Relationship Between Glucose Time in Range in Diabetic and Non-Diabetic Patients and Mortality in Critically Ill Patients

Abstract

Background: Shorter time spent in specific blood glucose ranges is associated with mortality benefit in critically ill patients. However, various time in range values are reported, each based on a specific blood glucose range. Objective: To evaluate relationship between percentage of time spent at various blood glucose ranges (TIR) and mortality in critically ill patients. Methods: Single-center, retrospective, cohort study that included adult patients admitted to ICU for at least one day. We evaluated the relationship between TIR at prespecified blood glucose ranges and hospital mortality in diabetic and non-diabetic patients Results: Of the 5287 patients included, 3705 (70.0%) were non-diabetic and 1582 were diabetic (29.9%). Diabetic patients had higher in-hospital mortality rate (15.8%) compared to non-diabetic patients (11.3%), p < 0.0001, and with higher incidence of hyperglycemia (77.8% vs. 39.4%) and hypoglycemia (14.3% vs. 10%) compared to non-diabetic patients, p < 0.0001. The highest median TIR for both diabetic [76% (49.1 − 97.8%)] and non-diabetic patients [100% (92.3--100%)] was at blood glucose range of 70-180 mg/dL. In non-diabetic cohort, the only optimal TIR of 40% at blood glucose range of 70-120 mg/dL was identified. Non-diabetic patients stratified into TIR 70-120 mg/dL > 40% reported significantly lower mortality (7.0%) rate compared to patients with TIR 70-120 mg/dL < 40% (15.7%), OR 0.52, 95% CI 0.27-0.97, adjusted-p = 0.03. In diabetic patients, no relationship was detected between TIR at all predefined glucose ranges and hospital mortality. Conclusion: Critically ill non-diabetic patients who spent at least 40% of time in blood glucose range of 70-120 mg/dL had improved survival. This association was not observed in diabetic patients.

Keywords

glucose diabetes dysglycemia critically ill patients intensive insulin therapy glycemic variability

Introduction

Intensive monitoring of plasma glucose levels and insulin treatment in critically ill patients has been a standard of care in the intensive care unit (ICU) and an area of ongoing research. Large randomized control trials have focused on target blood glucose levels when evaluating outcomes in the ICU.^1‐3 Earlier trials supported the use of intensive insulin therapy (IIT) to maintain a tight blood glucose range of 80-110 mg/dL, to avoid organ failure and mortality complications. However, subsequent studies were unable to show similar results.^4,5 A major, randomized, multicenter study (NICE-SUGAR) reported significant mortality risk associated with intensive insulin targets establishing the widely adopted conservative guideline of maintaining glucose levels around 140-180 mg/dL.⁶

Within the last decade, the concept of “dysglycemia” in ICU has been explored and used to explain the disparity in results seen in the previous randomized control trials.^7,8 This concept includes four domains: hyperglycemia, hypoglycemia, glycemic variation (GV), and time in range (TIR).⁷ TIR represents the percentage of time when a patient's glycemic level remains within the targeted blood glucose range and integrates information obtained by the other three dysglycemia domains into a single value.⁷ This unifying metric of glycemic control could have contributed to inconsistent outcomes in many major randomized controlled studies in IIT since it was found that many participants might have spent less time in their targeted glucose range.^7,8

Increased TIR has been associated with mortality benefit, reduction in organ failure, shorter ICU length of stay, shorter mechanical ventilation time, significant reduction in postoperative atrial fibrillation, and lower incidences of infection.^9–12 The reported TIR values in literature vary from 30-80% based on blood glucose range evaluated, making TIR use in the clinical setting less pragmatic. We sought to investigate the relationship between TIR and hospital mortality using various blood glucose ranges for diabetic (DM) and non-diabetic (NDM) patients. We hypothesize that optimal TIR varies based on blood glucose ranges evaluated.

Material and Methods

Study Design

This study was a non-interventional, retrospective, cohort study that included patients admitted to the intensive care unit (ICU) of Yale New Haven Hospital. A query of the electronic medical record was used to identify all blood glucose assays performed in Medical, Surgical, and Neuro-critical ICU patients over five years (January 1, 2013 – January 1, 2018). Patients were included if they were adults (at least 18 years old) and admitted to the ICU for at least one day. Patients were excluded if they had diabetic ketoacidosis, hyperosmolar hyperglycemic state, or expired within 48 hours from ICU admission. For patients with multiple ICU admissions within the same hospital admission, data were included from the first admission only. ICD-10 codes were utilized to identify DM patients versus NDM patients. The study was approved by the Yale New Haven Hospital Institutional Review Board and was exempted from the need for informed consent due to minimal risk status. During the study period, this institution used a hyperglycemia management protocol that was previously evaluated and targets a blood glucose range of 120-160 mg/dL for all patients admitted to the ICU.¹³ Nurses performed blood glucose monitoring using Strat Strip Glucose Meter (Nova Biomedical Corporation, Waltham, MA) to test capillary, venous, or arterial blood (Supplemental Appendix 1).

Study Outcomes

The primary objective was to explore the optimal TIR value associated with survival benefit in DM and NDM critically ill patients in predefined blood glucose ranges (70-100 mg/dL, 70-120 mg/dL, 70-140 mg/dL, 70-160 mg/dL, 70-180 mg/dL, 120-140 mg/dL, 120-160 mg/dL, 120-180 mg/dL, 140-160 mg/dL, and 140-180 mg/dL). DM and NDM patients were stratified into TIR-Hi (patients with TIR greater than optimal TIR identified) and TIR-Lo (patients with TIR less than optimal TIR identified) cohorts. Our secondary objective was to evaluate clinical outcomes between the TIR-Hi and TIR-Lo cohorts for both DM and NDM patients.

