We next examine changing patterns of care in the dialysis population to see how they may relate to this minimal change in first-year mortality, and how competing factors may blunt any potential improvement. Nearly 80 percent of prevalent dialysis patients, for instance, received vitamin D in 2005, more than double the level in 1993. Some investigators have suggested that vitamin D use is associated with a protective effect on mortality, yet surveillance data show little change in the death rate, suggesting that other factors may be at work. IV iron use has also increased substantially. Some researchers suggest that iron may be a pro-oxidant, increasing cardiovascular inflammatory markers and adverse events. Possible competing effects between these changes in practice should be considered.

Anemia correction over the last decade has been dramatic, and accompanied by a large increase in epoetin dosing. Both high hemoglobin levels and high ESA dosing have received attention as possibly having an adverse effect on survival. Recent results for randomized clinical trials in the CKD and oncology populations have shown adverse cardiac mortality and morbidity when hemoglobin levels exceed 13 g/dl. These trials and other information have led the FDA to issue alerts to providers on the risks of higher hemoglobin levels, and guidelines to practice within the recommendations of the package inserts. The FDA has also issued a “Black Box” warning for ESAs related to mortality effects of hemoglobins above 13 g/dl, warning providers not to exceed levels over 12 g/dl.

In last year’s ADR we demonstrated the common phenomenon of overshooting target hemoglobin levels. This year, both in this chapter and in Chapters Five, Eight, and Ten, we present trends in overshooting and look at the time patients spend over the FDA-recommended limit of 12 g/dl. Overshooting is now 3–4 times more common than in 1997, and the average time above 12 g/dl has grown from 2.3 months in 1995 to 3.3 months in 2004. For those exceeding 14 g/dl, the time to return to 12 g/dl has grown from 3.0 months to 3.6 months. Providers also appear to be reaching and exceeding the target more quickly. The time to reach higher hemoglobin levels has been reduced by as much as 2–3 months, a particular concern for those with an initial level below 10 g/dl at initiation. And ESA dosing, even that used to reach the previous target of 11 g/dl, has increased by 3–4 times. The lack of progress in first-year mortality rates suggest that safety concerns arising from clinical trials in the CKD and dialysis populations may need to be carefully considered.

Catheter use has been shown to be associated with increased infectious complications and mortality, yet use continues to be high in the U.S. Indications for use typically include the need for emergency dialysis and an inability to establish an internal access. Information from the 2005 revision of the ME form (presented in Chapter Three) shows that 81 percent of incident hemodialysis patients are using a catheter at their first outpatient dialysis session. This may be an important factor in the growing rates of infectious hospitalization and in the lack of progress in the first-year mortality.

We conclude this chapter with new data on mortality in the first months on dialysis, adjusting for information from the Medical Evidence form. Striking here is the lack of progress over the last ten years across almost all categories of age, gender, race, and cause of ESRD. The highest death rates occur in months 2–6 after initiation, corresponding to a period of rising hemoglobin levels and ESA doses, and growing use of IV iron, IV vitamin D, and dialysis catheters.

This new chapter presents only a small part of the data addressing early mortality and the potential effect on survival of anemia correction and other practices. Increased use of all intravenous injectables (some of which have been associated with reduced mortality), and hemoglobin trials showing adverse events, raise questions as to causal relationships in the lack of progress in first-year survival. The very high rate of catheter use is also of great concern. Many investigators have pointed out the late referral of patients to nephrologists; primary care physicians may not realize the implications of an acute dialysis start or of catheter use. Alternatively, emergency need of dialysis may be precipitated by an acute event, requiring the start of dialysis even when planning for an access has been anticipated. Events leading to the initiation of dialysis are not well defined, and require more research. Other countries clearly use catheters at far lower rates, so late referral may be only one element of the issue.

The USRDS will continue to address timely issues related to the public health of those with kidney disease, using both this new chapter and other sections of the Annual Data Report. ½

Treatment of bone & mineral disease & anemia  

Vtamin D hormone is indicated in the treatment of low levels of calcium and elevated PTH levels, and its use in kidney patients promotes normal bone growth. In 2005, just over half of the children on dialysis were treated with IV vitamin D, in contrast to 80 percent in the adult population (Figure ei.2). In the early 1990s and up to the turn of the century, approximately 60 percent of dialysis patients were treated with the IV vitamin D injectable Calcijex. Thereafter, alternative forms of vitamin D such as Zemplar and Hectorol began to find favor among renal practitioners; in 2005, 52.2 percent of patients were treated with Zemplar and 23.8 percent received Hectorol.

