The effect of renal function on the pharmacokinetics and pharmacodynamics of enavogliflozin, a potent and selective sodium-glucose cotransporter-2 inhibitor, in type 2 diabetes.

AIMS: To explore the effect of renal function on the pharmacokinetic (PK) and pharmacodynamic (PD) profile and safety of enavogliflozin, a selective sodium-glucose cotransporter 2 (SGLT2) inhibitor, in patients with type 2 diabetes mellitus (T2DM). METHODS: An open-label, two-part clinical trial was conducted in T2DM patients, stratified by renal function: Group 1, normal renal function; Group 2, mild renal impairment (RI); Group 3, moderate RI; and Group 4, severe RI. In Part A, Groups 2 and 4 received enavogliflozin 0.5 mg once. In Part B, Groups 1 and 3 received enavogliflozin 0.5 mg once daily for 7 days. Serial blood and timed urine samples were collected to analyse the PK and PD characteristics of enavogliflozin. Pearson's correlation coefficients were calculated to assess the correlations between PK or PD parameters and creatinine clearance (CrCL). RESULTS: A total of 21 patients completed the study as planned. The area under the curve (AUC) for enavogliflozin was not significantly correlated with CrCL, although the maximum concentration slightly decreased as renal function decreased. By contrast, daily urinary glucose excretion (UGE) was positively correlated with CrCL after both single- (r = 0.7866, p < 0.0001) and multiple-dose administration (r = 0.6606, p = 0.0438). CONCLUSIONS: Systemic exposure to oral enavogliflozin 0.5 mg was similar among the patients with T2DM regardless of their renal function levels. However, the glucosuric effect of enavogliflozin decreased with RI. Considering the UGE observed and approved therapeutic use of other SGLT2 inhibitors, the efficacy of enavogliflozin with regard to glycaemic control could be explored in patients with mild and moderate RI (estimated glomerular filtration rate ≥30 or ≥45 mL/min/1.73 m2) in a subsequent larger study.

Altered Lung Heat Shock Protein-70 Expression and Severity of Sepsis-Induced Acute Lung Injury in a Chronic Kidney Disease Rat Model.

Enhanced heat shock protein-70 (HSP-70) expression in the lungs is associated with attenuated acute lung injury (ALI) in a sepsis model. Chronic kidney disease (CKD) significantly contributes to the poor prognosis of patients with sepsis. This study examined the relationship between sepsis-induced ALI severity and altered lung HSP-70 expression in CKD. Experimental rats underwent a sham operation (control group) or 5/6 nephrectomy (CKD group). Sepsis was induced with cecal ligation and puncture (CLP). Laboratory tests and lung harvest were performed in the control group (without CLP and after 3, 12, 24, and 72 h of CLP) and in the CKD group (without CLP and after 72 h of CLP). ALI was the most severe after 12 h of sepsis. The mean lung injury score at 72 h after sepsis was significantly higher in the CKD group than in the control group (4.38 versus 3.30, p < 0.01). Nonetheless, enhanced lung HSP-70 expression was not observed in the CKD group. This study shows that altered lung HSP-70 expression is associated with the worsening of sepsis-induced ALI in patients with CKD. Enhancing lung HSP-70 is a novel treatment target for patients with CKD and sepsis-induced ALI.

THE NATIONWIDE INCIDENCE OF RETINAL ARTERY OCCLUSION FOLLOWING DIALYSIS AS A RESULT OF END-STAGE RENAL DISEASE: 2004 Through 2013.

