Modified Cleveland Clinic insulin infusion protocol versus physician-directed intravenous insulin adjustment after cardiac surgery: a randomized controlled trial (NUSULIN Trial)

Introduction

Postoperative hyperglycemia is a common metabolic derangement after cardiac surgery and is associated with increased risks of complications and mortality (1). This disturbance typically manifests early, particularly during the first 48–72 hours after cardiac surgery (2-4). Initial reports demonstrated that postoperative hyperglycemia in diabetic patients (plasma glucose >200 mg/dL) in the first two postoperative days was associated with a higher incidence of deep sternal wound infection (5). Early evidence from a surgical intensive care unit (ICU) suggested that intensive insulin therapy targeting blood glucose concentrations of 80–110 mg/dL could improve clinical outcomes, leading to widespread interest in tight glycemic control (6). However, subsequent studies did not consistently reproduce these benefits. In particular, the NICE-SUGAR trial demonstrated that intensive glucose control targeting 81–108 mg/dL increased mortality and severe hypoglycemia compared with a target of ≤180 mg/dL (7). Consequently, contemporary recommendations generally favor a more moderate target of 140–180 mg/dL for critically ill adults. This target seems to reduce the risks of sternal wound infection, atrial fibrillation, recurrent ischemic events, mortality, as well as shortening the duration of ICU and hospital stays (1,7). The Society of Thoracic Surgeons (STS) guidelines for blood sugar levels during postoperative cardiac surgery recommend a level of <180 mg/dL, and <150 mg/dL if the ICU stay is expected to exceed 3 days. Therefore, achieving glycemic control within the optimal target range during this period is clinically meaningful and can substantially influence outcomes (8-10).

Although intravenous insulin permits rapid dose titration, its effectiveness depends on timely blood glucose measurement and consistent adjustment of the infusion rate. Under conventional physician-directed care, insulin changes may vary among clinicians and may be delayed while awaiting assessment and prescription, particularly during periods of high clinical workload. Such inconsistency may prolong the time required to reach the target range and contribute to hyperglycemia, hypoglycemia, and glycemic variability (GV). A standardized insulin infusion protocol allows trained nurses to adjust insulin promptly according to predefined criteria, potentially improving the consistency and efficiency of postoperative glycemic management (11-13). Evidence regarding standardized insulin infusion protocols in Thai cardiac surgical ICUs remains limited. Existing Thai cardiac-surgery research has primarily described perioperative glucose patterns and GV rather than randomized comparisons of insulin infusion protocol and physician-directed insulin adjustment (14). In addition, ethnic variation has been reported in the relationship between insulin sensitivity and pancreatic β-cell insulin response, providing a rationale for evaluating Western-derived insulin protocols in Asian clinical populations (15-17). These physiological differences, together with differences in ICU staffing, workflow, and available resources, may affect insulin requirements, protocol performance, and the risk of hypoglycemia. Therefore, protocols developed in Western populations should be evaluated in the local clinical setting before broader implementation.

The original Cleveland Clinic cardiovascular ICU insulin protocol, which was reported by Olansky et al. is a table-based algorithm that determines insulin adjustments according to both the current blood glucose concentration and its change from the preceding measurement. Glucose is categorized as decreasing, stable, or increasing according to a change of approximately 30 mg/dL, and the insulin infusion is continued, reduced, increased, or withheld accordingly (18). The original protocol targeted 80–150 mg/dL on the day of surgery and 80–120 mg/dL thereafter, with specific actions for glucose concentrations ≤70 mg/dL. For the present study, we modified the protocol by using a single postoperative target range of 80–150 mg/dL, defining hypoglycemia as blood glucose <70 mg/dL, and incorporating earlier reduction or interruption of insulin when glucose approached the lower limit. These modifications were intended to provide a greater safety margin against hypoglycemia while maintaining timely and standardized insulin titration. We therefore conducted a randomized trial comparing the efficacy and safety of the modified Cleveland Clinic insulin infusion protocol adjustment with conventional physician-directed insulin adjustment in adults with diabetes after open-cardiac surgery. We hypothesized that the outcome of the insulin infusion protocol would be superior to physician-directed adjustment. We present this article in accordance with the CONSORT reporting checklist (available at https://jeccm.amegroups.com/article/view/10.21037/jeccm-2026-0017/rc).

Methods

Study design and setting

This study was a prospective, randomized, controlled, parallel 1:1, superiority trial, with participant and primary investigator blinding, that was conducted at Naresuan University Hospital, a teaching hospital in Thailand.