Data Collection and Metrics

Patient baseline characteristics were collected on ICU admission, including baseline demographics, Simplified Acute Physiology Score (SAPS), mechanical ventilation, renal replacement therapy, and positive infection (one or more positive cultures obtained from blood, normally sterile fluids, or sputum).¹⁴ Patient Rothman Index¹⁵ and diabetes status were collected at hospital admission. Glucose metrics measured included glucose levels measured via blood capillary, venous, and arterial blood sampling. TIR was expressed as the percentage of time when a patient's glycemic level remained within the specified blood glucose range while in ICU.⁷ TIR was calculated for each patient using reported blood glucose values without extrapolation. Other glucose metrics evaluated included GV, glycemic penalty index (GPI), and mean amplitude of glucose excursions (MAGE).^16,17

Statistical Analysis

To determine optimal TIR associated with survival benefit for predefined blood glucose ranges, a classification and regression tree (CART) analysis was utilized. CART analysis is a non-parametric statistical method that uses a decision tree to solve classification and regression problems in a binary recursive partitioning manner.¹⁸ The advantage of CART models is that they do not make assumptions about the underlying distribution of values of the independent variables, thus appropriate for highly skewed data.¹⁹ CART models are suited to the generation of clinical decision rules and have been successfully used in various clinical prediction models.^20–22 TIR is represented by a node in the decision tree that can only be split into two groups. This binary partitioning process is applied repeatedly until a significant predefined outcome is identified. The CART-defined breakpoints were then used to determine two cohorts for DM and NDM patients, TIR-Hi and TIR-Lo, and outcomes were compared between the two cohorts.

The discrimination, accuracy, and statistical significance of the optimal TIR thresholds as determined from the CART analysis in determining hospital mortality, was tested in the presence of other patient attributes. Using the identified optimal TIR cutoff, TIR-Hi and TIR-Lo cohorts were identified. Multivariable logistic regression was conducted to adjust for factors that may have confounded the effect of cohort assignment on hospital mortality. Variables that influence TIR and hospital mortality were tested for multicollinearity, and if two factors showed collinearity, only one factor was included in the multivariable regression analysis (Supplemental Table 1A-D).

To verify that the above procedure of identifying optimal TIR thresholds from various predefined blood glucose ranges does not overfit the dataset, the generalization capability of the procedure was tested. Patient cohort was randomly split into a validation and derivation set. The procedure was applied on the derivation set to identify the optimal TIR cutoff, which together with other clinical attributes as independent variables, was used to construct a multivariable logistic regression model of hospital-mortality. The area under the receiver–operating characteristic curve (AUROCC)²³ values for this model were evaluated and compared on both the derivation and validation set. Since the validation set was not involved in the construction of the model, the AUROCC of the model as applied to the validation set indicates the model's performance on unseen data.

A sensitivity analysis was constructed to evaluate the interaction of the optimal TIR threshold with ICU-specific severity of illness score (SAPS), type of ICU, and ICU length of stay.

Continuous data were analyzed with Student's t-test or Mann-Whitney U test, and categorical data were analyzed with Fisher's exact test or Chi-square test, as appropriate. All analyses were performed using R software, Version 3.5.2 (R Foundation for Statistical Computing, Vienna, Austria). A two-sided p-value of 0.05 or less was considered statistically significant. The Bonferroni correction was used to correct for multiple testing while keeping the overall type-I error within 5%. Bonferroni adjusted p-value of 0.05 or less was considered statistically significant.

Results

A total of 5287 patients met inclusion criteria without meeting exclusion criteria and were included in this evaluation. Baseline characteristics and glycemic control attributes of evaluated patients are summarized in Table 1. The majority of patients included were NDM (n = 3,705, 70.0%) versus DM (n = 1,582, 29.9%). DM patients were older, more obese, and sicker than NDM patients. More patients with diabetes were in the medical ICU (n = 958, 60.6%) than the NDM cohort (n = 860, 23.3%). Overall, DM patients had a higher hospital mortality rate (15.8%) compared to NDM patients (11.3%), p < 0.0001. Patients in the DM cohort had higher hyperglycemia and hypoglycemia incidence than NDM (hyperglycemia, 77.8% vs. 39.4%, p < 0.0001 and hypoglycemia, 14.3% vs. 10%, p < 0.0001). Mean serum blood glucose level was higher in the DM cohort (158.3 ± 41.2 mg/dL) compared to the NDM cohort (131.1 ± 26.3 mg/dL), p < 0.0001. Median number of blood glucose evaluations per 24-hour period was 6 (5-7) in the NDM cohort and 7 (5--10) in the DM cohort, p < 0.0001. Additionally, DM patients had more variation in their serum blood glucose levels than NDM patients with higher GV (24 ± 10% vs. 18 ± 10%, p < 0.0001), higher GPI (49.7 ± 21.6 vs. 29.8 ± 17.6, p < 0.0001), and higher MAGE (158.3 ± 41.2 vs. 131.1 ± 26.3, p < 0.0001).

Table 1.

Baseline Characteristics and Glycemic Control Attributes Stratified Based on Diabetes status.

