Glycaemic Variability and Its Outcome in Intensive Care Unit Patients with Sepsis

Abstract

Background:

Sepsis causes an uncontrollable activation of pro- and anti-inflammatory responses, leading to metabolic derangements, notably glucose variability (GV). Both hyperglycaemia and hypoglycaemia can occur in septic patients, regardless of the diabetes status. Hyperglycaemia control using insulin has been shown to reduce morbidity and mortality. Current guidelines suggest maintaining glucose levels between 140 and 180 mg/dL for better clinical outcomes. Managing glycaemic variability is crucial in reducing mortality in intensive care unit (ICU) patients with sepsis.

Materials and Methods:

The study included patients aged ≥ 18 years, admitted for ≥ 24 h, focussing on the first 5 days of ICU stay. Data on patient characteristics, glucose values, comorbidities, organ failures, and outcomes were collected. GV was assessed, and comorbidities were determined using diagnostic codes.

Results:

Among a group of 100 patients (mean age 54.16 ± 18.5 years; 66% male), diabetes mellitus (84%) and hypertension (57%) were the most common comorbidities. Pneumonia (25%) and urosepsis (22%) were the primary sources of sepsis. Patients with multiple organ dysfunction syndrome (MODS) had significantly higher mean glucose levels (MGLs) than those without (P < 0.05). Higher glucose levels were also observed in non-survivors compared to survivors (P < 0.05). Glycaemic variability, measured by the coefficient of variation, was significantly higher in non-survivors. Insulin requirements were higher in unresolved cases (P < 0.05).

Conclusion:

Higher glycaemic variability and MGLs were associated with increased mortality and MODS in ICU patients with sepsis. Improved outcomes were observed in patients with lower glycaemic variability, highlighting the need for insulin protocols to maintain optimal glucose control and reduce variability in critical care settings.

Introduction

Sepsis, a dysregulated host response to infection leading to organ dysfunction, remains a major global healthcare challenge, with significant morbidity and mortality rates, particularly in critically ill patients requiring intensive care unit (ICU) admission.[1,2] Despite advances in medical care, sepsis continues to impose a substantial burden on healthcare systems worldwide, necessitating ongoing research efforts to improve management strategies and outcomes.

Glycaemic control, the regulation of blood glucose levels within a physiological range, is a cornerstone of critical care management, particularly in patients with sepsis.[3] Both hyperglycaemia and hypoglycaemia have been associated with adverse outcomes in sepsis, prompting clinicians to strive for tight glycaemic control within a narrow range. However, recent attention has shifted towards glycaemic variability (GV), defined as fluctuations in blood glucose levels over time, as a potential independent predictor of outcomes in critically ill patients.[3,4,5]

The variation in blood glucose levels can arise from various factors, including pre-existing diabetes or stress-induced hyperglycaemia.[6,7,8] Extensive research has consistently linked hyperglycaemia with increased mortality rates in critically ill patients.[6,7,8] Glycaemic Variability encompasses various metrics, including standard deviation (SD), coefficient of variation (CV), and measures derived from continuous glucose monitoring (CGM). These metrics capture the dynamic nature of blood glucose fluctuations, which may reflect underlying pathophysiological processes such as stress response, hormonal fluctuations, and alterations in insulin sensitivity and secretion.[4] The pathophysiological mechanisms underlying glycaemic variability in sepsis are multifactorial. Sepsis triggers the release of a cascade of inflammatory mediators, including cytokines, chemokines, and reactive oxygen species, which contribute to insulin resistance and impaired glucose utilisation. Concurrently, stress hormones such as cortisol and catecholamines are released, leading to increased hepatic glucose production and peripheral insulin resistance. These hormonal and metabolic alterations result in dysregulated glucose homeostasis and contribute to glycaemic variability in septic patients.[4,5] The implications of glycaemic variability in septic patients admitted to the ICU have garnered increasing attention in recent years, with studies exploring its association with clinical outcomes such as mortality, length of stay, and development of organ dysfunction. However, the literature remains heterogeneous, with conflicting findings and methodological limitations complicating interpretation and generalisability.

This study aims to clarify the association between glycaemic variability and clinical outcomes in ICU patients with sepsis. By reviewing existing literature and analysing relevant data, we seek to provide insights that can inform clinical practice and guide future research on ICU and in-hospital mortality in septic patients.