The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. This study was approved by the institutional review board of Naresuan University Hospital (NU-IRB, certificate of approval No. 187/2023), and informed consent was obtained from all individual participants. The trial was registered in Thaiclinicaltrials.org (TCTR20230719001). https://www.thaiclinicaltrials.org/export/pdf/TCTR20230719001. The full trial protocol, prespecified statistical methods, individual de-identified data, and statistical code can be shared from the corresponding author upon reasonable request. A separate statistical analysis plan was not prepared.

Participants

All adult patients who underwent cardiac surgery between July 2023 and August 2025 were screened for enrollment. Inclusion criteria were (I) aged 20–90 years; (II) diagnosis of diabetes mellitus (DM); and (III) patients who underwent open cardiac surgery. Exclusion criteria were (I) patients who underwent closed-cardiac surgery; (II) patients who underwent emergency surgery. Screening and enrollment were performed by one of the nursing staff (D.B., S.M., C.P., and K.S.). Consenting for all patients was achieved. All patients were preoperatively evaluated by the anesthesiologists. Intraoperative management was conducted according to standard protocols. After transferring to the ICU postoperatively, patients were randomly assigned to either the control group, which received physician-directed insulin infusion adjustment, or the insulin infusion protocol group (modified Cleveland Clinic insulin infusion protocol) (IIP gr.) by using a block randomization method (Figure 1). Glycemic control in both groups was maintained by intravenous continuous insulin infusion. Regular insulin (Actrapid®) 100 units in 100 mL of normal saline mixture.

Figure 1 Study flow diagram. All randomized participants were analyzed in the groups to which they were originally assigned. The intention-to-treat and per-protocol populations were identical. CABG, coronary artery bypass grafting; IIP, insulin infusion protocol; ITT, intention-to-treat analysis; PP, per-protocol.

Randomization and blinding

The random allocation sequence was generated by P.A. (MD), who was not involved in participant recruitment, enrolment, clinical care, or treatment assignment. The sequence was generated using STATA version 18.0, with a 1:1 allocation ratio and a fixed block size of four. No stratification was applied. P.A. prepared the allocation assignments in sequentially numbered, opaque, sealed envelopes, which were stored in a secure location and were not accessible to the primary investigator (N.P.). After participant enrolment and admission to the postoperative ICU, the next envelope in sequence was retrieved by an independent nurse who was not involved in the participant’s clinical care (D.B., S.M., C.P., or K.S.). Before opening the envelope, the nurse confirmed that it remained intact and sealed, with no evidence of prior opening or tampering. The envelope was then opened, and the treatment allocation was disclosed to the bedside clinical team. Participants were blinded to treatment allocation. They were not informed of their assigned group. Blinding of the principal investigator was maintained by restricting access to the randomization sequence, sealed envelopes, and allocation records. The principal investigator did not open the envelopes, did not participate in bedside insulin adjustment, and was not informed of individual treatment assignments during participant enrolment, data collection, and outcome assessment. Therefore, the allocation was concealed from the patients and the primary investigator (N.P.). Both remained blinded to group assignment throughout the trial. The formal assessment of blinding was not conducted.

Because the insulin-adjustment procedures differed between the two groups, bedside nurses and treating physicians could not be blinded to treatment allocation. The outcome assessor and data analyst (P.A.) was not blinded to treatment allocation.

Intervention and control

During the ICU postoperative period, blood tests were drawn from the arterial line every 1–2 hours. In the control group, decisions regarding frequency of blood tests and insulin infusion rate were made at the physician’s discretion. In the IIP group, the frequency of blood tests and insulin infusion rate were managed according to the modified Cleveland Clinic insulin infusion protocol by the ICU nursing staff (Appendix 1, Table S1). The target blood sugar in both groups was 80–150 mg/dL. Insulin infusion was discontinued if (I) 72 hours after ICU arrival or (II) oral feeding was started. The study observation window extended from the initiation of intravenous insulin infusion until its discontinuation. All glycemic outcomes reported in this study were assessed during this period. The subsequent transition to subcutaneous insulin occurred outside the study observation window and was not included in the outcome assessment or statistical analysis. Participants were followed until discharged.

Nurse training and protocol adherence

Before study initiation, ICU nurses involved in caring for participants assigned to the intervention group received training on the insulin infusion protocol. The training included the target blood glucose range, timing of blood glucose monitoring, interpretation of the insulin-adjustment algorithm, management of hypoglycemia, and documentation of blood glucose values and insulin infusion rates. The ICU nurses practiced on the pilot patients (data were not included in the study) to familiarize themselves with the protocol, and the protocol was available for reference during patient care. Protocol adherence was assessed by comparing documented blood glucose measurements, monitoring intervals, and insulin infusion-rate adjustments with the recommendations specified in the protocol. A protocol deviation was defined as an insulin-dose adjustment or monitoring interval that differed from the protocol without a documented clinical justification. These deviations were prospectively collected and documented in case record forms and reviewed by the research team.

