Feasibility and accuracy of continuous glucose monitoring in surgical intensive care unit patients: a single-center pilot study in South Korea

Hyojun Park; Eunmi Gil; Joon Ho Lee; Chi-Min Park

doi:10.4266/acc.004975

Articles

Page Path: HOME > Acute Crit Care > Volume 41(2); 2026 > Article

Original Article Surgery Feasibility and accuracy of continuous glucose monitoring in surgical intensive care unit patients: a single-center pilot study in South Korea: Acute and Critical Care 2026;41(2):378-386.
DOI: https://doi.org/10.4266/acc.004975
Published online: March 27, 2026

Department of Critical Care Medicine, Samsung Medical Center, Sungkyunkwan University College of Medicine, Seoul, Korea

Corresponding author: Chi-Min Park Department of Critical Care Medicine, Samsung Medical Center, Sungkyunkwan University College of Medicine, 81 Irwon-ro, Gangnam-gu, Seoul 06351, Korea Tel: +82-2-3410-1096, Fax: +82-2-2148-7088 Email: dr99.park@samsung.com

• Received: October 21, 2025 • Revised: November 26, 2025 • Accepted: December 31, 2025

This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/4.0/) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.

972 Views
41 Download

prev next

Full Article

Download PDF

Abstract
INTRODUCTION
MATERIALS AND METHODS
RESULTS
DISCUSSION
KEY MESSAGES
NOTES
SUPPLEMENTARY MATERIALS
REFERENCES

Abstract

Background
Continuous glucose monitoring (CGM) technology offers potential advantages over intermittent point-of-care testing in critically ill patients by providing real-time glucose trends and automated alerts. However, its accuracy and feasibility in intensive care settings require validation before widespread implementation.
Methods
We conducted a single-center observational pilot study, evaluating CGM feasibility in 11 surgical intensive care unit (ICU) patients, including nine post–liver transplant recipients. The G6 CGM system was applied for continuous monitoring. CGM readings were paired with point-of-care glucose measurements for accuracy assessment. Performance metrics included the mean absolute relative difference (MARD), bias, standard deviation of relative differences (SDRD), Surveillance Error Grid (SEG) analysis, and International Organization for Standardization (ISO) 15197:2013 criteria compliance.
Results
During a median monitoring period of 5 days (interquartile range [IQR], 3–9), we analyzed 326 paired glucose measurements. CGM demonstrated acceptable accuracy, with a MARD of 13.5% (95% CI, 11.43%–15.76%), bias of 2.79% (95% CI, –2.48 to 7.27%), and SDRD of 18.69% (95% CI, 13.75%–23.65%). SEG analysis confirmed 99.1% of readings were in clinically acceptable zones A and B. ISO 15197:2013 criteria showed 62.9% of measurements were within ±15 mg/dl or ±15%. The median patient-level mean glucose was 199.0 mg/dl (IQR, 162.0–248.0), reflecting substantial hyperglycemic exposure in patients receiving high-dose methylprednisolone despite protocolized insulin therapy targeting a range of 140–180 mg/dl.
Conclusions
CGM was feasible and acceptably accurate in ICU patients. Persistent hyperglycemia despite protocolized care indicates that CGM-derived data may help to identify opportunities for future protocol improvement. Its potential impact on the time-in-target range, hypoglycemia, and clinical outcomes should be evaluated in future multicenter studies.
Key Words: continuous glucose monitoring; critical care; hyperglycemia; hypoglycemia; intensive care unit; liver transplantation

INTRODUCTION

Glycemic dysregulation is common in critically ill patients and is associated with adverse outcomes. The stress response to critical illness triggers increased cortisol, catecholamine, and inflammatory cytokine release, leading to insulin resistance and hyperglycemia even in patients without pre-existing diabetes. Both hyperglycemia and hypoglycemia have been independently linked to increased mortality, infectious complications, and prolonged mechanical ventilation [1,2].