Characteristic	Non-Diabetic n = 3705	Diabetic n = 1582	p-value
Age (years)	59.7 ± 18.6	66.0 ± 13.7	<0.0001
Weight (kg)	80.6 ± 23.1	90.2 ± 28.7	<0.0001
Height (inches)	66.8 ± 4.3	66.7 ± 4.1	0.20
Gender, Male (%)	2138 (57.7)	995 (56.6)	0.50
BMI (kg/m²)	27.9 ± 7.3	31.4 ± 9.5	<0.0001
SAPS at ICU admission	40.7 ± 14.3	44.7 ± 14.3	<0.0001
Rothman Index at hospital admission	50.7 ± 19.0	45.3 ± 19.1	<0.0001
Steroids administered during ICU stay (%)	848 (22.9)	449 (28.4)	<0.0001
Invasive mechanical ventilation at baseline (%)	1697 (45.8)	693 (43.8)	0.20
Renal replacement therapy at baseline (%)	122 (3.3)	150 (9.5)	<0.0001
Positive cultures (%)	548 (14.8)	326 (20.6)	<0.0001
Number of blood glucose levels per 24-hour period	6 (5 − 7)	7 (5 − 10)	<0.0001
ICU
MICU (%)	860 (23.2)	959 (60.6)	<0.0001
SICU (%)	2719 (73.4)	592 (37.4)
NICU (%)	126 (3.4)	32 (2.0)
ICU length of stay (days)	2 (1 − 5)	2 (1 − 5)	0.02
Hospital length of stay (days)	8 (5 − 15)	9 (5 − 15)	0.003
Hospital mortality (%)	419 (11.3)	250 (15.8)	<0.0001
Hyperglycemia (%)	1460 (39.4)	1231 (77.8)	<0.0001
Hypoglycemia (%)	371 (10.0)	226 (14.3)	<0.0001
Incidence hyperglycemia per patient	2.4 ± 7.1	11.2 ± 19.5	<0.0001
Incidence hypoglycemia per patient	0.25 ± 1.1	0.37 ± 1.3	<0.0005
Hyperglycemic index (HGI, mg/dL)	2.6 ± 9.9	16.1 ± 24.6	<<0.0001
Hypoglycemic index (HoGI, mg/dL)	0.028 ± 0.3	0.037 ± 0.3	0.3
Mean blood glucose (mg/dL)	131.1 ± 26.3	158.3 ± 41.2	<0.0001
Glucose coefficient of variation (GV, %)	18 ± 10	24 ± 10	<0.0001
Glycemic penalty index (GPI)	29.8 ± 17.6	49.7 ± 21.6	<0.0001
Mean amplitude of glucose excursions (MAGE)	131.1 ± 26.3	158.3 ± 41.2	<0.0001

Abbreviations: BMI, body mass index; SAPS, Simplified Acute Physiology Score; ICU, intensive care unit; MICU, medical intensive care unit; SICU, surgical intensive care unit; NICU, neuro-intensive care unit.

Date presented in mean ± SD or Median (IQR) or n (%) as appropriate.

Hypoglycemia was defined as reporting at least one blood glucose value less than 70 mg/dL during ICU stay, hyperglycemia was defined as reporting at least one blood glucose value greater than 180 mg/dL during ICU stay, GV was defined as standard deviation (SD) of blood glucose levels divided by the mean blood glucose level during ICU stay, Glycemic penalty index (GPI) was defined as a penalty assigned to blood glucose values that deviate from normoglycemia (80-110 mg/dL) during ICU stay, and calculated as the average of all penalties that are individually assigned to each measured blood glucose value based on an optimized smooth penalty function, where the computation of this index returns a number between 0 (no penalty) and 100 (the highest penalty), Mean amplitude of glucose excursions (MAGE) during ICU stay was defined as mean of the absolute values of any delta blood glucose from consecutive measurements that were higher than the SD, where mean daily delta blood glucose was defined as the average of the difference between highest and lowest blood glucose value of each day.

Median TIR breakdowns based on the predefined serum blood glucose ranges for DM and NDM are reported in Supplemental Table 2. Both DM and NDM patients highest median TIR was at blood glucose range of 70-180 mg/dL, DM, 76.0% (49.1 − 97.8)% and NDM, 100.0% (92.3–100.0)%. Overall, DM patients spent less time in the lower blood glucose ranges than NDM patients.

CART analysis was performed on all of the prespecified blood glucose ranges to identify significant TIR breakpoints in NDM (Supplemental Appendix 2). Subsequently, NDM patients were stratified into TIR-Hi and TIR-Lo based on the CART-identified TIR, and the significance of this breakpoint was assessed in multivariable analysis (Supplemental Appendix 2). TIR 40% in blood glucose range 70-120 mg/dL was the only significant breakpoint identified in the NDM cohort (Figure 1). This breakpoint can be interpreted as follows: for NDM patients below threshold TIR (70-120 mg/dL) 40%, the mortality rate is 15.7% which is significantly higher than 7.0%, when TIR (70-120 mg/dL) is ≥40%.

Table 2 provides clinical characteristics and glycemic control attributes in the NDM patient cohort stratified based on achieving TIR (70-120 mg/dL) greater than 40% (TIR-Hi) and less than 40% (TIR-Lo). In the NDM cohort, mortality was almost twice that in the TIR-Lo compared to TIR-Hi (15.7% vs. 7.0%, p < 0.0001, respectively). Overall, hyperglycemic episodes were higher than hypoglycemia episodes in both cohorts. In comparison to the TIR-Hi cohort, more patients in the TIR-Lo cohort received steroids (32.7% vs. 13.6%, p < 0.0001) and had higher incidence of hyperglycemia (55.8% vs. 23.8%, p < 0.0001). In multivariable analysis (Table 3), TIR-Hi was independently associated with decreased odds of dying compared to TIR-Lo (OR 0.52, 95% CI 0.27-0.97, adjusted-p = 0.03). The sensitivity analysis demonstrated that TIR-Hi was associated with a significant decrease in mortality in NDM patients when stratified based on ICU severity of illness, type of ICU, and ICU length of stay (Figures 2A–C). The derivation-validation analysis of the multivariable logistic regression model of hospital-mortality demonstrated that the procedure to identify optimal TIR cutoffs is robust to overfitting, with AUROCC (validation cohort; 0.94 and derivation cohort; 0.96) Supplemental Appendix 3.

