Q&A: CGMs Unreliable for ICU Insulin Management

Two experts discuss why more research is needed before continuous glucose monitors can be used reliably for insulin management for critically ill patients.

Although factory-calibrated continuous glucose monitors (CGMs) are increasingly used to guide glucose management in critically ill hospitalized patients, their accuracy remains uncertain, according to an industry-supported study published in the Journal of Diabetes Science and Technology.

“Continuous glucose monitors show promise for use in inpatient settings. However, factory-calibrated CGMs may require accuracy validation and additional bedside calibration to ensure optimal performance, particularly in critically ill patients,” wrote lead study author Gautam Ramesh, MD, MPH, of the Children’s Hospital of Philadelphia, and coauthors. “Further trials are needed to refine inpatient CGM protocols and better understand their accuracy in this population.”

Variable Results & Underestimated Glucose

Between November 2019 and December 2021, Dr. Ramesh and colleagues studied 29 male and 11 female adult patients who required continuous intravenous insulin infusion in the medical, surgical, and cardiac intensive care units of one academic medical center. The nonrandomized, open-label, clinically blinded prospective trial compared 2 CGM systems: the Dexcom G6 Pro (G6P; Dexcom, Inc., San Diego, California) and the FreeStyle Libre Pro (FLP; Abbott Diabetes Care, Alameda, California).

The researchers also obtained reference serum (Lab) glucose and point-of-care (POC) glucose values. At the end of the study, they analyzed CGM and reference values to assess CGM accuracy.

They found that 1,015 matched G6P-Lab pairs had a mean absolute relative difference (MARD) of 22.7%; 2,369 G6P-POC pairs had a MARD of 22.9%; 1,006 matched FLP-Lab pairs had a MARD of 25.2%, and 2,353 FLP-POC pairs had a MARD of 27.0%.

Both CGM systems showed marked inter-patient variability in sensor accuracy, and both tended to underestimate glucose levels when compared to reference values. Low reference values were rarely overestimated by either sensor, according to study results.

Expert Perspectives

Jason M. Ng, MD, clinical associate professor of medicine and chair of the Diabetes Task Force at the University of Pittsburgh School of Medicine, and Shichun Bao, MD, PhD, professor of medicine and leader of the Diabetes Technology Program at Vanderbilt University Medical Center, spoke with Physician’s Weekly about CGM use in the ICU.

PW: How might differences in accuracy between 2 CGMs affect critically ill patients? 

Dr. Ng: The differences in accuracy may change the insulin dosing with respect to glucose values based on the reported glucose levels between the two CGMs. For example, if the Dexcom G6 CGM reports a slightly higher glucose level than the Freestyle Libre CGM, the care team may elect to give more insulin based on the Dexcom G6 CGM reading, whereas they may elect not to give more insulin based on the Freestyle Libre CGM reading.

Dr. Bao: Dexcom and Libre CGMs have similar overall accuracy; however, providers and patients should be aware that physiological, sensor, environmental factors, and medication interference can affect CGM accuracy, especially in hospital patients whose situations may be variable.

CGMs are not yet approved by the US Food and Drug Administration for inpatient use; however, a temporary exemption was granted during the COVID-19 pandemic to minimize healthcare provider and staff exposure and conserve PPE [personal protective equipment].

Unless contraindicated, hospitals usually allow patients already using CGMs as outpatients to continue their use in the hospital as an off-label adjunctive therapy. However, CGM data should not be used to adjust insulin therapy in hospitalized patients due to accuracy Issues. Insulin management still requires fingerstick glucose measurement.

Despite their limitations, CGMs are highly effective in offering continuous glucose monitoring and alerts, enabling remote data sharing with healthcare providers. The ability of healthcare providers to review patients’ CGM data remotely improves patients’ glucose control.

What are the potential consequences of CGMs underestimating glucose levels?

Dr. Bao: Caution is needed. CGMs tend to underestimate glucose levels in the first 1 to 2 days. Although this tendency can be protective against hypoglycemia, if CGM readings are used for insulin management, underestimation can lead to undertreatment.

Dr. Ng: Potentially, this means the patient may be getting less insulin than they need to maintain a goal glucose level. Remaining at goal as much as possible is very important in the critically ill setting.

How should CGMs currently be used in the ICU?

Dr. Bao: CGMs should be standard care for close glucose monitoring and to provide alerts that can prevent severe hypoglycemia and hyperglycemia. However, CGMs should not be used for insulin adjustment without accuracy validation and per-patient calibration for inpatient use in the ICU.

Regulatory changes are needed to enable wider CGM adoption for inpatient glucose monitoring, and technical improvements are necessary to enable the use of CGMs to guide insulin therapy in hospitalized diabetes.

Allowing patients to continue personal CGM use in hospitals is crucial but requires awareness of its limitations and caution in CGM data interpretation.

Dr. Ng: I can foresee CGMs becoming standard practice in the future because they can reduce the resources needed in the ICU setting. However, based on this study’s findings, the CGMs will need to show improved precision and accuracy (reduction in MARD). More studies are needed to provide reliable data that is precise for glucose measurements with CGMs in critically ill patients, as glucose levels in these patients can fluctuate rapidly.

Are any strengths or limitations of Dr. Ramesh and colleagues’ study especially worth noting? 

Dr. Ng: That the data were not known to the clinical ICU personnel, and that treatment decisions were not affected, allowing for clean data collection, are strengths. A limitation of the study is the gender distribution of the patients included. It would be good to include more women in future studies.

Dr. Bao: The study’s strength is the side-by-side comparison of two professional CGMs in critically ill patients. However, a limitation is the use of blinded CGMs in the study, which prevented bedside calibration and replacement of inaccurate or defective sensors. Unlike personal CGMs, which are already FDA-approved for outpatient use, blinded CGMs prevent users from knowing whether devices are working correctly until the data are downloaded. This delay in knowing a patient’s glucose data does not aid real-time glucose management.

What are your thoughts on the future of CGM use in the ICU?

Dr. Ng: Although CGM use in the critically ill setting is not ready for clinical practice, it is an exciting area for future research. Studies to further explore this possibility should be encouraged.

Dr. Bao: I’m a big advocate of CGMs. Personal CGMs should be allowed as adjunctive therapy for hospitalized patients, even if not yet approved by the FDA. We should allow patients to continue to use their CGMs in the hospital, but we need to interpret CGM data with caution, understanding their limitations and the factors that can affect their accuracy.