Contemporary guidelines recommend targeting blood glucose levels of 140–180 mg/dl in most intensive care unit (ICU) patients while avoiding hypoglycemia. Evidence from large randomized trials, including the landmark NICE-SUGAR trial, has shown harm with tight glycemic targets (81–108 mg/dl) and highlighted hypoglycemia as a major safety concern [3-5]. These trials revealed that intensive insulin therapy significantly increases the risk of severe hypoglycemia, with many episodes going undetected between scheduled measurements. Traditional point-of-care (POC) testing every 4–6 hours may miss significant glycemic excursions and delay hypoglycemia detection, particularly during nocturnal periods or when nursing workload is high [6]. Continuous glucose monitoring (CGM) offers potential advantages over intermittent POC testing by providing real-time glucose trends and alerts for impending hypoglycemia and hyperglycemia, which may enhance the safety of intensive glycemic management. During the coronavirus disease 2019 pandemic, the U.S. Food and Drug Administration exercised enforcement discretion in April 2020, temporarily authorizing in-hospital CGM use to reduce healthcare worker exposure and conserve personal protective equipment [7]. This policy change catalyzed numerous studies evaluating CGM accuracy and feasibility in ICU environments [8-12]. However, ICU-specific point accuracy and clinical-risk agreement—particularly among post-transplant patients receiving high-dose corticosteroids—remain incompletely characterized. The primary objective of this pilot study was to evaluate the feasibility of implementing real-time CGM in surgical ICU patients.

MATERIALS AND METHODS

The Institutional Review Board of Samsung Medical Center approved the study (No. 2025-10-052). The need for informed consent was waived owing to the observational design and use of de-identified data.

Study Design and Patient Selection

We conducted a single-center observational study in the surgical ICU at Samsung Medical Center (Seoul, Korea). Eligible participants included adults aged ≥18 years who were admitted to the surgical ICU with an anticipated length of stay exceeding 3 days. Patients with an ICU length of stay of <3 days were excluded from analysis (Figure 1). While no formal sample size calculation was performed for this pilot study, our sample of 11 patients aligns with established pilot study frameworks recommending 10–15 participants for feasibility assessment [13]. Despite the small sample size, continuous monitoring generated 326 paired measurements over a median of 5 days, providing sufficient data points for preliminary accuracy estimation with acceptable precision (mean absolute relative difference [MARD] 95% CI width, 4.3%).

Continuous Glucose Monitoring

We used the G6 real-time CGM system (Dexcom, Inc.) for glucose surveillance. For patients selected per the study protocol, trained nursing staff inserted sensors on the posterior upper arm and connected them to either the manufacturer's receiver device or a smartphone application for continuous bedside display. Sensors remained in place for a maximum of 10 days; however, analysis was restricted to data acquired during the ICU admission period only.

Alert thresholds were set at 70 mg/dl (3.9 mmol/L) for hypoglycemia and 250 mg/dl (13.9 mmol/L) for hyperglycemia. Only hypoglycemia alerts triggered an immediate response: bedside POC glucose was verified without delay, followed by protocolized intervention per institutional guidelines. Hyperglycemia alerts were recorded for surveillance and clinician review, without a mandated immediate protocolized action.

Reference Glucose Measurement

POC glucose determinations served as the reference standard for accuracy evaluation. Reference measurements were obtained through two methods: (1) arterial blood samples analyzed on a blood gas analyzer (GEM Premier 5000; Werfen) and (2) capillary or arterial samples measured using bedside glucose meters with test strips. POC glucose values and their timestamps were extracted from the electronic medical record.

Glycemic Management Protocol

Glycemic management followed a standardized intravenous regular-insulin order set targeting 140–180 mg/dl. Blood glucose monitoring was performed every 2–6 hours depending on glycemic status and insulin infusion rate, i.e., every 6 hours for stable patients, every 4 hours when insulin infusion exceeded 5 units/hr, and every 2 hours when blood glucose was outside the target range. Initiation thresholds, titration steps, and hypoglycemia management were protocolized and implemented as part of routine ICU care. Throughout the investigation, CGM served as a supplemental monitoring modality. In patients receiving insulin, dose modifications were directed primarily by POC glucose determinations according to protocol.

Data Collection

All data were retrospectively extracted from the electronic medical record. Baseline characteristics included age, sex, body mass index, comorbidities, diabetes status, and glycated hemoglobin. Clinical parameters included primary diagnosis, surgical procedures, illness severity scores such as the Sequential Organ Failure Assessment (SOFA) score, organ-support interventions (vasopressors, mechanical ventilation, renal replacement therapy), and concurrent medications [14].

CGM data were retrieved from the Dexcom Clarity platform. POC glucose values were extracted from the electronic medical record. Only CGM–POC pairs obtained during the ICU stay were analyzed; values collected outside the ICU or outside active CGM recording were excluded. Pairing used the recorded timestamps from CGM and the EHR; each reference value was matched to the nearest CGM reading within ±5 minutes. Feasibility was assessed by examining sensor adherence and continuity of data capture, which are summarized in Table 1. Sensor wear was maintained for a median of 5 days, and the median active monitoring time was 80%, indicating stable data acquisition with minimal signal interruption during ICU monitoring.