Table 2.

Clinical Characteristics and Glycemic Control Attributes in Non-Diabetic Patients Stratified Based Achieving TIR (70-120 mg/dL) Greater Than 40% (TIR-Hi) and Less Than 40% (TIR-Lo).

Characteristic	TIR-Hi (n = 1894)	TIR-Lo (n = 1811)	p-value
Age (years)	58.1 ± 19.4	61.4 ± 17.5	<0.0001
Gender, Male (%)	1076 (56.8)	1061 (58.6)	0.30
BMI (kg/m²)	27.4 ± 7.2	28.5 ± 7.5	<0.0001
SAPS at ICU admission	39.9 ± 14.2	41.6 ± 14.4	0.0001
Rothman Index at hospital admission	52.3 ± 18.0	49.0 ± 19.8	<0.0001
Steroids administered during ICU stay (%)	258 (13.6)	592 (32.7)	<0.0001
Invasive mechanical ventilation at baseline (%)	816 (43.1)	880 (48.6)	0.0009
Renal replacement therapy at baseline (%)	74 (3.9)	53 (2.9)	0.10
Positive cultures (%)	258 (13.6)	292 (16.1)	0.04
Number of blood glucose levels per 24-hour period	5 (4 − 7)	6(5 − 8)	<0.0001
ICU
MICU (%)	384 (20.3)	474 (26.2)	<0.0001
SICU (%)	1447 (76.4)	1275 (70.4)
NICU (%)	63 (3.3)	62 (3.4)
ICU length of stay (days)	2 (1 − 5)	2 (1 − 5)	0.07
Hospital mortality (%)	133 (7.0)	284 (15.7)	<0.0001
Hyperglycemia (%)	450 (23.8)	1010 (55.8)	<0.0001
Hypoglycemia (%)	251 (13.3)	120 (6.6)	<0.0001
Incidence hyperglycemia per patient	0.56 ± 1.74	4.4 ± 9.7	<0.0001
Incidence hypoglycemia per patient	∼0	∼0	-
Hyperglycemic index (HGI, mg/dL)	0.49 ± 2.5	4.77 ± 13.5	<0.0001
Hypoglycemic index (HoGI, mg/dL)	∼0	∼0	-
Mean blood glucose (mg/dL)	115.1 ± 13.9	147.9 ± 25.7	<0.0001
Glucose coefficient of variation (GV, %)	18 ± 10	18 ± 10	0.80
Glycemic penalty index (GPI)	17.4 ± 9.6	42.9 ± 14.3	<0.0001
Mean amplitude of glucose excursions (MAGE)	32.6 ± 26.2	42.9 ± 35.9	<0.0001

Date presented in mean ± SD or median (IQR) or percentage (%) as appropriate. Hypoglycemia was defined as blood glucose value less than 70 mg/dL and hyperglycemia was defined as blood glucose value greater than 180 mg/dL. HGI and HoGI is the area under the blood glucose curve spent in hyper/hypo-glycemic state divided by duration of ICU stay.

Table 3.

Multivariable Logistic Regression for Hospital Mortality in non-Diabetic Patients.

Variable	OR (95% CI)	p-value	*Adjusted OR (95% CI)	*Adjusted p-value
Age	1.01 (1.00-1.02)	0.03	1.01 (1.00-1.02)	0.35
Gender (male)	1.23 (0.91-1.67)	0.18	1.23 (0.80-1.91)	1.00
BMI	1.02 (1.00-1.04)	0.02	1.02 (1.00-1.05)	0.17
ICU (SICU)	1.43 (1.02-2.01)	0.04	1.43 (0.88-2.34)	0.41
SAPS	1.02 (1.01-1.03)	<0.0001	1.02 (1.01-1.04)	0.001
Rothman Index	0.88 (0.86-0.89)	<0.0001	0.88 (0.86-0.89)	<0.0001
Steroids	1.18 (0.83-1.66)	0.35	1.18 (0.72-1.93)	1.00
Invasive mechanical ventilation	2.12 (1.44-3.14)	<0.0001	2.12 (1.22-3.74)	0.002
Renal replacement therapy	0.88 (0.49-1.57)	0.68	0.88 (0.37-2.01)	1.00
Positive cultures	0.97 (0.69-1.36)	0.87	0.97 (0.59-1.58)	1.00
ICU length of stay	0.94 (0.92-0.97)	<0.0001	0.94 (0.91-0.98)	<0.0001
Number of blood glucose levels per 24-hour period	0.98 (0.93-1.03)	0.38	0.98 (0.92-1.05)	1.00
Mean blood glucose	1.01 (0.99-1.02)	0.34	1.01 (0.99-1.02)	1.00
Mean hyperglycemic index (HGI)	1.00 (0.98-1.02)	0.98	1.00 (0.97-1.03)	1.00
Mean hypoglycemic index (HoGI)	1.88 (0.91-3.57)	0.08	1.88 (0.75-4.58)	0.77
Glucose coefficient of variation (GV)	3.00 (0.60-14.29)	0.17	3.00 (0.29-27.64)	1.00
TIR-Hi (TIR 70-120 mg/dL > = 40%)	0.52 (0.33-0.80)	0.003	0.52 (0.27-0.97)	0.03

Abbreviations: OR, odds ratio; CI, confidence interval; BMI, body mass index; ICU, intensive care unit; SICU, surgical intensive care unit; SAPS, Simplified Acute Physiology Score; TIR-Hi, TIR above 40%.

ICU type has MICU as reference level.

