We performed a prospective study from April 2015 to July 2016. A total of 107 adult patients with ESRD and 25 health volunteers were enrolled in this study and underwent CMR examination. The inclusion criteria were patients with kidney damage lasting more than 3 months as assessed by a decline in kidney function with a constant GFR < 15 mL/min per 1.73 m2 who requiring dialysis [16]. Exclusion criteria included coronary artery obstruction identified by coronary CT angiography (n = 25); presence of other congenital cardiac disease, myocardial infarction or inherited cardiomyopathy (n = 4); incomplete CMR T1 mapping scanning (n = 7); and poor CMR images (n = 5). After exclusion, 66 ESRD patients remained. All patients underwent hemodialysis regularly (twice weekly). In addition, 25 individuals who had no chronic diseases, cardiovascular diseases, family history of cardiovascular disease, diabetes mellitus, hypertension, or renal diseases were enrolled as normal controls. The subjects were informed as to possible adverse reactions to magnetic resonance contrast, and written informed consent was obtained from all subjects prior to the examination.

The height, weight, and brachial blood pressure of the subjects were recorded, and body mass index (BMI) and body surface area (BSA) values were calculated. Fasting venous blood from the ESRD patients was drawn for routine blood examination, hemoglobin (Hb), hematocrit (Hct), and renal function. The estimated glomular filtration rate (eGFR) was calculated from serum creatinine by using the CKD-Epidemiology Collaboration formula [17].

A 3.0-T whole-body scanner with an 18-element body phased array coil (Skyra; Siemens Medical Solutions, Erlangen, Germany) was used for scanning. The manufacturer’s standard ECG-triggering device and the breath-hold technique were used to monitor each participant’s ECG and breathing changes throughout the examination. All participants were examined in the supine position. Each sequence was acquired within an end-expiration breath-holding period by using ECG-triggered acquisition. By using a TrueFISP sequence, transverse, coronal, and sagittal plane localizing images were obtained. SSFP sequences (TR 39.34 ms, TE 1.22 ms, slice thickness 8.0 mm, field of view 284.42 × 340.00, flip angle 60 deg) were performed to acquire the 8–12 continuous CMR cine sections located from the mitral-valve level to the left ventricular apex in the short-axis view. Vertical 2-chamber long axis and horizontal 4-chamber cine series were scanned using the same sequences as used with the short-axis images. Then, 0.15 mmol/kg gadolinium chelate contrast agent (Gadodiamide, GE Healthcare, Ireland) was intravenously injected. A contrast-enhanced Modified Look-Locker inversion recovery (MOLLI) T1 mapping sequence with inline motion correction was performed. Native MOLLI images were scanned before contrast administration. Post MOLLI images were acquired 10 min after contrast injection. Both pre-contrast (native) (field of view 360.00 × 306.56, section thickness 8.00, flip angle 35 deg, TR 348.56 ms, TE 1.12 ms) and post-contrast (field of view 360 × 306.55, section thickness 8.00, flip angle 35 deg, TR 426.5 ms, TE 1.12 ms) MOLL sequences were assessed from the three corresponding short-axis sections of the left ventricle, which were located at the basal, middle and apex level. Short-axis color-scale parametric native and post-contrast T1 maps were created. Late gadolinium enhancement IR-prepared images were scanned 15 min following the administration of contrast agent at the same section of the basal, middle, and apex level of T1 mapping. All subjects were stable throughout the examination.

All CMR images were analyzed by using commercially available software (cmr42, version 5.6.4; Circle Cardiovascular Imaging Inc., Calgary, Canada). To measure the segment heterogeneity, native T1 and post-contrast T1 values were calculated by drawing endo- and epi-cardial borders on a series of three short-axis pre- and post-contrast MOLLI images. The partition coefficient lambda (λ) and ECV were computed as follows: ECV = λ(1-hematocrit); λ = (1/T1 myocardium post-contrast-1/T1 myocardium-native)/(1/T1 blood post-contrast-1/T1 blood-native). ECV maps and values were automatically obtained. LGE extent (%) and volume (ml) were assessed on LGE images by tracing the endo- and epi-cardial boundaries, and 5 SD was used as the threshold defining the appearance of LGE in comparison with normal zones; LGE was defined as being diffuse and present only if it was identified by 2 independent viewers (Xu and Zhang) [18]. The LV function and strain parameters, including LV ejection fraction (LVEF), end diastolic volume (EDV), end-systolic volume (ESV), stroke volume (SV), mass, global radial strain (GRS), global circumferential strain (GCS), and global longitudinal strain (GLS) were calculated. LV dysfunction was defined as LVEF< 50%.