Statistical Analysis

All statistical analyses were performed using R version 4.4.2 (R Foundation for Statistical Computing). Continuous variables were expressed as median (interquartile range [IQR]) values and categorical variables were expressed using a number (percentage). CGM accuracy was assessed using the MARD, bias, and standard deviation (SD) of relative difference (SDRD). Agreement between CGM and reference measurements was evaluated using Bland-Altman analysis, with limits of agreement calculated as the mean difference ±1.96 SDs. Because each patient contributed multiple paired CGM–reference measurements, accuracy metrics—including MARD, relative bias, and SDRD—were calculated at the patient level rather than across pooled pairs, thereby reducing the impact of intra-patient correlation.

Accuracy according to International Organization for Standardization (ISO) 15197:2013 standards was assessed at ±15/15, ±20/20, ±30/30, and ±40/40 mg/dl or % thresholds (±X mg/dl for glucose ≤100 mg/dl or ±X% for glucose >100 mg/dl). Because ISO 15197:2013 was developed for capillary self-monitoring blood glucose systems rather than interstitial CGM devices, ISO thresholds in this study were used solely as supportive reference benchmarks to facilitate comparison with prior CGM feasibility studies, rather than as regulatory performance criteria. Clinical accuracy was evaluated using Clarke Error Grid analysis and additional error grid methods (DTS Error Grid, Surveillance Error Grid [SEG], and Parkes Error Grid) using the SEG risk matrix as originally described; zone proportions (A–E) were computed from paired CGM–reference values. Risk analysis and figure generation were performed via the Diabetes Technology Society online SEG calculator (https://www.diabetestechnology.org/dtseg/), which implements the SEG [15,16]. Continuous accuracy metrics (MARD, relative bias, and SDRD) were summarized at the patient level with cluster-bootstrap 95% CIs, proportions (SEG A/B and ISO 15197:2013 thresholds) were reported as overall pair–level percentages, and Bland-Altman limits were computed as the mean difference ±1.96×SD.

RESULTS

Patient Characteristics

A total of 11 patients were included in this study. The median age was 53 years (IQR, 48–65), and six patients (54.5%) were male. The median body mass index was 27.0 kg/m² (IQR, 24.6–31.9). Nine patients underwent liver transplantation, including seven living-donor liver transplantations and two deceased-donor liver transplantations. Among the liver transplant recipients, the median MELD score was 28.0 points (IQR, 14.0–31.0). The remaining two patients presented with septic shock and underwent emergency surgical interventions: one received small-bowel resection with anastomosis, and the other underwent right hemicolectomy with end ileostomy.

Baseline severity of illness was a median SOFA score of 6 points (IQR, 5–10). Common comorbidities included diabetes mellitus (5 patients, 45.5%), cancer (5 patients, 45.5%), and hypertension (4 patients, 36.4%). Chronic pulmonary disease, ischemic heart disease, heart failure, and stroke/transient ischemic attack were less frequent (1–2 patients each).

All nine liver transplant recipients received basiliximab for induction immunosuppression, along with high-dose methylprednisolone, mycophenolate, and tacrolimus. Regarding critical care interventions, four patients (36.4%) required norepinephrine support, three patients (27.3%) received mechanical ventilation, and two patients (18.2%) underwent continuous renal replacement therapy. The median ICU length of stay was 6.0 days (IQR, 4.0–11.0), and the median total hospital length of stay was 25.0 days (IQR, 17.0–45.0). Detailed baseline characteristics are presented in Table 2.

CGM Metrics

A total of 326 paired glucose measurements were obtained from 11 patients. The median duration of CGM sensor use was 5.0 days (IQR, 3.0–9.0), with a median active time of 80.0% (IQR, 74.0%–89.0%). The median patient-level mean glucose was 199.0 mg/dl (IQR, 162.0–248.0). Glycemic variability was assessed by SD (median, 58.0 mg/dl; IQR, 40.0–69.0) and coefficient of variation (median, 25.8%; IQR, 24.3%–32.2%). The median time in range (70–180 mg/dl) was 47.0% (IQR, 23.0%–64.0%). Detailed CGM metrics are presented in Table 1.

Accuracy Assessment and Alignment with International Standards

The overall accuracy of CGM was evaluated using multiple metrics. The MARD was 13.50% (95% CI, 11.43%–15.76%). The mean bias was 2.79% (95% CI, −2.48% to 7.27%), and the SDRD was 18.69% (95% CI, 13.75%–23.65%). Bland-Altman analysis revealed limits of agreement ranging from −33.85% to 39.43%, with the lower limit of agreement at −33.85% (95% CI, −46.38% to −20.67%) and the upper limit at 39.43% (95% CI, 30.87%–47.87%) (Table 3, Figure 2A).

