A brief editorial aside: Both Josh and Scott have commented on the BICAR-ICU Trial in previous posts, each presenting well thought out erudite analyses. I recommend reading/listening to them.
EMCrit: Acid Base Episode 7
The overwhelming need for euboxia has inspired a multitude of therapeutic endeavors, including the use of sodium bicarbonate infusions. The physiologic plausibility often cited to support these interventions, is protein function suffers in a severely acidotic milieu, leading to cardiovascular dysfunction, refractory hypotension, multi-organ failure and death. But these dictums have very little science to support their assertions. In fact, there is a body of evidence suggesting that the exact level of acidemia matters far less than the physiologic process driving the accumulation of the acid in the first place. Like most medical controversies with a paucity of supporting evidence, opinions have hardened into opposing extremes. Small concessions are made from either side,
Only use bicarb if the pH is less than 7.2.
Only use bicarb for bicarbonate wasting problems or renal failure.
Do not use sodium bicarbonate to correct a lactic acidosis.
Each allowance based on just as little evidence as that which supports the primary opinions. The publication of a large multicenter RCT by Jaber et al, the BICAR-ICU trial represents the first large high quality attempt to quantify the benefit of alkalinization therapy in a cohort of critically ill ICU patients with metabolic acidosis1. Despite these massive efforts, the results will likely do nothing more than reinforce established opinions.
The authors conducted a multicenter, randomized, open-label controlled trial at 26 ICUs in France. They enrolled adults,18 years or older, within 48-hrs of their admission with severe acidosis (defined as pH ≤7.20, PaCO2 ≤45 mm Hg, and sodium bicarbonate concentration ≤20 mmol/L), and either a SOFA score greater than 4 or a serum lactate greater than 2 mmol/L. Patients were excluded if they had respiratory acidosis, proven digestive or urinary tract loss of sodium bicarbonate, stage IV chronic kidney disease, ketoacidosis, and sodium bicarbonate infusion within 24 hr before screening. Patients randomized to receive sodium bicarbonate received a 4.2% sodium bicarbonate infusion with the aim of achieving an arterial pH of 7.30 or more during the 28-day ICU admission or until ICU discharge.
Over a two year period the authors enrolled 400 patients (201 in the control arm and 199 in the bicarbonate arm). There was no difference in their primary outcome, composite of death from any cause by 28 days after randomization, and the presence of at least one organ failure at 7 days after randomization (71% in the control group vs 66% in the bicarbonate group, ARR 5.5%, 95% CI 15.2%-4.2%; p=0.24). Nor did the authors demonstrate a statistically significant difference in 28-day mortality, 46% in the control group vs 55% in the bicarbonate group (p-value of 0.07).
The authors did report fairly notable secondary outcomes in favor of the bicarbonate group. In their a priori subgroup analysis of patients with acute kidney injury, the primary outcome occurred significantly more frequently in the control group (82% vs 70%). They also note a statistically significant decrease in 28-day mortality in the control group vs the bicarbonate group. In fact, the patients in the bicarbonate group did noticeably better when compared to the control group in almost every metric examined. While not statistically significant the authors found a 5.5% absolute decrease in their primary outcome, a 9% decrease in 28-day mortality, a 16.7% decrease in the need for renal replacement therapy, as well as a 12.5% decrease in the need for dialysis at ICU discharge.
Given these results one could argue that patients randomized to receive the bicarbonate infusion clearly demonstrated improved outcomes, but the trial was underpowered to detect a clinically important difference. We are unable to differentiate a 5.5% decrease in the rate of death or severe organ failure from statistical chance. How do we interpret a statistically negative study with strikingly positive results? Where does the burden of evidence fall?
The goal of this trial was to determine the efficacy of a sodium bicarbonate infusion in critically ill patients with metabolic acidosis. While this trial is not perfect (open label with a questionable primary outcome, with significant crossover between groups), let us for argument sake accept that the randomization process effectively limited the influence various confounders had on the results. There is no question patients randomized to the bicarbonate grouped fared better. The question is why was there such a difference between the two groups? How much did random error influence the observed results and to what degree do these results deviate from the underlying truth? Statistical tools, such as the p-value, are utilized in an attempt to quantify this risk of random error. In this case, while there is a clear difference between groups, the likelihood this difference is due to random chance is higher than what we traditionally view as an acceptable risk (p-value > 0.05).
To adequately assess the efficacy of bicarbonate therapy we not only need to grasp the potential for random sampling error in the current trial, we have to know the prior evidence of benefit. This includes a plausible biological explanation as well as previous clinical data demonstrating efficacy. In this case there is an underlying physiologic reasoning supporting the restorative properties attributed to the use of sodium bicarbonate. The thought process can get fairly convoluted, but at its core it is simple. Acidosis is associated with a significant amount of morbidity and mortality. Bicarbonate can potentially correct the serum pH. If you correct the acidosis you may limit the unwanted associated outcomes. But there is an equally plausible counterargument. The acidosis is simply a symptom of an underlying pathologic process that leads to the morbidity and mortality associated with acidosis. Fixing the acidosis without correcting the underlying cause may make the laboratory values appear better, but will not impact patient oriented outcomes.
Prior to the publication of the BICAR-ICU trial, the clinical data examining the use of sodium bicarbonate as a therapeutic buffer was at best sparse. There have been a number of small prospective controlled trials that found bicarbonate administration has no effect on patient hemodynamics (increased cardiac output or decreased need for vasopressor agents) 2,3. The BICAR-ICU trial revealed similar results, finding no difference in vasopressor free days (9 vs 19), in vasopressor free days in survivors (26 in each group), nor a difference in duration of vasopressor therapy (mean of 2 in each group).
We also have a significant body of evidence suggesting that acidosis in and of itself causes minimal physiologic unrest4,5,6. Frumin demonstrated patients tolerated acidosis from hypercapnea to extremely low levels (pH as low as 6.8) without any signs of physiologic distress or hemodynamic compromise7. Olympic rowers, shortly after a race have demonstrated serum pHs as low as 6.8 without any ill effects8. We have all seen patients with DKA with impressively severe acidosis often with minimal hemodynamic compromise. And yet the septic patient with a lactic acidosis and a pH of 7.1 is often in refractory shock and dying. A recent article in Critical Care Medicine by Masevicius et al supports this shared anecdotal experience. In a prospective observational cohort the authors enrolled 4901 ICU patients, 1609 meeting criteria for a metabolic acidosis9. The authors reported that for any given pH, an acidosis due to lactate or unmeasured anions had a much higher mortality than a SID acidosis, which demonstrated mortality rates similar to patients without acidosis.
Even more telling then the lack of previous evidence supporting bicarbonate therapy for severe acidosis, is the overwhelmingly impressive effect size observed in this cohort. It is unlikely the 9% improvement in 28-day mortality associated with the use of a bicarbonate infusion reported in this trial represents a true effect size, especially given that no prior evidence has ever found correcting acidosis improves patient oriented outcomes. This is due in part to random sampling error. If an attempt to reproduce these findings was performed we would likely see a significant regression to the mean.
If the results of the BICAR-ICU trial represented a true therapeutic effect from the use of bicarbonate, the ICU community would rejoice. No one would argue against a simple, low resource, low cost intervention that reduces ICU mortality and need for dialysis. But a single study, especially one with a negative primary endpoint is incapable of providing definitive evidence representing the underlying truth. Our reaction to such a study should not be to argue the number of zeroes following a decimal point on some secondary endpoint, or use these results to justify our preconceived physiologic beliefs. Rather, repeat these results, validate its findings, then and only then will we have an idea of the real value of a sodium bicarbonate infusion and alkalization therapy.