Development and internal validation of machine learning algorithms for end-stage renal disease risk prediction model of people with type 2 diabetes mellitus and diabetic kidney disease.

AIMS: Diabetic kidney disease (DKD) is the most common cause of end-stage renal disease (ESRD) and is associated with increased morbidity and mortality in patients with diabetes. Identification of risk factors involved in the progression of DKD to ESRD is expected to result in early detection and appropriate intervention and improve prognosis. Therefore, this study aimed to establish a risk prediction model for ESRD resulting from DKD in patients with type 2 diabetes mellitus (T2DM). METHODS: Between January 2008 and July 2019, a total of 390 Chinese patients with T2DM and DKD confirmed by percutaneous renal biopsy were enrolled and followed up for at least 1 year. Four machine learning algorithms (gradient boosting machine, support vector machine, logistic regression, and random forest (RF)) were used to identify the critical clinical and pathological features and to build a risk prediction model for ESRD. RESULTS: There were 158 renal outcome events (ESRD) (40.51%) during the 3-year median follow up. The RF algorithm showed the best performance at predicting progression to ESRD, showing the highest AUC (0.90) and ACC (82.65%). The RF algorithm identified five major factors: Cystatin-C, serum albumin (sAlb), hemoglobin (Hb), 24-hour urine urinary total protein, and estimated glomerular filtration rate. A nomogram according to the aforementioned five predictive factors was constructed to predict the incidence of ESRD. CONCLUSION: Machine learning algorithms can efficiently predict the incident ESRD in DKD participants. Compared with the previous models, the importance of sAlb and Hb were highlighted in the current model.HighlightsWhat is already known? Identification of risk factors for the progression of DKD to ESRD is expected to improve the prognosis by early detection and appropriate intervention.What this study has found? Machine learning algorithms were used to construct a risk prediction model of ESRD in patients with T2DM and DKD. The major predictive factors were found to be CysC, sAlb, Hb, eGFR, and UTP.What are the implications of the study? In contrast with the treatment of participants with early-phase T2DM with or without mild kidney damage, major emphasis should be placed on indicators of kidney function, nutrition, anemia, and proteinuria for participants with T2DM and advanced DKD to delay ESRD, rather than age, sex, and control of hypertension and glycemia.

Addition of glomerular lesion severity improves the value of anemia status for the prediction of renal outcomes in Chinese patients with type 2 diabetes.

We aimed to determine the utility of biopsy data and anemia for the prediction of renal outcomes in Chinese patients with type 2 diabetes. In total, 441 Chinese patients with type 2 diabetes and biopsy-confirmed diabetic nephropathy (DN) were enrolled in a retrospective study. Their renal pathology was assessed using the Renal Pathology Society system. Cox proportional hazards models were used to estimate hazard ratios (HRs) for end-stage renal disease (ESRD), and immunofluorescence staining was used to assess the expression of hypoxia-inducible factor (HIF)-α in patients' kidneys. We found that glomerular pathology classification was an independent pathological predictor of low hemoglobin concentration, according to linear and logistic regression analyses. Each 1 g/dL decrease in baseline hemoglobin concentration was associated with a 42% higher risk of an adverse renal outcome, after adjustment for clinical and pathologic covariates. In patients with severe glomerular lesions, the risk of progression to ESRD was significantly higher if mild or moderate/severe anemia was present, but in patients with mild glomerular lesions, the risk was only significantly higher in those with moderate or severe anemia than in the absence of anemia. Harrell's C Concordance was improved, but the Akaike information criterion was worsened by adding the glomerular pathology classification to the use of anemia status and clinical data. Immunofluorescence staining revealed that renal HIF-1α and HIF-2α expression was significantly higher in classes II-IV than class I. Thus, the addition of glomerular pathology classification increases the value of anemia status for the prediction of the progression to ESRD.

Prognostic value of metabolic syndrome in renal structural changes in type 2 diabetes.