PURPOSE: To examine the incidence and risk of retinal artery occlusion (RAO) in patients who have undergone dialysis in Korea. METHODS: A nationwide, population-based study using South Korean national health insurance data from 2004 to 2013 was used for analysis. All patients who began dialysis between 2004 and 2013 and the same number of control subjects were selected via propensity score matching. The incidence of RAO in the dialysis and control cohorts was calculated for 2004 to 2013 using washout data from 2003. The multivariable Cox proportional hazards model was used to evaluate the risk of developing RAO in dialysis patients. Cumulative RAO incidence curves were generated using the Kaplan-Meier method. Whether dialysis modalities influenced the incidence of RAO was also evaluated. RESULTS: Seventy-six thousand seven hundred and eighty-two end-stage renal disease patients on dialysis were included in the dialysis cohort, and 76,782 individuals were included in the control cohort. During the study period, 293 patients in the dialysis cohort and 99 patients in the control cohort developed RAO. The person-years incidence of RAO was significantly higher in the dialysis cohort than in the control cohort (dialysis = 1.1/1,000 person-years; control = 0.3/1,000 person-years; P < 0.001). The incidence of RAO was not significantly different between the two methods of dialysis (hemodialysis vs. peritoneal dialysis; P = 0.25, log-rank test). CONCLUSION: The current study provided epidemiological evidence that undergoing dialysis for end-stage renal disease was associated with an increased risk of developing RAO. The incidence of RAO rapidly increased as the duration of dialysis increased. These results strengthen the significant role of the renal function in retinal vascular disease.

The Nationwide Incidence of Retinal Vein Occlusion Following Dialysis due to End-stage Renal Disease in Korea, 2004 through 2013.

BACKGROUND: We investigated the incidence and risk of retinal vein occlusion (RVO) in end-stage renal disease (ESRD) patients on dialysis in Korea. METHODS: In this nationwide cohort study, we used Korean National Health Insurance Service data between 2004 and 2013 for analysis. ESRD patients who started dialysis from 2004 to 2013 and an equal number of controls were selected through propensity score matching. RVO incidence in both cohorts were calculated for 2004-2013 using washout data from 2003. The multivariable Cox proportional hazards model was used to assess the risk of RVO in dialysis cohort. The Kaplan-Meier method was used to generate the cumulative RVO incidence curve. Whether the dialysis modality affects the development of RVO was also evaluated. RESULTS: In this study, 74,551 ESRD patients on dialysis and the same number of controls were included. The incidence of RVO was significantly higher in the dialysis cohort than in the control cohort (dialysis = 7.3/1,000 person-years [PY]; control = 1.9/1,000 PY; P < 0.001). The cumulative-incidence of RVO was also significantly higher in the dialysis cohort than in the control cohort (P < 0.001; log-rank test). However, there was no significant difference in the incidence of RVO between the two dialysis methods (P = 0.550; log-rank test). CONCLUSION: This study provided epidemiological evidence that receiving dialysis for ESRD could increase the risk of developing RVO. We also found a rapid increase in the incidence of RVO with a longer dialysis period. These results strengthen the relationship between retinal vascular disease and renal function.

DPP-4 inhibition enhanced renal tubular and myocardial GLP-1 receptor expression decreased in CKD with myocardial infarction.

BACKGROUND: Chronic kidney disease (CKD) is strongly associated with cardiovascular disease and is a significant risk factor for increased morbidity and mortality. In contrast, GLP-1 receptor (GLP-1R) activation has been shown to confer both renal and cardiovascular protection, though its relationship with CKD and CKD with myocardial ischemia/reperfusion (MI/R) remains poorly understood. Here, we investigated changes in renal and myocardial GLP-1R expression in the CKD rat model with MI/R. METHODS: Male Sprague Dawley rats with 5/6 nephrectomy were used as a rat model of CKD and CKD with MI/R. For myocardial ischemia, the left coronary artery was ligated and released for 30 min 1 week after 5/6 nephrectomy. Dipeptidyl-peptidase 4 (DPP-4) inhibitors were administered orally with linagliptin once daily for 8 weeks. Renal cortical and myocardial GLP-1R expression were measured via immunohistochemistry and western blot analysis. RESULTS: DPP-4 activity was increased in CKD. Western blot density of GLP-1R in renal cortex extracts revealed increased abundance 2 weeks after 5/6 nephrectomy, followed by a decrease at 8 weeks. In contrast, CKD and CKD with MI/R rats showed decreases in renal and cardiac expression of GLP-1R; these effects were attenuated in rats treated with linagliptin. CONCLUSIONS: In CKD with MI/R, linagliptin attenuated renal injury and increased renal and myocardial GLP-1R expression. These data suggest that activation of renal and myocardial GLP-1R expression may provide both cardio- and renoprotective effects.

