Chapter  29:  Use of Epoetin for Anemia in Chronic Renal Failure: Evidence Report/Technology Assessment Number 29

Part I: Adult Patients with CRF

1. Based on evidence from interventional studies, what are the outcomes of maintaining target Hct above 36 percent compared with maintaining target Hct in the 30 to <36 percent range or, more importantly, in the 33 to <36 percent range?

2. Based on evidence from associational studies, what are the outcomes of maintaining target Hct above 36 percent compared with maintaining target Hct in the 30 to <36 percent range or, more importantly, in the 33 to <36 percent range?

Part II: Pediatric Patients with CRF

What is the effect on outcomes of maintaining an Hct above 30 compared with 27 to < 30 percent or, maintaining Hct above 33 compared with 27 to <33 percent?

Part III: Subpopulations of Interest (with or without CRF)

What is the effect on outcomes of maintaining the Hct level >36 percent as compared with 30 to <36 percent in the following patient subgroups (regardless of the presence of renal failure):

(1) Patients who have coronary artery disease,
(2) Patients who have congestive heart failure,
(3) Patients who live at high altitude,
(4) Patients who have arterial occlusive disease,
(5) Patients who have cerebrovascular disorders,
(6) Patients who have obstructive lung disease,
(7) Patients who are in the adolescent age group?

It has been hypothesized that the presence of one of the conditions listed in subgroups 1 through 6 may justify maintaining Hct level above 36 percent to minimize the morbidity associated with the specified condition. For the adolescent subgroup, it is hypothesized that growth and developmental outcomes would be improved by maintaining Hct above 36 percent.

Mortality

Two interventional studies, both full reports, and four cross-sectional analyses, three full reports and one abstract, described mortality results. The first interventional study was an RCT of 1,233 hemodialysis patients with documented congestive heart failure or ischemic cardiac disease. This study analyzed and reported results of both an intention-to-treat analysis and a cross-sectional analysis of the data. The second interventional study was a nonrandomized controlled trial of 115 hemodialysis patients free of comorbidity. The three additional cross-sectional studies, two full reports and one abstract, analyzed large databases of patients on dialysis.

The evidence does not permit a conclusion regarding the effect on mortality of maintaining Hct above 36 percent. To date, a survival advantage has not been demonstrated.

• There was increased mortality among the patients with cardiac disease who were randomized to normal target Hct (42 percent) as compared with low target Hct (30 percent), with 31.6 percent mortality versus 26 percent mortality, respectively (statistical significance not reported). The relative risk for the combined endpoint of mortality and first nonfatal myocardial infarction was 1.3 (95 percent confidence interval [CI] = 0.9 to 1.9) for the normal target Hct group compared with the low target Hct group. This trial was halted even though these results did not achieve statistical significance. It was determined that, even if the trial was completed, the results could not demonstrate a statistically significant benefit for the higher target Hct for the primary endpoint, and there was concern over the potential harm associated with attempting to maintain higher Hct in such patients.

• No deaths were observed in a population of CRF patients free of associated comorbidity who were maintained at Hct above 36 percent. However, the 6-month duration of followup in this study was too short to assess a long-term benefit or adverse effect on mortality.

• Of three large database studies, one study (n=16,153) found a significantly higher mortality among those patients with Hct above 36 percent; one study (n=75,283) found no significant difference in mortality between Hct above 36 percent compared with Hct 30 to <33 percent; and a third study (n=82,879), available only as an abstract, found a significantly reduced mortality rate for Hct above 36 percent.

• A fourth cross-sectional analysis was performed on the data from an RCT of 1,233 patients comparing target Hct of 42 percent with target Hct of 30 percent in hemodialysis patients with documented cardiac disease. So few patients in the 30 percent target Hct group achieved average Hct above 36 percent that none was included in the analysis. The lowest mortality rate was noted among those who achieved an average Hct 39 to 41.9 percent. However, no difference was apparent in mortality between patients who achieve Hct 36 to 38.9 percent compared with those who achieve Hct >30 to <36 percent. Among patients with average achieved Hct in the 30 to 36 percent range, mortality is lower in those patients assigned to the 30 percent target Hct group than those assigned to the 42 percent target Hct group.

Quality of Life

Four interventional studies, one full report and three abstracts, and two cross-sectional studies, both full reports, described quality-of-life findings. All studies suffered from relatively weak design or methodologic flaws that might introduce biases that overestimate effect on quality of life. These data did not provide strong and consistent evidence of a benefit on quality of life of maintaining the Hct above 36 percent compared with Hct >30 to <36 percent.

Hospital Utilization

Two interventional studies, both full reports, and four cross-sectional studies -- two full reports and two abstracts -- described hospital utilization. Overall, these data did not provide strong or consistent evidence of reduced hospital utilization in patients whose Hct is maintained above 36 percent as compared with those maintained at Hct >30 to <36 percent.

• One RCT in 1,233 hemodialysis patients found no significant difference in hospital utilization between patients randomized to either the lower or higher target Hct arms. A pre/post comparison trial in 115 hemodialysis patients found a significant reduction in hospital utilization when Hct was maintained above 36 percent, but the results of this unmasked study using a comparison with historical data prior to study entry must be viewed with caution.

• Of the cross-sectional studies, one full report found that hospital utilization in the Hct 33 to <36 percent group was lower than in either the Hct 30 to <33 percent or the Hct 36 percent groups. Another full report found a slightly lower hospitalization rate among patients with Hct above 36 percent, but no analysis of statistical significance was performed. Finally, two abstracts described favorable findings for the Hct >36 percent group but without analysis of the statistical significance.

Transfusion

One RCT reported utilization of red blood cell transfusion. This trial of 1,233 hemodialysis patients with cardiac disease found a significant reduction in red blood cell transfusion in the normal target Hct group compared with the low target Hct group, 21 percent transfused versus 31 percent transfused, respectively (p<0.001). In this study, the need for transfusion was largely associated with acute blood loss (e.g., from gastrointestinal bleeding or as a result of surgery).

Cardiac Events and Related Intermediate Cardiac Outcomes

Four full reports and three abstracts discussed cardiac outcomes. Two full reports addressed cardiac clinical events, and the remainder addressed cardiac intermediate outcome measures, including left ventricular hypertrophy (LVH) (as assessed by left ventricular mass index) and left ventricular dilatation (LVD) (as assessed by left ventricular cavity volume index).

There is limited evidence on cardiac events, and the available studies are insufficient to draw conclusions. However the evidence available does not demonstrate a reduction in cardiac events when Hct is maintained above 36 percent as compared with 30 to <36 percent.

• One RCT studied 1,233 hemodialysis patients who had documented congestive heart failure or ischemic cardiac disease. In this group of patients, although overall mortality was greater in the higher target Hct arm, no significant difference between the two study arms was seen in fatal or nonfatal cardiac events. A cross-sectional analysis of a Phase IV surveillance study in 324 patients reported zero myocardial infarctions (MI) in the Hct above 36 percent group, one MI in the Hct 33.1 to 36 percent group, and four MIs in the Hct 30 to 33 percent group. However, the number of events is small, and, relative to the Hct above 36 percent group, the number of patient-months of observation is almost three times higher in the Hct 33.1 to 36 percent group and almost seven times higher in the Hct 30 to 33 group. No statistical analysis is reported for these data thus limiting interpretation of the results.

The evidence on cardiac intermediate outcome measures described possible relationships between Hct above 36 percent and reduced left ventricular mass and left ventricular cavity size. Two full-report cross-sectional analyses suggested an association between Hct and cardiac outcome measures. However, data were insufficient from well-designed intervention studies to determine whether raising Hct above 36 from Hct 30 to 36 percent will result in a clinically meaningful improvement in left ventricular hypertrophy.

• A cross-sectional analysis of an observational, longitudinal study in 246 evaluable chronic renal failure patients who were not yet on dialysis was conducted. Because CRF is a progressive condition, Hct decreases over time as renal function deteriorates. Over a 12-month observation period, this study compared the relative changes in mean Hct level between the group of 55 subjects who met the criteria for significant left ventricular growth with the group of 191 subjects who did not demonstrate significant left ventricular growth. A significantly greater drop in Hct was observed in the group that showed significant left ventricular growth. Furthermore, each 1.5 percent decrease in Hct was associated with a 32 percent increased odds of showing significant left ventricular growth (odds ratio [OR]=1.32, 95 percent CI=1.1 to 1.58).

• A separate report by the same first author also reported a significant association between a lower Hct level and the presence of left ventricular hypertrophy (LVH) in predialysis patients.

• An abstract of a Canadian multicenter RCT described results in 125 evaluable hemodialysis patients with asymptomatic LVH or left ventricular dilatation (LVD) who were targeted to either an Hct 39 to 42 percent or Hct 28.5 to 31.5 percent. Changes in left ventricular measurements were compared after 40 weeks. No significant differences were observable in left ventricular mass index. Only one of four analyses reported achieved statistically significant findings. Left ventricular cavity size decreased to a greater extent in patients with LVH who were maintained in the higher target Hct group compared with those maintained at Hct approximately 30 percent (p=0.05).

• An abstract from a crossover study in 11 patients demonstrated statistically significant reductions in left ventricular mass index (p<0.01), left ventricular end-diastolic diameter (p<0.01), but not left ventricular end-systolic diameter.

• An abstract using a pre/post design in 13 patients reported no significant change in left ventricular mass after Hct was maintained above 36 for 4 months.

Exercise Performance

Four small interventional studies, two full reports and two abstracts, presented findings on intermediate outcome measures related to exercise performance. The two full reports included selected patients who were free of significant cardiovascular or musculoskeletal disease; patient selection criteria were not well described in the two abstracts. No cross-sectional analyses addressed this outcome.

The available evidence is suggestive of an improvement on physiologic measures of exercise performance when Hct is maintained above 36 percent compared with Hct 30 to 36 percent. However, whether these improvements in physiologic measures are predictive of clinically significant benefits needs to be established. In addition, such findings would need to be reproduced in large studies or populations more representative of the general population of patients with CRF.