IV iron replacement therapy assists in red blood cell production, and is often used in conjunction with epoetin alfa to treat anemia. Eighty-seven percent of all dialysis patients were given some form of IV iron in 2005, though it was used in only 55.8 percent of pediatric patients (Figure ei.3). Until 2000, INFeD was the IV iron of choice, with 68 percent of patients receiving the therapy that year. Starting in 2001, however, Ferrlecit and Venofer began to gain favor, and in 2005, 33.9 and 51.9 percent, respectively, of dialysis patients were treated with these types of IV iron.

 The probability of achieving a hemoglobin of 11 g/dl or greater by the sixth month of ESRD therapy is 0.92 among all patients (Figure ei.4). Not surprisingly, patients who are closest to a hemoglobin level of 11 g/dl at initiation reach this level in a shorter period of time. At month three, for instance, patients with a starting hemoglobin of 10–<11 g/dl have a probability of 0.82 of reaching the target level, compared to 0.69 in those beginning with a hemoglobin less than 7 g/dl.

In 2004, 93.8 percent of patients age 65 and older who began dialysis with a hemoglobin below 11 g/dl reached 11 g/dl or greater by the fourth month of therapy (Figure ei.5). And 82.9, 59.7, and 29.8 percent, respectively, reached levels of 12+, 13+, and 14+ g/dl by month four.

Recent clinical trials (CHOIR, CREATE) have raised safety concerns related to high hemoglobin levels. Among patients starting ESRD therapy and reaching a hemoglobin greater than 12 g/dl, the average months above this level have increased since 1995, but have remained relatively steady since 2000 (Figure ei.6). In 2004, dialysis patients with a level higher than 12 g/dl had an average of 3.3 months above this level, while those with a level of 14 g/dl or greater spent an average of 3.6 months above 12 g/dl.

In 2004, it took just two months for patients to reach a hemoglobin level of 11 g/dl or greater and 2.8, 3.8, and 5.0 months, respectively, to reach levels above 12, 13, and 14 g/dl, illustrating the time-dependent nature of changing levels (Figure ei.7). A patient’s initial hemoglobin, however, does not seem associated with the time to reach a particular level. In patients with an initial level less than 7.0 g/dl, for instance, it takes 3.9 months to reach 13 g/dl or above, while in patients starting at 10–<11 g/dl it takes 3.8 months. Likewise, patients with initial hemoglobins less than 7.0 g/dl take 3.0 months to reach 12 g/dl or greater, compared to 2.7 months in those initiating at 10–<11 g/dl. This may be explained by the higher ESA doses given to patient with lower hemoglobin levels.

As time on dialysis increases from month one to month six, the amount of EPO needed to reach a hemoglobin level of 11 g/dl or greater rises as well (Figure ei.8). In 2004, over 27,000 units of EPO were needed to reach a level of 11 g/dl or greater in month one, while in month six the dose increased by more than 18,000 units to 46,224. This large growth between month one and month six is not as apparent at hemoglobin levels greater than 12, 13, and 14 g/dl, with changes of 27,632 to 33,601 units; 27,100 to 31,369 units; and 26,543 to 30,113 units, respectively.

It is important to note the staggering rise in EPO dosing over the last decade. In 1995, for example, a mean weekly EPO dose of 11,420 units at month one raised hemoglobin levels to 11 g/dl or greater. In 2004, the average dose given to achieve the same level was 27,634 units. It is difficult to explain the rationale for such a significant increase, as it appears that the case mix of the ESRD population has not changed dramatically over this period.

figures ei.2–3 period prevalent dialysis patients who survive on dialysis for the entire calendar year.

figures ei.4–8 Incident dialysis patients, age 65 & older, with a first service date between January 1, 1995, & December 31, 2004, with Medicare as primary payor, & receiving EPO during the first six months after incidence. Initial hemoglobin obtained from Medical Evidence form; patients with a missing initial hemoglobin or an initial hemoglobin greater than 11 g/dl were omitted. Probabilities calculated using the Kaplan-Meier method.

