Umbrella Review. Frontiers in Medicine. 2022. Cui N, Yan X, Zhang Y, et al.

Revisiting Unplanned Endotracheal Extubation and Disease Severity in Intensive Care Units. PloS One. 2015. Chuang ML, Lee CY, Chen YF, Huang SF, Lin IF.

Critical care nurses’ experiences of physical restraint in intensive care units: A qualitative systematic review and meta‐synthesis. Journal of Clinical Nursing. 2023. Lao Y, Chen X, Zhang Y, et al.SR

CMS regulatory code

A 72-year-old man who resides in a nursing home is brought to the ED because of worsening lethargy and fevers. He has a history of recurrent urinary tract infections that is presumed due to prostatic hypertrophy, and was admitted with sepsis attributed to a UTI last month.

You review his urine culture data for the E. coli isolated on his admission last month:

MRSA PCR screen from a nasal swab was negative at that time.

He is tachycardic, febrile, and has borderline hypotension. He is confused compared to his baseline. WBC is 20K, and creatinine is 2.9. In the ED, he is diagnosed with sepsis and provided a liter of I.V. crystalloid, piperacillin-tazobactam 3.5 g, and vancomycin 1.5 g I.V.

A CT of the abdomen shows no hydronephrosis or calculi.

Extended-spectrum beta-lactamase (ESBL) is an enzyme that can crack open the four-membered β-lactam ring, rendering the antibiotic ineffective.

ESBLs are encoded by genes located on large transferable modules (plasmids) that are easily shared between bacterial populations, most often Escherichia coli, Klebsiella pneumoniae, or Proteus mirabilis.

Infections by ESBL-producing gram-negative bacilli are notoriously lethal in the ICU, but as they tend to afflict the most vulnerable patients with longer ICU stays, more severe illness, and higher expected mortality, and also may often be present as colonizers alongside pathogenic bugs, their independent contribution to mortality is difficult to tease out precisely.

Generally speaking, urinary infections with ESBL-producing organisms are less often fatal than respiratory infections or primary bacteremias. Klebsiella is usually more virulent than E. coli. Septic shock with ESBL organisms has a higher mortality rate than with sensitive bacteria.

Microbiology labs may test for and report ESBL production, but it should be assumed anytime a gram-negative rod on the “usual suspects” list (E. coli, Klebsiella pneumoniae, Klebsiella oxytoca, and Proteus mirabilis) shows resistance to ceftriaxone (and often ceftazidime and aztreonam) but susceptibility to carbapenems.

Ceftriaxone resistance in Enterobacter cloacae, Klebsiella aerogenes, Citrobacter freundii, Serratia, etc., often suggests AmpC, not necessarily ESBL.

Narrower resistance (e.g., to ampicillin and/or cefazolin) does not imply ESBL, as many E. coli isolates are resistant to these agents.

Failure to adequately cover ESBL early is considered a likely major contributor to the excess mortality. As such, early appropriate antibiotic coverage for patients at risk is vital.

Meropenem is recommended for critically ill patients at high risk for ESBL organisms. Ertapenem is also effective, but does not cover Pseudomonas, which is often a concern in severe infections. Meropenem’s dose is reduced for renal impairment.

Although beta-lactamase inhibitors like tazobactam do thwart the ESBL enzyme in many patients, tazobactam has an unpredictable failure rate in ESBL-producing Enterobacteriaceae (aka Enterobacterales) in serious infections, especially bacteremia, sepsis, pneumonia, or ICU-level illness.

Gentamicin is considered an effective and appropriate agent against ESBL, but due to its risks for nephrotoxicity and ototoxicity, is considered second-line to meropenem for severe ICU-level infections.

Multiple newer β-lactam + β-lactamase inhibitor combination agents have entered the U.S. market in recent years. Ceftazidime-avibactam, meropenem-vaborbactam, imipenem-cilastatin-relebactam, ceftolozane-tazobactam, and cefiderocol all exhibit activity against ESBL organisms. These are generally reserved for carbapenem-resistant organisms to preserve their utility.

Oral nitrofurantoin or TMP-SMX may be a good option for uncomplicated ESBL UTIs, but are not recommended for severe infections. Fosfomycin is not recommended for K. pneumoniae or other organisms that may carry fosA genes, after a randomized trial found single-dose fosfomycin had more clinical failures than a 5-day course of nitrofurantoin.

Reflect to earn CME with Learner+

Sample reflection: I reviewed and reflected on infections due to known or suspected extended-spectrum beta-lactamase-producing gram-negative bacilli in the ICU, and the considerations for appropriate antimicrobial therapy.

References

Infectious Diseases Society of America 2023 Guidance on the Treatment of Antimicrobial Resistant Gram-Negative Infections. Clinical Infectious Diseases : An Official Publication of the Infectious Diseases Society of America. 2023. Tamma PD, Aitken SL, Bonomo RA, et al.

Infectious Diseases Society of America 2024 Guidance on the Treatment of Antimicrobial-Resistant Gram-Negative Infections. Clinical Infectious Diseases : An Official Publication of the Infectious Diseases Society of America. 2024. Tamma PD, Heil EL, Justo JA, et al.

Association Between Minimum Inhibitory Concentration, Beta-Lactamase Genes and Mortality for Patients Treated With Piperacillin/­Tazobactam or Meropenem From the MERINO Study. Clinical Infectious Diseases : An Official Publication of the Infectious Diseases Society of America. 2021. Henderson A, Paterson DL, Chatfield MD, et al.

Ceftolozane-Tazobactam Versus Meropenem for Definitive Treatment of Bloodstream Infection Due to Extended-Spectrum Beta-Lactamase (ESBL) and AmpC-producing Enterobacterales (”MERINO-3”): Study Protocol for a Multicentre, Open-Label Randomised Non-Inferiority Trial. Trials. 2021. Stewart AG, Harris PNA, Chatfield MD, Littleford R, Paterson DL.

Multidrug-Resistant Gram-Negative Bacterial Infections. Lancet. 2025. Macesic N, Uhlemann AC, Peleg AY.

Extended-Spectrum Β-Lactamase-Producing Enterobacteriaceae: Update on Molecular Epidemiology and Treatment Options. Drugs. 2019. Peirano G, Pitout JDD.

Carbapenem Use in Extended-Spectrum Cephalosporin-Resistant Enterobacterales Infections in US Hospitals and Influence of IDSA Guidance: A Retrospective Cohort Study. The Lancet. Infectious Diseases. 2024. Walker MK, Diao G, Warner S, et al.

Preserving the Antimicrobial Arsenal: Exploring Alternatives to Carbapenems in ESBL Battles Within the Southeast of Ireland. Journal of Medical Microbiology. 2025. Ali S, Tobin A, Lapthorne S, et al.

