The trial is a multicentre, prospective, randomised, un-blinded, controlled trial. Patients will be randomised 1:1 to use a dialysate temperature of 37°C for 12months or an individualised cooled dialysate. The latter will be set at 0.5°C less than the patient’s own temperature, determined from the mean of 6 prior treatment sessions with a tympanic thermometer, up to a maximum of 36°C. Protocol adherence will be regularly checked. Inclusion criteria are incident adult HD patients within 180days of commencing in-centre treatment 3 times per week with capacity to consent for the trial and without contra-indications for magnetic resonance imaging. Exclusion criteria include not meeting inclusion criteria, inability to tolerate magnetic resonance imaging and New York Heart Association Grade IV heart failure. During the study period, resting cardiac and cerebral magnetic resonance imaging will be performed at baseline and 12months on an inter-dialytic day. Cardiovascular performance during HD will also be assessed by continuous cardiac output monitors, intra-dialytic echocardiography and biomarkers at baseline and 12months. The primary outcome measure is a 5% between-group difference in left ventricular ejection fraction measured by cardiac magnetic resonance imaging at 12months compared to baseline. Analysis will be by intention-to-treat. Secondary outcome measures will include changes in cerebral microstructure and changes in cardiovascular performance during HD. A total of 73 patients have been recruited into the trial from four UK centres. The trial is funded by a Research for Patient Benefit Grant from the National Institute of Healthcare Research. AO is funded by a British Heart Foundation Clinical Research Training Fellowship Grant. The funders had no role in the design of the study.

This investigator-initiated study has been designed to provide evidence to help nephrologists determine the optimal dialysate temperature for preserving cardiac and cerebral function in HD patients.

In the setting of coronary artery disease without kidney disease, transient myocardial ischaemia may lead to left ventricular dysfunction that can persist for several hours after the return of normal perfusion, but is eventually followed by functional recovery. This reversible post-ischaemic dysfunction is known as myocardial stunning. Repetitive episodes of stunning are cumulative eventually leading to hibernation then irreversible contractile dysfunction [1]. The ability of HD to induce myocardial ischaemia had been long suspected with reports of silent, reversible ischaemic ECG changes during HD older than twenty years. Other studies showed elevated cardiac troponins were common and up to 30% of HD treatments were complicated by significant intra-dialytic hypotension [2]. These data suggested HD might induce myocardial ischaemia but were unable to prove this or elucidate the pathophysiology. Subsequent proof-of-concept studies by our group and others in subjects with normal coronary arteries have demonstrated that HD is characterised by a 30% reduction in regional myocardial blood flow by H₂¹⁵O positron emission tomography [3]. The decreased myocardial blood flow localises to the same segments as regional wall motion abnormalities (RWMAs) using simultaneous 2D-echocardiography at peak-stress, 15 minutes before the end of a HD treatment compared to pre-HD [3]. This supported use of these stress-RWMAs as a practical surrogate for myocardial ischaemia. These changes in the absence of coronary artery disease can be explained by decreased coronary flow reserve which is well described in dialysis populations [4] and is in turn promoted by a high prevalence of autonomic dysfunction, endothelial dysfunction, left ventricular hypertrophy and arterial stiffness. These factors may underlie the reduced sensitivity and specificity of ischaemia stress testing in HD patients [5]. The HD procedure involves a rise in core temperature, extracorporeal circuit, fluid removal and exposure to a dialysis membrane which combine to form a haemodynamic stress provoking demand ischaemia [6]. Our natural history study of 70 prevalent HD patients showed that compared to pre-HD echocardiography, new RWMAs developed at peak-stress in 45 participants and this associated with 30% mortality and 13% reduction in mean left ventricular ejection fraction by 1year [7,8]. RWMAs persisted at follow-up in the 36/45 survivors of the subgroup. In the cohort who had no abnormal segments at the start of the study, new segmental RWMAs had also developed. Multivariable analysis showed the significant determinants of these stress-RWMAs were age, ultrafiltration volume, intra-dialytic hypotension and serum troponin-T concentration and this displaced all other variables including diabetes status and history of ischaemic heart disease [7]. In short, the RWMAs characterise those who suffer ischaemia due to the haemodynamic stress of HD. Regardless of the balance of micro- and macrovascular disease in the pathophysiology, reducing these RWMAs represents a valid target and may lead to detectable amelioration of global ventricular function.

It is increasingly recognised that HD patients exhibit higher levels of functional and cognitive deficits. Dementia is highly prevalent in the HD population with a reported 7-fold relative risk of developing multi-infarct dementia within a year of commencing dialysis compared to age-matched controls; delirium and depression were also found to be common in this population when followed for a period of 4years [17,18]. Furthermore, even HD patients with normal dementia scores develop neurocognitive dysfunction affecting retention of complex verbal information, memory, executive function, language and mental processing speed [19,20]. Few studies have linked cognitive impairment with findings on brain imaging. Cerebral atrophy is a common finding in HD patients [18,21]. However, the predominant morphological abnormalities reported are white matter hyperintensities termed leukoaraiosis. These areas of high signal on T2-weighted MRI represent white matter ischemia characterised by neuronal loss and demyelination. Although leukoaraiosis develops with aging as a reflection of the burden of cardiovascular disease, many studies report a greater volume of leukoaraiosis in dialysis patients compared with age-matched controls [22,23]. A recent meta-analysis showed leukoaraiosis was associated with a higher risk of stroke, dementia and death in the general population [24]. Adequate perfusion of brain white matter depends on blood pressure and the brain’s autoregulatory capacity [25]. HD can induce haemodynamic and circulatory stress and HD patients have impaired autonomic function and autoregulatory capacity rendering them vulnerable to those haemodynamic changes [26]. Hence the HD process might play an important role in the pathogenesis of brain injury in HD patients.

