The Glycemic Ratio Is Strongly and Independently Associated With Mortality in the Critically Ill

Abstract

Background:

Interventional studies investigating blood glucose (BG) management in intensive care units (ICU) have been inconclusive. New insights are needed. We assessed the ability of a new metric, the Glycemic Ratio (GR), to determine the relationship of ICU glucose control relative to preadmission glycemia and mortality.

Methods:

Retrospective cohort investigation (n = 4790) in an adult medical-surgical ICU included patients with minimum four BGs, hemoglobin (Hgb), and hemoglobin A1c (HbA1c). The GR is the quotient of mean ICU BGs (mBG) and estimated preadmission BG, derived from HbA1c.

Results:

Mortality displayed a J-shaped curve with GR (nadir GR 0.9), independent of background glycemia, consistent for HbA1c <6.5% vs >6.5%, and Hgb >10 g/dL vs <10 g/dL and medical versus surgical. An optimal range of GR 0.80 to 0.99 was associated with decreased mortality compared with GR above and below this range. The mBG displayed a linear relationship with mortality at lower HbA1c but diminished for HbA1c >6.5%, and dependent on preadmission glycemia. In adjusted analysis, GR remained associated with mortality (odds ratio = 2.61, 95% confidence interval = 1.48-4.62, P = .0012), but mBG did not (1.004, 1.000-1.009, .059). A single value on admission was not independently associated with mortality.

Conclusions:

The GR provided new insight into malglycemia that was not apparent using mBG, or an admission value. Mortality was associated with acute change from preadmission glycemia (GR). Further assessment of the impact of GR deviations from the nadir in mortality at GR 0.80 to 0.99, as both relative hypo- and hyperglycemia, and as duration of exposure and intensity, may further define the multifaceted nature of malglycemia.

Keywords

glycemic ratio mortality malglycemia hyperglycemia hypoglycemia diabetes critical illness

Background

Stress-induced hyperglycemia is common in the critically ill, due to insulin resistance resulting in increased gluconeogenesis associated with increased circulating cortisol, glucagon, catecholamines, and inflammatory cytokines.¹ While stress-induced hyperglycemia is associated with increased mortality and morbidity in postoperative patients as well as those admitted for medical indications,^1-9 iatrogenic factors such as the concurrent use of dextrose-containing fluids and corticosteroids may independently contribute to hyperglycemia. This association has been reported in patients with and without diabetes, with mortality rates higher in patients without diabetes patients with new hyperglycemia than in known patients with diabetes patients with similar hyperglycemia.^2,5,8,10,11

There is evidence of a J-shaped curve for the relationship between blood glucose (BG) and outcome.^11-13 Relative hypoglycemia, described as a decrease in BG greater than or equal to 30% below prehospital admission levels, has been associated with higher mortality in the critically ill.^13,14 While definitive causality remains lacking,^15-20 observational studies recognize hyperglycemia, relative hypoglycemia, and absolute hypoglycemia (BG <70 mg/dL) as being independently associated with mortality.^2-14

Controlled studies randomizing groups to near-normal BG ranges in the critical care setting have led to beneficial effects in one study¹⁵ but not confirmed in others,^16-20 or upon meta-analysis.²¹ Changes from preadmission glycemia are more strongly associated with outcomes than BG levels.^22-31 Inadequate BG time in range and high rates of absolute hypoglycemia have contributed to this inconsistency. It is likely the unrecognized presence of relative hyperglycemia and relative hypoglycemia, in part facilitated by the sporadic nature of intensive care unit (ICU) BG monitoring, also affected outcomes. Recent findings highlight the developing understanding of acute changes from preadmission glycemia and the relationship with outcomes.³²

Changes from preadmission glycemia can be measured using the stress hyperglycemia ratio (SHR). Nathan et al³³ developed an HbA1c-derived estimation of average BG (eAG [as mg/dL] = [1.59 × HbA1c – 2.59] × 18), while Beck et al³⁴ developed a variation on this through the use of continuous glucose monitoring (eAG [as mg/dL] = [HbA1c – 3.38] / 0.02345), either of which can be used as an estimate of preadmission glycemia. The SHR is calculated as the quotient of admission BG and eAG. Anemia may affect HbA1c and it is unknown whether this significantly affects the prognostic ability of SHR.³⁵

The association of elevated SHR with mortality has been previously investigated in both the critical care setting²³ and other clinical settings,^22,24-31 and consistently displays superior prognostic insight compared with BG. However, this does not address the association of glucose control across the entirety of the ICU admission with mortality.

Malglycemia (the collective of absolute hypoglycemia, relative hypoglycemia, and hyperglycemia) is driven by both increases and decreases from preadmission glycemia across the admission.^{13,14,22,24-31} Compared with a single point at admission, a metric assessing malglycemia across the entire admission may provide superior insight into the nature of malglycemia. In recognition of the bidirectional nature of malglycemia over the course of an entire admission, we use the term Glycemic Ratio (GR). This is defined as the quotient of mean BG (mBG) during ICU admission and eAG, distinct from the SHR which refers to the quotient of the admission BG (admBG) and eAG.