• A full report of a double-masked, crossover study in 14 hemodialysis patients and an abstract of a crossover RCT trial in 27 hemodialysis patients both showed a significant improvement in several physiologic measures of physical work. However, the abstract reported that, depending on the outcome being considered, approximately one-half to two-thirds of the improvement was a result of conditioning and training and one-half to one-third of the total improvement was a result of correction of anemia.

• The two remaining studies, one abstract and one full report, also described favorable results from maintaining Hct above 36 percent. However, the selection of cases in the full report, nonrandomized control comparison study appeared open to substantial bias; and details regarding actual numerical results were very limited in the abstract.

Cognitive Function, Sleep Pattern, and Nutrition

Two full reports -- one interventional study in 20 hemodialysis patients performing neurophysiologic testing and another interventional study in 10 selected hemodialysis patients evaluating sleep patterns -- described testing conducted before and after Hct was raised to approximately 42 percent. A relationship between nutritional status and Hct level was described in an abstract reporting a cross-sectional analysis comparing serum levels of 3 different amino acids among 75 hemodialysis patients on epoetin therapy grouped according to Hct ranges. The average amino acid levels for each Hct group were compared with each other and with healthy control subjects.

Favorable results on physiologic measures of cognitive function, sleep patterns, or nutrition were observed with Hct levels above 36 percent in each of these studies. However, whether these improvements in physiologic measures are predictive of clinically significant benefits needs to be established. In addition, such findings need to be reproduced in large studies or populations more representative of the general population of patients with CRF.

Treatment-Related Morbidity

Eight studies -- four full reports and four abstracts -- described adverse events associated with maintaining Hct above 36 percent compared with Hct >30 to <36 percent. Seven of these were interventional studies, and one full report was a cross-sectional analysis of a Phase IV surveillance study without statistical analysis of results. Three of the four abstracts included fewer than 15 patients.

Blood Pressure

The evidence, derived from four full reports and three abstracts, showed no significant difference in overall blood pressure measurements when Hct was maintained above 36 percent compared with Hct >30 to <36 percent. However, several studies suggested that some patients required intensified medical management to maintain blood pressure control at Hct above 36 percent, but these studies did not permit estimation of the magnitude of the frequency of this occurrence. An RCT in 1,233 hemodialysis patients employed well-designed concurrent controls and reported no significant differences between target Hct groups in the use of various cardiovascular medications. The listed classes of medications included several antihypertensive agents; however, the reported details of data and methods of analysis did not clarify whether medication dosage might have been increased differentially in the high target Hct group.

Vascular Access Thrombosis

Five studies -- two full reports and three abstracts -- reported on vascular access thrombosis with Hct above 36 percent compared with Hct >30 to <36 percent. Targeting Hct at 42 percent significantly increased the rate of vascular access thrombosis in a population of patients with documented cardiac disease compared with targeting Hct at 30 percent. There is insufficient evidence in other groups of hemodialysis patients to determine the effect of Hct above 36 on vascular access thrombosis.

• The best quality evidence was from an RCT that included hemodialysis patients with documented congestive heart failure or ischemic cardiac disease. There was a statistically significant increase in vascular access thrombosis in the normal target Hct group (42 percent) as compared with the low target group (30 percent), 39 versus 29 percent, p=0.001. This group of patients may have been at particularly high risk of vascular access thrombosis because of associated cardiovascular comorbidity.

• Abstracts of two small concurrent control studies in hemodialysis patients reported no significant difference in vascular access thrombosis with Hct of 42 percent as compared with Hct 30 to 31 percent.

• Two pre/post comparisons in hemodialysis patients -- one full report and one abstract -- reported the observed rate of vascular access; however, the lack of a comparison control group limits interpretation of these findings.

Other Treatment-Related Morbidity

The evidence describing other treatment-related morbidities is limited to two full reports and one abstract. Overall, these studies did not suggest any significant increase in other adverse events such as cerebrovascular accident, transient ischemic attack, peripheral gangrene, intestinal ischemia, or seizure with Hct above 36 percent.

3. The evidence is not sufficient to compare the outcomes of maintaining target Hct above 30 percent with outcomes from the 27 to 30 percent range in pediatric patients with CRF.

No studies of pediatric patients with CRF were identified that met the study selection criteria for inclusion in this systematic review. The primary reason for exclusion of most of the studies reviewed in full-text was that the baseline or control Hct with which the normalized group was being compared was below 27 percent. Thus, the available evidence does not address the key question as set out in this systematic review, particularly with regard to outcomes of treatment efficacy.

4. The evidence is not sufficient to determine whether target Hct above 36 percent is more beneficial than target Hct 30 to <36 percent for subpopulations of interest with clinical characteristics thought to warrant Hct above 36 percent.

Patients Who Have Cardiovascular Disease

The most robust evidence on CRF patients with cardiovascular disease was from an RCT of 1,233 hemodialysis patients who had documented ischemic cardiac disease or congestive heart failure. This trial was halted because of increased mortality in the group assigned to 42 percent target Hct when compared with a 30 percent target Hct group. Although the results for a combined Endpoint -- death or first nonfatal MI -- had not achieved statistical significance (relative risk=1.3, 95 percent CI 0.9 to 1.9), it was determined that a benefit could not be demonstrated even if the trial was completed.

A second study provided a cross-sectional analysis in patients without CRF and compared perioperative morbidity and mortality for those with or without cardiovascular disease. No statistically significant difference was seen between patients who have a preoperative Hct above 36 percent compared with those who had an Hct >33 to <36 percent.

Patients Who Have Cerebrovascular Disease

A single cross-sectional analysis in patients without CRF but with cerebrovascular disease reported that, compared with patients with Hct at 30 to 35 percent, cerebral oxygen delivery was greater in those patients who had Hct levels between 40 and 45 percent, but not in those whose Hct was between 35 and 40 percent. No intervention study meeting selection criteria reported outcomes of maintaining Hct above 36 percent in a population with CRF and cerebrovascular disease.

Patients With Other Clinical Characteristics Thought to Warrant Hct Above 36 Percent

There were no studies that met selection criteria for inclusion in this systematic review for patients who live at high altitude, who have obstructive lung disease, or who are in the adolescent age group.

Conclusions

The published literature does not provide strong or consistent support that maintaining Hct above 36 percent is beneficial to patients with CRF. Although limited evidence derived from physiologic measures of functional status or left ventricular mass index suggests improvements in some selected patients, the potential for benefit should be tested in well-designed intervention studies.

Part I: Adult Patients with Chronic Renal Failure

1. Based on evidence from interventional studies, what are the outcomes of maintaining target Hct above 36 percent compared with maintaining target Hct in the 30 to <36 percent range or, more specifically, in the 33 to <36 percent range?

2. Based on evidence from associational studies, what are the outcomes of maintaining target Hct above 36 percent compared with maintaining target Hct in the 30 to <36 percent range or, more specifically, in the 33 to <36 percent range?

Part II: Pediatric Patients with Chronic Renal Failure

What is the effect on outcomes of maintaining an Hct in the following ranges:

• Above 30 compared with 27 to <30 percent, or

• Above 33 compared with 27 to <33 percent?

Part III: Subpopulations of Interest (with or without Chronic Renal Failure)

What is the effect on outcomes of maintaining the Hct level >36 percent as compared with 30 to <36 percent in the following patient subgroups (regardless of the presence of renal failure): (1) patients who have coronary artery disease, (2) patients who have congestive heart failure, (3) patients who live at high altitude, (4) patients who have arterial occlusive disease, (5) patients who have cerebrovascular disorders, (6) patients who have obstructive lung disease, and (7) patients who are in the adolescent age group?

Incidence, Prevalence, Costs, and Burden of Illness

The passage of the 1972 Social Security Amendment (P.L. 92-603) provided for Federal funding of health care coverage for dialysis and renal transplantation. Medicare's ESRD program has continued to be the only nationally funded, disease-specific program in the United States (Frankenfield, Prowant, Flanigan, et al., 1998). During the past 10 years, however, the incidence of ESRD has almost doubled. From 1987 to 1996, the rate grew from 142 per 1 million population to 276 per 1 million population, which translated in real numbers to 34,797 new cases of ESRD in 1987 increasing to 74,116 new cases in 1996 (National Institute of Diabetes and Digestive and Kidney Diseases, 2000). One major component in this increase appears to be an increased incidence of diabetes mellitus, particularly type 2. The ESRD incidence due to diabetes was 113 per 1 million population in 1996, which was a 150 percent increase over the 45 per 1 million population observed in 1987 (National Institute of Diabetes and Digestive and Kidney Diseases, 2000).

According to the U.S. Renal Data System, 335,014 U.S. patients in 1996 either received dialysis or a kidney transplant as life-saving treatment (National Institute of Diabetes and Digestive and Kidney Diseases, 2000). Particular minority populations are disproportionately afflicted with CRF. Compared with Caucasians, African-Americans and Native Americans have 4.5 and 3.7 times higher prevalence, respectively (National Institute of Diabetes and Digestive and Kidney Diseases, 2000). Although the number of cadaveric transplantations has only increased by 20 percent since 1987 (8,500 cases in 1996), the number of patients with ESRD receiving dialysis has doubled, and the number of patients on the transplantation waiting list has nearly tripled (Wolfe, Held, Hulbert-Shearon, et al., 1998). Studies have found lower rates of kidney transplantation and longer waiting times for minorities and women compared with whites and men (National Institute of Diabetes and Digestive and Kidney Diseases, 2000).

As the numbers of patients with ESRD have grown, so have the costs of the program. Medicare spent an estimated $10.96 billion in 1996, which was a 12.4 percent increase from the $9.74 billion spent in 1995. Although the ESRD population represented only 0.6 percent of the total Medicare population in 1994, their expenditures were 5.1 percent of the Medicare budget (National Institute of Diabetes and Digestive and Kidney Diseases, 2000).

