Consequences of Frequent Hemodialysis: Comparison to Conventional Hemodialysis and Transplantation

Abstract

The average life expectancy of a person on hemodialysis is less than 3 years and hasn't changed in 20 years. The Hemodialysis (HEMO) trial, a randomized trial to determine whether increasing urea removal to the maximum practical degree through a 3-times-a-week schedule, showed no difference in mortality in the treatment and control groups. Investigators speculated that the increment in functional waste removal in the HEMO study was too small to produce improvements in mortality. To test this hypothesis, the NIDDK funded the Frequent Hemodialysis Network, a consortium of centers testing whether patients randomized to intensive dialysis would demonstrate improved (reduced) left ventricular LV mass and quality of life. The trial has two arms: the daily (in-center) and the home (nocturnal) arms. Each arm has patients randomized to conventional dialysis or 6 days (or nights) of dialysis. The results of the HEMO trial will be reported in the fall of 2010.

The expected mean survival of a 55-year-old American is 26 years. According to the United States Renal Data System (USRDS) 2009 report (1), the expected survival for a 55-year-old person with a kidney transplant is 15 years, but the expected survival of a 55-year-old person on dialysis is only 5 years (Figure 1). The mean survival for all people in America who start dialysis is 3 years. The shorter life expectancy for people starting dialysis is because the mean age of people starting dialysis is somewhat over 65 years. These generally grim statistics for people on hemodialysis have not changed substantially in more than 20 years. What is the basis for this shortened life expectancy, and is there a way to improve survival for those on hemodialysis?

Fig. 1.

Expected survival of a 55-year-old person. Data from USRDS in 2009 (ref 1).

BACKGROUND OF HEMODIALYSIS THERAPY

In 1993, Owen et al. (2) published an analysis of mortality of patients on hemodialysis, their albumin concentrations, and the dose of dialysis measured as a fraction of urea removed. This analysis showed, as others had, that serum albumin was a powerful predictor of mortality—with the lower the albumin, the higher the mortality. The data also indicated that the amount of urea removed during dialysis was related to mortality, with greater urea removal associated with lower mortality. A urea reduction ratio of 0.60 (i.e., 60% of the serum urea content removed) appeared to be a breakpoint. This urea reduction ratio remained a standard for dialysis “adequacy” for many years. Formulae were developed whereby nephrologists could quantitate the dose of dialysis using derivatives of the fraction of urea removed during a single dialysis session. The working jargon for dialysis dose became Kt/V, a value related to urea removal as computed with a constant (for the dialyzer), the time on dialysis (t), and the volume of distribution of urea (V).

The use of urea reduction as a reflection of the adequacy of dialysis was questioned by many nephrologists, but no convincing data have emerged to supplant its widespread use. Medicare, which Congress authorized to pay for dialysis in 1973, had established that thrice-weekly dialysis seemed adequate, and regulators set a limit of reimbursement for dialysis frequency. There were some exceptions, but thrice weekly dialysis became the dominant practice. During this period, different manufacturers developed dialysis membranes of different compositions, each claiming to improve the “biocompatibility” and solute-removal characteristics over those of the modified cellophane membranes used initially. The poor prognosis of patients on hemodialysis led to the HEMO trial to test whether a greater dialysis dose and/or improved removal of “middle molecules”—of ∼5,000 daltons—could reduce mortality. The study had a 2 × 2 design with patients being randomized to high-flux membranes versus cellophane, and a high versus conventional dialysis dose.

The HEMO trial showed that neither the higher dialysis dose nor greater removal of middle molecules improved mortality (3). These results were initially met with skepticism and disbelief. Nephrologists “knew” that more dialysis was better. They also “knew” that removal of middle molecules would reduce morbidity and mortality. It turned out that neither prediction was supported by the data.

Response to the Results of the HEMO Trial

There emerged two general theories on why more intensive dialysis did not improve mortality. The first theory was that kidneys make essential factors that dialysis can't replace. The second theory was that the increment in solute removal was insufficient to alter mortality. These theories are not mutually exclusive.

Support for the idea that normal kidneys produce factors that are vital to survival includes the well-established concept that the kidney is an endocrine organ. It produces renin, the active form of vitamin D, erythropoietin, and other potentially critical substances, such as renalase (4) and klotho (5). Renal production of many substances might singly or in concert have beneficial effects on survival. In support of this idea is that people who receive a kidney transplant live, on average, three times as long as those on dialysis, with factoring for age and gender.

