This preferred type of vascular access is created by connecting a vein to an artery, which requires that the two vessels be in proximity to each other. Fistula creation is most commonly a surgical procedure that can often be done with local or regional block anesthesia. Both the artery and vein must have adequate luminal diameter for the procedure to be successful, though other complex, poorly understood physiologic requirements are also needed for AVF maturation. In theory, the increased arterial pressure and flow into the venous system leads to compensatory changes in the vascular system, including dilation and elongation of the vein, thickening of the venous walls, and augmented artery-like flow rates through the fistula and outflow vein. When these venous changes occur (the term “arterialized” is often used), an AVF is deemed mature (ready to use) and should be easily accessible by physical examination. It may take 1 to 2 months for an AVF to mature enough for use; however, there should be early examination findings within the first few weeks to indicate progression. Functionally, a mature AVF should provide a robust blood flow (400–500 mL/min) that is equal to or more than the anticipated HD blood flow rate (Q _b ), with enough distance between the two needle cannulation sites so that recirculation—dialyzed blood from the venous return needle pulled back into the arterial needle—does not occur. If the HD prescription only calls for a Q _b of 200 to 300 mL/min, as is typical in Japan, then an AVF with lower flow rate would be acceptable. The “rule of 6” serves as an easy way to remember physical examination guidelines to determine if an AVF is ready for use: 1. ideally 6 mm in diameter; 2. 6 cm of overall needle-accessible length; and 3. no more than 6 mm below the skin surface.

The original procedure described by Brescia and colleagues was a side-to-side (side of vein onto side of artery) AVF using the radial artery and the cephalic vein at the wrist; it has become the preferred access when these vessels are adequate. However, patients starting dialysis are generally older with more comorbidities that lend to poor vascular health. Thus upper arm veins provide alternatives when distal vessels are felt to be inadequate in size or quality or when a distal AVF has failed. The cephalic and basilic veins are considered superficial veins of the arm and often used in AVF creation. The basilic vein courses up the arm and perforates the deep fascia to join the brachial vein somewhere between the midbiceps and axilla. The basilic vein AVF often uses the brachial artery as its inflow and typically requires an additional transposition procedure (moving the vessel out of its native bed and location) because of its awkward location for needle cannulation along the medial side of the biceps. The brachio-basilic fistula can be performed in stages, with anastomosis followed by a transposition surgery after the vein has matured. This procedure appears to have advantages over an AVG in terms of longevity. The basilic vein has also been transposed to the forearm with successful outcomes. Alternatives to using the cephalic or basilic vein as outflow for an AVF are the brachial vein (considered a deep vein of the arm) and median antecubital vein. The former is anastomosed to the brachial artery and requires a second procedure to superficialize and transpose, whereas the latter (variants of the Gracz fistula) has potential dual outflow through both the cephalic and basilic veins. These AVFs have advantages and disadvantages that are beyond the scope of this chapter.

In upper arm AVFs, end-to-side anastomoses are often the case. Studies suggest similar patency rates of side-to-side versus end-to-side anastomoses but possibly a higher incidence of arterial steal phenomenon and venous hypertension in the arm with the side-to-side technique. Computational flow dynamics suggest the more steep (sometimes near 90 degrees) angle of the anastomosis seen in an end-to-side AVF results in turbulence that may lead to undesired remodeling of the vein. In native veins, the higher turbulence from steeply angled connections predisposes to stenosis. For example, the cephalic arch at the confluence of the cephalic vein and the axillary vein is a common location to find recurrent stenosis requiring intervention in patients with an AVF or AVG with outflow through the cephalic vein ( Fig. 62.9 ). Typical vascular anatomy of the arm and the location of commonly placed AVFs are shown in Fig. 62.10A and 62.10B .

Angiography of a cephalic arch stenosis (white arrow) that developed in a patient with a left upper arm AV fistula.

The high blood flow rate in a location where the vein turns 90 degrees leads to turbulent flow, leading to intimal hyperplasia. Lesions like this need repeated angioplasty and sometimes benefit from stent placement or use of drug-eluting balloons.

Photo courtesy of Dr. Suresh Appasamy, Interventional Nephrology, University of California, Davis.

(A) Vasculature of the right upper limb outlining superficial and deep veins and major arteries.

(B) Location of anatamoses for typical wrist (radiocephalic) and upper arm (brachiocephalic and brachiobasilic) arteriovenous (AV) fistulae. (C) Location of typical AV grafts.

Medical device and procedural advancements have offered another avenue for creating an AVF. These percutaneously created AVFs use novel devices placed in the appropriate vessels under ultrasound and fluoroscopy guidance to connect an artery and adjacent vein by using a short pulse of radiofrequency energy. The common ulnar or radial artery and its respective veins are often target vessel candidates for this type of AVF. This technique potentially allows interventional radiologists and nephrologists to participate in AVF creation. Currently, there are limited data on outcomes, either short or long term. However, it is apparent that patient selection for percutaneous AVF is critical to success. Early reports suggest that additional follow-up procedures, such as angioplasty, are often needed to achieve maturation.

Despite the advantages of AVFs, the prevalence of AVF use in the United States has remained behind that of other countries. The primary failure rate of AVFs is high in the United States, around 25% to 37% in observational studies. To lower this rate, a good physical examination along with use of preoperative ultrasonographic imaging to assess vascular anatomy are used in the evaluation and planning of fistula placement. ^, Commonly termed “vein mapping,” this study should interrogate veins and arteries. Lower limits for artery and vein internal diameters are 2 mm and 2 to 3 mm, respectively. However, even in situations where vessel size appears to be appropriate, maturation of the AVF may still not occur. Intimal hyperplasia not readily seen with imaging and arterial lesions that limit augmentation of flow after AVF creation are likely additional factors that lead to a nonmaturing AVF. ^, Surgical technique is also a well-recognized variable in AVF outcomes. Data from the Dialysis Outcomes and Practice Patterns Study (DOPPS) have shown a marked difference in the training of surgeons in Europe compared with the United States. The much greater volume of procedures completed by European surgeons in training may lead to more expertise and comfort in performing AVF procedures.

To increase AVF use in HD in the United States, a “fistula first” initiative began in 2006, aimed to educate providers and patients on the advantages of this form of HD access. The dogmatic placement of AVFs in all patients, however, has disadvantages. A meta-analysis of more than 12,000 patients found the failure rate for AVFs increasing over time, casting concern about overly aggressive “fistula-first” approaches. The major disadvantage of attempting an AVF in all patients when primary failure rates are high and additional procedures are needed to reach AVF maturation (assisted patency) is the longer time with a CVC, during which a higher mortality is observed. ^, Therefore patient planning for HD access should include consideration of the next step, if an AVF is deemed unlikely to succeed from the onset or if there is primary failure of an AVF attempt.

When an AVF has failed or is unlikely to succeed upon initial access planning, there are two common alternatives to consider: 1. attempt an AVF at another site (often in the upper arm), or 2. place an artificial AVG. Vascular access using ePTFE conduits was the most common type of AV vascular access in the United States in the 1980s and 1990s, partly due to ease of placement, the short time required between placement and use, and perhaps factors related to reimbursement. An AVG still requires an arterial inflow vessel and a patent venous outflow. Grafts can be placed as a straight conduit or as a loop between the vessels. Common vascular connections include radial artery to basilic or cephalic vein, brachial artery to basilic or cephalic vein, and brachial artery to brachial vein ( Fig. 62.10C ). Cannulation for HD is into the graft material itself, so dynamic change of the vein is not needed. Thus AVGs have distinct advantages: They do not need to mature in the same manner as AVFs, so they can be used earlier, and have a lower primary failure rate than AVFs. Although the recognized low primary failure rate for AVGs is an early advantage, the primary patency rate at 12 months is only around 50% (i.e., half of AVGs will have failed or need intervention to remain patent at 1 year). When considering the assisted patency (or secondary patency) rate of AVGs, there is greater parity with AVFs. Another factor to consider is the greater access-related infection with an AVG compared with an AVF, though access-related infection is still highest with CVC use. Grafts also allow surgeons more possible sites for HD access, including the femoral vessels and even the chest wall using subclavian vessels. Placing an AVG in the leg may appear extreme, but an observational report suggests that a thigh AVG may still be better than a CVC. Regardless of location, most ePTFE grafts require about 4 weeks of healing after placement before cannulation is recommended, to allow for fibrosis or incorporation of the graft into the surrounding soft tissue (so that a surrounding hematoma does not occur upon needle cannulation) and for endothelization of the intraluminal surface.

Other AVG materials include alternative synthetic materials, modified biological materials, and tissue-engineered products. An example of an alternative plastic for AVG is polyurethane, known for its self-sealing property when punctured, obviating the need for graft incorporation, which could allow almost immediate cannulation after surgical placement. This graft may play a role in limiting or avoiding the use of a CVC. Seemingly an ideal material, the clinical experience with polyurethane AVGs has been mixed, with some reports of increased infections and difficulty with cannulation, and others of equivalency to traditional grafts. ^, A commercially available modified biological AVG includes the Artegraft, a collagen graft from acellular bovine carotid artery, whose maker touts earlier cannulation, lower thrombosis, and similar primary and secondary patency rates compared with traditional ePTFE grafts. Tissue-engineered AVGs coming onto the market are created from human cells stimulated to develop collagen and other structures of vascular tissue, grown on a scaffold that eventually dissolves. The cells are then washed away, leaving an acellular vascular graft. Possible benefits include lower risk of thrombosis, less infection, and lower immunogenicity. Outcome studies of these engineered grafts are still limited, and the overall cost-to-benefit balance of using AVGs from this technology is still unclear.

Even with an AVG as a vascular access option, sometimes the lack of a good venous outflow limits placement of an AV shunt. An available device for patients with limited venous outflow options uses a typical ePTFE graft at the point of the arterial anastomosis and transitions to a large-bore, single-lumen catheter at the venous outflow end. The catheter end is then tunneled up the arm and placed into the central venous circulation, much like a typical CVC, establishing a central venous drainage. This Hemodialysis Reliable Outflow (HeRO) device has a potential role in some patients as an AV access of last resort.