We hypothesize that GR will provide greater prognostic power than does measurement of admBG or SHR, and deviation from preadmission glycemia across the length of stay (LOS) using this metric will allow deeper insight into the nature of malglycemia.

Aim

The aim of this study was to establish the relationship of the GR with in-hospital mortality.

Research Design and Methods

This observational cohort study included patients admitted to the intensive care unit (ICU) between October 11, 2011, and November 30, 2019, and was approved by the institutional review board of Stamford Hospital (Western IRB #1- 1167662-1) with no need for informed consent.

Stamford Hospital is a university-affiliated teaching hospital with a 20-bed ICU that treats a heterogeneous population of medical and surgical patients. Eligible patients required a minimum of four BG measurements during ICU stay, hemoglobin at admission, and HbA1c obtained at ICU admission or within 3 months prior to admission. Patients admitted for diabetic ketoacidosis or postcardiovascular surgery monitoring (due to their low associated mortality) were excluded.

Details of the BG target ranges, management approach, monitoring, nutritional support, data abstraction techniques, and associated instrumentation for this cohort have been described in detail elsewhere, along with details of the glucose and HbA1c assay techniques and specifications.¹⁴

Statistical Plan

Differences in attributes between groups that were measured at baseline were assessed by t tests or Wilcoxon rank-sum tests for normally distributed or skewed data, respectively, and χ² tests for proportions.

Mean BG (mBG) was calculated as the mean of all BG obtained while in ICU. It was not time-weighted due to the frequency of BG testing, which was a maximum duration of three hours between samples. The GR was calculated as the quotient of mBG and eAG, and defined preadmission glycemia in mg/dL ([1.59 × HbA1c – 2.59] × 18).³³ The SHR was calculated as the quotient of the first ICU BG (admBG) and eAG.

The primary endpoint was all-cause mortality, including mortality on the general ward after ICU discharge. We assessed risk-adjusted mortality using observed:expected mortality ratios, the quotient of hospital mortality, and the Acute Physiology and Chronic Health Evaluation IV predicted mortality (APIV PM).³⁶

The GR was assessed in increments of 0.10, with the lowest strata <0.70, and the highest ≥1.30. The mBG was grouped into categories of <100 mg/dL, increasing by 20 mg/dL for each category, to a maximum ≥180 mg/dL. Concurrent mean values for HbA1c and APIV PM were calculated for each category and their relationship to mortality and risk-adjusted mortality was assessed.

Locally Weighted Scatterplot Smoothing (Lowess) was used to describe the relative relationships between GR and mBG separated for HbA1c <6.5% and >6.5%, Hgb <10 g/dL and >10 g/dL, and for medical and surgical cohorts. Lowess modeling fits simple models to localized subsets of the data so that each outcome datum point is replaced by a smoothed value, to build up a function that describes the deterministic part of the variation in the data. This does not require specification of a global function to fit a model to the data. Each smoothed value is given by a weighted linear least squares regression over the span.

Logistic regression analysis was used to assess the relationship of mBG and GR to mortality and for risk-adjusted mortality, as was admBG and SHR. Covariates (mBG, GR, and APIV PM) were assessed for collinearity using variance inflation factors (designated threshold 5).

A P value of less than .05 (two-sided) was statistically significant. Ninety-five percent confidence intervals (95% CIs) are presented throughout. Analyses were performed using Stata 16 (Stata Corp, College Station, Texas).

Results

From an initial cohort of 9401 patients admitted to the ICU over the study period, there were 4790 eligible patients who had 145 792 BG tests obtained during their ICU stay (Supplementary Figure 1).

For the entire cohort, there were 7.6 (5.0) (mean, SD) BG tests per 24 hours and the patients had a median ICU LOS of 1.9 (1.1-3.9) days (Table 1). There were 1386 (28.9%) patients with admission Hgb <10 g/dL, and 3828 (79.9%) with HbA1c <6.5% (Table 1).

Table 1.

Patient Characteristics (Mean ± SD Unless Otherwise Stated).