Aim of the study

Glucose variability (GV) rather than the glucose level has also been shown to be an important factor associated with mortality, morbidity, and improvement outcomes in patients with sepsis under ICU care. This study aims to determine the association between GV and its outcome in sepsis patients in the ICU.

Materials and Methods

Study design

A prospective observational study.

Study setting

Conducted in the ICU setting at Meenakshi Medical College Hospital and Research Institute, Kanchipuram, a tertiary care hospital.

Study period

November 2022 to April 2024.

Inclusion criteria

• (1)Patients >18 years of age.
• (2)
  Patients with sepsis admitted to the ICU with at least two of the sepsis criteria.

  • (I)
    Temperature <35°C or > 38°C.
  • (II)
    Respiratory rate >20 cpm.
  • (III)
    White cell count >12 or <4 × 10[9] cells/L.
  • (IV)
    Altered mental status.
• (3)Patients who have been admitted in the ICU for >24 h.
• (4)Patients who have given informed consent to participate in the study.

Exclusion criteria

• (1)Patients </= 18 years of age.
• (2)Patients with sepsis admitted in the ward.
• (3)Patients who have not given consent to participate in the study.

Sampling

The validated method was used to identify the patient and their record consistent with sepsis. Subjects were excluded if they were <18 years and duration of hospital stay <24 h. Only first 5 days of ICU stay is included in the analysis. Data on all patient characteristics, resources use, outcomes data, glucose values, and management were collected.

Comorbidities and organ failures are determined through the validated method using diagnosis and procedural codes. Insulin protocols were followed at the discretion of the doctor and individual variations in approaching patients with sepsis.

Mean blood glucose level (MGL) was calculated as the arithmetic mean of all recorded glucose values for the given patient. SD of MGL and CV of glucose (derived as a percentage of SD to mean blood glucose) were calculated for each patient, as described earlier. The primary endpoint for the analysis was an all-cause hospital mortality (defined as death before hospital discharge).

Statistical analysis

After data collection, data were entered in Excel. Data analysis was done with the help of statistical software GraphPad InStat v3.0. Categorical data are shown as absolute numbers and percentage, whereas continuous variables are shown as mean ± SD or median and interquartile range (IQR). CV of glucose (SD/MGL, [%]) was derived for each patient. Prior to statistical analysis, the normality of the data was examined using the Shapiro–Wilk test. The unpaired t test was utilised to compare continuous variables that were normally distributed, whereas the Mann–Whitney U test was employed for comparison of variables that were not. Either the Fisher’s exact test or the chi-square test was used to analyse categorical data. A P value of less than 0.05 was considered to be statistically significant.

Results

The study included 100 patients with a mean age of 54.16 ± 18.5 years, ranging from 20 to 89 years. The majority of patients were in the age group of 30–39 years (18%), followed by those aged 40–49 years (17%). There was male preponderance, with 66% of patients being male and 34% female. The mean age for males was 56.29 ± 18.3 years, while for females, it was 50.02 ± 18.1 years; however, this difference was not statistically significant (P = 0.11 on the unpaired t test).

In terms of presenting symptoms, the most common were fever (96%), burning micturition (84%), and cough with expectoration (79%). A detailed history was also taken to assess comorbidities, with diabetes mellitus (84%) being the most prevalent, followed by systemic hypertension (57%) and coronary artery disease (26%) [Figure 1].

Figure 1.

Distribution of comorbidities

Regarding personal history, 14 patients reported alcohol consumption, while 26 had a history of smoking. Diabetic ketoacidosis was noted in six patients, and hyperosmolar hyperglycaemic state was observed in one. The most frequent source of sepsis identified was pneumonia (25%), followed by urosepsis (22%).

Patient outcomes were categorised into various groups, with details provided in Table 1, which summarises the different outcomes encountered in this cohort.

Table 1.

Outcomes encountered by sepsis patients

Outcomes encountered	Resolved	MODS/sepsis	Death
Yes (n)	58	35	7
No (n)	42	65	93

Glycated hemoglobin (HbA1c) levels and patient outcomes

In assessing HbA1c levels in relation to patient outcomes, there was no statistically significant difference in mean HbA1c levels between cases that resolved versus unresolved cases, nor between patients with or without multiple organ dysfunction syndrome (MODS) or sepsis (P > 0.05 on unpaired t test). However, a significant difference was observed when comparing HbA1c levels between survivors and non-survivors. Non-survivors had higher HbA1c levels (9 ± 2.2) compared to those who survived (7.8 ± 1.4; P = 0.03 on unpaired t test), suggesting that elevated HbA1c levels may be a risk factor for death in patients with sepsis [Table 2].