The Incidences and Characteristics of Various Cancers in Patients on Dialysis: a Korean Nationwide Study.

BACKGROUND: The numbers of patients on dialysis and their life expectancies are increasing. Reduced renal function is associated with an increased risk of cancer, but the cancer incidence and sites in dialysis patients compared with those of the general population require further investigation. We investigated the incidences of various cancers in dialysis patients in Korea and used national health insurance data to identify cancers that should be screened in dialysis clinics. METHODS: We accessed the Korean National Health Insurance Database and excerpted data using the International Classification of Disease codes for dialysis and malignancies. We included all patients who commenced dialysis between 2004 and 2013 and selected the same number of controls via propensity score matching. RESULTS: A total of 48,315 dialysis patients and controls were evaluated; of these, 2,504 (5.2%) dialysis patients and 2,201 (4.6%) controls developed cancer. The overall cancer risk was 1.54-fold higher in dialysis patients than in controls (adjusted hazard ratio, 1.71; 95% confidence interval, 1.61-1.81). The cancer incidence rate (incidence rate ratio [IRR], 3.27) was especially high in younger dialysis patients (aged 0-29 years). The most common malignancy of end-stage renal disease patients and controls was colorectal cancer. The major primary cancer sites in dialysis patients were liver and stomach, followed by the lung, kidney, and urinary tract. Kidney cancer exhibited the highest IRR (6.75), followed by upper urinary tract (4.00) and skin cancer (3.38). The rates of prostate cancer (0.54) and oropharyngeal cancer (0.72) were lower than those in the general population. CONCLUSION: Dialysis patients exhibited a higher incidence of malignancy than controls. Dialysis patients should be screened in terms of colorectal, liver, lung, kidney and urinary tract malignancies in dialysis clinics.

Renal Tubular Glucagon-Like Peptide-1 Receptor Expression Is Increased in Early Sepsis but Reduced in Chronic Kidney Disease and Sepsis-Induced Kidney Injury.

Acute kidney injury (AKI) is common in patients with sepsis and causes renal ischemia. Glucagon-like peptide-1 (GLP-1) protects the vascular system and the kidney, and GLP-1 receptor (GLP-1R) is expressed in the kidney. Renal GLP-1R activity is decreased in chronic kidney disease (CKD), but is increased by the inflammatory response; however, the effect of AKI on GLP-1R expression is unknown. We investigated the role of GLP-1 by assessing GLP-1R expression in the renal cortex in animals with AKI-related sepsis, CKD, and CKD-with-sepsis. We generated a model of CKD by 5/6 nephrectomy, and sepsis induced by cecal perforation, in male Sprague-Dawley rats. We compared renal GLP-1R expression at 3, 6, 12, 24, and 72 h after cecal perforation, and in CKD and CKD-with-sepsis. We performed blood and urine tests, western blotting (WB), and immunohistochemistry (IHC) to assay GLP-1R expression in renal tubules. The CKD-with-sepsis group showed the lowest kidney function, urine volume, and serum glucose and albumin levels. GLP-1R expression in renal tubules was decreased at 3 h, increased at 24 h, and decreased at 72 h after sepsis induction. GLP-1R expression was decreased at 8 weeks after CKD and was lowest in the CKD-with-sepsis group. The WB results were verified against those obtained by IHC. GLP-1R expression in renal tubules is increased in early sepsis, which may explain the protective effect of endogenous GLP-1 against sepsis-related inflammation.