Possibly stimulated by areas of turbulent blood flow or changes in wall shear stress, a host of complex biological responses occur, particularly in the outflow vein distal to the AVG anastomosis. The resulting hyperplasia leads to a high incidence of stenosis in the vein immediately downstream from the anastomosis with the graft ( Fig. 62.11 ) and is the leading cause of graft thrombosis. This high incidence of stenosis is a rationale for monitoring AVG flow to allow preemptive intervention before thrombosis, though studies have not universally concluded that ultimate AVG survival benefits from monitoring. The inflammatory response and hyperplasia in AVGs have focused attention on devices and drugs that may prevent or reduce eventual stenosis. In addition to balloon angioplasty, stents are frequently used as a tool to prevent restenosis and maintain patency. Whether drug-eluting stents and drug-delivering angioplasty balloons will have a role in AVG survival has yet to be determined. Use of drug elution from the graft material itself is also under investigation in animals.

Angiography of stenosis in the outflow vein of a left upper arm AV graft (gray arrow).

This typical lesion in patients with an AV graft is felt due to turbulent flow as blood moves from graft to vein. These areas of intimal hyperplasia generally respond to angioplasty but tend to recur, oftentimes requiring stenting or drug-eluting balloons to maintain long-term patency. There are also numerous intragraft stenoses.

Photo courtesy Dr. Suresh Appasamy, Interventional Nephrology, University of California, Davis.

Starting HD with an AVG to avoid CVC time and converting to an AVF at the first sign of graft dysfunction combines the merits of the AVG (its higher primary patency rate) and AVF (improved longevity and resistance to infection). By starting with an AVG, the outflow vein is subjected to the higher pressure and flows that an AVF would similarly produce, thus maturing the vein. When the patient encounters the first episode of AVG dysfunction, the outflow vein is converted to a traditional AVF by connecting the already “arterialized” outflow vein to an artery, taking the AVG out of the circuit. The new AVF may be ready for cannulation immediately or soon after creation.

Catheters have transformed the application of HD in both acute kidney failure and long-term ESKD settings. Technologic advances have moved CVC designs from large, single-lumen catheters to the present dual-lumen devices. With advances in plastics, CVCs are now more flexible, leading to less vascular trauma, yet are more resistant to degradation by cleansing solutions such as alcohols. Catheters are placed within the venous system and do not require arterial flow, so HD can be provided to those who otherwise would not be able to obtain or sustain an AV shunt. Most importantly, the development of the subcutaneous Dacron cuff allowed a portion of the catheter to be tunneled subcutaneously before exiting the skin, which provides for longer use and protection against infection. Tunneled CVCs have a twofold to threefold lower infection rate than nontunneled CVCs.

Catheters have been both a blessing and a curse. Without a doubt, many patients would face death from kidney failure without the availability of catheter technology for emergent HD access. However, the ease of placement, ability for immediate use, and general convenience for the patient (fear of needles is common) have led to overreliance on CVCs. The USRDS report analyzing 2021 data showed that even in patients with >12 months of nephrology care before incidence of HD, about 72% of patients started with a CVC. Observational studies, such as those using data from DOPPS, relate the high prevalence of CVC use in the United States to some, but not all, of the higher HD mortality compared with similar patients in Europe. ^, It is also quite evident that health care is extremely fragmented in the United States, so predialysis planning is often difficult if not outright impossible.

Low Q _b s through the CVC are a common problem faced by dialysis personnel, resulting in a negative “arterial pressure” and machine alarm that stops the blood pump. Catheter Q _b s, even with some of the large-lumen catheters, reach a realistic maximum of 350 mL/min. In some countries, this may be perfectly fine, but in the United States with shorter treatment times and larger patients, any decrease in blood flow may result in inadequate HD clearance. The use of intraluminal tissue plasminogen activator (tPA), 2 mg in one or both ports, is often the approach to a catheter with poor flow. This usually adds about 30 to 45 minutes to a patient’s treatment time, as the tPA is allowed to dwell before trying the CVC again. The success rate for tPA varies and is probably influenced by patient and facility factors. A more chronic problem that can result in poor CVC flow is the development of a fibrin sheath that envelopes the intravascular portion of the CVC. Fibrin sheaths are likely not affected by anticoagulants or antiplatelet agents and cannot be remedied with tPA. Treating a fibrin sheath generally involves stripping the tough scar tissue–like material off the CVC or removing the CVC and using a large angioplasty balloon to disrupt the sheath. Fibrin sheaths often recur.

To avoid thrombosis, catheter lumens are filled or locked with various substances, most commonly high-dose unfractionated heparin solutions (1000–10,000 units per mL) with enough volume to fill both lumens at the termination of HD. Many studies have shown that a fraction of the heparin lock leaks into the circulation and can rarely result in bleeding. Some studies have shown no difference in catheter patency between 1000 units per mL versus 5000 units per mL heparin solutions. An alternative to locking with heparin is sodium citrate, whose calcium chelating effect may theoretically decrease thrombosis. One study demonstrated that locking the catheter with tPA once a week instead of heparin, compared with heparin all 3 days per week, reduces both thrombosis and the incidence of bacteremia. ^, However, implementing this strategy would be cost prohibitive. Interestingly, using a silicone catheter port cap (Tego) that does not require removal when attaching the HD blood lines and is flushed with saline was comparable with a more traditional catheter lock.

Catheter-related infection is a major complication of CVC use in HD, resulting in high morbidity, costly hospitalizations, and mortality. The frequency of bacteremia associated with HD CVCs has been estimated to be 2 to 4 episodes per 1000 patient-catheter days, a frequency 10- to 20-fold higher than rates of infection estimated in patients with AVF. These catheter-related infections can result in more complex infections such as osteomyelitis, endocarditis, or septic arthritis despite antibiotic therapy.

Patients can present in various fashions with a catheter-related infection, and notably ESKD patients are considered immunocompromised and may not have classic signs and symptoms of infection. Some patients simply feel poorly when they show up for HD, while others will present with tachycardia and hypotension. Fevers in any HD patient with a CVC should prompt a workup with high suspicion for a catheter-related infection. Skin-associated bacteria are the most common cause of tunneled HD catheter infections, with gram-positives accounting for 83% of organisms including Staphylococcus epidermidis (48%) and Staphylococcus aureus (28%) in a large observational study. Since most CVCs used in HD patients outside of the hospital are tunneled, the point of entry for bacteria may be intraluminal during the uncapping of the ports through touch contamination or via bacteria already on the port hubs. Therefore a “scrub-the-hub” protocol, using an antiseptic pad to vigorously remove residue and bacteria, is championed by most HD providers.

When a CVC infection is diagnosed, timely and appropriate antibiotic therapy, adequate duration of therapy, and determining if the catheter requires removal or exchange are critical for eradication of infection, as well as preventing metastatic infections. The type of organism from blood cultures (we feel cultures drawn through catheter ports and peripherally should ideally be part of the initial workup) dictates whether antibiotics alone will lead to full resolution; for example S. epidermidis may clear with antibiotics alone. Other bacteria, such as S. aureus, generally require CVC exchange or removal with a “line holiday” before reinsertion of a new one. The combined use of antibiotic instillation into catheter lock solutions, with systemic antibiotics, may increase success in preserving the existing CVC when removal is not an option. When the infection is isolated to the exit site and does not involve the tunnel, studies have suggested that with antibiotics, CVC exchange can be successfully undertaken.

Given the high infection rate of CVCs, attempts at prevention have been studied and a few prophylactic interventions show promise. The use of mupirocin ointment at the exit site may reduce the incidence of infection. ^, Locking CVCs with antibiotic-containing solutions (such as gentamicin) after all HD treatments or even just once weekly may also be helpful in reducing the incidence of CVC-related infections. With the prophylactic use of antibiotics, the issue of antibiotic-resistant bacterial selection is always a concern. Catheters impregnated with bacterial-inhibiting agents have not been as promising in preventing infection. Since many CVC infections are theorized to arise from bacteria found on the hub of the catheter ports, the use of a silicone cap that does not require removal each HD treatment (Tego, mentioned earlier) and chlorhexidine impregnated catheter hub caps (ClearGuard HD) are also innovations needing more real-world data.

Even if the thrombotic and infectious complications could be solved, CVCs may not support the Q _b s needed to achieve an adequate HD dose in some patients. Additionally, long-term CVC use results in vascular damage that often leads to central vein stenosis and in situ thrombosis. In the end, CVCs must remain an interim solution or the access of last resort in the patient on long-term HD.

The rationale for an AV access monitoring protocol is based on a few assumptions: 1. that the natural history of AVF or AVG dysfunction and impending failure can be predicted by observable changes in access pressure or flow that occur beforehand, 2. that if detectable changes in either pressure or flow were reliable and observed regularly, an intervention to prevent emergent loss of access function is available, and 3. that intervening on the problem will salvage the AV access and extend the lifespan of the access.

A consistently applied program of AV access monitoring is recommended by most guidelines. The first step is a good physical examination of the AV shunt. This can be taught to dialysis technicians, nurses, and nephrologists. For example, if an AVF has a bounding pulsation with increasing aneurysm size (or does not flatten when the arm is raised above the head in a lower arm, radiocephalic fistula), then a venous outflow or central vein stenosis should be suspected. In some teaching interventional nephrology programs, the physical examination findings correlate well to angiographic findings. However, the reality is that most dialysis clinics run a tight schedule between patient shifts, and a thorough physical examination of the AV access is rarely performed.

Other hints to access dysfunction may come from information gathered during patient rounds. For example, if a decline in HD urea clearance cannot be readily explained, a poor access flow rate may be the cause. Also, if there is routine difficulty in cannulation or prolonged hemostasis time (particularly if bleeding from the AV shunt at home is reported), arterial inflow or venous outflow stenosis may be the culprits, respectively.

Access pressure can be monitored by measuring static (blood pump not running) or dynamic (blood pump running at a predetermined rate) venous pressures when the patient is on HD by noting the venous chamber pressure. To be a useful tool, AV access pressure monitoring must be done serially using the same needle size, blood tube set, and Q _b (if dynamic venous pressure is being used) each time. The change or trend in measurement over time is more important in predicting a stenosis than an absolute threshold value. Clinical studies suggest that AVGs may benefit from monitoring of venous pressures, since most problems in AVGs occur in the native vein within the first few centimeters from the anastomosis. Conversely, arterial inflow problems in an AVG are less readily identified with venous pressure monitoring. Venous pressures are less useful in detecting AVF problems since these have multiple draining side veins and the fistula itself continues to remodel over time.