CGM accuracy was assessed according to ISO 15197 criteria. Of the 326 paired measurements, 205 (62.9%) were within ±15 mg/dl or ±15%, 253 (77.6%) were within ±20 mg/dl or ±20%, 302 (92.6%) were within ±30 mg/dl or ±30%, and 321 (98.5%) were within ±40 mg/dl or ±40% (Table 3).

SEG analysis demonstrated that 272 measurements (83.4%) fell within zone A, while 51 measurements (15.6%) fell within zone B, resulting in 99.1% of all paired measurements being in zones A and B combined. Zone C contained one measurement (0.9%), while no measurements fell within zones D and E or were excluded from analysis (Figure 2B). Similar results were observed across different error grid methods, with zones A and B totaling 99.1%–99.7% of measurements obtained using the DTS Error Grid, SEG, Parkes Error Grid, and Clarke Error Grid (Table 3).

Alarm Performance and Hypoglycemia Safety

During the monitoring period, the CGM system generated hypoglycemia alerts (<70 mg/dl) in three patients. In two of these patients, these alerts corresponded to confirmed hypoglycemic events (blood glucose approximately 60 mg/dl), enabling prompt clinical intervention. In the third patient, the hypoglycemia alert was determined to be a false alarm due to sensor displacement; the sensor was repositioned, and subsequent paired measurements showed appropriate concordance with POC values.

DISCUSSION

This pilot study evaluated the feasibility of CGM in ICU patients. Despite the small sample size of 11 subjects, the CGM system demonstrated acceptable performance, with a MARD of 13.5% and SEG analysis confirming 99.1% of readings were present in clinically acceptable zones A and B. These accuracy metrics establish the feasibility of CGM implementation in this critically ill population. Feasibility was also supported by sustained sensor use and a high proportion of active monitoring time (Table 1), suggesting reliable CGM performance during ICU monitoring. Because two different reference methods were used, we performed a stratified analysis separating arterial blood gas analysis (ABGA)–CGM and point-of-care testing (POCT)–CGM pairs. CGM accuracy was comparable between the two datasets (ABGA MARD 12.3% vs. POCT MARD 14.3%), and the error-grid distributions were also similar (Supplementary Figure 1). These findings suggest that the use of mixed reference values did not materially influence our overall accuracy estimates.

Although CGM generated two hypoglycemia alerts that were confirmed by reference measurements, the very small number of events limits any meaningful assessment of CGM performance in the hypoglycemic range. These episodes are better interpreted as illustrative clinical observations rather than evidence of CGM performance in the hypoglycemic range.

Our results demonstrate comparable accuracy to previously published CGM feasibility studies. Prior investigations in various clinical settings have reported MARD values ranging from 9.4% to 18.0%, which aligns with our observed MARD of 13.5%. Similarly, SEG analyses in previous studies showed 96.8%–99.9% of values falling within clinically acceptable zones, consistent with our finding of 99.1%. Regarding ISO 15197:2013 criteria, our achievement of 62.9% compliance is comparable to findings in other CGM studies, reflecting the inherent challenges of applying blood glucose monitoring standards to interstitial glucose measurements. However, our bias (2.79%) and SDRD (18.69%) values showed greater variability compared to some previous reports, with 95% limits of agreement ranging from −33.85% to 39.43%. This increased variability likely reflects our limited sample size and the physiological differences between interstitial and blood glucose measurements during periods of rapid glucose change [8,11,12,17]. CGM measures interstitial glucose using an enzymatic electrochemical sensor, a principle that naturally introduces a short physiological lag behind blood glucose and may accentuate discrepancies during abrupt fluctuations. In addition, unlike most prior CGM study populations, our cohort largely consisted of post-liver transplant recipients receiving high-dose corticosteroids. This population exhibits unique glycemic patterns—including marked variability and steroid-associated hyperglycemia—that remain underrepresented in existing ICU CGM research. Our findings therefore provide population-specific feasibility and accuracy data that complement and extend current evidence [18].