Medicare's reimbursement for erythropoietin, which was approved in June 1989, has been in effect for a decade (Collins, Ma, Xia, et al., 1998). During this time, a number of studies have reported that increasing the Hct level from under 30 percent to more than 30 percent has decreased the occurrence of left ventricular hypertrophy (LVH) and has resulted in improved cognitive function, exercise tolerance, and quality of life (Collins, Ma, Xia, et al., 1998). Initially, Medicare's approved reimbursement for erythropoietin was capitated at $40 per administration. Within 3 years of Medicare coverage of erythropoietin treatment, the percentage of those with Hct values under 30 percent had decreased from nearly 60 percent of the patient claims to 40 percent.

Medicare then changed reimbursements to $11 per 1,000 units (U), which was comparable to 2,500 to 2,800 U/administration at the $40 cap. Costs began rising when the dosage pattern started to increase from 2,700 to almost 4,000 U/administration. Medicare responded by reducing reimbursement to $10 per 1,000 U. The agency also started denying payment to patients who had Hct levels over 36.5 percent. By the end of 1997, dosage patterns began to stabilize. Subsequently, Medicare has taken a more flexible stance and has allowed patients to sustain Hct levels up to 37.5 percent, while continuing to track patterns of usage and effects on morbidity and mortality (Collins, Ma, Xia, et al., 1998). However, medical justification (e.g., angina) is required in order to routinely maintain Hct above 36 percent.

Disease Biology, Natural History, Pertinent Clinical Features, Patient Populations, and Settings

Chronic renal failure may result from a variety of medical conditions that produce progressive, irreversible damage to the kidneys. The most common causes for CRF are diabetes mellitus and hypertension; other major causes include glomerulonephritis, polycystic kidney disease, and other urologic conditions (Brenner and Lazarus, 1994). Progressive renal injury follows a variable time course with gradual reduction in renal function often occurring over a period of years (Brenner and Lazarus, 1994). Ultimately CRF may result in ESRD requiring renal replacement therapy (i.e., hemodialysis or peritoneal dialysis) for patient survival.

Patients with CRF (either those receiving dialysis or those who are not yet on dialysis), commonly present with a normocytic, normochromic anemia (Bunn, 1994b; Faulds and Sorkin, 1989). A combination of factors appears to be responsible for this anemia. First, patients have an inappropriately low endogenous erythropoietin production, probably as a result of poorly functioning renal tissue (Bunn, 1994a, 1994b). Also, erythropoiesis appears to be inhibited by the presence of uremic solutes; these solutes also appear to shorten erythrocyte circulation half-life by increasing the osmotic fragility of the cells. The degree of anemia is roughly proportional to the degree of azotemia (Bunn, 1994b). Patients with endstage renal failure may demonstrate hemolysis, possibly because of metabolic or mechanical stresses from renal dialysis (Bunn, 1994b). Also, ferrokinetic studies have shown impaired incorporation of iron into circulating erythrocytes (Bunn, 1994b). Up to 25 percent of patients receiving chronic dialysis can develop a severe, transfusion-dependent anemia (Faulds and Sorkin, 1989).

Endstage renal disease is the final common pathway for a variety of pathologic conditions. Thus the ESRD population comprises a heterogeneous group of patients with a spectrum of clinical comorbidities. The type and severity of associated comorbidity may influence the type and severity of vascular pathology present in patients with ESRD. For example, a patient with ESRD resulting from severe diabetes mellitus would likely have more extensive systemic vascular abnormalities than a patient with ESRD resulting from chronic urinary obstruction or another process that selectively injures the kidneys and causes no systemic injury. Heterogeneity in the clinical spectrum of CRF and ESRD patients is a relevant consideration in analyzing the outcomes associated with increasing the Hct using epoetin.

NKF-DOQI Guidelines

The National Kidney Foundation (NKF) established the Dialysis Outcomes Quality Initiative (DOQI™) in March of 1995. This initiative was developed following a series of conferences occurring in 1993 and 1994 that highlighted controversies in the quality of dialysis care. One major activity of the NKF-DOQI has been the development of evidence-based practice guidelines. The completed NKF-DOQI guidelines were published in the September and October 1997 supplement issues of the American Journal of Kidney Diseases and can also be accessed through the NKF Web site, www.kidney.org. Four topics were identified for initial guideline development: (1) hemodialysis adequacy, (2) peritoneal dialysis adequacy, (3) vascular access, and (4) anemia management. The NKF-DOQI guidelines were developed by four independent Work Groups composed of renal experts from diverse clinical disciplines. The Anemia Work Group comprised physician members representing the nephrology, hematology, and pediatric nephrology specialties and included other medical professionals such as nurses and pharmacists. The anemia guideline includes 28 separate recommendations addressing seven different management dimensions: anemia workup, target Hct/hemoglobin, iron support, administration of epoetin, inadequate epoetin response, role of red blood cell transfusion, and possible adverse effects related to epoetin therapy.

The NKF-DOQI guideline report on anemia provides a thorough review of the scientific evidence and expert opinion describing how epoetin should be administered to CRF patients as well as when adjuvant therapies may be necessary to optimize treatment response. Based on a review of the evidence, the NKF-DOQI work group concluded that target Hct should be 33 to 36 percent.

The NKF-DOQI guidelines recommend that iron status should be monitored and supplemental iron is necessary in order to maintain iron stores in a range that supports optimal erythropoiesis. Although oral iron may be administered initially, the work group noted that many CRF patients may be unable to maintain sufficient iron stores using oral iron alone. In most hemodialysis patients, ongoing blood loss necessitates the use of intravenously administered iron supplementation. Intravenous iron has much greater bioavailability than orally administered iron, and the evidence reviewed by the NKF-DOQI report showed that changing from oral to intravenous iron can result in increased Hct while using smaller amounts of epoetin.

The route of administration for epoetin appears to influence the efficiency in terms of the total dose of epoetin necessary to achieve and maintain a desired Hct level. Many studies have examined the effect of using subcutaneous versus intravenous epoetin, and the NKF-DOQI work group concluded that the evidence supported a recommendation to use subcutaneous epoetin as the preferred route of administration because of its increased efficiency. Because epoetin is more slowly absorbed when administered subcutaneously, elevated serum levels of epoetin are sustained over a longer period of time. The resulting effect is that target Hct level may be maintained using approximately 15 to 50 percent lower doses of epoetin when subcutaneous epoetin is substituted for intravenous epoetin.

Canadian Evidence-Based Recommendations

In 1995, a set of evidence-based recommendations for the clinical use of epoetin were published (Muirhead, Bargman, Burgess, et al., 1995). Over 200 scientific papers were reviewed by a group of Canadian medical professionals in the process of developing these recommendations that address five major questions: (1) Who should receive epoetin? (2) What should the target Hct be? (3) What is the best route of epoetin administration? (4) How should iron status be evaluated and managed? and (5) How should patients be monitored and followed up? The Canadian recommendations are similar with regard to route of epoetin administration and management of iron status and supplementation. However, the recommended target Hct for hemodialysis patients was 31.5 to 34.5 percent in the Canadian review, which is slightly lower than the target Hct recommended in the NKF-DOQI report.

European Best Practice Guidelines

The European renal community has established a Working Party including representatives from the European Renal Association/European Dialysis and Transplantation Association (ERA-EDTA) as well as from a cross-section of the national nephrology societies from European countries. The European Best Practice Guidelines for the Management of Anaemia in Patients with Chronic Renal Failure were published in 1999 (Working Party for European Best Practice Guidelines for the Management of Anaemia in Patients with Chronic Renal Failure, 1999). The published guidelines include 28 separate sections covering a broad range of issues relating to anemia in CRF patients. The specific guideline addressing target hemoglobin concentration recommends that Hct should be maintained above 33 percent; however, the guideline does not specify an upper limit for target Hct and suggests that such targets should be individualized (Working Party for European Best Practice Guidelines for the Management of Anaemia in Patients with Chronic Renal Failure, 1999).

Unpublished International Clinical Trials

During the course of conducting this systematic review, the EPC team followed the status of two additional international studies that evaluated the effects of increasing Hct above 36 percent in patients with CRF (Jacobs, 1999): (1) a multicenter randomized controlled trial (RCT) conducted in Canada by Foley, Parfrey, Morgan, et al. (1998) comparing normalization of Hct (40.5 percent target) with partial correction of anemia (30 percent target) in 146 hemodialysis patients with asymptomatic cardiomyopathy; and (2) a multicenter RCT conducted in Scandinavia comparing normalized Hct (target of 43.5 to 48 percent in males, 40.5 to 45 percent in females) with partial correction of anemia (target 27 to 36 percent) in 416 patients with CRF (293 hemodialysis, 51 continuous ambulatory peritoneal dialysis, and 72 predialysis) ("Scandinavian Multicentre Trial," see description in Jacobs, 1999). At the time this report was completed, results of the Canadian study had been published in two abstracts (Foley, Parfrey, Morgan, et al. 1998; Wells, Coyle, Lee, et al. 1998), and these findings are described in this systematic review. No results of the Scandinavian study were available.

Types of Participants

The primary population of interest for this systematic review is patients with CRF who have anemia secondary to the renal disease. The report includes studies addressing CRF patients of all ages and divides the discussion of Results and Conclusions into two parts: Part I includes adult patients with CRF, and Part II includes pediatric patients with CRF. Chronic renal failure includes the full spectrum of disease severity from those patients on hemodialysis or peritoneal dialysis to those patients with progressive CRF but not yet receiving dialysis at the time of treatment for anemia.