	All	Survival status		HbA1c		Hemoglobin		Diagnostic group^a
	All	Died	Survived	HbA1c < 6.5	HbA1c ≥ 6.5	HgB < 10	Hgb ≥ 10	Medical	Surgical
N (%)	4790	604 (12.6)	4186 (87.4)	3828 (79.9)	962 (20.1)	1386 (28.9)	3404 (71.1)	3548 (74.1)	975 (20.4)
HbA1c, %	6.0 ± 1.4	6.0 ± 1.2	6.0 ± 1.4	5.5 ± 0.5	8.09 ± 1.8*	5.9 ± 1.3	6.1 ± 1.4*	6.0 ± 1.4	6.0 ± 1.3
Age, years	65.4 ± 17.7	72.3 ± 15.4	64.4 ± 17.8*	64.7 ± 18.5	68.3 ± 13.9*	68.1 ± 16.3	64.3 ± 18.1*	65.5 ± 17.6	65.3 ± 18.1
Male, n (%)	2824 (59.0)	351 (58.1)	2472 (59.1)	2282 (59.6)	541 (56.2)	730 (52.7)	2094 (61.5)	2126 (59.9)	526 (53.9)*
LOS, d, median (IQR)	1.9 (1.1-3.9)	2.8 (1.4-6.5)	1.9 (1.0-3.5)*	2.0 (1.1-4.0	1.8 (1.0-3.5)	2.4 (1.3-5.1)	1.8 (1.0-3.4)*	2.1 (1.2-4.3)	1.8 (1.0-3.2)*
GFR, mL/min/1.73 m²	78 ± 48	59 ± 43	81 ± 47*	82 ± 48	64 ± 41	65 ± 48	84 ± 46*	75 ± 49	83 ± 43*
GFR < 30, n (%)	790 (16.5)	180 (29.8)	610 (14.6)*	558 (14.6)	232 (24.1)	391 (28.2)	399 (11.7)*	661 (18.6)	104 (10.7)*
Hemoglobin, g/dL	11.5 ± 2.5	10.7 ± 2.5	11.6 ± 2.4*	11.5 ± 2.5	11.4 ± 2.4	8.6 ± 1.0	12.7 ± 1.8*	11.6 ± 2.5	11.0 ± 2.2*
APIV PM, median (IQR)	10 (3-26)	52 (27-81)	8 (3-19)*	2 (1-4)	11 (4-30)*	18 (7-41)	7 (2-21)*	12 (4-30)	6 (2-15)*
Prior insulin	449 (9.4)	52 (8.7)	397 (9.8)	114 (3.0)	335(34.8)*	186 (13.4)	263 (7.7)*	332 (9.4)	96 (9.8)
Death, n (%)	605 (12.6)	—	—	476(12.4)	129 (13.4)	247 (17.8)	358 (10.5)*	520 (14.7)	73 (7.5)*
Med/Surg (%)	77.8/22.2	85.9/14.1	72.4/27.6*	73.5/26.5	76.2/23.8	76.7/23.3	78.2/21.8	—	—
Ventilated, n (%)	1693 (35.3)	428 (70.9)	1256 (30.2)*	1371(35.8)	322 (33.5)	591 (42.6)	1102 (32.4)*	1334 (37.6)	329 (3.7)*
Admission BG, mg/dL	146 ± 69	160 ± 79	144 ± 67*	130 ± 45	210 ± 104*	152 ± 75	144 ± 67*	148 ± 74	144 ± 52
Mean BG, mg/dL	128 ± 30	138 ± 36	127 ± 29*	120 ± 23	158 ± 36*	132 ± 30	126 ± 30*	127 ± 31	131 ± 27*
SHR (admission)	1.18 ± 0.43	1.28 ± 0.54	1.16 ± 0.41*	1.19 ± 0.40	1.15 ±.50*	1.25 ± 0.47	1.15 ± 0.40*	1.18 ± 0.45	1.18 ± 0.36
Mean GR	1.06 ± 0.24	1.13 ± 0.32	1.04 ± 0.23*	1.10 ± 0.23	0.89 ± 0.23*	1.11 ± 0.27	1.03 ± 0.223*	1.05 ± 0.25	1.09 ± 0.23*
Hypoglycemia, n (%)	658 (13.7)	168 (27.8)	490 (11.7)	503 (13.1)	155 (16.1)	279 (20.1)	379 (11.1)	548 (15.4)	110 (8.9)
CV (%)	20 ± 10	24 ± 11	19 ± 10*	18 ± 9	27 ± 12*	21 ± 10	19 ± 10*	20 ± 10	19 ± 9*
No. BGs, median (IQR)	14 (7-32)	30 (12-67)	13 (7-27)*	13 (7-30)	16 (8-41)*	18 (9-48)	12 (7-26)*	15 (8-37)	12 (7-28)*

Hypoglycemia defined as experiencing BG <70 mg/dL during the time in ICU.

Abbreviations: LOS, length of stay in ICU; IQR, interquartile range; GFR, glomerular filtration rate; APIV PM, Apache IV Predicted Mortality; BG, blood glucose; SHR, stress hyperglycemia ratio; GR, glycemic ratio; CV, coefficient of variation; ICU, intensive care unit.

267 patients classified as trauma.

P value < .05.

