A reliable method to assess the water permeability of a dialysis system: the global ultrafiltration coefficient.

Background: Recent randomized controlled trials suggest that sufficiently high convection post-dilutional haemodiafiltration (HC-HDF) improves survival in dialysis patients, consequently this technique is increasingly being adopted. However, when performing HC-HDF, rigorous control systems of the ultrafiltration setting are required. Assessing the global ultrafiltration coefficient of the dialysis system [GKD-UF; defined as ultrafiltration rate (QUF)/transmembrane pressure] or water permeability may be adapted to the present dialysis settings and be of value in clinics.
Methods: GKD-UF was determined and its reproducibility, variability and influencing factors were specifically assessed in 15 stable patients routinely treated by high-flux haemodialysis or HC-HDF in a single unit.
Results: GKD-UF invariably followed a parabolic function with increasing QUF in dialysis and both pre- and post-dilution HC-HDF (R2 constantly >0.96). The vertex of the parabola, GKD-UF-max and related QUF were very reproducible per patient (coefficient of variation 3.9 ± 0.6 and 3.3 ± 0.3%, respectively) and they greatly varied across patients (31–42 mL/h−1/mmHg and 82–100 mL/min, respectively). GKD-UF-max and its associated QUF decreased during dialysis treatment (P < 0.01). The GKD-UF-max decrease was related to weight loss (R2 = 0.66; P = 0.0015). Conclusions: GKD-UF is a reliable and accurate method to assess the water permeability of a system in vivo. It varies according to dialysis setting and patient-related factors. It is an objective parameter evaluating the forces driving convection and identifies any diversion of the system during the treatment procedure. It is applicable to low- or high-flux dialysis as well as pre- or post-dilution HDF. Thus, it may be used to describe the characteristics of a dialysis system, is suitable for clinical use and may be of help for personalized prescription.

INTRODUCTION

Two randomized controlled trials (RCTs) testing the supposed benefits of haemodiafiltration (HDF) on survival observed such beneficial effects only in a post hoc analysis when convection volumes were high (>17 or 22 L following the studies) [1, 2]. In a third RCT by Maduell et al. [3], applying high convection as the treatment of choice was associated with observed a significant improvement in survival over classical high-flux dialysis. These reports and subsequent confirmatory work have definitely influenced the opinion of the renal community, and the belief is growing that post-dilutional online HDF (OL-HDF) with high convection volumes(HC-HDF) is the best treatment, at the present time, to improve patient's survival prospects [4, 5].

High convection volumes can be obtained only with high-flux/highly permeable dialyzers and require increased transmembrane pressure (TMP). During the treatment procedure, particularly when high convection is requested, fouling of the membrane may occur, altering the efficacy of the system [6] and provoking a sustained increase of the TMP necessary to obtain the requested volumes. This results in alarms and system instability. Some attempts have been made to control this situation, and several systems automatically decrease the ultrafiltration flow when TMP is considered too high [6-12]. Dialysis stability is then obtained at the price of decreasing the total convection volume below that initially prescribed, without informing the physician in charge of the treatment. Therefore, new approaches to increase the stability of the system and minimize its deviation in terms of water permeability are needed.

The recently described GKD-UF and GKD-UF-max [13] are promising parameters to support maintaining the system at its optimal filtration conditions. GKD-UF follows a parabolic function when increasing convection flow, defining a maximum level of GKD-UF, which is the vertex of the parabola. The QUF at which GKD-UF-max is observed is the highest ultrafiltration flow obtained per TMP unit in that system [14].

Since GKD-UF is an objective parameter of the water permeability of a dialysis system, it can be used to monitor convection flow and help in identifying any potential diversion of the system during the treatment procedure when high convective volumes are requested.

To deepen our understanding of this parameter, we assessed the reproducibility of GKD-UF, GKD-UF-max and its associated QUF and observed that these parameters are accurate and reproducible enough to be used in clinics.

MATERIALS AND METHODS

Patients

Fifteen stable dialysis patients treated in the dialysis centre of Néphrologie Dialyse St Guilhem in Sète (France) were included in the study (Table 1). They were dialyzed three times a week with online HDF Dialog+ (BBraun, Melsungen, Germany) and DBB 07 (NIKKISO, Tokyo, Japan) machines, using ultrapure double reverse osmosis water. Their vascular accesses were native arteriovenous fistulas (14 patients) and jugular catheters (1 patient). They had been on dialysis for >3 months and had no active disease during the study. They were able to understand the study and gave signed informed consent to participate in it. The study protocol was approved by the Comité de Protection des Personnes of Nîmes (2011.10.05 bis; registration number at the French Agency AFSSAPS 2011-A01092-39).

