Sodium Bicarbonate for Critically Ill Adults with Metabolic Acidosis and Shock

Welcome to AudioScholar. Today we're covering Sodium Bicarbonate for Critically Ill Adults with Metabolic Acidosis and Shock, by Ary Serpa Neto and colleagues, published in New England Journal of Medicine.

Metabolic acidosis in the shocked patient is one of the oldest reflexes in critical care. The acidemia is associated with worse myocardial contractility, blunted catecholamine response, and higher mortality, and giving bicarbonate makes the numbers on the gas look better within an hour. Whether it makes the patient better has never been settled. The early trials from the nineteen nineties corrected the pH without moving hemodynamics or survival. The BICAR-ICU trial in 2018 found no overall mortality benefit but a signal in the subgroup with acute kidney injury, both for death and for dialysis. That subgroup finding is what drove the 2021 Surviving Sepsis guidance to endorse bicarbonate in septic shock with acute kidney injury. The problem is that BICAR-ICU was open label, and so was its successor, BICAR-ICU 2. Deciding to start renal replacement therapy is a judgment call, not a hard endpoint, and a clinician who knows the patient is already receiving bicarbonate has every reason to wait a little longer before dialyzing. A reduction in dialysis generated that way is at least as consistent with performance bias as with a real kidney effect, and an open label trial cannot tell you which one you're looking at.

That is the gap SODa-BIC was built to close. It is a pragmatic, adaptive, double-blind, placebo-controlled phase three trial run across fifty-five intensive care units in seven countries, predominantly Australia and New Zealand, between April 2023 and December 2025. Adults were eligible if they were on vasopressors to maintain a mean arterial pressure above 65, and had a pH below 7.30, a base excess of minus four or worse, and a carbon dioxide tension no higher than 45 millimetres of mercury, or 50 if they were intubated — all three criteria within the two hours before randomization, which together ensure the acidosis is genuinely metabolic and not respiratory. Patients who had met criteria for more than forty-eight hours were excluded, as were those with a clear secondary cause such as diabetic ketoacidosis or gastrointestinal losses, an estimated glomerular filtration rate below thirty, significant electrolyte derangement, or dialysis already underway or planned within three hours.

The intervention was 8.4 percent sodium bicarbonate diluted in dextrose to a final concentration of 600 milliequivalents per litre, titrated by a protocolized algorithm at one, three and five hours toward a pH of 7.30 or above with a base excess of zero or better. The comparator was 5 percent dextrose, macroscopically indistinguishable from the active drug, and the person preparing the infusion was separate from the blinded investigator collecting outcomes. The primary endpoint was major adverse kidney events at thirty days — the composite of death from any cause, initiation of kidney support therapy, or persistent renal dysfunction. Five hundred patients gave ninety percent power to detect a fourteen percentage point absolute reduction from an assumed control event rate of forty percent, with a blinded sample size re-estimation after the first hundred patients that could push enrolment as high as seven hundred if the event rate came in low.

Five hundred patients were randomized and 498 analyzed. This was a sick cohort — median age sixty-six, median APACHE two score of twenty-one, median SOFA of seven, and when the investigators compared them to the general intensive care population at the same units over the same period, the trial patients had substantially higher illness severity and roughly double the hospital mortality. Sepsis or septic shock was the cause of acidosis in about half. Importantly, this was moderate rather than severe acidemia: the median pH at enrolment was 7.26, with a median base excess of minus nine, and only about forty percent had a pH below 7.25.

The intervention did what it was supposed to do biochemically. Patients in the bicarbonate arm reached the pH target earlier, and recurrence of metabolic acidosis over the first seven days was roughly halved — about a third of the bicarbonate group versus more than half of the placebo group. Rescue open-label bicarbonate was needed about twice as often in the placebo arm. So there was real separation in acid-base physiology, at least early. By day seven, the pH curves in the two groups had converged.