Comorbidity & disease severity in incident hemodialysis patients  

Average body mass index (BMI) has risen nearly 13 percent since 1995 in patients starting ESRD therapy, and at the end of 2006 measured 28.5 kg/m2 (Figure ei.9). In the mid-1990s, higher BMIs were associated with lower estimated glomerular filtration rates (eGFRs), but this phenomena has progressively disappeared, and in 2006 average BMI was similar in patients with eGFRs below 15 and those with rates of 15 or greater—28.6 versus 28.3. BMI appears to have little association with hemoglobin levels, in 2006 ranging from 27.9 kg/m2 in patients with hemoglobins of 12 g/dl and higher to 28.6 in those with levels less than 10. The average BMI in those with five or more comorbid conditions was 28.9 kg/m2 in 2006, compared to 27.5 in those with no comorbidity.

Since 1995, the average eGFR (MDRD equation) in incident patients has risen 40 percent, measuring 10.5 ml/min/1.73 m2 at the end of 2006. This rise may indicate that patients are initiating therapy earlier in their progression to ESRD. Average eGFR is highest in patients with BMIs less than 18.5 kg/m2, at 11.1, and lowest in those with a BMI of 25–<30. Patients with hemoglobins of 12 g/dl or greater had the highest eGFRs in 2006, at 11.3 compared to 9.9, 10.8, and 11.2, respectively, in those with hemoglobins of <10, 10–<11, and 11–<12. Severity of disease appears to play an important role in eGFR levels at initiation. In patients listing five or more comorbidities, eGFR levels measures nearly 12 ml/min 1.73 m2, compared to 9.8 in those listing zero comorbidities­—providing further evidence that patients are carrying a greater burden of morbidity as they initiate therapy.

In 2006, the mean hemoglobin level in patients starting ESRD therapy was 10.2 g/dl, a difference of less than 1.0 g/dl from 1995. Levels vary little by BMI, ranging from 10.0 to 10.2 g/dl across all BMI categories. Higher hemoglobins do appear to have some association with higher eGFRs. In patients with eGFRs of 15–30 ml/min/1.73 m2, for instance, the mean hemoglobin in 2006 measured 10.5 g/dl, compared to 10.1 in patients with rates less than 15 ml/min. In terms of disease severity, hemoglobin levels show little difference, measuring 10.16 g/dl in patients with five or more comorbid conditions and 10.19 in those with no comorbidity.

Patients initiating in 2006 carried an increased burden of disease compared to patients in 1995, with an average of 3.1 comorbidities compared to 1.7. The number of comorbidities is highest in patients with eGFRs of 15–30.

Only one-third of new ESRD patients now begin therapy with a BMI in the normal range of 18.5–<25 kg/m2, compared to 44 percent in 1995 (Figure ei.10). The percentage beginning with a hemoglobin less than 10 g/dl has fallen from 67.5 percent to 48.4. And patients are initiating with more comorbidities, though the spike in 2005 reflects the introduction of the new Medical Evidence form, with additional comorbidity categories.

Congestive heart failure is the most common type of cardiovascular comorbidity in new ESRD patients with diabetes, hypertension, and glomerulonephritis, at 42, 36, and 16 percent, respectively (Figure ei.11).

Comorbidity & disease severity in incident hemodialysis patients age 67+  

On this spread we look further at comorbidity and disease severity in new hemodialysis patients, focusing here on patients who are age 67 and older when they begin therapy. Because Medicare coverage starts at age 65, limiting the data to this cohort allows us to look at the comorbidities and hospitalizations of these patients in the two years prior to their ESRD diagnosis.

The great majority of these new ESRD patients have had five or more comorbid conditions during these two years—88.1 percent in 2005, up from 81.9 percent in 1995 (Figure ei.12). They have also spent considerable time in the hospital. In 2005, for instance, 29.0 percent had been in the hospital for 22 days or more in the two-year period (down slightly from 34 percent in 1995), and 28.1 percent had been admitted at least four times (Figures ei.13–14).

The nature of this population’s comorbidity is illustrated in part in Figure ei.15. In the two years preceding ESRD, 59.6 percent of white patients age 67 and older now carry a diagnosis of diabetes; this rises to 70.9 percent of African Americans, and 73.5 percent of patients of other races. These numbers, moreover, have been rising steadily since 1995. Cardiovascular comorbidity has been growing at a slightly slower pace, but is still on the rise. Among African American patients beginning ESRD therapy in 2005, for example, 56.9 percent had ASHD in the two prior years, up from 47.5 in 1995, while 68.3 percent had congestive heart failure, up from 63.9.