There was a J-shaped relationship between GR with both mortality and risk-adjusted mortality, with the lowest mortality evident for those with GR 0.80 to 0.99 (Figure 1a and 1c). The GR <0.70 category consisted primarily of those with markedly elevated HbA1c (mean HbA1c 9.9%). As the GR category increased, the mean HbA1c for each category gradually decreased, culminating with a mean HbA1c of 5.2% in those with GR ≥1.30, but with an associated mBG of 153 mg/dL. Preadmission glycemia was also calculated using the equation from Beck et al³⁴ and relationships between GR and mortality remained unaffected (Supplementary Figure 2).

Figure 1.

(a) Mean glycemic ratio and mortality. (b) Mean blood glucose and mortality. (c) Mean glycemic ratio and risk-adjusted mortality. (d) Mean blood glucose and risk-adjusted mortality. Risk-adjusted mortality calculated as actual mortality divided by Apache IV predicted mortality.

The relationships between mBG with mortality (Figure 1b) and mBG with risk-adjusted mortality (Figure 1d) were largely linear when graphed categorically. In contrast to GR, the HbA1c associated with the lower categories of mBG was lowest, increasing from a mean HbA1c 5.4% in those with mBG <100 mg/dL to 8.4% in those with mBG ≥180 mg/dL.

Lowess modeling also indicated a J-shaped relationship of GR with mortality with a nadir of 0.90 (Figure 2a), followed by a linear relationship between GR and mortality above this point. The GR was associated with higher mortality rates compared with SHR. Lowess modeling revealed a J-shaped relationship between mBG and mortality with a nadir at 100 mg/dL, and a linear relationship above this. There was little change in mortality between admBG and mBG.

Figure 2.

(a) Relationship between admission and mean GR and BG with mortality. (b) Relationship between mean GR and BG with mortality separated for HbA1c <6.5% and HbA1c ≥6.5%. (c) Relationship between mean GR and BG with mortality separated for Hgb <10 g/dL and Hgb ≥10 g/dL. (d) Relationship between mean GR and BG with mortality separated for medical and surgical cohorts. Abbreviations: GR, glycemic ratio; BG, blood glucose; SHR, admission value; mBG, mean blood glucose; admBG, admission blood glucose.

The relationship between GR and mortality was not influenced by HbA1c (Figure 2b). Those with HbA1c ≥6.5% had an identical GR-mortality response curve that sat marginally higher than the curve for those with HbA1c <6.5%, consistent with the differing APIV PM for each. By contrast, there was no synchrony between the mBG-mortality curves separated for HbA1c above or below 6.5%.

Figure 2c indicates the relationship between GR and mortality was similar for patients with Hgb <10 g/dL and ≥10 g/dL, with a marginal displacement of the GR-mortality curve to the right for those with Hgb <10 g/dL. By contrast, there was a major displacement in the mBG-mortality relationship for those with Hgb <10 g/dL and above 10 g/dL.

The relationship between medical and surgical patients with mortality was similar for both GR and BG (Figure 2d). Surgical patients had lower mortality associated with any given GR or BG value compared with medical patients, consistent with the difference in APIV IV (6 vs 12).

Using multivariable regression analysis, neither SHR nor admBG was independently associated with mortality (Table 2). In the total cohort, GR, mBG, and APIV PM all remained independently associated with mortality after adjustment. Separate analysis was performed on the cohort of patients with GR above and below the nadir of 0.9. For those with GR >0.9 (n = 3662), GR remained associated with mortality (odds ratio [OR] = 2.61, 95% CI = 1.48-4.62, P = .0010), but mBG did not (OR = 1.004, 95% CI = 1.000-1.009, P = .059).

Table 2.

Univariate and Multivariate Relationships With Mortality.

	Univariate		Multivariate
	OR (95% CI)	P	OR (95% CI)	P
At admission (n = 4790)
SHR	1.74 (1.46-2.08)	<.0001	1.04 (0.73-1.48)	.83
Admission BG (mg/dL)	1.003 (1.002-1.004)	<.0001	0.999 (0.997-1.001)	.46
APIV PM	1.056 (1.052-1.060)	<.0001	1.056 (1.052-1.060)	<.0001
Entire admission (n = 4790)
Mean GR	3.86 (2.81-5.32)	<.0001	1.82 (1.16-2.85)	.0097
Mean BG (mg/dL)	1.010 (1.008-1.013)	<.0001	1.006 (1.002-1.010)	.0078
APIV PM	1.056 (1.052-1.060)	<.0001	1.055 (1.051-1.059)	<.0001
mGR > 0.9 (n = 3622)
Mean GR	5.80 (3.92-8.61)	<.0001	2.60 (1.47-4.60)	.0010
Mean BG (mg/dL)	1.012 (1.009-1.015)	<.0001	1.004 (0.999-1.009)	.0583
APIV PM	1.054 (1.050-1.059)	<.0001	1.053 (1.049-1.058)	<.0001
mGR < 0.9 (n = 1168)
Mean GR	0.104 (0.021-0.530)	.0064	0.29 (0.03-2.64)	.27
Mean BG (mg/dL)	1.003 (0.997-1.008)	.42	0.998 (0.990-1.007)	.70
APIV PM	1.062 (1.053-1.071)	<.0001	1.062 (1.053-1.071)	<.0001

Abbreviations: OR, odds ratio; CI, confidence interval; SHR, stress hyperglycemia ratio; BG, blood glucose; APIV PM, Apache IV Predicted Mortality; GR, glycemic ratio.