Polysulfone high-flux dialyzers (Xevonta Hi 18, Amembris and Diacap Hips 18, 1.8 m2, B Braun Avitum, Melsungen, Germany) were used. Total dialysate production flow was checked for every dialysis monitor and set at 600 mL/min in post-dilution. In pre-dilution; it was set at 500 mL/min plus the infusion flow (maximum 700 mL/min).

Convection flows assessed

GKD-UF was determined for all patients at increasing convection flows. To establish GKD-UF-max, the infusion flow rate was set at 0 mL/min and then modified stepwise by 10 mL/min from 50 to 100 or 110 mL/min. After ∼1-min stabilization, TMP was recorded and GKD-UF was calculated with QUF:

To prevent excess haemoconcentration in post-dilution, the last step was limited to a QUF value of 30% of the blood flow (Qb). The vertex of the parabolic function (GKD-UF/QUF) is GKD-UF-max. The corresponding total convection flow is GKD-UF-max associated QUF (and corresponds to the x value of the GKD-UF-max point). A specific software was developed to quickly determine GKD-UF-max and its associated QUF at bedside.

TMP was given by the dialysis machines with three (B BRAUN Dialog+) or four (NIKKISO DBB 07) pressure sensors. The pressure sensors to assess TMP were located at the inlet and outlet of the blood side, the outlet of the dialysate side of the dialyser and, in the case of four-point readings, the dialysate inlet (Figure 1B). In order to increase the accuracy of TMP measurements and standardize them, in vitro experiments were performed.

FIGURE 1

Methods and results of KUF and GKD-UF-max determinations. (A and B) Schematics of the setting to determine KUF. (A) The in vitro setting proposed by the FDA [15, 16] to assess KUF for high-permeability dialyzers. It is an isolated ultrafiltration system (with no dialysate), where KUF is determined by the slope of QUF/TMP points (see C). (B) The in vivo setting presently used in clinics, which is closed with an ultrafiltration controller (balancing chambers). (C) QUF increases linearly with TMP in isolated ultrafiltration, the in vivo KUF of the dialyzer is the slope of this line (KUF = 1.414 × 60 = 85 mL/h/mmHg) (open squares). The straight line is shifted to the right and parallel when introducing a dialysate flow (same slope and therefore same KUF according to Keshaviah's calculations). The shift may be explained, at least in part, by the hydrostatic pressure linked to dialysate flow and the oncotic pressure modifications linked to blood flow. When measurements of QUF higher than those proposed by Keshaviah's were performed, the QUF–TMP relationship no longer followed a straight line function. It bent and tended to plateau. (D) Plotting the values of GKD-UF (QUF/TMP)/QUF for these in vivo measurements described the parabolic distribution of GKD-UF. The vertex of the parabola is GKD-UF-max and the corresponding QUF is the highest QUF that can be obtained for the minimal TMP.

The in vitro studies consisted in putting the dialysate tubing in an open volume (a laboratory plastic beaker, at the same height as the dialyzer where pressure = 0) and reading the measurements of the monitor for TMP. A correction factor for each machine could then be obtained, which was the deviation of the TMP readout of the machine from zero during these calibration studies. Following these studies, we decided to incorporate our correction factor to correct the readouts given by the machines during GKD-UF determinations at bedside.

All pressures were measured outside the dialyzer, and the resultant given by the dialysis monitor was corrected as described to obtain the TMP value. Although the precise values of hydrostatic and oncotic pressures whithin the dialyzer are not determined, they are incorporated in the TMP readings. As a result, the parabolic function between QUF and GKD-UF holds true regardless of the oncotic pressure or haematocrit levels, which influence the absolute value of the vertex but not the shape of the curve.

Statistics

Statistical analyses were performed using SAS 9.2 (SAS, Cary, NC, USA). P-values <0.05 were considered significant. Values are given as mean ± standard error of the mean (SEM).

RESULTS

Parabolic distribution of the GKD-UF and QUF relationship in high-flux settings with ultrafiltration control

The GKD-UF determinations were repeatedly performed at the beginning of dialysis sessions in 15 patients. The parabolic distribution of GKD-UF was systematically observed with high correlation indexes (R2 = 0.995 ± 0.001 N = 150 determinations). The worst fit that was observed had an R2 value of 0.958 and the best was 0.999. The parabolic function held true in both post-dilutional and pre-dilutional HDF. Figure 1 shows the schematics of the setting to measure KUFin vitro (Figure 1A) and in vivo with an ultrafiltration controller (Figure 1B). In isolated ultrafiltration (Figure 1C), the P/Q QUF over TMP function describes a straight line when limited to 50 mL/min (the US Food and DRug Administration proposes 30 mL/min [15]), the slope of which is KUF based on Keshaviah et al. [17]. Adding a dialysate flow shifted the straight line to the right and increasing the filtration rate bent the line towards a plateau (Figure 1C). Finally, when GKD-UF was calculated and plotted over QUF, the parabolic function appeared with its vertex, GKD-UF-max (Figure 1D).