The mean glucose level of 199 mg/dl observed in our cohort reflects substantial hyperglycemic exposure, consistent with the inclusion of post–liver transplant recipients receiving high-dose methylprednisolone. Corticosteroids are well-documented drivers of inpatient hyperglycemia through increased insulin resistance and enhanced hepatic gluconeogenesis [19]. Notably, despite strict adherence to our institutional insulin protocol, CGM data revealed persistent hyperglycemic exposure throughout the monitoring period. This continuous monitoring capability provided clear insights into the extent of hyperglycemic burden that may not have been fully appreciated through intermittent POCT alone. Post-transplant hyperglycemia has been associated with an increased risk of allograft rejection and infectious complications, highlighting the importance of optimizing glycemic control in this vulnerable population [20]. Given the particular vulnerability of liver transplant recipients receiving high-dose steroids to dysglycemia, CGM may offer additional information that could be useful when considering future adjustments to glycemic-management practices [21].

This pilot study has several important limitations. First, its small sample size (n=11) limits the precision of our accuracy estimates and the overall generalizability of the findings. The single-center design may not reflect practices at other institutions. In addition, because nine of our patients were post–liver transplant recipients receiving high-dose corticosteroids, the results largely reflect a highly specific postoperative cohort with steroid-driven glycemic variability. This distinctive metabolic profile differs from that of general ICU populations, and therefore the accuracy and feasibility estimates should be interpreted within this clinical context. Finally, the limited number of hypoglycemic events prevented a robust evaluation of CGM performance in the hypoglycemic range.

Despite these limitations, CGM provided valuable clinical insights that support its potential utility in critical care settings. The system enhanced awareness of hyperglycemic exposure patterns and enabled real-time hypoglycemia detection. Two instances of hypoglycemia at approximately 60 mg/dl were identified through CGM alerts, allowing for prompt clinical intervention that might have been missed with intermittent monitoring alone. Studies have demonstrated that real-time CGM can significantly reduce severe hypoglycemic episodes [21]. The observation of persistent hyperglycemia in patients receiving high-dose steroids, despite protocolized insulin therapy, suggests that CGM could facilitate protocol refinement and improve glycemic management. Future larger-scale studies are warranted to evaluate whether CGM-guided insulin protocols can improve the time-in-target range (140–180 mg/dl), reduce the hypoglycemia burden, and ultimately improve clinical outcomes in this high-risk population [18,22,23].

In this pilot study, CGM demonstrated acceptable accuracy (MARD, 13.5%) and clinical feasibility in ICU patients, with 99.1% of readings existing in clinically acceptable zones. CGM provided valuable real-time hypoglycemia detection and revealed persistent hyperglycemia in post–liver transplant recipients despite protocolized insulin therapy. These findings support the need for larger studies to evaluate whether CGM-guided insulin protocols can improve glycemic outcomes in critically ill patients.

KEY MESSAGES

▪ In a surgical intensive care unit pilot enriched with liver-transplant recipients on high-dose steroids, real-time continuous glucose monitoring (CGM) showed clinically acceptable accuracy versus point-of-care testing (mean absolute relative difference, 13.5%; Surveillance Error Grid A/B, 99.1%).

▪ Continuous trends data revealed substantial hyperglycemic exposure despite protocolized insulin, suggesting opportunities to refine glycemic protocols and increase the time in range while safeguarding against hypoglycemia.

▪ These feasibility data support the conduct of larger trials to determine whether CGM-guided insulin strategies improve glycemic metrics and patient-centered outcomes in critically ill populations.

NOTES

CONFLICT OF INTEREST

No potential conflict of interest relevant to this article was reported.

FUNDING

None.

ACKNOWLEDGMENTS

None.

AUTHOR CONTRIBUTIONS

Conceptualization: HP. Formal analysis: HP. Investigation: HP, EG. Data curation: JHL. Supervision: CMP. Writing - original draft: HP. Writing - review & editing: CMP, EG, JHL. All authors read and agreed to the published version of the manuscript.

SUPPLEMENTARY MATERIALS

Supplementary materials can be found via https://doi.org/10.4266/acc.004975.

Supplementary Figure 1.

Stratified continuous glucose monitoring accuracy by reference method. (A) Surveillance Error Grid (ABGA-only). (B) Surveillance Error Grid (point-of-care testing [POCT]-only). (C) Summary of stratified accuracy metrics.

acc-004975-Supplementary-Figure-1.pdf

Figure 1.

Flowchart of this study.

Figure 2.

Accuracy assessment of glucose measurements. (A) Bland-Altman plot of relative differences (%), showing mean bias (2.79%) and 95% limits of agreement. Different symbols represent individual patients (n=11). (B) Surveillance Error Grid (SEG) analysis with International Organization for Standardization (ISO) 15197:2013 criteria. Color zones indicate clinical risk levels A–E. Dotted lines represent ISO 15197:2013 error limits (relative error bands of ±15%, ±20%, ±30%, and ±40%; absolute mg/dL limits applied for glucose levels <100 mg/dl). SD: standard deviation.