There was no interaction between GR, mBG, and APIV PM (respective variance inflation factors mBG 1.19, GR 1.18, and APIV PM 1.03).

When the identified optimal GR range of 0.80 to 0.99 was utilized as a reference range, the adjusted odds ratios for mortality in categories above and below this target range were significantly higher (Figure 3).

Figure 3.

Mortality odds ratios for glycemic ratio categories outside 0.8 to <1.0 reference range.

Discussion

This investigation establishes a new metric, the GR, the quotient of the mean BG during the entire ICU admission and estimated preadmission glycemia. The salient finding is the identification of a clear J-shaped relationship of GR with in-hospital mortality for the entire cohort of patients, with an optimal GR range of 0.80 to 0.99. This relationship remained intact regardless of HbA1c <6.5% versus ≥6.5%, Hgb <10 g/dL versus ≥10 g/dL, and medical versus surgical admission category. Relationships remained unchanged regardless of whether the Nathan et al³³ approach or the Beck et al³⁴ approach was used to calculate preadmission glycemia and hence GR. By contrast, the relationship of mBG with mortality was not consistent, with marked discrepancies in the relationship with mortality when separated for HbA1c and Hgb. A single measurement at ICU admission enabled no prognostic insight using either admBG or SHR.

An optimal target range of GR 0.80 to 0.99 was established. The GR below this identified patients exposed to relative hypoglycemia (GR <0.70), a cohort consisting of those with elevated HbA1c (mean HbA1c 9.9%). The right-hand side of the J-curve (GR ≥1.0) identified relative hyperglycemia, largely in patients with low HbA1c, often in conjunction with mBG below 180 mg/dL. Further investigation is needed to determine the association of deviations above and below this range with mortality, including the duration and severity of exposure.

In the cohort of those above the GR nadir of 0.90, GR, but not BG, was associated with mortality. Moreover, those with GR ≥1.30 experienced the highest mortality rate despite an associated mBG of only 153 mg/dL. Increased mortality associated with stress-induced hyperlycemia is related to the acute increase from preadmission glycemia, rather than the BG level. In the cohort with GR <0.9, GR was associated with outcome only in the unadjusted model. The impact of hypoglycemia was driven by two elements, namely, absolute (<70 mg/dL) and relative (GR <0.70) hypoglycemia.

Importantly, while GR remained independent of HbA1c, mBG did not. The relationship between GR and mortality was largely unaffected by the presence of Hgb < 10 g/dL, or for medical or surgical origins.

Relationship to Prior Literature

The increasing mortality above the GR nadir of 0.90 occurred primarily in those with HbA1c <6.5%. This is explained by the detection of ongoing exposure to relative hyperglycemia using GR—an acute clinically significant increase from preadmission glycemia but remaining below the conventionally recognized threshold of 180 mg/dL. These results confirm previous reports in non-ICU populations.²¹ The mBG associated with each of the GR categories above GR 1.0, including the GR ≥1.3 category, remained well below 180 mg/dL.

Previous studies have reported a lack of association between increasing BG and outcome in those with diabetes and/or those with HbA1c ≥6.5%.^2,5,8,10,11 In this study, in those with HbA1c ≥6.5%, Lowess modeling indicated no association of hyperglycemia with mortality until mBG exceeded 200 mg/dL.

The existence of relative hypoglycemia, and the restriction of its impact to those with HbA1c ≥8%, has previously been identified in this cohort using a categorical approach and by others.^13,14 While GR identified those at risk of relative hypoglycemia, conventional use of BG did not.

The use of SHR (a single measurement at ICU admission) has previously been associated with mortality but this was not observed in this study.²³ The rapid and aggressive glucose intervention of hyperglycemia and hypoglycemia at the onset of critical care admission may lead to a disconnect between admission BG markers and markers that subsequently assess BG management over the course of the admission. Numerous other factors such as nutrition, steroid use, pressor use, and the natural decline in the acute physiology will also contribute. The continuous exposure to the differing forms of malglycemia over the entire course of the admission will continue to influence outcome. This highlights the importance of the utility of a marker such as the GR that provides insight into ongoing exposure to malglycemia during this period.

