Part2: Diffusion-magnetic Resonance Imaging Predicts Decline Of Kidney Function in Chronic Kidney Disease And in Patients With A Kidney Allograft
Jul 01, 2022
For more info. contact tina.xiang@wecistanche.com
RESULTS
Characteristics of the study population
From August 2013 to October 2018, we included 197 CKD patients, mainly White(90%) and male(68%), with a mean age of 54(±14)years, undergoing kidney biopsy for clinical reasons. Of the 197 patients, 154(78%)were kidney allograft patients and 43(22%) were native kidney patients (Figure 1 and Supplementary Figure S1).
Baseline characteristics are presented in Table 3. Biopsy indications were made by the nephrologist in charge of the patients, as clinically justified, and independently of this study. For native kidney disease, most of the indications were abnormal urinary microscopy and proteinuria, and/or acute or chronic renal dysfunction. For allograft patients, biopsy indications were routine biopsies (at 1 year, after steroid withdrawal), the elevation of creatinine levels, and the apparition of proteinuria or de novo donor-specific antibodies. MRI was done within 1 week of the biopsy.△ADC was available in 188 of the 197 total patients.
Click here to get your cistanche experience
Univariable and multivariable analysis of predictors of renal function decline
In this cohort, the median follow-up time from the biopsy was 2.2 years (interquartile range,1.1-3.7 years). Diagnosis of the rapid decline of renal function was defined as eGFR decline>30 ml/min per 1.73 m²or initiations of dialysis during follow- Rapid decline of renal function occurred in 54 pa-up.
patients after a median time of 1.1 years (interquartile range, 0.9-2.1 years). Median follow-up in the non-rapid decline of renal function patients was 2.9 years(interquartile range,1.8-4.0 years). During the follow-up,11 patients died, including 5 patients classified as having a rapid decline of renal function before their death. The 6 remaining patients were considered the non-rapid decline of renal function responders based on the last available eGFR.
Recognized clinical predictors of the rapid decline of renal function, such as sex, age,eGFR, and proteinuria, were included in a Cox survival analysis. By univariate analysis, eGFR at baseline and proteinuria were associated with the rapid decline in renal function (Table 4 and Supplementary Figure S2). Moreover, a negative △ADC was associated by univariate analysis with the rapid decline of renal function(HR, 5.4;95% CI,2.29-12.58; P<0.001; Table4 and Figure 2). This result was confirmed in both kidney allograft patients (HR,3.88;95% CI,1.81-10.9;P=0.003)and CKD patients (HR,4.7;95% CI,1.45-15.5;P = 0.01;Supplementary Table S1).A decrease of △ADC was more associated(HR, 5.4 ys. 0.70)to rapid decline of renal function than cortex ADC（cortex ADC >1735 and ≤1891 × 10~6mm²/s∶HR,0.70;95% CI,0.37-1.33;P= 0.273;cortex ADC>1891 ×10~mm²/s∶ HR，0.39;95% CI，0.19-0.78;
P = 0.008). In Figure 3,2 representative examples are shown: 2 patients presented with a creatinine of 110 to 120 μmol/L of creatinine and no proteinuria at baseline. Patient 1 displayed a positive △ADC and had a good evolution at 3 years follow-up（creatinine level of 119 μmol/L）， whereas patient 2 displayed a negative △ADC and had an increase of creatinine to 178 umol/L at 4 years follow-up. Thus, our tool may identify patients with a worse prognosis despite similar baseline clinical characteristics.
By multivariable analysis, a negative △ADC, low eGFR, and high proteinuria were independently associated with the rapid decline of renal function(Table 5). The HR was the highest for a negative baseline △ADC. In another multivariable analysis, the absolute cortex ADC value was not associated with the rapid decline in renal function (Supplementary Table S2).
The composite score ranged from 0 to 3.5(Supplementary Table S3). The higher the score, the higher the risk of renal function decline: from 6.6% for a score of 0 to 74.5% for a score of 3(Figure 4a).