PURPOSE: To investigate the prognostic value of metabolic syndrome (MetS) and its relationship with renal structure changes in patients with type 2 diabetes and associated diabetic nephropathy (DN). METHODS: 411 Chinese patients with type 2 diabetes and biopsy-confirmed DN were enrolled in this retrospective study. MetS was defined according to the modified criteria of the 2005 International Diabetes Federation. Baseline demographics and clinical information at the time of renal biopsy were extracted from the hospital's electronic medical records system. Renal pathological findings were assessed according to Renal Pathology Society system. Univariate and multivariate logistic regression analyses were performed to define the pathological covariates associated with MetS. A competing risk model, with death as the competing risk, was used to estimate the sub-distribution hazard ratio (SHR) of MetS for end-stage kidney disease (ESKD). RESULTS: 224 (55%) patients had MetS. Patients with MetS had poor renal function and more severe interstitial fibrosis tubular atrophy scores (IFTA) than those without MetS. Multivariate logistic regression analysis revealed that IFTA was significantly associated with MetS (odds ratio per score increase 1.45, 95% confidence interval [CI] 1.02-2.05). Of the patients with DN at risk, 40% of patients progressed to ESKD. After adjusting for renal function and pathological parameters, the presence of MetS was an independent predictor for progression to ESKD (SHR 1.93, 95% CI 1.34-2.79). The SHRs for progression to ESKD also increased as the number of MetS components increased. Additionally, adding the IFTA scores improved the prognostic power of a model that only contained MetS and clinical covariates for predicting future ESKD. CONCLUSION: MetS is an independent prognostic predictor of ESKD in patients with T2D and DN, while adding the IFTA scores increased the prognostic value of MetS for renal outcome.

Whether Renal Pathology Is an Independent Predictor for End-Stage Renal Disease in Diabetic Kidney Disease Patients with Nephrotic Range Proteinuria: A Biopsy-Based Study.

Aims: To investigate whether renal pathology is an independent predictor for end-stage renal disease (ESRD) in diabetic kidney diseases (DKD) with nephrotic range proteinuria. Methods: A total of 199 DKD patients with nephrotic range proteinuria underwent renal biopsy and were divided into an ESRD group and a non-ESRD group. A Kaplan−Meier analysis was used to compare renal survival rate, and univariate and multivariate Cox proportional hazard analyses were used to determine the predictors of the ESRD. Results: The mean age of included patients was 51.49 ± 9.12 years and 113 patients (56.8%) progressed to ESRD. The median follow-up period was 16 (12−28) months. The glomerular pathology class III is the most common type (54.3%). In the Kaplan−Meier analysis, compared with patients without ESRD, patients with ESRD had a longer duration of diabetes (≥6 years), lower eGFR (<60 mL/min/1.73 m2), lower albumin (<30 g/L), lower hemoglobin (<120 g/L), and a higher grade of glomerular stage (class III + IV vs. class I + II) (p < 0.05). The hemoglobin and e-GFR, but not the histopathological damage, were significantly associated with a higher risk of ESRD in both the univariate and multivariate Cox analyses. Conclusions: In patients with diabetic kidney disease characterized by nephrotic range proteinuria, histopathological damage (glomerular alterations, interstitial fibrosis and tubular atrophy (IFTA), interstitial inflammation, and arteriolar hyalinosis) is not associated with poor renal outcomes, but hemoglobin and e-GFR could predict poor renal outcomes.

Association between serum uric acid and renal outcome in patients with biopsy-confirmed diabetic nephropathy.

OBJECTIVE: To investigate the relationship between serum uric acid (SUA) level and renal outcome in patients with type 2 diabetes mellitus (T2DM) and diabetic nephropathy (DN). METHODS: A total of 393 Chinese patients with T2DM and biopsy-proven DN and followed at least 1 year were enrolled in this study. Patients were stratified by the quartiles of baseline level of SUA: Q1 group: 286.02 ± 46.66 µmol/L (n = 98); Q2 group: 358.23 ± 14.03 µmol/L (n = 99); Q3 group: 405.50 ± 14.59 µmol/L (n = 98) and Q4 group: 499.14 ± 56.97µmol/L (n = 98). Renal outcome was defined by progression to end-stage renal disease (ESRD). Kaplan-Meier survival analysis and Cox proportional hazards model were used to analyze the association between SUA quartiles and the renal outcomes. RESULTS: During the median 3-year follow-up period, there were 173 ESRD outcome events (44.02%). No significant difference between SUA level and the risk of progression of DN (P = 0.747) was shown in the Kaplan-Meier survival analysis. In multivariable-adjusted model, hazard ratios for developing ESRD were 1.364 (0.621-2.992; P = 0.439), 1.518 (0.768-3.002; P = 0.230) and 1.411 (0.706-2.821; P = 0.330) for the Q2, Q3 and Q4, respectively, in comparison with the Q1 (P = 0.652). CONCLUSIONS: No significant association between SUA level and renal outcome of ESRD in Chinese patients with T2DM and DN was found in our study. Besides, the role of uric acid-lowering therapy in delaying DN progression and improving ESRD outcome had not yet been proven. Further study was needed to clarify the renal benefit of the uric acid-lowering therapy in the treatment of DN.

