Albumin handling in different hemodialysis modalities

ABSTRACT

Hypoalbuminemia is a major risk factor for morbidity and mortality in dialysis patients. With increasing interest in highly permeable membranes and convective therapies to improve removal of middle molecules, transmembrane albumin loss increases accordingly. Currently, the acceptable upper limit of albumin loss for extracorporeal renal replacement therapies is unknown. In theory, any additional albumin loss should be minimized because it may contribute to hypoalbuminemia and adversely affect the patient’s prognosis. However, hypoalbuminemia-associated mortality may be a consequence of inflammation and malnutrition, rather than low albumin levels per se. The purpose of this review is to give an overview of albumin handling with different extracorporeal renal replacement strategies. We conclude that the acceptable upper limit of dialysis-related albumin loss remains unknown. Whether enhanced middle molecule removal outweighs the potential adverse effects of increased albumin loss with novel highly permeable membranes and convective therapies is yet to be determined.

INTRODUCTION

With conventional low-flux (LF) hemodialysis (HD), removal of middle molecules and protein-bound uremic toxins is insufficient. Enhancing dialysis dose (based on urea clearance, Kt/V) and use of high-flux (HF) membranes did not improve clinical outcome [1]. Current dialysis practice increasingly makes use of convective therapies and highly permeable membranes to enhance middle molecule removal. High-volume hemodiafiltration (HDF) removes middle molecules more efficiently and may decrease the risk for all-cause and cardiovascular mortality compared with LF- and HF-HD [2–5]. This has led to the increasing use of HDF in Europe over the past decade [6, 7].

However, convective therapies and highly permeable membranes are associated with higher transmembrane albumin (66.4 kDa) loss than the previously routinely used LF-HD. Any additional albumin loss is a concern in the dialysis population in whom albumin levels are lower than in the general population, and may limit the application of treatment strategies with enhanced middle molecule removal. Hypoalbuminemia is a strong predictor of cardiovascular disease and all-cause mortality in dialysis patients [8–14], although it is unclear whether albumin levels per se influence outcomes. The risk associated with hypoalbuminemia can be linked to the cause of hypoalbuminemia, including malnutrition and inflammation, each of which has its own adverse effect on outcomes [15–17]. It is unknown whether treatment-related albumin loss has an effect on serum albumin levels or clinical outcome, independently of inflammation or malnutrition [18–22].

To give an overview of albumin loss with different dialysis membranes and different HD modalities and its influence on serum albumin levels, a comprehensive literature search was performed using combinations of the following keywords (including synonyms, abbreviations and different spellings): albumin, protein, dialysis, hemodialysis, hemodiafiltration, hemofiltration, convection, high cut-off (HCO), membrane and dialyzer. Studies were excluded when transmembrane albumin loss per treatment was not reported. Only studies in humans published in English were included.

ALBUMIN HOMEOSTASIS IN END-STAGE KIDNEY DISEASE

Serum albumin concentration is a function of its rate of synthesis by the liver, the fractional catabolic rate (FCR, the fraction of the vascular pool catabolized/unit of time), external loss (renal, gastro-intestinal and transmembrane loss during dialysis), hydration status and redistribution from the vascular to the extravascular space (or vice versa) (Figure 1). The predominant cause of hypoalbuminemia in dialysis patients is a reduction in albumin synthesis rate and an increase in FCR, both driven by the acute phase response [15, 16]. Albumin, prealbumin and transferrin are all negative acute phase proteins [23]. The reduced albumin synthesis rate in the presence of inflammation [24, 25] is mediated at the transcriptional level by cytokines, interleukin-6 (IL-6), interleukin-1β and tumor necrosis factor-α [26]. Adverse outcomes associated with any of these reflect in large part the level of inflammation mediated by these cytokines, also illustrated by the inverse relationship of albumin with positive acute phase proteins such as C-reactive protein (CRP), serum amyloid A and fibrinogen [27]. In addition to the decreased albumin synthesis rate and the increased FCR, extreme activation of the inflammatory response following sepsis or severe trauma results in altered vascular permeability to albumin, allowing translocation of albumin from the vascular compartment into the extravascular compartment, further reducing serum albumin concentrations [28].

The response to extracorporeal albumin loss (or albumin pool dilution secondary to expansion in plasma volume) is a transfer of albumin from the extravascular compartment to the vascular compartment [29–35]. Normally, ∼70% of the albumin pool is in the vascular compartment (Figure 2). This distribution is altered in patients on continuous ambulatory peritoneal dialysis (CAPD) and in nephrotic patients so that extravascular albumin is transferred to the vascular compartment, supporting serum albumin concentration. Albumin distribution in patients on HD is similar to that observed in control subjects and is not different among HD patients with mild hypoalbuminemia and those with normal serum albumin level [38, 39].

