A novel diagnostic model for insulinoma

Abstract

The aim is to describe a simple and feasible model for the diagnosis of insulinoma. This retrospective study enrolled 37 patients with insulinoma and 44 patients with hypoglycemia not due to insulinoma at the First Affiliated Hospital of Guangxi Medical University. General demographic and clinical characteristics; hemoglobin A1c (HbA1c), insulin and C-peptide concentrations; and the results of 2-h oral glucose tolerance tests (OGTT) were recorded, and a logistic regression model predictive of insulinoma was determined. Body mass index (BMI), HbA1c concentration, 0-h C-peptide concentration, and 0-h and 1-h plasma glucose concentrations (P < 0.05 each) were independently associated with insulinoma. A regression prediction model was established through multivariate logistics regression analysis: Logit p = 7.399+(0.310 × BMI) - (1.851 × HbA1c) - (1.467 × 0-h plasma glucose) + (1.963 × 0-h C-peptide) - (0.612 × 1-h plasma glucose). Using this index to draw a receiver operating characteristic (ROC) curve, the area under the curve (AUC) was found to be 0.957. The optimal cut-off value was - 0.17, which had a sensitivity of 89.2% and a specificity of 86.4%. Logit P ≥ - 0.17 can be used as a diagnostic marker for predicting insulinoma in patients with hypoglycemia.

Keywords: Diagnostic predictive model; Hypoglycemia; Insulinoma.

Fig. 1

Flowchart of the sample selection

Fig. 2

Diagnostic efficacy of the Logit P model, Fajans’ index, Turner’s index and verification of two models. Fajans’ index = immunoreactive insulin/glucose; Turner’s index = insulin * 100/(glucose − 30). Liao’s model = 8.305–0.441 * insulin 2 h/0 h ratio − 1.679 * C-peptide 1 h/0 h ratio. FPG*HBA1C index = FPG*HBA1C. PPV, positive predictive value; NPV, negative predictive value; AUC, area under the curve; CI, confidence interval. ^#: Validated model