*Bonferroni correction was used to correct for multiple testing. Since ten blood glucose ranges were considered, and ten hypothesis tests for significance of TIR cutoffs were performed, the Bonferroni adjustment rejects at significance level 0.05/10 = 0.005, which keeps the overall type-I error within 5%. The Bonferroni adjustment constructs 99.5% confidence intervals (1-0.05/10 = 0.995) to ensure overall confidence levels are at least 95%. Bonferroni adjusted p-value of 0.05 or less was considered statistically significant.

Multivariable analyses evaluating CART analysis-identified TIR breakpoints for other prespecified blood glucose ranges in NDM patients showed no significant relationship between TIR within the prespecified blood glucose ranges and in-hospital mortality (Supplemental Table 3 and Supplemental Appendix 2). Similarly, CART analysis was performed on all prespecified blood glucose ranges to identify significant breakpoints in DM patients (Supplemental Appendix 4). Multivariable analyses evaluating CART analysis-identified TIR breakpoints for prespecified blood glucose ranges in DM patients showed no significant relationship with in-hospital mortality (Supplemental Appendix 4).

Discussion

In this analysis, we identified that high TIR was associated with survival benefit in NDM critically ill patients. However, this relationship was not observed in DM patients. NDM patients spending more than 40% of the time in the ICU within blood glucose range of 70-120 mg/dL were associated with 48% decrease in odds of death (OR 0.52, 95% CI 0.27-0.97, adjusted-p = 0.03). This association was persistent regardless of severity of illness, type of ICU, and ICU length of stay. In a multivariable analysis, this relationship was independent of other predictors, such as ICU length of stay, severity of illness, and other domains of glycemic control such as GV, hyperglycemic index, and hypoglycemic index.

TIR has not been analyzed as a variable in large randomized control trials that have been published to date.⁷ However, observational studies that evaluated TIR have reported various TIR values. These studies used inconsistent blood glucose ranges and glucose management protocols that targeted different glucose goals. Signal and colleagues assessed the relationship between a TIR (72-126 mg/dL) and hospital mortality and reported that TIR greater than 30%, 50%, and 70% were all associated with an increase in odds of survival. However, TIR 70% was associated with the highest increase odds of survival compared with lower thresholds of TIR (≥30% and ≥50%).²⁴ Notably, patients were only evaluated for five days, and length of blood glucose control was measured from first recognition of hyperglycemia versus ICU admission.²⁴ The same group performed a post-hoc analysis of the prematurely stopped Glucontrol trial and included patients from both the intervention (79-110 mg/dL) and the control groups (140-180 mg/dL). They reported that the presence of TIR greater than 50% at blood glucose range 72-126 mg/dL was associated with significant survival regardless of the blood glucose target.²⁵ Omar et al evaluated the role of TIR in cardiac surgery patients only and reported that TIR (108-146 mg/dL) greater than 80% was associated with a significant decrease in postoperative atrial fibrillation, shorter mechanical ventilation, lower incidence of operative wound infection, and shorter ICU length of stay.⁹ Most recently, the relationship between TIR (70-140 mg/dL) and mortality in DM and NDM patients was evaluated in a well-designed observational study by Krinsley and Preiser. In this evaluation, mortality of NDM patients doubled when TIR was less than 80%. However, TIR was not associated with mortality in DM patients.¹¹

Our findings corroborate the importance of evaluating glucose TIR in critically ill patients. However, the optimal TIR in existing literature ranges from 30% to 80%. This variation in TIR doesn't provide sufficient guidance for the practical use of this domain.^{9,11,12,24,25} We hypothesize that this could be due to lack of sensitivity in the strategies used to identify the optimal TIR breakpoints; some evaluations used empiric TIR cutoff values^9,24,25 while others used the median TIR of their evaluable cohort.^11,12 In the current analysis, we used the CART method, which strengthened our findings by exploring various TIR breakpoints and identifying a significant TIR value in this cohort.¹⁸ The cutoff value identified in this analysis is lower than what has been previously reported.^9,11,25 This could potentially be due to different target blood glucose ranges utilized in the study glucose management protocols. This institution insulin protocol targets bringing blood glucose to a level between 120-160 mg/dL¹³, while Krinsley and Preiser's protocol targeted blood glucose range of 90-120 mg/dL, Omar, et al targeted blood glucose range 108-146 mg/dL, and Penning, et al targeted blood glucose range of 79-110 mg/dL (intervention cohort) and 140-180 mg/dL (control cohort).^9,11,13,25

Furthermore, we hypothesize that since the existing literature used various glucose ranges to determine TIR, it led to discrepancies in the optimal TIR breakpoints. Krinsley and Preiser used a glucose range 70-140 mg/dL as the authors indicated that ICU nurses could achieve this range.¹¹ Given that glucose control in the ICU is managed with proactive protocols, it's expected that patients spend more time at specific glucose ranges targeted by these treatment protocols than others. As such, selecting a specific glucose range for the TIR analysis might result in a selection bias as TIR would differ based on the glucose range selected.^{9,11,12,24,25} Additionally, the glucose ranges for DM are expected to be broader than the glucose ranges in NDM patients.²⁶ This was demonstrated in our patient cohort as we identified that compared to NDM patients, DM patients spent less time in the lower blood glucose ranges. Conversely, DM patients spent comparatively more time in the higher ranges 140-160 mg/dL and 140-180 mg/dL. As such, it would be more pragmatic to identify different TIR glucose ranges for DM and NDM patients and evaluate TIR with various blood glucose ranges. We addressed these limitations in our current study by evaluating various blood glucose ranges and NDM and DM patients separately.

