Native T1 mapping: inter-study, inter-observer and inter-center reproducibility in hemodialysis patients

There is an increased risk of cardiovascular (CV) mortality in chronic kidney disease (CKD), and end stage renal disease (ESRD) populations [1]. This increased risk can be attributed to the stereotyped changes that lead to the development of uremic cardiomyopathy and include, left ventricular (LV) hypertrophy, LV dilatation and myocardial fibrosis within the extracellular matrix [2]. To date, LV mass is the most commonly used surrogate end-point of mortality in clinical trials [3], as observational studies of HD patients have shown LV mass is good predictor of CV outcomes [4]. However, a recent systematic review and meta-analysis in patients of all stages of CKD suggested that there is no clear association between intervention-induced LV mass reduction and mortality [5]. Novel imaging biomarkers that can robustly and reliably measure pathological CV changes that link strongly to outcomes are required. Post-mortem studies of patients with CKD and ESRD on hemodiaysis (HD) demonstrate that uremia is a highly significant, independent determinant of extent of myocardial fibrosis [6]. Furthermore endomyocardial biopsy studies have shown that extent of myocardial fibrosis is the only independent predictor of death (mean follow-up period 3.1 years) for these patents [7]. As degree of myocardial fibrosis is the strongest predictor of increased CV mortality, defining a reliable measure of myocardial fibrosis in HD patients is a priority.

Cardiovascular Magnetic Resonance (CMR) with Late Gadolinium Enhancement (LGE) is an imaging biomarker used to assess myocardial fibrosis in many populations. Whilst gadolinium based contrast agents have previously been used to assess cardiac disease in HD patients [8] this is no longer possible due to the rare, but serious complication of nephrogenic systemic fibrosis [9]. Moreover, whilst LGE is a sensitive and reproducible way of assessing focal myocardial fibrosis, there are limitations in using gadolinium to assess diffuse myocardial fibrosis due to the reliance of the technique on demonstrating a difference between signal intensity of normal and fibrotic myocardial tissue [10]. Native T1 mapping is a novel, non-contrast CMR technique that correlates well with biopsy measured myocardial fibrosis in aortic stenosis [11, 12] and can differentiate patients with hypertrophic cardiomyopathy from hypertensive cardiac disease [13]. The inter-study repeatability and inter-observer variability of native T1 mapping has been shown to be very good in patients with aortic stenosis [14], patients with LV hypertrophy or dilated cardiomyopathy [15] and patients with Anderson-Fabry disease [16]. Myocardial native T1 times have been shown to be significantly higher in HD patients compared to control subjects [17, 18] and to associate with circulating markers of cardiac disease [17] and measures of myocardial systolic strain [18], but the reproducibility of native T1 mapping has not been assessed in HD patients who are prone to shifts in extracellular volume. Native T1 times are prolonged with increasing water content of tissue and the presence of intermittent myocardial edema from alterations in fluid status may in theory affect native T1 time confounding results and reducing native T1 time reproducibility. Concerns remain about the use of this technique to assess myocardial fibrosis in patients with ESRD on HD, who are subject to significant changes in fluid status and who may potentially have intermittent myocardial oedema [19].

In this study we aim to assess the reproducibility and reliability of native T1 mapping in patients with ESRD on HD by assessing: i) the inter-study, inter-observer and intra-observer variability of native T1 times in HD patients at 3 Tesla (3 T); ii) the associations between changes in markers of fluid status and changes in native T1 times to assess the effect of fluid status on native T1 times; iii) the inter-centre reproducibility of native T1 values between two UK cardio-renal imaging centres, comparing analysis techniques of native T1 maps and native T1 values with ‘phantom’ analysis [20].