As in the overall incident ESRD population (Figure ei.9, on page 38), the initial eGFR of patients age 67 and older has been growing—40 percent since 1995, to 11.2 ml/min/1.73 m2—suggesting that patients are starting therapy earlier in the course of their disease (Figure ei.16). The rate is greatest in those with five or more comorbid conditions in the two years before ESRD (because of the few patients with 0–2 comorbidities, results for these groups should be interpreted with caution). The number of prior comorbidities, in turn, rises with eGFR. Patients with an eGFR at initiation of less than 15 ml/min/1.73 m2 had, in 2005, an average of 7.2 comorbidities in the previous two years, compared to 8.6 among those with an eGFR of 20 or above.

The odds of having an eGFR at initiation that is greater than the gender-specific mean have changed considerably over time, with a patient initiating in 2005 more than 311 percent more likely than one starting in 1995 (Table ei.a).

Figure ei.17 shows how the increasing complexity of the incident ESRD population is associated with rising eGFR levels at the initiation of therapy. Adjustments for age, gender, and race have little effect on eGFR. Adjustments for the number of comorbidities and for primary diagnosis, however, lower the eGFR, suggesting that eGFR appears to be highly correlated with comorbidity and may reflect severity of disease.

Catheter use & placements at initiation  

In the 2004 ESRD CPM dataset, 57 percent of patients age 75 and older had a catheter as their first vascular access, compared to 47.9 percent of those age 65–74 (Figure ei.18). Despite guidelines recommending the use of fistulas in hemodialysis patients, the percentage of patients dialyzing with a catheter at initiation has grown since 1995, from 65.3 to 73.6 (Figure ei.19). A slightly greater percentage of women than men begin with a catheter, at 76.1 compared to 71.3.

Among patients age 67 and older, catheter use is slightly more common in those not referred to a nephrologist during the 6–24 months prior to initiating therapy—60.7 percent of these patients have a catheter inserted during the two years prior to ESRD, compared to 53.4 percent of those who see a nephrologist (Figure ei.20). Catheters are also, not surprisingly, found more often in patients who begin hemodialysis therapy with an acute start—after being admitted to the hospital from the emergency room (Figure ei.21). In 2005, 84.9 percent of these patients had a catheter, compared to 65.4 percent of those without an acute start.

The percentage of incident patients with a catheter placed during the two years prior to ESRD varies little by primary diagnosis or cardiovascular comorbidity, but does differ dramatically by the number of comorbidities (Figure ei.22). In 2005, for example, only 31.9 percent of patients with one or two comorbidities had a catheter placed during this period, compared to 72.8 percent of those with five or more comorbid conditions.

Table ei.b shows the association of various patient characteristics with the likelihood of having a catheter placed prior to initiation. In 2005, patients with pre-ESRD nephrologist care were 46 percent less likely to have a catheter than were their counterparts not receiving this specialty care. Women were 32 percent more likely than men to have a catheter placed, while patients with congestive heart failure or cancer were 56 and 28 percent, respectively, more likely to have this type of vascular access than those without the diagnoses. And despite recent guidelines, the likelihood of catheter use overall was 35 percent greater in 2005 than in 1996, possibly indicative of disease severity.

Mortality in the first year of ESRD  

Trends in incident mortality rates have been analyzed by the USRDS for a number of years. As presented in the Précis (Figure p.24), rates for hemodialysis patients after the first year of therapy have declined 16–26 percent since 1985. Mortality in the first year, however, has remained relatively stable since 1994.

On this spread we present detailed mortality rates in the first and following months of dialysis treatment, to determine if varying degrees of mortality exist at different times during the initial year of ESRD therapy.

Since 1995, the mortality pattern within each month of the first year demonstrates a similar stability (Table ei.c). Mortality within the first month, for example, has remained relatively constant, at 233 deaths per 1,000 patient years at risk in both 1995 and 2004. A similar pattern is evident for each one- and two-month period examined in this table, and is illustrated in Figure ei.24.

The pattern of mortality across months of the first year of therapy has received little attention, but may provide insight into the changing risk of a dialysis patient during this period. Mortality rates increase 73 percent between the first and second months, from 233 to 403 deaths per 1,000 patient years at risk, and rise 65 percent between the first and third months, to 386. After six months, rates return to levels of the first month, demonstrating a substantial early hazard to incident patients as they stabilize on dialysis therapy.

As noted in Figure 3.5 (Chapter Three), showing that 81 percent of patients are using a catheter at their first outpatient hemodialysis treatment, this changing early hazard is associated with high initial rates of catheter use. Dialysis catheters have been associated with infectious complications, including sepsis (Figure 5.48), which may contribute to higher mortality rates in the first three months.