We determine 3 levels of risk based on the composite score value. Values of <0.80 correspond to a 3-year risk of<20%(low-risk category), values between 0.81 and 2.00 correspond to risk between 20% and 50%(intermediate-risk category), and values of >2.00 correspond to a risk of ≥50%(high-risk category).In these categories, the observed 3-year risk of decline in renal function (Figure 5a) matched with the predicted risk based on the composite score values(Figure 5b).
The relationship between each component of the composite score and the 3-year free decline survival is shown in Figures 4b through d. The risk estimation based on proteinuria only increases importantly when proteinuria increases from 0 to 3 g/24 h but remains around 60% for proteinuria values >3 g/24 h(Figure 4d).eGFR also displayed a threshold effect at ≈45 ml/min per 1.73 m*. Only our combined risk score displayed a mostly linear relationship to actual survival.
Proteinuria is a major predictive factor for CKD prognosis. To better determine the potential value of our composite score, we studied its performance in comparison to proteinuria. Taking into account only the predictive value of proteinuria, patients with lower than 0.17 g/24 h have a low risk of progression of 20%. However, in this subgroup, △ADC varied importantly across patients(AADC: from -184 to 296× 10-°mm2/s). Therefore, despite low proteinuria, the composite score also varied importantly: from 0 to 2.50 (corresponding respectively to a 3-year risk based on composite scores of 6.6% and 51.8%). Thus, the risk assessment based on proteinuria and based on the composite score leads to different risk classifications in some patients. Of the 95 patients classified at low risk by proteinuria, 50% were classified at intermediate risk by the composite score and 43%were still classified at low risk and 2 at high risk. The difference in survival between these 2 groups of patients was confirmed to be statistically significant (log-rank test, P= 0.04).
To further determine whether the composite score increased performance to predict a rapid decline of renal function compared with proteinuria alone, we compared the observed survival with the survival predicted by proteinuria or the composite score. As shown in Figure6, the observed survival was closer to the survival predicted by the composite score than to the survival predicted by the proteinuria. In low-risk patients at 3 years, proteinuria predicts risk of 14.2% and the composite score predicts a risk of 9.4% for an observed risk of 7.5%. At 4 years, proteinuria predicts a risk of 20.53%, the composite score predicts risk of 9.4%, and the observed risk is 7.5%.In intermediate-risk patients at 3 years, proteinuria predicts a risk of 14.2% and the composite score predicts risk of 24.7% for a risk observed of 24.5%. At 4 years, proteinuria predicts risk of 19.9%, the composite score predicts a risk of 33.5%, and the risk observed is 37%.
As described in previous studies,△ADC correlated with IF (r = -0.56; P<0.001; Supplementary Figure S3). IF was associated with rapid decline in renal function(fibrosis>25%and≤50%:HR,1.95;95% CI,1.05-3.64; P=0.035;fibrosis >50%:HR,7.82;95% CI,3.90-15.69;P<0.001).Adding fibrosis to the multivariable analysis did not improve the prediction of the model (Supplementary Table S4).
DISCUSSION
In a mixed population of 197 patients, including 154 kidney allograft patients and 43 patients with CKD, followed up for 5 years (median of 2.2 years), a negative △ADC value was predictive of a worse renal outcome. This predictive value was observed in both patients with native kidney disease and kidney allograft recipients and was independent of baseline age, sex, eGFR, and proteinuria.