Our finding that a blood glucose range of 70-120 mg/dL was the only range identified in NDM with a significant TIR could further highlight the results characterized by the earlier IIT trial that demonstrated significant reductions in mortality and morbidity in patients targeting blood glucose range 80-110 mg/dL.¹ Subsequent trials failed to show similar findings due to increased risk of death related to hypoglycemia.^3,4 However, these trials failed to report TIR. The Glucontrol trial was the only trial that reported TIR. Notably, this study reported TIR of 42.8% for the IIT cohort (79-110 mg/dL), which is comparatively similar to this study's findings.⁴ Additionally, the post-hoc analysis of this trial reported no difference in organ failure development based on intention-to-treat to different glucose ranges (79-110 mg/dL and 140-180 mg/dL). Interestingly, achieving 50% TIR in blood glucose range of around 72-126 mg/dL was independently associated with increased survival for patients in both intensive and conventional arms.²⁵ Our findings further corroborate the results from the Glucontrol trial and its post-hoc analysis.^4,25 Multiple blood glucose ranges were assessed in this evaluation, potentially increasing the risk of identifying a significant blood glucose range by chance. However, the derivation-validation analysis on the NDM cohort shows that the use of CART-analysis on multiple blood glucose ranges to identify TIR cutoffs does not overfit to the data, as the AUROCC values of the logistic regression model for hospital mortality, constructed using the discovered optimal TIR cutoff and adjusted for other possibly confounding variables, remains similar between derivation and validation cohorts.

Similar to previous reports, in this study, DM patients had higher mean blood glucose and more fluctuations in blood glucose levels (GPI, MAGE, and GV values) compared to NDM.^27,28 In contrast to NDM patients, there was no apparent relationship between TIR and mortality in DM. These results were similar to what has been reported in the Krinsley and Preiser study.¹¹ It has been postulated that DM patients have persistent hyperglycemia, resulting in downregulation of GLUT 1 and GLUT 3 transporters. The activity of these transporters leads to cellular toxicity.²⁹ This cellular conditioning effect protects DM patients against the damage caused by acute hyperglycemia during critical illness, potentially explaining our findings.²⁹ A recent study by Lanspa and colleagues reported an association between TIR and survival in DM patients. A possible explanation for this discordance in findings is that Lanspa and colleagues included patients with good antecedent glycemic control based on pre-admission hemoglobin A1C (HbA_1c) levels. These patients, similar to NDM patients, lack the cellular conditioning benefit of hyperglycemia observed in DM with poor antecedent glycemic control.³⁰

Even though this is one of the largest studies to evaluate the relationship between TIR and mortality, there are some limitations to this study. First, this was a retrospective medical record review and thus, our findings don't prove causation. Our findings are hypothesis-generating, and the ideal TIR blood glucose target can only be derived from a randomized controlled trial. However, designing a prospective study that randomizes patients to a low TIR would be challenging due to lack of adopting continuous glucose monitoring in the ICU.³¹ Additionally, it might be unethical to randomize patients to a low TIR. Therefore, investigators must rely on observational cohort studies, often of retrospective nature similar to this study, to evaluate the optimal TIR.^11,12 Even though, we adjusted for several potential confounding variables in the multivariable logistic regression, residual confounding might still exist. This was a single-center study with a specific protocolized approach to manage blood glucose levels and our findings might not apply to other institutions that utilize a different protocol. Additionally, our study included mixed medical and surgical patients, with the NDM cohort being predominantly surgical patients while the DM cohort was primarily medical patients. As such, our contrasting findings in DM and NDM cohorts might have been impacted by the type of patients evaluated in each cohort. Furthermore, due to the retrospective nature of the study, we were not able to report on baseline HbA_1c, which has a prognostic value in critically ill patients.^32,33 Additionally, we could not report on the amount of insulin administered. However, this institution's insulin protocol has been previously evaluated.¹³ The median blood glucose in this study was similar to what has been previously reported, indicating the consistent performance of this institution protocol.¹³ Even though blood glucose values were mainly from capillary glucose meters, which might have analytic inaccuracies, they are considered the standard of care in most ICUs.³⁴ Finally, we relied on ICD-10 codes to identify DM versus NDM patients, which might lead to misclassification. However, the proportion of DM and NDM patients reported in this analysis is similar to what has been previously reported.¹¹

This study has several strengths, including accounting for other glucose variation domains that could potentially impact outcomes^16,17,35 and reporting on clinically significant variables at baseline, such as use of corticosteroids, need for invasive mechanical ventilation, and presence of positive cultures. Additionally, the high frequency of blood glucose sampling in our study strengthens the study's design by assuming that the sampling times between blood glucose levels used in evaluating the glucose metrics were constant. Finally, accounting for diabetes status at baseline in this study is consistent with recent literature showing association between glycemic control measures and ICU mortality differs between patients without and with DM.²⁷

Conclusion

In the current study, critically ill NDM patients who spent at least 40% of time in blood glucose range of 70-120 mg/dL had improved survival compared with those who did not. This association was not observed in DM patients. Future prospective evaluations of intensive glucose therapy in the ICU should incorporate TIR as a component of glycemic control.

