Development of a novel algorithm for detecting glucocorticoid-induced diabetes mellitus using a medical information...

Area of Science:

• Endocrinology
• Pharmacoepidemiology
• Medical Informatics

Background:

• Glucocorticoid-induced diabetes mellitus (GIDM) poses risks for diabetes mellitus (DM)-related complications.
• Current diagnostic criteria for GIDM are not well-established, making clinical detection challenging.
• Medical Information Databases (MIDs) offer potential for post-marketing surveillance (PMS) studies.

Purpose of the Study:

• To develop and validate an algorithm for detecting GIDM using MIDs.
• To identify reliable screening factors for GIDM detection.
• To improve the diagnostic accuracy for GIDM.

Main Methods:

• A pharmacoepidemiological study was conducted using MID data from 1214 inpatients newly prescribed prednisolone.
• GIDM screening initially used fasting blood glucose (FBG) and HbA1c levels against Japan Diabetes Society (JDS) criteria.
• Expert review of medical records validated GIDM diagnoses, and receiver operating characteristics analysis identified screening factors.

Main Results:

• The JDS criteria identified 63 patients, with expert review confirming 33 as definite GIDM cases (positive predictive value 52.4%).
• Patients with GIDM showed significantly higher maximum HbA1c and relative increase in maximum HbA1c (RIM-HbA1c) compared to non-GIDM patients.
• A 13% increase in RIM-HbA1c effectively distinguished between GIDM and non-GIDM patients, outperforming FBG levels alone.

Conclusions:

• Patients with GIDM exhibit significantly elevated RIM-HbA1c levels.
• An algorithm incorporating RIM-HbA1c into MIDs offers superior sensitivity and specificity for GIDM detection compared to existing JDS criteria.
• Monitoring HbA1c changes is crucial for effective GIDM detection and may enhance current type 2 DM diagnostic standards.