Table 2.

HbA1c and its outcomes in patients studied

HbA1c value	Mean ± SD	Coefficient of variation of HbA1c	P value (unpaired t test)
Resolved (n = 58)	7.63 ± 1.4	0.18	0.07 (non-significant)
Unresolved (n = 42)	8.19 ± 1.7	0.2
Multiple organ dysfunction syndrome (MODS)/sepsis present (n = 35)	8.1 ± 1.4	0.17	0.2 (non-significant)
MODS/sepsis absent (n = 65)	7.7 ± 1.6	0.2
Death in patient (n = 7)	9 ± 2.2	0.24	0.03* (significant)
Alive patient (n = 93)	7.8 ± 1.4	0.18

GV in patients with MODS/sepsis

Analysis of GV among patients with and without MODS/sepsis revealed significantly higher mean glucose levels (MGLs) on all days in patients with MODS/sepsis compared to those without (P < 0.05 on unpaired t test). This finding indicates that higher glycaemic values increase the risk for developing MODS or sepsis [Table 3]. Additionally, the CV of glucose levels, when compared across tertiles of MGLs, showed a significant increase in patients with MODS/sepsis compared to those without (P < 0.05 on unpaired t test). This association suggests that a higher CV of MGL is linked to a greater incidence of MODS/sepsis [Table 4, Figure 2]. A further comparison of outcomes (MODS/sepsis) with tertiles of CV of glucose based on MGL revealed a significantly higher number of patients with MGL between 150 and 200 mg/dL and CV between 15% and 30% and > 30% in those with MODS/sepsis (P < 0.001 on Fischer exact test). Additionally, a significantly higher number of patients with MGL > 200 mg/dL and CV > 15% was noted in the MODS/sepsis group (P < 0.001 on Fischer exact test) [Table 5].

Table 3.

Glycaemic variability and presence of multiple organ dysfunction syndrome (MODS)/sepsis

Days of observation	Parameters	MODS/sepsis present (n = 35)	MODS/sepsis absent (n = 65)	P value (unpaired t test)
Day 1	Mean ± SD	243.9 ± 83.5	198.5 ± 59.3	0.002* (significant)
	Coefficient of variation (CV)	0.17	0.29
Day 2	Mean ± SD	202.19 ± 35.1	183.1 ± 49.28	0.04* (significant)
	CV	0.20	0.26
Day 3	Mean ± SD	194.9 ± 42.5	179.01 ± 43.3	0.02* (significant)
	CV	0.21	0.24
Day 4	Mean ± SD	192.4 ± 23.8	176.7 ± 29.9	0.008* (significant)
	CV	0.16	0.17
Day 5	Mean ± SD	186.1 ± 31.4	172.4 ± 27.4	0.02* (significant)
	CV	0.16	0.15

Table 4.

Comparison of coefficient of variation (CV) of mean glucose level (MGL) as per the outcome

MGL	Multiple organ dysfunction syndrome (MODS)/sepsis present (n = 35) Mean CV ± SD	MODS/sepsis absent (n = 65) Mean CV ± SD	P value (unpaired t test)
MGL (< 150)	22.45 ± 8.2	8.3 ± 4.5	0.0018* (significant)
MGL (150–200)	18.8 ± 8.9	8.11 ± 4.7	0.0001* (significant)
MGL (> 200)	21.3 ± 10.6	8.25 ± 4.7	0.009 * (significant)

Figure 2.

Comparison of CV of glucose as per outcome in tertiles of MGL

Table 5.