End-stage Renal Disease and Risk of Active Tuberculosis: a Nationwide Population-Based Cohort Study.

BACKGROUND: The converging epidemics of tuberculosis (TB) and end-stage renal disease (ESRD) have generated a significant public health burden, however, previous studies have been limited to a small number of patients. This nationwide cohort study aimed to assess the rate of developing active TB among patients receiving dialysis for ESRD. METHODS: The Korean national health insurance database was used to identify patients receiving dialysis for new-onset ESRD during 2004-2013, who were propensity score matched to an equivalent number of non-dialysis subjects from the general population. The incidences of active TB in the ESRD and control cohorts were calculated for 2004-2013, and multivariable Cox proportional hazards model was used to evaluate the ESRD-related risk of active TB. RESULTS: During 2004-2013, 59,584 patients received dialysis for newly diagnosed ESRD. In the dialysis and control cohorts, 457 (0.8%) and 125 (0.2%) cases of active TB were detected, respectively. Patients with ESRD were associated with a significantly higher risk of active TB compared to the controls (incidence rate ratio, 4.80). The ESRD cohort had an independently elevated risk of active TB (adjusted hazard ratio, 4.39; 95% confidence interval, 3.60-5.37). CONCLUSION: We found that patients receiving dialysis for ESRD had an elevated risk of active TB. These results highlight the need for detailed and well-organised guidelines for active TB screening among patients with ESRD.

Erythropoiesis-stimulating Agents and Anemia in Patients with Non-dialytic Chronic Kidney Disease.

Anemia is common in patients with advanced chronic kidney disease (CKD). Though erythropoiesis-stimulating agents (ESAs) have been strongly endorsed in guidelines, it is of particular financial interest. Recently, the reimbursement of ESAs in non-dialytic patients was started by the Korean National Health Insurance System. Thus, we investigated the impact of the reimbursement of ESAs on the anemia care in non-dialytic CKD patients. Medical records of patients with advanced CKD (estimated GFR <30 mL/min/1.73 m(2)) were reviewed. Use of ESAs, blood transfusion, and hemoglobin concentrations were analyzed from one year prior to reimbursement to three years following. We used multivariable modified Poisson regression to estimate the utilization prevalence ratio (PRs). A total of 1,791 medical records were analyzed. The proportion of patients receiving ESAs increased from 14.8% before reimbursement to a peak 33.6% in 1 yr after reimbursement; thereafter, ESA use decreased to 22.4% in 3 yr after reimbursement (compared with baseline; PR, 2.19 [95% CI, 1.40-3.42]). In patients with Hb <10 g/dL, the proportion of receiving ESAs increased from 32.1% before reimbursement to 66.7% in 3 yr after reimbursement (compared with baseline; PR, 2.04 [95% CI, 1.25-3.32]). Mean hemoglobin concentrations were 10.06±1.54 g/dL before reimbursement and increased to 10.78±1.51 g/dL in 3 yr after the reimbursement change (P=0.001). However, the requirement of blood transfusion was not changed over time. With the reimbursement of ESAs, the advanced CKD patients were more likely to be treated with ESAs, and the hemoglobin concentrations increased.

Is Renal Denervation Effective in Treating Resistant Hypertension?