Mechanical thrombectomy (aka endovascular thrombectomy or EVT) within 24 hours can improve outcomes after ischemic occlusion of a large cerebral vessel in the anterior circulation. Thrombectomy has become standard care for ischemic strokes with LVO, with ongoing refinement over the optimal time, methods for patient selection, and which large posterior-circulation occlusions also benefit.

Through rapid triage and transport networks, endovascular thrombectomy has been effectively deployed across swaths of developed countries over a relatively short period (since 2015).

Investigators then shifted focus to thrombectomy for medium-sized cerebral vessels (so-called medium-vessel occlusions or MeVO).

MeVOs involve the M2 and M3 segments of the middle cerebral artery, the A2 and distal segments of the anterior cerebral artery, and the P2 and distal segments of the posterior cerebral artery. These account for a large proportion of ischemic strokes (>25%).

Three high-quality randomized trials tested EVT against medical therapy in patients with medium-vessel or distal occlusion strokes:

ESCAPE-MeVO (Canada, n=530); no improvement in functional outcome and significantly higher mortality (13.3% vs. 8.4%, adjusted HR 1.82) and more symptomatic intracranial hemorrhages (5.4% vs. 2.2%). (Goyal et al., NEJM 2025)

DISTAL (Europe, n=543), no reduction in disability or mortality with EVT. Also showed numerically higher mortality (15.5% vs. 14.0%) and double the rate of symptomatic intracranial hemorrhage (5.9% vs. 2.6%). (Psychogios et al., NEJM 2025)

DISCOUNT, a smaller French trial of 161 patients, was halted early for futility and safety in ~2025, with worse functional outcomes and higher rates of intracranial hemorrhage in the thrombectomy arm.

These three trials enrolled patients with mildly disabling strokes (e.g., median NIHSS of 6), so one interpretation was that the patients did not have large enough occlusions to benefit from mechanical intervention (and to overcome its potential harms).

A trial conducted in China then reported strikingly positive benefits from EVT for medium-vessel ischemic occlusions.

In ORIENTAL-MeVO, 563 patients at 48 centers across China were randomized to EVT plus medical therapy or medical therapy alone.

Enrollment was limited to patients with moderate to severe neurologic deficits at presentation (NIHSS ≥6).

Patients receiving thrombectomy had good neurologic outcomes (mRS 0-2) much more often (~59% vs ~47%) than those receiving medical therapy alone. An absolute 15% more patients had excellent outcomes. Number needed to treat was 8.

Symptomatic intracranial hemorrhage was doubled (4.7% vs 2.2%), however.

It’s possible that the improvements with thrombectomy were due to the larger strokes and salvageable ischemic brain in this trial population.

MeVO Thrombectomy in the USA: The DUSK Trial

Due to major differences between populations, the delivery of health care and possible other unmeasurable factors, randomized trial data from China should usually not be extrapolated directly to Western systems of care.

Guidelines do not advise (nor do they explicitly advise against) endovascular thrombectomy for ischemic stroke due to occlusions in medium-sized vessels.

The DUSK trial is currently enrolling patients in the United States, centered at the U. of Pittsburgh, testing thrombectomy specifically in a Western MeVO population among an intended n=584 patients.

More severe strokes will likely be included (NIHSS ≥6), but patients with NIHSS <6 are allowed in if they have aphasia, neglect, or other less-disabling deficits, if considered significant by the treating team.

If chance brings enough patients with genuinely disabling strokes to the participating centers, DUSK could provide a much clearer answer in either supporting the previous three negative trials, or by demonstrating a clear benefit from mechanical thrombectomy for ischemic stroke due to MeVO.

Given the moderately strong signal of harm among patients undergoing MeVO for less disabling strokes, the former seems more likely.

Results from DUSK may arrive sometime in 2028.

References

ESCAPE-MeVO

DISTAL

Summary of DISCOUNT data

DUSK Trial protocol

A new sepsis alert is hitting the market after receiving 510(k) clearance by the U.S. FDA. It’s more accurate than many others in internal testing, and big-time customers are signing up. Could this be “the one” — the algorithm that will save patients’ lives without compromising clinicians’ sanity?

Meet TREWS

Johns Hopkins researchers, led by computer scientists Suchi Saria and Katharine Henry, developed the Targeted Real-Time Early Warning System (TREWS) more than a decade ago. The homegrown algorithm was integrated inside Hopkins’ instance of Epic in an ongoing, pragmatic testing/deployment rollout. Like other modern sepsis flaggers, it lurks in modular code within the EMR, where its machine learning algorithm integrates streams of routinely collected lab data, clinician notes, context of care, medication history, and more, firing when a mysterious threshold of statistical associations within its training data has been exceeded.

TREWS’ Performance On Paper …

In a 2015 paper, they showed TREWS had relatively high discriminative power to help identify or rule out sepsis. Its area under the receiver operating characteristic curve was 0.83, which for a sepsis tool is very good.

For example, when tuned to a specificity of 0.67 (a one-third false positive rate), it correctly identified 85% of patients who went on to develop septic shock, usually more than 28 hours before hypotension developed. In a follow-up paper describing a wider rollout, its performance was even better: a reported sensitivity of 82% and an even higher specificity (>90%) and AUROC (0.97).

In Hopkins’ data, TREWS’ performance crushed that of Epic Systems (as evaluated by U. of Michigan researchers in 2018), which was only moderately better than chance (AUROC 0.63), and which “despite generating alerts on 18% of all patients, did not detect sepsis in 67% of patients with sepsis,” according to the U of M paper. (Epic argued that in broader third-party use, the AUROC was higher at ~0.76).

Does this mean TREWS could solve the problems of alarm fatigue, operational drain, and antibiotic overuse that sepsis alerts create?

Here’s what TREWS’ superior performance looked like from the clinician's perspective at Hopkins, using the waterfall from their Nature Medicine 2022 paper.

… And In the Hospital

During a two-year deployment at five hospitals, the TREWS system screened ~469,000 encounters, and fired ~32,000 times: ~7% of all patient encounters fired alerts.

Among the 469,000 patients, ~9,800 were retrospectively adjudicated to have had sepsis (through an automated method described here).

This corresponded to a 2% (adjudicated, inferred) prevalence of sepsis in the overall cohort (n~469K). For the purposes of the study and calculation of test performance characteristics, these 9,800 were considered “positives” (but read the gray box).

TREWS flagged ~8,000 (82%) of the ~9,800 cases that the case-identifying algorithm would later call sepsis: an 82% sensitivity and ~95% specificity.

But even at that impressive performance, the test fired about three times in non-septic patients for every patient correctly flagged as sepsis.

That was because of the low 2% prevalence of sepsis in the cohort (positive predictive value, which depends heavily on disease prevalence, was ~0.25). With an increasing prevalence of sepsis, the alert would fire correctly more often.

Validating in Circles

One reason why TREWS performs so well: The same team of computer science and clinical faculty created both TREWS (the algorithm to identify sepsis in real time by EMR surveillance) and the automated algorithm that identified cases as “sepsis,” also based on EMR-extracted data.