Comparison of outcome multiple organ dysfunction syndrome (MODS/sepsis) with tertiles CV of glucose as per mean glucose

Mean glucose level (MGL)	Coefficient of variation (%)	MODS/sepsis present (n = 35)	MODS/sepsis absent (n = 65)	P value (Fischer exact)
MGL (< 150 mg/dL)	< 15	1	14	0.12 (non-significant)
	15–30	1	0
	> 30	0	0
MGL (150–200 mg/dL)	< 15	10	41	< 0.001* (significant)
	15–30	12	3
	> 30	2	0
MGL (> 200 mg/dL)	< 15	0	7	< 0.001* (significant)
	15–30	3	0
	> 30	6	0

GV in mortality outcomes

When analysing GV in relation to mortality, patients who did not survive exhibited significantly higher MGLs on all days compared to those who survived [Table 6]. Furthermore, a significant increase in the CV of glucose was noted among non-survivors compared to survivors (P < 0.05 on unpaired t test), indicating that a higher CV of MGL is associated with increased mortality [Tables 7 and 8].

Table 6.

Glycaemic variability and mortality

Days of observation	Parameters	Dead (n = 7)	Alive (n = 93)	P value (unpaired t test)
Day 1	Mean ± SD	280.69 ± 113.1	211.57 ± 68.7	0.01 * (significant)
	Coefficient of variation (CV)	0.4	0.32
Day 2	Mean ± SD	228.1 ± 38.9	188.7 ± 45.7	0.02* (significant)
	CV	0.14	0.24
Day 3	Mean ± SD	232.4 ± 53.1	184.2 ± 43.4	0.01* (significant)
	CV	0.22	0.25
Day 4	Mean ± SD	191.02 ± 57.7	181.5 ± 25.3	0.01* (significant)
	CV	0.26	0.13
Day 5	Mean ± SD	199.1 ± 32.5	170.9 ± 24.4	0.004* (significant)
	CV	0.16	0.14

Table 7.

Comparison of coefficient of variation (CV) of glucose with mortality

Mean glucose level (MGL)	Alive (n = 93) Mean CV ± SD	Dead (n = 7) Mean CV ± SD	P value (unpaired t test)
MGL (< 150)	12.4 ± 7.23	26.3 ± 0	0.03* (significant)
MGL (150–200)	10.95 ± 7.8	21.9 ± 0	0.05* (significant)
MGL (> 200)	7.6 ± 3.6	18.9 ± 7.4	0.01* (significant)

Table 8.

Comparison of coefficient of variation (CV) of glucose as per the outcome (mortality) in tertiles of mean glucose level (MGL)

MGL	CV (%)	Alive mean CV ± SD	Dead mean CV ± SD	P value (Fischer exact)
MGL (< 150 mg/dL)	< 15	15	0	0.16 (non-significant)
	15–30	1	1
	> 30	1	0
MGL (150–200 mg/dL)	< 15	54	0	< 0.04* (significant)
	15–30	12	1
	> 30	2	0
MGL (> 200 mg/dL)	< 15	6	3	0.18 (non-significant)
	15–30	0	2
	> 30	0	0

Insulin requirements

1. Resolved versus unresolved cases: Comparison of insulin requirements between resolved and unresolved cases demonstrated a statistically significant increase in insulin usage in unresolved cases on all days (P < 0.05 on Mann–Whitney U test) [Table 9].

2. Patients with and without MODS/Sepsis: Patients with MODS/sepsis required significantly higher insulin doses compared to those without MODS/sepsis across all days (P < 0.05 on Mann–Whitney U test) [Table 10].

3. Mortality outcomes: Comparison of insulin requirements between survivors and non-survivors showed a statistically significant increase in insulin usage in patients who succumbed to mortality on all days (P < 0.05 on Mann–Whitney U test) [Table 11].

Table 9.

Insulin requirement (IU/mL) in resolved/ unresolved cases

Days of observation	Parameters	Resolved (n = 58)	Unresolved (n = 42)	P value (Mann–Whitney U test)
Day 1	Median (IQR)	16 (10–20)	20 (16–30)	0.002* (significant)
Day 2	Median (IQR)	16 (10–20.5)	22 (15–30.5)	0.003* (significant)
Day 3	Median (IQR)	16 (10–22)	24 (18–32)	0.006* (significant)
Day 4	Median (IQR)	16 (10–22)	23 (17.5–30.5)	0.002* (significant)
Day 5	Median (IQR)	17 (10–22.5)	23 (16–30.5)	0.003* (significant)

Table 10.

