Part1: Diffusion-magnetic Resonance Imaging Predicts Decline Of Kidney Function in Chronic Kidney Disease And in Patients With A Kidney Allograft

For more info. contact tina.xiang@wecistanche.com

Kidney cortical interstitial fibrosis is highly predictive of kidney prognosis and is currently assessed by evaluation of a biopsy. Diffusion-weighted magnetic resonance imaging is a promising non-invasive tool to evaluate kidney fibrosis. We recently adopted a diffusion-weighted imaging sequence for discrimination between the kidney cortex and medulla and found that the cortico-medullary difference in apparent diffusion coefficient (△ADC) correlated with histological interstitial fibrosis. Here, we assessed whether △ADC, as measured with diffusion-weighted magnetic resonance imaging, is predictive of kidney function decline and dialysis initiation in chronic kidney disease (CKD) and patients with a kidney allograft in a prospective study encompassing 197 patients. We measured AADC in 43 patients with CKD (estimated GFR(eGFR)55ml/min/1.73m)and 154 patients with a kidney allograft (eGFR 53ml/min/1.73m).Patients underwent a kidney biopsy and diffusion-weighted magnetic resonance imaging within one week of biopsy; median follow-up of 2.2 years with measured laboratory parameters. The primary outcome was a rapid decline of kidney function (eGFR decline over 30% or dialysis initiation) during follow-up. Significantly, patients with a negative △ADC had 5.4 times more risk of the rapid decline of kidney function or dialysis (95% confidence interval: 2.29-12.58). After correction for kidney function at baseline and proteinuria, low ADC still predicted significant kidney function loss with a hazard ratio of 4.62(95%confidence interval 1.56-13.67) independent of baseline age, sex,eGFR, and proteinuria. Thus, low AADC can be a predictor of kidney function decline and dialysis initiation in patients with native kidney disease or kidney allograft, independent of baseline kidney function and proteinuria.

The reduction of renal impairment is a crucial parameter to individualize treatment and follow-up in patients with chronic kidney disease (CKD). Tools using biological parameters, such as estimated renal function (estimated glomerular filtration rate [eGFR]) and proteinuria, can help in this prediction. Although valuable, these tools probably lack some personalized aspects of kidney disease. Renal lesions assessed by biopsies are predictive of the renal evolution, independently of renal function, and are usually more informative on individual prognosis than biochemical parameters only.' However, renal biopsies present a sampling bias, and they carry some risks and can therefore not be performed on all patients, nor repeatedly to serve as a prognostic tool. A need exists for non-invasive methods to evaluate renal parenchyma, but also to predict the individual renal evolution.

Diffusion-weighted magnetic resonance imaging (MRI) is an imaging method sensitive to the Brownian motion of water molecules in the tissue that can be used to assess tissue structure in multiple organs.'The apparent diffusion coefficient (ADC)obtained from diffusion-weighted MRI has emerged over the past years as an important measure to evaluate kidney interstitial fibrosis (IF) noninvasively.* A negative correlation between ADC and the renal fibrosis assessed by biopsy has been reported by several groups. " As supported by a recent meta-analysis, diffusion-weighted MRI may be a promising tool to diagnose and classify early CKD diseases. ' However, whether diffusion-weighted MRI could also predict the renal function evolution is currently not known. In the 122 participants of the CKD Optimal Management With Binders and Nicotinamide trial, baseline ADC was associated with a decrease in egg rover time in the l-year observation period. "In a 5-year follow-up of 91 patients with the various stage of CKD, the eGFR decline was not associated with the baseline ADCbut with the baseline eGFR.2 The difference between both these studies may be related to the patient population and study design as well as the MR method used. Substantial improvement in the assessment of renal fibrosis by diffusion-weighted MRI can be obtained by using the corticomedullary difference of ADC(AADC) instead of cortical or renal ADC. Subtracting the medullary from the cortical ADC in each patient allows for lower interindividual variability of the measured index.AADC was better correlated to IF than any other histologic parameters, including inflammation." When △ADC measurements were repeated in a CKD patient,△ADC variation over time was better associated with IF and tubular atrophy progression than eGFR, which is a relatively late marker of parenchymal kidney loss.

In this study, we thus assessed the role of diffusion-derived AADC in the prediction of renal evolution in a mixed cohort of 197 patients with native kidney or kidney allograft patients followed up over 5 years. During follow-up, laboratory parameters (creatinine and proteinuria)were measured.

METHODS 

Patients

We designed a prospective study, including adult kidney allograft recipients and CKD patients. Every patient aged ≥18 years followed up in our hospital and planned for a kidney biopsy for clinical purposes from August 2013 to October 2018, was eligible for enrollment in studies assessing the role of MRI in noninvasive kidney diagnosis, as previously described. Exclusion criteria were the presence of a pacemaker or other magnetic resonance incompatible device, pregnancy, claustrophobia, and patient refusal. MRI was scheduled on the same day as the biopsy whenever possible, or within 1 week. In all patients, additional fasting serum and urine were collected and stored at -80°C. Patients were then followed up at least yearly by the physicians of the outpatient clinic of Nephrology of our hospital. The study was approved by our local ethical committee for human studies(Commission Ethique de la Recherche [CER] l1-160) and performed according to the Declaration of Helsinki principles. All the patients were contacted to provide written informed consent to participate in this prospective study. None of the patients was from a vulnerable population, and all patients provided written informed consent, which was freely given. In the present study, all patients included in our initial study and having benefitted from research MRI were included, and their follow-up was analyzed (Figure 1).