Fluid retention may cause hypoalbuminemia in dialysis patients [40], although the effect of plasma volume expansion might be partly offset by an appropriate increase in albumin synthesis rate. In HD patients, albumin synthesis rate correlates with plasma albumin levels [41], and in the absence of malnutrition or inflammation, HD patients are able to maintain albumin levels in the normal range by increasing albumin synthesis [42].

In addition to altered albumin distribution, albumin synthesis rate increases in case of extracorporeal albumin loss in patients on CAPD and patients with the nephrotic syndrome [36]. In the nephrotic syndrome, alteration in plasma oncotic pressure is followed by increased albumin gene transcription, which is accompanied by increased transcription of genes encoding both positive (fibrinogen) and negative (transferrin) acute phase proteins. Similarly, increased albumin synthesis rate in HD patients in response to expanded plasma volume is accompanied by an increase in the rate of fibrinogen synthesis [41, 42]. Thus, albumin synthesis varies inversely with that of fibrinogen and other positive acute phase proteins in the presence of inflammation, but varies positively with that of fibrinogen when the driving force for albumin synthesis is reduced plasma oncotic pressure. Fibrinogen is prothrombotic and associated with cardiovascular morbidity [43, 44] and graft thrombosis [45] in HD patients. Since fibrinogen is too large (340 kDa) to be filtered by HD, its plasma concentration increases [42]. Additionally, the synthesis of lipoprotein (a) [Lp(a)] is increased in parallel with that of albumin in the nephrotic syndrome [31] and Lp(a) concentrations are increased in dialysis patients in conjunction with external loss of albumin [46]. Lp(a) is also associated with increased cardiovascular risk and represents a potential hazard at any level of increased loss of albumin.

ALBUMIN LOSS WITH DIFFERENT DIALYSIS MEMBRANES AND HD MODALITIES

Classification of different membranes and their application

Membranes vary greatly in terms of removal characteristics of middle molecules and albumin loss depending on the ultrafiltration coefficient (K_UF), mass transfer coefficient and protein adsorptive capacity. Dialysis membranes can be classified as ‘low-flux’, ‘high-flux’ or ‘super-flux’ (SF) based on the K_UF (<10, >20 and >50 mL/h/mmHg, respectively) and the level of albumin loss in g/4 h of treatment (0, <2 and >2 g, respectively) [47].

Conventional HF membranes have a molecular weight cut-off (MWCO, i.e. the molecular weight at which the sieving coefficient is 0.1) of ∼10–20 kDa. SF membranes were developed to enhance removal of middle molecules and have a MWCO closer to that of the native kidney (∼65 kDa). HCO membranes (MWCO ∼50–60 kDa) were initially designed to remove large proinflammatory cytokines in patients with severe sepsis syndrome and are currently primarily applied for removal of monoclonal-free light chains [FLC, kappa (κ) 22.5 kDa, lambda (λ) 45 kDa] in myeloma cast nephropathy and of myoglobin (17.6 kDa) in rhabdomyolysis. HCO membranes have been applied for maintenance HD in pilot trials lasting up to 3 weeks [48–51]. In 2016, medium cut-off (MCO) membranes, tailored to the removal of large middle molecules while retaining albumin, were designed for routine use in maintenance HD patients [52–54].

In Japan, a different classification is used to identify five types of dialyzers based on β₂-microglobulin (β₂-MG) clearance (<10, ≥10, ≥30, ≥50, ≥70 mL/min, respectively) [55]. Types IV and V dialyzers correspond with SF membranes. High performance membranes (HPM) were developed in Japan in the 1980s to improve middle molecule removal and biocompatibility of conventional LF and HF membranes. They are characterized by a high permeability (β₂-MG clearance >50 mL/min at a blood flow rate of 200 mL/min), high flux-rate and more favorable biocompatibility [56].