Resistant hypertension is diagnosed in patients whose blood pressure target is unmet despite the use of three or more antihypertensive medications. Systemic sympathetic hyperactivation is associated with the development of resistant hypertension. As the kidney is largely pervasive of the sympathetic nervous system renal denervation procedure was developed to control blood pressure by attenuating the renal and systemic sympathetic hyperactivity. Renal denervation is a minimally invasive procedure that uses radiofrequency or ultrasound energy waves to reduce the activity of the renal artery nerves. Previous clinical trials have shown conflicting results regarding the efficacy of the procedure. Symplicity HTN-1 and -2 trials showed effective blood pressure lowering results in the renal denervation group with a good safety profile. However, the Symplicity HTN-3 trial showed no difference in blood pressure lowering effect between the renal denervation and control Sham procedure groups. Notwithstanding, some recent clinical trials with Sham control and meta-analysis showed clinical benefits of renal denervation. Other clinical benefits of renal denervation include glucose control, cardiovascular protective effect, reduction of obstructive sleep apnea, and neuralgia control. A subset of patients with satisfactory blood pressure control response to the procedure may experience improved glucose control due to the overall reduced sympathetic activity and insulin resistance. Sympathetic activity control after renal denervation has cardioprotective effects, especially for those with arrhythmia and left ventricular hypertrophy. Also, renal denervation could be helpful in renalorigin pain control. Renal denervation is an effective, safe, non-invasive procedure with many clinical benefits beyond blood pressure control. Further development in the procedure technique and selection of target patients are needed for wider clinical use of renal denervation in resistant hypertension.

The Association Among Post-hemodialysis Blood Pressure, Nocturnal Hypertension, and Cardiovascular Risk Factors.

Background: Most hemodialysis (HD) patients suffer from hypertension and have a heightened cardiovascular risk. While blood pressure (BP) control is essential to end-stage kidney disease (ESKD) patients, overly stringent control can lead to intradialytic hypotension (IDH). This study aimed to examine BP variations during and after HD to determine whether these variations correlate with IDH risk. Methods: BP measurements during dialysis were taken from 28 ESKD patients, and ambulatory BP monitoring was applied post-dialysis. Laboratory parameters and risk factors, including diabetes, coronary disease, and LV mass index, were compared between IDH and non-IDH groups using an independent t-test. Results: Of the 28 patients with an average age of 57.4 years, 16 (57.1%) had diabetes, 5 (17.9%) had coronary artery disease, and 1 (3.6%) had cerebrovascular disease. The mean systolic blood pressure (SBP) during and post-HD was 142.26 mmHg and 156.05 mmHg, respectively (p=0.0003). Similarly, the mean diastolic blood pressure (DBP) also demonstrated a significant increase, from 74.59 mmHg during HD to 86.82 mmHg post-HD (p<0.0001). Patients with IDH exhibited a more substantial SBP difference (delta SBP, 36.38 vs. 15.07 mmHg, p=0.0033; age-adjusted OR=1.58, p=0.0168) and a lower post-dialysis BUN level (12.75 vs. 18.77 mg/dL, p=0.0015; age-adjusted OR=0.76, p=0.0242). No significant variations were observed in daytime and nocturnal BP between the IDH and non-IDH groups. Conclusion: Hemodialysis patients exhibited a marked increase in post-dialysis BP and lacked a nocturnal BP dip, suggesting augmented cardiovascular risks. This highlights the importance of more stringent BP control after hemodialysis.

Ceftiaxone-induced neurotoxicity: case report, pharmacokinetic considerations, and literature review.

Ceftriaxone is widely used in patients for the treatment of serious gram-negative infections. Ceftriaxone can induce some potential side effects, including neurotoxicity, however, nonconvulsive status epilepticus has rarely been reported. We report a case of acute reversible neurotoxicity associated with ceftriaxone. A 65-yr-old woman with chronic kidney disease developed altered consciousness during ceftriaxone treatment for urinary tract infection. The electroencephalogram demonstrated continuous bursts of generalized, high-voltage, 1 to 2 Hz sharp wave activity. Neurologic symptoms disappeared following withdrawal of ceftriaxone. The possibility of ceftriaxone-induced neurotoxicity should be considered in patients developing neurological impairment during ceftriaxone use, and the discontinuation of the drug could lead to complete neurological improvement.

Incidence of acute cholecystitis underwent cholecystectomy in incidence dialysis patients: a nationwide population-based cohort study in Korea.

