Predicting hypoglycemia after treatment of hyperkalemia with insulin and glucose (Glu-K60 score)

Abstract

Background: Hyperkalemia can lead to fatal cardiac arrhythmias. Ten units of intravenous (IV) regular insulin with 25 g of glucose is the mainstay for treating hyperkalemia. However, the most important complication of this treatment is hypoglycemia. We aimed to develop a scoring model to predict hypoglycemia after the treatment of hyperkalemia. Methods: A retrospective study was conducted at a university-based hospital between January 2013 and June 2021. We included the hyperkalemic patients (> 5.3 mmol/L) who were ≥ 18 years old and treated with 10 units of IV regular insulin with 25 g of glucose. Incomplete data on posttreatment blood glucose, pregnancy, and diabetes mellitus were excluded. Endpoint was posttreatment hypoglycemia (≤ 70 mg/dL or ≤ 3.9 mmol/L). Multivariable logistic regression was used to establish a full model and a subsequently reduced model using the backward elimination method. We demonstrated the model performance using the area under the receiver operating characteristic curve (AuROC), calibration plot, and Hosmer–Lemeshow goodness-of-fit test. Internal validation was done with a bootstrap sampling procedure with 1000 replicates. Model optimism was estimated. Results: Three hundred and eighty-five patients were included, with 97 posttreatment hypoglycemia (25.2%). The predictive model comprised the following three criteria: age > 60 years old, pretreatment blood glucose ≤ 100 mg/dL (≤ 5.6 mmol/L), and pretreatment potassium > 6 mmol/L. The AuROC of this model was 0.671 (95% confidence interval [CI] 0.608 to 0.735). The calibration plot demonstrated consistency with the original data. Hosmer–Lemeshow goodness-of-fit test showed no evidence of lack-of-fit (p 0.792); therefore, the model was also fit to the original data. Internal validation via bootstrap sampling showed a consistent AuROC of 0.670 (95% CI 0.660 to 0.670) with minimal model optimism. A high risk for posttreatment hypoglycemia was indicated if the patient met at least one of those criteria. Sensitivity and specificity were 95.9% and 14.9%, respectively. Conclusion: High risk was indicated when at least one of the criteria was met: age > 60 years old, pretreatment blood glucose ≤ 100 mg/dL (≤ 5.6 mmol/L), and pretreatment potassium > 6 mmol/L. Blood glucose levels should frequently check in the high-risk group. Trial registration: TCTR20210225002 (www.thaiclinicaltrials.org).