Objective: To evaluate the efficacy and safety of potassium replacement infusions in critically ill patients.
Design: Prospective cohort study.
Setting: Multidisciplinary critical care unit.
Patients: Forty-eight critically ill adult patients, age 25 to 86 yrs. Patients entered the study when hypokalemia (potassium less than 3.5 mmol/L) was noted on routine laboratory blood analysis. Most common primary diagnoses on ICU admission included postoperative cardiac surgery (n = 9), sepsis and multiple organ system failure (n = 9), complicated myocardial infarction (n = 7), and respiratory failure (n = 5).
Intervention: Potassium chloride infusions (20, 30, or 40 mmol in 100 mL normal saline over 1 hr) were administered to patients for serum potassium levels of less than 3.5 but greater than 3.2 mmol/L (n = 26), 3.0 to 3.2 mmol/L (n = 11), and less than 3.0 mmol/L (n = 11), respectively. Serum and urine potassium levels were monitored during and for 1 hr after the infusion.
Measurements and results: All patients tolerated the infusions without evidence of hemodynamic compromise, ECG change, or new dysrhythmia requiring treatment. The mean maximum potassium increase was 0.5 +/- 0.3 mmol/L, 0.9 +/- 0.4 mmol/L, and 1.1 +/- 0.4 mmol/L in the 20-, 30-, and 40-mmol groups, respectively. The increase in serum potassium was maximal at the completion of the infusion and was significant (p less than .05) compared with baseline in all groups. Peak potassium levels were the same in patients with normal renal function (n = 33) compared with those with renal insufficiency (n = 15). Urinary excretion of potassium increased in all groups during the infusion and was significant (p less than .05) in the 30- and 40-mmol groups, but was no greater in those patients who had received diuretics (n = 8) compared with those patients who had not (n = 40).
Conclusions: In the select group of hypokalemic patients studied, potassium infusions of 20 to 40 mmol delivered over 1 hr were safe to administer and effectively increased serum potassium levels in a dose-dependent and predictable fashion. Furthermore, these results were independent of the patient's underlying renal function or associated diuretic administration.