The Role Of The Renal Arterial Resistance Index For The Differential Diagnosis Of Diabetic Kidney Disease And Non-diabetic Kidney Disease

Background:Globally,the epidemic of diabetes mellitus(DM)and its complications have emerged as an increasing major health challenge to all countries.Diabetic kidney disease(DKD)is now one of the most frequent and severe complications of diabetes and continues to be a major cause of end-stage kidney disease(ESKD)worldwide.However,non-diabetic kidney disease(NDKD)is also very common in diabetic patients with chronic kidney disease(CKD).The therapy and prognosis of NDKD are pretty different from DKD.It is believed that the renal outcomes of patients with DKD are significantly worse compared with their counterparts with biopsy-proven NDKD because renal pathological lesions in DKD are deemed difficult to reverse.NDKD patients are potentially misdiagnosed with DKD and could not get proper treatment.Therefore,the differential diagnosis between DKD and NDKD is crucial.Currently,the renal pathological diagnosis is the gold standard to discriminate DKD from NDKD.However,the kidney biopsy is invasive and not impracticable in patients with contraindications,such as pyknotic kidney,bleeding tendency,solitary kidney,uncontrolled hypertension,or severe anemia.Moreover,a renal biopsy could not be routinely performed in some primary hospitals.Thus,the diagnosis and appropriate treatment were usually based on clinical indicators,such as diabetes duration,hematuria,diabetic retinopathy(DR),and other indices.However,those markers are not perfect enough to meet clinical requirements.Therefore,it is necessary to find a newly precise and sensitive non-invasive marker for clinical differentiation of DKD from NDKD.Doppler ultrasound is a low-cost and non-invasive tool in detecting kidney diseases.The Renal atrial resistance index(RI),measured by Doppler ultrasound,has been extensively used to evaluate renal blood flow as a semi-quantitative parameter in the differential diagnosis of numerous kidney diseases.Previous studies noticed that RI in patients with DKD is significantly higher when compared with non-diabetic controls,which might be helpful for the identification and prediction of DKD.However,the potential role of RI in the clinical differentiation of DKD from NDKD and the optimal cutoff value remains to be illuminated.Thus,the present study was conducted to evaluate whether RI can be helpful in the differential diagnosis of DKD from NDKD and establish a better-quantified differential diagnostic model.Contents and methods:We consecutively reviewed 469 patients with type 2 DM who underwent a kidney biopsy at the department of nephrology at Xinqiao Hospital,Army Medical University,from January2014 to September 2020.All patients had received echo-color-Doppler examination of renal vessels,systematic screening for DR,and renal biopsy.Eligible patients were classified into the DKD group and the NDKD group based on the kidney biopsy results.The optimal cutoff point of the RI to predict DKD and the diagnostic implications of RI were explored by receiver-operating characteristic(ROC)curve analysis.Univariate and multivariate logistic regression analyses(stepwise forward)were performed to identify independent risk factors for the occurrence of DKD.Then,we used two multivariate logistic regression analyses to establish two differential diagnostic models(with or without RI value)to explore the value of RI for the clinical differentiation between DKD and NDKD.Then a novel diagnostic model was selected using Delong test performed by R language.Finally,a back-substitution and a validation test(a validation cohort of 96 cases)were conducted to evaluate the new model.The main results:1.A total of 469 eligible patients were categorized into two groups according to the renal biopsy results(337 in the DKD group and 132 in the NDKD group).Significantly higher levels of RI were observed in the DKD group as compared with the NDKD group(0.70 vs.0.63,p< 0.001).2.The area under the curve(AUC)of RI was 0.785.The optimal cutoff value of RI for predicting DKD was 0.66 with high specificity(80.9 %)but relatively low sensitivity(69.2%).3.RI ≥ 0.66 showed statistical significance in the multivariate logistic regression analysis as an independent risk factor for the DKD diagnosis,as well as the duration of diabetes ≥ 60 months,BMI,DR,and Gh ≥ 7%.The area under ROC curve of the traditional model was 0.889.After adding RI,the area under the ROC curve of the RI-based model increased to 0.912.The statistical significance of two ROC AUC by De Long’s test(Z =2.5964,p< 0.01)implied that the RI-based model has improved the efficiency of differential diagnosis.4.The validation cohort test demonstrated that the RI-based model had a sensitivity of81.5% and a specificity of 78.6%.The ROC curve(AUC=0.857)based on the validation cohort indicated that the RI-based model had good diagnostic efficiency.Conclusion:RI might serve as a potential tool for identifying DKD from NDKD,and the best predictive cutoff value of RI for DKD diagnosis was 0.66.The newly established model,which contains RI,showed good performance in the differential diagnosis of DKD and NDKD.