Performance of the DEXCOM G7 CGM system during and after major surgery

Perioperative hyperglycaemia is associated with adverse surgical outcomes,1 underscoring the need for glucose monitoring to maintain glucose within recommended targets.2 Continuous glucose monitoring (CGM) has garnered increasing interest as a tool to overcome barriers to efficient and effective glycaemic management. Currently, CGM devices are not approved for perioperative use due to concerns regarding interference from diathermy, certain medications and physiological changes like temperature and fluid shifts.3 Nevertheless, recent recommendations from the Endocrine Society endorse CGM use for glycaemic monitoring in hospitalized patients.4 While accumulating evidence supports satisfactory CGM accuracy in general hospital wards and intensive care units,5-7 and various studies have evaluated older-generation CGM systems in the intraoperative setting,8, 9 the performance of newer-generation devices such as the Dexcom G7 during surgery remains unexplored. Here, we provide first insights into Dexcom G7's accuracy during and after major surgery. We performed a post hoc analysis of clinical studies in adults undergoing major surgery (>90 min, complex, invasive procedure)10 at Bern University Hospital, Switzerland. Participants provided written informed consent, and studies were approved by the Ethics Committee Bern (2020–01690, 2022–01054, 2024–01483). Patients were fitted with Dexcom G7 CGM (Dexcom, San Diego, CA, USA) up to a week before surgery. Intraoperative reference glucose measurements (capillary, arterial, central venous and heart lung machine samples) were analysed using Accu-Chek® Inform II meter (Roche Diagnostics GmbH, Mannheim, Germany), ABL800 blood gas analyser (Radiometer, Brønshøj, Denmark) and GEM Premier 5000 blood gas analyser (Werfen GmBH, München, Germany) from anaesthesia induction to emergence. Postoperative capillary measurements used the Accu-Chek® Inform II device. Calibration occurred at anaesthesia induction and, for some sensors, on postoperative day 1 (using Accu-Chek®). Mean absolute relative difference (MARD) was the main accuracy metric. We further calculated bias, 95% limits of agreement, median absolute relative difference, mean absolute difference for values less than 100 mg/dL, the proportion of CGM values within ±15%, 20% and 30% of reference values for glucose levels >100 mg/dL and ±15 mg/dL, ±20 mg/dL, or ±30 mg/dL for reference glucose levels ≤100 mg/dL (%15/15, %20/20, %30/30).11 Error grid analysis assessed clinical reliability using Diabetes Technology Society Error Grid (DTS EG)12 and Clarke Error Grid (CEG).13 Accuracy metrics were calculated for the intraoperative (from induction to end of anaesthesia) and postoperative (until discharge or up to 10 days after surgery) periods. Other outcomes of interest were the intraoperative CGM availability, the CGM-reference value difference at calibration (mmol/L) and device deficiencies (premature sensor failures requiring replacement and adverse device effects). Accuracy was also assessed separately during extracorporeal circulation (ECC) and for postoperatively calibrated versus non-calibrated devices. CGM measurements were linearly interpolated to a 1-min grid (excluding gaps >10 min). Reference-CGM pairs were matched by timestamp, discarding those >5 min apart. All metrics were calculated at the sensor level, and summary statistics appropriate to the distribution are reported. The limits of agreement were estimated using a random effects model with participants included as a random effect, accounting for both between- and within-sensor variability. Participant-level data (number of pairs, proportion of values in DTS EG and CEG zones) are summarized using mean ± SD or median [25th; 75th percentile] in case of deviation from normal distribution. Results are reported as mean ± SD, median [25th; 75th percentile] or n (%), unless specified otherwise. Analysis was performed using R version 4.4.1. A total of 619 paired reference-CGM measurements were collected from 29 individuals (HbA1c 5.9 [5.4; 6.8]%, BMI 27.5 ± 4.9 kg/m2) and used for the present analysis. The patients underwent the following surgeries: open cardiac surgery with ECC (62.1%), abdominal (20.7%), orthopaedic (10.3%) and urologic surgery (3.4%). More details on the clinical context are provided in Table S1. At the induction of anaesthesia, CGM wear time was 0.8 [0.2; 1.8] days, with minimum and maximum wear time of 0.6 h and 6.8 days, respectively. There were 265 intraoperative CGM/blood glucose reference pairs available, with a median time interval between reference values of 22.0 [19.0; 35.0] min (minimum interval 1 minute, maximum interval 229 min). The details of intraoperative reference measurements (frequency of blood compartment and device used) are displayed in the Table S2. Intraoperative MARD was 12.5 [8.7; 15.8]% with a bias of −0.6 [−1.3; 0.3]mmol/L. Computing DTS EG analysis, 66.7 [44.4; 92.9]% of pairs were in Zone A and 0.0 [0.0; 16.7]% in Zone B (Figure 1A). The median bias of individual sensors varied considerably, with the 90% deviation ranges extending from approximately −4.0 mmol/L to 2.0 mmol/L (Figure 1B). Intraoperative CGM data were available continuously (100%). During ECC (lowest body temperature 28.6°C), MARD was 15.5 [9.7; 19.9]% with a bias of −1.7 [−2.1; −0.9]mmol/L (95 data pairs, 89 heart lung machine and 6 arterial reference measurements). A case example of glucose trajectories (CGM and blood) during cardiac surgery with ECC is shown in Figure