Association between glycemic variability and short-term mortality in patients with acute kidney injury: a retrospective cohort study of the MIMIC-IV database – Scientific Reports

In this retrospective cohort study, we found that MBG and CV were independently associated with poor outcomes in ICU patients with AKI. There is a U-shaped relationship between the MBG levels and ICU 30-day mortality rate in AKI patients, with an inflection point of 111.3 mg/dl. When the MBG levels were < 111.3 mg/dl, a 1-mg/dl decrease in MBG was associated with a 1.1% reduction in the ICU 30-day mortality rate among AKI patients (HR 0.989, 95% CI 0.980–0.998, P = 0.0169). Conversely, when the MBG levels were ≥ 111.3 mg/dl, the ICU 30-day mortality rate increased by 0.3% for every 1-mg/dl increase in MBG (HR 1.003, 95% CI 1.002–1.004, P < 0.001). Moreover, we observed a linear relationship between the CV levels and ICU 30-day mortality, with the ICU 30-day mortality rate increasing by 11% for every 1-SD (17.86%) increase in CV (HR 1.11, 95% CI 1.06–1.16, P < 0.001). Elevated CV levels were significantly associated with lower survival rates in AKI patients. Consequently, maintaining BG levels within a reasonable range and minimizing glycemic variability represent effective strategies for mitigating the risk of mortality in AKI patients.

AKI has long been acknowledged as a complication of critical illness that is independently associated with mortality23,24. AKI is characterized by a variable etiology and complex pathogenesis. Previous studies have shown that infection, age, AKI stage, severity of circulatory shock, RRT treatment, comorbidities, and SOFA score are important factors that could affect the outcome of patients with AKI23,25,26,27. However, the evidence from studies on the relationship between BG levels and AKI outcomes remains limited. Bagshaw et al. analyzed the data of adult patients at 24 ICUs from 2000 to 2005 and found that both early hypoglycemia (BG < 4.5 mmol/L) and early variability in BG (< 4.5 and > or = 12.0 mmol/L) independently predicted an increased mortality risk16. Despite its large sample size, the study did not deeply explore the impact of AKI as a co-morbid condition on outcomes, and it defined BG variability solely based on the range of two randomly measured glucose values. The study by Gorelik et al. found that hyperglycemia (> 10 mmol/L) was positively associated with the incidence of AKI and the severity of acute illness, as well as an increase in 30-day mortality among non-diabetic inpatients28. A retrospective cohort study by Li et al. found that AKI patients should maintain their BG at a reasonable range, and should not drop lower than 5.52 mmol/L10. However, none of the aforementioned studies establish a conclusive link between various BG levels and mortality in AKI patients, and it is difficult to represent the overall BG levels during hospitalization based solely on the first BG values on admission. Therefore, the optimal BG control range still needs further study.

Recent studies have found that increased perioperative glucose variability is significantly associated with a heightened risk of postoperative AKI29,30. Consistently, a high glucose variability has been consistently associated with adverse prognosis in critically ill patients during their ICU stay31,32. Despite these findings, there remains a lack of consensus on the impact of glucose variability on mortality among AKI patients. In this study, we used all biochemical BG records from the first day of ICU admission to calculate the MBG and CV levels, thereby capturing the overall status of BG on the first day. Our results suggested that increased BMG and CV levels were associated with elevated mortality in AKI patients. Specifically, high BMG levels (≥ 167.2 mg/dl) and high CV levels (≥ 24.8%) were independently associated with an increased risk of ICU 30-day mortality among patients with AKI. In line with our findings, a secondary analysis from the Normal versus Augmented Level of RRT study showed that higher CV and SD values were associated with an increased risk of ICU 90-day death in severe AKI patients33. Unfortunately, this study focused on the relationship between baseline BG levels, glycemic variability and RRT-related outcomes in patients with AKI, which did not clearly elucidate other confounding factors that could influence clinical outcomes, such as diabetic status and the use of hypoglycemic drugs, nor did it establish a definitive BG control range.

There is inconsistency among guidelines regarding BG control goals for patients with AKI. The latest American Diabetes Association guidelines recommend a target glucose range of 7.8–10.0 mmol/L for patients with severe disease34. The KDIGO guidelines previously recommended a BG control target at 6.1–8.3 mmol/L for AKI patients35. In our study, the inflection point of the MBG levels in AKI patients was 111.3 mg/dl (6.18 mmol/L), indicating that these patients have the highest survival rate at this point. Notably, the inflection points in our study occurred in the T1 (< 125.0 mg/dl) group, while more patients had higher MBG levels, so only a weak protective effect was shown in the T2 (125.0–167.2 mg/dl) group, and the lowest survival rate was observed in the T3 (≥ 167.2 mg/dl) group due to higher MBG levels. We speculated that the reason for this phenomenon is that critically ill patients in the ICU often have elevated stress glucose, hence most patients have glucose values above the inflection point.