Patients from centre 1 were recruited as a part of the CYCYLE-HD study (ISRCTN 11299707) [21]. The study was given ethical approval by the NHS Research Ethics Committee East Midlands (Northampton; REC ref: 14/EM/1190). Patients from centre 2 were recruited as part of the observational cardiac uraemic fibrosis detection in dialysis patients study (CUDDLE study ISCRTN99591655). The study was approved by the West of Scotland Ethics committee (WoS 13/WS/0301). All participants gave written and informed consent.

The demographic data for all patients from both centres are shown in Table 1. There were no statistically significant differences between the demographics of the groups. There was a trend towards longer dialysis vintage in participants from centre 1 compared to those in centre 2, but these difference did not reach statistical significance. There were no differences between the basal or the mid-ventricular native T1 times for HD patients between centre 1 and centre 2 respectively (basal T1 1280.5 ms ± 36.6 vs 1276.3 ms ± 32.3, P = 0.62 and mid T1 1282.6 ms ± 38.1 vs 1283.4 ± 39.3, P = 0.42). All scans were analysable and included in the analysis.

In this study we have shown, for the first time in patients with ESRD on HD, that the reproducibility of native T1 values are excellent. The reproducibility values for T1 mapping using the MOLLI sequence compare favourably with those previously reported. Previous CoVs for inter-study reproducibility have been between 1.5 and 8.4%, [14‐16, 29] compared with 0.7% in this report. As previously described, native T1 is related to water content of the relevant tissue [30] and the finding in this study that ΔT1 between test-re-test scans was unrelated to changes in fluid status suggests that changes in fluid status at the levels we have seen do not directly affect native T1 values in HD patients, although we accept that this study is only in 10 patients. Whilst the absolute changes in inter-study fluid status that we report are not large, they are the types of fluctuations in fluid you would expect to see in HD patients scanned on a non-dialysis day on week apart and therefore clinically relevant. This is a crucial observation in defining the reliability of T1 time as a potential measure of myocardial fibrosis in this patient group.

We have demonstrated that standardization of native T1 mapping analysis between centres is eminently possible. The agreement between the analysis techniques of centres is encouraging. Both techniques define segmental native T1 values for basal- and mid-ventricular slices as per the American Heart Association model (Fig. 1), with the average of these segments taken to be the mean native T1 for each slice. When areas of artefact are identified, affected segments can be excluded with the mean native T1 calculated from remaining unaffected segments. This way of assessing native T1 is likely to be more reproducible than alternative methods, where single regions of interest are drawn in the inter-ventricular septum [31]. Such methods are prone to sampling differences in native T1 times depending on where the region of interest is drawn. We have previously demonstrated that septal native T1 is significantly higher than non-septal native T1 in HD patients [18]. Moreover, a recent paper by Rauhalammi et al showed that native T1 values are significantly higher in the interventricular septum of health volunteer patients than non-septal myocardium [27] so analysis techniques that measure only the interventricular septal native T1 cannot be thought of as equivalent to mean circumferential ventricular native T1 values. The phantom analysis conducted at centre 1 and centre 2 confirmed the scanners and scan sequences at the different centres gave very similar results, with the differences in absolute values being clinically irrelevant. The mean basal- and mid-ventricular native T1 values between the two centres were virtually identical. Given that subjects from both centres were well matched in terms of demography and medical co-morbidity, it appears that the native T1 values generated at both centres are commensurate with each other. It should be noted that both centres used Siemens 3 T CMR scanners and whilst they were not the same model scanner, inter-vendor comparisons of myocardial native T1 values in HD patients are required. Ideally assessment of the reproducibility of native T1 times at different field strengths and between vendors and centres should also be carried out.

Although native T1 time has been shown to correlate well with histological levels of myocardial fibrosis in diseases of pressure overload [12] we still do not have histological confirmation that this is the case for HD patients; indeed low grade inflammation may increase native T1 [32]. Low grade inflammation is known to be a feature of CKD and for patients with ESRD on HD [33], so histological studies are required to confirm increased T1 times are due to myocardial fibrosis and not low grade inflammation or an additional disease process such as amyloidosis [34]. Whilst our results suggest inter-compartmental fluid shifts do not affect native T1 values, further, confirmatory studies should be considered, including studies that look at the acute effects of HD and ultrafiltration on native T1 values.