The J-shaped curve relationship between GR and mortality was similar for patients with Hb above and below 10 g/dL, with the curve displaced to the right-hand side by a factor of GR 0.10 for those with Hg <10 g/dL, consistent with a slightly lowered HbA1c. Outside long-term BG control, a number of factors may influence HbA1c and hence accuracy of GR. Anemia may spuriously increase or decrease concurrent HbA1c values, and other factors, both genetic and nongenetic, may also affect HbA1c.^37,38 A lower Hgb in the critical care setting may also be driven by acute blood loss, which is not expected to result in altered HbA1c.

Strengths and Limitations

We have examined both unadjusted and adjusted mortality in a variety of graphical and analytical approaches. This allowed a broad and balanced perspective into the prognostic ability of the GR. The high-frequency three-hourly monitoring enabled deeper insight into glycemic patterns.

These findings are hypothesis-generating and variables not considered in the analysis such as body mass index (BMI) and scoring bias across ethnicities may have had created minor bias.^39,40 Glucose control outside ICU may have affected in-hospital mortality.⁴¹ We have delineated the presence of preadmission glycemia based on HbA1c. While we have excluded those not known as having diabetes but with HbA1c >6.5%, the impact of those known as having diabetes and with HbA1c <6.5% remains unclear. We did not have access to detailed therapeutic data during the ICU stay such as insulin dosing, nutrition, use of steroids, or the cause of death, nor BG data for the general wards after ICU discharge. We have used the BG mean without time-weighting based on the very high frequency of BG monitoring. Some patients with BG instability may have received more frequent monitoring during periods of instability, potentially causing bias in mean BG.

Clinical Implications

A target range of GR 0.80 to 0.99 in the critical care population was associated with reduced mortality related to malglycemia. Variation just above or below this proposed target range was associated with a marked increase in mortality.

Exposure to hyperglycemia, relative hypoglycemia, and absolute hypoglycemia (BG <70 mg/dL) occurs at any time across the critical care admission. While some detection of patients at risk of malglycemia is enabled through the use of BG >180 mg/dL and <70 mg/dL, it excludes detection of relative hypoglycemia and relative hyperglycemia. Results of this study indicate the increased mortality associated with hyperglycemia is due to a relative increase from preadmission glycemia rather than BG. The use of GR addresses both these issues by recognizing acute changes from preadmission glycemia, including relative hyperglycemia occurring at BG <180 mg/dL, and the ongoing exposure across the admission.

The convention of reliance on BG for recognition of malglycemia and for shaping glucose control targets in the critical care setting needs to be challenged. This needs to be from a position of optimal clinical insight. There are now three recognizable forms of malglycemia in the acute hospital setting, namely, absolute hypoglycemia, relative hypoglycemia, and relative hyperglycemia. The interaction of coefficient of variation (CV) with these, and if CV is simply a marker for these different forms of malglycemia, remains unknown and requires further investigation. Hyperglycemia requires further clarification regarding the role of relative increases in preadmission glycemia and the impact of duration of exposure above and below preadmission glycemia. The nature of malglycemia must be clearly understood to interpret the generally confounding results of previous glucose interventional studies and to properly inform future studies. Further study is needed to clearly define the impact of GR deviations during the course of hospitalization, especially the duration and intensity of malglycemia exposure above and below the optimal target range of GR 0.80 to 0.99 suggested by this study.

Conclusions and Future Directions

The use of GR provided valuable new insights into the relationship between malglycemia and mortality in the critical care setting, beyond those offered by the use of mBG.

Deviations in GR across the admission, as both relative hypo- and hyperglycemia, assessed as both duration and intensity of exposure, require further study. Clarifying their role in our developing understanding of malglycemia will improve our ability to better construct and interpret glucose management interventional studies.

Supplemental Material

sj-docx-1-dst-10.1177_19322968221124114 – Supplemental material for The Glycemic Ratio Is Strongly and Independently Associated With Mortality in the Critically Ill

Supplemental material, sj-docx-1-dst-10.1177_19322968221124114 for The Glycemic Ratio Is Strongly and Independently Associated With Mortality in the Critically Ill by Greg Roberts, James S. Krinsley, Jean-Charles Preiser, Stephen Quinn, Peter R. Rule, Michael Brownlee, Michael Schwartz, Guillermo E. Umpierrez and Irl B. Hirsch in Journal of Diabetes Science and Technology

Supplemental Material

sj-docx-2-dst-10.1177_19322968221124114 – Supplemental material for The Glycemic Ratio Is Strongly and Independently Associated With Mortality in the Critically Ill

Supplemental material, sj-docx-2-dst-10.1177_19322968221124114 for The Glycemic Ratio Is Strongly and Independently Associated With Mortality in the Critically Ill by Greg Roberts, James S. Krinsley, Jean-Charles Preiser, Stephen Quinn, Peter R. Rule, Michael Brownlee, Michael Schwartz, Guillermo E. Umpierrez and Irl B. Hirsch in Journal of Diabetes Science and Technology

Footnotes

Abbreviations

admBG, admission blood glucose; APIV IV, Acute Physiology and Chronic Health Evaluation IV predicted mortality; BG, blood glucose; eAG, estimated average glucose; GR, glycemic ratio; HbA1c, hemoglobin A1c; Hgb, hemoglobin; ICU, intensive care unit; Lowess, locally weighted scatterplot smoothing; mBG, mean blood glucose; SHR, stress hyperglycemia ratio.

