The past decade has witnessed important developments in the care of patients with kidney disease. Incidence rates of end-stage renal disease (ESRD) have generally stabilized in industrialized countries. In contrast, in developing countries, incidence rates of treated ESRD have increased dramatically, although this is actually a welcome development because it reflects increased access to maintenance dialysis in these countries. As many such nations continue to develop, the world will face a prodigious economic challenge in delivering ESRD care, principally maintenance dialysis, to these patients. In industrialized countries, relative gains in life expectancy for dialysis patients have substantially increased over a comparatively short time, although absolute differences in life expectancy between patients receiving dialysis and their age-matched, nondialysis counterparts remain high, most strikingly for the youngest patients. Much of the improvement in mortality rates is due to falling rates of cardiovascular disease–related mortality, which is now comparable to infection-related mortality in the United States. Unfortunately, sudden cardiac death rates have improved little over the past decade. To continue the trend toward improved mortality rates in prevalent dialysis patients, reduction in the rates of sudden cardiac death and infection-related mortality is critical. Costs for maintenance dialysis vary substantially by country. It is least expensive in less-developed counties, as expected, but still constitutes a huge expense in relative terms. As such, therapies designed to forestall or slow the development of kidney disease in at-risk individuals are as important as ever, and perhaps more so. The discovery and implementation of novel treatment paradigms is essential if the incidence of kidney disease is to be reduced and the care of kidney disease patients improved.
Keywordschronic kidney disease, costs, diabetes, epidemiology, etiology, hypertension, incidence, outcomes, prevalence
Development of End-Stage Renal Disease: Conceptual Considerations, 311
Epidemiology of End-Stage Renal Disease, 311
Causes of End-Stage Renal Disease and Indications for Maintenance Dialysis, 311
Incidence and Prevalence of End-Stage Renal Disease, 313
Timing of Dialysis Initiation, 317
Outcomes in End-Stage Renal Disease, 319
Mortality Trends in Patients Receiving Maintenance Dialysis, 319
Cause-Specific Mortality, 321
Morbidity and Hospitalizations, 321
Dialysis Modality, 331
Costs of End-Stage Renal Disease, 335
End-stage renal disease (ESRD), defined as loss of kidney function such that life is unsustainable in the absence of renal replacement therapy (RRT), is an enormous public health challenge. RRT, provided in the form of a transplanted kidney (renal allograft) or, much more commonly, long-term (“maintenance”) dialysis, exacts an immense toll on the affected individual, the healthcare system, and society in general. Dialysis, as an “organ replacement therapy,” has been rightly hailed as one of the greatest advancements in the history of medicine ; however, it is expensive, burdensome, and far from an ideal solution to kidney failure. Since ESRD is no longer limited to industrialized countries, nations at all levels of economic development will struggle to meet the growing worldwide need for RRT. One positive development is that ESRD, which has long been recognized within the nephrology community as a problem of immense proportions, has gradually come into greater focus in the broader public health community. The US Healthy People (HP) 2020 initiative, for example, outlines specific public health targets for both chronic kidney disease (CKD) and ESRD. In this chapter, we discuss the epidemiology, outcomes, and costs of ESRD, particularly as they pertain to maintenance dialysis.
Development of End-Stage Renal Disease: Conceptual Considerations
Ongoing loss of functioning renal parenchyma has been termed the sine qua non of progressive kidney disease leading from CKD to ESRD. CKD does not inevitably lead to ESRD, and ESRD is not invariably a consequence of CKD. For example, for some patients, CKD is only slowly, if at all, progressive, as can occur due to a transient renal insult earlier in life. In contrast, some patients can develop ESRD after acute kidney injury (AKI) resulting from a serious illness or trauma, perhaps via a hastening of progressive CKD. Although CKD progression and advancement to ESRD are by no means inevitable, especially in people with less-advanced CKD, ESRD is generally the result of chronic processes that characterize CKD and that lead to progressive nephron loss. Nephron loss results in compensatory hyperfiltration in the remaining nephrons, an injurious process that itself contributes to CKD advancement. Depending on the underlying cause of CKD, proteinuria often is manifest, which itself is injurious to the kidney through a variety of biological mechanisms.