Other subpopulations of interest for this systematic review (Results and Conclusions, Part III,) include patients who have at least one of a predefined set of clinical characteristics thought to warrant maintaining Hct >36 percent. These predefined clinical characteristics included coronary artery disease, congestive heart failure, arterial occlusive disease, cerebrovascular disease, obstructive lung disease, patients living at high altitude, and patients in the adolescent age group. Maintaining Hct >36 in the adolescent age group might favorably influence growth and development. Living at high altitude exposes the patient to oxygen at lower levels in the air. Normal subjects adapt to living at high altitude by increasing their Hct level; however, patients with ESRD are unable to respond appropriately because of deficient erythropoietin production. In each of the other clinical circumstances, maintaining Hct >36 is postulated to alleviate symptoms from these comorbidities.

Studies of patients with primary hemoglobinopathies (e.g., sickle cell disease or thalassemia) were excluded because of the potential for outcomes associated with the primary disease to confound the analysis of outcomes resulting from differences in Hct level. These subjects were not required to have anemia secondary to renal disease.

Types of Studies

Studies were required to report a controlled comparison of a health outcome (or appropriate intermediate outcome) for the included patient groups at the relevant levels of Hct defined for each key question. Controlled comparisons could be derived from clinical studies using randomized or nonrandomized control designs. Studies reporting cross-sectional analyses describing the relationship between outcome and Hct level that mathematically controlled for confounding factors were included as well.

Hematocrit Range Being Studied⁴

The primary focus of this systematic review was to compare the outcomes of maintaining the Hct >36 percent as compared with 33 to <36 percent. However, evidence derived from studies comparing outcomes of maintaining Hct >36 percent with outcomes of maintaining Hct in the range of 30 to <36 percent was also included in this systematic review because of the small number of studies addressing the primary comparison of interest.

Because the normal range of Hct in the preadolescent population⁵ is lower than that for the postadolescent population, the primary Hct range of interest was different for the pediatric population. In the pediatric population of preadolescent patients, the relevant comparisons were Hct >30 compared with Hct 27 to <30 or Hct >33 percent as compared with Hct 27 to <33 percent.

Sample Size

Studies were excluded if there were data on fewer than 10 patients in each group being compared. Since the prevalence of chronic renal disease is high, the requirement of at least 10 patients per analyzed group was considered reasonable by the various experts consulted in developing this protocol. Information derived from studies with fewer than 10 patients per analyzed group was considered of negligible statistical and clinical value and was not considered to be an adequate basis for informed decisionmaking in managing anemia in this clinical setting.

Type of Outcome Reported

In this systematic review, the EPC team was primarily interested in outcomes that would be considered health outcomes, those that are readily recognizable to patients as being meaningful in their lives. For example, health outcomes would include death rate or myocardial infarction (MI) rate. However, in the interest of completeness, the team considered a variety of intermediate outcomes. For an intermediate outcome to be relevant for this systematic review, the intermediate outcome measure must be known or believed to be predictive of a health outcome.

Studies were included if they reported at least one of the following outcomes. Results for each outcome were analyzed separately in the evidence tables:

• Mortality.

• Quality of life.

• Hospital utilization.

• Red blood cell transfusions.

• Cardiac outcomes (including clinical cardiac events as well as intermediate outcomes such as echocardiographic measures of left ventricular size).

• Functional status (including physiologic measures related to various aspects of functional status such as physical performance, cognitive performance, sleep quality, or nutritional status).

• Adverse effects.

• Growth and development (pediatrics).

• School-related performance (pediatrics and adolescents).

Stage 1: Screening Titles and Abstracts

Initially, one of two independent reviewers evaluated each title (and abstract if available) against the predefined screening criteria. References were grouped into two categories: "retrieve" or "hold." Those references labeled as "hold" by the first reviewer were also evaluated by the second reviewer. Full-text copies of articles were obtained if either reviewer labeled the reference "retrieve."

The screening criteria used in Stage 1 were designed to be as inclusive as possible. In other words, articles were selected for further review in full text unless it was clear from the information provided in the title or abstract that it did not relate to the scope of this review.

If insufficient information was provided in the abstract to determine the study's potential relevance, the study was selected for further review. References that only had titles available were marked for retrieval unless the title clearly identified the article as outside the scope of this review.

Studies were rejected from further review if any of the following criteria were met:

• The number of subjects in each group being compared was clearly stated to be fewer than 10 per group.

• The target or achieved Hct was clearly stated to be less than 36 percent (hemoglobin [Hb] 12 g/dL) for studies of adolescent or adult patients or less than 30 percent (Hb 10 g/dL) for studies of pediatric patients.

• The abstract clearly demonstrates that the study will not include an analysis of the relationship between Hct level (or hemoglobin level) and a health outcome or a relevant intermediate outcome.

• The patient population did not have CRF or one of the prespecified clinical characteristics listed in the key question for Part III.

• The patients in the study were selected for having a primary hemoglobinopathy (e.g., sickle cell disease or thalassemia).

The screening process produced 1,144 articles to be reviewed in full text.

Stage 2: Review and Categorization of Articles in Full Text

The full-text articles were categorized as follows: (1) potential data article, (2) review/editorial/letter, or (3) non-English article. Approximately 200 non-English articles were identified from the articles retrieved in full text. After discussion with the TAG, external experts, and the Task Order Officer, it was concluded that the non-English literature was unlikely to contain data relevant to the key issues that was not reported elsewhere within the English literature. Thus, the decision was made to exclude non-English articles from further review. Approximately 550 publications were identified as potential data articles, and 360 as review/editorial/letter. Articles identified as potential data articles were further evaluated in Stage 3.

Stage 3: Evaluation of Potential Data Articles and Identifying Studies Eligible for Analysis of Evidence

Table 2. Reasons for study exclusion

Table 2. Reasons for study exclusion

Reason for Exclusion	Criterion
Not in English	Full-length journal article was not published in the English language.
No primary data	Review article, editorial, commentary, letter, or position statement that did not report primary outcomes data from a clinical trial or cross-sectional analysis.
Wrong Hct range	The Hct range being studied did not correlate with the Hct ranges prespecified in the key questions.
No relevant outcome	The results of the study did not include information on health outcomes or relevant intermediate outcome.
Hct and outcomes not related	The outcomes reported in the article were not reported in relation to the Hct level of patients in the study.
N <10 per study arm	One or more of the groups being analyzed included fewer than 10 patients.
Wrong patient population	Population did not have renal disease (Parts I and II). Population did not have one of the prespecified clinical characteristics (Part III). Population included patients with primary hemoglobinopathy (e.g., sickle cell anemia, thalassemia).
No control comparison	The study did not include a control group or make a controlled comparison with patient data at an appropriate Hct level.

N=number of patients.

Key Questions

1. Based on evidence from interventional studies, what are the outcomes of maintaining target Hct above 36 percent compared with maintaining target Hct in the 30 to <36 percent range or, more specifically, in the 33 to <36 percent range?

2. Based on evidence from associational studies, what are the outcomes of maintaining target Hct above 36 percent compared with maintaining target Hct in the 30 to <36 percent range or, more specifically, in the 33 to <36 percent range?

Overview of the Evidence

The literature search and selection process identified 13 full reports and 10 scientific abstracts that reported findings on the relationship between outcomes associated with an Hct level above 36 percent and outcomes associated with an Hct level between 30 and 36 percent in patients with CRF (see Evidence Tables 1 through 3). Some of these reports originate from the same authors and/or institutions, and the degree of overlap of patients across separate reports cannot be determined from the information provided.⁷

Quality of Study Design and Conduct

The study selection criteria for this systematic review required that studies use some form of control in conducting the analysis; however, multiple methodologies and study designs were considered to meet this criterion. The primary assessment of study quality that was used in this systematic review is based on study design with a three-tier hierarchy of quality. Accordingly, each section regarding study quality is subdivided into three categories (listed in order of strongest to weakest design): RCT, nonrandomized controlled clinical trial, or cross-sectional analysis. The majority of nonrandomized controlled trials were single-arm intervention trials, where patients served as their own controls and outcome measures at baseline Hct levels were compared with those observed after raising the Hct level above 36 percent (pre/postdesign). Discussion of results within each category of study design addresses both results described in full reports and results described in abstracts. Results from full reports are viewed with greater confidence than are those from abstracts presented at scientific meetings.

Table 3. Overview of study quality and relevance: Results (more...)

Table 3. Overview of study quality and relevance: Results and conclusions, Part I

Study	N	Better Studies Are Checked		Worse Studies Are Checked		Method of Case-Mix Adjustment
		Uses Double-Masked Design	Compares Hct >36% with 33-36%	Loss to Followup or Missing Data for >10%	Potential Bias in Control Group Comparison
Randomized controlled clinical trial
Besarab, Bolton, Browne, et al., 1998	1,233					N/A
McMahon, McKenna, Sangkabutra, et al., 1999	14					N/A
Abstract Foley, Parfrey, Morgan, et al., 1998	146			? Not reported		N/A
Abstract Wells Coyley, Lee, et al., 1998	130			? Not reported		N/A
Abstract Stray-Gundersen, Sams, Gookin, et al., 1997	27					N/A
Nonrandomized controlled clinical trial
Moreno, Sanz-Guajardo, Lopez-Gomez, et al., 2000	115				Hospital utilization Quality of life	N/A
Pickett, Theberge, Brown, et al., 1999	20					N/A
Benz, Pressman, Hovick, et al., 1999	10					N/A
Suzuki, Tsutsui, Yokoyama, et al., 1995	10 cases; 39 control				Exercise Performance	N/A
Abstract Mason, Skinner, Sangkabutra, et. al., 1997	10	?				N/A
Abstract Mason and McMahon 1997	10	?			Quality of life	N/A
Abstract Eschbach, Glenny, Robertson, et al., 1993	10-13			Quality of life	Quality of life	N/A
Cross-sectional analysis
Besarab, Bolton, Browne, et. al., 1998	1,233				Quality of life	N/A
Levin, Thompson, Ethier, et. al., 1999	246		N/A			Age, gender, race, underlying renal disease, blood pressure, renal function, laboratory values, and hemoglobin level
Levin, Singer, Thompson, et al., 1996	175	N/A	N/A		N/A	None
Lowrie, Huang, Lew, et al., 1994	16,153	N/A		Not reported	N/A	Case mix: cause of renal failure Lab values
Ma, Ebben, Xia, et. al., 1999	75,283	N/A		Not reported	N/A	Age, gender, race, prior dialysis exposure, comorbidity, primary renal diagnosis, and severity of disease (# vascular access procedures, # blood transfusions, hospital length of stay)
Xia, Ebben, Ma, et. al., 1999	71,717	N/A		Not reported	N/A	Age, gender, race, prior dialysis exposure, comorbidity, presence of diabetes, and severity of disease (# vascular access procedures, # blood transfusions, hospital length of stay)
Levin, Lazarus, and Nissenson, 1993	324	N/A			N/A	None
Moreno, Lopezgomez, Sanz-Guajardo, et al., 1996	1,013	N/A			N/A	Age, gender, comorbidity index, presence of diabetes, socioeconomic level, educational level
Abstract Collins, Hao, Ebben, et. al., 1990a	6,650	N/A				Age, gender, race, 10 comorbid conditions, 3 severity of disease measures, diabetes
Abstract Collins, Hao, Ebben, et al., 1999b	85,473	N/A				Age, gender, race, 10 comorbid conditions, 3 severity of disease measures, diabetes
Abstract Collins, Hao, Ebben, et. al., 1999c	82,879	N/A				Age, gender, race, renal diagnosis, and entry year
Abstract Riedel, Hampl, Nundel, et. al., 1996	75	N/A		Not reported	N/A	Not reported

