Statistical Analysis of Glucose and Insulin Responses to Intravenous Tolbutamide: Evaluation of Hypoglycemic and Hyperinsulinemic States

Abstract

Computer-programmed multivariate statistical methods were applied to analyze data from 474 intravenous tolbutamide tolerance tests (TTT). A covariable was introduced to adjust for the effect of obesity on each of the variables. An initial 270 patients seen from 1964 to 1971 were classified into 13 clinical groups, including normals, insulinoma, chemical diabetes, reactive hypoglycemia, myotonic dystrophy, postgastrectomy hypoglycemia, and hyperparathyroidism. Multivariate analysis of variance was employed to statistically compare TTT data between each group. Discriminant analysis was used to calculate probabilities that an individual test result fitted each of the 13 clinically defined groups; thus it was possible to “interpret” the results of each of the TTTs from the subsequent 204 patients seen from 1971 to 1974 by calculating probabilities for each of the more recent patients fitting the 13 groups defined on past experience. Patients with reactive hypoglycemia displayed greater insulin secretions and impaired glucose disposal compared with normal individuals. By multivariate analysis of variance, the insulinoma and normal groups were significantly different from all other groups, including reactive hypoglycemia. Of eight patients with proven insulinoma, all except one was assigned maximal probability by discriminant analysis for possessing an insulinoma, and this single “false negative” might have been diagnosed had the method been designed to also consider from 60 to 180 min. One hundred of the 204 newer patients were undergoing evaluation for symptoms of hypoglycemia and were not found to have islet cell tumors. None of these 100 patients with symptoms of reactive hypoglycemia had any probability for their TTT data fitting the insulinoma group. Of the other 102 newer patients who did not have symptoms of reactive hypoglycemia, there were two presumed “false positives” for insulinoma, one normal member of a family with multiple endocrine adenoma, type I, and the other with parathyroid adenoma. Of the 21 variables used in the multivariate data analysis, one variable, the insulin value at 60 min, contributed most to the discrimination between insulinoma and other groups. A covariable-adjusted 60-min insulin greater than 41 indicated a cause of hyperinsulinism other than reactive hypoglycemia. The full 21-variable analysis was necessary to differentiate insulinoma from the other hyperinsulinemic states, myotonic dystrophy, hyperparathyroidism, and chemical diabetes. Criteria are also presented for differentiating insulinoma patients on the basis of fasting glucose and insulin. These results indicate that (1) patients with symptoms of reactive hypoglycemia possess abnormal insulin and glucose responses to intravenous tolbutamide; (2) the TTT is a highly accurate diagnostic test for insulinoma, differentiating insulinoma patients from other patients presenting with symptoms of hypoglycemia primarily on the basis of a prolonged hyperinsulinemic response; (3) other hyperinsulinemic states, including hyperparathyroidism, myotonic dystrophy, and chemical diabetes, can be differentiated from insulinoma using multivariate analysis of TTT data.