The risk that CKD will advance to ESRD can be underappreciated because mild to moderate CKD generally produces no signs or symptoms, and compensatory nephron hyperfiltration makes estimated glomerular filtration rate (eGFR) a “lagging indicator” of disease. Once serum creatinine levels have overtly increased (i.e., once eGFR has manifestly decreased), more than half of nephron mass may have been lost. The unfortunate reality is that the symptoms associated with kidney failure often occur only when ESRD is impending. An additional difficulty is that CKD progression is variable and often unpredictable for individual patients, so the urgency to act and the calculation of risks and benefits for particular therapies may be uncertain for patients and their healthcare providers.
Epidemiology of End-Stage Renal Disease
Causes of End-Stage Renal Disease and Indications for Maintenance Dialysis
The cardinal causes of CKD are diabetes, hypertension, glomerulonephritis, and polycystic kidney disease. As would be expected, the major causes of CKD are also the major causes of ESRD. Large ESRD registries, such as the United States Renal Data System (USRDS), provide information about causes of ESRD. As noted in the USRDS Annual Data Report, attempting to determine the cause of ESRD at the level of the individual patient or at the population level presents considerable difficulties. First, attribution of the cause of ESRD, which must be reported to the USRDS, relies on the physician’s best judgment, and it can be difficult to determine with confidence that hypertension, for example, is the specific cause. This issue is particularly acute when a patient’s condition (such as hypertension or diabetes) rarely results in a renal biopsy, or when a patient has received little pre-ESRD care. Second, presence of a comorbid condition, such as diabetes, does not mean that it is the sole or even major cause of ESRD. Finally, many patients have more than one medical condition that can contribute to ESRD, so choosing a single condition as the primary cause is a somewhat artificial exercise.
With these important caveats, the USRDS Annual Data Report provides information on temporal trends in the major causes of ESRD. As shown in Fig. 21.1 , the adjusted rate of diabetes as a cause of ESRD declined between 2001 and 2014, from 175.2 to 156 per million per year (PMPY), whereas the rate of hypertension was relatively stable (102.3 PMPY, 2001; 100.9 PMPY, 2014). The rate declined substantially for glomerulonephritis (from 38 to 27.2 PMPY), and only slightly for cystic disease. As the overall adjusted rate of ESRD in 2014 was approximately 354 PMPY, diabetes is attributed to 44% of individuals who develop ESRD and hypertension to 29%; the two leading causes therefore contribute to nearly three-quarters of the total.
Although much is known about the causes of ESRD, at least in the United States and other developed countries where standardized forms are completed for new dialysis initiates, information about the indications for dialysis initiation is surprisingly sparse. Data from large registries or other administrative data sets are unlikely to provide granular detail about bedside indications for dialysis, making the generation of broad-based epidemiological conclusions problematic. However, a recent systematically conducted chart review of >450 dialysis initiates from a single center between 2004 and 2012 rendered important insights. Development of uremic symptoms was the largest primary indication for dialysis initiation, at 47%, followed by volume overload/hypertension (21%) and laboratory evidence of abnormalities (18%); the indication was uncertain for 14%. Initiation for volume overload/hypertension was associated with the highest unadjusted risk for subsequent (postinitiation) mortality (hazard ratio [HR], 2.2 compared with the referent group of laboratory abnormalities); HRs were 1.2 for both uremic symptoms and unknown indications. After full adjustment, the magnitude of the HRs was much attenuated and significant differences ablated between the groups, yet the adjusted point for volume overload/hypertension remained notably higher than for the other etiologies. Although 69% of all initiations were judged as being urgent, this was the case for 88% of the patients initiating for volume overload but only 55% of the patients initiating for laboratory abnormalities. With the caveat that the study was a relatively modestly sized sample of US patients, this report appears to be the first to rigorously quantify the percentage distribution of indications for dialysis initiation and their associations with outcomes.