Table 1.

Continuous glucose monitoring metrics

CGM metric	Value (n=11)
Patient-level mean glucose (mg/dl)	199.0 (162.0–248.0)
Sensor use (day)	5.0 (3.0–9.0)
Active time (%)	80.0 (74.0–89.0)
Standard deviation (mg/dl)	58.0 (40.0–69.0)
Coefficient of variation (%)	25.8 (24.3–32.2)
Time in range (70–180 mg/dl, %)	47.0 (23.0–64.0)

Values are presented as median (interquartile range).

CGM: continuous glucose monitoring.

Table 2.

Baseline characteristics of study population

Characteristics	Value (n=11)
Age (yr)	53 (48–65)
Sex
Male	6 (54.5)
Female	5 (45.5)
Body mass index (kg/m²)	27.0 (24.6–31.9)
SOFA score	6.0 (5.0–10.0)
MELD score^a)	28.0 (14.0–31.0)
Comorbidity
Diabetes mellitus	5 (45.5)
Hypertension	4 (36.4)
Cancer	5 (45.5)
Chronic pulmonary disease	1 (9.1)
Ischemic heart disease	1 (9.1)
Heart failure	1 (9.1)
Stroke/transient ischemic attack	2 (18.2)
ICU length of stay (day)	6.0 (4.0–11.0)
Hospital length of stay (day)	25.0 (17.0–45.0)
Norepinephrine use	4 (36.4)
Mechanical ventilation	3 (27.3)
Continuous renal replacement therapy	2 (18.2)
Acetaminophen use	4 (36.4)
Induction immunosuppression (Basiliximab)	9 (81.8)
Methylprednisolone use	9 (81.8)

Values are presented as median (interquartile range) or number (%).

SOFA: Sequential Organ Failure Assessment; MELD: model for end-stage liver disease; ICU: intensive care unit.

^a)MELD score is reported only for the liver transplant subset.

Table 3.

Accuracy metrics of continuous glucose monitoring

Parameter	Value	95% CI
Sample size (n)
Patients	11	-
Total pairs	326	-
Accuracy metrics
Bias (%)	2.79	−2.48 to 7.27
MARD (%)	13.50	11.43 to 15.76
SDRD (%)	18.69	13.75 to 23.65
Bland-Altman analysis
LoA (%, range)	−33.85 to 39.43	-
LoA lower limit (%)	−33.85	−46.38 to −20.67
LoA upper limit (%)	39.43	30.87 to 47.87
ISO 15197:2013 criteria (No. %)
Within ±15 mg/dl or ±15%	205 (62.9)	-
Within ±20 mg/dl or ±20%	253 (77.6)	-
Within ±30 mg/dl or ±30%	302 (92.6)	-
Within ±40 mg/dl or ±40%	321 (98.5)	-
Error grid analysis, zone A+B (%)
Surveillance Error Grid	99.1
Clarke Error Grid	99.1	-
Parkes Error Grid	99.7	-
DTS Error Grid	99.7	-

MARD: mean absolute relative difference; SDRD: standard deviation of relative difference; LoA: limits of agreement; ISO: International Organization for Standardization; DTS: Diabetes Technology Society.