TREWS’ machine learning was very likely trained to optimize for the output of “sepsis” as generated by the companion sepsis case-identifying algorithm.

(This is largely unavoidable today in a condition with no gold standard diagnostic method, and all sepsis flaggers are probably designed similarly and subject to similar circularity.)

In other words, TREWS is very good at predicting the outputs of another algorithm it had been trained on (with both designed by the same group).

TREWS is in use by multiple health systems, but has not yet been independently or publicly critiqued in the way that Epic’s model was.

Ready, Set, Monetize

Sometime before 2021, Johns Hopkins and the two lead investigators (Henry and Saria) spun up a for-profit corporation, Bayesian Health, to market AI-based alerting software to health systems.

It emerged from stealth to reportedly receive an initial funding round of $15 million by AndreesenHorowitz in 2021 and has had multiple funding rounds since then, according to Crunchbase.

Its website lists numerous major health systems as customers (e.g., Cleveland Clinic), some of whose executives’ testimonials report many fewer false sepsis alerts than before using TREWS.

The “TREWS Saves Lives” Claim

One of Bayesian Health’s claims is that TREWS reduces sepsis-associated mortality.

The source for this seems to be a separate Nature Medicine 2022 paper, which found that patients whose caregivers responded to their alerts within three hours had lower mortality than those whose caregivers ignored their alerts for at least three hours.

Obviously, the “ignoring alerts” behavior might be associated with many other potential factors and confounders of a worse outcome, independent of the alert itself.

TREWS might indeed save lives. But Hopkins and the investigators knew how to actually establish that: through a cluster-randomized trial at their (and their collaborators’) many hospitals. That wasn’t done.

A Small Step Forward

Sepsis is sneaky and often goes unrecognized until it’s too late. Sepsis alerts promise to help with this problem, and deployed at scale might even save thousands of lives annually. This comes at the price of their high false-alarm rates, producing alarm fatigue, clinician frustration, operational drain, and excess antibiotic use.

With a ~20% false negative rate (~80% sensitivity) and 75% false-alarm rate (~0.25 PPV), TREWS does not solve these problems by any stretch, and it has not been evaluated independently and critically in the way that Epic’s model was. It can’t replace clinical judgment. TREWS has not and will likely never be tested in a randomized trial. Its creators and home institution have major (disclosed) financial interests in its success.

Ironically, that last piece—the U.S.’s capitalistic approach to health care—may in the end provide the best evidence we will get for TREWS’ possible advantages over its competition. Investors and health system executives seem to believe in the technology, and the latter are starting to pay for it with their hard earned revenues—whoops, I meant with your insurance premiums and Medicare taxes.

“It’s not perfect, but it’s a lot better than Epic” isn’t Bayesian Health’s slogan, but they are welcome to use it (no charge).

References

Adams R, Henry KE, Sridharan A, Soleimani H, Zhan A, Rawat N, Johnson L, Hager DN, Cosgrove SE, Markowski A, Klein EY, Chen ES, Saheed MO, Henley M, Miranda S, Houston K, Linton RC, Ahluwalia AR, Wu AW, Saria S. Prospective, multi-site study of patient outcomes after implementation of the TREWS machine learning-based early warning system for sepsis. Nat Med. 2022 Jul;28(7):1455-1460. doi: 10.1038/s41591-022-01894-0. Epub 2022 Jul 21. PMID: 35864252.

Henry KE, Hager DN, Pronovost PJ, Saria S. A targeted real-time early warning score (TREWScore) for septic shock. Sci Transl Med. 2015 Aug 5;7(299):299ra122. doi: 10.1126/scitranslmed.aab3719. PMID: 26246167.

Wong A, Otles E, Donnelly JP, et al. External Validation of a Widely Implemented Proprietary Sepsis Prediction Model in Hospitalized Patients. JAMA Internal Medicine. 2021;181(8). doi:https://doi.org/10.1001/jamainternmed.2021.2626

‌Henry KE, Adams R, Parent C, Soleimani H, Sridharan A, Johnson L, Hager DN, Cosgrove SE, Markowski A, Klein EY, Chen ES, Saheed MO, Henley M, Miranda S, Houston K, Linton RC 2nd, Ahluwalia AR, Wu AW, Saria S. Factors driving provider adoption of the TREWS machine learning-based early warning system and its effects on sepsis treatment timing. Nat Med. 2022 Jul;28(7):1447-1454. doi: 10.1038/s41591-022-01895-z. Epub 2022 Jul 21. PMID: 35864251.

A 48-year-old man presents to the ED with worsening dyspnea, fever, and an increase in his chronic leg edema. He receives diagnoses of sepsis and decompensated heart failure, with suspected pneumonia based on nonspecific infiltrates on chest radiograph. His initial blood pressure is 100/70 with a normal heart rate. His SpO2 is 94% on 5 liters of nasal oxygen. Lactic acid is normal. Creatinine is 2.1 (baseline 1.5 mg/dL). Furosemide and ceftriaxone are provided.

Over the subsequent hours, his blood pressure falls to 85/45 (MAP 58) with pulse 110/min. His SpO2 is now 90% on high-flow nasal oxygen at FiO2 0.50, and respirations are 25 with increased work of breathing. ABG shows pH 7.35 / pCO2 33 / paO2 60.

A bedside ultrasound shows a non-collapsing inferior vena cava during respirations. A passive leg raise is performed; two minutes later, his heart rate is 100/min (a 9% decrease). A jugular venous waveform is not well seen; you note a BMI of 40 and a thick neck.

Should patients with a diagnosis of septic shock (perhaps better termed sepsis-induced hypotension) first receive fluids, vasopressors, or both simultaneously? Few questions in critical care inspire as diverse and strongly held opinions as this.

Read more

More than probably any other group of people on earth, intensivists are aware that as people get older, things fall apart. We see firsthand that on average, 85-year-olds are at greater risk for bad outcomes from critical illness than 45-year-olds. After seeing this pattern play out hundreds or thousands of times over years, it's understandable that clinicians unconsciously form heuristics.

But we also see that senescence happens to people at highly variable rates and patterns, mostly based on their behaviors and baseline health (which are closely linked to socioeconomic status).

This dispersion produces “young” octogenarians and “old” fortysomethings. A marathon-running, never-smoking 85-year-old may have a better chance of surviving critical illness or emergency surgery than a 45-year-old with cirrhosis and ESRD who drinks and smokes heavily daily.

Age-based heuristics tend to align with clinical reality, and the pattern-recognizing clinician-brain may be unable to fully resist the pressure to make a suite of assumptions about an older patient, especially one who cannot communicate.

However, robust elderly patients are no longer a rarity, becoming far more common in more affluent communities especially. Any ageist bias lingering from training or experience poses a risk to the healthiest elderly.

