Introduction

The international diabetes federation predicted that the global population of diabetes will reach to 454 million in 2030; furthermore, the number will increase to 783 million in 20451. Type 2 diabetes mellitus (T2DM) and chronic kidney disease (CKD) are both chronic diseases. A study showed that the prevalence of T2DM patients complicated with CKD reached to 64–81.6/10,000, and it will be further increased2. CKD is one of the most debilitating and expensive complication of T2DM3. T2DM complicated with CKD will reduce life expectancy by 16 years than without CKD, which will increase higher medical cost burden and mortality4,5. T2DM and CKD patients often need to restrict their diet during their daily treatment, and thus are prone to disease-related malnutrition, which further leads to poor prognosis before admission to ICU6,7.

T2DM patients with CKD are accompanied by protein-energy malnutrition and inflammation before admission to ICU. Blood urea nitrogen (BUN) not only reflects the status of kidney function, but also as an important indicator of protein nutrient metabolism level and inflammation8. A study found that high BUN levels were positively associated with increased adverse renal outcomes9. Albumin is the main component of plasma proteins, and as a commonly used biomarker in clinical practice, its slight changes can affect the prognosis of patients in various diseases. A study showed that low serum albumin levels accelerated the decline of kidney function10. Studies have reported that BAR can be used as a useful biomarker for poor prognosis of many diseases, such as gastrointestinal bleeding, sepsis, COVID-19, and acute kidney injury11,12,13.

However, since the impact of BAR among T2DM patients with CKD has not yet been investigated, in the study, we evaluated the relationship between BAR and the prognosis of T2DM patients with CKD in ICU.

Materials and methods

Data source

This was a retrospective cohort study that involved Medical Information Mart for Intensive Care III (MIMIC-III) (https://physionet.org/content/mimiciii/1.4/) database. MIMIC-III database contains clinical information on over 50,000 ICU patients at the Beth Israel Deaconess Medical Center from 2001 to 201214. After completing the online training course of the National Institutes of Health, the author (S.L.) obtained access to this database (certification number: 42883491). Accessed to database was approved by the Institutional Review Boards of Beth Israel Deaconess Medical Center (Boston, MA) and the Massachusetts Institute of Technology (Cambridge, MA). We extracted patient demographics, laboratory findings, length of hospital stay, 30- and 90-day mortality, and other clinical variables by PostgreSQL V.10.0. Our study was performed in accordance with the Declaration of Helsinki (as revised in 2013).

Inclusion and exclusion criteria

Patients admitted to the ICU for the first time were included in this study. Patients were excluded according to: (1) Less than 18 years old; (2) ICD code is not T2DM. (3) Without CKD. According to the KDIGO clinical practice guidelines, CKD was diagnosed that glomerular filtration rate (GFR) below 60 mL/min/1.73 m2 for 3 months or more15. (4) Missing BUN or serum albumin values; (5) Less than 48 h in ICU; (6) Missing data for more than 5% of patients. Finally, a total of 1920 patients were included in this study (Fig. 1). We had listed the top five diagnosed diseases and top five diseases that were first diagnosed in this population at admission to ICU (Supplementary Table 1, 2).

Figure 1
figure 1

Flow diagram of the study.

Full size image

Data extraction

We extracted the variables as follows: age, gender, weight, systolic blood pressure (SBP), diastolic blood pressure (DBP), coronary artery disease (CAD), chronic heart failure (CHF), hypertension, CKD stage, sequential organ failure assessment (SOFA), hemoglobin (HGB), white blood cell (WBC) count , platelet count, potassium, sodium, phosphate, creatinine, blood urea nitrogen (BUN), albumin (ALB), alkaline phosphatase (ALP), alanine transaminase (ALT), aspartate transaminase (AST), glucose, lactate level, arterial oxygen partial pressure (PaO2), partial thromboplastin time (PTT), prognostic nutritional index (PNI) and estimate-GFR (eGFR) by structured query language in PostgreSQL. We had listed some diseases that might influence the prognosis and clinical data itself (Supplementary Table 3). BAR was calculated by dividing the BUN by the albumin.