Insulin requirement in patients with and without multiple organ dysfunction syndrome (MODS)/sepsis

Days of observation	Parameters	MODS/sepsis present (n = 35)	MODS/sepsis absent (n = 65)	P value (Mann–Whitney U test)
Day 1	Median (IQR)	22 (18–30)	16 (10–20)	0.003* (significant)
Day 2	Median (IQR)	24 (18–30)	16 (9–21)	0.0004* (significant)
Day 3	Median (IQR)	25 (20–32)	16 (10–22)	< 0.00001* (significant)
Day 4	Median (IQR)	24 (18–30)	16 (9–22)	0.0002* (significant)
Day 5	Median (IQR)	24 (20–30)	16 (8–23)	0.0002* (significant)

Table 11.

Insulin requirement in patients dead/ alive cases

Days of observation	Parameters	Dead (n = 7)	Alive (n = 93)	P value (Mann–Whitney U test)
Day 1	Median (IQR)	20 (20–50)	18 (10–25)	0.02* (significant)
Day 2	Median (IQR)	40 (20–42)	18 (10–25.5)	0.008* (significant)
Day 3	Median (IQR)	40 (18–44)	20 (12–26)	< 0.03* (significant)
Day 4	Median (IQR)	28 (28–44)	18 (11–26)	0.005* (significant)
Day 5	Median (IQR)	30 (28–44)	20 (12–26)	0.001* (significant)

These findings suggest that increased insulin requirements are associated with unresolved cases, MODS/sepsis, and mortality.

Discussion

This prospective observational study was carried out among 100 adult patients admitted to ICU with sepsis to assess the glucose variability rather than the glucose level to be an important factor associated with mortality, morbidity, and improvement outcomes in patients with sepsis under ICU care. This section describes the study results and compares-contrasts with the existing body of evidence.

Demographic details

In the current study, the mean age of the patients was 54.16 ± 18.5 years with the majority of the patients belonging to the age group 30–39 years (18%). The study showed a male preponderance (66%) in ICU admission with a diagnosis of sepsis. Assessing the co-morbidities, it was seen that diabetes mellitus (84%) was the most common disease followed by systemic hypertension. The most common source of sepsis was pneumonia in 25%, followed by urosepsis in 22%. The demographic findings of the current study are consistent with the data found in the literature. We found in this study that elderly patients had a greater risk of sepsis, which resulted in ICU hospitalisation. Ageing has been found as the single most important risk factor for encountering sepsis among the elderly. Co-morbid conditions including cancer, diabetes, obesity, and human immunodeficiency virus infection, among others, may be responsible for this. These are all far more important for elderly patients. Prior concomitant conditions such as lung or kidney disease are frequently linked to a higher risk of sepsis. However, co-morbidities by themselves are insufficient; additional elements like different medications (polypharmacy), instrumentations, malnutrition, endocrine deficiency states, and frequent hospital stays also contribute to compromising an already weakened immune.[9] Several studies including the current research found a male preponderance in sepsis. However, the exact reason for the same is still under investigation. Nevertheless, It is still unclear, if estrogen or testosterone deficiency is a factor in severely sick women’s increased survival. Testosterone has been shown to have a critical role in immunological depression that follows trauma and bleeding, ultimately leading to sepsis. Furthermore, for future therapeutic approaches in sepsis, the potential difference in the ratio of proinflammatory to anti-inflammatory mediators in males and females may be significant.[10] The current study highlighted that comorbidities like diabetes mellitus and systemic hypertension were most encountered among patients with sepsis. Numerous lines of evidence suggest that diabetes patients are more likely to become infected, have a two to six times higher chance of developing sepsis than age-matched non-diabetic individuals, and have higher rates of sepsis-related morbidity and death. In comparison to non-diabetics, diabetic individuals are also likely to have greater rates of colonization by resistant microorganisms, such as methicillin-resistant Staphylococcus aureus. These factors provide credence to the observation that diabetes is a comorbidity that septic patients are experiencing more frequently.[11]

HbA1c levels and outcomes:

When we compared the mean HbA1c levels across various outcomes, we found that there was no statistically significant difference between HbA1c levels and patients with or without MODS/sepsis, cases that were resolved or not (P > 0.05 on unpaired t test). Nonetheless, there was a statistically significant rise in the HbA1c level among patients who had experienced death (9 ± 2.2 versus 7.8 ± 1.4; P = 0.03 on unpaired t test) when comparing their levels to those of patients who are still living. This suggests that in sepsis patients, a greater HbA1C may be a risk factor for mortality.

