Hold the potassium

It was not an unusual event during my clinical fellowship to get a call from the cardiac team asking for dialysis in a patient with hyperkalemia and acute renal failure. On a few occasions the hyperkalemia seemed disproportionate to the level of renal function and on further investigation, this was found to the as a result of the zealous correction of the potassium level to >4 mEq/L in patients in the cardiac unit. I have often wondered about this practice and whether or not there was firm evidence for keeping the potassium in the 4-5 mEq/L range (or even 4.5-5.5 as suggested by some authors) and so it was with interest that I read this paper that just appeared in JAMA.

As mentioned in the paper, the data suggesting that low potassium levels are associated with increased mortality are relatively old and date from an era when ventricular arrhythmias were more common following an MI. They were also generally relatively small studies. The authors of this study used a database of patients presenting to 67 US hospitals with an ICD9 code for MI and increased cardiac biomarkers. In total >39,000 patients were included.

They were looking primarily at post-admission potassium levels and their relationship with in-hospital mortality and the occurrence of arrhythmias. AS one would expect, there was a U-shaped curve for the relationship between in-hospital mortality and the potassium level. What was unexpected, however, was that the lowest mortality was seen in patients with a potassium between 3.5 and 4.5 mEq/L and that the mortality doubled in patients with a potassium between 4.5 and 5 mEq/L.

This was borne out in the fully adjusted model. The OR for in-hospital mortality was 1.96 (CI 1.64-2.34) for patients with a potassium between 4.5 and 5 mEq/L. Interestingly, the risk of a ventricular arrhythmia was the same in patients in the midrange of potassium values and only increased in patients with a potassium of below 3 or above 5 mEq/L. This contrasted with the mortality data and the authors suggested that this might be in part a result of incorrect coding of ventricular arrhythmias and is a potential limitation of the study. Also, this study certainly does not prove that replacing potassium to a level above 4.5 mEq/L is dangerous. This could only be answered by a randomized trial. There may be some residual confounders that have not been accounted for in the model. Still, as the authors point out, this study challenges current guidelines and suggest that a better target for potassium in patients following an acute MI would be 3.5-4.5 mEq/L. Perhaps we might see less of this particular consult in the future?