Effect of ultrafiltration during hemodialysis on hepatic and total-body water: an observational study

This observational study was approved by the Health Sciences Research Ethics Board at the University of Western Ontario (HSREB106868) and conducted according to the GCP/ICH guidelines and the Declaration of Helsinki.

Patients from the prevalent chronic HD population cared for in the London Regional Renal Program (Ontario, Canada) were recruited to a cross-sectional study and written consent was obtained from all patients. For inclusion, patients had to have been receiving dialysis for at least 3 months. Exclusion criteria included the presence of liver disease or known risk-factors for liver disease (viral hepatitis, history of significant alcohol intake), contraindications to MR scanning (for the sub-study) and history of lower limb amputation. In addition, if the patients had any signs of inflammation, infection, maglignancy, and/or hemodynamic instability, they were excluded.

Baseline information was obtained for each subject, including age, gender, height, blood pressure, underlying renal disease, dialysis information (frequency, blood flow rate, dialysate flow rate, ultrafiltration volume, duration of dialysis session, and single-pool Kt/V urea measurements), and most recent residual renal function measurement.

Prior to and after their routine HD treatments, all subjects underwent clinical assessment of fluid status 30–60 min prior to their HD treatment. Multi-frequency bioimpedance analysis of total body water and magnetic resonance imaging (MRI) scans in supine position were also performed during this period. In addition, blood samples were taken to measure biomarkers such as hematocrit, creatinine, urea, bilirubin, liver enzymes, and sodium levels.

Bioimpedance analyzer were performed using the standard tetrapolar technique, with electrodes placed at both wrists and both ankles after subjects lay supine for at least 10 min (50 frequencies (5 to 1000 kHz), Body Composition Monitor®, Fresenius Medical Care Deutschland GmbH) [13].

Organ water content was quantified pre- and post- dialysis using 3D chemical shift encoded MRI (CSE-MRI) pulse sequence implemented on a 3 Tesla MRI scanner (MR 750, GE Healthcare, Waukesha, WI) [14]. This technique has been validated in clinical studies and employed in large randomized-controlled trials [15‐17]. Imaging parameters were selected to cover the abdomen with proton density contrast: Field Of View (FOV) = 48x34x58 cm [3], matrix = 128x90x72, repetition time (TR) = 4.7 ms, flip angle = 3 deg. The acquisition was accelerated using ARC (autocalibration reconstruction for Cartesian imaging) in the phase and slice directions with a 32 channel Torso array for an overall acceleration factor of R = 3.9, permitting a complete volume to be acquired in a sub 20s breath hold. The images at the 6 echoes (echo time (TE) (ms) = [0.37 0.76 1.14 1.152 1.90 2.28]) were receive sensitivity corrected and processed to give water only and fat only images. Organ water content was then evaluated by drawing regions of interest (ROI) in each organ and normalizing by comparing with a water phantom that was included in the field of view [18]. For the hepatic ROI, we select the right lower lobe with 1 cm margin to exclude splanchnic vascular bed (Fig. 1). We also avoided the portal and hepatic vasculature. For the peripheral muscle water content, unaccelerated, high resolution, 3D CSE-MRI images (FOV = 48x38x48, matrix = 192x160x80) were also acquired of the thigh with otherwise similar parameters and processing, to quantify muscle water content. Finally, the infectior vena vaca (IVC) diameter was measured in an axial T2 weighted fast spin echo slice (FOV = 48 cm, thickness = 8 mm, matrix = 384 × 160, TE = 80 ms) approximately 5 cm below the diaphragm. Diameters were measured by two readers and averaged.

Measures of liver stiffness were obtained by a single, trained operator (CG) using transient elastography (Fibroscan® GE Healthcare) [19]. Measurements were obtained by the conventional clinical method, with the subject supine and 10 valid measurements obtained with the median value recorded. Unreliable results were deemed as those with an IQR of less than 33% of the median or in those in whom fewer than 10 valid results could be obtained.

All these measurements and assessments were repeated 30–60 min after their HD treatment.

Eighteen patients consented to be part of the study from February 2015 to April 2016. Two subjects were excluded due to the presence of exclusion criteria. One subject was excluded from the analysis due to the diagnosis of metastatic cancer. Poor data quality meant that liver MRI data was not available for one subject. Fourteen subjects for whom liver MRI data was available formed the study population.