Statistical analysis

Patients were divided into three groups: BAR < 9.2, 9.2 ≤ BAR ≤ 21.3 and BAR > 21.3, according to the interquartile ranges (IQRs) of BAR value. Continuous variables were presented as mean ± standard deviations or IQRs, and classification variables were presented as totals and percentage (%). The chi-square test was used for classified variables between groups. The Wilcoxon rank-sum test was used for non-normally distributed continuous variables, and student t-test was used for normally distributed continuous variables. Univariate and multivariate cox regression were used to assess the independent factors associated with 90-days mortality, which presented as the hazard ratio (HR) and 95% confidence interval (CI). In order to reduce the impact of confounding factors, we constructed three Cox regression models to identify the potential clinical usefulness of BAR by including covariates with p values < 0.05 in the univariate Cox analyses or for importance of clinical concern. Kaplan–Meier survival analysis was used to determine the difference in 90-day mortality between the three groups. Stratification analyses was used to assess the association of BAR with 90-day mortality. P < 0.05 was considered statistically significant. The statistical analyses were performed using the Stata software version 16.0 (Stata Corp. LLC, TX, US).

Ethical approval and consent to participate

The informed consent was waived by the Institutional Review Boards of Beth Israel Deaconess Medical Center (Boston, MA) and the Massachusetts Institute of Technology (Cambridge, MA). The patient's information has been standardized and the project did not affect clinical care, so requirement for individual patient consent was waived.

Results

Baseline characteristics

A total of 1920 patients were enrolled and were divided into the three groups: low-BAR group (BAR < 9.2, n = 486), mid-BAR group (9.2 ≤ BAR ≤ 21.3, n = 954), and high-BAR group (BAR > 21.3, n = 480). The number of man was significantly higher in the high-BAR group with lower proportion of hypertension. A higher proportion of CHF, CKD 4 stage and CKD 5 stage, along with higher levels of SOFA score, WBC, potassium, phosphate, creatinine, ALP, lactate and PTT in the high-BAR group; Patients with higher BAR had lower age, SBP, DBP, HGB, sodium, PaO2, PNI, and eGFR (all p < 0.05) (Table 1).

Table 1 Comparisons of demographics within three BAR levels.
Full size table

BAR levels and outcome

Compared with low-BAR group (BAR < 9.2), patients with higher BAR had significantly higher the length of stay, in-hospital mortality, 30-day mortality and 90-day mortality (p < 0.05) (Table 2).

Table 2 BAR level and clinical outcome.
Full size table

Association between the BAR and 90-day mortality

Univariate and multivariate Cox regression analyses were utilized to evaluate the significance of BAR in predicting 90-day mortality. Univariate regression analysis showed that age, weight, SBP, DBP, CAD, CKD stage, SOFA score, WBC, phosphate, lactate, PTT, eGFR, BAR, mid-BAR and high-BAR were significantly associated with 90-day mortality (p < 0.05) (Table 3). Unadjusted in model 1, high BAR level was significantly associated with higher risk of 90-day mortality (BAR as continuous variable, HR 1.018, 95% CI 1.011–1.027; mid-BAR, HR 1.261, 95% CI 1.021–1.557; high-BAR, HR 1.768, 95% CI 1.409–2.218). Adjust for age, gender, weight, SBP, DBP, CAD, CHF, hypertension, CKD stage and SOFA score in model 2, high BAR level was also significantly associated with higher risk of 90-day mortality (BAR as continuous variable, HR 1.021, 95% CI 1.012–1.030; mid-BAR, HR 1.304, 95% CI 1.043–1.631; high-BAR, HR 1.934, 95% CI 1.489–2.511). Furthermore, Adjust for model 2 plus HGB, WBC, platelet, potassium, sodium, phosphate, ALP, ALT, AST, glucose, PaO2, PTT and eGFR in model 3, high BAR level remained a greater risk of 90-day mortality (BAR as continuous variable, HR 1.018, 95% CI 1.009–1.028; mid-BAR, HR 1.277, 95% CI 1.011–1.618; high-BAR, HR 1.864, 95% CI 1. 399–2.487) (all p < 0.05) (Table 4).