N = number of patients; N/A = not available.

The quality and relevance of each study included in this systematic review is further described along five dimensions (Table 3). For the purpose of this report, the first two dimensions serve to define better studies (uses double-masked design, or compares Hct above 36 percent with 33 to < 36 percent⁸); whereas the second two dimensions serve to define worse studies (loss to followup or missing data for >10 percent, or potential bias in control group comparison). Finally, the extent to which each cross-sectional analysis study identified and controlled for potential confounding influences in the data analysis is described under the method of case-mix adjustment.

Outcomes

Table 4. Overview of outcomes reported in each study: (more...)

Table 4. Overview of outcomes reported in each study: Results and conclusions, Part I

Author(s) Year	Sample Size	Study Design	Efficacy Outcomes						Treatment-Related Morbidity
			Mortality	Quality of Life	Hospital Utilization	RBC Transfusion	Cardiac	Other	Increased Blood Pressure	Vascular Access Thrombosis	Other
Besarab, Bolton, Browne, et al., 1998	1,233	Randomized, prospective, open-label trial
McMahon, McKenna, Sangkabutra, et al., 1999	14	Randomized, double-blind, crossover study						Exercise performance
Abstract Foley, Parfrey, Morgan, et al., 1998	146	RCT (Canadian Study)
Abstract Wells, Coyle, Lee, et al., 1998	130	RCT (Canadian Study)
Abstract Stray-Gundersen, Sams, Goodkin, et al., 1997	27	RCT						Exercise Performance
Abstract Mason Skinner, Sangkabutra, et al., 1997	10	Blinded crossover study
Abstract Mason and McMohan, 1997	10	Blinded crossover study								?
Moreno, Sanz-Guajardo, Lopez-Gomex, et al., 2000	115	Pre/post
Pickett, Theberge, Brown, et al., 1999	20	Pre/post						Cognitive function physiologic testing
Benz, Pressman, Hovick, et al., 1999	10	Pre/post						Sleep pattern
Abstract Eschbach, Glenny, Robertson, et al., 1993	10-13	Pre/post						Exercise performance
Suzuki, Tsutsui, Yokoyama, et al., 1995	10 cases; 39 controls	Non-randomized open label study with historical control group						Exercise performance
Besarab, Bolton, Browne, et al., 1998	1,233	Cross-sectional analysis
Levin, Thompson, Ethier, et al., 1999	246	Cross-sectional analysis
Levin, Singer, Thompson, et al., 1996	175	Cross-sectional analysis
Lowrie, Huang, Lew, et al., 1994	16,153	Cross-sectional analysis
Ma, Ebben, Xia, et al., 1999	75,283	Cross-sectional analysis
Xia, Ebben, Ma, et al., 1999	71,717	Cross-sectional analysis
Levin, Lazarus, and Nissenson, 1993	324	Cross-sectional analysis
Moreno, Lopez-Gomez, Sanz-Guajardo, et al., 1996	1,013	Cross-sectional analysis
Abstract Collins, Hao, Ebben, et al., 1999a	6,650	Cross-sectional analysis
Abstract Collins, Hao, Ebben, et al., 1999b	85,473	Cross-sectional analysis
Abstract Collins, Hao, Ebben, et al., 1999c	82,879	Cross-sectional analysis
Abstract Riedel, Hampl, Nundel, et al., 1996	75	Cross-sectional analysis						Nutritional status

RBC = red blood cell; RCT = randomized controlled trial.

The outcomes identified for this report include mortality, quality of life, hospital utilization, red blood cell transfusion, treatment-related morbidity, and cardiac events as well as cardiac-related intermediate outcomes, including left ventricular mass index(LVMI) as a measure of LVH and left ventricular cavity volume as a measure of left ventricular dilatation (LVD). Several additional intermediate outcomes were identified including physiologic measures of exercise performance, physiologic measures of cognitive function, physiologic measures of sleep quality, and serum levels of amino acids as a measure of nutritional status. An overview of which studies report each of the various outcome measures is provided in Table 4.

Results

Mortality

Overview

Four full reports and one abstract described the relationship between mortality and maintaining an Hct above 36 percent compared with maintaining an Hct > 30 and <36 percent (Besarab, Bolton, Browne, et al., 1998; Collins, Hao, Ebben, et al., 1999c [abstract]; Lowrie, Huang, Lew, et al., 1994; Ma, Ebben, Xia, et al., 1999; Moreno, Sanz-Guajardo, Lopez-Gomez, et al., 2000). The overall quality of the evidence was strengthened by the inclusion of a large RCT although the remaining studies were weak in design for evaluating effect on mortality. However, only the three cross-sectional analyses actually provided evidence relating Hct above 36 percent compared with Hct 33 to <36 percent, the primary Hct ranges of interest in this systematic review.

Evidence from randomized controlled clinical trials

Besarab, Bolton, Browne, et al. (1998; n=1,233) was the largest, prospective, RCT designed to examine the outcomes associated with targeting the Hct to 42 percent ("normal target group") compared with targeting the Hct at 30 percent ("low target group") in long-term hemodialysis patients with documented cardiac disease (congestive heart failure or ischemic heart disease). The primary endpoint was combined death or first nonfatal MI. The report compared overall mortality as well as cause-specific mortality between the two study groups on an intent-to-treat basis and reported results of a cross-sectional analysis of the data as well.

At the third interim data analysis, higher mortality was observed in the normal target group, although this difference was not statistically significant. Nevertheless, the independent data monitoring committee recommended that the study be halted after 29 months because, in light of the increased death rate in the normal target group, the trial could not demonstrate a benefit for the normal target Hct group.

The primary analyses were performed on an intention-to-treat basis so that all 1,233 patients were included in the analysis with median followup of 14 months. When the study was halted (29 months), the normal target group had 183 deaths and 19 first nonfatal MIs whereas the low target group had 150 deaths and 14 first nonfatal MIs. These differences did not reach statistical significance and correspond to a risk ratio of 1.3 and 95 percent confidence interval (CI) of 0.9 to 1.9. Additional analyses adjusting for a variety of prespecified baseline characteristics did not affect these results. Overall mortality at 1 and 2 years was reported to be 7 percentage points higher in the normal target group compared with that in the low target group. Data presented in Evidence Table 4 showed slightly higher rates of infectious mortality (5.2 percent compared with 3.6 percent) and cardiac mortality (20.2 percent compared with 18.2 percent) in the normal target group compared with those in the low target group, but statistical significance of these comparisons was not provided in the article.

Besarab, Bolton, Browne, et al. (1998) also reported a cross-sectional analysis of mortality based on the average Hct value achieved for each patient over the course of the study regardless of target Hct group assignment. Average Hct was calculated using all Hct values for each patient up to the point of death, loss to followup, or March 31, 1996. Each patient was categorized into one of five ranges for average Hct: 27 to 29.9; 30 to 32.9; 33 to 35.9; 36 to 38.9; or 39 to 41.9 percent.

Figure 1. Mean Mortality Rate as a Function of the (more...)

   Figure 1. Mean Mortality Rate as a Function of the Average Hematocrit Value

(Adapted from Besarab, Bolton, Browne, et al., 1998 (Copyright (c) 1998 Massachusetts Medical Society. All rights reserved.)

Figure 1 provides a comparison of the observed death rate versus the average Hct category, stratified by assigned target group.

In the study by Besarab and coworkers (1998) there was an insufficient number of patients randomized to the low target group that maintained an average Hct above 36 percent to allow comparison between study groups at the higher levels of Hct. The lowest mortality was seen in those who achieved Hct 39 to 41.9 percent. However, there was no difference in mortality between patients who achieved Hct 36 to 38.9 compared with those who achieved Hct 33 to 35.9. When the normal target group was compared with the low target group within the same range of achieved Hct, mortality was actually lower in the low target group. Moreover, mortality in the low target group patients who achieved an average Hct of 33 to 35.9 percent was only slightly higher than that for those normal target level patients who achieved Hct 39 to 41.9 percent, roughly 19 percent versus 14 percent, respectively.