Incidence and Prevalence of End-Stage Renal Disease
The epidemiology of disease is best considered using the familiar concepts of incidence and prevalence. Incidence is classically defined as the new diagnosis or other manifestation of disease in individuals previously without evidence of disease, and prevalence as the number of individuals who have the disease in the general population at any one time. Incidence is typically expressed as the number of new cases divided by person-time at risk, and prevalence as the number of individuals with a disease divided by the number alive at any given point in time (point prevalence) or the number diagnosed within a certain period of time (period prevalence). Prevalence often is expressed as a percentage, and therefore can be considered a cross-sectional perspective of how common a disease is. Prevalence is the product of disease incidence multiplied by disease duration. Disease duration reflects the natural history of the disease, the effectiveness of treatment for it (both the “inherent” effectiveness of treatment and how effectively it is rendered, in real-world settings, to individuals with the disease), and other diseases and factors that contribute to death.
Consideration of ESRD incidence differs from that of many other diseases. As with many medical conditions, ESRD incidence depends in part on the presence and severity of ESRD risk factors (hypertension, diabetes, vascular disease, etc.). However, ESRD is an unusual disorder in that the treatments—dialysis or kidney transplant—often are electively withheld. Information on ESRD incidence therefore reflects not true incidence, but rather treated incidence. In countries where maintenance dialysis therapy is universally available, decisions to forgo dialysis initiation are likely driven primarily by personal choice, comorbidity burden, and a conversation between patient and physician about the utility of dialysis compared with the burden it imposes on the patient. As detailed more fully later, however, in countries where dialysis is a limited healthcare resource, ESRD is infrequently treated. Thus registries that track patients initiating maintenance dialysis or undergoing kidney transplant identify only patients for whom ESRD treatment is offered, not those for whom a decision is made not to offer RRT. Even in developed countries, the number of individuals who develop ESRD and are not offered RRT is uncertain, meaning the true ESRD incidence is unknown.
Incidence of Treated End-Stage Renal Disease
Establishing the incidence and prevalence of treated ESRD is not especially difficult in industrialized countries due to comprehensive registries, typically funded by governments, such as the USRDS, the United Kingdom Renal Registry, the Canadian Organ Replacement Register, the Japanese Society for Dialysis Therapy Renal Data Registry, the Australia and New Zealand Dialysis and Transplant Registry, and others. As the definition of ESRD includes not only patients receiving maintenance dialysis (in-center hemodialysis, home hemodialysis, or peritoneal dialysis [PD]), but also those who have undergone kidney transplant (preemptively or after initiating dialysis), estimates of ESRD are incomplete in the absence of data on transplant recipients.
The USRDS, which contains near-universal identification of patients receiving dialysis or undergoing kidney transplant, provides important information. Even this registry is imperfect, however, as approximately 2% of patients are thought to be missing from it. In particular, early death (i.e., death soon after initiating maintenance dialysis before registration can occur) appears to be missed. An additional source of difficulty for any such registry is ambiguity concerning recovery of renal function (i.e., misclassification of AKI that subsequently recovers as ESRD); however, these challenges likely do not seriously undermine the broad conclusions drawn from research endeavors or, ultimately, the formation of public policy.
Data from the most recent USRDS registry demonstrate that, in the United States, approximately 121,000 patients developed ESRD in 2014, representing an unadjusted rate of 370 persons PMPY and a demographically adjusted rate of 354 PMPY. Of these, slightly >106,000 initiated hemodialysis, >11,000 initiated PD, and slightly >3000 underwent preemptive kidney transplant. Since 2006, when adjusted incidence rates peaked at 387 PMPY, adjusted rates declined to about 354 PMPY over 2012 to 2014 ( Fig. 21.2 ).
Incidence of ESRD varies by demography. Rates are higher for older than for younger individuals: 1556 PMPY, aged ≥75 years; 1265, ages 65 to 74; 573, ages 45 to 64; and 129, ages 22 to 44 years. Rates are much higher for blacks/African Americans (877 PMPY) than for whites (286 PMPY); values are intermediate for Asians (357 PMPY) and Native Americans (333 PMPY). Adjusted incidence rates are substantially higher for individuals of Hispanic ethnicity (466 PMPY) than for non-Hispanics (346 PMPY). Rates are substantially higher for men than for women.
Comprehensive examination of the demography of ESRD incidence is incomplete, however, without an examination of temporal trends. Incidence rates have been declining more for older than for younger individuals ( Fig. 21.3 ); more for members of racial minorities, with little change for whites ( Fig. 21.4 ); and more for Hispanic than for non-Hispanic individuals.