Early-onset of type 2 diabetes mellitus is a risk factor for diabetic nephropathy progression: a biopsy-based study.

Several studies show that patients with early-onset diabetes have higher risk of diabetic complications than those diagnosed in middle age. However, whether early-onset of type 2 diabetes mellitus (T2DM) is a risk factor for diabetic nephropathy (DN) progression remains unclear, especially a lack of data in biopsy-confirmed cohort. In This study, we enrolled 257 patients with T2DM and biopsy-confirmed DN to investigate the role of early-onset T2DM in DN progression. Participants were divided into two groups according to the age of T2DM diagnosis: early-onset group (less than 40 years) and later-onset group (40 years or older). We found that patients with early-onset T2DM had higher glomerular grades and arteriolar hyalinosis scores than those in later-onset group. After adjusted for confounding factors, early-onset of T2DM remained an independent predictor of end-stage renal disease (ESRD) for patients with DN. In conclusion, although with the comparable renal function and proteinuria, patients with early-onset T2DM and DN had worse renal pathological changes than those with later-onset. Early-onset of T2DM might be an important predictor of ESRD for patients with DN, which called more attention to early supervision and prevention for patients with early-onset T2DM and DN.

A Modified MTS Proliferation Assay for Suspended Cells to Avoid the Interference by Hydralazine and ß-Mercaptoethanol.

The 3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium (MTS) assay is one of the most commonly used tests of cell proliferation. Hydralazine has been reported to interfere with the performance of the MTS assay when used on adherent cells. This study aimed to investigate whether hydralazine interferes with the performance of the MTS assay on suspended cells. THP-1 (a monocytic leukemia cell line) cells were cultured in the presence or absence of hydralazine (0, 10, 50, 100, and 500 µM) for 2 or 24 h. Cell numbers were analyzed using the MTS, trypan blue exclusion, or microscopic assays. A modified version of the standard MTS assay was established by centrifuging the cells and replacing the test medium with fresh culture medium immediately before the addition of the MTS reagent. Culture of THP-1 cells with hydralazine at concentrations of 50, 100, and 500 µM for 2 h increased absorbance (p < 0.001) in the standard MTS assay, whereas both the trypan blue exclusion assay and microscopy suggested no change in cell numbers. Culture of THP-1 cells with 100 and 500 µm hydralazine for 24 h increased absorbance (p < 0.05) in the standard MTS assay; however, trypan blue exclusion and microscopy suggested a decrease in cell numbers. In a cell-free system, hydralazine (100 and 500 µM) increased absorbance in a time- and concentration-dependent manner. The modified MTS assay produced results consistent with trypan blue exclusion and microscopy using THP-1 cells. In addition, the modified MTS assay produced reliable results when K562 and Jurkat cells were incubated with hydralazine or ß-mercaptoethanol (ßME). In conclusion, a simple modification of the standard MTS assay overcame the interference of hydralazine and ßME when assessing suspended cells.

An Improved 3-(4,5-Dimethylthiazol-2-yl)-5-(3-Carboxymethoxyphenyl)-2-(4-Sulfophenyl)-2H-Tetrazolium Proliferation Assay to Overcome the Interference of Hydralazine.