BACKGROUND: Patients on dialysis have numerous gastrointestinal problems related to uremia, which may represent concealed cholecystitis. We investigated the incidence and risk of acute cholecystitis in dialysis patients and used national health insurance data to identify acute cholecystitis in Korea. METHODS: The Korean National Health Insurance Database was used, with excerpted data from the insurance claim of the International Classification of Diseases code of dialysis and acute cholecystitis treated with cholecystectomy. We included all patients who commenced dialysis between 2004 and 2013 and selected the same number of controls via propensity score matching. RESULTS: A total of 59,999 dialysis and control patients were analyzed; of these, 3,940 dialysis patients (6.6%) and 647 controls (1.1%) developed acute cholecystitis. The overall incidence of acute cholecystitis was 8.04-fold higher in dialysis patients than in controls (95% confidence interval, 7.40-8.76). The acute cholecystitis incidence rate (incidence rate ratio, 23.13) was especially high in the oldest group of dialysis patients (aged ≥80 years) compared with that of controls. Dialysis was a significant risk factor for acute cholecystitis (adjusted hazard ratio, 8.94; 95% confidence interval, 8.19-9.76). Acute cholecystitis developed in 3,558 of 54,103 hemodialysis patients (6.6%) and in 382 of 5,896 patients (6.5%) undergoing peritoneal dialysis. CONCLUSION: Patients undergoing dialysis had a higher incidence and risk of acute cholecystitis than the general population. The possibility of a gallbladder disorder developing in patients with gastrointestinal problems should be considered in the dialysis clinic.

Predialysis Urea Nitrogen Is a Nutritional Marker of Hemodialysis Patients.

End-stage renal disease (ESRD) patients on hemodialysis have poor nutritional status and associated problems such as inflammation and sarcopenia. Blood urea nitrogen (BUN) is an important measure of uremic toxins, and urea reduction is a marker of hemodialysis efficacy. However, a low protein diet for lower BUN could aggravate malnutrition in patients, and optimal pre-dialysis BUN is not defined. We investigated the association of pre-dialysis BUN with patients' comorbidities and the relationship between pre-dialysis BUN and serum albumin as a nutrient marker. Among the 67 patients, the average pre- and post-dialysis BUN were 59.2 and 15.0 mg/dL, respectively, serum creatinine was 10.1 mg/dL, and the average serum albumin was 4.0 g/dL. Patients' age was negatively correlated with serum creatinine (r=-0.277, p<0.05) and albumin (r=-0.453, p<0.001). Predialysis BUN showed a significant positive correlation with serum albumin (r=0.287, p<0.05) and creatinine (r=0.454, p<0.001). However, the predialysis BUN was not significantly related to diabetes, coronary artery disease, congestive heart failure, or cerebrovascular disease. Hemodialysis patients with high pre-dialysis BUN and high serum creatinine could be regarded as having good nutritional status. The significance of this study lies in the potential utility of pre-dialysis blood urea nitrogen as an indicator of the nutritional status of patients. Liberal protein intake might be recommended to adequately dialyzed patients.

Influence of Renal Function on the Single-Dose Pharmacokinetics of Besifovir, a Novel Antiviral Agent for theTreatment of Hepatitis B Virus Infection.

Per the well-known resistance of hepatitis B virus to nucleoside/nucleotide analogs, alternative treatment options with higher resistance barriers have been approved for use in both treatment-naïve and lamivudine-resistant hepatitis B virus infections. This phase I study was conducted in adults with normal and impaired renal function to evaluate the effect of renal impairment on the pharmacokinetics of besifovir, a prodrug of an acyclic nucleotide phosphonate, that is mainly cleared via renal excretion. An open-label, single-dose parallel-group clinical study was conducted in subjects with normal renal function and mild, moderate, and severe renal impairment. Subjects received a single oral dose of besifovir dipivoxil 150 mg, and serial blood and urine samples were collected for up to 72 hours after dosing to assess the pharmacokinetic characteristics of besifovir. The extent of plasma exposure of besifovir, detected as its major and active metabolites, LB80331 and LB80317, respectively, increased with worsening renal function. Compared to the subjects with normal renal function, the mean areas under the concentration-time curves of LB80331 increased by 1.5-, 2.5-, and 4.5-fold in subjects with mild, moderate, and severe impairment, respectively. LB80317 showed a 1.8-, 3.2-, and 6.2-fold increase in subjects with mild, moderate, and severe renal impairment compared to those with normal function. The ratios of LB80331 renal clearance and the average estimated glomerular filtration rate of each renal impairment group with respect to the normal group were similar. The increase in plasma exposure and decrease in renal clearance suggest the need to adjust dosage regimens in patients with moderate to severe renal impairment.