The idea that the increment in dialysis dose might be insufficient to demonstrate improved survival requires rejection of the notion that dialysis dose is linearly related to survival. An assessment of how solute removal in standard and high-dose dialysis compares with normal kidney function is shown in Figure 2. Expressed in units of glomerular filtration rate (GFR), which for normal people is 120 ml/min, patients on conventional dialysis have the equivalent of 8 ml/min and the intensively treated group about 11 ml/min. Thus, while intensive dialysis administered in this way can provide about 40% more solute removal than conventional dialysis, this increment may not be large enough to alter mortality.

Fig. 2.

Comparison of estimated GFR values in dialysis and normal patients. GFR values for dialysis patients are computed by extrapolating urea removal rates over time. Kt/V values are computed from urea kinetic measurements and are estimates of urea removal during a single dialysis session. The lower value of Kt/V (1.1) is typical for conventional treatment. The higher value (1.5) was the target for more intensive dialysis using a thrice weekly schedule in the HEMO trial.

TESTING THE DIALYSIS HYPOTHESIS

The hypothesis that the increment in solute removal was too small to register an appreciable affect on mortality was the one that was easier to test than the hypothesis of unknown beneficial renal humors. The general strategy was to test the hypothesis that would provide the largest dose of dialysis possible—dialysis all night long for 6 nights a week. The corollary hypothesis was that increasing the frequency of dialysis to daily dialysis, even without providing the extent of solute removal that would be possible by dialyzing all night, would provide improvements in outcomes. The NIDDK funded the Frequent Hemodialysis Network (FHN) trial, a randomized trial designed to examine two types of frequent dialysis as compared to conventional dialysis (Figure 3). The nocturnal arm compared nocturnal dialysis 6 nights a week of the FHN trial to conventional home hemodialysis 3 days a week. The daily arm compared dialysis conducted in a center on 6 days per week for 2.5–3 h per day with conventional dialysis in a center on 3 days per week.

Fig. 3.

Design of the Frequent Hemodialysis Network (FHN) trial. Dialysis in the nocturnal arm was conducted at home; dialysis in the daily arm was conducted in dialysis centers. Conventional dialysis was always given 3 times a week. Frequent dialysis was targeted for 6 times a week. The duration of the dialysis in the nocturnal arm was much longer than the duration in the daily arm.

The idea that nocturnal dialysis could provide more effective therapy than conventional dialysis in a dialysis center has been most thoroughly tested in Canada. In uncontrolled observational studies, patients undergoing nocturnal dialysis reportedly feel better, reduce some of their medication requirements, and liberalize their dietary restrictions. In general, patients on short daily dialysis also seem to feel better but don't have as dramatic changes in medications and diet.

Designing the Trial

In designing the FHN trial, we realized that recruitment and adherence could be a problem. In general, recruiting patients to a randomized trial in which the treatment options are very different is challenging. For example, recruiting patients to be randomized to medical or surgical therapy is extremely difficult, because many physicians and patients have preferences that make them unwilling to engage in randomization. The Cardiovascular Outcomes with Renal Atherosclerotic Lesions (CORAL) trial (6, 7), in which patients with atherosclerotic renal artery stenosis are recruited to be randomized to medical management or physical intervention, represents a recent example of this. Another example is the attempt to recruit patients to be randomized to peritoneal dialysis therapy versus conventional hemodialysis therapy. That trial failed to recruit adequate numbers of patients.

The major question relating to the efficacy of intensive dialysis is its effect on mortality. To answer this question, given the underlying mortality and current estimates of effect size, would require about 1000 patients studied for 4–5 years, a population sample similar to that in the HEMO trial. Several considerations led to the conclusion that a study of this size and complexity would not be feasible. First, with respect to the frequent daily dialysis arm, maintaining patient adherence to 6-times-weekly dialysis sessions in a center would probably be very difficult, and few patients would agree to be randomized for a study lasting up to 5 years. The perceived uncertainties and patient burden were considered too onerous. Second, even with a modest increase in reimbursement from the CMS for a fourth dialysis session each week, the economic considerations were too negative for commercial dialysis centers to commit to a study lasting up to 5 years.

For the nocturnal dialysis arm, we perceived two major barriers to recruitment. First, it seemed likely that informed patients would be unwilling to be randomized to conventional therapy when they believed, given the existing literature (and testimonials), that nocturnal dialysis therapy was superior to conventional dialysis. Second, several centers in North America already had programs of frequent home hemodialysis, and many patients who wanted such therapy could find a way to get it. For these reasons, we decided to design a shorter study and to measure surrogate outcomes rather than mortality.