Incidence rates vary substantially by geography, from a low of 250 PMPY in New England (Connecticut, Rhode Island, Massachusetts, New Hampshire, Vermont, and Maine) to 432 PMPY in Texas, a nearly 1.4-fold difference. As these rates are adjusted, demographic case mix alone, at least as determined by traditional variables, cannot account for the difference. Gross geographical location, such as a crude division into north versus south, also cannot fully account for the difference; rates in Pennsylvania and Delaware are 352 PMPY, for example, but 325 PMPY in North Carolina, South Carolina, and Georgia. The reasons for this are not fully understood, but likely relate to socioeconomic, historical, cultural, and perhaps even geographical differences, and to systems of care delivery.
One notable trend in ESRD incidence is the eGFR at which patients initiate dialysis. Since the late 1990s, the mean eGFR at maintenance dialysis initiation has increased. In 2001 for example, approximately 22% of patients initiated dialysis with eGFR <5 mL/min/1.73 m 2 ; this decreased to 14% in 2014. The percentage of patients initiating with eGFR 10 to <15 mL/min/1.73 m 2 increased from 18% to 27%, whereas the percentage initiating with eGFR ≥15 mL/min/1.73 m 2 increased from 8% to 12%. This trend appeared to peak in 2010, and has moderately reversed in the years since ( Fig. 21.5 ), possibly influenced in part by clinical trials and observational studies, as discussed later in this chapter.
Prevalence of ESRD
In contrast to incidence, prevalence counts and rates continue to increase substantially, due primarily to increasing longevity for ESRD patients, especially those receiving maintenance dialysis. Prevalence counts increased from approximately 570,000 in 2009 to 678,000 a mere 5 years later. Adjusted prevalence rates increased from 1598 PMPY in 2001 to 1977 in 2014 ( Fig. 21.6 ), a 19% increase over that period, with year-on-year increases averaging 1% to 2%.
Of prevalent ESRD patients in 2014, approximately 429,000 were receiving maintenance hemodialysis, 47,000 were receiving PD, and 201,000 had undergone kidney transplantation ( Fig. 21.7 ). Prevalence is higher for blacks/African Americans than for whites, for older than for younger individuals, for Hispanics than for non-Hispanics, and for men than for women. As would be expected, prevalence rates vary by geography: Adjusted rates are lowest in the Pacific Northwest (Alaska, Washington, Oregon, Montana, and Wyoming), at 1419 PMPY, and highest in the South Central region (Tennessee, Alabama, and Mississippi), at 2339 PMPY, a roughly 1.5-fold difference.
Temporal trends demonstrate important findings. Between 2001 and 2014, adjusted prevalence rates increased more for older than for younger individuals (from 4284 PMPY to 6243, or about 46%, aged ≥75 years; 2973 PMPY to 3658, or about 23%, ages 45 to 64 years) and for whites than for blacks/African Americans (from 1108 to 1507 PMPY, or 36%, whites; 4839 to 5605 PMPY, or 16%, blacks/African Americans).
Several recent studies have investigated the prevalence of maintenance dialysis worldwide ( Fig. 21.8 ). Because only a small minority of patients with ESRD undergo kidney transplant, especially in developing countries, receipt of maintenance dialysis is generally a good proxy for ESRD, or what might be more precisely termed the need for RRT. One study reported a 70% increase in prevalent dialysis patients between 1990 and 2010 alone. The authors estimated that 284 persons per million population received dialysis as of 2010, or roughly 2 million people worldwide. This estimate was comparable to that of another recent study. Global incidence of ESRD more than doubled during this period, from 44 to 93 per million population, corresponding to roughly 700,000 new cases per year. Interestingly, this increase appears not to be fueled so much by population growth and structure (mainly aging), estimated to have been responsible for about one-third of the increase, or by increases in diabetes or hypertension (two main drivers of ESRD), as by increases in the availability of maintenance dialysis. This would suggest a substantial increase in governmental support of maintenance dialysis programs around the world, including in developing countries, constituting a major welcome development. However, this also indicates a likely huge unmet need for RRT, and as governmental programs expand, many more patients will be recognized as requiring RRT. These authors projected the need for RRT out to 2013, by region ( Fig. 21.9 ). This likelihood is highlighted by estimates of the number of people who die prematurely because they cannot access RRT; using conservative and liberal approaches, between 2.3 and 7.1 premature deaths have likely occurred because patients need maintenance dialysis that has not been provided. In Asia alone, at least 1.9 million people have likely died due to lack of access to RRT. The RRT gap could grow dramatically as China, India, and other nations develop.