Study endpoints

The primary outcome was the percentage of time blood glucose is in range during insulin infusion (TIR). The TIR was calculated by using the points in range (PIR) formula (19), which is the percentage of total valid readings that fall within the target range (80–150 mg/dL): TIR = (total number of blood glucose readings within target range/total number of blood glucose readings) × 100.

Secondary outcomes were time to target (TTT), which is defined as time to reach blood sugar of <150 mg/dL after insulin infusion started, GV, and adverse events, which included hypoglycemia and severe hypokalemia.

Hypoglycemia was defined as any blood glucose measurement <70 mg/dL during the insulin-infusion period. Blood glucose was routinely measured every 1–2 hours in both study groups. (Interval checks were performed according to the protocol in IIP gr., and at physicians’ discretion in control gr.) Both groups, therefore, were not assessed solely on the basis of clinical suspicion. Severe hypokalemia was defined as any serum potassium concentration <3.0 mmol/L. Serum potassium was measured every 2–6 hours in both study groups (According to protocol in IIP gr., and according to the routine postoperative ICU laboratory schedule in the control gr.), and additionally whenever clinically indicated. Adverse events were assessed prospectively from the initiation of insulin infusion until insulin infusion was discontinued, oral feeding was commenced, or 72 hours after ICU admission, whichever occurred first.

GV indices were calculated separately for each individual participant using all blood glucose measurements (20). GV was assessed using the coefficient variation (CV), mean amplitude of glycemic excursions (MAGE), J-index, high blood glucose index (HBGI), and average daily risk rate (ADRR). The CV was calculated as the standard deviation (SD) of glucose values divided by the mean glucose concentration and expressed as a percentage: CV = (SD/mean glucose) × 100. The J-index was calculated as 0.001 × (mean glucose + SD)². MAGE was defined as the mean absolute difference between consecutive glucose peaks and nadirs for excursions exceeding 1 SD of the patient’s mean glucose concentration, according to the method described by Service et al. (21). HBGI was calculated using the glucose-risk transformation proposed by Kovatchev et al. (22), with only the positive, hyperglycemia-related components of the risk function included in the calculation. ADRR, another index proposed by Kovatchev et al. (23), was calculated from the same transformed risk function and was used to represent the combined daily risk of extreme low and high glucose values.

Each GV indices of participants from the same groups were then analyzed as a continuous variable.

Sample size determination

Sample size calculation was based on two significant outcomes: TIR and TTT. From our pilot study, we found that TIR of the control group =29.4% vs. intervention group =67%, whereas TTT of the control group =7.8 hrs. vs. intervention group =4.4 hrs. At a power =0.80 and alpha =0.05, if we use TIR as the outcome, a sample size of 8 per group allows detection of significant differences, but if TTT was used as the outcome, the calculated sample size was 28 per group would be needed, corresponding to a total planned sample size of 56 participants. After 50 participants had been randomized and completed the study, an unplanned between-group analysis was conducted. Recruitment was discontinued following the observation of statistically significant between-group differences. No statistical adjustment for the unplanned interim analysis was prespecified or applied.

Variables and data collection

The allocation to the patients and the primary investigator (N.P.) was blinded until after the statistical analysis was complete. Demographic data were obtained. The data on types of surgery, duration of cardiopulmonary bypass time (CPB time), and aortic cross-clamp time (CCT) were collected. Postoperative magnitude of inotropic support was collected by using the vasoactive inotropic score (VIS) as a surrogate. Insulin administration data included duration of insulin infusion, TTT, TIR, total insulin dose administered, and number of blood tests recorded. Incidence of hypoglycemia (blood glucose <70 mg/dL), severe hypokalemia (serum K <3.0 mmol/L), numbers of protocol deviations were collected.

Statistical analysis

Data were summarized according to their distribution and type. Continuous variables were assessed for normality using the Shapiro-Wilk test or visual inspection of histograms. Data were expressed as mean ± SD if the data of a variable had a normal distribution, and expressed as median [IQR] if the data did not. Categorical variables were reported as absolute frequencies and percentages.

The primary analysis was conducted according to the intention-to-treat (ITT) principle. All randomized participants were included in the analysis according to their groups. For comparisons between the IIP group and the control group, the independent-sample t-test was employed for normally distributed continuous variables, including age, total cumulative insulin dose, and average insulin infusion rate. The Mann-Whitney U test was used for non-normally distributed continuous variables, including body weight, HbA1c, TTT glucose, time in range (TIR) and MAGE. Categorical variables were analyzed using Fisher’s exact test, given the small sample size (n=25 per group) and low expected cell counts in several comparisons.

