Clinical Characteristics and Development of a Risk-Identification Model for Incidentally Detected Cancer in Patients with Diabetes: A Case–Control Study

Nengi Cheang, Xueru Chen, Hongmei Zhang, Qing Su, Shichun Du Department of Endocrinology, Xin Hua Hospital Affiliated to Shanghai Jiao Tong University School of Medicine, Shanghai, 200092, People’s Republic of ChinaCorrespondence: Shichun Du, Email [email protected] Qing Su, Email [email protected]: In hospitalised adults with diabetes, asymptomatic cancers may be overlooked during admission, delaying diagnosis and potentially contributing to poorer prognosis by missing opportunities for earlier treatment. Although cross-sectional imaging can detect lesions, it is not systematically performed in asymptomatic patients. A risk signal from routine measurements could prompt targeted imaging and enable diagnosis.Methods: We conducted a single-centre retrospective case-control study at Xinhua Hospital. Cases were adults with diabetes with first, pathology-confirmed cancer (any site) detected incidentally during index admission without cancer-related symptoms. Controls were inpatients matched by age and diabetes duration. We developed two models. Model A predictors were age, sex, body mass index, HbA1c, mean glucose, mean amplitude of glycaemic excursions (MAGE) from the first 72 h of capillary glucose, 2-hour C-peptide, apolipoprotein A-I (ApoA-I), albumin, and sodium–glucose cotransporter 2 (SGLT2) inhibitor use. Model B added carcinoembryonic antigen (CEA), carbohydrate antigen 19– 9 (CA19-9), carbohydrate antigen 125 (CA125), and cytokeratin 19 fragment (CYFRA21-1); albumin was not retained. Performance was assessed by area under the receiver operating characteristic curve (AUC), calibration, Brier score, decision curve analysis (DCA), and net reclassification improvement (NRI).Results: Of 281 patients, 140 were cases and 141 were controls. Model A used 219 participants and Model B 156 participants. Higher 2-hour C-peptide and ApoA-I were associated with lower odds of cancer, whereas higher CYFRA21-1 was associated with higher odds. Model B outperformed Model A (AUC 0.740 to 0.852; Brier 0.199 to 0.147).Conclusion: Using inpatient measurements, we developed a risk-identification model for incidentally detected, asymptomatic cancer among hospitalised adults with diabetes, which may help prioritise in-hospital diagnostic work-up. Adding tumour markers, particularly CYFRA21-1, improved discrimination. Prospective multicentre validation and assessment of clinical and economic impact are warranted before implementation.Keywords: diabetes mellitus, incidentally detected cancer, inpatients, glycaemic variability