REFERENCES

1. Finfer S, Liu B, Chittock DR, Norton R, Myburgh JA, McArthur C, et al. Hypoglycemia and risk of death in critically ill patients. N Engl J Med 2012;367:1108-18.Article PubMed
2. Evans L, Rhodes A, Alhazzani W, Antonelli M, Coopersmith CM, French C, et al. Surviving Sepsis Campaign: international guidelines for management of sepsis and septic shock 2021. Intensive Care Med 2021;47:1181-247.Article PubMed PMC
3. van den Berghe G, Wouters P, Weekers F, Verwaest C, Bruyninckx F, Schetz M, et al. Intensive insulin therapy in critically ill patients. N Engl J Med 2001;345:1359-67.Article PubMed
4. Van den Berghe G, Wilmer A, Hermans G, Meersseman W, Wouters PJ, Milants I, et al. Intensive insulin therapy in the medical ICU. N Engl J Med 2006;354:449-61.Article PubMed
5. Finfer S, Chittock DR, Su SY, Blair D, Foster D, Dhingra V, et al. Intensive versus conventional glucose control in critically ill patients. N Engl J Med 2009;360:1283-97.Article PubMed
6. Incidence of hypoglycemia in hyperkalemia patients after treatment with insulin and dextrose in the emergency department of a tertiary care hospital in India: a prospective observational study. Acute Crit Care 2024;39:499-506.Article PubMed PMC PDF
7. U.S. Food and Drug Administration. Enforcement policy for non-invasive remote monitoring devices used to support patient monitoring during the coronavirus disease 2019 (COVID-19) public health emergency (revised). U.S. Food and Drug Administration; 2020.
8. Voglová Hagerf B, Protus M, Nemetova L, Mraz M, Kieslichova E, Uchytilova E, et al. Accuracy and feasibility of real-time continuous glucose monitoring in critically ill patients after abdominal surgery and solid organ transplantation. Diabetes Care 2024;47:956-63.Article PubMed PMC PDF
9. Agarwal S, Mathew J, Davis GM, Shephardson A, Levine A, Louard R, et al. Continuous glucose monitoring in the intensive care unit during the COVID-19 pandemic. Diabetes Care 2021;44:847-9.Article PubMed PDF
10. Guerrero-Arroyo L, Faulds E, Perez-Guzman MC, Davis GM, Dungan K, Pasquel FJ, et al. Continuous glucose monitoring in the intensive care unit. J Diabetes Sci Technol 2023;17:667-78.Article PubMed PMC PDF
11. Friman O, Soltani N, Lind M, Zetterqvist P, Balintescu A, Perner A, et al. Performance of subcutaneous continuous glucose monitoring in adult critically ill patients receiving vasopressor therapy. Diabetes Technol Ther 2024;26:763-72.Article PubMed
12. Lee JH, Kim K, Jo YH, Rhee JE, Lee JC, Kim KS, et al. Feasibility of continuous glucose monitoring in critically ill emergency department patients. J Emerg Med 2012;43:251-7.Article PubMed
13. Hertzog MA. Considerations in determining sample size for pilot studies. Res Nurs Health 2008;31:180-91.Article PubMed
14. Vincent JL, Moreno R, Takala J, Willatts S, De Mendonça A, Bruining H, et al. The SOFA (Sepsis-related Organ Failure Assessment) score to describe organ dysfunction/failure. On behalf of the Working Group on Sepsis-related Problems of the European Society of Intensive Care Medicine. Intensive Care Med 1996;22:707-10.Article PubMed PDF
15. Kovatchev BP, Wakeman CA, Breton MD, Kost GJ, Louie RF, Tran NK, et al. Computing the surveillance error grid analysis: procedure and examples. J Diabetes Sci Technol 2014;8:673-84.Article PubMed PMC PDF
16. Klonoff DC, Lias C, Vigersky R, Clarke W, Parkes JL, Sacks DB, et al. The surveillance error grid. J Diabetes Sci Technol 2014;8:658-72.Article PubMed PMC PDF
17. Huang W, Li S, Lu J, Shen Y, Wang Y, Wang Y, et al. Accuracy of the intermittently scanned continuous glucose monitoring system in critically ill patients: a prospective, multicenter, observational study. Endocrine 2022;78:470-5.Article PubMed PMC PDF
18. Nielsen CG, Grigonyte-Daraskeviciene M, Olsen MT, Møller MH, Nørgaard K, Perner A, et al. Accuracy of continuous glucose monitoring systems in intensive care unit patients: a scoping review. Intensive Care Med 2024;50:2005-18.Article PubMed PDF
19. Peláez-Jaramillo MJ, Cárdenas-Mojica AA, Gaete PV, Mendivil CO. Post-liver transplantation diabetes mellitus: a review of relevance and approach to treatment. Diabetes Ther 2018;9:521-43.Article PubMed PMC PDF
20. Wallia A, Parikh ND, Molitch ME, Mahler E, Tian L, Huang JJ, et al. Posttransplant hyperglycemia is associated with increased risk of liver allograft rejection. Transplantation 2010;89:222-6.Article PubMed PMC
21. Holzinger U, Warszawska J, Kitzberger R, Wewalka M, Miehsler W, Herkner H, et al. Real-time continuous glucose monitoring in critically ill patients: a prospective randomized trial. Diabetes Care 2010;33:467-72.Article PubMed
22. American Diabetes Association Professional Practice Committee. 16. Diabetes care in the hospital: standards of care in diabetes-2025. Diabetes Care 2025;48:S321-34.Article PubMed PMC
23. Honarmand K, Sirimaturos M, Hirshberg EL, Bircher NG, Agus MS, Carpenter DL, et al. Society of Critical Care Medicine guidelines on glycemic control for critically ill children and adults 2024. Crit Care Med 2024;52:e161-81.Article PubMed

Figure & Data

References

Citations

Citations to this article as recorded by

PubReader
ePub Link

Cite

CITE: export Copy Format; Close

Download Citation

Download a citation file in RIS format that can be imported by all major citation management software, including EndNote, ProCite, RefWorks, and Reference Manager.