To analyze temporal trends in blood glucose levels during the postoperative period, hourly mean values were calculated. To minimize signal noise and highlight underlying glycemic patterns, Savitzky-Golay smoothing (polynomial order 3, window size of 20% of data points) was applied to the time-series data. All statistical tests were two-tailed, and a P value of <0.05 was defined as the threshold for statistical significance.

To account for baseline imbalances between groups, a rank-based ANCOVA (Conover-Iman method) was performed for all glycemic outcomes, adjusting for covariates with standardized mean differences (SMDs) greater than 0.2: body weight, aortic CCT, and HbA1c. Vasoactive-inotropic score (SMD =0.19) did not meet this threshold and was not included as a covariate in the primary adjusted model (Table S2).

All statistical analyses and data visualizations were performed using Stata version 18.0 (StataCorp, College Station, TX, USA) and Python version 3.12 (utilizing pandas, SciPy, and Matplotlib libraries).

AI assistance

The authors utilized ChatGPT solely for grammatical editing and enhancing the readability of this manuscript. Claude (Anthropic) was used to provide limited assistance with statistical programming, including suggesting or troubleshooting code that the authors subsequently implemented and verified. All AI-assisted outputs were independently checked against the raw data. The authors take full responsibility for the accuracy, integrity, and interpretation of the reported results.

Results

Patient characteristics

Initially, 52 patients who had underlying disease of DM underwent cardiac surgery. Two patients were excluded from the study due to non-open cardiac surgery (off-pump coronary artery bypass grafting), hence, 50 patients remained eligible. Of the planned 56 participants, 50 participants were randomized and included in the analysis, with 25 participants in each group. Recruitment was discontinued before achievement of the prespecified sample size following an unplanned analysis of accumulating comparative outcome data showing a clear separation in glycemic outcomes between groups. Demographic data were comparable between IIP and control groups, including age (mean 65.7 vs. 65.3 years; P=0.85), hemoglobin A1c (HbA1c) (median 6.6 % vs. 7.1 %; P=0.23), coronary artery bypass surgery proportion (68% vs. 72%; P=0.92), CPB time (179.4 vs. 170.7 minutes; P=0.39), and aortic CCT (140.8 vs. 127.2; P=0.09). Postoperatively, both groups received similar crystalloid fluids (P=0.17) (Table 1).

Insulin administration data

The primary outcome, TIR, in the IIP group was higher than in the control group (83.3% vs. 58.4%, respectively, P<0.001). One of the secondary outcomes, TTT, the IIP group reached target faster than the control group {median 2.0 hrs [interquartile range (IQR), 1.9–4.0 hrs] vs. 5.0 hrs (IQR, 4.0–7.8 hrs), respectively, P<0.001}. Mean duration of insulin infusion was about 1.5 days in both groups. Average total insulin dose administered in both groups was not significantly different (P=0.52). Average insulin rate per hour in the IIP group was higher than the control group, though this was not statistically significant (1.49 units/hr in IIP group vs. 1.24 units/hr in the control group, P=0.35). The number of blood tests was not different (P=0.48). Data are reported in Table 2 and Figure 2. Insulin administration patterns are shown in Figure 3. Trends of mean glucose level over time were lower in the IIP group (Figure 4).

Figure 2 Glycemic outcomes between groups. (A) Time to target glucose. (B) Percentage of time within target range.

Figure 3 Insulin administration pattern. (A) Mean insulin infusion rate over time. (B) Cumulative insulin dose over time.

Figure 4 Trend of mean glucose over time. To visualize glucose trends over time, we calculated mean glucose levels and 95% confidence intervals at each hourly time point for both treatment groups. Smoothed trend lines were generated using Savitzky-Golay filtering (polynomial order =3, window size =20% of data points) to reduce noise while preserving the underlying temporal pattern. The analysis was performed using Python (version 3.12) with pandas, scipy, and matplotlib libraries.

Adverse events

Another secondary outcome, the incidence of hypoglycemia, was higher in terms of the number of patients who developed hypoglycemia but was not statistically significant (12% in the IIP group vs. 4% in the control group, P=0.61). The magnitude of difference was smaller in terms of the number of blood tests (0.9% in IIP group vs. 0.2 % in control group, P>0.99). Incidence of severe hypoglycemia (blood sugar <60 mg/dL) was not different as well. (Numbers of patients: 4% vs. 0%, Blood tests: 0.19% vs. 0%, P>0.99) (Table 3).

The last secondary outcome, incidence of severe hypokalemia, was not significantly different. (4% in the IIP group vs. 0% in the control group, P>0.99). Protocol deviations in the IIP group occurred in three patients (3 patients from 25, 12%), but only 0.56% from blood readings (3 events from 540 readings, P>0.99). These events were three failures to follow the recommended insulin doses due to misinterpretation of data (Table 3). These 3 patients were not excluded, remained in the IIP gr. and were analyzed using ITT analysis.