The current investigation demonstrated a favorable correlation between greater HbA1c levels and mortality in sepsis patients, despite its inability to establish a link between HbA1c and the existence of MODS. The mean plasma glucose level for the previous three months is shown by the HbA1c test (e.g., 6% equals to 126 mg/dL, 7% to 154 mg/dL, and 8% to 183 mg/dL). Furthermore, the HbA1c level is a valid indicator of premorbid hyperglycemia since it remains unchanged when a chronic disease manifests. Sepsis patients have demonstrated endothelial glycocalyx injury because of persistent hyperglycemia. Consequently, the degree of initial glycocalyx degradation is correlated with elevated HbA1c levels, which signify persistent hyperglycemia and may subsequently promote the development of organ failure.[12]

Glycaemic variability and outcomes

Analysing the glucose variability among patients with and without MODS/ sepsis, it was seen that on all days there was a significantly increased mean glucose level in patients with MODS/sepsis compared with patients without MODS/sepsis (P < 0.05 on unpaired t test). Hence, higher glycaemic values posed a risk for the occurrence of MODS/sepsis. Additionally, from the results, we could infer that the higher CV of MGL was associated with a greater incidence of MODS/sepsis.

Analysing the glucose variability among patients who encountered death and those alive, it was seen that statistically significant increase in mean glucose levels at all days in patients who encountered mortality than the ones alive. The study results from the current research were in line with the existing evidence in the literature. GV stands for glyceamic variation, a typical stress reaction in blood glucose levels. But as of right now, there’s no universal agreement on how to define GV. Higher glycaemic variability was shown to be linked to worse outcomes in a trend that was seen. Whether this is an epiphenomenon or if glycaemic variability and death in critically sick individuals are causally related is yet unknown. Increasing oxidative stress, brain damage, mitochondrial damage, and irregular coagulation brought on by fluctuations in glucose levels are some of the hypothesized processes. One noteworthy finding was that improved outcomes were linked to mild hyperglycemia with lower glycaemic variability rather than stricter glycaemic control with higher variability.

Insulin requirement in outcomes of sepsis

Furthermore, our study found that a statistical increase in the number of units of insulin consumed was noted in unresolved cases of sepsis, MODS patients, and deceased patients of sepsis. This might be attributed to the stress hyperglycemia. In stressful circumstances it is believed that the body stimulates the neuroendocrine axis and central nervous system, releasing hormones including cortisol, glucagon, and catecholamines that are known to increase the generation of glucose in the liver and cause hyperglycemia. Rather than peripheral insulin resistance, hepatic gluconeogenesis, and glycogenolysis are the primary causes of stress hyperglycemia. Moreover, it is believed that hyperglycemia is at least partially physiological and appropriate for the organism from a survival perspective: All cells require glucose, and although transporters like glucose transporters help, glucose absorption is totally dependent on a concentration gradient. Because hypo-perfusion and decreased blood flow are present in circumstances like sepsis, glucose must cross interstitial space to reach its target—an under-perfused cell. Under such circumstances, a greater concentration of glucose in the root, or hyperglycemia, must be seen as adaptive to hypo-perfusion.[13] Hence, our study is consistent with the hypothesis that sepsis being a stressful event requires higher insulin therapy.

Conclusion

In conclusion, this study underscores the need for glycaemic management strategies that address both blood glucose levels and variability to improve survival and reduce complications in septic patients. By highlighting the role of stable glycaemic control in critical care, these findings contribute to the development of optimised management protocols for ICU patients with sepsis, particularly among those with diabetes.

Limitations of the study

This study’s small sample size and focus on immediate ICU outcomes limit its generalizability, indicating the need for larger, multi-centre studies. Confounders such as sepsis severity, treatment variations, and CGM accuracy may have influenced results. Standardised protocols and extended follow-up are needed to confirm these findings.

Ethical issues

The study proposal was submitted for approval by the ethics review committee of the institution. The purpose of the study and study protocol were explained to the patient in the language that the patient understands, and written informed consent was obtained if the patient is willing to participate in the study.

Conflicts of interest

No speculated issue or conflict of interest for the thesis is foreseen.

Funding Statement

Nil.