Table 3 Univariate Cox regression analyses to assess risk factors associated with 90-day mortality in T2DM patients with CKD.
Full size table
Table 4 Association between BAR and 90-day mortality.
Full size table

Prediction of 90-day mortality

The receiver operating characteristic (ROC) curve generated using the indicator variables (BAR, ALB, and BUN) were shown in Fig. 2. The AUC value of BAR was 0.708, which showed significantly higher AUC value than the ALB and BUN (p < 0.05). Similarly, the c-statistic of BUN, ALB, and BAR showed that BAR had the highest c-statistic (Supplementary Table 4).

Figure 2
figure 2

Receiver operating characteristic curves for the prediction of 90-day mortality.

Full size image

Subgroup analyses

Subgroup analyses were used to determine the consistency of association between BAR and risk of 90-day mortality (Table 5). Subgroup analyses showed that patients with age > 75(HR 1.024, 95% CI 1.013–1.034), man (HR 1.022, 95% CI 1.012–1.032), hypertension (HR 1.023, 95% CI 1.010–1.037), CHF (HR 1.019, 95% CI 1.008–1.030), CAD (HR 1.022, 95% CI 1.007–1.038), SOFA score ≥ 5 (HR 1.014, 95% CI 1.005–1.023), weight ≥ 77 (HR 1.027, 95% CI 1.016–1.038), CKD 3 stage (HR 1.023, 95% CI 1.011–1.035), phosphate < 4 (HR 1.019, 95% CI 1.005–1.033), WBC ≥ 11.3 (HR 1.019, 95% CI 1.010–1.029), lactate ≥ 2.4 (HR 1.024, 95% CI 1.013–1.035), ALP ≥ 103 (HR 1.022, 95% CI 1.012–1.032) and PaO2 < 155 (HR 1.019, 95% CI 1.009–1.029) had a significantly higher risk of 90-day mortality with high BAR level. The results also showed that high BAR was significantly associated with increased 90-day mortality in these patients with CKD stage 5 (Supplementary Table 5).

Table 5 Subgroup analysis of the associations between BAR and 90-day mortality.
Full size table

Kaplan–Meier analysis

The Kaplan–Meier survival curve analysis showed that BAR > 21.3 had worst prognosis. Patients in higher BAR group had significantly higher risk of 90-day mortality than low and mid-BAR groups (BAR > 21.3 vs 9.2 ≤ BAR ≤ 21.3 vs BAR < 9.2; 40.2% vs 30.6% vs 25.1%, respectively; log-rank test p value < 0.001) (Fig. 3).

Figure 3
figure 3

Kaplan–Meier curve was used to evaluate the difference between BAR levels and 90-day mortality in T2DM patients with CKD in the ICU. In the Kaplan–Meier analysis, the log-rank test P value < 0.001.

Full size image

Discussion

With the increasing prevalence of T2DM and CKD in worldwide, timely identification of prognostic risk factor is particularly important in clinical work. In our study, we found that higher BAR on admission to ICU was significantly associated with an increased risk of 90-day mortality in T2DM patients with CKD. And BAR could serve as an independent predictive factor of 90-day mortality. Further, the K-M curve also presented that the high BAR group had a worse prognosis. Our study was the first largest study to explored the relationship between BAR and prognosis in T2DM patients with CKD in the ICU.