Supplemental Material

sj-docx-1-jic-10.1177_08850666221098383 - Supplemental material for Relationship Between Glucose Time in Range in Diabetic and Non-Diabetic Patients and Mortality in Critically Ill Patients

Supplemental material, sj-docx-1-jic-10.1177_08850666221098383 for Relationship Between Glucose Time in Range in Diabetic and Non-Diabetic Patients and Mortality in Critically Ill Patients by Mahmoud A. Ammar, Abdalla A. Ammar, Timothy Wee, Ranjit Deshpande, Matthew Band and Shamsuddin Akhtar in Journal of Intensive Care Medicine

Supplemental Material

sj-docx-2-jic-10.1177_08850666221098383 - Supplemental material for Relationship Between Glucose Time in Range in Diabetic and Non-Diabetic Patients and Mortality in Critically Ill Patients

Supplemental material, sj-docx-2-jic-10.1177_08850666221098383 for Relationship Between Glucose Time in Range in Diabetic and Non-Diabetic Patients and Mortality in Critically Ill Patients by Mahmoud A. Ammar, Abdalla A. Ammar, Timothy Wee, Ranjit Deshpande, Matthew Band and Shamsuddin Akhtar in Journal of Intensive Care Medicine

Supplemental Material

sj-docx-3-jic-10.1177_08850666221098383 - Supplemental material for Relationship Between Glucose Time in Range in Diabetic and Non-Diabetic Patients and Mortality in Critically Ill Patients

Supplemental material, sj-docx-3-jic-10.1177_08850666221098383 for Relationship Between Glucose Time in Range in Diabetic and Non-Diabetic Patients and Mortality in Critically Ill Patients by Mahmoud A. Ammar, Abdalla A. Ammar, Timothy Wee, Ranjit Deshpande, Matthew Band and Shamsuddin Akhtar in Journal of Intensive Care Medicine

Supplemental Material

sj-docx-4-jic-10.1177_08850666221098383 - Supplemental material for Relationship Between Glucose Time in Range in Diabetic and Non-Diabetic Patients and Mortality in Critically Ill Patients

Supplemental material, sj-docx-4-jic-10.1177_08850666221098383 for Relationship Between Glucose Time in Range in Diabetic and Non-Diabetic Patients and Mortality in Critically Ill Patients by Mahmoud A. Ammar, Abdalla A. Ammar, Timothy Wee, Ranjit Deshpande, Matthew Band and Shamsuddin Akhtar in Journal of Intensive Care Medicine

Supplemental Material

sj-docx-5-jic-10.1177_08850666221098383 - Supplemental material for Relationship Between Glucose Time in Range in Diabetic and Non-Diabetic Patients and Mortality in Critically Ill Patients

Supplemental material, sj-docx-5-jic-10.1177_08850666221098383 for Relationship Between Glucose Time in Range in Diabetic and Non-Diabetic Patients and Mortality in Critically Ill Patients by Mahmoud A. Ammar, Abdalla A. Ammar, Timothy Wee, Ranjit Deshpande, Matthew Band and Shamsuddin Akhtar in Journal of Intensive Care Medicine

Supplemental Material

sj-docx-6-jic-10.1177_08850666221098383 - Supplemental material for Relationship Between Glucose Time in Range in Diabetic and Non-Diabetic Patients and Mortality in Critically Ill Patients

Supplemental material, sj-docx-6-jic-10.1177_08850666221098383 for Relationship Between Glucose Time in Range in Diabetic and Non-Diabetic Patients and Mortality in Critically Ill Patients by Mahmoud A. Ammar, Abdalla A. Ammar, Timothy Wee, Ranjit Deshpande, Matthew Band and Shamsuddin Akhtar in Journal of Intensive Care Medicine

Supplemental Material

sj-docx-7-jic-10.1177_08850666221098383 - Supplemental material for Relationship Between Glucose Time in Range in Diabetic and Non-Diabetic Patients and Mortality in Critically Ill Patients

Supplemental material, sj-docx-7-jic-10.1177_08850666221098383 for Relationship Between Glucose Time in Range in Diabetic and Non-Diabetic Patients and Mortality in Critically Ill Patients by Mahmoud A. Ammar, Abdalla A. Ammar, Timothy Wee, Ranjit Deshpande, Matthew Band and Shamsuddin Akhtar in Journal of Intensive Care Medicine

Footnotes

Abbreviations

Author Contributions

M.A., A.A., R.D., M.B., and S.A. conceived the idea and designed the study. M.A. and A.A acquired the data. M.A., A.A., and T.W. analyzed the data. The manuscript was drafted by M.A., A.A., T.W., R.D., M.B., and S.A. All authors contributed to the data interpretation and edited the manuscript.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Ethical Approval

Not applicable, because this article does not contain any studies with human or animal subjects.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iDs

Mahmoud A. Ammar

Timothy Wee

Supplemental Material

Supplemental material for this article is available online.

References

Van den Berghe

Wouters

Weekers

, et al. Intensive insulin therapy in critically ill patients. N Engl J Med. 2001;345(19):1359-1367.

Van den Berghe

Wilmer

Hermans

, et al. Intensive insulin therapy in the medical ICU. N Engl J Med. 2006;354(5):449-461.

Investigators

N-SS

Finfer

Chittock

, et al. Intensive versus conventional glucose control in critically ill patients. N Engl J Med. 2009;360(13):1283-1297.

Preiser

Devos

Ruiz-Santana

, et al. A prospective randomised multi-centre controlled trial on tight glucose control by intensive insulin therapy in adult intensive care units: The Glucontrol study. Intensive Care Med. 2009;35(10):1738-1748.

Brunkhorst

Engel

Bloos

, et al. Intensive insulin therapy and pentastarch resuscitation in severe sepsis. N Engl J Med. 2008;358(2):125-139.

Jacobi

Bircher

Krinsley

, et al. Guidelines for the use of an insulin infusion for the management of hyperglycemia in critically ill patients. Crit Care Med. 2012;40(12):3251-3276.

Krinsley

. Glycemic control in the critically ill: What have we learned since NICE-SUGAR? Hosp Pract (1995). 2015;43(3):191-197.

Krinsley

Preiser

. Is it time to abandon glucose control in critically ill adult patients? Curr Opin Crit Care. 2019;25(4):299-306.

Omar

Salama

Allam

, et al. Association of time in blood glucose range with outcomes following cardiac surgery. BMC Anesthesiol. 2015;15(1):14.