It is notable that the group of patients in the normal target group that maintained an average Hct of 39 to 41.9 percent had the lowest mortality; however, multiple factors should be considered in interpreting this finding. First, a favorable patient selection process may account for this finding (i.e., those patients able to achieve and maintain a normal Hct level may tend to be the healthiest patients to begin with and those with the lowest a priori likelihood of death independent of their average Hct level). Second, Besarab and colleagues (1998) noted that, at least in some patients, "the Hct fell considerably before they died." Study therapy was prematurely terminated in some patients weeks or even months before they died, which could affect the calculated average Hct category to which the patient was assigned for this analysis. Thus, a patient who ultimately died during the study may have had a period of time before death with a relatively low Hct that would serve to lower his average Hct. Third, intercurrent illness, particularly infection, may lower the achieved Hct during the time of illness because of development of resistance to epoetin. Since patients who died during the study may reasonably be expected to have a higher frequency of such illness, the effect would be to further depress the average Hct in such patients. Each of these factors would serve to categorize sicker patients into a lower Hct group, leaving only relatively healthy and robust patients surviving in the highest average Hct group.

A Cox regression analysis was performed on the primary endpoint (combined endpoint of death or nonfatal MI) using the average Hct instead of randomized group assignment in the model. This model adjusted for the 11 prespecified baseline conditions including: age, gender, race, adequacy of dialysis, type of vascular access, hypertension, diabetes mellitus, peripheral vascular disease, congestive heart failure, ischemic heart disease, and New York Heart Association class III cardiac disability. Based on this analysis, the risk ratio associated with Hct was 0.7 (95 percent CI, 0.6 to 0.8 p<0.001). That is, a 10 point higher Hct was associated with a 30 percent lower risk of death or MI. However, this type of cross-sectional analysis adjusts for some but not all of the differences in patient characteristics that may influence the primary study endpoint, and the potential for additional factors not included in the model to confound the analysis remains.

Interpretation of this analysis raises several questions. The full report did not specify the manner in which average Hct was modeled. It appears that the model may have assumed a linear relationship between average Hct and the primary endpoint; however, the data presented in Figure 1 do not clearly support using a linear assumption. Moreover, the analysis assumes that a single Hct value for each patient (calculated by averaging all Hct values over the entire study period) accurately describes the relationship between Hct and mortality. The actual Hct values observed over time may not be well represented by this average value.⁹

Analyzing the results using this single average Hct value might obscure a potential relationship between actual Hct and outcome. Personal communication with one of the coauthors (Goodkin D, January 2000) revealed that multiple additional unpublished analyses were performed by the investigators' modeling Hct in numerous ways. Reportedly, none of these analyses demonstrated a relationship between higher achieved Hct level and increased mortality. If this is the case, then having a higher Hct per se may not be toxic, but some other factor associated with trying to normalize Hct could be causing the increased morbidity and mortality in these cardiac patients.

Besarab, Bolton, Browne, et al. (1998) noted that there were differences between the two treatment groups other than the assigned target Hct level. Two particular factors distinguishing the two treatment groups were discussed. These two factors were a higher use of intravenous iron dextran and lower values of Kt/V, a measure of the adequacy of dialysis, in the normal target group. The authors discuss several theories on how intravenous iron dextran may be related to the unfavorable outcomes associated with the higher Hct target observed in this study. Iron may be implicated in free-radical generation, damage of the myocardium, worse cardiac outcomes in men, and a greater predisposition to infection in hemodialysis patients.

Evidence from nonrandomized controlled clinical trials

Moreno, Sanz-Guajardo, Lopez-Gomez, et al. (2000) studied 156 selected hemodialysis patients who were observed over a 6-month period after Hct was raised from a baseline mean ± standard deviation (SD) of 31 ± 2 percent up to 38.5 ± 2.5 percent. The authors reported no deaths observed in this cohort during the study. However, it is important to acknowledge the stringent selection criteria that were used to enroll patients into this study when interpreting this favorable observation. These selection criteria are detailed in the study description table (Evidence Table 2), but briefly, patients with diabetes, uncontrolled hypertension, or history of severe comorbidity were excluded. This type of cohort was quite different and considerably more healthy than the patients enrolled in the study by Besarab and colleagues (1998). In addition, the length of followup in the study by Moreno and coworkers (6 months) was much shorter than that in the study by Besarab and coworkers (median 14 months), and a 6-month followup is probably too short to assess long-term benefits or adverse effects.

Evidence from cross-sectional analyses

Ma, Ebben, Xia, et al. (1999) performed a cross-sectional analysis on 1993-94 administrative data from the Health Care Financing Administration ESRD Program Management and Medical Information System (PMMIS). The analysis included 75,283 hemodialysis patients and categorized patients into the following Hct ranges: <27, 27 to <30, 30 to <33, 33 to <36, and above 36 percent. All-cause mortality was significantly decreased in the 33 to <36 percent group compared with that in the reference group, 30 to <33 percent (relative risk [RR]=0.9, 95 percent CI 0.85 to 0.95). Risk of cardiac death was also decreased in this comparison (RR=0.92, 95 percent CI=0.85 to 0.99), but risk of infectious death was not significantly decreased. In contrast, mortality in the above 36 percent group was compared with that in the 30 to <33 percent reference group, and no significant differences in mortality risk were identified. In the Hct >36 percent group, all-cause mortality had a relative risk of 1.06 (95 percent CI=0.89 to 1.27), cardiac mortality had a relative risk of 1.15 (95 percent CI=0.9 to 1.47), and infectious mortality had a relative risk of 1.04 (95 percent CI=0.62 to 1.73). The full report did not directly compare mortality in the above 36 percent group with that observed in the 33 to <36 percent group; however, the 95 percent confidence intervals for each of these two groups compared against the same reference group (30 to < 33 percent) could be compared with each other. When the results for Hct above 36 percent were compared with those for Hct 33 to 36 percent, no mortality advantage was seen. There is some suggestion of greater mortality in the Hct above 36 group, but this may be related to population selection rather than any adverse effect of using epoetin to maintain Hct above 36 percent.

Ma, Ebben, Xia, et al. (1999) controlled for covariates including: age, gender, race, prior dialysis exposure, comorbidity, primary renal diagnosis, and severity of disease (number of vascular access procedures, number of blood transfusions, hospital length of stay); however, no R² was reported in this paper. Since the subjects included in the analysis by Ma and colleagues (1999) were managed in the Medicare ESRD program, it should be noted that Medicare requires a medical justification (i.e., the patient must have a medical condition such as angina) to obtain reimbursement for maintaining an Hct above 36 percent. Thus, patients in this cohort with Hct above 36 percent likely would have been sicker than the general ESRD population and have had a mixture of comorbidities that justified a target Hct above 36 percent. This type of unfavorable selection process would bias upward the observed mortality in the higher Hct group, assuming these selection factors were not fully accounted for in the regression analysis.

Some of the same investigators as those who worked on the study by Ma, Ebben, Xia, et al. (1999) published an abstract reporting an analysis restricted only to patients newly on dialysis and in the Medicare ESRD population (incident Medicare patients) as opposed to their prior analysis, which included all eligible prevalent Medicare patients (Collins, Hao, Ebben, et al., 1999c [abstract], n=82,879). This abstract showed that, when compared with that for patients with Hct 30 to <33 percent, the relative risk of mortality was significantly decreased for hemodialysis patients with Hct 33 to <36 percent (RR=0.85, p=0.0001) or for those with Hct above 36 percent (RR=0.81, p=0.0001). There was insufficient information provided in the abstract to determine if the improvement at Hct above 36 percent represented any significant difference when compared with that at Hct 33 to <36 percent. In addition, peritoneal dialysis patients with Hct 33 to <36 percent had significantly lower mortality (RR=0.85, p=0.01) than those who had Hct 30 to <33 percent. When the relatively small number of peritoneal dialysis patients who had Hct above 36 percent (n=305) was compared with the reference group (Hct 30 to <33 percent), no significant difference in mortality was seen (RR=0.95, p=0.67). Although use of an incident population may be more homogenous with respect to severity and duration of CRF than a prevalence population, the followup period (which was not clearly stated in the abstract and was assumed to be less than or equal to 1 year) may have been too short to discern any mortality differences.

Lowrie, Huang, Lew, et al. (1994) performed a cross-sectional analysis on 1991 administrative data from National Medical Care. The analysis included 16,153 hemodialysis patients and categorized patients into the following Hct ranges: <20, 20 to 25; 25 to 30; 30 to 35; 35 to 40, and >40 percent. The authors found that mortality was statistically significantly higher when Hct increased above the reference range of 30 to 35 percent. The odds ratio of mortality for Hct of 35 to 40 percent compared with 30 to 35 percent was 1.45, p<0.0001. Similarly, the odds ratio of mortality for Hct >40 percent compared with 30 to 35 was 1.8, p<0.01.

The analysis by Lowrie and colleagues adjusted for case-mix predictors of death such as age, gender, diabetic status, and renal diagnosis. In addition, a variety of laboratory variables were noted to be predictors of death and were adjusted for in the model. However, the R² for this model suggests that the model accounts for no more than 20 percent of the variation in mortality among these hemodialysis patients.

Cognitive Performance

Overview

One full report by Pickett, Theberge, Brown, et al. (1999) described results of cognitive function testing in 20 dialysis patients. Patients were fully evaluated at a baseline mean Hct of 31.6 percent and reevaluated 2 to 3 weeks after Hct was normalized to a mean level of 42.8 percent. A variety of physiologic measures were recorded using electroencephalogram (EEG) spectra and cognitive event-related potentials (ERPS) during the Auditory Oddball Task and the Continuous Performance Task. These neurophysiologic tests were listed and discussed in increasing order of difficulty such that the highest level of cognitive function is required to perform the difficult Continuous Performance Task. Results are summarized in Evidence Table 11.

Evidence from nonrandomized controlled clinical trials

EEG spectra were obtained at baseline (T1) and followup (T2) in both the eyes-open and eyes-closed conditions using three different electrode locations: parietal (Pz), central (Cz), and frontal (Fz). After normalization of Hct, a decrease in EEG slowing was observed. The authors noted that this neurophysiologic result is consistent with a "general improvement in mental status." EEG results are reported in Evidence Table 11 using a power ratio:

Using within-subjects analysis of variance (ANOVA), the effect from T1 to T2 was significant in the eyes-open condition (p<0.02), but not the eyes-closed condition. According to the authors, this "was predictable because alpha tends to dominate the EEG spectrum (including theta) during eyes-closed conditions, obscuring any changes in theta power."