Glycemic variations (CV)

CV was acceptable in the IIP group (21.9 in the IIP group vs. 18.4 in the control group, P=0.14). MAGE in the IIP group was not increased (P=0.63). IIP group had a lower J-index as compared to the control group (24.1 vs. 31.2, respectively, P=0.002). HBGI in the IIP group is lower than the control group. (1.57 vs. 3.54, P<0.001). ADRR was not significantly different (10.19 vs. 10.60, P=0.72) (Table 4).

Discussion

Efficacy

Perioperative glycemic control by continuous insulin infusion is the most effective method and should be based on a protocol, which can be paper-based or a computerized algorithm (24). Our protocol was modified from the Cleveland Clinic insulin infusion protocol to mitigate the risk of postoperative hypoglycemia. First, our protocol initiates hypoglycemia management at a blood glucose threshold of <70 mg/dL, whereas the Cleveland Clinic insulin infusion protocol utilizes a cutoff of <60 mg/dL. Second, in the insulin infusion adjustment table, we modified the criteria for maintaining the current infusion rate at the blood glucose range of 101–120 mg/dL, whereas the original protocol used a range of 101–115 mg/dL when blood glucose is stable (defined as a change of <30 mg/dL). The other part of the insulin infusion adjustment rate was similar to the original protocol.

Previous studies have shown that insulin infusion protocols provide several advantages over physician-directed adjustment. First, a shorter TTT means earlier blood sugar control, resulting in less cumulative exposure to stress hyperglycemia, which correlates with worse outcomes (4,25). The TTT reported for effective protocols is approximately 4–8 hours, depending on the initial blood glucose level and the strictness of the target range (26-28). In our study, the TTT of the control group was 5.0 hrs., whereas the protocol group was 2.0 hrs.

Second, TIR reflects both hyperglycemic burden and stability better than a single mean/peak glucose level. TIR >80% was associated with a lower incidence of wound infection, shorter ventilation time, and shorter ICU stay (29). Effective protocols have reported TIR around 40–80% (28-30). Our protocol achieves an acceptable TIR of 83.3% compared to 58.4% in the control group.

Third, GV, or the fluctuation of glucose levels, has been linked to oxidative stress, endothelial dysfunction, and inflammatory signaling; clinically related to arrhythmia and organ injury. While mechanisms are still being elucidated, the associations with adverse outcomes are well established (31-34). Several metrics are used to measure GV, such as CV, MAGE, and J-index. Most studies have supported the CV, which is the most extensively studied metric and the strongest predictor of poor outcomes (35). Our study did not demonstrate significant differences in CV and MAGE between the protocol and control groups. However, we found that the J-index was lower in the IIP group. Several factors may explain these findings.

Regarding CV formula, this metric may remain unchanged when both the mean blood glucose level and the standard deviation decrease proportionally, as observed in our study.

Regarding MAGE, one limitation is that it may remain stable when few excursions exceed the threshold, even if the overall glucose profile shifts downward.

In contrast, J-index is more sensitive to changes in overall glycemic control. If the protocol group has either a lower mean glucose level or a lower standard deviation, the J-index will decrease even when CV or MAGE does not change.

In our study, the mean blood glucose level in the IIP group was significantly lower than that in the control group (127.7 vs. 150.0 mg/dL, P<0.001), whereas the standard deviation was similar between groups (31.4 vs. 31.8 mg/dL, respectively, P=0.99) (Table 4). This likely explains why the J-index showed a significant difference, whereas CV and MAGE did not.

Finally, insulin infusion protocols may reduce ICU nursing workload by decreasing the need for physician notification, bedside decision-making, and documentation time through structured decision support. Although our study did not objectively measure this outcome, ICU nursing staff subjectively reported that the protocol was more convenient, suggesting a favorable impact on workflow. There was some concern regarding protocol deviations caused by misinterpretation of data, which led to incorrect infusion rates in three of 540 readings (0.56%) in the protocol group. Although no serious adverse events occurred, this issue should not be overlooked.