Reproducibility and variability of GKD-UF-max and its associated QUF:

Reproducibility for a given patient

GKD-UF-max and its related QUF were reproducible within a dialysis session. GKD-UF showed a coefficient of variation (CV) of 1.9 ± 0.7% when consecutively determined at the beginning of the dialysis session and 1.0 ± 0.3% at the end of the dialysis session (Figure 2A). The reproducibility of GKD-UF-max associated QUF was good, with CVs of 1.3 ± 0.6 and 2.3 ± 0.2% at the beginning and end of the dialysis session, respectively (Figure 2A).

FIGURE 2

GKD-UF-max and QUF at GKD-UF-max reproducibility. (A) GKD-UF was consecutively determined three times at the initiation and just before the end of the dialysis session in three patients. The coefficient of variation (CV) was calculated for each patient and the mean ± SEM of individual CVs is plotted. It can be observed that coefficients of variation the CV was <3%. The variability of the measure of the value of GKD-UF-max decreased with the dialysis session, whereas the variability of the QUF at which GKD-UF-max is obtained was remarkably low at the beginning. (B) The points represent the mean of a minimum of four GKD-UF measures for 12 patients and the bars are the SEM. The measurements were performed at the beginning of the first session of the week during four consecutive weeks. The blood flow was 370 ± 33 mL/min (mean ± SD), with a range of 300–400 mL/min. It can be observed that the cross-patient variation may be important (>30%), while the values observed for a given patient are in a narrow range (small SEM lines).

GKD-UF-max and its related QUF determined at the initiation of dialysis were reproducible from one dialysis session to the following one for every patient (Figure 2B). The average CV for GKD-UF-max was 3.9 ± 0.6% (highest 6.7%). The average CV for GKD-UF-max associated QUF was even lower (3.3 ± 0.3%; highest 5.1%; table in Figure 2B).

Variability across patients

GKD-UF-max varied across patients, from 31 to 42 mL/h−1/mmHg (36% increase; Figure 2B). GKD-UF-max associated QUF ranged from 82 to 100 mL/min (22% increase; Figure 2B).

Factors influencing GKD-UF-max and its associated QUF:

Patient characteristics

Across patients, the mean GKD-UF-max was negatively associated with plasma protein concentration (Spearman ρ = −0.77; P = 0.004), haematocrit (ρ = −0.63; P = 0.03) and haemoglobin (ρ = −0.58; P = 0.04).

Table 1

Patient characteristics

Characteristics	Patients (N = 15)
Sex ratio	8 males/7 females
Age (years), mean ± SEM	73 ± 12
Body weight after dialysis (kg), mean ± SEM	71 ± 2
Serum proteins (g/L), mean ± SEM	62.8 ± 1.2
Haematocrit (%), mean ± SEM	35.5 ± 1.4
Haemoglobin (g/dL), mean ± SEM	11.1 ± 0.3
Initial renal disease, n
Diabetic nephropathy	4
Glomerulonephritis	2
Nephroangiosclerosis	3
Polycystic renal disease	2
Other/unknown	4
Vascular access, n
Native arterio-venous fistula	14
Jugular catheter	1
Blood flow (mL/min), mean ± SEM	373 ± 8

Time of dialysis session

The GKD-UF parabola assessed at the start of the dialysis session was repeated after 1 and 3 h of dialysis, showing a significant decrease in GKD-UF-max both during haemodialysis and HDF (P < 0.001 for both; Figure 3A). More importantly, this decrease affected the absolute value of GKD-UF-max more than the associated QUF. Correlation studies showed that the GKD-UF-max change was significantly correlated to weight loss (R2 = 0.65; P < 0.001; Figure 3B). These data show that variations in GKD-UF during the dialysis session are patient dependent.

FIGURE 3

Mean GKD-UF-max variations during the treatment. (A) The influence of the dialysis technique and time on GKD-UF-max. The same patients were treated with either haemodialysis (HD) or HDF. GKD-UF was measured at the beginning of the treatment and at 60 and 180 min; HD and HDF are plotted. The absolute value of GKD-UF-max decreased during the treatment and, although not significantly different, there was a trend towards a greater decrease for HDF. (B) The influence of body weight loss on GKD-UF-max variation during dialysis. The change in GKD-UF-max (% difference of initial and 180 min) was significantly correlated with body weight loss] in percentage of total body weight; R2 = 0.65 (P < 0.001)], showing that patient factors (among which, probably refilling) influence the decrease of GKD-UF-max during the dialysis session.