Characteristics of the study population are shown in Table 1. The majority of patients were male and had either diabetic or hypertensive nephropathy as the underlying etiology of their renal disease. Other etiologies included unspecified nephrotic syndrome (1) obstructive uropathy (1), lupus nephritis (1), renal dysplasia associaved with Prune-Belly syndrome (1) and IgA nephropathy (1). The majority of patients (86%) were dialyzed via central catheter access with the remainder via peripheral fistula access. Echocardiographic of all these patients showed normal ejection fraction (EF) > 55–60% except one patient who had EF 35–40%.

All subjects underwent their prescribed length of dialysis session as part of the study. Ultrafiltration volumes ranged from 0 to 4090 mL (mean 8.13 ± 4.4 mL/kg/hr). Two patients experienced symptomatic intra-dialytic hypotension (IDH), with dizziness or severe cramps. Three subjects had IDH defined as a nadir systolic blood pressure of less than 100 mmHg, whilst six subjects experienced a fall in systolic blood pressure of at least 40 mmHg during dialysis.

Mean pre-dialysis hepatic water content was 0.63 ± 0.19 g/mL (range 0.34 g/mL to 1.05 g/mL). There was no association between hepatic water and clinically or bioimpedance-derived measures of hypervolemia. There was also no significant relationship between liver stiffness and hepatic water content (r = − 0.02, p = 0.93; Table 2). Higher hepatic water content was, however, associated with a smaller IVC diameter (r = − 0.57, p = 0.03) (Fig. 2) and with a higher pre-dialysis hematocrit (r = 0.61, p = 0.048).

The mean liver stiffness values were 5.0 ± 2.07 kPa and 6.3 ± 2.31 kPa pre and post-hemodialysis (p = 0.20). There was no statistically significant change in the mean liver water content after HD (from 0.63 ± 0.2 g/mL to 0.70 ± 0.2 g/mL, p = 0.34) (Fig. 3). The liver water content increased in eight subjects but fell in the remaining six subjects. The magnitude of the change in liver water content pre- and post- hemodialysis was not associated with ultrafiltration volume (r = − 0.04, p = 0.88) (Fig. 4) or the change in liver stiffness values (r = 0.06, p = 0.84). There were no clear clinical factors which were significantly associated with the direction of change in liver water content, including age, co-morbidity or fluid status before dialysis. However, liver water content rose in all patients with intra-dialytic hypotension, which was defined as > 40 mmHg drop in blood pressure with/without symptoms and/or a nadir blood pressure < 100mHg based on previous studies [20]. Furthermore, there was a greater increase in liver water content in those whose systolic blood pressure fell by more than 40 mmHg during dialysis (0.21 ± 0.2 g/mL vs. -0.04 ± 0.2 g/mL, p = 0.05) (Fig. 3).

IVC diameter did not change significantly after dialysis. Nevertheless, the correlation between liver water content and IVC diameter persisted (r = − 0.60, p = 0.03). Pre-dialysis IVC diameter was lower in those who suffered IDH, with a nadir blood pressure < 100mHg, (2.02 ± 0.17 cm vs 1.69 ± 0.14 cm, p = 0.009).

Pre-dialysis, total body water was 39.0 ± 9.4 L with intra-dialytic weight gains of between − 3 and 5.9 kg with median weight gain of 1.65 kg. One subject was 3 kg under their target weight prior to dialysis and bioimpedance measured that he was 1.2 L below his ideal fluid status. Full bioimpedance analysis was unobtainable in the one subject who demonstrated clinically significant fluid overload, being almost 6 kg over the target weight with profound peripheral edema. Pre-dialysis MRI-measured limb (thigh) water content was associated with a higher total body water (r = 0.59, p = 0.06), higher intracellular water volume (r = 0.64, p = 0.02) and a higher diastolic blood pressure (r = 0.70, p = 0.02). After dialysis, total body water fell by a mean of 1.59 L to 37.4 ± 8.4 L and thigh water content decreased from 0.85 ± 0.21 g/mL to 0.83 ± 0.18 g/mL, although this was not statistically significant (p = 0.55).

The liver enzymes were measured pre- and post-dialysis. There were significant changes in aspartate aminotransferase, alkaline phosphatase, hematocrit and total bilirubin values. The alanine aminotransferase and gamma-glutamyl transferase values were not significantly altered. After using the correction factor based on the published data by Schneditz et al., however, we still found statistically significant changes in some of the liver enzymes, including aspartate transaminase, alkaline phosphatase and gamma-glutamyl transpeptidase [21]. Please see Table 3. However, these values did not correlate with the changes in liver water.