But also, let’s be honest, how many 85-year-olds are running marathons?

Chronological Age and Frailty (“Biological Age”)

ICU mortality increases with increasing age—almost tautologically—because older people usually have more medical problems and frailty. Mortality after ICU discharge is also associated with age, which is also tautological: older people have a shorter life expectancy.

Reviews that attempt to control for comorbidities and frailty, however, have found an inconsistent association between age and mortality in the ICU. That seems to be because enough healthier older people do better, thanks to their younger “biological age,” to weaken the association.

Frailty also correlates well with ICU mortality, about as strongly as chronological age, but also not well enough to use as an accurate predictor of outcome.

Experienced clinicians incorporate considerations of age, chronic disease burden, acute disease severity, and frailty or disability in forming a gestalt prognostic assessment of each individual patient.

And to date, there has been no model shown to perform better in predicting outcomes in individual patients after ICU admission.

But even well-tuned clinician gestalt is very frequently inaccurate, especially among trainees, nurses, and younger physicians.

All that is to say that any heuristic that is predominantly based on age (or any other single variable) will too often be wrong, and could easily result in under- or overtreatment.

We’ve all gotten very used to regular overtreatment in the ICU. But those ethical challenges aside, undertreatment—not providing lifesaving testing and interventions to a patient incorrectly believed to be moribund based on age—should worry us more.

A Geriatrician in Every ICU?

A major critical care society released their “Guidelines on Caring for Older Adults in the ICU.”

Wisely, and cautiously, they avoided any direct mention of the possibility of overtreatment (harm) or undertreatment (rationing, ageism) of the elderly in ICUs.

Instead, the panel recommended a “geriatric model of care” for all older adults admitted to the ICU. By this they meant a geriatric consult—for every patient aged 65 or older in the ICU.

They acknowledged that the evidence they cite does not support this suggestion. But beyond that, it’s mathematically impossible.

There are about 6,400 practicing geriatricians in the U.S. They are all already working in other settings—clinics and hospital wards, mostly.

Adults aged 65 and older account for about half of the 5 million ICU admissions annually, or ~2.5 million. That’s about 350 ICU consults per geriatrician per year, about 30 per doc per month. Seeing each patient twice more in the unit would be 90 encounters per geriatrician per month.

Implementing this guideline would immediately require a quarter to half the geriatrician workforce working full-time in ICUs (or all of them working 25% in the unit), or training 3,000 more geriatricians. That would cost about $1 billion per year.

And since there’s no geriatrician coming, guess who is supposed to be the geriatrician in this “model of care”?

ICU Care is Already Geriatric Care

The panel knew all this. And since 65-and-older patients already comprise half the ICU, and probably 75% of its sickest patients, encouraging us to consider them as a separate group is an odd suggestion.

So what is the panel getting at?

They explicitly declined to say:

The panel was intentionally ambiguous regarding the specific geriatric models of care to be implemented as part of this recommendation and acknowledged interventions may vary depending on hospital, resources, and expertise available. At a minimum, the included studies intentionally incorporated geriatric principles into the care of older adults, such as removal of unnecessary tethering devices (urine catheters and restraints), addressing hearing impairment, and a focus on functional and cognitive outcomes through occupational therapy.

But we should be removing restraints and urinary catheters as soon as possible in all patients, regardless of age.

The only other recommendation they made (other than the impossible-geriatric-consult) was not to give antipsychotics to prevent delirium in the elderly.

Antipsychotics already have a black-box warning against doing this, although it certainly doesn’t hurt to re-emphasize it.

It feels like there was a sense somewhere in the organization that there should be a guideline for older adults, but then it turned out there was nothing much to say in it.

And that’s fine.

So, Just Carry On, Then

Older adults comprise half of ICU patients, and a majority of the sickest, frailest, or both. A minority, meanwhile, were vigorous and robust until the onset of their acute illness.

The clinician’s job is to learn who is which, and treat each according to her individual situation and needs.

A 75-year-old marathon runner may get an open aortic aneurysm repair and CRRT.

The 75-year-old cachectic, contracted, and catatonic nursing home resident with end-stage dementia may get most of his meds discontinued and a palliative care consult.

As long as each patient’s plan is devised thoughtfully and compassionately with consideration of their personhood, that’s not ageism. It’s good medicine.

Reference

Ferrante LE, Chaudhuri D, Laiya Carayannopoulos K, Jain S, Tate JA, Álvarez-Espinoza E, Austin CA, Burry L, Devinney MJ, Ehlenbach WJ, Happ MB, Hope AA, Hua M, Kho ME, Palakshappa JA, Scheunemann LP, Sinvani L, Stahl B, Wang S, Wunsch H, Rochwerg B, Brummel NE. Society of Critical Care Medicine Guidelines on Caring for Older Adults in the ICU. Crit Care Med. 2026 May 1;54(5):1060-1072. doi: 10.1097/CCM.0000000000007085. Epub 2026 Mar 20. PMID: 41860322.

Last week's analysis of the randomized and retrospective evidence base for hypothermia after cardiac arrest generated several dissenting emails and comments.

Including this especially thoughtful one:

Reading this comment gave me an acute pang of the anxiety I periodically get from publishing this newsletter. I recalled reading meta-analyses, but I did not include any in the post. What did they say? Did I leave out something vital from this discussion? Am I going to need to apologize again?

ICYMI, a quick recap: Hypothermia as a treatment after out-of-hospital shockable cardiac arrest showed massive benefits in preserving brain in two randomized trials published in NEJM in February 2002. One (HACA, n=275) reported a 23% absolute improvement in good neurologic outcome and a 14% absolute reduction in mortality; the other (Bernard et al, n=77) found a 16% absolute improvement in good neurologic outcomes.

In critical care, these are huge, almost unheard-of magnitudes of benefit.

Unfortunately, these findings were not replicated in 10 subsequent larger randomized trials. (HYPERION, finding a neurologic benefit from hypothermia for nonshockable rhythms, was the outlier.)

The last ongoing hypothermia trial in the U.S., ICECAP, quietly shut down for statistical futility in 2025 after enrollment of about 1,150 patients (short of the intended enrollment of 1,800). Futility was determined in both the shockable and nonshockable rhythm arms of the trial.

Altogether, hypothermia was tested in randomized trials enrolling over 9,140 patients.

And after 2002 (other than HYPERION), it didn’t work.

But what if it did?

What if it does?

Bring in the Meta-Analysts

Read more

A 51-year-old woman is brought to the ED after a witnessed seizure at home after she complained of headaches and then became confused. She had another generalized seizure in the ambulance and a third in the ED. She is intubated for status epilepticus and loaded with levetiracetam. Neurology are consulted.

According to her family, she was diagnosed with rheumatoid arthritis this year, for which she now takes a tumor necrosis factor-alpha antagonist, with good disease control. She also takes metoprolol for hypertension. She has never had a seizure before today.