BUN is a nitrogen-containing compound that it is influenced by renal function, neurohormone, and sympathetic nervous activity. BUN is mainly filtered through the glomeruli and excreted through urine. When glomerular filtration function decreases, BUN concentration will increase. BUN can not only be used to estimate glomerular filtration function, but also to assess the body's nutritional status, low blood volume, protein metabolism and others8. Many studies have found that BUN was a powerful predictor of prognosis in patients with heart failure, and its efficacy was even better than GFR and serum creatinine16,17. A large study in China found that BUN levels were positively associated with the risk of developing T2DM in Chinese adults18. Studies in T2DM patients showed that the increased BUN level will significantly increase the risk of diabetes retinopathy and diabetes nephropathy19,20. Elevated BUN indicated poor prognosis for patients in the ICU13,21,22. In the ICU, T2DM patients with CKD had circulatory dysfunction and neuroendocrine system dysfunction, which further aggravated kidney injury. At this time, high levels of BNU may predict a worse prognosis for patients.

ALB is not only a nutritional marker, but also plays an important role in anti-inflammatory, antioxidant and others aspects23. As an important antioxidant in plasma, ALB inhibits apoptosis of renal tubular cells by clearing oxygen free radicals24. It was found that ALB not only improves renal perfusion and glomerular filtration by prolonging renal vasodilation, but also selectively inhibits the expression of tumor necrosis factor-α-induced vascular cell adhesion molecule 1 and the activation of nuclear factor kB and monocyte adhesion in human endothelial cells to prevent kidney injury13,25.In T2DM patients, ALB level was negatively correlated with the incidence of diabetes retinopathy26. Besides, a study also showed that hypoproteinemia significantly accelerated the risk of renal failure in patients with diabetes nephropathy27. Low ALB level was caused by insufficient nutrition intake and a state of inflammatory stress in ICU patients. Numerous studies had suggested that hypoalbuminemia was a risk factor for poor prognosis in ICU patients28,29,30.

High BAR levels are caused by high BUN or low ALB. BAR has been proven to be a more reliable predictor than BUN or ALB. In our study, ROC curve also showed the AUC value of BAR was significantly higher than the ALB and BUN. Studies suggest that high BAR can significantly increase the mortality of patients with sepsis, acute myocardial infarction, acute pulmonary embolism, heart failure and others11,21,31,32. A study found that high BAR significantly increased in-hospital mortality and the incidence of AKI for patients with cerebral hemorrhage in the ICU13. Our study also found similar results. We found that when BAR > 21.3, patients' stay in the ICU, hospital mortality, and 30-day and 90-day mortality were significantly increased. After adjusting for confounders such as serum creatinine, CKD and others, elevated BAR was still positively associated with poor prognosis in patients undergoing cardiac surgery33. Similarly, in our study, after controlled for confounding variables by multivariate Cox regression analysis, we found that high BAR was an independent risk factor for 90-day mortality. Besides, in subgroup analysis, we demonstrated that BAR was an effective predictor of 90-day mortality in T2DM patients with CKD under various specific conditions. Therefore, in clinical practice, we may be able to reduce the BUN value by improving glomerular filtration rate (such as increasing renal perfusion by maintaining normal volume load), and also by increasing albumin (infusion of human serum albumin, etc.) to reduce BAR, thereby reducing the risk of 90 day mortality in patients. In routine clinical practice, BAR could be calculated easily and quickly, and is more stable and conducive to clinical application compared with a single indicator.

In our study, several limitations should be highlighted to interpret the results as follows: First, this study as a single-centric retrospective study, we couldn’t avoid to selection bias. Second, the data used in the manuscript were recorded from 2001 to 2012, which were old. Third, due to the limited contents of this database, some unrecorded clinical information was missing, may affect the outcome. Fourth, this study only included BUN and albumin records of patients at their first admission, so that the prognostic impact of dynamic changes in BAR was still unclear. Finally, the underlying mechanism of BAR affected the prognosis could not be determined. Therefore, it is necessary to establish a large multicenter prospective study to confirm the above results and further to investigate the mechanism.

Conclusion

A higher BAR was significantly associated with an increased risk of 90-day mortality. BAR could be served as a prognostic predictor for 90-day mortality in T2DM patients with CKD in ICU, due to its inexpensive and readily available nature.