Safety

Hypoglycemia may result from tight glycemic control and can increase the risk of major morbidity (36). In studies evaluating the efficacy of insulin infusion protocols, the incidence of hypoglycemia (<70 mg/dL) has been reported using three different denominators. First, incidence per patient refers to the proportion of patients who experienced at least one hypoglycemic event. Marvin et al. reported a patient-based incidence of 17.6% (37). In our study, the incidence per patient was 12% in the protocol group, compared with 4% in the control group. Second, incidence per readings refers to the proportion of blood glucose measurements that were hypoglycemic. In previous insulin infusion protocol studies, the reported incidence was generally less than 3% (38,39). In our study, the incidence per reading was 0.9% in the protocol group and 0.2% in the control group. Third, refers to the number of hypoglycemic episodes occurring during every 1,000 hours of insulin infusion. Juneja et al. reported 20.5 readings per 1,000 infusion-hours for blood glucose <70 mg/dL and 6.35 readings per 1,000 infusion-hours for blood glucose <60 mg/dL (40). In our study, the protocol group had 5.3 readings per 1,000 infusion-hours for blood glucose <70 mg/dL and 1.1 readings per 1,000 infusion-hours for blood glucose <60 mg/dL.

Hypokalemia is another potential adverse event. Although many studies have evaluated insulin infusion protocols, the incidence of hypokalemia has been infrequently reported, likely because it is not usually a primary outcome. In a study by Hoekstra et al., the incidence of hypokalemia (<3.5 mmol/L) was 1.7% of blood tests in the protocol group. In our study, the incidence of severe hypokalemia (<3.0 mmol/L) in the protocol group was 0.2% of blood tests. The low incidence observed in our study is likely attributable to our ICU’s prophylactic potassium replacement protocol, in which potassium is supplemented when the serum potassium level falls below 4.0 mmol/L to prevent post-cardiac surgery arrhythmias. We still aim to further reduce this complication by refining the protocol in its next version.

Limitations

There are several limitations of this study. First, this is a small-sample-size study in a single center. The study participants consist entirely of Thai patients, which cannot directly address inter-ethnic differences, so the generalizability to other ethnic groups is uncertain. The results may not be generalizable to other ICUs, different case-mixes, or other clinical settings. Additionally, the high frequency of blood tests performed in this study may only be feasible in well-resourced ICUs and may not apply to settings with limited nursing staff. This may limit external validity. Second, physician-directed adjustment is often heterogeneous, which makes the comparator less standardized. Third, recruitment was discontinued before the prespecified sample size was reached after an unplanned analysis showed statistically significant between-group differences. Because no prespecified interim-analysis plan, stopping boundary, or alpha-spending procedure was used, early termination may have increased the risk of a type I error and may have overestimated the magnitude of the observed treatment effects. Fourth, the question regarding the burden of nursing care was not directly answered. Finally, due to limited sample size, we cannot study the clinical effect of this protocol. Future studies with a larger number of patients are needed to determine the magnitude of clinical benefits.

Conclusions

The Modified Cleveland Clinic insulin infusion protocol was associated with improved glycemic control during the postoperative critical period, as demonstrated by a shorter TTT, higher percentage of time in range, and a lower J-index. Hypoglycemia and severe hypokalemia were infrequent in both groups; however, the study was not adequately powered to establish safety between the two approaches. Larger multicenter trials are required to confirm safety and improvement in clinical outcomes.

Acknowledgments

The authors would like to express their sincere gratitude to Dr. Peerapol Wong for his valuable guidance in the preparation of this manuscript. The authors utilized chatGPT solely for grammatical editing and enhancing the readability of this manuscript. Claude (Anthropic) was used to provide limited assistance with statistical programming, including suggesting or troubleshooting code that the authors subsequently implemented and verified. All AI-assisted outputs were independently checked against the raw data. The authors take full responsibility for the accuracy, integrity, and interpretation of the reported results.

Footnote

Reporting Checklist: The authors have completed the CONSORT reporting checklist. Available at https://jeccm.amegroups.com/article/view/10.21037/jeccm-2026-0017/rc

Data Sharing Statement: Available at https://jeccm.amegroups.com/article/view/10.21037/jeccm-2026-0017/dss

Peer Review File: Available at https://jeccm.amegroups.com/article/view/10.21037/jeccm-2026-0017/prf

