C‐peptide index at 2h post‐meal is a useful predictor of endogenous insulin secretory capacity and withdrawal from insulin therapy in subjects with type 2 diabetes

• Published in: Diabetes, obesity & metabolism Vol. 26; no. 7; pp. 2761 - 2773
• Main Authors: Iwamoto, Yuichiro, Kimura, Tomohiko, Shimoda, Masashi, Morimoto, Yuichi, Watanabe, Yuki, Itoh, Takashi, Sasaki, Taku, Mori, Shigehito, Kubo, Masato, Takenouchi, Haruka, Kazunori, Dan, Iwamoto, Hideyuki, Sanada, Junpei, Fushimi, Yoshiro, Katakura, Yukino, Nakanishi, Shuhei, Tomoatsu Mune, Kaku, Kohei, Kaneto, Hideaki
• Format: Journal Article
• Language: English
• Published: Oxford Wiley Subscription Services, Inc 01.07.2024
• Subjects: Alanine transaminase; Creatinine; Diabetes; Diabetes mellitus (non-insulin dependent); Fasting; Hemoglobin; Insulin; Learning algorithms; Machine learning; Patients; Peptides
• ISSN: 1462-8902; 1463-1326
• DOI: 10.1111/dom.15595

AimsTo evaluate the correlation between C‐peptide index (CPI) at 2 h post‐meal and endogenous insulin secretory capacity and to develop clinical models to predict the possibility of withdrawal from insulin therapy in patients with type 2 diabetes.MethodThis was a single‐centre retrospective study of patients with type 2 diabetes admitted to our hospital. Patients were divided into a withdrawal group (n = 72) and a non‐withdrawal group (n = 75) based on whether they were able to withdraw from insulin therapy at discharge, and the correlation between CPI at 2 h after meal and diabetes‐related parameters was evaluated. In addition, we created two clinical models to predict the possibility of withdrawal from insulin therapy using machine learning.ResultsThe glycated haemoglobin values of the study participants were 87.8 ± 22.6 mmol/mo. The CPI at 2 h post‐meal was 1.93 ± 1.28 in the non‐withdrawal group and 2.97 ± 2.07 in the withdrawal group (p < 0.001). CPI at 2 h post‐meal was an independent predictor of withdrawal from insulin therapy. In addition, CPI at 2 h post‐meal was a better predictor than fasting CPI. Six factors associated with insulin therapy withdrawal (age, duration of diabetes, creatinine, alanine aminotransferase, insulin therapy until hospitalization, and CPI at 2 h post‐meal) were used to generate two clinical models by machine learning. The accuracy of the generated clinical models ranged from 78.3% to 82.6%.ConclusionThe CPI at 2 h post‐meal is a clinically useful measure of endogenous insulin secretory capacity under non‐fasting conditions.