Access flow rates can also be measured while the patient is on HD. Shunt blood flow measurement that uses an ultrasound-based indicator dilution method (Transonic; Transonic, Ithaca, N.Y.) has proven to be quite accurate. Another method for measuring access flow uses dialysate-side changes in conductivity, the sodium-dialysance methodology built into some HD machines. Both methods are also capable of measuring recirculation. Flow monitoring and its benefit in long-term access outcomes, like with venous pressure monitoring, is not clearcut. A nonrandomized report comparing three monitoring schemes in a single HD program demonstrated significant improvement in short-term AVG and AVF survival using blood flow monitoring versus dynamic pressure monitoring or no monitoring. In addition to access survival, there were fewer missed HD sessions and reduced overall patient costs when using the blood flow monitoring protocol. Some studies of routine flow monitoring found benefit in secondary patency in AVFs but not in AVGs. However, a review of surveillance trials did not definitively support routine monitoring of HD access. Therefore the role of access surveillance is a topic of ongoing research. Nonetheless, a formal monitoring program puts more care team “eyes” on the HD access.

Access loss occurs from thrombosis and hyperplasia-related stenosis, but agents well known to prevent thrombosis in other vascular diseases have not consistently improved vascular access survival. A multicenter trial in the Veterans’ Affairs Health System with clopidogrel plus aspirin was discontinued because of bleeding problems in the treatment group. The Dialysis Access Consortium (DAC) Study Group supported RCTs in both AVFs and AVGs. For AVFs, the group reported that clopidogrel did not increase the number of usable fistulas. For AVGs, dipyridamole and low-dose aspirin (compared with aspirin alone) started immediately after placement of an AVG did improve graft patency at 1 year but did not result in any longer-term benefits. ^, Thus there are no clear guidelines on general systemic antiplatelet or anticoagulation use in patients with AVFs or AVGs for access survival.

The myointimal proliferation and resulting stenosis that occur, particularly with AVGs, have also been a focus of prevention, especially because this process may be the ultimate final pathway of access failure. Strategies to prevent cell proliferation in AVGs have been adapted from studies of other vascular diseases, such as cardiac disease. For example, a retrospective study using angiotensin-converting enzyme (ACE) inhibitors in patients with AVGs suggested a reduced risk for graft loss compared with patients not receiving this class of medication. Interventions that have been shown to reduce proliferation in other vascular diseases are being applied to HD vascular access. Brachytherapy has been used for some time—it is safe and has some promise in reducing stenosis in AVGs. Far-infrared light therapy, which improves endothelial function, was shown in an RCT to improve patency rates of new AVFs when administered repeatedly and may become a future tool to improve AVG survival. The use of antiproliferative medications such as paclitaxel, delivered locally by drug-eluting stents or by medication-loaded angioplasty balloons, holds promise for AVG failures related to venous outflow vein hyperplasia, which accounts for a large proportion of lesions that lead to graft thrombosis.

The clearance of a solute must be distinguished from its absolute removal rate. Clearance is best envisioned as a measure of removal expressed as a fraction of the remaining solute and is therefore independent of the concentration (see Eq. 1 in the prior section). Two substances may have the same clearance, but if one is present at half the concentration of the other, the removal rate will also be half. In practical terms, it is impossible to compare dialyzers by measuring removal rates alone because removal depends on the solute concentration. Measurement of clearance eliminates this requirement, allowing use of a single term to make valid comparisons among dialysis modalities and native kidney function.

Similarly, finding a lower concentration of a solute in a patient does not indicate that the clearance is higher; it may simply reflect a lower generation rate. Fig. 62.13 shows two theoretical patients, both with the same 36 kg total body water but different urea generation rates (40 and 30 g/hour). At a steady-state urea clearance of 20 mL/min, the patients have different urea removal rates and serum urea concentrations. If the clearance were to drop suddenly to 15 mL/min in both patients, a new steady-state serum urea concentration (mg/mL) would be noted. Once steady state is achieved again, the removal rate of urea is simply a reflection of its generation rate in the two patients. These concepts are similar in intermittent clearance by the dialyzer on HD.

Clearance versus removal rate of a substance is illustrated here, in two patients with the same 36 L volume of distribution of urea with the same constant clearance of 20 mL/min.

Patients 1 and 2 have urea generation rates of 40 g/hour and 30 g/hour, respectively. When the clearance of both patients suddenly drops to 10 mL/min, absolute removal rate of urea drops but at a different rate despite the same change in clearance. See text for more information.

The serum urea concentration has proven to be a poor surrogate for uremic toxicity. Urea concentration is a balance between generation and clearance. Whereas urea clearance by the dialyzer should correlate with the clearance of other small, dialyzable solutes that are presumably responsible for uremic toxicity, the generation of urea as an end product of protein catabolism correlates poorly with uremic toxicity—cell culture and animal studies have suggested that urea only mildly disrupts cellular function and promotes metabolic processes. In fact, patients on HD with higher urea generation rates have better outcomes, probably as a reflection of better appetite and higher protein intake. It is difficult to dissect the clearance factor from the generation factor in a single blood urea nitrogen (BUN) measurement, but as explained later, this can be differentiated by modeling the change in BUN during a HD treatment. For purposes of measuring the dose and adequacy of HD, only the relative change in urea concentration during HD is used to model clearance; the absolute concentrations are ignored. Thus despite its lack of intrinsic toxicity, urea measurements during HD can be used to assess dialysis adequacy as a surrogate for the clearance of other small, easily dialyzed solutes, some of which must be toxic.

Diffusive clearance in a flowing system depends on the rates of blood and dialysate flow, as well as the targeted solute’s membrane permeability. The membrane permeability constant (K ₀ ) for a given solute is a function of the biomaterial characteristics, including median pore size, pore size distribution, and membrane thickness. Multiplying K ₀ by the surface area of diffusion (A) yields the permeability or mass transfer-area coefficient (K ₀ A) of the dialyzer. K ₀ A is expressed in mL per minute and is independent of solute concentration. The predictable exponential decline in concentration gradient of a readily diffusible solute along the dialyzer membrane from blood inflow to outflow ( Fig 62.14A ) is the basis for the calculation of K ₀ A and is widely used in mathematic models of solute kinetics. For countercurrent dialysate and blood flow, where K _d is dialyzer clearance, the following equation is applicable:

(A) Flow-limited clearance.

A logarithmic decline in small solute concentrations, indicated by the arrows on either side of the membrane, is depicted from the dialyzer inlet at the left to the blood outlet on the right . This decline is based on rapid diffusion of solute across the membrane and forms the basis for Eq. 5 . Removal is maximized by countercurrent flow of blood and dialysate. (B) Membrane-limited clearance: The diffusive force is still the solute concentration gradient. Solute concentrations along the membrane from dialyzer inflow to outflow for both blood and dialysate are relatively constant because transport across the membrane is limiting and relatively low.

Analogous to clearance, which expresses the dialyzer removal rate normalized to the inflow solute concentration, K ₀ A is an expression of dialyzer performance normalized to specific blood flow (Q _b ) and dialysate flow (Q _d ) rates. K ₀ A is sometimes called the “intrinsic clearance” of a dialyzer and can be viewed as the maximum clearance possible for a particular solute at infinite Q _b and Q _d . Note that K ₀ A is both dialyzer and solute specific. It is the best parameter for comparing dialyzers, with higher values indicating more effective solute removal. Modern high-efficiency dialyzers have K ₀ A urea values of >500, with many well over 1000 mL/min.

A useful rearrangement of this equation, solving for K _d , provides a measure of dialyzer clearance at any Q _b and Q _d , as shown in the following:

As one can see, Q _b has a direct effect on clearance, and when one is attempting to get the most clearance per unit time, optimizing Q _b plays an important role. Realistically, HD Q _b rate reaches a maximum at around 500 mL/min given the limitations of needle gauge and tubing caliber. We also observe that increase in Q _d increases K _d. However, the effect of increasing Q _d on clearance K _d is modest. For example, assuming dialyzer K ₀ A urea of 1000 with Q _b of 400 mL/min, increasing Q _d from 600 to 800 mL/min only increases K _d by 15 mL/min ( Fig. 62.15 ).

For small, flow-limited solute removal, like urea, across a membrane, an increase in blood flow rate increases dialyzer clearance to a much greater extent than a similar increase in dialysate flow rate.

This model assumes no convective clearance from ultrafiltration and assumes a dialyzer mass transfer-area coefficient ( K ₀ A ) for urea of 1000, like what we expect for a modern, high-flux dialyzer.

The expression of clearance in Eq. 6 does not include the contribution of convective solute removal by ultrafiltration (Q _f ), which most patients on HD need to remain euvolemic. Convective clearance results from bulk movement of solute across the membrane driven as plasma water is removed by hydrostatic pressure. Simultaneous filtration across the same membrane used for HD removes additional solute, but the amount removed is inversely related to the efficiency of diffusion for that solute. For example, if the dialyzer removes a solute efficiently by diffusion, with an extraction ratio approaching 100%, addition of Q _f contributes minimally to the overall solute removal rate, which cannot exceed 100% of the blood inflow. Mathematically, the effect of Q _f on clearance is expressed as follows :

where Q _b is the blood flow rate, C _in is the inflow concentration, C _out is the outflow concentration, and Q _f is the ultrafiltration rate, in mL per minute. As C _out approaches zero, the dialysis component of clearance maximizes and the Q _f component extinguishes. Nonetheless, any quantification of solute clearance on HD should include the additional clearance provided by Q _f since the additive effect of convective clearance can be substantial over the course of a typical HD treatment.

Treatment-related variables already discussed include the permeability of the membrane to solutes based primarily on size, membrane surface area, blood and dialysate flow rates, and, of course, treatment time. A few other variables that affect clearance during HD are worth mentioning ( Table 62.4 ). Solute-related variables include the physical and chemical properties of the substance to be removed (size, charge, protein binding) and its distribution in the body (intracellular, extracellular, interstitial).

Solute molecular size determines its membrane permeability but in flowing systems, it also determines where along the dialyzer diffusion takes place. Smaller molecules tend to be cleared rapidly at the proximal end of the dialyzer, leaving the more distal end available for further enhancement of clearance as Q _b is increased, making small solute clearance flow-limited (see Fig. 62.14A ). The concentration of larger molecules tends to remain constant along the length of the dialyzer, and thus their clearance is minimally influenced by blood or dialysate flow rates ( Fig. 62.14B ). Note that flow-limited clearance for small solutes and membrane-limited clearance for larger solutes occur at the same time in a dialyzer. The relation between flow and clearance is shown graphically in Fig. 62.16 for both limiting scenarios.