Funding: None.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://jeccm.amegroups.com/article/view/10.21037/jeccm-2026-0017/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the institutional review board of Naresuan University Hospital (NU-IRB, certificate of approval No. 187/2023) and informed consent was obtained from all individual participants.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Jin X, Wang J, Ma Y, et al. Association Between Perioperative Glycemic Control Strategy and Mortality in Patients With Diabetes Undergoing Cardiac Surgery: A Systematic Review and Meta-Analysis. Front Endocrinol (Lausanne) 2020;11:513073. [Crossref] [PubMed]
2. Garg R, Grover A, McGurk S, et al. Predictors of hyperglycemia after cardiac surgery in nondiabetic patients. J Thorac Cardiovasc Surg 2013;145:1083-7. [Crossref] [PubMed]
3. Lazar HL. Glycemic Control during Coronary Artery Bypass Graft Surgery. ISRN Cardiol 2012;2012:292490. [Crossref] [PubMed]
4. Polito A, Thiagarajan RR, Laussen PC, et al. Association between intraoperative and early postoperative glucose levels and adverse outcomes after complex congenital heart surgery. Circulation 2008;118:2235-42. [Crossref] [PubMed]
5. Zerr KJ, Furnary AP, Grunkemeier GL, et al. Glucose control lowers the risk of wound infection in diabetics after open heart operations. Ann Thorac Surg 1997;63:356-61. [Crossref] [PubMed]
6. van den Berghe G, Wouters P, Weekers F, et al. Intensive insulin therapy in critically ill patients. N Engl J Med 2001;345:1359-67. [Crossref] [PubMed]
7. Finfer S, Chittock DRNICE-SUGAR Study Investigators, et al. Intensive versus conventional glucose control in critically ill patients. N Engl J Med 2009;360:1283-97. [Crossref] [PubMed]
8. Ascione R, Rogers CA, Rajakaruna C, et al. Inadequate blood glucose control is associated with in-hospital mortality and morbidity in diabetic and nondiabetic patients undergoing cardiac surgery. Circulation 2008;118:113-23. [Crossref] [PubMed]
9. Bhamidipati CM, LaPar DJ, Stukenborg GJ, et al. Superiority of moderate control of hyperglycemia to tight control in patients undergoing coronary artery bypass grafting. J Thorac Cardiovasc Surg 2011;141:543-51. [Crossref] [PubMed]
10. Furnary AP, Wu Y. Eliminating the diabetic disadvantage: the Portland Diabetic Project. Semin Thorac Cardiovasc Surg 2006;18:302-8. [Crossref] [PubMed]
11. Tamaki M, Shimizu T, Kanazawa A, et al. Efficacy and safety of modified Yale insulin infusion protocol in Japanese diabetic patients after open-heart surgery. Diabetes Res Clin Pract 2008;81:296-302. [Crossref] [PubMed]
12. Zimmerman CR, Mlynarek ME, Jordan JA, et al. An insulin infusion protocol in critically ill cardiothoracic surgery patients. Ann Pharmacother 2004;38:1123-9. [Crossref] [PubMed]
13. Vongsumran N, Buranapin S, Manosroi W. Standardized Glycemic Management versus Conventional Glycemic Management and Postoperative Outcomes in Type 2 Diabetes Patients Undergoing Elective Surgery. Diabetes Metab Syndr Obes 2020;13:2593-601. [Crossref] [PubMed]
14. Sindhvananda W, Surit M. Variations of Perioperative Cardiac Surgery Blood Glucose and Glycemic Variability in Diabetes and Non-Diabetes. J Med Assoc Thail 2020;103:777-84.
15. Kodama K, Tojjar D, Yamada S, et al. Ethnic differences in the relationship between insulin sensitivity and insulin response: a systematic review and meta-analysis. Diabetes Care 2013;36:1789-96. [Crossref] [PubMed]
16. Inaishi J, Meguro S, Itoh A, et al. Efficiency and Safety of New Insulin Infusion Protocol for Japanese Patients After Open-Heart Surgery. Diabetes and Endocrinology 2018;1:4.
17. Cao Y, Widyahening I, Sun X, et al. Pathophysiology of type 2 diabetes: A focus on the metabolic differences among southeast Asian, Chinese and Indian populations and how this impacts treatment. Diabetes Obes Metab 2025;27 Suppl 10:3-14. [Crossref] [PubMed]
18. Olansky L, Sam S, Lober C, et al. Cleveland Clinic cardiovascular intensive care unit insulin conversion protocol. J Diabetes Sci Technol 2009;3:478-86. [Crossref] [PubMed]
19. Imrisek SD, Lee M, Goldner D, et al. Effects of a Novel Blood Glucose Forecasting Feature on Glycemic Management and Logging in Adults With Type 2 Diabetes Using One Drop: Retrospective Cohort Study. JMIR Diabetes 2022;7:e34624. [Crossref] [PubMed]
20. Lazar S, Ionita I, Reurean-Pintilei D, et al. How to Measure Glycemic Variability? A Literature Review. Medicina (Kaunas) 2023;60:61. [Crossref] [PubMed]
