To evaluate the reliability of flash continuous glucose monitoring(CGM) and itsaccuracy in supporting clinical decision-making after pancreatic surgery.

This was a single-center retrospective secondary analysis derived from a multicenter prospective cohort study. Adult patients undergoing elective pancreatic surgery at Peking Union Medical College Hospital between March 2024 and March 2025 were enrolled. All patients had a CGM sensor placed on the left upper arm one day before surgery, with monitoring continued until at least postoperative day 7. The sensor measured interstitial fluid glucose every minute and allowed real-time scanning to obtain instant glucose values. In parallel, all patients underwent point-of-care(POC) blood glucose testing postoperatively. Paired glucose measurements at the same time points were identified using the Gplus glucose management system and the hospital information system. The mean absolute relative deviation(MARD) was calculated to assess the degree of deviation between the two methods. Agreement was evaluated using Bland-Altman analysis, and the clinical accuracy of CGM for decision support was assessed using Parkes error grid analysis.

A total of 255 patients were ultimately included, with a median CGM wear time of 276.3(204.8, 314.1) hours. A total of 829 paired CGM and POC glucose values were obtained. The MARD between CGM values and POC glucose measurements was 14.10%. Bland-Altman analysis showed a mean difference of -18.72 mg/dL between the two methods, with 95% limits of agreement ranging from -69.88 to 32.44 mg/dL. The dispersion of differences widened progressively with increasing glucose levels, indicating marked heteroscedasticity. Parkes error grid analysis demonstrated that 99.88% of the paired glucose deviations fell within zones A and B, suggesting high accuracy of CGM in supporting correct clinical decisions.

Flash CGM demonstrates generally satisfactory reliability for postoperative glucose monitoring following pancreatic surgery, with a low risk of harmful clinical decisions attributable to monitoring deviations. However, the measurement bias increases significantly with higher glucose levels, indicating that in the setting of severe hyperglycemia, CGM readings should be clinically calibrated and interpreted with caution in conjunction with POC blood glucose testing.