The final study design (8) included the intent to recruit 250 patients in each of the two arms of the FHN trial—the frequent daily arm and the nocturnal arm. Patients were randomized to conventional therapy or intensive therapy. At the completion of the trial, the daily arm was able to recruit its full complement of patients, but the nocturnal arm was able to recruit only 87 patients.

We analyzed the reasons for difficulty in recruiting patients for home dialysis (9); the results are summarized in the table. The predominant barrier to getting patients to agree to undergo home hemodialysis was lack of motivation. It was clear to the recruiters that those patients who had settled into a routine for a year or more were generally reluctant to consider a change in the way their dialysis was conducted.

TABLE.

Barriers to Patients Attempting Home Hemodialysis. Data Excerpted from Reference 9 and Listed According to the Frequency with Which Each Barrier Was Observed. All Reasons were Evaluated as Significantly Contributing to Patients' Ability to Conduct Home Hemodialysis

Psychological Barriers to Home Hemodialysis

Lack of motivation

Fear of self-cannulation

Fear of needles falling out or catheter disconnecting

Fear of machine

Fear of inability to learn procedure

Situational Barriers to Home Hemodialysis

Lack of partner or family support

Age greater than 70 years

Inadequate dwelling

Increased water and electric bills

Inadequate water supply

Reasons for patients being willing to consider home hemodialysis generally involved increased flexibility in scheduling, encouragement of a partner, and less travel to a dialysis unit.

Design Outcomes for the FHN Trial

Because the limited power of the FHN trial precluded using mortality as a primary outcome, we identified two surrogate outcomes. The first was the combined mortality and change in left ventricular mass measured by magnetic resonance imaging (MRI). The co-primary outcome was the combined mortality and change in quality of life as measured with the SF-36 RAND Physical Health Composite (PHC) (10). The use of two combined primary outcomes is an unusual design, but given the fact that the endpoints are surrogates, the combined endpoints added an element of certainty to a conclusion that frequent dialysis had a positive effect. The quality of life was chosen because of its obvious importance in a treatment that imposes a heavy patient burden. In addition, quality of life has a predictive value for mortality and hospitalization (11). The left- ventricular-mass measurement was selected because of its strong correlation with mortality, as documented in the dialysis population (12).

Outcomes of the FHN Trial

At the time of this writing, we do not know the outcomes of the FHN trial. However, there are outcomes from other trials that can be reviewed here. First, we have solid information regarding the magnitude of the effect on Kt/V with frequent daily dialysis and nocturnal dialysis. According to measurements made by Suri et al. (13), frequent daily dialysis improves urea removal by about 25%, while nocturnal dialysis improves urea removal by at least 2-fold. Others point out that a major benefit of nocturnal dialysis may be more related to time on dialysis than to the amount of urea removed.

Outcomes of Frequent Hemodialysis Studies

There are several reports of improved surrogate outcomes in patients undergoing nocturnal dialysis. Observational studies report regression of left ventricular hypertrophy (14), as well as improvements in blood pressure control (15, 16), mineral metabolism (17), cognition (18) and anemia (19).

There is only one randomized trial of patients on nocturnal dialysis. Culleton et al. (20) randomized patients to conventional home hemodialysis versus nocturnal dialysis and measured quality of life and left ventricular (LV) mass by MRI. Within 6 months of randomization, LV mass was reduced by an average of 15 grams while patients assigned to conventional dialysis did not have a reduction in LV mass (Fig 4). Culleton et al. also documented improvement in some kidney-specific measures of quality of life. The importance of LV mass in mortality has been recognized for decades in the general population. A similar association appears to hold for dialysis patients (21). Perhaps more importantly, dialysis patients who demonstrate the greatest increase in LV mass over 18 months have a much greater risk of death than those whose increase in LV mass is less (12). The rate of change in LV mass appears to be one of the most powerful predictors of mortality of all the known risk factors in patients undergoing hemodialysis.

Fig. 4.

Effect of nocturnal and conventional home dialysis on left ventricular mass. Patients on nocturnal dialysis for 6 months had a reduction in LV mass while those on conventional dialysis did not. Data from ref (20).