Pickett and coworkers (1999) tested cognitive event-related potentials in three different manners: Auditory Oddball Task, simple Continuous Performance Task (sCPT), and difficult Continuous Performance Task (dCPT). These three tasks were listed in increasing order of difficulty, each requiring successively more complex cognitive functioning. These tasks evoked an event-related potential termed P300 which has two characteristics, namely, P300 amplitude and P300 latency.

Table 5. Summary of results for Pickett, Theberge, (more...)

Table 5. Summary of results for Pickett, Theberge, Brown, et al., 1999

Pickett, Theberge, Brown, et al., 1999	Time Effect	Rx Effect	Electrode Effect	Time-by-Electrode Interaction	Rx-by-Electrode Interaction	Correlation with Hct
Auditory Oddball Task
P300 latency	ns	ns	p<0.05	??	ns	Yes, in 1 of 3 electrodes
P 300 amplitude	? not reported	ns	p<0.01	??	ns	No
sCPT
P300 amplitude	ns		p<0.003	p<0.002		No
P300 latency	ns					No
dCPT
P300 amplitude	ns		ns	p<0.05		No
P300 latency	ns					No

Rx = treatment; Hct = Hct; ns = not specified; SCPT = simple Continuous Performance Task; dCPT = difficult Continuous Performance Task.

Results of a statistical analysis using within-subjects ANOVA of P300 latency and P300 amplitude are summarized in Table 5. Terminology used in this article to describe effects included time effect, treatment effect, electrode effect, and several interaction terms -- time-by-electrode and treatment-by-electrode. In interpreting the results of cognitive event-related potential testing, the lack of an overall time or treatment effect when comparing mean Hct 31.6 percent (T1) to mean Hct 42.8 percent (T2) is somewhat concerning. Although some statistically significant results were reported for other comparisons, the numerous statistical comparisons made in analyzing this data made the importance of the various reported statistically significant results uncertain.

On cognitive event-related potential testing, the P300 amplitude was significantly reduced in the fz electrode after normalization of Hct for the simpler tasks (Auditory Oddball Task and sCPT). Pickett and colleagues (1999) attributed this finding to an "improved ability to sustain an attentional set." Changes in P300 amplitude on the dCPT after Hct normalization were suggested to reflect an improved ability to "maintain a memory trace of the last occurring stimulus" and to make the subject "more certain in detecting the target at the greater Hct level."

Conclusions

In adult patients with CRF, what are the outcomes of maintaining target Hct above 36 compared with maintaining target Hct in the 33 to <36 percent range?

Evidence comparing the outcomes of maintaining target Hct above 36 percent compared with maintaining target Hct in the 33 to <36 percent range consisted of four full reports (Besarab, Bolton, Browne, et al., 1998; Levin, Lazarus, and Nissenson, 1993; Ma, Ebben, Xia, et al., 1999; Xia, Ebben, Ma, et al., 1999;) and three abstracts (Collins, Hao, Ebben et al., 1999a, 1999b, 1999c).

No interventional studies directly addressed this question. The main source of evidence is derived from multiple cross-sectional analyses of a large Medicare dataset consisting of over 70,000 individuals. Because Medicare reimburses for maintenance of Hct over 36 percent only for patients with a comorbid condition, the findings of such associational studies are likely to reflect the underlying comorbidity in this population. Such database analyses cannot anticipate the results of adequately controlled interventional studies where the effects of using epoetin to maintain Hct above 36 percent could be directly compared with maintaining Hct in the 33 to < 36 percent range. The study described in Besarab, Bolton, Browne, et al. (1998) was an RCT; however, data specific to the Hct 33 to 36 percent group were available only from a cross-sectional analysis performed on the entire study population. The report by Levin, Lazarus, and Nissenson (1993) was a Phase IV surveillance study that reported on adverse events of epoetin. However, all comparisons among target Hct levels were based on cross-sectional analysis, and no tests of statistical significance were reported for differences between these subgroups.

The evidence is not adequate to compare the outcomes of maintaining target Hct above 36 compared with maintaining target Hct in the 33 to <36 percent range in adult patients with CRF. The evidence available on each of the specific outcomes of interest is summarized below.

• Three cross-sectional analyses, two full reports (Besarab, Bolton, Browne, et al., 1998; Ma, Ebben, Xia, et al. 1999) and one abstract (Collins, Hao, Ebben, et al. 1999c), provided data on the association between Hct and mortality. These data did not provide strong or consistent evidence of a mortality benefit in adult patients with CRF whose Hct is maintained above 36 percent as compared with those maintained at Hct 33 to <36 percent.

• Four cross-sectional analyses, two full reports (Levin, Lazarus, Nissenson, et al., 1993; Xia, Ebben, Ma, et al., 1999) and two abstracts (Collins, Hao, Ebben, et al., 1999a and 1999b), provide data on the association between Hct and hospital utilization. One full report and one abstract described favorable results but without analysis of statistical significance. One full report found higher hospital utilization in the group with Hct > 36 percent, and one abstract found no difference. These data did not provide strong or consistent evidence of reduced hospital utilization in adult patients with CRF whose Hct was maintained above 36 percent as compared with those maintained at Hct 33 to <36 percent.

• One cross-sectional analysis (Levin, Lazarus, and Nissenson, 1993; n=324) provided data on the association between Hct and cardiac outcomes. This study also reported on hypertension, CVAs, TIAs, and seizures; however, the absolute number of adverse events observed in each group being compared was quite small and no statistical analysis was reported. Thus, no conclusions can be drawn on the basis of this data.

No studies specifically compared the Hct ranges of primary interest to this systematic review with respect to the following outcomes: quality of life, red blood cell transfusion, exercise performance, cognitive function, sleep patterns, or nutrition.

2. In adult patients with CRF, what are the outcomes of maintaining target Hct above 36 compared with maintaining target Hct in the >30 to <36 percent range?

Evidence comparing the outcomes of maintaining target Hct above 36 percent compared with maintaining target Hct in the >30 to <36 percent range consisted of 13 full reports (Benz, Pressman, Hovick, et al., 1999; Besarab, Bolton, Browne, et al., 1998; Levin, Lazarus, and Nissenson, 1993; Levin, Singer, Thompson, et al., 1996; Levin, Thompson, Ethier, et al., 1999; McMahon, McKenna, Sangkabutra, et al., 1999; Moreno, Lopez-Gomez, Sanz-Guajardo, et al., 1996; Moreno, Sanz-Guajardo, Lopez-Gomez, et al., 2000; Lowrie, Huang, Lew, et al., 1994; Pickett, Theberge, Brown, et al., 1999; Suzuki, Tsutsui, Yokoyama, et al., 1995; Ma, Ebben, Xia, et al., 1999; Xia, Ebben, Ma, et al., 1999) and 10 abstracts (Collins, Hao, Ebben, et al., 1999a, 1999b, 1999c; Eschbach, Glenny, Robertson, et al., 1993; Foley, Parfrey, Morgan, et al., 1998; Mason and McMahon, 1997; Mason, Skinner, Sangkabutra, et al., 1997; Riedel, Hampl, Nundel, et al., 1996; Stray-Gunderson, Sams, Goodkin, et al., 1997; Wells, Coyle, Lee, et al., 1998). Some of these reports originated from the same authors and/or institutions, and the degree of overlap of patients across separate reports could not be determined from the information provided. Of the 12 interventional studies, only 4 included more than 100 patients. Two of these four were full reports (Besarab, Bolton, Browne, et al., 1998, n=1,233; Moreno, Sanz-Guajardo, Lopez-Gomez, et al., 2000, n=115) and two were abstracts and were both actually drawn from the same Canadian trial (Foley, Parfrey, Morgan, et al., 1998, n=146; Wells, Coyle, Lee, et al., 1998, n=130).

Additional evidence is derived from multiple cross-sectional analyses, with five of these publications, two full reports and three abstracts, coming from the same research group that analyzed a large Medicare dataset consisting of over 70,000 individuals. Because Medicare reimburses for maintenance of Hct over 36 percent only for patients with comorbid conditions, the findings of such associational studies using Medicare data are likely to reflect the underlying comorbidity in this population. Levin, Thompson, Ethier, et al., (1999) used an observational design to examine the natural history of predialysis patients.

Overall, the evidence is not sufficient to determine whether maintaining target Hct above 36 percent is more beneficial than maintaining target Hct in the 30 to <36 percent range in adult patients with CRF.

Key Question

What is the effect on outcomes of maintaining an Hct in the following ranges:

• Above 30 compared to 27 to <30 percent?

• Above 33 compared to 27 to <33 percent?

Results

The available literature evaluating the relationship between Hct in the relevant ranges and outcomes was sparse. None of the studies identified by the systematic review met the study selection criteria to be included in this evidence review. The key question for Part II is framed to compare outcomes associated with raising the Hct above 30 percent with outcomes achieved by maintaining the Hct between 27 and 30 percent. The minimum threshold for baseline Hct of 27 percent used in the systematic review protocol was prospectively defined based on current medical practice in pediatric nephrology (personal communication, Sandra Watkins, MD, April 8, 1999; Jabs and Harmon, 1996). The primary reason for exclusion of most of the studies reviewed in full text was that the baseline or control Hct to which normalization of Hct was being compared was below 27 percent. Evidence from these studies does not address the key question as set out in this systematic review, particularly with regard to the efficacy outcomes.

Table 6. Summary of observed effect on blood pressure (more...)