Laboratory measurement

Baseline characteristics, including medical history, comorbidities, and treatment, were collected through patient records. Patients'blood pressure, weight, and size were measured routinely during follow-up visits. Serum creatinine and other standard laboratory values were measured during routine follow-up visits or hospitalizations. Standard biochemical analyses were performed in the hospital laboratory using routine automated analyzers. The eGFR was calculated using the CKD Epidemiology Collaboration equation. Creatinine was measured by Jaffe kinetics using isotopic dilution mass spectroscopy-traceable methods. When 24-hour urine protein was not available, the protein-to-creatinine ratio was used to estimate proteinuria.

Outcomes

The primary outcome was a decline of eGFR of >30% or renal replacement therapy. For participants who died, the last available eGFR was used to assess the primary outcome.

Histologic fibrosis quantification

Renal fibrosis was assessed quantitatively on the kidney biopsy specimen by the Pathology Department of our hospital, using Masson trichrome-stained kidney sections. The expert pathologist (SM)was blinded to the other results, including eGFR and MRI. Expert evaluation of fibrosis is recommended to evaluate IF and is reproducible. This is the current gold standard in most pathology services.5 The severity of renal fibrosis was scored from 0% to 100% for each patient and reported on the clinical biopsy report independently of our study. The 0% to 100% of renal fibrosis refers to the percentage of cortical IF (extent of fibrosis in the cortical tissue present in the renal biopsy) assessed in a semiquantitative manner by the pathologist using the trichrome staining. To verify the reproducibility of this evaluation, 60 random sections were evaluated blindly by 2 experienced nephrologists(SDS and LB).This repeated fibrosis evaluation displayed a good correlation to pathologic evaluation (intraclass correlation coefficient, 0.92;95% confidence interval [CI],0.87-0.95). Furthermore, renal fibrosis was quantified using the Banff criteria in renal allograft patients: ci(IF) and ct (tubular atrophy)with a minimum score of 0 and a maximal score of 6. Because of a good correlation between the 2 methods(r=0.86; P<0.001), we used subjective histologic renal fibrosis as a continuous variable (0%-100%)for all analyses.

Magnetic resonance imaging. Patients were scanned on a Prisma 3-T MR(Siemens AG) with the standard 32-element spine coil and the 18-element phased-array abdominal coil. MRI protocol parameters are summarized in Table 1. The analysis of the MRI images was blinded to all other markers. Regions of interest were determined on the T1 map （about 1 cm² for the cortex and 20 mm²in 3 medullae zones), as previously described, and copied on the ADC map, produced by the Siemens MR system, which uses a monoexponential fitting model. This model was used based on previous studies.2 △ADC, the corticomedullary difference, was used to reduce intersubject variability and provide an MRI index of fibrosis. AADC was calculated as follows: cortical ADC-medullary ADC. AADC was separated into 3 subgroups based on previous studies.32] All focal pathologic areas(cyst, scar, and hematomas)were avoided in the region of interest placement, aiming to cover a large and representative part of the cortex and medulla.

Reproducibility has been assessed in our protocol. Strong reproducibility of ADC in the cortex and medulla was found between 2 readers. For each patient independently, all intraclass correlation coefficients were superior to 0.91(95% CI,0.92-0.99)for ADC cortex, ADC medulla, and △ADC. Correlation coefficients between the 2 readers were R'=0.96 for the ADC evaluation in the cortex, R²=0.97 in the medulla， and R2=0.95 for the △ADC（P<0.05）.

Statistical analysis

Continuous variables are expressed as mean ± SD or median and interquartile range, according to the distribution. Categorical variables are expressed as numbers and percentages. The statistical significance was determined as P<0.05, and all tests were 2-sided. To test the hypothesis that MRI and clinical parameter tests could predict a decline in renal function or new dialysis, we performed log-rank tests for trends comparing each variable to the risk of renal function loss. Time-to-event data were evaluated using Kaplan-Meier estimates and Cox proportional-hazard models. Survival curves were compared using the log-rank test. Hazard ratios(HRs), 95% CIs, and P values were calculated with the use of Cox models. Proportionality of hazard was graphically verified by plotting log minus log of survival against time. The selection of covariates(eGFR and proteinuria) in multivariable analyses was based on prior knowledge from the scientific literature. The discriminative performance of clinical parameters (eGFR and proteinuria) and △ADC to predict a decline in renal function was assessed by a composite score. For a specific patient, the composite score is calculated as the sum of the regression coefficients corresponding to his/her characteristics (Table 2). The higher the prognostic index, the higher the level of risk predicted by the model. The 3-year free eGFR decline survival, according to the value of a predictor (composite score or proteinuria), was assessed by using a nonparametric method based on Kaplan-Meier's approach and nearest neighbor's approach with the package problem for R (version 2019.11.13).

Statistical analyses were performed using STATA 13.1(StataCorp)and R(R Foundation for Statistical Computing).