Survival on Nocturnal Dialysis

There are no randomized trials comparing the effect of frequent dialysis with conventional dialysis on mortality. However, one study compared a highly matched group of 177 nocturnal dialysis patients with patients receiving a deceased-donor or living-donor kidney transplant. This analysis demonstrated that patients undergoing nocturnal dialysis (5–6 nights per week) had a survival similar to that of patients receiving a deceased-donor kidney transplant. The group receiving a living-donor transplant had a somewhat better survival than either of the other two groups (22). This remarkable survival in the nocturnal dialysis group was even more striking because the nocturnal dialysis group had significantly greater comorbidities.

How Does Survival with a Kidney Transplant Compare with Survival on Dialysis?

The data from USRDS show that survival of patients with a kidney transplant is far superior to that of patients on dialysis. However, there is a large selection bias for transplant patients; most patients with significant comorbidities are not eligible for a transplant. To account for these differences, Wolfe et al. (23) analyzed the survival of transplant patients matched with dialysis patients who were listed for a transplant but who didn't get one. This method accounted not only for age, gender, and diabetes status, but also for confounding comorbidity. They found that up to 3 months after transplantation the risk of death was greater for transplant patients (i.e., surgical mortality), but that by 1 year the relative risk of death was 3-fold greater for dialysis patients than for transplant patients. Overall, the survival of patients with transplants is about twice that of matched patients on hemodialysis.

Mortality of Patients on Short Daily Dialysis

The evaluation of patients on short daily dialysis is less robust than that of patients on nocturnal dialysis. However, there have been some efforts to assess the long-term outcomes of short daily dialysis. Blagg et al. (24) compared the standardized mortality rate in 117 patients with matched patients from the USRDS. Their data showed a 61% better survival in those undergoing short daily hemodialysis. An analysis by Kjellstrand et al. (25) concluded that the duration of short daily dialysis was positively associated with improved survival.

There is one study comparing USRDS data for patients using short daily hemodialysis, nocturnal dialysis, and conventional dialysis. Johansen et al. (26) found that as compared with matched patients undergoing conventional dialysis, those using nocturnal dialysis had significantly lower mortality, while those undergoing short daily dialysis tended to have lower mortality.

PUTTING THE DATA TOGETHER–CONCLUSIONS FOR TODAY

The nephrology community is on the verge of demonstrating that more intensive dialysis—beyond conventional dialysis—can prolong life. From a broad perspective, we are close to concluding that improved removal of substances that accumulate in kidney failure can improve life expectancy. This conclusion addresses one of the major questions raised by the HEMO trial: Can any intensive dialysis therapy improve survival? It looks like the answer is yes, but proving this answer to the harshest skeptics will require substantially more work. There are three major barriers to making this conclusion with certainty.

First, our experience with the FHN trial shows that we will not be able to recruit enough patients in a randomized trial to compare mortality with nocturnal dialysis and that with conventional dialysis. Given patient preferences, we will probably not be able to convert most prevalent dialysis patients to intensive dialysis (27, 28).

Second, we need to devise ways to assess outcomes in patients undergoing different kinds of intensive dialysis. From the existing data, it appears that nocturnal dialysis will produce better survival outcomes than will short daily dialysis. However, the patient acceptance of these therapies will play a major role in how they are prescribed and who will receive them. Randomized controlled trials are not a practical way to address the effect of dialysis dose on mortality.

Third, it seems unlikely that nephrologists will be able to offer intensive dialysis to patients unless reimbursement policies are made friendlier or the costs of intensive dialysis are reduced. In order for nephrologists to understand the magnitude of the advantages offered by intensive dialysis in all its various forms, there will need to be large numbers of patients undergoing such therapies. Whereas some countries have adopted reimbursement policies that accommodate large numbers of patients on intensive home dialysis therapy, CMS has shown reluctance to develop policies that encourage home dialysis, especially intensive home hemodialysis.

DISCUSSION

Cohen, Washington, D.C.: John, very provocative data. I wonder how these data compare with continuous peritoneal dialysis.

Stokes, Iowa City: It turns out that that's actually been studied really very well, and that people undergoing peritoneal dialysis—which by the way is done almost always at home—have a virtually identical survival with people who have conventional center hemodialysis. Some studies show a little better survival and some show a little bit worse, but by and large it's a wash.

Sedor, Cleveland: I'm looking forward to the results of a trial next month, but I have one question: As you know, I work at the Old City Hospital in Cleveland. How applicable do you think setting up at-home nocturnal dialysis is to the population at large with end-stage renal disease?