Her blood pressure was initially 200/110 and is now 165/85 with a nicardipine infusion. Other vital signs and labs are unremarkable. She is sedated on propofol and poorly responsive.

A CT of her head showed no acute process. An EEG is ordered. High-dose ceftriaxone, acyclovir, and vancomycin are infused IV.

A stat brain MRI is obtained:

Read more

Is it too easy to say it was all a fever dream?

There was good reason to be optimistic in 2002 that induced hypothermia would improve survival and neurologic outcome after cardiac arrest. Animal studies aligned with anecdotal evidence from victims of drowning in frigid waters: colder temperatures preserved brain. Two “landmark” trials on out-of-hospital cardiac arrest seemed to confirm the preclinical data. Practice changed quickly; guidelines enshrined the practice of cooling all cardiac arrest patients.

Twenty-four years and 11 large randomized trials later (all negative except the equivocal HYPERION), guideline authors walked back from “cool them all” to “prevent all fevers” by actively maintaining normothermia (≤37.5°C) with external cooling systems like Arctic Sun™.

Those guidelines, however, still tacitly encouraged hypothermia by specifying a range of 32° to 37.5°C as appropriate temperature targets.

The awkward truth was that maintenance of normothermia was not itself known to be beneficial, because it had never been tested. After the original trials favoring hypothermia (likely spuriously, in hindsight), allowing fevers was presumed to be unethical.

The supposed benefits of fever prevention were inferential, based on highly confounded observational studies correlating fevers with worse outcomes in strokes. The obvious problem: larger strokes cause both fevers and bad outcomes.

Randomized trials testing fever prevention in strokes (most notably the INTREPID trial, Greer et al JAMA 2024), showed no benefit of the practice.

It might be too embarrassing and heterodox for academics to run a large randomized trial that could establish that doing little or nothing temperature-wise is the best approach.

Thus, we will likely only have the findings of observational studies to consider.

Temperature Management for In-Hospital Cardiac Arrest

The most recent data for IHCA comes from the “Discover In-Hospital Cardiac Arrest” cohort (Andrea et al Crit Care Med 2026).

In 2023-2024, a major U.S. critical society guided 46 participating hospitals in multiple nations in a prospective observational cohort study assessing the use of targeted temperature management and outcomes among 1,006 patients achieving ROSC after in-hospital cardiac arrest, among whom 615 remained alive and comatose and eligible for TTM.

Adherence with TTM was generally low, and had no observed association with mortality or improved neurologic status at discharge.

• In more than half, the treating teams did not document any strategy to manage temperature.

• 30% of patients had targeted normothermia (≤37.5°C) or fever avoidance as their strategy.

• In 14% (85 patients), hypothermia was targeted, usually 36°C.

There was only one consistent outcome resulting from fever prevention: fewer fevers.

“Patients with documented strategies received more temperature management therapies and had a lower burden of fevers, but there was no association with outcomes.”

But Many Single-Center Observational Studies Suggest Benefits of TTM …

Yet it’s easy to find multiple retrospective observational series that found an association between hypothermic temperature management (e.g. 33°C) and improved outcomes after out-of-hospital and in-hospital cardiac arrest.

One of the most prominent and recent was a single-center retrospective study at U. of Pittsburgh (Callaway et al JAMA Ntwk Open 2020).

Among 1,319 patients admitted to one center after out-of-hospital cardiac arrest, cooling to 33°C was associated with better survival and neurologic outcomes compared to TTM at 36°C, in more severely ill patients (shock and/or deeper comas).

However, in the less severely ill, 36°C was associated with improved outcomes.

The magnitudes of effects were large in both subgroups—in opposite directions: 14% “harm” of 33°C in the less-ill vs 22% “benefit” of 33°C in the sicker patients. Almost two-thirds of patients died after withdrawal of life support.

In the context of negligible or no benefit of TTM in multiple randomized trials, this data pattern strongly suggests clinician bias in selecting which patients were targeted at which temperatures. Authors noted that (based on their interviews with clinicians),

Treating physicians were more likely to select TTM at 36°C for patients they believed had nonsurvivable illness or with antecedent goals of care that limited critical care support … Assignment of a few more moribund patients to TTM at 36°C would bias our results to favor TTM at 33 °C.

The higher proportion of deaths in TTM at 36°C vs TTM at 33°C attributed to multiple organ failure (39.5% vs 27.3%) or WLST for nonneurologic reasons (25.3% vs 11.0%) suggests that this bias is present.

The transparency of this discussion is laudable. The Pitt data highlights the limitations of retrospective observational data from any single institution when clinician discretion has an overarching influence on the final outcomes.

… While Larger Registries Tell a Different Story

The most robust and compelling data regarding temperature management after in-hospital cardiac arrest come from a large retrospective observational cohort derived from the AHA’s Get With the Guidelines registry.

The outcomes for over 26,000 patients at 355 U.S. hospitals who achieved ROSC after an in-hospital cardiac arrest between 2002 and 2014, who were either treated hypothermia or not were analyzed.

One interesting finding is just how few hospitals were cooling IHCA patients during this time: only 1,568 (6%) were treated with hypothermia, although most were comatose.

Patients treated with hypothermia were more likely to die in the hospital or to have a poor neurologic outcome (about a 2% absolute difference for either), compared to propensity-matched patients not treated with hypothermia.

Bias was certainly also present in these decisions, but the direction and magnitude of the finding observed across many more systems is more informative than any single-center cohort.

Why Prevent Fevers After Cardiac Arrest?

Guidelines still advise temperature management to prevent the supposedly harmful fevers after cardiac arrest.

With the benefit of hindsight, this should have been the first intervention to test, not the last—i.e., active maintenance of normothermia vs. usual, reactive care only.

After the massive efforts poured in, the guidelines written and unwritten, a total lack of benefit here might be too much to accept, though. “Temperature management” still provides us with something to do in a usually terrible situation that might be helpful.

Accordingly, no randomized trials are underway with a control arm allowing for a “pre-2002 usual care” response to fevers (e.g., acetaminophen only).

That being said, meta-analyses indirectly suggest a benefit from maintenance of normothermia, in the sense that later trials whose control arm included normothermia were more firmly negative, while earlier trials that permitted more fevers were the ones in which a potential benefit of hypothermia was observed.

The hope for a benefit from hypothermia is still alive, albeit on life support. The SAVE-J NEUROTHERM and the R-CAST OHCA studies are ongoing in Japan. The Europeans have PRINCESS2 approaching coronation (after the first PRINCESS was deposed). And in the U.S., although ICECAP quietly shut down in 2025 for futility, its pediatric version continues.

Hope is always good. Just don’t call hypothermia “therapeutic” anymore, please.

