Development and validation of a scoring system for prediction of insulin requirement for optimal control of blood glucose during glucocorticoid treatments

Abstract

AimsWe have developed and validated a novel scoring system to predict insulin requirement for optimal control of blood glucose during glucocorticoid (GC) treatments, by retrospective analyses of clinical parameters before GC treatment. MethodsThree hundred-three adults (the Developing set) undergoing their first treatment of prednisolone (PSL) were divided into two groups, depending on treatment with or without insulin. Independent risk factors for insulin requirement were identified by a stepwise logistic regression analysis after univariate analyses between the two groups. We constructed a point-addition scoring system consisting of several categories and their coefficients in each risk factor derived from another logistic regression analysis. We validated it to two validation sets, A and B. ResultsMale, higher levels of fasting plasma glucose (FPG), HbA1c, and serum creatinine (CRE) and a higher initial dose of PSL were identified as the risk factors. The sensitivity, specificity, and accuracy were 90.0%, 88.1%, and 88.4%; 87.5%, 66.7%, and 70.5%; 83.3%, 76.1%, and 76.6% in the Developing set, Validation set A, and Validation set B, respectively, when the scoring system was applied. ConclusionsThe scoring system is a valid and reliable tool to predict insulin requirements in advance during GC treatment.