Timing of Dialysis Initiation
Timing of dialysis initiation has long been a matter of great uncertainty in nephrology, but work published since 2010 has greatly improved understanding and, it seems, palpable trends recently observable in the epidemiological data. The timing of dialysis initiation typically involves consideration of the symptom burden associated with uremia, fluid status (volume overload) and its consequent hypertension, and laboratory-based values (eGFR usually derived from creatinine, blood urea nitrogen [BUN], potassium, and, perhaps to a lesser degree, other analytes such as albumin and bicarbonate). From 1996 to 2007, the mean eGFR at which patients initiated maintenance dialysis increased from approximately 7.7 to 11.2 mL/min/1.73 m 2 , a trend that does not seem to be accounted for by changes in case-mix, even when markers of frailty or detailed assessments of symptomology are explicitly considered. The potential reasons for this increase are unknown. The widespread introduction of automated eGFR laboratory reporting, an obvious potential explanation, does not appear to be responsible, nor do reimbursement incentives, regional variation in the nephrology workforce, or regional differences in dialysis facility practice patterns.
However, this trend appears to be reversing; as of 2013, the mean eGFR at which patients initiated maintenance dialysis had decreased to 10.3 mL/min/1.73 m 2 . This may be due to a confluence of results from a key clinical trial and multiple observational studies. In 2010 the Initiating Dialysis Early and Late (IDEAL) trial was published. This trial randomized patients to initiate dialysis at 10 to 15 mL/min/1.73 m 2 or much later, at 5 to 7 mL/min/1.73 m 2 . Although the control arm did initiate within the range intended (12 mL/min/1.73 m 2 ), the late-start group initiated at a mean eGFR of 9.8 mL/min/1.73 m 2 ; a major reason appeared to be the perceived threat of hyperkalemia in the patients randomized to start at lower eGFRs. Nevertheless, this seemingly small difference of slightly over 2 mL/min/1.73 m 2 represented a nontrivial difference in start times of fully 6 months between the two groups. The main finding was no difference in all-cause mortality a mean of 3.6 years later, providing strong support for the benefits of initiating dialysis later rather than earlier. The study found no major differences in secondary endpoints such as cardiovascular events or hospitalizations, markers of cardiac structure or function, or overall costs. Gratifyingly, these results are largely concordant with multiple large observational studies, many of which used advanced design and analytical techniques such as stratification, propensity-score matching, inverse probability of treatment weighting, and marginal structural models.
As a result of this evidence, clinical practice guidelines from the United States, Canada, and Europe have become much more favorable toward initiating maintenance dialysis on the basis of signs, symptoms, and laboratory values other than creatinine, BUN, and eGFR, and generally appear to condone later dialysis initiation. Only in the Canadian guidelines is an eGFR threshold for initiation (specifically, 6 mL/min/1.73 m 2 ) explicitly recommended.
Outcomes in End-Stage Renal Disease
Mortality Trends in Patients Receiving Maintenance Dialysis
Since the late 1990s, mortality rates have declined substantially for maintenance dialysis patients and kidney transplant recipients, easily exceeding the HP 2020 goal. Adjusted mortality rates in 2014 for these populations were 166 and 30 per 1000 patient-years, respectively, representing decreases of 32% and 44%, relative to 1996 ( Fig. 21.10 ). Changes in mortality by dialysis modality over several eras are shown in Fig. 21.11 . In all four cohorts, death rates for hemodialysis patients are high immediately after initiation and decline precipitously thereafter, reflecting the fact that the sickest, and likely frailest, patients die soon after initiation. This can be conceptualized as the first stage of a “survivor cohort effect.” Death rates then increase progressively until about 8 years after dialysis initiation, as shown by the oldest two cohorts (1996 and 2001); this will likely be the case for younger cohorts, such as the 2008 cohort, which is likely close to a leveling off in mortality rates. After approximately 8 years, death rates decline and perhaps even show a small annualized decline, likely representing continued survival of only the healthiest or most resilient patients. However, the number of survivors 8 years after dialysis initiation is, unfortunately, only a small fraction of the initiating cohort.