10.

Chase

Pretty

Pfeifer

, et al. Organ failure and tight glycemic control in the SPRINT study. Crit Care. 2010;14(4):R154.

11.

Krinsley

Preiser

. Time in blood glucose range 70 to 140 mg/dl >80% is strongly associated with increased survival in non-diabetic critically ill adults. Crit Care. 2015;19(1):179.

12.

Lanspa

Krinsley

Hersh

, et al. Percentage of time in range 70 to 139 mg/dL is associated with reduced mortality among critically ill patients receiving IV insulin infusion. Chest. 2019;156(5):878-886.

13.

Shetty

Inzucchi

Goldberg

Cooper

Siegel

Honiden

. Adapting to the new consensus guidelines for managing hyperglycemia during critical illness: The updated Yale insulin infusion protocol. Endocr Pract. 2012;18(3):363-370.

14.

Viasus

Pano-Pardo

Pachon

, et al. Factors associated with severe disease in hospitalized adults with pandemic (H1N1) 2009 in Spain. Clin Microbiol Infect. 2011;17(5):738-746.

15.

Rothman

. Development and validation of a continuous measure of patient condition using the electronic medical record. J Biomed Inform. 2013;46(5):837-848.

16.

Van Herpe

De Brabanter

Beullens

De Moor

Van den Berghe

. Glycemic penalty index for adequately assessing and comparing different blood glucose control algorithms. Crit Care. 2008;12(1):R24.

17.

Meyfroidt

. Blood glucose amplitude variability in critically ill patients. Minerva Anestesiol. 2015;81(9):1010-1018.

18.

Breiman

Friedman

Olshen

Stone

. Classification And Regression Trees. 1st ed. CRC Press; 1984, https://doi.org/10.1201/9781315139470.

19.

Lewis

. An introduction to classification and regression tree (CART) analysis. Available at: https://citeseerx.ist.psu.edu/viewdoc/download?doi = 10.1.1.95.4103&rep = rep1&type = pdf. Accessed October 23, 2021.

20.

Fonarow

Adams

Jr Abraham

, et al. Risk stratification for in-hospital mortality in acutely decompensated heart failure: Classification and regression tree analysis. JAMA. 2005;293(5):572-580.

21.

Ebell

Afonso

Geocadin

. American heart association's get with the guidelines-resuscitation I. Prediction of survival to discharge following cardiopulmonary resuscitation using classification and regression trees. Crit Care Med. 2013;41(12):2688-2697.

22.

Afonso

Ebell

Gonzales

Stein

Genton

Senn

. The use of classification and regression trees to predict the likelihood of seasonal influenza. Fam Pract. 2012;29(6):671-677.

23.

Robin

Turck

Hainard

, et al. pROC: An open-source package for R and S + to analyze and compare ROC curves. BMC Bioinformatics. 2011;12:77.

24.

Signal

Le Compte

Shaw

Chase

. Glycemic levels in critically ill patients: Are normoglycemia and low variability associated with improved outcomes? J Diabetes Sci Technol. 2012;6(5):1030-1037.

25.

Penning

Chase

Preiser

, et al. Does the achievement of an intermediate glycemic target reduce organ failure and mortality? A post hoc analysis of the glucontrol trial. J Crit Care. 2014;29(3):374-379.

26.

Agiostratidou

Anhalt

Ball

, et al. Standardizing clinically meaningful outcome measures beyond HbA1c for type 1 diabetes: A consensus report of the American association of clinical endocrinologists, the American association of diabetes educators, the American diabetes association, the endocrine society, JDRF international, The Leona M. And Harry B. Helmsley charitable trust, the pediatric endocrine society, and the T1D exchange. Diabetes Care. 2017;40(12):1622-1630.

27.

Sechterberger

Bosman

Oudemans-van Straaten

, et al. The effect of diabetes mellitus on the association between measures of glycaemic control and ICU mortality: A retrospective cohort study. Crit Care. 2013;17(2):R52.

28.

Krinsley

Egi

Kiss

, et al. Diabetic status and the relation of the three domains of glycemic control to mortality in critically ill patients: An international multicenter cohort study. Crit Care. 2013;17(2):R37.

29.

Klip

Tsakiridis

Marette

Ortiz

. Regulation of expression of glucose transporters by glucose: A review of studies in vivo and in cell cultures. FASEB J. 1994;8(1):43-53.

30.

Plummer

Bellomo

Cousins

, et al. Dysglycaemia in the critically ill and the interaction of chronic and acute glycaemia with mortality. Intensive Care Med. 2014;40(7):973-980.

31.

Finfer

Wernerman

Preiser

, et al. Clinical review: Consensus recommendations on measurement of blood glucose and reporting glycemic control in critically ill adults. Crit Care. 2013;17(3):229.

32.

Kompoti

Michalia

Salma

Diogou

Lakoumenta

Clouva-Molyvdas

. Glycated hemoglobin at admission in the intensive care unit: Clinical implications and prognostic relevance. J Crit Care. 2015;30(1):150-155.

33.

Egi

Bellomo

Stachowski

, et al. The interaction of chronic and acute glycemia with mortality in critically ill patients with diabetes. Crit Care Med. 2011;39(1):105-111.

34.

Kanji

Buffie

Hutton

, et al. Reliability of point-of-care testing for glucose measurement in critically ill adults. Crit Care Med. 2005;33(12):2778-2785.

35.

Vogelzang

van der Horst

Nijsten

. Hyperglycaemic index as a tool to assess glucose control: A retrospective study. Crit Care. 2004;8(3):R122-R127.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.03 MB

0.01 MB

0.18 MB

0.64 MB

0.18 MB