21. Service FJ, Molnar GD, Rosevear JW, et al. Mean amplitude of glycemic excursions, a measure of diabetic instability. Diabetes 1970;19:644-55. [Crossref] [PubMed]
22. Kovatchev B, Umpierrez G, DiGenio A, et al. Sensitivity of Traditional and Risk-Based Glycemic Variability Measures to the Effect of Glucose-Lowering Treatment in Type 2 Diabetes Mellitus. J Diabetes Sci Technol 2015;9:1227-35. [Crossref] [PubMed]
23. Kovatchev BP, Otto E, Cox D, et al. Evaluation of a new measure of blood glucose variability in diabetes. Diabetes Care 2006;29:2433-8. [Crossref] [PubMed]
24. American Diabetes Association Professional Practice Committee for Diabetes*. 16. Diabetes Care in the Hospital: Standards of Care in Diabetes-2026. Diabetes Care 2026;49:S339-55. [Crossref] [PubMed]
25. Greco G, Ferket BS, D'Alessandro DA, et al. Diabetes and the Association of Postoperative Hyperglycemia With Clinical and Economic Outcomes in Cardiac Surgery. Diabetes Care 2016;39:408-17. [Crossref] [PubMed]
26. Shetty S, Inzucchi SE, Goldberg PA, et al. Adapting to the new consensus guidelines for managing hyperglycemia during critical illness: the updated Yale insulin infusion protocol. Endocr Pract 2012;18:363-70. [Crossref] [PubMed]
27. Barth MM, Oyen LJ, Warfield KT, et al. Comparison of a nurse initiated insulin infusion protocol for intensive insulin therapy between adult surgical trauma, medical and coronary care intensive care patients. BMC Emerg Med 2007;7:14. [Crossref] [PubMed]
28. Passarelli AJ, Gibbs H, Rowden AM, et al. Evaluation of a Nurse-Managed Insulin Infusion Protocol. Diabetes Technol Ther 2016;18:93-9. [Crossref] [PubMed]
29. Omar AS, Salama A, Allam M, et al. Association of time in blood glucose range with outcomes following cardiac surgery. BMC Anesthesiol 2015;15:14. [Crossref] [PubMed]
30. Giovannetti ER, Lee RO, Thomas RL, et al. Continuous Glucose Monitoring-Guided Insulin Infusion in Critically Ill Patients Promotes Safety, Improves Time Efficiency, and Enhances Provider Satisfaction. Endocr Pract 2025;31:1143-9. [Crossref] [PubMed]
31. Chang S, Xu M, Wang Y, et al. Acute Glycemic Variability and Early Outcomes After Cardiac Surgery: A Meta-Analysis. Horm Metab Res 2023;55:771-80. [Crossref] [PubMed]
32. Subramaniam B, Lerner A, Novack V, et al. Increased glycemic variability in patients with elevated preoperative HbA1C predicts adverse outcomes following coronary artery bypass grafting surgery. Anesth Analg 2014;118:277-87. [Crossref] [PubMed]
33. Sim MA, Liu W, Chew STH, et al. Wider perioperative glycemic fluctuations increase risk of postoperative atrial fibrillation and ICU length of stay. PLoS One 2018;13:e0198533. [Crossref] [PubMed]
34. Ogawa S, Okawa Y, Sawada K, et al. Continuous postoperative insulin infusion reduces deep sternal wound infection in patients with diabetes undergoing coronary artery bypass grafting using bilateral internal mammary artery grafts: a propensity-matched analysis. Eur J Cardiothorac Surg 2016;49:420-6. [Crossref] [PubMed]
35. Guo Y, Qiu Y, Xue T, et al. Association between glycemic variability and short-term mortality in patients with acute kidney injury: a retrospective cohort study of the MIMIC-IV database. Sci Rep 2024;14:5945. [Crossref] [PubMed]
36. Johnston LE, Kirby JL, Downs EA, et al. Postoperative Hypoglycemia Is Associated With Worse Outcomes After Cardiac Operations. Ann Thorac Surg 2017;103:526-32. [Crossref] [PubMed]
37. Marvin MR, Inzucchi SE, Besterman BJ. Computerization of the Yale insulin infusion protocol and potential insights into causes of hypoglycemia with intravenous insulin. Diabetes Technol Ther 2013;15:246-52. [Crossref] [PubMed]
38. Boord JB, Sharifi M, Greevy RA, et al. Computer-based insulin infusion protocol improves glycemia control over manual protocol. J Am Med Inform Assoc 2007;14:278-87. [Crossref] [PubMed]
39. Clergeau A, Parienti JJ, Reznik Y, et al. Impact of a Paper-Based Dynamic Insulin Infusion Protocol on Glycemic Variability, Time in Target, and Hypoglycemic Risk: A Stepped Wedge Trial in Medical Intensive Care Unit Patients. Diabetes Technol Ther 2017;19:115-23. [Crossref] [PubMed]
40. Juneja R, Roudebush CP, Nasraway SA, et al. Computerized intensive insulin dosing can mitigate hypoglycemia and achieve tight glycemic control when glucose measurement is performed frequently and on time. Crit Care 2009;13:R163. [Crossref] [PubMed]