References

Andrea L, Berg KM, Johnson NJ, et al. Temperature Control After In-Hospital Cardiac Arrest: Outcomes From the Discover In-Hospital Cardiac Arrest Cohort. Critical Care Medicine. Published online March 20, 2026. doi:https://doi.org/10.1097/ccm.0000000000007121

‌Chan PS, Berg RA, Tang Y, Curtis LH, Spertus JA; American Heart Association’s Get With the Guidelines–Resuscitation Investigators. Association Between Therapeutic Hypothermia and Survival After In-Hospital Cardiac Arrest. JAMA. 2016 Oct 4;316(13):1375-1382. doi: 10.1001/jama.2016.14380. PMID: 27701659; PMCID: PMC5486217.

Callaway CW, Coppler PJ, Faro J, Puyana JS, Solanki P, Dezfulian C, Doshi AA, Elmer J, Frisch A, Guyette FX, Okubo M, Rittenberger JC, Weissman A. Association of Initial Illness Severity and Outcomes After Cardiac Arrest With Targeted Temperature Management at 36 °C or 33 °C. JAMA Netw Open. 2020 Jul 1;3(7):e208215. doi: 10.1001/jamanetworkopen.2020.8215. PMID: 32701158; PMCID: PMC7378753.

A 64-year-old woman re-presents to the ED with persistent GI symptoms. She was seen four days prior with nausea, non-bloody diarrhea, and abdominal pain. A CT of her abdomen showed mild small-bowel wall thickening. Vitals were normal. After receiving IV fluids, she felt somewhat better and was discharged home from the ED.

She returns with worsening abdominal pain that is now severe, and diarrhea that is now bloody. She is febrile, with a pulse 110/min in new-onset atrial fibrillation, blood pressure 85/55, and respirations of 22/min.

She reports regularly taking aspirin-containing analgesics for the pain. Salicylate level is 35 mg/dL (upper limit of therapeutic level = 30 mg/dL).

On exam, her abdomen is soft with diffuse mild tenderness.

Labs show:

Nephrology are consulted. She receives 30 cc/kg of crystalloid resuscitation fluids.

Read more

Microaxial flow pumps (Impella™ being the only available product) were shown in the Danger-Shock trial (NEJM 2024) to reduce mortality in patients with cardiogenic shock due to ST-elevation myocardial infarction. The survival benefit came with the cost of significant risks for limb ischemia, bleeding, and acute kidney injury.

Almost as soon as Danger-Shock was published, pressure grew for the expansion of Impella beyond its evidence-based use case for STEMI-related cardiogenic shock.

Cardiologists commonly face gray-zone situations in which patients deteriorate during or after PCI, without frank shock. Impella was FDA-approved for mechanical circulatory support during procedures way back in 2012. After Danger-Shock, interventional cardiologists used Impellas much more frequently during high-risk procedures (and maybe some not-so-high-risk ones).

Reimbursements in the tens of thousands of dollars per Impella (to the hospital) incentivized the expansion. Cardiologists don’t receive big per-device payments for placing Impellas, but they gain massive leverage in salary negotiations with their centers based on the revenue they bring in.

Abiomed, acquired by J&J’s Medtech division in 2024 for $16B, surfed this wave well. Under the slogan “Impella Protected PCI,” the device was marketed as a safety hedge to cardiologists traumatized by patients dying on their cath lab tables, and sales grew dramatically. The PROTECT I, II, and III trials were cited to support the practice (but did so weakly, if at all).

However, to really secure the expansion into this market, i.e., patient population, higher-quality randomized trial data would be needed to establish the benefit of Impella in MI without shock.

In a refreshing (or depressing, or both) abandonment of any pretense, J&J MedTech’s own chief medical science officer served as a lead investigator of a randomized trial (‘Door to Unload’) that, if successful, could expand Impella’s FDA indication (or just clinician willingness) to MIs without shock.

Contemporaneously, the U.K. funded its own smaller trial (‘Chip’) testing Impellas for MI without shock—presumably as a due diligence counterweight against the massive expenditure that would be required of the NHS if pressure rose to include Impellas as standard care in this much larger patient population.

The ‘Door to Unload’ and ‘Chip’ Trials

In the “Door to Unload” trial, 527 patients with acute anterior STEMI were randomized to Impellas for 30 minutes before PCI vs. immediate PCI alone.

(Yes, you read that correctly: in patients with STEMI, PCI was deferred to place an Impella and run it for 30 minutes before revascularizing. Patients remained ischemic for about 47 minutes on average.)

In the Chip trial, Impellas were placed immediately before PCI in the intervention arm. Three hundred patients with STEMI and severe LV dysfunction (<35%) were randomized.

Neither trial showed a benefit in the Impella arm over usual care:

• In Chip, there was a concerning trend toward harm (hazard ratio for death 1.54 in the Impella arm).

• In DTU, Impellas did not improve infarct size (the primary outcome) or secondary endpoints (win ratio 1.04). All-cause mortality and cardiovascular mortality at 30 days did trend toward a benefit from Impellas (P=0.08).

A Takeaway

The Chip trial argues strongly against routine use of Impellas for patients with STEMI without shock.

Door-to-Unload’s trend toward a mortality benefit left the door cracked open for use cases for Impella in this population, which aligns with interventionalists’ strong anecdotal experience of Impella’s benefits in selected patients.

Identifying who will benefit, though, remains as difficult as ever.

An editorialist noted:

[FDA] approval and the commercial framing of these devices under the “Protected PCI” designation has reinforced a perception of procedural safety for the operator…[but] the criteria for the selection of appropriate patients remain poorly defined despite years of use.

These [trial] findings force the interventional cardiologist to occupy an uncomfortable position between two unsatisfying options — the potential burden of perceived undertreatment when a patient’s condition deteriorates or recognition that the act of doing more may itself constitute harm. Any honest appraisal must begin with that tension.

This dilemma pervades medicine, but lands hardest on proceduralists. Even if financial considerations were eliminated, this tension would remain.

References

Perera D, Ryan M, Ezad SM, et al. Left Ventricular Unloading in High-Risk Percutaneous Coronary Intervention. New England Journal of Medicine. Published online March 29, 2026. doi:https://doi.org/10.1056/nejmoa2515704

‌Kapur NK, Mangner N, Aghili N, et al. Left Ventricular Unloading in Anterior STEMI without Shock: The STEMI Door to Unload (DTU) Randomized Controlled Trial. JACC. Published online March 2026. doi:https://doi.org/10.1016/j.jacc.2026.03.071

‌

For most patients who’ve been in the ICU more than a few days who then develop a new suspected bacteremia, empiric treatment with very broad-spectrum antibiotics like piperacillin-tazobactam, cefepime, or carbapenems is almost a clinical reflex. After that, an organism’s sensitivity usually isn’t known for at least a few days, during which broad coverage is continued in most patients.

A new generation of rapid antimicrobial sensitivity tests may reduce that timeframe. Evidence is emerging that these assays can direct optimized antibiotic therapy earlier, in either direction (broader, for resistant organisms, or narrower, when resistance is unlikely).