Declaration of Conflicting Interests

The author(s) declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: JSK and J-CP disclosed that they are advisory board members and consultants for Edwards Life Sciences, OptiScan Biomedical, and Roche Diagnostics. IBH’s institution received funding from Insulet and Beta Bionics; he disclosed that he is a consultant for Abbott, Roche, LifeScan, and GWave. GEU’s institution received funding from Dexcom, Abbott, and Baxter. The remaining authors have disclosed that they do not have any potential conflicts of interest.

Funding

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

ORCID iDs

Greg Roberts

James S. Krinsley

Jean-Charles Preiser

Guillermo E. Umpierrez

Irl B. Hirsch

Supplemental Material

Supplemental material for this article is available online.

References

Dungan

Braithwaite

Preiser

JC.

Stress hyperglycaemia. Lancet. 2009;373:1798-1807.

Umpierrez

Isaacs

Bazargen

You

Thaler

Kitabchi

AE.

Hyperglycaemia: an independent marker of in-hospital mortality in patients with undiagnosed diabetes. J Clin Endocrinol Metab. 2002;87:978-982.

Krinsley

Association between hyperglycaemia and increased hospital mortality in a heterogeneous population of critically ill patients. Mayo Clin Proc. 2003;78:1471-1478.

Baker

Janaway

Philips

, et al. Hyperglycaemia is associated with poor outcomes in patients admitted to hospital with acute exacerbations of chronic obstructive pulmonary disease. Thorax. 2006;61:284-289.

Roberts

Sires

Chen

, et al. A comparison of the stress hyperglycemia ratio, glycemic gap, and glucose to assess the impact of stress-induced hyperglycemia on ischemic stroke outcome. J Diabetes. 2021;13(12):1034-1042. doi:10.1111/1753-0407.13223.

Iglesias

Polini

Muñoz

, et al. Fasting hyperglycaemia and in-hospital mortality in elderly population. Int J Clin Pract. 2011;65:308-313.

Barsheshet

Garty

Grossman

, et al. Admission blood glucose level and mortality among hospitalized nondiabetic patients with heart failure. Arch Intern Med. 2006;166:1613-1619.

Capes

Hunt

Malmberg

Gerstein

HC.

Stress hyperglycaemia and increased risk of death after myocardial infarction in patients with and without diabetes: a systematic overview. Lancet. 2000;355:773-778.

Vriesendorp

Morélis

Devries

Legemate

Hoekstra

JB.

Early post-operative glucose levels are an independent risk factor for infection after peripheral vascular surgery. Eur J Vasc Endovasc Surg. 2004;28(5):520-525.

10.

Krinsley

Egi

Kiss

, et al. Diabetic status and the relationship of the 3 domains of glycemic control to mortality in critically ill patients: an international multi-center cohort study. Crit Care. 2013;17:R37.

11.

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:R52.

12.

Ntaios

Egli

Faouzi

Michel

J-shaped association between serum glucose and functional outcome in acute ischemic stroke. Stroke. 2010;41:2366-2370.

13.

Kwan

Zwakman-Hessels

Marhoon

, et al. Relative hypoglycemia in diabetic patients with critical illness. Crit Care Med. 2020;48(3):e233-e240. doi:10.1097/CCM.0000000000004213.

14.

Krinsley

Rule

Roberts

, et al. Relative hypoglycemia and lower hemoglobin A1c-adjusted time in band are strongly associated with increased mortality in critically ill patients. Crit Care Med. 2022;50:e664-e673. doi:10.1097/CCM.0000000000005490.

15.

Van den Berghe

Wouters

Weekers

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

16.

Van den Berghe

Wilmer

Hermans

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

17.

NICE-SUGAR Study Investigators. Intensive versus conventional glucose control in critically ill patients. N Engl J Med. 2009;360:1283-1297.

18.

Johnston

Bruno

Pauls

, et al; for the Neurological Emergencies Treatment Trials Network and the SHINE Trial Investigators. Intensive vs standard treatment of hyperglycemia and functional outcome in patients with acute ischemic stroke the SHINE randomized clinical trial. JAMA. 2019;322:326-335.

19.

Bohé

Abidi

Brunot

, et al. Individualised versus conventional glucose control in critically-ill patients: the CONTROLING study-a randomized clinical trial. Intensive Care Med. 2021;47(11):1271-1283. doi:10.1007/s00134-021-06526-8.

20.

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.

21.