Flow-limited versus membrane-limited clearances.

Flow limits clearance when the membrane is not fully exposed to inflow solute concentrations (see Fig. 62.12). This typically occurs for small, easily dialyzed solutes. In contrast, larger solutes tend to saturate the membrane along its entire length at lower blood (or dialysate) flow rates (see Fig. 62.13). For these less easily dialyzed solutes, further increases in flow have no influence on clearance. K ₀ A, Mass transfer-area coefficient.

The molecular activity of a solute determines its capacity for movement across the HD membrane. About 90% of whole blood volume is water, which is where water-soluble solutes can be dialyzed. Thus Q _b through the dialyzer for water-soluble solutes should be expressed as blood water flow, or about 90% of whole blood flow. Similarly, blood concentrations should be expressed as blood water concentrations, which are about 7% higher than whole serum concentrations. Note that BUN is a misnomer; serum urea nitrogen is measured.

For charged molecules, the Gibbs-Donnan effect may limit the diffusion of a charged solute by the electrostatic effect reducing the effective blood activity. This effect for blood equilibrated with dialysate is attributable to nondialyzable plasma proteins, mostly albumin, which has a net negative charge (≈17 mEq/mmol albumin). The asymmetric charge distribution across the membrane effectively “captures” a small fraction of the positively charged sodium (Na) ions on the plasma side, reducing their potential for diffusion. Correcting for this charge gradient, the effective Na concentration on the blood side of the membrane is about 3 mEq/L lower than its actual concentration, thereby affecting the movement or dialysance of Na and other positively charged solutes.

In general, the size of the molecule is the most important intrinsic physical feature governing its removal. The rate of movement, or flux (J), of smaller molecules is higher than the flux of larger molecules. Other factors such as plasma protein binding, shape, charge, and sequestration in the intracellular compartment must be considered when attempting to predict solute clearance.

A distinction must be made between clearance across the dialyzer and clearance across the patient. For each of these, the removal rate ( Eq. 1 ) is the same, but the denominator differs. Dialyzer clearance is an expression of solute removal as a fraction of the blood concentration (adjusted for blood water content, about 90%) at the dialyzer inflow port. In contrast, whole-body clearance is an expression of removal as a fraction of average concentrations throughout the body. The average whole-body concentration is substituted for dialyzer inflow concentration in the denominator of the standard clearance formula. Whole-body concentrations are higher than serum concentrations during HD because of solute disequilibrium, or a delay in the diffusive movement of solute from remote body compartments to the patient’s circulating blood during HD. There may also be nonuniform distribution of solutes. In general, whole-body clearances are always lower than dialyzer clearances.

Typical solutes that exhibit disequilibrium distribute preferentially in the intracellular compartment and diffuse slowly across the cell membrane to the extracellular compartment. Such solutes are sometimes labeled as “difficult to dialyze.” For example, HD is not recommended for removal of digoxin in patients with digoxin intoxication, even though it is water-soluble and easily removed in vitro by dialysis. Removal in vivo is limited by sequestration in remote tissue compartments. Solutes such as digoxin have a large apparent volume of distribution, often larger than total body water volume.

In normal physiologic conditions, urea is often termed an “ineffective osmole” and is considered to distribute evenly into total body water (V). Even though this solute normally diffuses rapidly across body membranes, it exhibits a concentration difference between water compartments during the high clearance rate of HD, displaying disequilibrium. Whole-body urea clearance can be calculated by using the equilibrated postdialysis BUN in the denominator in Eq. 1 , requiring the patient to wait 30 to 60 minutes after the completion of HD ( Fig. 62.17 ). Experience has shown that urea sequestration is predictable and has led to the development of correction factors to allow the use of the immediate post-HD BUN, obviating the need to wait for equilibration to take place.

Equilibrated postdialysis blood urea nitrogen (BUN), the basis for eKt/V.

Precise measurements of BUN every 15 minutes during a typical patient’s 2.5-hour hemodialysis shows a logarithmic decrease in concentration during the treatment and a rapid rebound that is complete approximately 1 hour later. The double-pool model shown in Fig. 62.23 and the solid line in the graph predict the concentrations accurately. The equilibrated postdialysis BUN is an extrapolated value shown as the large solid circle .

In quantifying HD by way of urea clearance (K), we express dialysis clearance relative to volume of distribution of urea, which is the patient’s total body water (V). Since HD is intermittent, we must enter a unit of time (t). Incorporating these factors, we arrive at the familiar term, Kt/V, which can be viewed as a fractional clearance of urea per HD treatment. Note that Kt/V has no units, since all unit terms cancel out. Also, Kt/V can be larger than total body water (Kt/V can be >1.0) since HD is a recirculating process.

In the blood compartment, solutes may diffuse slowly or not at all out of RBCs during transit through the dialyzer. ^, Water content of RBCs is about 64% by volume, so this may account for the additional transport of water-soluble toxins during HD. Solute clearance depends on how readily they move out of RBCs. For example, creatinine and uric acid are small molecules with clearances similar to that of urea when measured in vitro in a saline solution. Measured in vivo, however, urea clearance is higher than that of creatinine and uric acid because RBCs possess urea channels that provide facilitated transport across their membrane, augmenting the available urea for clearance across the dialyzer.

Potassium (K) and phosphorus (P) are cleared easily in vitro, but their whole-body removal is limited by cellular and bone sequestration in vivo, explaining the need for dietary restriction and the use of intestinal binders to diminish absorption. Phosphorus is rapidly removed from the intravascular compartment, causing intermittent hypophosphatemia in patients during conventional HD (high Q _b and Q _d , relatively short duration of treatment of 3–4 hours). However, P moves from sequestered locations back into the blood space in the 2 to 4 hours after HD, leading to a rapid increase in serum P, and returning close to predialysis levels ( Fig. 62.18 ). Thus sequestration and consequent postdialysis rebound account for the failure of conventional HD alone to normalize interdialytic concentrations of P. Intermittently low concentrations of serum P and other sequestered solutes during HD may also contribute to postdialysis disequilibrium symptoms. One can speculate that the magnitude of sequestration for the toxic solutes responsible for “uremia” cannot be as pronounced as for P; otherwise, HD would not be successful.

Effect of sequestration on serum phosphate levels and removal during dialysis.

Measurements of serum inorganic phosphorus concentrations were taken at 15-minute intervals during both high-flux and standard hemodialysis. The rapid flux of phosphorus due to its relatively high mass transfer-area coefficient caused levels to fall into the hypophosphatemic range (below the lower dotted line ) during most of the 4-hour dialysis. A marked rebound continued for 4 hours after dialysis ended.

Modified from DeSoi CA, Umans JG. Phosphate kinetics during high-flux hemodialysis. J Am Soc Nephrol. 1993;4:1214−1218.

The biomaterials used to make the hollow fiber dialyzer dictate its clearance and UF characteristics, as well as its biocompatibility. Two major classes of membrane material are available commercially: 1. cotton fiber, or cellulose-based membranes and 2. synthetic membranes. Unmodified cellulose-based membranes contain many free hydroxyl groups, which activate blood components (see “Membrane Biocompatibility”). Treating the cellulose polymer with acetate and tertiary amino compounds improves membrane biocompatibility, presumably through the covalent binding of the hydroxyl groups to form acetylated cellulose and laminated cellulose, such as Cellosyn or Hemophan. ^,

The major polymers in synthetic membranes are polyacrylonitrile, polysulfone, polycarbonate, polyamide, polyether sulfone, and polymethylmethacrylate. Although these membranes are thicker, they can be rendered more permeable than the cellulose membranes, yielding greater fluid and solute removal. The larger pore sizes in the synthetic membranes more efficiently remove higher molecular weight substances such as β-2 microglobulin (β ₂ M).

Dialyzer membranes may activate white blood cells, platelets, and the complement cascade via the alternative pathway to generate anaphylatoxins C3a and C5a. ^, The degree to which the membrane activates blood components determines its “biocompatibility.” Bioincompatible membranes may cause allergic reactions, hypoxemia, transient neutropenia (due to leukosequestration), altered immunity, tissue damage, anorexia, protein catabolism, or an inflammatory state. Because HD is repetitive, the effects of low-grade, subclinical membrane interactions during each treatment may be cumulative, eventually resulting in adverse clinical outcomes such as infection, accelerated atherosclerosis, frequent hospitalization, and death.

A membrane’s absorptive capacity can also influence its biocompatibility. Some synthetic membranes, such as polyacrylonitrile, polyamide, and polymethylmethacrylate, are more hydrophobic, bind proteins to a greater extent, and may ameliorate bioincompatible inflammatory reactions through their ability to bind anaphylatoxins and cytokines. ^, Therefore measurements of these elements in blood may not reflect the true capability of a membrane for inciting the complement cascade and inducing an inflamed state. In general, though, synthetic membranes and modified or “substituted” cellulose membranes are more biocompatible than unmodified cellulose membranes, ^{, ,} but issues may still arise, such as increased blood levels of bisphenol A (BPA) in patients treated with a polysulfone membrane.

Because bacterial contaminants in product water (see “Water Treatment”) can also activate blood components, it is difficult to determine the relative contribution from bioincompatible membranes versus nonsterile or inadequately purified water to the inflamed state seen in HD patients. With increasing use of modified cellulose and synthetic membranes and closer attention to water quality, the distinction has become even more difficult. Studies evaluating the relative biocompatibility of substituted cellulose versus synthetic membranes have reported no difference. ^{, ,} Ongoing efforts to improve biocompatibility include coating the membrane with heparin or vitamin E. ^,

Dialyzers perform two important functions: elimination of unwanted solutes and removal of excess fluid. The thickness, porosity, composition, and surface area of a membrane determine its ability to clear solutes and remove water. In general, thinner and more porous membranes provide more efficient transport of solutes and fluid. The dialyzer urea K ₀ A (or clearance) describes its ability to eliminate low-molecular-weight substances, while the vitamin B ₁₂ and β ₂ M K ₀ A (or clearance) describe the capacity to remove higher molecular weight substances and the ultrafiltration coefficient (K _UF ) is its ability to remove water ( Table 62.6 ).