Preliminary Study of Different Treatment Responses between Bevacizumab, Aflibercept and Dexamethasone Implant According to Renal Function in Diabetic Macular Edema Patients.

Background: The purpose of this study was to investigate the association between responses to intravitreal bevacizumab injection and renal function in diabetic macular edema (DME) patients. Methods: A retrospective study of the medical records of 104 treatment-naïve DME patients who received intravitreal bevacizumab injection (IVBI) was conducted. Based on the estimated glomerular filtration rate (eGFR, mL/min/1.73 m2), the participants were classified into three groups. Intergroup comparisons of the best-corrected visual acuity (BCVA) and central subfield retinal thickness (CST) changes were performed after three-monthly consecutive IVBIs. In the groups with decreased renal function, the response to further treatment with a different drug was investigated. Results: A total of 104 participants were included in the study: 60 participants in the preserved renal function group (eGFR ≥ 60), 25 participants in the moderate chronic kidney disease (CKD) group (30 ≤ eGFR < 60), and 19 participants in the severe CKD group (eGFR < 30). After three-monthly consecutive IVBIs, BCVA (p < 0.001) and CST (p < 0.001) were significantly improved only in the preserved renal function group. Following further treatment of patients with decreased renal function, the treatment results were significantly better in those who were switched to aflibercept or dexamethasone implant than in those who were maintained on IVBI. Conclusions: From this preliminary study, we observed that renal function might affect the response to IVBI treatment in patients with DME. In the case of a poor response to initial IVBI treatment for DME in patients with moderate to severe CKD, our study supports switching to the aflibercept or dexamethasone implant.

Comparison between Statistical Models and Machine Learning Methods on Classification for Highly Imbalanced Multiclass Kidney Data.

This study aims to compare the classification performance of statistical models on highly imbalanced kidney data. The health examination cohort database provided by the National Health Insurance Service in Korea is utilized to build models with various machine learning methods. The glomerular filtration rate (GFR) is used to diagnose chronic kidney disease (CKD). It is calculated using the Modification of Diet in Renal Disease method and classified into five stages (1, 2, 3A and 3B, 4, and 5). Different CKD stages based on the estimated GFR are considered as six classes of the response variable. This study utilizes two representative generalized linear models for classification, namely, multinomial logistic regression (multinomial LR) and ordinal logistic regression (ordinal LR), as well as two machine learning models, namely, random forest (RF) and autoencoder (AE). The classification performance of the four models is compared in terms of accuracy, sensitivity, specificity, precision, and F1-Measure. To find the best model that classifies CKD stages correctly, the data are divided into a 10-fold dataset with the same rate for each CKD stage. Results indicate that RF and AE show better performance in accuracy than the multinomial and ordinal LR models when classifying the response variable. However, when a highly imbalanced dataset is modeled, the accuracy of the model performance can distort the actual performance. This occurs because accuracy is high even if a statistical model classifies a minority class into a majority class. To solve this problem in performance interpretation, we not only consider accuracy from the confusion matrix but also sensitivity, specificity, precision, and F-1 measure for each class. To present classification performance with a single value for each model, we calculate the macro-average and micro-weighted values for each model. We conclude that AE is the best model classifying CKD stages correctly for all performance indices.