The MTS [3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium] assay is one of the most commonly used assays to assess cell proliferation and cytotoxicity, but is subject to interference by testing compounds. Hydralazine, an antihypertensive drug, is commonly investigated in multiple fields such as heart failure, cancer, and blood pressure research. This study reported interference of the MTS assay by hydralazine and a simple modification overcoming this interference. Vascular smooth muscle cells were cultured in the presence or absence of hydralazine (0, 10, 50,100, and 500 µM) for 2 or 24 h. Cell numbers were analyzed using MTS, trypan blue exclusion, or microscopic assays. A modified version of the standard MTS assay was established, in which an additional step was added replacing the test medium, containing hydralazine, with fresh culture medium immediately before the addition of the MTS reagent. Culture with hydralazine at concentrations of 50, 100, and 500 µM for 2 h increased absorbance (p < 0.05) in the standard MTS assay, whereas microscopy suggested no change in cell numbers. Culture with 500 µm hydralazine for 24 h increased absorbance (p < 0.05) in the standard MTS assay, however, trypan blue exclusion and microscopy suggested a decrease in cell numbers. In a cell-free system, hydralazine (≥10 µM) increased absorbance in a concentration-dependent manner. The modified MTS assay produced results consistent with trypan blue exclusion and microscopy. In conclusion, a simple modification of the standard MTS assay overcame the interference of hydralazine and may be useful to avoid interference from other tested compounds.

Diabetes Reduces Severity of Aortic Aneurysms Depending on the Presence of Cell Division Autoantigen 1 (CDA1).

Diabetes is a negative risk factor for aortic aneurysm, but the underlying explanation for this phenomenon is unknown. We have previously demonstrated that cell division autoantigen 1 (CDA1), which enhances transforming growth factor-ß signaling, is upregulated in diabetes. We hypothesized that CDA1 plays a key role in conferring the protective effect of diabetes against aortic aneurysms. Male wild-type, CDA1 knockout (KO), apolipoprotein E (ApoE) KO, and CDA1/ApoE double-KO (dKO) mice were rendered diabetic. Whereas aneurysms were not observed in diabetic ApoE KO and wild-type mice, 40% of diabetic dKO mice developed aortic aneurysms. These aneurysms were associated with attenuated aortic transforming growth factor-ß signaling, reduced expression of various collagens, and increased aortic macrophage infiltration and matrix metalloproteinase 12 expression. In the well-characterized model of angiotensin II-induced aneurysm formation, concomitant diabetes reduced fatal aortic rupture and attenuated suprarenal aortic expansion, changes not seen in dKO mice. Furthermore, aortic CDA1 expression was downregulated ∼70% within biopsies from human abdominal aortic aneurysms. The identification that diabetes is associated with upregulation of vascular CDA1 and that CDA1 deletion in diabetic mice promotes aneurysm formation provides evidence that CDA1 plays a role in diabetes to reduce susceptibility to aneurysm formation.

Role of Cell Division Autoantigen 1 (CDA1) in Cell Proliferation and Fibrosis.

Cell Division Autoantigen 1 (CDA1) was discovered following screening a human expression library with serum from a patient with Discoid Lupus Erythematosus. CDA1, encoded by TSPYL2 on the X chromosome, shares anti-proliferative and pro­fibrotic properties with TGF-b. It inhibits cell growth through p53, pERK1/2 and p21­mediated pathways and is implicated in tumorigenesis and the DNA damage response. Its pro-fibrotic property is mediated through cross-talk with TGF-b that results in upregulation of extracellular matrix proteins. The latter properties have identified a key role for CDA1 in diabetes associated atherosclerosis. These dual properties place CDA1 as an attractive molecular target for treating tumors and vascular fibrosis including atherosclerosis and other vascular disorders associated with enhanced TGF-ß action and tissue scarring.

Antiproliferative autoantigen CDA1 transcriptionally up-regulates p21(Waf1/Cip1) by activating p53 and MEK/ERK1/2 MAPK pathways.