Albumin loss with different membranes

Albumin leakage and middle molecule removal per single dialysis treatment for different membranes and modalities are shown in Supplementary data, Table S1 and Figures 3–5. With conventional LF-HD albumin leakage is usually absent and removal of middle molecules is very low [57–60]. Also, with HF-HD albumin loss is usually absent or low (<2.4 g/4 h treatment), but removal of middle molecules is higher than with LF-HD [16, 19, 39, 51, 58, 61–76]. Albumin loss with HPM-HD is usually below 3 g/4 h treatment, although greater albumin leakage up to 8 g/4 h treatment has been reported [56, 71]. With SF-HD, albumin loss is generally higher than with HF-HD (range: 1–5 g/4 h treatment), but the removal of middle molecules and protein-bound toxins increases as well [55, 57, 59, 64, 77–80]. MCO membranes improve middle molecule removal, including λFLC, compared with HF membranes, but at the expense of increased albumin loss [53]. HCO membranes outperform MCO membranes with respect to λFLC removal, but albumin loss is also much higher [53, 81, 82]. Reported albumin loss with HCO-HD ranges from 6 to 9 g/4–5 h treatment [21, 51, 72] and up to 12 g/ 8 h treatment (as is applied in myeloma cast nephropathy) [50]. Lower blood flow rates (150 mL/min) may lower albumin loss, even at extended treatment times (1.7–4.8 g/10–12 h), while maintaining high middle molecule removal [20, 81]. Increasing membrane surface area (2.2 m²) increases albumin loss as compared with a small surface area membrane (1.1 m²) [50].

Albumin loss with convective therapies

Convective therapies significantly increase middle molecule removal compared with diffusive therapies, especially when high transmembrane pressures (TMPs) are applied to obtain high convective volumes [83–85]. As expected, albumin loss with convective therapies is greater as well, especially in the post-dilution mode (range: 0.08–7 g/4 h treatment) [51, 62, 65, 68, 70, 81, 84–90], and increases with higher convective volumes [85]. Albumin loss with pre-dilution HF-HDF is generally smaller (range: 0.3–4.8 g/4 h treatment) as a consequence of the diluted albumin concentration available for convection [55, 80, 87, 90–94]. Mixed- and mid-dilution HDF may improve middle molecule removal and enhance albumin loss as compared with pre-dilution HDF [91, 95, 96]. Increasing convective volume from 20 to 30 L/4 h session moderately increases albumin loss from 0.08–0.4 g to 0.4–1.8 g/4 h treatment [85]. This indicates that increasing convective volume to 30 L is not limited by unacceptably high albumin loss.

To improve middle molecule removal and reduce albumin loss with conventional convective therapies, novel technologies have been developed. Normally, albumin loss is greatest within the first 30–60 min of treatment as a result of the high TMP applied to the intact membrane [20, 50, 80, 84, 88, 94, 97]. Further albumin loss is limited by the formation of a secondary protein layer caused by the deposition of proteins such as fibrinogen on the dialysis membrane, a phenomenon referred to as ‘fouling’. Albumin loss is reduced when blood flow and TMP remain low at the beginning of the session until fouling is complete [94, 98]. Pedrini et al. [95] used a TMP feedback control system that modulates filtration pressure and showed that slowly increasing TMP at the start of the session to a maximum value limited albumin loss both at initiation and during the whole session compared with operating at a constant TMP, while β₂-MG removal increased as a result of improved membrane preservation. Push-pull HD is another method that reduces aggressive filtration in the early phase of an HDF session. It is a form of internal HDF, characterized by alternate repetition of short fore- and back-filtration over the dialysis membrane. Shinzato et al. [78] reported a 66% reduction in albumin loss during push-pull HDF as compared with conventional post-dilution HDF, while β₂-MG and myoglobin removal were greater. Of note, albumin loss was still twice as high as during HD. Thus, aggressive filtration in the early phase of an HDF session should be avoided to minimize albumin leakage and enhance middle molecule removal.

Effect of HD-related albumin loss on serum albumin levels

Single-session studies show that dialysis-related albumin loss up to 26.4 g/4 h treatment do not lower post-treatment serum albumin levels compared with pre-treatment levels, possibly as a result of net fluid removal and albumin redistribution from the extravascular into the vascular compartment [51, 57, 61, 72, 84, 86, 87, 89, 92, 99]. Short-term follow-up studies did find a significant decrease (∼2–4.5 g/L) in serum albumin levels after 2–3 weeks of SF- and HCO-HD (albumin loss: 3.4–9.0 g/4–5 h treatment thrice weekly) [21, 59, 71, 79], whereas albumin levels remained stable after 2 weeks of HF-HD (albumin loss: 0.2 g/5 h treatment thrice weekly) [21] or 6 weeks of SF-HD with low albumin loss (1.2 g/4.5 h treatment thrice weekly) [64]. Long-term studies on the effect of dialytic albumin loss on serum albumin levels are scarce. Tsuchida and Minakuchi [71] report that serum albumin levels started to decrease after 1 month of HPM-HD and were reduced from 34.4 ± 3.0 to 32.2 ± 2.7 g/L after 3 years (albumin loss: 7.7 g/4 h treatment, frequency of treatments not reported). Hutchison et al. [100] suggest albumin replacement when using HCO membranes, although this is not routinely performed. Only three studies report that albumin was substituted during treatment with these membranes [20, 50, 100].