Most dialyzers have a membrane surface area of 0.8 to 2.1 m ² . The desirable increase in solute transport associated with larger membranes can be achieved by increasing the length and number or by decreasing the diameter of the hollow fiber, but each maneuver has undesirable effects when carried too far. Lengthening the fiber increases the shear rate and resistance to blood flow, increasing risk of hemolysis. Increasing the number of hollow fibers increases surface area but expands the extracorporeal blood volume, which can compromise the patient’s hemodynamic stability. Smaller-diameter fibers can offset this disadvantage, but as fiber diameter decreases, resistance to blood flow increases, enhancing not only filtration but also backfiltration and clotting. As fibers thrombose, the effective surface area decreases. The design and geometry of the hollow fiber dialyzer represent a delicate balance among these factors.

Historically, low dialyzer membrane permeability limited HD efficiency, requiring more than 6 hours per treatment. As dialyzer design improved, treatment times decreased. In the late 1980s with the advent of more permeable dialyzer membranes, more precise UF control, and more reliable vascular access to achieve adequate blood flow, treatment times decreased to 2 to 3 hours three times weekly in the United States. The improved dialyzers ushered in the era of high-efficiency and high-flux HD.

The distinction between high-efficiency and high-flux dialyzers is imprecise, and sometimes these terms are used interchangeably. Both types of dialyzers have improved solute and fluid clearance compared with standard dialyzers and take advantage of higher Q _b and Q _d to reduce dialysis time while maintaining an adequate dose (see Table 62.6 ). High-efficiency dialyzers have a high K ₀ A and high small molecule clearance (such as urea) compared with standard dialyzers. High-flux dialyzers have a highly permeable membrane for larger molecules, such as vitamin B ₁₂ and β ₂ M, and have a higher K _UF than high-efficiency dialyzers, but not necessarily high urea clearances.

High-efficiency and high-flux dialyzers contain a substituted cellulose or synthetic membrane. Both membranes improve dialyzer permeability; substituted cellulose membranes can be made thinner to increase porosity and surface area, and synthetic membranes can be manufactured with more and larger pores. Both high-efficiency and high-flux HD require the use of bicarbonate dialysate (see “Dialysate Composition”). Additionally, the high K _UF of these dialyzers creates the potential for hemodynamic instability with pressure-controlled filtration and thus requires volume-controlled filtration (see “Dialysate Circuit”).

Because of their relative porosity, high-flux dialyzers can remove larger molecules such as β ₂ M, which are not removed at all by standard cellulose dialyzers. Removal of β ₂ M reduces the risk of carpal tunnel syndrome and other complications in long-term patients. ^{, ,} Initial results suggest that removal of other large molecules may offer additional benefits, such as an enhanced response to EPO, greater leptin removal possibly leading to better appetite, and perhaps lower mortality and hospitalizations. ^, However, potential adverse consequences include more efficient removal of amino acids, albumin, and drugs such as vancomycin.

Despite early promise in patients undergoing maintenance HD, RCT or crossover trials comparing high-flux HD with standard HD have found no difference in the incidence of hypotension and intradialytic symptoms, ^, BP control, neuropsychologic function, hemoglobin (Hgb) concentration, and use of erythropoiesis-stimulating agents (ESAs) or markers of inflammation, oxidative stress, or nutritional status. Three large RCTs, the High-Flux Hemodialysis (HEMO) Study, ^, Membrane Permeability Outcome (MPO) Study, and EGE Study, on the impact of membrane permeability on CV outcomes detected no significant difference in mortality or morbidity between patients treated with standard versus high-flux membranes. Post hoc subgroup analyses demonstrated that high-flux HD reduced CV events in patients with longer dialysis vintage in the HEMO Study, ^, reduced mortality in patients with low serum albumin (<4.0 g/dL) and DM in the MPO Study, and reduced CV mortality in patients with an AVF or DM in the EGE study. A meta-analysis of available data concluded that high-flux HD may reduce CVD mortality by 15% but does not alter infection-related or all-cause mortality.

Further increases in dialyzer flux may not lead to improved middle molecule clearance or outcomes as some solutes that accumulate in kidney failure are protein bound or sequestered inside the cell and must diffuse across cell membranes. The removal of such solutes remains time dependent. Initial experience has suggested that hemofiltration (HF) and hemodiafiltration (HDF) may augment the removal of larger molecules and protein-bound solutes through increased convective clearance. However, RCTs comparing HF and HDF with HD have found no difference in Hgb or ESA resistance, ^, serum phosphate levels, health-related quality of life, cardiovascular (CV) parameters such as left ventricular (LV) mass and pulse wave velocity, or intradialytic hypotension (IDH), ^, although some studies have reported less IDH with HDF. ^, Two of the three largest RCTs, the Convective Transport Study (CONTRAST) and Turkish Online Haemodiafiltration Study (OL-HDF), have reported no difference in CVD and all-cause mortality, but subgroup analyses have suggested a benefit in patients treated with greater convective and replacement volumes (>22 L and 17.4 L, respectively, so-called “high efficiency” HDF). The ESHOL (Estudio de Supervivencia de Hemodiafiltracion On-Line) Study, which achieved about 23 L of convective volume per session, reported lower CVD mortality, all-cause mortality, and hospitalization in patients randomized to HDF. Meta-analyses have concluded that convective therapies may reduce IDH ^{, ,} and CVD mortality, ^, but additional high-quality RCTs addressing the effect of convective volumes are needed.

Further technologic advances have allowed the development of high-retention onset or mid- or high-cutoff dialyzer membranes, which have larger but more homogenous pore sizes that will clear molecules up to 45 kDa in size yet prevent the excessive loss of albumin. Clinical data have suggested that these membranes clear larger middle molecules, such as advanced glycation end products and inflammatory mediators, comparable with that achieved with HDF, and may improve the residual syndrome (symptoms that remain in patients despite adequate dialysis) and reduce CVD risk. An economic analysis of RCT data demonstrated cost savings with maintenance HD using a medium cutoff dialyzer, driven by fewer hospitalizations compared with the control group using a high-flux dialyzer.

Online monitoring of clearance may provide the best assessment of dialysis adequacy. Online monitors record urea clearance by measuring the urea concentration in the dialysate, either continuously or periodically, ^{, ,} determining dialyzer Na clearance by pulsing the dialysate with Na and measuring dialysate conductivity at the dialyzer inlet and outlet (ionic dialysance), ^{, , ,} or determining clearance of uremic solutes by measuring ultraviolet light absorbance of spent dialysate. ^{, ,} Most online methods for monitoring urea or Na kinetics provide Kt/V based on whole-body clearance in addition to dialyzer clearance. ^, Online urea monitoring requires repeated calibration and has not gained popularity. Online clearance monitoring removes the expense and risks of blood sampling, reduces personnel time, and allows frequent determination of delivered dose, and provides real-time measurements for instant feedback. However, reported clearance values may differ and adjustments are applied by the instrument’s software to match the urea Kt/V more closely. ^, Drawbacks of online clearance monitoring include the need for multiple measurements of K _d to obtain an average for the treatment, accurate monitoring of treatment time, the need for blood urea sampling to allow determination of protein catabolic rate (a marker of nutrition), and the need to measure or estimate V to allow the calculation of Kt/V from the online K _d measurements.

The HCT can be measured online during HD using an ultrasound determination of plasma protein concentration or, more commonly, an optical measure of Hgb concentration or HCT. ^{, , ,} Patients who are prone to IDH and cramping may benefit from online monitoring of HCT because their symptoms are usually caused by decreased circulating blood volume when the UF rate exceeds intravascular refilling. ^{, , , ,} The degree of hemoconcentration reflects the immediate magnitude of intravascular volume depletion. Theoretically, altering the filtration rate during HD to minimize excessive hemoconcentration may reduce symptoms during HD and optimize the dry weight (DW). ^{, ,} In practice, using just the relative blood volume to guide the filtration rate has not been very successful in ameliorating symptoms, likely because of inaccuracies in measurement, the varied compensatory CV responses to volume depletion within and among individual patients, and a dialysis-induced reduction in arteriolar tone and LV function (myocardial stunning). ^{, , , ,}

Online determination of relative blood volume varies significantly among devices and underestimates the true decline in blood volume because of intravascular translocation of blood from the microcirculation (capillaries and venules), which has a lower HCT to the larger vessels. ^{, ,} What has shown more promise is identifying the pattern of the relative blood volume decline and using a computer-controlled biofeedback system to modify the filtration rate continuously during HD, in combination with clinical assessment such as symptoms and bioimpedance spectrometry. ^{, , ,} The absolute or total blood volume may be more predictive of intradialytic symptoms; automated online monitoring of absolute blood volume is not currently available, although mathematic models may allow derivation of the total blood volume from relative blood volume. ^,

Solute removal during HD reduces plasma osmolarity, favoring fluid shift into cells, and opposes net fluid removal. Raising the dialysate sodium concentration (Na _D ) helps preserve plasma osmolarity, offers the theoretic benefit of reducing IDH and cramping, and may allow continued fluid removal. Computer-controlled Na modeling (“sodium ramping”) changes the Na _D automatically during HD, usually starting at 150 to 155 mEq/L, and steps down to 135 to 140 mEq/L near or at the end of HD. Because of predialysis hyponatremia and overall positive Na balance during HD despite the stepdown, this causes increased thirst, excessive interdialytic weight gain (IDWG), and HTN, although the latter is not a consistent finding. ^, Instead, these unwanted effects can be avoided by individualizing Na _D to maintain a gradient of 0 to −2 mEq/L with respect to the predialysis plasma Na concentration, including perhaps diminished intradialytic symptoms because of lower UF requirements. ^, Such individualization may be accomplished through the use of online conductivity monitoring or by estimating each patient’s inherent plasma Na concentration (Na set point) using an average predialysis Na. This can be measured with direct potentiometry or mathematically corrected for the Gibbs-Donnan effect—Na that is not available for diffusion because of trapping by negatively charged proteins. ^, In addition, measured Na _D deviates significantly from prescribed levels in more than 40% of assays, ranging from 13 mEq/L higher to 6 mEq/L lower than prescribed, making it difficult to interpret studies evaluating the clinical effects of varying Na _D . ^, Given our incomplete understanding of the effects of altering Na _D on morbidity and mortality, the indiscriminate use of Na ramping and Na modeling in its current form should be abandoned, and individualization of Na _D should be done cautiously (see later “Dialysate Composition—Sodium”). ^{, , ,}