Stokes, Iowa City: That is a very important question, and we've studied this as best we can, actually, throughout the world. In the United States, about 1% or so of the whole population is dialyzed at home. That includes peritoneal dialysis and home hemodialysis. It's a very small number compared to that in some other countries, like Canada, New Zealand, and Australia, where those dialyzed at home may be 15%, and their targets are 20%. So probably, 20% is an upper limit of the total population of dialysis patients that would be eligible for home therapy, and the point there is that not everybody with end-stage renal disease is going to be a candidate for this therapy. Having said that, we also don't know exactly how to do this best. It'll take us a long time to figure out where what I call a dose–response curve, and exactly the best way to do it for everybody. There are tradeoffs, as you well know. This is not a simple exercise for the family and for the patients who need dialysis. Dialysis is a very, very burdensome therapy. So everybody has to make a tradeoff.

Weir, Baltimore: John, that was very nice. As I mentioned last night, I think that an alternate hypothesis to this so-called uremic toxin may be that recurrent central pressure–volume overloading, from the inability to achieve dry weight, may lead to cardiac remodeling, which may explain some of the sudden death and arrhythmia that we see in these patients, and that one of the most unique observations about nocturnal hemodialysis is the ability to achieve dry weight, removal of medications, and regression of left ventricular hypertrophy. So, it may be a pressure–volume overload circumstance that is more effectively remedied by this type of modality; in fact, smaller studies now being published do show that achievement of dry weight, at least as measured with relative plasma-volume monitoring, is correlated with reduced mortality in dialysis patients.

Stokes, Iowa City: Yes, I think that's a really good hypothesis, and one of the things that all nephrologists will admit is that dialysis three times a week does not allow us or nurses or caregivers or patients to achieve what we think is really good dry weight. It's basically an economic problem, because we only have so much time to accomplish—say 3 or 4 hours—what a patient has accomplished in 48 or 72 hours at home, and we have to remove all of that fluid very quickly. So I think that's a very good hypothesis, and we'll see how that plays out with the data that is presented next month.

Henrich, San Antonio: Dr. Stokes, if the study is positive and the results presented next month are positive, one of the questions that will loom very quickly is the cost of nocturnal dialysis versus conventional dialysis, and given the strains on the healthcare budget, even with a benefit to patients, there may be some pressure in this regard. I wonder if you might comment on the relative costs, if you have any preliminary information about that and can tell us what you think is coming down the pike.

Stokes, Iowa City: Right. As Dr. Henrich astutely pointed out, this study, in part, has a political arm to it, because Medicare doesn't really want to pay for this effective therapy unless the CMS can be absolutely positive that first, the therapy is going to be more effective, and second, that it isn't going to break the bank. They'd hope to see that there would be some overall reduction in cost. I can tell you that at Iowa, we have a group of people that are trying to sort out the cost of this, and right now it's too complicated to say: “Yes, this is going to be, in the long run, more expensive or less expensive.” Hospitalization rates actually drive about half the total cost of dialysis, and the cost of dialysis itself is about half the total cost. So if you could reduce hospitalization rates and the horrible medical consequences, you could certainly save a lot of money, and that by itself might pay for the extra supplies it would take to do the dialysis. It turns out that if you do home hemodialysis, you can do four dialyses a week for about the same cost as three dialyses in a center. The break-even point depends on exactly how you do it, and it depends on how you justify it to CMS. Unfortunately, we are going into a system in January that is called “bundling,” and all the costs are going to be reorganized. I'm not sure we are going to be able to come up with a clear answer to this question right away. It's something we have to pay a lot of attention to.

Hochburg, Baltimore: My brief question will have three parts and you can answer them in a step-down fashion. First part: Are those two endpoints co-primary endpoints for the purposes of statistical analysis?

Stokes, Iowa City: Yes.

Hochburg, Baltimore: Second part, will you be doing cost per quality-adjusted life years as a secondary outcome, since you'll be able to calculate quality-adjusted life years by collecting the SF36 data?

Stokes, Iowa City: Maybe. That's not a primary or secondary aim, but there may be some attempts at that analysis.

Hochburg, Baltimore: Third, for the non-nephrologists, non-cardiologists in the audience, at least this one at the microphone, is the reduction in left ventricular mass as measured by MRI a surrogate outcome for survival?

Stokes, Iowa City: The answer to that one is yes, we think so. It's the best surrogate we can find. The cardiologists in the audience will realize that LV mass is, in fact, a pretty good surrogate for survival and also for bad outcomes, if you will. The situation in the dialysis population seems to be very similar. There are a lot of epidemiologic corrolations in dialysis that don't track with the rest of the population, but this one does seem to track, as best as we can tell.