Of note, an attenuated inflammatory state as a result of enhanced removal of proinflammatory substances such as cytokines must be taken into account when comparing serum albumin during treatment with highly permeable membranes and/or convective therapies with that of conventional HD. Subanalysis of the CONvective TRAnsport STudy (CONTRAST), a large randomized controlled trial comparing outcome in online postdilution HF-HDF with that in LF-HD, showed that serum albumin decreased at a similar rate in both groups [101]. Stable CRP and IL-6 levels in patients on HF-HDF, as opposed to increasing levels in LF-HD, suggest that inflammation-augmented albumin catabolism was stable in HDF and increased in LF-HD (note: albumin loss was not quantified). In CAPD patients, daily transperitoneal albumin loss varies between 2.7 and 6.6 g [102] and may cause a decrease in serum albumin levels. However, CAPD patients without signs of inflammation or malnutrition maintain stable serum albumin levels despite higher transperitoneal protein loss than observed in hypoalbuminemic patients [103].

DISCUSSION

The improved elimination of middle molecules by membranes with increased pore size and convective therapies occurs at the expense of elevated albumin loss, which may theoretically be harmful.

Albumin loss during HD with HF membranes is negligible. Among the membranes with increased permeability, characteristics of novel MCO membranes seem most favorable with only moderately elevated albumin loss in MCO-HD, and significantly improved removal of larger middle molecules [>22.5 kDa, such as alpha-1-microglobulin (α₁-MG) and FLCs], both comparable to HF-HDF. MCO-HD might, therefore, be an alternative for HF-HDF when the prerequisites for HF-HDF are not within reach (such as online production of substitution fluid and high blood flow). Future studies should evaluate the impact of MCO-HD on outcome.

The addition of convective transport to HF-HD results in both enhanced removal of middle molecules and increased albumin loss (to a degree comparable to that in CAPD). The possible favorable effects of high volume HDF, including prolonged survival, should be weighed against the possible harms resulting from increased albumin loss. The possible favorable effect of high efficiency HF-HDF on outcome [104] indicates that the potential adverse effect of increased albumin loss does not eliminate the beneficial effects of HF-HDF. Increasing convective volumes up to 30 L/HDF session only moderately increases albumin loss. This is relevant, since increasing convective volume beyond 23 L/session might be associated with improved clinical outcome [5]. Albumin loss in HF-HDF can be limited by keeping TMP low at the beginning of a dialysis session until fouling is complete.

Albumin loss >3.4 g/4 h treatment, as in thrice-weekly HD and HDF with SF and HCO membranes, is associated with a decrease in serum albumin levels within 2–3 weeks after treatment initiation, suggesting that albumin loss is too high to be compensated for by an increase in albumin synthesis and/or altered distribution. However, studies did not control for malnutrition and inflammation, although inflammation-augmented albumin catabolism may be relatively low with these membranes due to enhanced removal of proinflammatory cytokines. Albumin turnover studies with radioactively labeled iodine, as performed in LF- and HF-HD [16, 39], should be repeated with these membranes to properly differentiate between these factors. Albumin substitution may be considered during treatment with HCO membranes.

Based on this review of the literature, the question remains whether a decrease in serum albumin levels due to extracorporeal albumin loss is harmful. The adaptive mechanism to albumin loss is increased synthesis of albumin, which is accompanied by increased synthesis of positive acute phase proteins, which may theoretically adversely affect outcome [105]. On the other hand, losing a certain amount of albumin might be beneficial due to increased removal of albumin-bound toxins and the oxidized form of albumin that has lost its antioxidant effect. This may promote synthesis of new functional albumin with antioxidant properties [71].

This review is limited by the fact that comparison of albumin loss with different treatment modalities and dialysis membranes is hampered by variable operating conditions that may influence albumin loss such as TMP, blood and dialysate flow rate, and treatment time.

In conclusion, the acceptable upper limit of extracorporeal albumin loss per treatment remains unknown. Long-term controlled studies need to evaluate whether the beneficial effects of enhanced (larger) middle molecule removal with novel highly permeable membranes and convective therapies outweigh the potential adverse effects of increased albumin loss on patient outcomes.

SUPPLEMENTARY DATA

Supplementary data are available online at http://ndt.oxfordjournals.org.

CONFLICT OF INTEREST STATEMENT

None declared. The results presented in this article have not been published previously in whole or part.

REFERENCES