Ultrafiltration modeling provides a variable rate of fluid removal during HD according to a preprogrammed profile (e.g., linear decline, stepwise changes, and exponential decline of filtration rate with time). Theoretically, altering the UFR during HD allows time for the blood compartment to refill from the interstitial compartment, leading to less IDH and cramping. As with Na modeling, stand-alone UF modeling is relatively crude and altering the UFR in response to blood volume monitoring may be of more benefit (see previous, “Monitoring Hematocrit and Relative Blood Volume”). The effects of Na and UF modeling may be difficult to separate because they are often used together. ^{, , ,}

Technologic advances include the development of HD machines with biofeedback systems, allowing for computer-controlled adjustments of treatment parameters based on real-time input from the online monitors. The most common system used monitors blood volume and adjusts the UFR and dialysate conductivity to prevent them from decreasing below a preset value during HD. Small studies have demonstrated that this device ameliorates symptoms in hypotension- and nonhypotension-prone patients. ^{, , , , , , ,}

Automated control of dialysate temperature (T _D ) to maintain isothermic dialysis (constant body temperature) has shown promise and is superior to thermoneutral dialysis (using constant T _D ) in reducing IDH without incurring a Na load. ^{, , ,} Other studies have suggested that individualized dialysate cooling (0.5°C lower than body temperature) reduces episodes of IDH by 70%, raises mean arterial BP during HD, and abrogates both HD-associated cardiomyopathy and brain white matter changes. ^, Although these online monitors and automated biofeedback systems are expensive, they have the potential to reduce IDH, detect vascular access dysfunction, and increase dialysis efficiency while minimizing blood sampling. ^{, , , ,} By improving patient care, they may prove to be cost-effective in the long run.

Because HD patients are exposed to as much as 600 L of dialysate water/week, treating the water used to generate dialysate is essential to avoid exposure to harmful substances such as aluminum, chloramines, fluoride, endotoxins, and bacteria. Technical advances such as high-flux dialyzers, reuse or reprocessing of dialyzers, and bicarbonate-based dialysate have made high water quality even more imperative. To avoid complications, tap water is softened, exposed to charcoal to remove contaminants such as chloramines, filtered to remove particulates, and then filtered again under high pressure (reverse osmosis or RO) to remove other dissolved contaminants ( Fig. 62.20 ). A complete review of this topic is beyond the scope of this chapter, and readers are referred to reviews on the topic.

Schematic of a typical configuration of a reverse osmosis water treatment system.

Tap water undergoes filtration to remove gross particulate matter and then is softened before exposure to charcoal (carbon tanks) to remove contaminants such as chloramine. A second filtration process removes particulate matter, as well as microbiologic organisms. Finally, water is filtered under high pressure to remove dissolved contaminants such as aluminum (reverse osmosis). Product water is then either stored in a water tank or piped directly to each dialysis station.

Formal urea kinetic modeling considers the above aspects of urea and its clearance, using an iterative process of fitting known data into a model to achieve a good fit (usually <1% error). Changes in volume from UF, urea generation, and K _ru are incorporated in the modeling. Requiring a computer or programmable calculator, formal urea modeling provides not only the urea clearance Kt/V but also urea production, which can be translated to a normalized protein catabolic rate, useful in estimating nutrition. Urea modeling can be used as the monthly monitor of HD adequacy. Equations (such as the Daugirdas II equation) that give results accurate to that of formal modeling are often used instead.

The first step of urea kinetic modeling is ensuring accurate pre- and post-HD BUN measurements. Pre-HD sampling of blood is straightforward, but the post-HD BUN is prone to variability and measurement errors are more significant when the BUN is low, such as post-HD levels. ^, Although it decreases more slowly toward the end of HD, the BUN rebounds rapidly as soon as the blood pump is stopped (see Fig. 62.17 ). The early rapid phase of upward rebound is determined by both access recirculation and cardiopulmonary recirculation. Efforts should be made to sample after access-related rebound is complete but before cardiopulmonary rebound begins. The Kidney Disease Outcomes Quality Initiative (KDOQI) guidelines recommend slowing the blood pump to 100 mL/min for 15 seconds (to permit access rebound) or stopping the dialysate flow for 3 minutes and then drawing the sample from the dialyzer inflow port. Access recirculation dilutes the post-HD BUN, causing a falsely high Kt/V, which can endanger the patient because of inadequate dialysis. Sampling after cardiopulmonary rebound has begun gives a falsely low Kt/V.

In addition to measuring the HD adequacy, urea modeling allows the measurement of two patient parameters that independently influence the patient’s risk of mortality—urea generation (G) and the patient’s volume of urea distribution (V). Accumulation of urea results from both amino acid catabolism (a measure of protein nutrition) and failure of kidney excretion. Although these dual effects on urea concentrations complicate the interpretation of any single measured level, mathematic modeling of urea mass balance allows separation of the two and an estimate of urea distribution volume. Both higher urea generation rates and higher urea volumes are associated with lower mortality. ^, For patients dialyzed thrice weekly, diurnal variations in urea generation have little effect, but for nocturnal dialysis, the reduction in urea generation at night can cause a significant error, an overestimation of Kt/V, and underestimation of V if G is modeled as a constant.

Blood concentrations are the net effect of solute generation and elimination. If one attributes uremic toxicity to the concentration of accumulated solutes (concentration-dependent toxicity), it might seem logical that the clearance (Kt/V) should sufficiently balance the generation rate to maintain a safe low concentration. However, during the NCDS, attempts to demonstrate this relationship by reducing the dose of dialysis in patients who ate poorly caused an unfortunate vicious cycle of uremia-induced anorexia and malnutrition that eventually led to early discontinuation of the study. ^, Similarly, observational studies have shown consistently enhanced survival in patients who eat more, even when Kt/V is held constant, and patients who generate more creatinine have a similar higher rate of survival. ^, It appears that the relationship between diet and toxin generation and elimination is complex and poorly understood.

Urea modeling essentially provides a measure of the urea elimination constant, which can be considered as the fractional rate of urea disappearance during HD (K/V). To calculate K, one must know V or vice versa. Because the prescribed K should be the same as the delivered K, and prescribed K can be determined from Eq. 5 , modeled V is easily determined. By convention, V is expressed after dialysis because it is less variable. Comparison of modeled V from dialysis to dialysis can be used as a quality assurance measure, and values should not differ by more than 15%. ^, Causes of a discrepancy include access recirculation, dialyzer malfunction (from clotting or fouling of the hollow fibers), blood pump variances, and blood sampling and measurement errors. ^,

Several studies have shown that various measures of body size, including V, are associated independently with mortality ( Fig. 62.24 ). ^{, , ,} Survival rates in larger patients are higher than in smaller patients for reasons that are not entirely clear but may be related to nutrition and the caloric buffer afforded by muscle and fat. Because body size expressed as V is the size-normalizing factor for urea clearance in the Kt/V expression, larger patients require higher clearances and are therefore at higher risk for underdialysis. However, correction for the favorable influence of large size on mortality tends to mitigate this risk, as shown in Fig. 62.25A . Kt/V is a more powerful predictor of mortality than body size, and correction of Kt/V for the independent (and opposing) risk associated with body size renders it an even more powerful predictor of mortality ( Fig. 62.25B ).

Risk of death as a function of dialysis dose and body size.

The risk of death in hemodialysis patients decreases with increased dialysis dose (Kt/V) and may be further stratified by urea volume as a measure of body size. Larger patients, in general, have a lower death risk.

Risk of mortality related to body size and dialysis dose.

These data were obtained from a large observational study of 43,334 patients. (A) Hazard ratio analysis was adjusted for case mix. (B) Hazard ratio analysis included an interaction term between Kt and body mass index and was adjusted for case mix. BSA, Body surface area; L/Rx, liters per treatment.

From Lowrie EG, Li Z, Ofsthun N, Lazarus JM. Body size, dialysis dose and death risk relationships among hemodialysis patients. Kidney Int. 2002;62:1891−1897.

The HEMO study has uncovered a potentially size-independent effect of sex on the response to higher doses of Kt/V. Although mortality was not affected by administering a higher dialysis dose for the 1846 randomized patients as a whole, when females were analyzed post hoc, a borderline significant improvement in outcomes was seen at the higher dose. The counterbalancing effect was a nonsignificant higher mortality in males. However, sex was difficult to separate from size because the two are so closely linked, especially regarding V. If body surface area is considered to be the more appropriate denominator for dosing dialysis, females, and perhaps smaller males, would clearly require more dialysis than larger males when the dose is measured as Kt/V ( Fig. 62.26A and 62.26B ). Similarly, malnourished patients who lose weight have an automatic increase in Kt/V unrelated to the effort of dialysis, simply because the denominator in the Kt/V expression decreases. This dose increase in patients at higher risk of death may explain the reverse J-shaped relationship between Kt/V and survival in observational studies. Although the latest update of the KDOQI guidelines for HD adequacy raises the idea of normalizing Kt by body surface area, use of body weight, in essence, body water (V) to normalize dialysis dose, still predominates currently.

(A) Standard Kt/V in the conventional and high-dose HEMO study subjects, by sex.

(B) Surface area normalized standard Kt/V in the conventional and high-dose HEMO study subjects, by sex. Conversion to surface area was based on an anthropometric estimate of V in each patient. ^, V, Urea distribution volume.

From Daugirdas JT, Greene T, Chertow GM, Depner TA. Can rescaling dose of dialysis to body surface area in the HEMO study explain the different responses to dose in women versus men? Clin J Am Soc Nephrol. 2010;5(9):1628−1636.

Average HD treatment times for thrice-weekly HD are lowest in the United States compared with that in many other nations. This stems from the desire to be at the dialysis clinic for the shortest time possible. Other times, patients shorten their treatment times due to discomfort they experience toward the end of the procedure. Muscle cramps, fatigue, and general malaise increase in intensity as more fluid and solute are removed. Paradoxically, shortening the treatment typically accentuates these symptoms because the rate of removal must increase if the patient is to remain in solute and water balance. Extending treatment time (t) or increasing the dialysis frequency tends to alleviate these symptoms. Sometimes, a temporary trial of either an extended duration or increased frequency of HD is sufficient to persuade the patient.