There’s no proof they improve hard outcomes— but could insisting on that level of certainty be a disservice to patients and the antibiotic stewardship project?

FAST: It’s Faster (But Not That Fast)

The status quo

First, a recap of traditional sensitivity testing. After standard blood cultures have incubated in broth media (in the bottles) and “turned positive” (usually by detection of high CO2 production inside the bottle), the broth is inoculated onto agar plates. Once enough colonies grow, the organism can be identified through conventional methods and complementary molecular techniques.

Antibiotics can then be applied directly to the agar to determine sensitivity, with other more advanced techniques often used (e.g., dilution of the broth into multiple wells containing antibiotics).

Conventional sensitivity testing takes about 24 to 48 hours after the blood cultures turn positive, often two to five days after blood cultures are drawn.

Rapid susceptibility testing

Rapid susceptibility tests are performed on the blood culture broth in the inoculated bottles after the cultures have returned as positive—rapidly, without the necessity of waiting for growth on agar or other media.

They use techniques not dissimilar to conventional susceptibility testing (e.g., diluting broth across a matrix of antibiotic-impregnated microwells), but identify signs of resistance faster (e.g., detecting volatile metabolites from the wells, or microscopic optical changes, rather than waiting for the traditional slower signals).

Rapid AST assays can return susceptibility results in less than 8 hours.

It’s important to distinguish rapid AST (which test an actual cultured bacteria for observable resistance) from inference-based rapid molecular genetic tests (e.g., BioFire™, Cepheid Expert™) that identify bacterial DNA and resistance genes without actually observing resistance or susceptibility.

Six randomized trials have tested rapid AST in patients with bacteremia:

• Banerjee et al JAMA 2026, (the FAST trial, n=850)

• Banerjee et al Clin Infect Dis 2015 (n=617)

• Beuving et al Eur J Clin Microbiol Infect Dis 2015 (n=129)

• Bannerjee et al Clin Infect Dis 2021 (n=500)

• Kim et al, Clin Microbiol Infect 2021 (n=116)

• Caspar et al Eur J Clin Microbiol Infect Dis 2024 (n=112)

All found that use of rapid AST identified organisms and antimicrobial resistance faster, resulting in faster delivery of optimal antimicrobial therapy—usually by 12 to 48 hours faster than conventional sensitivity testing.

This usually led to more accurate targeting of antibiotics and less use of broad spectrum agents.

Rapid testing had high agreement (e.g., ~94%) with conventional testing.

However, none of the six RCTs found that use of rapid AST reduced mortality, hospital lengths of stay, or other “hard” clinically relevant outcomes.

The FAST Trial

As the first five trials were conducted in countries with relatively low rates of antibiotic-resistant organisms, the FAST trial was performed specifically to test the use of rapid AST at hospitals in nations with high antibiotic resistance.

At 7 medical centers in 4 countries with a high prevalence of multidrug-resistant gram-negative bacteria (Greece, India, Israel, and Spain), 899 patients with confirmed gram-negative bacteremia were randomized to standard susceptibility testing or fast AST using Biomerieux’s system (VITEK Reveal).

There was even less support for rapid AST in this trial than its predecessors: similar proportions of patients received appropriate antibiotics at enrollment, 24 hours, and at 3 days.

However, patients with resistant organisms had a shorter time to optimal targeting of their antimicrobials (escalation or de-escalation). For cephalopsporin-resistant organisms, rapid AST led to changes in ~12 vs. 25 hours with conventional testing; for carbapenem-resistant bugs, tailoring occurred after 13 vs. 47 hours.

Why No Benefit?

Rapid susceptibility testing occurs well after the critical decision and time window has passed for treatment of severe infections with antibiotics. Most patients with positive blood cultures have already received antibiotics for the two days while their cultures were incubating. Empiric therapy, not downstream tailoring does almost all the work in saving lives from sepsis.

Patients in clinical trials are often already receiving high-quality care at advanced centers, which for patients with suspected bacteremia in the ICU includes broad-spectrum antibiotics; this likely led to good outcomes in the control arms of these trials.

It’s also possible that most patients don’t suffer bad outcomes from being on suboptimal antibiotics for 1-2 days, on average.

Mortality is driven by many other factors besides antibiotic susceptibility. Anecdotally, the patient whose death is intuitively attributable to undertreatment with broad spectrum antibiotics is a rarity; overtreatment is by far the more common scenario in modern ICUs.

It’s Worth Being Bullish on Rapid AST Anyway

Accumulated evidence strongly suggests that rapid susceptibility testing can lead to faster appropriate antibiotic de-escalation, and likely leads to more rapid escalation to cover resistant organisms as well. It’s rare to find a new test that works so consistently well on an important, clinically-relevant measure.

Although pricing is opaque, rapid susceptibility testing appears to be expensive, both on a platform and per-unit basis, and in the absence of demonstrable clinical or financial benefit, adoption may be slow. This is an area of ongoing innovation and competition among vendors of the competing diagnostic platforms and assays.

For the practicing clinician, an important distinction at this stage of the game is between assays that test actual resistance on live, cultured bacteria (e.g. the rapid AST test we’ve been discussing and its competitors), which have been demonstrated to have a high concordance with conventional susceptibility testing — versus the molecular DNA-based tests that only infer or predict resistance based on genetic profiles. The latter class of tests can provide a lot of value in early targeting decisions, but must be complemented with phenotypic resistance testing (rapid and/or conventional AST).

References

Fast Antimicrobial Susceptibility Testing for Gram-Negative Bacteremia: Research Summary. JAMA. 2026 Apr 18:e265536. doi: 10.1001/jama.2026.5536. Epub ahead of print. PMID: 41999288.

• Banerjee et al JAMA 2026, (the FAST trial, n=850)

• Banerjee et al Clin Infect Dis 2015 (n=617)

• Beuving et al Eur J Clin Microbiol Infect Dis 2015 (n=129)

• Bannerjee et al Clin Infect Dis 2021 (n=500)

• Kim et al, Clin Microbiol Infect 2021 (n=116)

• Caspar et al Eur J Clin Microbiol Infect Dis 2024 (n=112)

A 46-year-old man who drinks large amounts of alcohol daily and who has a history of MRSA bacteremia presents to the ED after falling down stairs while intoxicated. Ethanol level is 80 mg/dL. He is evaluated by the trauma team and “cleared” after CT and exam show no fractures or hemorrhage. He is agitated and is considered “too much for the floor,” and is admitted to the ICU on a dexmedetomidine infusion and symptom-triggered therapy for potential alcohol withdrawal delirium.

Two hours later, he complains of severe neck pain. On exam, vitals are normal (BP 100/55, pulse 90, respirations 18/min, afebrile), but he has diminished (4/5) strength in the deltoids, biceps, and handgrip bilaterally. Plantar flexion and hip flexion are both 5/5 bilaterally.