Format:

RIS — For EndNote, ProCite, RefWorks, and most other reference management software
BibTeX — For JabRef, BibDesk, and other BibTeX-specific software

Include:

Citation for the content below
Citation and abstract for the content below

Feasibility and accuracy of continuous glucose monitoring in surgical intensive care unit patients: a single-center pilot study in South Korea

Acute Crit Care. 2026;41(2):378-386. Published online March 27, 2026

DOI: https://doi.org/10.4266/acc.004975

XML Download

Figure

Related articles

Trends and management of acute respiratory failure in hospitalized patients: a multicenter retrospective study in South Korea

Feasibility and accuracy of continuous glucose monitoring in surgical intensive care unit patients: a single-center pilot study in South Korea

Figure 1. Flowchart of this study.

Figure 2. Accuracy assessment of glucose measurements. (A) Bland-Altman plot of relative differences (%), showing mean bias (2.79%) and 95% limits of agreement. Different symbols represent individual patients (n=11). (B) Surveillance Error Grid (SEG) analysis with International Organization for Standardization (ISO) 15197:2013 criteria. Color zones indicate clinical risk levels A–E. Dotted lines represent ISO 15197:2013 error limits (relative error bands of ±15%, ±20%, ±30%, and ±40%; absolute mg/dL limits applied for glucose levels <100 mg/dl). SD: standard deviation.

Figure 1.

Figure 2.

Feasibility and accuracy of continuous glucose monitoring in surgical intensive care unit patients: a single-center pilot study in South Korea

CGM metric	Value (n=11)
Patient-level mean glucose (mg/dl)	199.0 (162.0–248.0)
Sensor use (day)	5.0 (3.0–9.0)
Active time (%)	80.0 (74.0–89.0)
Standard deviation (mg/dl)	58.0 (40.0–69.0)
Coefficient of variation (%)	25.8 (24.3–32.2)
Time in range (70–180 mg/dl, %)	47.0 (23.0–64.0)

Characteristics	Value (n=11)
Age (yr)	53 (48–65)
Sex
Male	6 (54.5)
Female	5 (45.5)
Body mass index (kg/m²)	27.0 (24.6–31.9)
SOFA score	6.0 (5.0–10.0)
MELD score^a)	28.0 (14.0–31.0)
Comorbidity
Diabetes mellitus	5 (45.5)
Hypertension	4 (36.4)
Cancer	5 (45.5)
Chronic pulmonary disease	1 (9.1)
Ischemic heart disease	1 (9.1)
Heart failure	1 (9.1)
Stroke/transient ischemic attack	2 (18.2)
ICU length of stay (day)	6.0 (4.0–11.0)
Hospital length of stay (day)	25.0 (17.0–45.0)
Norepinephrine use	4 (36.4)
Mechanical ventilation	3 (27.3)
Continuous renal replacement therapy	2 (18.2)
Acetaminophen use	4 (36.4)
Induction immunosuppression (Basiliximab)	9 (81.8)
Methylprednisolone use	9 (81.8)

Parameter	Value	95% CI
Sample size (n)
Patients	11	-
Total pairs	326	-
Accuracy metrics
Bias (%)	2.79	−2.48 to 7.27
MARD (%)	13.50	11.43 to 15.76
SDRD (%)	18.69	13.75 to 23.65
Bland-Altman analysis
LoA (%, range)	−33.85 to 39.43	-
LoA lower limit (%)	−33.85	−46.38 to −20.67
LoA upper limit (%)	39.43	30.87 to 47.87
ISO 15197:2013 criteria (No. %)
Within ±15 mg/dl or ±15%	205 (62.9)	-
Within ±20 mg/dl or ±20%	253 (77.6)	-
Within ±30 mg/dl or ±30%	302 (92.6)	-
Within ±40 mg/dl or ±40%	321 (98.5)	-
Error grid analysis, zone A+B (%)
Surveillance Error Grid	99.1
Clarke Error Grid	99.1	-
Parkes Error Grid	99.7	-
DTS Error Grid	99.7	-

Table 1. Continuous glucose monitoring metrics

Values are presented as median (interquartile range).

CGM: continuous glucose monitoring.

Table 2. Baseline characteristics of study population

Values are presented as median (interquartile range) or number (%).

SOFA: Sequential Organ Failure Assessment; MELD: model for end-stage liver disease; ICU: intensive care unit.