Editorial 

Will the Bicarbonate for In-Hospital Cardiac Arrest Trial Change Practice?

Clifton W. Callaway

JAMA Published Online: June 11, 2026

doi: 10.1001/jama.2026.10508

In this issue of JAMA, Granfeldt and colleagues share results of the Bicarbonate for In-Hospital Cardiac Arrest randomized clinical trial comparing routine administration of sodium bicarbonate with placebo during attempted resuscitation of 779 patients with in-hospital cardiac arrest.¹ The physiological rationale for administering sodium bicarbonate is to temporarily reduce the profound metabolic acidosis that develops during cardiac arrest to reverse myocardial depression and improve catecholamine action. Sodium bicarbonate is hypertonic and may increase intravascular volume. The primary outcome of return of spontaneous circulation was not different after sodium bicarbonate (39%) relative to placebo (37%). The intervention also did not change secondary outcomes, including organ dysfunction in the first 72 hours, 30-day survival (12% vs 9%), functional scores, or quality-of-life scores. Among patients who survived, those in the sodium bicarbonate group had higher pH and blood bicarbonate values, demonstrating successful delivery of the intervention.

The current American Heart Association guidelines have recommended against routine use of sodium bicarbonate during cardiac arrest since 2010, except in special circumstances, such as metabolic acidosis, hyperkalemia, or tricyclic antidepressant overdose, where the sodium load or alkalinization might be beneficial.^2,3 The few clinical trials that test sodium bicarbonate during cardiac arrest do not demonstrate benefits in long-term survival or neurological recovery, although most have methodological limitations.³ Guidelines transparently declare the evidence supporting the recommendation against routine use of sodium bicarbonate to be low or very-low certainty.⁴

Despite guidelines, sodium bicarbonate is one of the most frequently administered resuscitation drugs after epinephrine, with use among 40% to 50% of patients with in-hospital cardiac arrest⁵ and 19% of patients with out-of-hospital cardiac arrest.⁶ Bicarbonate use during in-hospital cardiac arrest has increased since 2001, including a modest increase since 2010 for patients with nonshockable rhythms.⁵ Clinicians vary widely in their enthusiasm for using sodium bicarbonate.^6,7 Many clinicians believe that this drug is useful and will require high certainty of evidence to remove it from their practice.

The current trial has numerous strengths in design and execution that increase trust in the results. Patients were selected with clear inclusion and exclusion criteria—more than 65% of eligible patients were enrolled. Randomization was concealed by using identically labeled study drug vials. The trial reached its target sample size of 778. The primary outcome was missing for only 1 patient. Results were identical for the comparison between groups both of unadjusted outcomes and of outcomes adjusted for strong confounders, such as age, witnessed status, and initial electrocardiographic rhythm. The results of this trial are strong evidence that the routine use of sodium bicarbonate in this population does not alter important outcomes.

A potential concern is that the primary analysis used a modified intention-to-treat approach, where the outcome for only patients who received the study drug was analyzed. Although this approach seems similar to a per-protocol analysis, in fact, most randomized patients who were excluded (57 of 70 in the bicarbonate group and 55 of 60 in the placebo group) were excluded because resuscitation ended prior to receiving the drug or the drug was otherwise not available. It is biologically implausible that the blinded study drug could alter the observed outcomes when it was not administered, and excluding those cases from numerator and denominator may have improved precision in the estimate of the average treatment effect.

Another potential concern is that the primary outcome was return of spontaneous circulation rather than long-term neurological outcomes. Scientific statements emphasize the primacy of patient-centered, long-term outcomes for guiding resuscitation practice.⁸ However, improving those long-term outcomes requires optimization of many individual and interacting treatment decisions and multitreatment pathways. Random allocation balances measured and unmeasured confounders that are present at baseline prior to treatment, but does not guarantee balance in those decisions and treatments that occur after randomization. When interpreting long-term outcomes, it is important to measure postrandomization confounders and consider sensitivity analyses to account for potential bias.⁹ In this trial, the investigators wisely selected return of spontaneous circulation as a primary outcome that is both important and proximate in time to the intervention. This minimizes the opportunity for postrandomization confounders to obscure any treatment effect. Furthermore, this article presents the secondary outcomes at 30 days not just as a binary outcome of survival, but as categories including survival, recurrent cardiac or respiratory arrest, and death after withdrawal of life support. This detailed reporting of the circumstances surrounding in-hospital death, which allows interpretation of postrandomization treatment decisions, should be routine in all trials of critical illness.

One reason clinicians might continue to use sodium bicarbonate despite the results of this trial is that they focus treatment on individual patients, not populations. Clinical trials are the best method for estimating the average treatment effect of an intervention for a population, but the effect of a treatment on a specific patient might differ from the average effect. Traditionally, trials look for clues about heterogeneity of treatment effects between individuals by including analyses within subgroups of patients who might have differential benefit or harm. Conventional wisdom is to define these subgroups a priori, based on biologically plausible variables, to avoid spurious findings. For example, did treatment effects vary between men and women, between old and young, or by etiology underlying the cardiac arrest? This trial examined treatment effect among several subgroups defined by time to drug administration, age, initial electrocardiographic rhythm, known prearrest metabolic acidosis, and witnessed arrest. None of these analyses identified a group with different response to the drug.

Recently, more sophisticated analytic approaches were proposed to understand heterogeneity of treatment effects between individuals, recognizing that each individual patient has multiple observable characteristics that might influence response to treatment. Machine learning methods can use multiple variables to estimate individual treatment effects and potentially identify specific individuals who might benefit from an intervention.^10,11 For example, older women with respiratory etiologies might benefit more than younger men with cardiac etiologies. This multivariable post hoc analysis of trial data requires subsequent testing in separate datasets or trials to reject false discoveries. Although this article did not present individual treatment effects, the trial collected sufficiently granular data to allow subsequent exploration of personalized approaches.

Can the results of this trial in patients with in-hospital cardiac arrest inform out-of-hospital cardiac arrest practice? The demographics, etiology, and outcomes differ substantially between in-hospital and out-of-hospital cardiac arrest. In general, out-of-hospital cardiac arrests require longer times to deliver drug therapy and may develop greater metabolic acidosis. Unfortunately, this trial did not detect a trend for increased benefit of sodium bicarbonate related to later timing of drug or acidosis, which may inform out-of-hospital cardiac arrest practice. Definitive evidence for the benefits of buffer therapy in out-of-hospital cardiac arrest may require separate testing.

Will this trial change practice and reduce sodium bicarbonate use during resuscitation attempts? It seems unlikely to do so, even though this rigorous and high-quality clinical trial reinforces and provides stronger evidence for the current guideline recommendation against routine use of sodium bicarbonate during in-hospital cardiac arrest. The administration of bicarbonate during cardiac arrest entered practice in the 1960s based on small preclinical studies that included the combination of bicarbonate and epinephrine.¹² Giving buffer therapy to a patient who is certain to have severe acidosis has remained intuitively appealing to clinicians, despite subsequent clinical trials and preclinical laboratory studies finding equivocal effects of sodium bicarbonate outside of special circumstances.¹³

Rather than stronger guidelines, clinicians will demand more exploration of individual treatment effects to change practice. It is imperative that we acknowledge the average treatment effect from trials like this one to advise what we should routinely do for all patients, but we must then use the high-quality data from trials to thoroughly look for any individual treatment effects. Even if those analyses cannot identify a subgroup of patients who do benefit from sodium bicarbonate, clinicians still may not stop using this drug as a heroic effort. If so, the neutral result of this trial is at least reassuring that sodium bicarbonate is not harmful.