Cervical spine CT shows degenerative cervical spondylosis without fracture. He is placed in a cervical immobilization collar. An MRI of the cervical spine is pending. Neurosurgery has been consulted.

Read more

Acute kidney injury is associated with worse outcomes after surgery and acute illness generally. Preventing it has proven to be much harder than predicting it, however.

There’s an irresolvable problem at the heart of AKI research today: it’s impossible to reliably differentiate which patients have true injury from those with a transient, essentially benign reduction in creatinine clearance.

A recent randomized trial found a huge reduction in the incidence of moderate to severe AKI after major surgery, with the application of a biomarker-guided supportive care bundle. This was remarkable in and of itself.

But after congratulating the authors, we should also ask: were any kidneys saved?

Bigpak-2 Trial

At 34 centers in Europe, 1,180 patients undergoing major surgery (90% elective, usually abdominal/general, cardiac, or vascular) and at high risk for AKI were randomized to receive guideline-compliant preventive care (advanced hemodynamic monitoring, optimized volume status and hemodynamics, avoidance of nephrotoxins and contrast, and prevention of hyperglycemia), or usual care.

Intervention patients’ angiotensin-converting enzyme (ACE) inhibitors and angiotensin II receptor blockers were stopped for at least 48 h postoperatively. Usual care patients’ were as well, but less often.

They got passive leg raises every 3 hours for at least 12 hours, with fluid boluses for predicted fluid responsiveness.

They also got invasive hemodynamic monitoring and/or regular ultrasound exams.

Less Moderate AKI in the Intervention Arm…

Within 72 hours after surgery, 8% fewer patients in the intervention group (14.4%) than in the usual care arm (22%) had developed moderate or worse AKI (at least a doubling of creatinine or urine output <0.5 mL/kg/hr, or about 840 mL/day in a 70 kg patient).

The positive finding was most often triggered by a temporary drop in urine output below 0.5 mL/kg/hour; the creatinine-doubling criterion was met in 6% (intervention) vs 10% (controls).

There was no difference in severe (stage 3) AKI: 16% in the intervention group vs 15.8% in controls.

…But No Difference in RRT

However, there was no difference in clinically relevant endpoints, most importantly the use of renal replacement therapy (dialysis or CRRT) at 30 days (5.1% intervention vs 5.8% control), or 90 days (5.1% vs. 5.9%). Neither was there a difference in deaths at 30 days (5.1% vs 4.6%) or 90 days (7% vs 7%).

There was no difference in the MAKE30 or MAKE90 rates (in-hospital death, new receipt of renal replacement therapy (RRT), or persistent renal dysfunction at 30 and 90 days), which were virtually identical between groups.

Patients in the bundle arm had a 2.4% absolute greater rate of renal recovery at day 90 (non-significant).

The Biomarker Angle

Bigpak-2 was funded by Biomerieux, a family-owned multi-billion-euro diagnostics company (e.g., bacterial cultures, procalcitonin, BioFire, et al) headquartered in France.

Biomerieux markets the NephroCheck™ test, which is promoted as an early-warning system for severe AKI. Nephrocheck measures serum levels of the biomarkers TIMP-2 and IGFBP7. Tissue inhibitor of metalloproteinases-2 (TIMP-2) and insulin-like growth factor-binding protein 7 (IFGBP7) are signaling proteins secreted by renal tubular cells under metabolic stress that lead them to slow mitosis as a conservation strategy.

Nephrocheck multiplies TIMP-2 and IGFBP7 and returns the product as a value indexed to purported renal stress. Results >0.3 /1000 ng/mL were used as an inclusion criterion in Bigpak-2—ostensibly “enriching” the population for patients at high risk for progressing to severe AKI and dialysis.

But patients receiving the intervention guided by Nephrocheck did not have a lower rate of renal replacement therapy, nor of severe AKI (16% vs 15.8%, non-significant).

Nor did they have a greater reduction in their assay level over time (the product of TIMP-2 and IGFBP), compared to the usual care group.

That would seem to pose a problem for Nephrocheck’s use-case story, which is to identify patients at high risk for dialysis and extended stays so these outcomes can be prevented. When the interventions are working, Nephrocheck levels should fall, representing reduced risk for severe AKI.

Earlier randomized trials, including the first Bigpak trial and others, showed the same general finding: bundle-driven care in postop populations at high risk for AKI after “enrichment” with TIMP-2 x IGFBP7 (and clinical factors), led to a reduced rate of AKI but no significant reduction in need for dialysis.

But based on those earlier trials showing similar findings as Bigpak-2, use of Nephrocheck was recommended in the Enhanced Recovery After Surgery (ERAS) guidelines.

Surgeons and their hospitals have signed on, in part because it’s easy to run the seemingly advanced Nephrocheck assay on a machine platform (VIDAS) they already own.

And there’s no strong evidence that it doesn’t help in a clinically meaningful way.

Discussion

“Our findings show that the prevention of hypotension and discontinuation of ACE inhibitors and ARBs had the strongest association with the primary outcome.” —Bigpak-2 authors

The consistent relative improvement in creatinine clearance seen with biomarker-guided AKI prevention could either be from the reduction of genuine injury or better hydration and volume management (and a higher rate of ACEI/ARB discontinuation).

Intervention patients were better monitored and had much higher compliance with KDIGO bundle guidelines. They had half the rate of “hypotension for longer than 5 minutes with no intervention performed to prevent this” (an unusual, post-hoc circular metric).

This might be meaningful, but could alternatively represent a transient improvement in urine output or creatinine clearance by kidneys that were destined for the same outcome, either way.

The lack of a signal showing a reduction in the need for renal replacement therapy—even in an “enriched” population at high risk for AKI—suggests the latter.

All that said, there is a vaguely persistent signal that seems to pervade these sorts of trials. Patients who get more clinical attention do better, in one modest way or another.

That mysterious and scarce resource—focused time and cognitive effort at the bedside—may not always be measurable, but it always matters.

In terms of observable outcomes, however, in Bigpak-2 it resulted in better-managed volume status and reduced transient prerenal azotemia, without saving kidneys or preventing dialysis.

Reference

Zarbock A, et al ; BigpAK-2 study group. A preventive care strategy to reduce moderate or severe acute kidney injury after major surgery (BigpAK-2); a multinational, randomised clinical trial. Lancet. 2025 Dec 13;406(10521):2782-2791. doi: 10.1016/S0140-6736(25)01717-9. Epub 2025 Nov 13. PMID: 41242333.

Göcze I, et al. Biomarker-guided Intervention to Prevent Acute Kidney Injury After Major Surgery: The Prospective Randomized BigpAK Study. Ann Surg. 2018 Jun;267(6):1013-1020. doi: 10.1097/SLA.0000000000002485. PMID: 28857811.

Enhanced Recovery After Surgery (ERAS) guidelines.

von Groote et al, meta-analysis in ICM