This is an un-blinded study. Therefore, after randomisation, the investigators, HD unit staff and patient will be aware of group allocation. This is necessary to ensure adherence to the study protocol. There will be blinding of treatment allocation for analysis of cardiac images for the primary outcome.

Patients in each trial arm will undergo standard management, as per local HD unit protocols with the exception of dialysate temperature.

Contrast-free cardiac and cerebral MRI will be performed at baseline and 12months when the patient has been established on HD with a stable dry weight unchanged for at least 4weeks. Studies will occur on non-dialysis days at two centres using a 1.5T MRI scanner (GE Signa HDxt 1.5T, GE Healthcare, Milwaukee,US). Acquisition will include serial contiguous short-axis cines piloted from vertical or horizontal long axis of the left ventricle using ECG-gated steady-state in free precession sequences in accordance with validated methodologies [27]. Analysis will be performed with the anonymised images off-line by planimetry using commercially available software for assessment of ventricular volumes and mass. Brain MRI will examine the integrity and microstructure of white matter using diffusion tensor imaging. The anonymised images will be analysed off-line for diffusivity and anisotropy corrected for brain volume.

Cognitive assessment will be performed on the same day as the MRI scans at baseline and 12months. The Montreal Cognitive Assessment is a validated tool for the detection of mild cognitive impairment with higher sensitivity and specificity than the Folstein Mini-Mental State Examination for the detection of mild cognitive impairment [28]. Trail Making Tests A and B will be used to examine attention and mental flexibility.

Detailed characterisation of cardiovascular performance will be assessed during HD at baseline and 12months. This occurs during the patient’s usual HD session in their own centre using the same portable data collecting instruments across all centres. 2-dimensional echocardiography will be performed using commercially available equipment (1.5-4Mhz probe, Vivid-I, GE Healthcare) by one of two trained technicians in the left lateral position. Sector width, frequency and depth will be adjusted to produce optimal border definition and a frame-rate of 50–80 per second. Standard apical 2 and 4-chamber image loops will be recorded pre-dialysis (rest) and 15 minutes before the end of dialysis (stress). Image loops will be anonymised and analysed off-line in random order for regional deformation using dedicated software (EchoPac, GE Vingmed). Continuous non-invasive cardiac output monitoring will be recorded using a thoracic bioreactance monitor (NICOM™, Cheetah Medical). This uses phase changes in alternating electrical current to derive cardiac output. This endows a greater precision and signal-to-noise ratio than impedance cardiography. The monitor is validated in both healthy and co-morbid adults [29]. The Finometer™ (TNO Instruments) utilises a finger-cuff to detect beat-to-beat changes in digital arterial pressure and calibrates this against oscillometric brachial artery pressure. The Finometer™ is validated in critical care and HD settings [30].

Arterial stiffness measured by pulse wave velocity is independently associated with increased risk of cardiovascular disease in HD patients [31]. Carotid-Femoral pulsewave velocity will be measured by a volume displacement technique [32]. Pre-HD body composition and hydration status will be estimated by segmental multi-frequency bioimpedance [33].

Change in resting LVEF by cardiac magnetic resonance (CMR) at 12months compared to baseline between the intervention and control group.

We selected LVEF measured by CMR as the primary outcome because previous observational study showed mean reduction in LVEF by echocardiography of 13% over 12months in patients who displayed RWMAs at baseline. RWMAs were abrogated in the short-term by a dialysate cooling strategy in a separate cohort [10,12]. We also selected LVEF as an objective measure which is predictive of mortality and hospitalisations [34]. A change in LVEF greater than 5% has been shown to be an important cut-off approaching the minimum clinically important difference [35,36]. CMR is the gold-standard for measurement of left ventricular mass and volumes, allowing determination of the primary outcome to a much higher precision compared to alternative methods such as echocardiography. This markedly reduces the sample size required to detect between-group differences in an intervention trial, permitting lower costs and faster reporting of final results [37,38]. Sample estimation is from a recent reference range study which used the same CMR acquisition and analytic techniques planned for this study [27] and were performed by a medical statistician using commercially available software (Nquery Advisor v6).

The primary aim in this study is to detect a 5% difference in LVEF as measured by CMR between the control and intervention group after 12months. Using a two sample t test and assuming a mean LVEF of 67% in the control arm and equal Standard Deviation in each group of 6% [27], a study of 64 participants (32 per group) would permit the detection of a 5% difference in LVEF from baseline to 12months between the intervention and control groups, with 90% power at 5%, 2-sided significance level. Allowing for study attrition and death of 10%, we aim to recruit 72 participants to achieve an evaluable sample of 64 participants.

All the continuous variables will be tested for normality using their histograms and normality tests. The primary endpoint will be compared between the two groups at specific time points using the Mann U Whitney or Independent T-test. Repeated measures analysis will be used for testing the difference on primary endpoint between tests across the whole period. Poisson regression analysis will be used for comparing the frequency of intra-dialytic RWMAs between the two groups. Haemodynamic variables and other secondary endpoints will be analysed using repeated measures and Bonferroni’s test for multiple comparisons. Paired tests, Wilcoxon Signed test or Paired T-test, will be used for comparing the primary and secondary endpoints between two time points within each group. An alpha error at 0.05 will be judged as significant. Analysis will be performed using SPSS for Windows, v16 or Stata v10. Statistical analysis will be based on an intention-to-treat principle with all data analysed according to the patient’s original study group allocation regardless of death, crossover or early withdrawal.

The number and proportion of participants who report treatment-emergent adverse events will be summarized for each treatment group. Treatment emergent events include events that start on or after allocation to the study assigned dialysate temperature.