We began to follow up the ESRD patients every 3 month immediately after the CMR examination. When HF was happened, this time was defined as endpoint time. If no HF was happened, continuous follow-up would be performed every 3 month until HF occurrence. All the patients were followed up to December 2018. If some patients did not have HF until December 2018, these patient’s data were censored. Follow up data were acquired each time. We followed up the patients by telephone and called them to come back hospital for the typical symptoms, signs and echocardiography data collecting of HF according to the definition in the “2016 ESC Guidelines for the diagnosis and treatment of acute and chronic heart failure: The Task Force for the diagnosis and treatment of acute and chronic heart failure of the European Society of Cardiology (ESC) Developed with the special contribution of the Heart Failure Association (HFA) of the ESC [19].” Complete follow-up was available for all patients in this study.

All statistical analyses were performed by using commercially available software packages (SPSS version 21.0, Armonk, NY; GraphPad version 7.00, San Diego, CA). All data were assessed for normality by using the Kolmogorov–Smirnov test and presented as the mean ± SD or median (quartile). The independent t-test and Mann–Whitney U-test were used to compare characteristics between the normal and ESRD groups. Comparisons among multiple groups were performed by using one-way analysis of variance with post hoc Bonferroni correction. Bivariate correlations were calculated by using the Pearson or Spearman method as appropriate. HF event times were measured from the date of the CMR study. To identify independent predictors of HF and determine the associations of CMR findings and variables with HF, we considered all the variables significantly associated (p < 0.05) with HF in the univariate analyses and sought the best overall multivariate Cox regression model by using a forward stepwise model. Hazard ratios (HRs) with 95% confidence intervals are presented. HF event curves were determined by using the Kaplan-Meier method, and comparisons of event rate and risk stratification between the high- and low-ECV groups were performed by using log-rank tests. Receiver operating characteristic (ROC) curves were constructed to compare the diagnostic accuracy of T1 mapping variables for detecting myocardial fibrosis. Two-tailed p values < 0.05 were considered statistically significant.

A total of 66 ESRD patients were enrolled in the cohort, as summarized in Table 1. These patients were aged 56.44 ± 15.19 years, and 37.87% of the patients were male. Age and gender was matched between the ESRD patients and normal controls. Height, weight, blood pressure and heart rate did not differ between the normal and ESRD cohorts. The leading causes of ESRD were adult polycystic kidney disease (24.24%), primary glomerular nephropathy (53.03%), vasculitis (4.54%), genitourinary tuberculosis (3.03%) and urethral tumor (1.51%). Patients had experienced renal dysfunction for 0.25 to 19 years and had been treated with regular hemodialysis for 0.08 to 19 years. Extremely severe renal function manifesting as decreased eGFR (6.35 ± 2.77 ml/min/1.732 m²) was found. The levels of uremic toxins such as urea (909.00 [651.00–1052.25] mmol/l) and creatinine (824.07 ± 290.44 μmol/l) were very high, and 15 (22.72%) patients suffered from secondary hyperparathyroidism. The average parathyroid hormone (PTH) level was up-regulated (33.55 [21.62, 73.245] pmol/l) relative to the normal range. A total of 25 (37.88%) patients were on angiotensin-converting enzyme inhibitor (ACEI) or angiotensin-receptor blocker diuretic (ARB) therapy, whereas 38 (57.57%) patients were on oral calcium channel blocker (CCB) medication. A few patients were being treated with ɑ (15, 46.88%) or β (10, 15.16%) blockers.

The CMR findings are summarized in Tables 2 and 3. The follow-up time interval range from 11 to 30 months. The first HF cases were happened at the 11 month after CMR performed. Over a median follow-up duration of 18 months, 26 (39.39%) HF patients were documented. Regarding the functional parameters, LV dysfunction of some HF patients was identified at the time of CMR scanning, with LVEF lower in HF patients than in patients free from HF (45.77 ± 17.04% vs. 58.10 ± 6.99%, p < 0.05) and normal controls (45.77 ± 17.04% vs. 64.99 ± 4.43%, p < 0.05). LV mass measurements revealed the cardiac hypertrophy of ESRD patients with HF, with these patients having the highest LV mass among the three cohorts (p < 0.05). In analyzing myocardial deformation, the strain parameters GRS, GCS and GLS were found to be deteriorated in the group of ESRD patients with HF compared with both non-HF patients and normal controls (all p < 0.05). However, the strain variables did not differ between the non-HF patients and normal controls.