Our results are in apparent disagreement with the recent study by Sugiyama et al. They did not observe a correlation of eGFR decline with cortical ADC over 5 years in a single-center, longitudinal, retrospective observational design of 91 patients. The CKD stage of the patients could not explain the difference as the baseline eGFR was close (eGFR 53.8± 24 ml/min per 1.73 m²in our study vs.49.2± 28.9 ml/min per 1.73 m² in the study of Sugiyama et al.）. The inclusion of both native and kidney transplant recipients in our study could not explain the difference as our results are still valid in the subgroup of patients with native kidneys. Important methodological differences between both studies may explain the different results. First, the study by Sugiyama et al. was conducted on a 1.5-T MR system, whereas our study was performed on a 3-T MR system equipped with the strongest clinical gradients available on the market. The strength of the static magnetic field and gradients has been well recognized as a key factor in the improvement of diffusion MRI. Furthermore, we used a RESOLVE diffusion sequence based on segmented acquisition rather than the single-shot approach of Sugiyama et al., according to the best overall performance of this sequence for prostate and renal The RESOLVE sequence allows much shorter imaging." echo train with reduced signal blurring due to T2* decay and yields a better correlation of ADC and renal fibrosis than traditional single-shot echo-planar imaging MR sequences.1 Finally, the increased image quality of the RESOLVE diffusion sequence allows a combined ADC measurement from both the cortex and medulla (namely, the corticomedullary ADC difference△ADCD), which corrects for some extended interindividual ADC variations that are known for absolute ADC values. We have shown previously that △ADC was reproducible and better associated with IF in CKD patients than cortical ADC alone.3,4 In the present study,△ADC also outperformed ADC for the prediction of renal function evolution and was more robust in multivariable analysis. Therefore, we believe that the technical improvements used in our study are the main explanation for the different results from the study by Sugiyama et al. The recent study of Srivastava et al,' performed also on a 3-T MR system, confirmed an association between eGFR decline and cortical ADC and supports our results. In their study, albuminuria abrogated the predictive value of cortical ADC. Similar results were observed in our study with ADC but not AADC, which remained an independent predictor of renal function evolution. These observations call for a better uniformization of diffusion MRI techniques to allow optimal use of this technique in the clinical setting with the systematic use of the corticomedullary difference for the measure of ADC.
Histologically, IF is the parameter displaying the best association to renal function and has the largest predictive value for renal function evolution.1Because ADC and AADC are surrogates for IF, it is not surprising that they also predict renal function evolution. Correcting for IF decreased slightly the independent predictive value of △ADC(from 4.69 to 4.32)but did not abrogate it. This result suggests that diffusion MRI is also dependent on factors other than IF. It may be affected by inflammation, for example, and renal perfusion, as well as other unknown parameters. It may also suggest that the MRI assessment of IF is different from the one performed by biopsy given the largest sample of cortex analyzed using MRI. △ADC is therefore a good predictor of renal function evolution, because it may cumulate the effects of several parameters involved in CKD progressions, such as capillary rarefaction, perfusion, and fibrosis, for example.
Previous studies have shown that blood oxygen level-dependent (BOLD)-MRI could also predict the evolution of the renal disease in native kidneys. Given our observations, the value of BOLD-MRI for prognosis assessment in kidney allograft recipients would be valuable. Moreover, the added value of multiparametric MRI for renal prognosis would be of interest.
One limitation of our study is its monocentric design. Another limitation is the follow-up in the median of 2.5 years and not 5 years, but we included a mixed population of patients (CKD and kidney allograft patients) and the number of patients included is high. We used only one modality of MRI; multiparametric MRI could be of interest for further studies. We use a monoexponential fit for the analysis of our data as our previous study showed a better correlation of AADC than diffusion coefficients(AD)obtained from intravoxel incoherent motion. Moreover, △ADC can be measured directly from the Siemens ADC maps of the RESOLVE sequence and could therefore improve the reproducibility of our results by other groups. However, evaluation of more advanced diffusion models on our data could be of interest in further studies. The last limitation is that we included fewer patients with emergency biopsies and acute kidney injury, given the design of our study.
Altogether, we show herein that AADC is predictive of kidney outcome independently of biochemical parameters. Diffusion MRI may be of value in better assessing the individual renal prognosis, also in patients in whom biopsy is difficult, or not indicated. We showed that baseline AADC was predictive of the worst evolution, independently of biochemical parameters, including eGFR. We propose that AADC could be used in addition to biochemical parameters to predict the individual outcome and tailor follow-up of given patients with renal disease, in native and allograft kidneys.