Although the NCDS has shown only a borderline significant effect of HD duration, most population studies have shown that a longer duration is associated with enhanced survival. Like the NCDS, the HEMO study failed to show a significant benefit from longer time but it also did not specifically target dialysis duration, and the range of treatment times in the study was limited. Prospective observational studies and clinical experience favor prolonged treatments, which also result in lower UF rates. ^{, ,} A prospective, cluster-randomized pragmatic study to look at usual care (3–4 hours per HD) versus long (>4.25 hours per HD) duration in incident patients on thrice-weekly HD was attempted, but the study’s inability to sufficiently recruit into the longer duration arm resulted in termination of the trial.

Urea reduction ratio (URR) is simply the ratio of the change in BUN by HD to the pre-HD BUN:

where C ₀ and C are the pre- and post-HD BUN levels, respectively. This ratio has the advantage of simplicity. However, it is the least accurate measure of HD adequacy with many limitations. For example, as the frequency of HD increases, and presumably its efficiency improves, URR decreases. It is not possible to add URR values to show a cumulative weekly effect and, as the frequency extends to continuous dialysis, URR extinguishes to zero. URR does not take into account interdialytic fluid accumulation, urea generation, RKF, or the additional clearance afforded by UF. On the positive side, in addition to its simplicity, URR has a curvilinear relationship with Kt/V ( Fig. 62.27 ), paralleling the relationship between outcome and Kt/V. Although efforts have been made to convert Kt/V to a URR equivalent or to use the solute removal index, a more reliable index of HD dose, these approaches have not been popular. Other efforts to report the reciprocal of Kt/V as a concentration equivalent, targeting low concentrations instead of high clearances, have not been applied, partly because Kt/V has become ingrained in the practice of dialysis quantification.

Curvilinear relationship between urea reduction ratio (URR) and Kt/V, stratified by degree of ultrafiltration during dialysis.

Whereas the URR (see text) falls with increasing fluid removal during dialysis (ΔWt) from 0% to 10% of body weight, Kt/V increases. The latter more appropriately accounts for the increase in clearance caused by ultrafiltration. Curves are derived from formal urea modeling.

The average clearance of small, dialyzable solutes is easily derived from measurement of the pre- and post-HD BUN levels, using urea as the representative molecule. The instantaneous clearance of urea could also easily be measured by drawing blood samples from the dialyzer inlet and outlet, though this is not needed when using formal kinetic modeling or equations to determine treatment Kt/V. A novel approach to estimating instantaneous urea clearance across the dialyzer, which can then be used to calculate the treatment Kt/V, is by utilizing the conductivity clearance method.

This method, which is completely dialysate sided, takes advantage of the fact that Na moves across the modern dialyzer membrane nearly equivalent to urea. Thus understanding the dialyzer Na or ionic dialysance (capacity for ionic movement across the membrane), we can assume that urea is being cleared in a similar manner. Machines with this capability measure the change in electrical conductivity at the dialysate inlet and outlet, before and after short but abrupt changes in the dialysate concentration. ^, With these measures of Na ionic dialysance, usually performed three to six times per HD session, we then have instantaneous clearance (K). With treatment time and an input for V, we can arrive at the Kt/V for that treatment. This method has the advantage of not needing blood specimens and the results are immediately available. The use of an estimated V, usually by one of many available anthropometric calculators, proves to be the shortcoming of this method. Currently, guidelines do not fully support the use of conductivity clearance for routine HD adequacy monitoring.

Because Na is the major determinant of tonicity of extracellular fluids, the dialysate sodium concentration (Na _D ) influences CV stability during HD. Historically, Na _D was kept lower than serum Na concentration (SNa) at 130 to 135 mEq/L to facilitate diffusive Na loss during HD and prevent interdialytic HTN, exaggerated thirst, and excessive IDWG. However, with the advent of high-flux dialyzers in the late 1960s and more efficient solute removal, headaches, nausea, vomiting, seizures, hypotension, and cramps became more common and were attributed to the hyponatric dialysate ^{, ,} but were more likely caused by the use of acetate as a source of base (see later). This prompted a progressive increase in Na _D —first to that of SNa and subsequently higher than it—with an improvement in symptoms. The pendulum now has swung back, with some, but not all studies, demonstrating that a high Na _D leads to thirst, IDWG, and HTN, leading to a resurgent interest in reducing Na _D concentrations and abandoning the use of Na modeling or “ramping” (see “Components of the Extracorporeal Circuit—Computer Controls”) in most patients. ^{, , , , , ,} To complicate matters further, data from DOPPS have suggested a differential effect of Na _D on outcomes; HD patients with the lowest predialysis SNa have a lower mortality when dialyzed against a high Na _D , despite an increase in IDWG. ^,

The ongoing debate and uncertainty regarding optimal Na _D are due to multiple confounding factors, such as lack of RCTs, inaccuracy of Na _D delivery compared with prescribed, differences in dietary Na intake and urinary Na excretion, varying tissue Na stores (which affect vascular stiffness), and disparate predialysis plasma osmolarity not due to plasma Na that results in hypotension during HD as osmolarity decreases. ^, Conflicting data have suggested that one size does not fit all, leading to a growing interest for an individualized approach. Although a computer-controlled biofeedback system using conductivity to lower the plasma Na level to 135 mEq/L may offer the added benefits of improved fluid balance and BP control without sacrificing hemodynamic stability, these methods add complexity and/or increase demand on staff time. ^{, , ,} A reasonable alternative may be to apply a constant Na _D of 136–138 mEq/L empirically or align the Na _D with the average predialysis SNa to achieve the same ends, because many dialysis patients tend to be hyponatremic. ^{, ,}

Unlike Na, only 2% of the 3000 to 3500 mEq of total body K is distributed in the extracellular space. Although colonic excretion of K increases threefold in patients with ESKD and eliminates about 30% of dietary intake, the remaining K accumulates between HD treatments and can become life threatening. By using a dialysate potassium concentration (K _D ) lower than that of plasma, excess K is removed during HD, mainly through diffusion down its concentration gradient. ^{, , ,} However, K flux from the intracellular to extracellular compartment is usually slower than the efflux of K into the dialysate. The use of a high dialysate bicarbonate concentration (BIC _D ) and/or creating a large bicarbonate gradient during HD may enhance K shifting into the intracellular compartment, potentially creating significant intradialytic hypokalemia; this is followed by ∼30% rebound in the serum K concentration (SK) 3 to 4 hours after completion of HD. ^, Life-threatening hypokalemia typically occurs during the first 2 hours of HD, when a high predialysis K favors a precipitous decline in its concentration, leading to arrhythmias through hyperpolarization of cardiac membrane potential, QT prolongation, and increase in ventricular late potentials. ^{, , , ,}

Minimizing the risk for intradialytic hypokalemia and postdialysis rebound is made more complex by the high variability in K removal among patients and between treatments for the same patient despite an identical HD prescription. The intracellular distribution of K leads to a variable volume of distribution (V _D ) such that the greater the total body K content, the lower the V _D and the higher the fractional decline in K concentration during HD. Factors such as amelioration of acidosis, stimulation of insulin release by dialysate glucose, release of catecholamines in response to hemodynamic events, and decline of plasma tonicity all favor intracellular shifting, thus reducing the gradient for K removal during HD. ^{, , , ,}

Several large epidemiologic studies have sought to clarify the risks associated with varying K _D . ^, The optimal predialysis SK with respect to survival appears to be 4.6 to 5.3 mEq/L, with poor nutritional status contributing to death at lower SK and fatal arrhythmias at predialysis SK higher than 5.6 to 6 mEq/L. Dialyzing against a K _D lower than 2 or 3 mEq/L appears to increase the risk for sudden death, especially in patients with a predialysis SK <5 mEq/L. ^, A more recent DOPPS study did not find a difference in risk for sudden death between K _D concentrations of 2 mEq/L versus 3 mEq/L and was unable to interpret data for K _D lower than 2 mEq/L because of the small number of patients in this category, likely reflecting changes in practice since the previous studies. Although some studies have suggested a survival benefit in hyperkalemic patients dialyzed against low K _D , ^, other studies have found no difference in survival, even in patients with predialysis SK levels >6.5 mEq/L. ^, In addition, lower K _D seems to have little effect on predialysis SK, given the 2- to 3-day interval in between sessions. Individualizing the K _D , depending on the unique situation of each patient, may be crucial in navigating between increased mortality from predialysis hyperkalemia and sudden death from hypokalemia during and after each HD session. ^{, ,}

The prescribed K _D is guided by the predialysis SK and the considerations discussed earlier. ^{, , ,} Because of increased sudden death during and after HD when using K _D of 0 mEq/L, presumably due to a rapid decline in SK, its use should be abandoned. Most patients should dialyze against a K _D of 2 or 3 mEq/L. Patients with increased total body K from diet, medications, hemolysis, tissue breakdown, catabolism, or gastrointestinal (GI) bleeding may require a lower K _D . However, concentrations of 1 mEq/L should be used only when a compelling reason exists because of the higher risk for arrhythmias and death, and only after exhausting all efforts targeting dietary K restriction, use of K exchange resins, and discontinuing medications that interfere with aldosterone production and GI elimination of K (e.g., ACE inhibitor, angiotensin receptor blocker [ARB], and aldosterone antagonists). Patients on digoxin in particular must dialyze against a K _D of at least 2 mEq/L because of the greater propensity for digoxin toxicity and death with predialysis SK <4.3 mEq/L and intradialytic hypokalemia. Potassium modeling with a gradual stepdown in K _D , thus keeping the blood to K _D gradient constant, may optimize rate of removal and minimize the risk for arrhythmias. ^{, , , ,} However, experience with and data for this approach are scant and consist of small studies using electrocardiography to detect repolarization abnormalities (e.g., prolonged QT interval or QT dispersion) as surrogate markers for sudden death. The validity of these tools as surrogate markers has been questioned in the cardiology literature. ^, Randomized trials are needed to inform this dilemma.

With the availability of new oral K binding resins—patiromer sorbitex calcium and sodium zirconium cyclosilicate—usually taken on nondialysis days, many patients are able to remain on ACE inhibitors, ARBs, and aldosterone antagonists. ^, The use of oral sodium polystyrene sulfonate has largely fallen out of favor given the rare but serious side effect of intestinal necrosis and generally poor tolerability.