Alternatively, hemoglobin levels rise during the first months of therapy, from an average of 10.4 g/dl, as reported on the Medical Evidence form, to 12.2 g/dl by month three (Figures 3.8 and 5.4). This can be attributed to use of ESA therapies, with peak doses used in the second month of treatment (Figure 5.5). These high doses may reflect provider interest in correcting anemia to recommended target levels. They may also, however, indicate ESA resistance from infectious complications related to the high rate of catheter use.

Concomitant with the use of ESA therapy to correct anemia is the use of IV iron to correct iron deficiency, providing substrate for red blood cell production. This may, however, contribute to the infectious complications associated with the use of dialysis catheters, and the higher mortality hazard in the early months after initiation of ESRD therapy.

Although first-year mortality by gender and race has been relatively stable since 1993, rates have been rising for patients age 75 and older (Figure ei.23) and for patients with causes of ESRD other than diabetes, hypertension, and glomerulonephritis (Figure ei.26). The increasing mortality in the first months of dialysis therapy in these latter patients may reflect more frequent treatment of individuals with cancer, such as multiple myeloma.

Mortality rates presented here are adjusted only for age, gender, race, and primary cause of ESRD. Information presented here and in earlier ADRs has shown increasing severity of disease in the incident population, which should also be addressed in the adjusted rates. In spite of these limitations, it is clear that, while rates after the first year have been steadily decreasing,  there has been little progress since the early 1990s in reducing early mortality rates in the dialysis population. Detailed investigations are needed to assess the relative impact on mortality of dialysis catheters, anemia correction, iron therapy, and other treatments used in the first months of dialysis therapy.

Mortality rate adjustments  

In the previous spread we investigate trends in mortality during the first year of dialysis therapy, adjusting for basic characteristics of the dialysis population—age, gender, race, and primary cause of ESRD. These adjustments, however, do not take into consideration the growing complexity of the dialysis population.

Here we look at the impact of measures of comorbidity and disease severity, as captured in the revised Medical Evidence form introduced in April, 1995. This form includes information on patient height, weight, comorbidity, hemoglobin level, and serum creatinine. Using data on these measures, we compute, for each incident patient from 1996 to 2004, a body mass index (BMI) and an estimated glomerular filtration rate (eGFR), calculated with a modified MDRD equation. Although there may be limitations in calculating an eGFR in patients approaching dialysis, there is a relationship between increasing comorbidity and higher levels of eGFR at initiation, as shown in Figure ei.16 earlier in this chapter.

In Figure ei.27 we present first-year mortality rates in the hemodialysis and peritoneal dialysis populations, using increasingly complex adjustments. In the first graphs for each modality, for example, we compare unadjusted rates to those with basic adjustments, and also add comorbidity as reported on the Medical Evidence form (adjustment B1). In the hemodialysis population, the basic and B1 rates, as expected, reduce the unadjusted mortality rates, but once again show little change over the last nine years. First-year rates for peritoneal dialysis patients have declined over this period, and adjustments here confirm the growing selection bias reflected by a decreasing patient population.

Corrections for BMI, noted in the second graphs for each modality, change the absolute value of the mortality rates, but not the overall trends over the last nine years. The addition of hemoglobin level has little impact on the rates, while the addition of eGFR changes the absolute value, but not the trends themselves.

The composite adjustments, which add comorbidity, BMI, hemoglobin level, and eGFR, appear to have little impact on adjusted mortality rates during the first year in the hemodialysis population, but in the peritoneal dialysis population dampen the effect of the basic adjustment.

Table ei.d illustrates the impact of each risk factor used in adjusting the mortality rates shown in Figure ei.27. Increasing BMIs, for instance, are associated with lower death rates, reported here as well as by other investigators. Higher hemoglobin levels at initiation have only modest effects on outcomes, compared to higher eGFRs, which are associated with as much as a 50 percent increase in mortality.

This detailed analysis of reported measures of disease severity suggests that factors other than the usual, patient-related conditions traditionally used to adjust mortality rates should be considered in analyses of early mortality. As suggested earlier, issues such as the high use of catheters, rapid anemia correction, and the growing use of IV iron may be more important than previously thought. More complex analytical methods, such as marginal structural models and instrumental variable analyses, may be required to address unmeasured, confounding-by-indication treatments.