The tissue characteristics, including T1 mapping values and LGE, were as follows: Among the three groups, the HF group had the largest native T1 value (1360.10 ± 50.14 ms) and ECV (35.42 ± 4.42%), and non-HF patients had higher values than the normal controls (both p < 0.05) (Table 2, Fig. 1). LGE most frequently occurred in the HF group, and 19 (73.08%) of the patients with HF with a higher total LGE enhanced mass and extent also had diffused LGE. Among the subjects free from HF, 52.50% were affected by diffused LGE, with lower total LGE enhanced mass and extent (both p < 0.05). As shown in Tables 3, 40 ESRD patients were found to have diffused LGE, with significantly increased LGE mass and extent (p < 0.001). The ECV and native T1 value of the diffuse LGE patients were increased relative to those of patients without diffused LGE (ECV, 34.03 ± 3.74% vs. 32.07 ± 3.74%, p = 0.041; native T1, 1352.45 ± 48.60 ms vs. 1309.23 ± 59.08 ms, p < 0.001). Correspondingly, GRS, GCS and GLS were all reduced in those ESRD patients with diffuse LGE (p < 0.05). Although LGE, native T1 and ECV showed significant differences in all comparisons, post-contrast T1 did not significantly differ among the groups. In investigating the relationships of T1 mapping with LGE, the native T1 value showed a moderate correlation with ECV (r = 0.59, p < 0.001). No relationship was detected between ECV and either LGE parameter. We performed bivariate correlation analysis among the duration of CKD/dialysis and parameters in myocardial strain, T1mapping, and LGE. We found that the during of the dialysis is positively related with the LGE extent (r = 0.330, p = 0.007); and the duration of CKD show positive relationship with the native T1 value (r = 0.262, p = 0.034).

ECV was moderately correlated with GRS (r = − 0.501, p = 0.009), GCS (r = 0.553, p = 0.005) and GLS (r = 0.507, p = 0.008) (Fig. 2). Native T1, LGE mass and extent showed no significant relationship with any of the above strain variables. To screen the influencing factors of MF, all physical and clinical data and biochemical and uremic toxin data were analyzed using Pearson’s or Spearman’s correlation tests to determine their correlations with ECV and native T1 value. However, the only association was between PTH secreted by the parathyroid and ECV (r = 0.406, p = 0.001).

All patients were divided into two groups according to ECV value; i.e., above (high-ECV group) or below (low-ECV group) the median value of 32.86%. Kaplan-Meier analysis revealed a significantly different HF event occurrence curve between the high- and low-ECV groups. The high-ECV group had a shorter median survival time than the low-ECV group (18 months vs. 20 months; log-rank p = 0.046) (Fig. 3a). Seven (21.21%) ESRD patients in the low-ECV group developed HF during follow-up (occurrence time, 16–21 months); in the high-ECV group, 19 (57.57%) ESRD patients developed HF (occurrence time, 11–29 months). Risk stratification analysis by hazard function curve indicated that ESRD patients with high ECV were more likely to develop HF (Fig. 3b). Additionally, ECV (r = 0.491, p < 0.0001) was more strongly associated with HF events than were other variables, such as LGE and strain parameters, in the univariate analyses (Table 4). Cox proportional hazard regression model analysis revealed that ECV (hazard ratio [HR] = 1.160, 95% confidence interval: 1.022 to 1.318, p = 0.022) was the only independent predictor of HF events among ESRD patients (Table 4), and ECV had higher diagnostic accuracy for detecting severe MF (area under the curve [AUC] = 0.936; 95% confidence interval: 0.864 to 0.976, criterion> 28.89%) than did native T1 or post T1 value (all p < 0.05) (Fig. 4).

In this research, we found ECV as identified on T1 mapping may be a risk factor useful for predicting HF. However, the present study has a few limitations. The immunohistochemical validations of Native T1, Post T1 and ECV were lacking in this research. However, previous studies in humans have demonstrated a consistent relationship between a variety of T1-based indices (Native T1, Post T1 and ECV), and T1 mapping assumed to be a precise modality for measuring MF [13, 41‐43]. In this research, no significant correlations were found between LGE and the T1 mapping parameters, and this lack of correlation might reflect the facts that MF is mostly diffuse in MF patients and that LGE has been shown to be insufficient for calculating diffuse MF because of its dependence on the normal reference area [26]. Although we found HF had been occurred in our study cohort, the follow-up time of 11–30 months which may be a little be short. A longer follow-up time is needed to obtain more cardiac alterations on MRI and clinical evolution of CKD. Unfortunately, all ESRD patients in this study were stage five and undergoing dialysis; thus, the MF state of early-stage CKD patients was not considered in this work. Future work is warranted to investigate whether MF in different stages of CKD affects clinical outcomes or HF.