Blood flow and infusion site

For six patients, GKD-UF-max was determined in four different conditions (pre- and post-dilution HDF each with 250 and 400 mL/min blood flows). The parabolic shape was always observed. Increasing blood flow significantly increased GKD-UF-max and its associated QUF in post-dilution HDF, whereas the opposite was observed in pre-dilution (Figure 4).

FIGURE 4

Effect of blood flow and infusion site on GKD-UF. The upper panels display the measurements of GKD-UF in a patient treated with two different blood flows [Qb = 250 mL/min (full squares) and Qb = 400 mL/min (open diamonds)] and using two different infusion sites [post-dilution (left-hand side) and pre-dilution (right-hand side)]. The lower panels show the effect of increasing blood flow on GKD-UF-max and its associated QUF. Both significantly increased in post-dilution HDF (left-hand side), while they decreased in pre-dilution HDF(right-hand side) (N = 6 and P < 0.05 for all), although to a lesser extent.

DISCUSSION

Determining GKD-UF-max is a new method to assess the convection characteristics of a dialysis system that is more adapted to the presently used technology (high convection flows, high-permeability dialysers, closed ultrafiltration circuit and ultrafiltration controllers) [13] than the ones advised by certain regulatory authorities [16], which were designed for low-permeability dialyzers and open systems [17].

Establishing the value of GKD-UF-max and its associated QUF at the beginning of the dialysis procedure provides an objective method to identify the best situation in terms of convection individually for every patient. Since it is a global measure in vivo [13], it takes into account all the parameters known to modulate ultrafiltration internally, alongside the dialyzer (haematocrit, total protein and elicited oncotic pressure) [18]. By determining GKD-UF-max, one can identify the setting with the highest convection for the minimal TMP constraints.

Given the instability observed when requesting very high ultrafiltration flows, the use of GKD-UF-max in clinics is promising to minimize the increase in TMP and consequent alarms while maintaining a high ultrafiltration flow. However, before the expected contribution of GKD-UF-max to high-flux haemodialysis and HDF is proved, it was important to address the reproducibility and/or variability of the method, as well as the factors influencing this variability. The present work provides all this information and shows that determining GKD-UF is easily performed, reliable, reproducible and has very low coefficients of variation. It is patient-specific, showing that the convection characteristics of a dialysis system may vary by a patient effect, and indeed stresses the value of a personalized prescription of convection. It further shows that the parabolic function also holds true in pre-dilution HDF. The opposite effect on GKD-UF following the increase in blood flow observed in pre- and post-dilution HDF, while maintaining the same infusion flow, is certainly influenced by variations in viscosity [19] at the dialyzer entrance (an increase in blood flow in pre-dilution results in an increase in viscosity by changing the volume/volume blood–infusate proportion, thereby decreasing GKD-UF). While this explains the observed decrease in water permeability, it does not indicate total removal efficacy of the system, which is decreased in pre-dilution HDF [20].

The values of convection obtained in post-dilution at GKD-UF-max in the example given in Figure 4 were ∼80 mL/min when Qb was 400 mL/min (20% blood processed), whereas they were 62 mL/min when Qb was 250 mL/min (25% blood processed). These results may be somewhat lower than those usually obtained with automated systems [9, 11, 12]. If the target convection volume (or flow) is higher than that obtained in the GKD-UF-max situation in a given setting, it is possible for the prescriber to increase the dialyzer surface area, to change the dialyzer and/or, if the patient is treated with post-dilution HDF and the vascular access allows it, to increase blood flow (as shown in Figure 4). Doing so, the dialysis system can be maintained in the GKD-UF-max situation while allowing higher convection volumes. Alternatively, prescribers may want to obtain the aimed convection volume by setting the system at a QUF exceeding that of the GKD-UF-max. Determining GKD-UF still informs the prescriber on the level of pressure constraints the system will undergo to obtain the prescribed convection volume. Using GKD-UF determinations is a completely different approach than limiting the convection to be prescribed (or obtained) to a percentage of blood flow or imposing a TMP threshold that may be used by other automated systems. We would propose to measure GKD-UF at the beginning of the session to use the value for prescription. The physician will prescribe at the GKD-UF-max, or lower or even higher than GKD-UF-max, and subsequently GKD-UF determinations may be repeated at any time during the dialysis session to identify any modification appearing during the treatment time.

In presently used clinical settings, determining GKD-UF is a promising tool guiding how to increase convection volume while maintaining system stability as long as possible. GKD-UF is the first objective parameter that has been proved to be applicable to both pre- and post-dilution HDF. Thus it may be of assistance for physicians prescribing high convection post-dilution HDF as well as for those aiming to further increase convection volume using pre-dilution HDF. Finally, GKD-UF may also be of assistance to describe the convective characteristics of a dialysis system very much in line with what is required by the regulatory bodies (FDA, EMA).