Historically, patients with ESKD were dialyzed against a higher Ca concentration (Ca _D ) of 3 to 3.5 mEq/L (1.5–1.75 mmol/L) to help control hyperparathyroidism and prevent Ca and subsequent bone mineral loss. ^, However, with the shift from aluminum to Ca-containing phosphate binders and increased use of vitamin D analogs in the 1980s and 1990s came more frequent hypercalcemia and concern for accelerated vascular calcification, so the trend shifted to a lower Ca _D . While KDIGO and KDOQI guidelines currently recommend a Ca _D of 2.5–3 mEq/L (1.25–1.5 mmol/L), they are mostly opinion based, reflecting limited understanding of Ca mass balance in patients undergoing HD. ^{, ,} Data from retrospective and small randomized studies have suggested that 2.5 mEq/L is the fulcrum for Ca _D , below which Ca is removed from the patient and above which Ca diffuses into the patient during HD, although there is wide variability among patients. ^{, , ,} Patients whose Ca _D was reduced from 3 to 3.5 mEq/L down to ≤2.5 mEq/L had decreased serum calcium (SCa) and higher parathyroid hormone (PTH) concentrations, correction of low bone turnover (adynamic bone disease), and improvement in vascular calcification, although other studies have reported unchanged or worsening vascular calcification with lower Ca _D . ^, Vascular calcification was assessed variably with measures of aortic calcification, coronary artery calcification, arterial stiffness, carotid intima-media thickness, and carotid femoral pulse wave pressures. None of the studies addressed interdialytic Ca mass balance (which may be variable given the selection of available phosphorus binders, vitamin D analogs, and calcimimetic agents) and its contribution to vascular calcification.

Dialysate Ca may also affect hemodynamic stability during HD through lowering ionized Ca concentrations, resulting in impaired LV contractility and possibly peripheral vasoconstriction. ^{, ,} Levels <2.5 mEq/L or a higher SCa to Ca _D gradient are associated with an increased risk of IDH, ^{, , , ,} potentially placing patients at risk of myocardial stunning and increased risk of sudden death. Using a higher BIC _D may exacerbate IDH because an increase in pH during HD may further reduce ionized Ca concentration, as well as affect the proportioning of Ca _D from concentrate. ^{, ,}

Whether or how vascular calcification and IDH relate to morbidity and mortality is not clear. Reports from studies of patients treated with lower Ca _D (≤2.5 mEq/L) have ranged from increased sudden death or increased heart failure hospitalization to enhanced or no difference in survival. ^{, , ,} Most of these studies were small; the largest study involved more than 43,000 patients but used a case-control methodology using retrospective data with potential for residual confounding. Phosphorus is a potential confounder that likely contributes to vascular calcification yet has not been adequately addressed in studies on Ca _D .

Given the complexities discussed and the observed wide variations in predialysis SCa, individualizing Ca _D may be ideal. Patients prone to IDH or at risk for sudden death may benefit from increasing the Ca _D to 3–3.5 mEq/L at the expense of increased risk for hypercalcemia, vascular calcification, and low bone turnover. ^{, ,} Until a Ca kinetic model accounting for the various factors discussed is available to guide optimal Ca _D prescription, one rational approach would be to maintain near-neutral Ca mass balance during HD, which can be accomplished with a Ca _D of 2.5 mEq/L in patients with predialysis SCa <8.75 mg/dL and a Ca _D of 3 mEq/L in those with predialysis SCa >9.15 mg/dL.

Similar to K, only 1% to 2% of Mg is in the extracellular compartment. Because two-thirds is in bone, Mg flux during HD is difficult to predict. With current practice in the United States of using a dialysate Mg concentration (Mg _D ) of 0.75 to 1 mEq/L, up to one-third of HD patients have low serum Mg concentrations (SMg), sometimes exacerbated by concurrent proton pump inhibitor use. Raising the Mg _D to 1.5 mEq/L will normalize predialysis SMg in most patients and will make some mildly hypermagnesemic. ^, Traditional practice has been to avoid Mg supplementation in patients with advanced CKD because of the reduced ability to eliminate it, given concern for life-threatening hypermagnesemia with bradyarrhythmias and neurotoxicity (decreased deep tendon reflexes and muscle weakness); this rarely occurs with an SMg below 4 to 5 mg/dL. Higher SMg may inhibit vascular calcification and suppress PTH but may also reduce bone mineralization, leading to adynamic bone disease. Low SMg may predispose to IDH, increase osteoclast activation, increasing risk of osteitis fibrosa, and increase the propensity for ventricular arrhythmias through depolarizing resting membrane potentials. Epidemiologic studies have suggested that low SMg is associated with mortality, but the absolute level differed among studies and ranged from <1.3 to <2 mg/dL, with a large study suggesting that the optimal SMg may be between 2.7 and 3.1 mg/dL. Adjusting for comorbidities and indices of malnutrition, inflammation, and atherosclerosis markedly attenuated the risk conferred by low SMg, but residual increased risk persisted in malnourished and/or inflamed HD patients. Only one study found an increased risk of death with high SMg above 3.1 mg/dL. Whether a higher Mg _D would reduce mortality is unclear. For now, consider the use of higher Mg _D for patients with persistent IDH and/or those at risk of cardiac arrhythmias or vascular calcification and lower Mg _D if adynamic bone disease is a concern.

Correction of metabolic acidosis during HD is achieved by increasing the dialysate concentration of a base equivalent to promote diffusion into the blood. ^{, ,} Historically, bicarbonate was introduced into dialysate by bubbling carbon dioxide through it to lower the pH and prevent precipitation of Ca and Mg salts. In the 1960s, acetate was introduced as a source of bicarbonate and became the standard for 2 decades. Acetate offered the advantage of a low incidence of bacterial contamination, lack of precipitation with Ca and Mg, and ease of storage. However, it became a hemodynamic stressor when high-efficiency and high-flux HD was introduced in the 1980s because the higher rate of acetate diffusion into blood exceeded the metabolic capacity of the liver and skeletal muscle. Acetate accumulation led to acidosis, vasodilation, and hypotension. These complications prompted a resurgence of bicarbonate-based dialysate, which persists today.

The major complications of bicarbonate dialysate are bacterial contamination and the precipitation of Ca and Mg salts. ^{, ,} Gram-negative halophilic rods require sodium chloride or sodium bicarbonate to grow and thrive in bicarbonate dialysate. ^, With regular disinfection of bicarbonate containers, these bacteria have a latency period of 3 to 5 days, an exponential growth phase at 5 to 8 days, and maximal growth rate at 10 days, which compare favorably with a latency of 1 day, exponential growth at 2 to 3 days, and maximal growth by 4 days in a contaminated container. Thus disinfecting the containers and mixing the bicarbonate daily help prevent bacterial contamination. The use of commercially available dry powder cartridges offers an alternative solution to this problem.

Bicarbonate and the acid concentrate, which contains all solutes other than bicarbonate, are separated until use to minimize the formation of insoluble Ca and Mg salts with bicarbonate. ^{, ,} The acid concentrate derives its name from the small amount of acetic acid (4–8 mEq/L in the final dilution) used to ensure the solubility of divalent cations. The dialysate delivery system draws up the components separately and mixes them proportionately with purified water to form the final dialysate. This technologic advance allowed the widespread reintroduction of bicarbonate as a dialysate buffer in the 1970s. Because some precipitation of Ca and Mg salts still occurs, the dialysate delivery system must be rinsed periodically with an acid solution to eliminate any buildup.

In many HD centers, bicarbonate concentration is fixed at 32, 35, or 38 mEq/L and allows the use of a central bicarbonate delivery system that pumps the bicarbonate concentrate from a central tank to individual patient stations. An advantage of this is fewer personnel back injuries, but a major disadvantage is the inability to individualize the BIC _D . By using dry powder cartridges or individual bicarbonate containers at each patient station, prescriptions can be individualized.

Although correction of metabolic acidosis is desirable to reduce protein catabolism, bone demineralization, inflammation, and insulin resistance, overcorrection to generate metabolic alkalosis during HD may predispose patients to hemodynamic instability, paresthesias, muscle twitching, cramping, and reduce cerebral blood flow possibly through alkalosis-induced lowering of the serum K and ionized Ca levels, as well as increased tissue calcium phosphate deposition. ^{, ,} Several large observational studies have reported that very low (<17 mEq/L) and very high (>27 mEq/L) predialysis bicarbonate levels were associated with mortality and hospitalization, but after adjusting for case mix and markers of inflammation and malnutrition, only the association between very low bicarbonate levels and adverse outcomes remained. A higher predialysis bicarbonate is likely a marker for poorer nutritional status. Patients with mild to moderate acidosis (bicarbonate 20–23 mEq/L) appear to have the best survival, perhaps reflecting more robust dietary protein intake, although newer studies have found no association between serum bicarbonate levels and mortality. Instead, a predialysis serum pH of ≥7.40 is associated with death from CV causes and unexpectedly, a higher BIC _D is associated with infection-related mortality. ^{, , ,} It is difficult to reconcile these findings and no large-scale, event-driven RCTs have compared high versus low BIC _D or oral bicarbonate supplementation in ESKD.

Although lowering BIC _D for patients with predialysis hyperbicarbonatemia is prudent, especially if postdialysis metabolic alkalosis is present, it may not improve outcomes. Instead, causes of malnutrition and inflammation should be identified and corrected if possible. ^, Whether patients with very low serum bicarbonate concentrations would benefit from raising BIC _D is unclear. In addition, kinetic studies have indicated that raising BIC _D increases serum bicarbonate levels at the end of and for 2 hours after HD but does not affect predialysis levels. Postdialysis metabolic alkalosis and its effect on serum K and ionized Ca levels, however, could explain a fraction of the excess mortality observed after the first HD treatment of the week. ^,

Residual confounding from inaccurate reporting of comorbidities, variability of serum bicarbonate measurements, disparity between prescribed and delivered BIC _D , and incomplete accounting of total base (acetate or citrate in the acid bath are converted to bicarbonate in the body) may have contributed to the disparate findings discussed previously. ^, Data have suggested that the base equivalent in the acid bath has negligible influence on the serum bicarbonate. If mortality risk is related to a high dialysate-to-blood bicarbonate gradient and abrupt changes in serum bicarbonate levels, bicarbonate modeling and individualizing BIC _D may be of benefit, in theory. ^, Until a definitive answer is available, oral bicarbonate supplementation may be preferable to raising the BIC _D to correct very low predialysis serum bicarbonate levels (<18 mEq/L).