S1. In the postoperative period, MARD was 9.0 [6.6; 11.1]% (based on 354 data pairs). All intra- and postoperative CGM accuracy metrics are in Table 1. The median absolute deviation of the CGM reading from the reference value at the time of calibration was 0.4 [−0.5; 1.0]mmol/L at anaesthesia induction (n = 29 patients) and 0.1 [−0.1; 0.4]mmol/L on postoperative day 1 (n = 9 patients). CGM devices calibrated on postoperative day 1 showed a postoperative MARD of 6.9 [6.7; 7.7]% (186 data pairs) and the non-calibrated CGM devices yielded a postoperative MARD of 9.9 [7.6; 11.6]% (168 data pairs). Additional CGM accuracy metrics are in Tables S3 and S4. No premature sensor failure or adverse device effects occurred during or after surgery. Full hospitalization CGM metrics are in Table S5. We evaluated the intra- and postoperative performance of the Dexcom G7 CGM, providing early insights into its role in perioperative glucose management. The intraoperative MARD was 12.5% with 100% sensor availability, suggesting reasonable accuracy for surgeries >90 min. Postoperative MARDs were lower for calibrated devices (6.9%) compared to non-calibrated devices (9.9%), indicating greater intraoperative challenges. To our knowledge, this is the first evaluation of Dexcom G7's intraoperative accuracy. While comparable studies on the Dexcom G6 during surgery reported slightly lower sensor availability (98.6% vs. 100%) but a lower mean MARD (12.7%–12.9% vs. 13.8%) in similar patient populations (BMI 27.3–32.5 kg/m2, HbA1c 5.6%–7.0%), no direct comparisons exist for Dexcom G7.9, 14 However, in the current study, sensor wear time at the surgery start was shorter (0.8 vs. 7 days), potentially challenging accuracy due to early phase problems, such as host immune reactions.15 Additionally, over half of the participants underwent cardiac surgery with ECC, a condition known to impair CGM accuracy.8 Notably, the Dexcom G7 performed better in this challenging setting than the Dexcom G6 (mean MARD 17.6% vs. 23.8%).8 Consistent with Dexcom G6 studies, we observed intraoperative CGM under-reading (median bias −0.6 mmol/L), with no sensor failures. While the exact causes remain unclear, potential contributors include diathermy, perfusion changes, pH/temperature fluctuations and chemical interference.16 These findings support current recommendations to verify hypoglycaemia with point-of-care testing and calibrate for relevant deviations (20/20%).6, 17 Limitations of our study include the small sample size, heterogeneous reference measurements (multiple compartments and devices), limited variability in CGM wear time and inability to identify specific factors contributing to inaccuracies. We used hospital point-of-care devices as the reference (rather than gold-standard methods like YSI) and for sporadic calibrations to reflect real-world clinical practice, where CGM would supplement intermittent capillary testing.6 Consequently, our accuracy findings are specific to this clinical application. Additionally, our study design could not assess trend accuracy and within-person sensor variability, which represent important dimensions of CGM performance evaluation.18 While including challenging procedures like cardiac surgery with ECC is a strength, further studies with larger sample sizes and diverse surgeries are needed. Future research should also differentiate results by glycaemic levels and contextualize factors affecting CGM accuracy. In conclusion, our results suggest that the Dexcom G7 CGM system exhibits adequate accuracy with no signal losses during surgery. Further exploration of the new-generation CGM systems is needed to assess their perioperative clinical utility. LB and DH designed the study procedures and supervised the analysis. GK coordinated the study, performed data analysis and wrote the first draft of the article. All authors contributed to the interpretation of the results, critically revised the article and approved the final manuscript. All authors made the decision to submit for publication. We are grateful to study participants for their participation. We also acknowledge support from the clinical research nursing assistants Joana Filipa Rodrigues Cunha Freitas and Vera Zenklusen from the Department of Diabetes, Endocrinology, Nutritional Medicine and Metabolism, Inselspital, Bern University Hospital in the data collection. We also thank Corinne Zahnd and Carlos Biedermann (Department of Anaesthesiology and Pain Medicine, Inselspital, Bern University Hospital, University of Bern) for their contribution to data collection. Dexcom Inc., USA provided product support. This work is supported by the Swiss National Science Foundation grant number 33IC30_201298/1 (24 June 2021). LB reports having received speaker and advisory board honoraria from Dexcom Inc. For the remaining authors, none were declared. LB is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. The peer review history for this article is available at https://www.webofscience.com/api/gateway/wos/peer-review/10.1111/dom.16462. The data that support the findings of this study are available from the corresponding author upon reasonable request. Table S1. Clinical context. Table S2. Intraoperative glucose reference measurements. Table S3. Postoperative CGM accuracy metrics. Table S4. CGM accuracy additional metrics. Table S5. CGM metrics during hospitalization. Figure S1. Case example of glucose excursions during an open cardiac surgery with extracorporeal circulation in a non-diabetic patient. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.