Changes in the unadjusted (crude) death rates are somewhat smaller, primarily because the mean ages of dialysis patients and transplant recipients have increased. The rate of decrease in adjusted mortality has not been monotonic; for hemodialysis patients, for example, adjusted mortality rates declined by 4% from 1996 to 2003, but by fully 24% from 2004 to 2014. Overall, during this period, adjusted mortality rates decreased by 29% in hemodialysis patients and by 44% in PD patients.
Mortality in the first year after maintenance dialysis initiation deserves special scrutiny. As noted, death rates among hemodialysis patients are highest soon after initiation. Because of a marked interaction of mortality with age, it is most useful to consider patients aged ≥65 years separately from those aged <65 years. For example, for patients aged ≥65 years incident to hemodialysis in 2013, the mortality rate was highest in the second month after initiation, at 620 deaths per 1000 patient-years, decreasing to 297 by month 12; corresponding values for patients aged <65 years were 214 and 117. However, the death rate in the first month of dialysis is almost certainly a substantial undercount, as deaths in the first month are likely not fully identified if they occur before patients are registered with the Centers for Medicare & Medicaid Services (CMS) as having ESRD.
The pattern of deaths for PD patients is quite different, as rates increase throughout the first year, from 32 to 54 per 1000 patient-years among patients aged <65 years, and from 115 to 214 for patients aged ≥65 years. These differential patterns of first-year deaths may be attributable to the fact that patients initiating PD are generally healthier than those initiating hemodialysis, and are likely to have greater social support.
Mortality rates also differ by race, with an interaction with age. Death rates are similar between younger whites and blacks/African Americans (aged ≤22 years) but are substantially lower for blacks/African Americans than for whites in older age groups. Regarding age and sex, death rates are lower for men than for women aged ≤44 years, comparable for ages 45 to 64 years, and higher for men than for women aged ≥65 years. These rates are shown in Table 21.1 , which demonstrates the age-×-race and age-×-sex interactions in death rates among prevalent dialysis patients in 2014.
|Age (y) and Race||ESRD||Dialysis||Transplant|
|Age (y) and Sex|
Two related quantities, survival probabilities and expected remaining years of life, may constitute intuitive ways of conceptualizing dialysis survival, at least compared with annualized mortality rates. The associated unadjusted increase in life span can be graphically illustrated, as in Fig. 21.12 , which represents unadjusted increases in life span for prevalent dialysis patients between 2004 and 2010. Life spans increased for all age groups. On a relative scale, 5-year survival probabilities increased by approximately 17% for hemodialysis patients and 31% for PD patients between 2001 and 2009. However, in absolute terms, 5-year survival probabilities remain low, increasing from 36% to only 42% for hemodialysis patients and from 39% to 51% for PD patients over this period. Younger individuals experienced both higher absolute mean increases and higher proportional increases, but this would be expected since they have more potential remaining years of life than older individuals do. However, in absolute terms, the increases have been modest; this is especially concerning regarding younger individuals, who, were it not for ESRD, could expect many decades of remaining life.
Expected remaining years of life for patients receiving maintenance dialysis can be compared with the general population or with relevant disease-specific populations. The former comparison strikingly demonstrates the reality of lower survival among dialysis patients relative to the general population. As expected, survival probabilities are far worse than for age- and sex-matched individuals in the general population; overall 3-year survival is 56% and 67% for hemodialysis and PD patients, respectively, but 92% to 94% for the general population. The difference in expected remaining years of life is >43 years for women aged 20 to 24 years (16.4 vs. 59.5 years), and >6.5 years for men aged 75 to 79 years (3.2 vs. 9.8), to illustrate two possible examples from among all age groups ( Table 21.2 ).
|ESRD patients, 2013|
|General US population, 2013|