The inter-study reproducibility of traditional CMR measures of LV structure and function, including LV mass, LVEDV and LVEF were also excellent as has been shown previously for patients with ESRD [35]. The reproducibility of LV mass is particularly important as LV mass is known to be overestimated when imaged with echocardiography [36]. These findings confirm previous studies in HD patients showing conservation of measurement of LV mass with CMR in patients with differing LVEDV’s and fluid and loading status [35].

In 2011, Sado et al outlined a framework for development of imaging techniques to measure myocardial fibrosis, to ensure the robustness of the techniques [37]. They suggested that to be considered a reliable imaging biomarker of myocardial fibrosis, techniques should: detect changes in established disease states compared with controls; correlate with cardiac markers fibrosis (e.g., diastolic function, LVH); correlate with blood biomarkers of cardiac fibrosis; be able to track changes over time; be standardized in the way they are carried out (inter-vendor/inter-centre); be proven to compare closely with histological specimens from human subjects; and changes in the imaging biomarker should track changes in the disease after treatment. Work by our respective groups has already shown that native T1 values are higher than healthy and co-morbidity matched controls [17, 18]. Rutherford et al demonstrated that global native T1 values in HD correlate with LV mass indices (r = 0.452, P = 0.008), and that septal T1 values correlate with pre-dialysis high-sensitive troponin T (r = 0.397, P = 0.027) [17]. In addition to this, Graham-Brown et al showed that native T1 values correlate with measures of global circumferential and longitudinal strain (r = 0.41, P = 0.002, r = 0.55, P < 0.001) and septal native T1 correlated with septal systolic strain (r = 0.46, P < 0.001) [18].

Following the framework outlined by Sado et al, histological confirmation of the relationship between myocardial fibrosis and native T1 mapping now appears to be the final step required before longitudinal and interventional studies can be undertaken to define whether native T1 mapping is able to: track progression of myocardial fibrosis over time; assess how baseline or changes in native T1 values relate to outcomes; whether interventions that attempt to modify/improve fibrosis are measurable by native T1 mapping; and whether this modification improves outcomes for patients. The excellent reproducibility of native T1 mapping makes it a biomarker with great potential for the measurement of myocardial fibrosis in HD patients. Based on the results we have presented, only small numbers of patients would be required to adequately power interventional clinical trials that seek to reduce myocardial fibrosis. Whilst historically LV mass has been the surrogate end-point for clinical trial work that sought to improve CV outcomes, a recent systematic review and meta-analysis has called into question whether interventions that improve LV mass actually improve outcomes for patients with CKD and ESRD [5]. It should be noted that the neutral result of this meta-analysis might be due to inadequate power. Post-mortem histo-pathological studies have shown us that extent of myocardial fibrosis is the best predictor of outcome in patients with CKD and ESRD [6] and there is evidence in animal models, humans and even HD patients that regression of LVH is accompanied by reductions in measures of myocardial fibrosis [38‐41]. Given that multiple studies have suggested that reducing LV mass improves outcomes for patients with ESRD [42], it could be that previous trials that have shown reductions in LV mass have actually just reflected a crude measure of a reduction in myocardial fibrosis content and this has led to the conflicting results about the importance of LV mass reduction and CV risk that has recently been reported. Given the central association between fibrosis, LVH, morbidity and mortality in HD patients, being able to reliably measure diffuse myocardial fibrosis is crucial to understanding its potential as a research and clinical end point.

Native T1 mapping is an extremely reproducible technique in HD patients. Native T1 times seem to be unaffected by the changes in fluid status to which HD patients are prone. Further work is required to determine whether native myocardial T1 is related to prognosis in ESRD and whether interventions that reduce T1 are associated with improved outcomes.