None of that translated. Major adverse kidney events at thirty days occurred in about forty percent of patients in each arm — essentially superimposable, with an adjusted absolute difference of just over one percentage point favouring placebo. Death in hospital by day thirty was about a quarter in both groups, and there was no separation in the survival curves at ninety days either. Prespecified sensitivity analyses adjusting for age, sex, pre-randomization bicarbonate use and illness severity gave the same answer, as did an alternative filtration-rate-based definition of persistent renal dysfunction. A confidence distribution analysis put the probability that bicarbonate reduces the primary outcome to any degree at just under forty percent — in other words, the weight of the evidence tips very slightly toward no benefit rather than toward benefit.

The kidney-specific components are worth dwelling on, because this is where the prior literature made its claim. Every one of them moved numerically in bicarbonate's favour, and none reached significance. Kidney support therapy within thirty days was about seventeen percent with bicarbonate versus about twenty-one percent with placebo. Persistent renal dysfunction was fourteen versus eighteen percent. Dialysis dependence at day thirty was under six percent versus nine percent. Three point estimates all leaning the same way is either an underpowered trial, a genuinely absent effect, or both, and this trial cannot distinguish between them. What it can say is that once you blind the clinician making the dialysis decision, the dialysis signal from BICAR-ICU 2 shrinks to something that no longer clears the bar. In the subgroup with a baseline pH below 7.25 there was still an eleven point difference in dialysis use, but it was not significant, and there was no evidence of treatment effect modification across any of the eight prespecified subgroups — not septic shock, not stage two or three acute kidney injury, not anion gap versus non-anion gap acidosis, not illness severity.

Everything else was neutral. Acute kidney injury within seven days occurred in about sixty-five percent of both arms. Vasopressor-free days, dialysis-free days, intensive care unit-free days at thirty days, and hospital-free days at ninety days were all indistinguishable. Harm was minimal: adverse events in under two percent of the bicarbonate group and none in placebo, driven by hypokalemia requiring correction, with a single case of hypernatremia. No serious adverse events in either arm.

The strengths here are real. This is the first adequately sized, genuinely double-blind trial of bicarbonate in this population, with a patient-centred composite endpoint, a prespecified blinded sample size re-estimation, and fifty-five centres across multiple health systems. The blinding is the point — it removes exactly the bias that made the earlier renal signals uninterpretable.

The limitations deserve equal honesty. Open-label bicarbonate was permitted when clinically indicated, and other sodium- and buffer-containing fluids were not controlled, so treatment separation was attenuated; about fourteen percent of the bicarbonate arm and nine percent of the placebo arm had already received bicarbonate before randomization. The correction of acidemia, while rapid, was brief, and it is fair to ask whether a few hours of buffering could plausibly alter a thirty-day outcome. Blinding may have been imperfect, since a clinician watching the bicarbonate level climb can guess the allocation. Physiological data at the moment of dialysis initiation were not collected, so we cannot fully audit whether the threshold for starting differed. And the trial was powered for a moderate to large effect. It convincingly excludes a major benefit; it does not exclude a smaller, still clinically meaningful kidney-specific one. Finally, the phenotype differs from the earlier trials — this was moderate acidosis with vasopressor dependence, not the severe acidemia with severe acute kidney injury that BICAR-ICU 2 studied.

So what should change on Monday. Routine sodium bicarbonate infusion to prevent major adverse kidney events in the vasopressor-dependent patient with moderate metabolic acidosis is now difficult to justify. It corrects the gas and it reduces recurrence of acidosis, but it does not reduce death, does not reduce dialysis, and does not shorten time on pressors or in the unit. If your practice has been to reach for bicarbonate in the septic patient with a pH of 7.25 and a rising creatinine on the strength of the earlier subgroup data, that rationale has weakened considerably. What this trial does not address is the patient with profound acidemia below 7.10, or the patient with hyperkalemia, or established renal tubular acidosis — none of those questions are answered here. And there remains a defensible pragmatic use: buffering a deteriorating patient as a temporizing measure while you treat the underlying cause. Just be clear with yourself about what you're doing. The takeaway is not that bicarbonate is useless. It is that you should not expect it to change the prognosis.

That's your AudioScholar summary. The full transcript and reference are on the episode page. Until next time.

This is an automated summary generated by artificial intelligence, which can make mistakes. Always review the original source materials.