We previously reported that overexpression of cell division autoantigen 1 (CDA1) in HeLa cells arrests cell growth and inhibits DNA synthesis at S-phase. Here we show that CDA1-induced arrest of cell growth is accompanied by increases in protein and mRNA levels of the cyclin-dependent kinase (Cdk) inhibitor protein, p21(Waf1/Cip1) (p21). Both p21 induction and cell growth arrest are reversed when CDA1 expression is inhibited. CDA1 also increases p53 protein, but not its mRNA, in a time- and dose-dependent manner. MDM2, a ubiquitin ligase regulating p53 degradation, is inactivated by CDA1, suggesting that p53 protein accumulation is due to decreased protein degradation. Knockdown of p53, using siRNA targeting two sites of p53 mRNA, abrogates transcriptional induction of p21 by CDA1. Deletion of the p53 responsive element in the distal region of p21 promoter attenuates promoter activity in response to CDA1. DNA damage caused by camptothecin treatment increases mRNA and protein levels of CDA1, accompanied by induction of p53. The DNA damage-induced p53 induction is markedly attenuated by CDA1 knockdown. CDA1 induces phosphorylation of ERK1/2(p44/42), an activity blocked by PD98059 and U0126, inhibitors of the upstream kinase MEK1/2. The MEK inhibitors also block induction of p21 mRNA and abrogate p21 promoter activity stimulated by CDA1. Cell cycle kinases, Cdk1, -2, -4, and -6 are inhibited by CDA1 overexpression. We conclude that CDA1 induces p53- and MEK/ERK1/2 MAPK-dependent expression of p21 by acting through the p53 responsive element in the p21 promoter and that this contributes to its antiproliferative activity.

17 beta-hydroxysteroid dehydrogenase type XI localizes to human steroidogenic cells.

We searched expressed sequence tag databases with conserved domains of the short-chain alcohol dehydrogenase superfamily and identified another isoform of 17 beta-hydroxysteroid dehydrogenase, 17 beta HSDXI. This enzyme converts 5 alpha-androstane-3 alpha, 17 beta-diol to androsterone. The substrate has been implicated in supporting gestation and modulating gamma-aminobutyric acid receptor activity. 17 beta HSDXI is colinear with human retinal short-chain dehydrogenase/reductase retSDR2, a protein with no known biological activity (accession no. AAF06939). Of the proteins with known function, 17 beta HSDXI is most closely related to the retinol-metabolizing enzyme retSDR1, with which it has 30% identity. There is a polymorphic stretch of 15 adenosines in the 5' untranslated region of the cDNA sequence and a silent polymorphism at C719T. A 17 beta HSDXI construct with a stretch of 20 adenosines was found to produce significantly more enzyme activity than constructs containing 15 or less adenosines (43% vs. 26%, P < 0.005). The C719T polymorphism is present in 15% of genomic DNA samples. Northern blot analysis showed high levels of 17 beta HSDXI expression in the pancreas, kidney, liver, lung, adrenal, ovary, and heart. Immunohistochemical staining for 17 beta HSDXI is strong in steroidogenic cells such as syncytiotrophoblasts, sebaceous gland, Leydig cells, and granulosa cells of the dominant follicle and corpus luteum. In the adrenal 17 beta HSDXI, staining colocalized with the distribution of 17 alpha-hydroxylase but was stronger in the mid to outer cortex. 17 beta HSDXI was also found in the fetus and increased after birth. Liver parenchymal cells and epithelium of the endometrium and small intestine also stained. Regulation studies in mouse Y1 cells showed that cAMP down-regulates 17 beta HSDXI enzymatic activity (40% vs. 32%, P < 0.05) and reduces gene expression to undetectable levels. All-trans-retinoic acid did not affect 17 beta HSDXI expression or activity, but addition of the retinoid together with cAMP significantly decreased activity over cAMP alone (32% vs. 23%, P < 0.05). Cloning and sequencing of the 17 beta HSDXI promoter identified the potential nuclear receptor steroidogenic factor-1 half-site TCCAAGGCCGG, and a cluster of three other potential steroidogenic factor-1 half-sites were found in the distal part of intron 1. Collectively, these results suggest a role for 17 beta HSDXI in androgen metabolism during steroidogenesis and a possible role in nonsteroidogenic tissues including paracrine modulation of 5 alpha-androstane-3 alpha, 17 beta-diol levels. 17 beta HSDXI could act by metabolizing compounds that stimulate steroid synthesis and/or by generating metabolites that inhibit it.