Marik

Oreiser

Toward understanding tight glycemic control in the ICU: a systematic review and metaanalysis. Chest. 2010;137:544-551.

22.

Roberts

Quinn

Valentine

, et al. Relative hyperglycemia a marker of critical illness: introducing the stress hyperglycemia ratio. J Clin Endocrinol Metab. 2015;100:4490-4497.

23.

Lee

Drake

Roberts

, et al. Relative hyperglycemia is an independent determinant of in-hospital mortality in patients with critical illness. Crit Care Med. 2020;48(2):e115-e122. doi:10.1097/CCM.0000000000004133.

24.

Zhao

Zhang

T-Y

Cheng

Y-J

, et al. Prognostic significance of relative hyperglycemia after percutaneous coronary intervention in patients with and without recognized diabetes. Curr Vasc Pharmacol. 2021;19(1):91-101.

25.

Chen

Wen

, et al. Association between stress hyperglycemia ratio and in-hospital outcomes in elderly patients with acute myocardial infarction. Front Cardiovasc Med. 2021;8:698725. doi:10.3389/fcvm.2021.698725.

26.

Gao

Liu

Chen

Predictive value of stress hyperglycemia ratio for the occurrence of acute kidney injury in acute myocardial infarction patients with diabetes. BMC Cardiovasc Disord. 2021;21:157. doi:10.1186/s12872-021-01962-2.

27.

Meng

Zhu

Liu

, et al. The stress hyperglycaemia ratio is associated with left ventricular remodelling after first acute ST-segment elevation myocardial infarction. BMC Cardiovasc Disord. 2021;21:72. doi:10.1186/s12872-021-01889-8.

28.

Di Luzio

Dusi

Mazzotti

Petroni

Marchesini

Bianchi

. Stress hyperglycemia and complications following traumatic injuries in individuals with/without diabetes: the case of orthopedic surgery. Diabetes Metab Syndr Obes. 2020;13:9-17.

29.

Yang

Liao

Tang

, et al. Glycated hemoglobin A1c-based adjusted glycemic variables in patients with diabetes presenting with acute exacerbation of chronic obstructive pulmonary disease. Int J Chron Obstruct Pulmon Dis. 2017;12:1923-1932.

30.

Marenzi

Cosentino

Milazzo

, et al. Acute kidney injury in diabetic patients with acute myocardial infarction: role of acute and chronic glycemia J Am Heart Assoc. 2018;7:e008122. doi:10.1161/JAHA.117.008122.

31.

Yuan

Chen

Ruan

, et al. The stress hyperglycemia ratio is associated with Hemorrhagic Transformation in patients with acute ischemic stroke. Clin Interv Aging. 2021;16:431-442.

32.

Krinsley

Rule

Pappy

, et al. The interaction of acute and chronic glycemia on the relationship of hyperglycemia glucose variability to mortality in the critically ill. Crit Care Med. 2020;48:1744-1751.

33.

Nathan

Kuenen

Borg

, et al; A1c-Derived Average Glucose Study Group. Translating the A1C assay into estimated average glucose values. Diabetes Care. 2008;31:1473-1478.

34.

Beck

Bergenstal

Cheng

, et al. The relationships between time in range, hyperglycemia metrics, and HbA1c. J Diabetes Sci Technol. 2019;13(4):614-626.

35.

English

Idris

Smith

Dhatariya

Kilpatrick

John

WG.

The effect of anaemia and abnormalities of erythrocyte indices on HbA1c analysis: a systematic review. Diabetologia. 2015;58(7):1409-1421. doi:10.1007/s00125-015-3599-3.

36.

Zimmerman

Kramer

McNair

Malila

FM.

Acute Physiology and Chronic Health Evaluation (APACHE) IV: hospital mortality assessment for today’s critically ill patients. Crit Care Med. 2006;34(5):1297-1310.

37.

Leong

Wheeler

Genetics of HbA1c: a case study in clinical translation. Curr Opin Genet Dev. 2018;50:79-85.

38.

Nitin

HbA1c and factors other than diabetes mellitus affecting it. Singapore Med J. 2010;51(8):616-622.

39.

Deliberato

Serpa Neto

Komorowski

, et al. An evaluation of the influence of body mass index on severity scoring. Crit Care Med. 2019;47(2):247-253. doi:10.1097/CCM.0000000000003528.

40.

Sarkar

Martin

Mattie

Gichoya

Stone

Celi

LA.

Performance of intensive care unit severity scoring systems across different ethnicities in the USA: a retrospective observational study. Lancet Digit Health. 2021;3(4):e241-e249. doi:10.1016/S2589-7500(21)00022-4.

41.

Krinsley

Maurer

Holowinski

, et al. Glucose control, diabetes status and mortality in critically ill patients: the continuum from ICU admission to hospital discharge. Mayo Clin Proc. 2017;92:1019-1029.

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

0.30 MB