Glycemic variability represents a crucial aspect of glucose homeostasis, with its elevation signaling potential for extreme glycemic excursions, thereby increasing the risk for both hyperglycemia and hypoglycemia36. The CV is widely recognized as a key measure of glycemic variability37,38. In our study, we observed a higher mortality rate in the highest tertile of glycemic variability (T3, CV ≥ 24.8%), which suggested that reducing glycemic variability could improve survival outcomes in patients with AKI. This finding also confirmed the wide variability in glycemic control associated with significantly higher mortality, as suggested in previous studies12,16,39. Considering the potential effect of diabetic status on CV, we further performed multifactorial regression analysis by dividing the entire cohort into diabetic and non-diabetic groups. Our results showed that elevated glycemic variability is independently associated with increased risk of mortality among patients without diabetes mellitus, but this association was not observed in patients with diabetes mellitus (Table S4 and Table S5). This suggests that diabetic status may act as a protective effect against their adverse prognosis. We propose two possible explanations for this observation: (1) Patients with diabetes experience chronic hyperglycemia, which allows their bodies to adapt to fluctuations in BG levels and enhances their capacity to counteract stress-induced BG fluctuations40,41. However, non-diabetic patients with stress hyperglycemia might be less capable of regulating BG levels. This discrepancy can lead to acute oxidative stress production and release of large amounts of inflammatory factor42,43, which potentially cause pancreatic β-cell toxicity and irreversible impairment of pancreatic function44. (2) The increased risk of hypoglycemia due to elevated CV could mask the association between CV and mortality in AKI patients with diabetes. Further investigation needs to be done to uncover the underlying mechanisms responsible for this effect.

It is well known that the kidney, as an organ of gluconeogenesis, plays an important role in maintaining glucose metabolic homeostasis through glucose reabsorption and utilization45. The kidney is a mitochondria-rich organ46. In the high glucose state, mitochondria experience swelling and expansion, but cannot be cleared promptly. This malfunction leads to the release of large amounts of reactive oxygen species, contributing to the oxidative stress response in renal tubules. This oxidative stress plays a pivotal role in the progression of fibrosis and apoptosis of renal tubular epithelial cells, which in turn leads to reduced glucose reabsorption capacity of renal tubules, inducing renal fibrosis and failure47,48. Meanwhile, mitochondrial damage, characterized by fragmentation and the release of damage-associated molecular patterns has been shown to induce and accelerate the development of AKI49. In addition, hyperglycemia exacerbates renal ischemia–reperfusion injury/hypoxia-reperfusion injury (HRI) induced AKI50. This condition has been strongly associated with all-cause mortality in patients with critical illness51,52,53. Consistently, this conclusion was also confirmed in in vitro experiments54. The primary pathophysiological mechanism is closely associated with severe hyperglycemia leading to dehydration, reduced circulating blood volume, acidosis, infection, and ultimately resulting in rhabdomyolysis and multiple organ dysfunction55,56. Hypoglycemia is comparable to diabetic ketoacidosis and hyperosmolar nonketotic diabetic coma, with acute cardiovascular and cerebrovascular disease due to insufficient energy supply to the heart and brain cells as the main mortality event57. There is insufficient evidence regarding the direct cause of kidney injury by hypoglycemia, and the possible mechanisms need to be further explored.

Our study offers several notable advantages. First, this is a real-world study with a large sample, which is highly generalizable. Second, by utilizing the MBG and CV on the first day of admission as variables to study the effect on the outcome, we mitigated the risk of false inferences on outcome caused by one random BG fluctuation—thereby enhancing the stability of our findings. Third, we employed rigorous statistical methods to minimize the impact of residual confounders on our findings. Finally, to the best of our knowledge, this is the first study demonstrating an independent association between glycemic variability and short-term mortality in ICU patients with AKI. The results of this study will help to improve the prognosis of AKI patients by informing more precise glycemia management strategies.

However, there are some limitations to our study that should be considered. First, due to the inherent nature of retrospective cohort studies, multiple confounding factors are inevitable. Thus, we conducted a multifactorial analysis to adjust for the effect of confounding factors on the study outcome as much as possible. Second, as an observational study, causality cannot be fully deduced. Third, for patients with multiple ICU admissions, we only analyzed baseline information from the first admission, which may have created a selection bias. Fourth, the inclusion of AKI patients was incomplete due to less than three BG tests on the first day of admission, which may have influenced the study’s outcomes. Finally, although high CV was related to the presence of diabetes mellitus, the lack of detailed information about insulin dosage, duration of diabetes, and other possible diabetes-related complications, which means that confounding related to diabetes mellitus may not have been fully addressed.