Table 6. Summary of observed effect on blood pressure in selected studies using epoetin to raise Hct over 30 percent in pediatric patients

Study	N	Age	Subjects	Effect on Blood Pressure
Yalcinkaya, Tumer, Cakar, et al., 1997	20	10.55 ± 2.93 range 5-16 y	CAPD	Increases in BP observed in high-dose epoetin group (150 U/kg) requiring increased medication, particularly in patients with hypertension at baseline. Mean Hct achieved = 32.4%.
Brandt, Avner, Hickman, et al., 1999	44	9.3 ± 6.25 range 4 m-20 y	HD, PD, predialysis	New or worsening hypertension was found in 13 (30%) and was more common in HD patients (66%) than PD (33%) or predialysis (16%) (p=0.02). Hct target was lower end of normal range for age.
Navarro, Alonso, Avilla, et al., 1991	23	8.3 + 5.7 y	CAPD, HD, predialysis	4 normotensive children developed mild hypertension during epoetin and required antihypertensive therapy. None had seizures or other major complications. Mean Hct target 30-36%.
Scigalla, Bonzel, Bulla, et al., 1989	51	12.6 + 3.9 y	HD, CAPD	16% developed new hypertension, 50% of prior hypertensive patients needed increased medication. 4% had to discontinue epoetin because of uncontrolled hypertension. Approx. 8% of patients had improved BP or decreased medication. Target Hct was greater than 30%.
Scharer, Klare, Braun, et al., 1993	15	7.8 + 5.3 y range 0.6-17 y	Predialysis	Increase in blood pressure observed with increased medication required in about one-half of patients. One patient developed uncontrolled hypertension requiring discontinuation of epoetin. Mean Hct achieved = 34.2%.

N = number of patients; BP = blood pressure; HD = hemodialysis; PD = peritoneal dialysis; CAPD = continuous ambulatory peritoneal dialysis.

Conclusions

1. In pediatric patients with CRF, what are the outcomes of maintaining target Hct above 30 compared with maintaining target Hct in the 27 to <30 percent range?

2. In pediatric patients with CRF, what are the outcomes of maintaining target Hct above 33 compared with maintaining target Hct in the 27 to <33 percent range?

No studies of pediatric patients were identified that met the study selection criteria to be included in this systematic review. The primary reason for exclusion of most of the studies reviewed in full text was that the baseline or control Hct to which the normalized group was being compared was below 27 percent. Thus, the available evidence does not address the key question as set out in this systematic review, particularly with regard to outcomes of treatment efficacy.

Key Question

What is the effect on outcomes of maintaining the Hct level >36 percent compared with 30 to <36 percent in the following patient subgroups (regardless of the presence of renal failure): (1) patients who have coronary artery disease, (2) patients who have congestive heart failure, (3) patients who live at high altitude, (4) patients who have arterial occlusive disease, (5) patients who have cerebrovascular disorders, (6) patients who have obstructive lung disease, and (7) patients who are in the adolescent age group?

Results

The literature search identified three full reports and no abstracts that met the selection criteria for this key question. Two of these studies (Besarab, Bolton, Browne, et al., 1998; Carson, Duff, Poses, et al., 1996) reported evidence relevant to patients with cardiovascular disease (including patients with coronary artery disease, congestive heart failure, arterial occlusive disease, subgroups 1, 2, and 4); one study (Kusunoki, Kimura, Nakamura, et al., 1981) addressed patients with cerebrovascular disease (subgroup 5). Only the study by Carson, Duff, Poses, et al. (1996) reported data comparing Hct above 36 percent compared with Hct 33 to <36 percent. These studies are described in Evidence Tables 15 and 16. There were no studies addressing patients in subgroup 3, 6, or 7 that met selection criteria for inclusion in this systematic review.

Quality of Study Design and Conduct

Table 7. Overview of study quality: Results and conclusions, (more...)

Table 7. Overview of study quality: Results and conclusions, Part III

Study	N	Better Study Quality		Worse Study Quality		Method of Case-Mix Adjustment
		Uses Double-Masked Design	Compares Hct >36% with 33-36%	Loss to Followup or Missing Data for >10%	Potential Bias in Control Group Comparison
Randomized controlled clinical trial
Besarab, Bolton, Browne, et al., 1998	1,233					N/A
Cross-sectional analysis
Carson, Duff, Poses, et al., 1996	1,958	N/A			N/A	Age; gender; race; cancer in past 5 years; diabetes mellitus; hypertension history; angina pectoris; congestive heart failure; atherosclerosis; cardiovascular disease; cardiopulmonary disease; APS+age score; Charlson score; intraperitoneal, intrathoracic, or aortic procedure; emergency surgery; general anesthesia
Kusunoki, Kimura, Nakamura, et al., 1981	27	N/A			N/A	None

N/A = not available; Hct = Hct; APS = acute physiology score.

Outcomes

Table 8. Overview of outcomes reported in each study: (more...)

Table 8. Overview of outcomes reported in each study: Results and conclusions, Part III

Author(s) Year	Sample Size	Study Design	Efficacy Outcomes						Treatment-Related Morbidity
			Mortality	Quality of Life	Hospital Utilization	RBC Transfusion	Cardiac Events	Other	Increased Blood Pressure	Vascular Access Thrombosis	Other
Besarab, Bolton, Browne, et al., 1998	1,233	Randomized, prospective, open-label trial
Besarab, Bolton, Browne, et al., 1998	1,233	Cross-sectional analysis
Carson, Duff, Poses, et al., 1996	1,958	Cross-sectional analysis
Kusunoki, Kimura, Nakamura, et al., 1981	27	Cross-sectional analysis						Cerebral blood flow; arterial oxygen content; oxygen delivery

RBC = red blood cell.

Evidence Pertaining to Patients Who Have Cardiovascular Disease

Evidence Pertaining to Patients Who Have Cerebrovascular Disease

Conclusions

1. What is the effect on health outcomes of maintaining the Hct level above 36 percent versus 30 to 36 percent in the following patient subgroups (regardless of the presence of renal failure): (1) patients who have coronary artery disease, (2) patients who have congestive heart failure, (3) patients who live at high altitude, (4) patients who have arterial occlusive disease, (5) patients who have cerebrovascular disorders, (6) patients who have obstructive lung disease, and (7) patients who are in the adolescent age group?

Technical Advisory Group

• Joseph W. Eschbach, MD

• University of Washington

• Alan E. Lichtin, MD

• Hematology/Oncology

• Cleveland Clinic Foundation and Co-Chair, ASH/ASCO Expert Panel on the Use of Erythropoietin

• Allen R. Nissenson, MD

• University of California in Los Angeles

• Earl Steinberg, MD, MPP

• Covance Health Economics and Outcomes Services, Inc.

Task Order Officer

• Martin Erlichman, MS

• Center for Health Care Technology

• AHRQ

Health Care Financing Administration Liaison

• John Whyte, MD, MPH

• Medical Officer

• HCFA

External Expert Reviewers/Advisors

• John W. Adamson, MD

• Blood Research Institute

• Tom Anderson, MD

• Division of Hematology/Oncology

• Medical College of Wisconsin

• Barbara Alving, MD

• Director, Division of Blood Diseases and Resources

• National Heart, Lung and Blood Institute

• Hanan S. Bell, PhD

• Independent Consultant; formerly Clinical Policies Analyst/Co-Director

• American Academy of Family Physicians

• Allan J. Collins, MD, FACP

• Associate Professor of Medicine

• Division of Nephrology, University of Minnesota

• Helen Danko, MS, RN, CNN

• Winthrop Hospital, Minneola, New York

• Peter DeOreo, MD

• Associate Clinical Professor of Medicine

• Case Western Reserve University, Division of Nephrology

• University Hospitals of Cleveland

• Steve Fishbane, MD

• Director, Dialysis Services

• Winthrop University Hospital, Minneola, New York

• Thomas Golper, MD

• Professor of Medicine

• Division of Nephrology, University of Arkansas for Medical Sciences

• David Goodkin, MD

• Amgen, Inc.

• Lawrence Goodnough, MD

• Division of Laboratory Medicine

• Washington University

• Paul Kimmel, MD

• Nephrology Program Director

• National Institute of Diabetes and Digestive and Kidney Diseases

• National Institutes of Health

• Carolyn E. Latham, President

• American Nephrology Nurses' Association

• Glenn Ramsey, MD

• Northwestern Memorial Hospital Blood Bank

• Mitchell Slavin, PharmD

• Ortho Biotech, Inc.

• Alice Thurston, Esq.

• Patient Advocate

• American Association of Kidney Patients

• Neil Powe, MD, MPH, MBA

• Johns Hopkins University

• School of Hygiene and Public Health

• Fernando Valderrabano, MD

• Chief, Department of Nephrology

• Hospital General Universitario

• Gregorio Maranon

• Madrid, Spain

• David van Wyck, MD

• Tucson, Arizona

• Sandra L. Watkins, MD

• Department of Pediatrics

• University of Washington Children's Hospital

• Jay Wish, MD

• Coordinator, Patient-based Program of Hemodialysis Services

• University Hospital, Division of Nephrology, Cleveland, Ohio

Blue Cross and Blue Shield Association Medical Advisory Panel

• Peter C. Albertsen, MD

• University of Connecticut Health Center

• Division of Urology

• James. O. Armitage, MD

• Internal Medicine

• University of Nebraska Medical Center

• Wade Aubry, MD (Chair)

• Blue Cross and Blue Shield Association

• National Medical Consultant

• J. Lawrence Colley, MD

• Corporate Medical Director

• Trigon BlueCross BlueShield

• Helen Darling, MA

• Practice Leader, Group Benefits & Health Care

• Watson Wyatt & Company

• David M. Eddy, MD, PhD

• Kaiser Permanente

• Southern California Region

• Scientific Advisor for Health Policy and Management

• Alan M. Garber, MD, PhD

• Director, PCOR and CHP

• Stanford University

• Michael A. Hattwick, MD

• Woodburn Internal Medicine Associates, Ltd.

• I. Craig Henderson, MD

• University of California, San Francisco

• Professor of Medicine, Chief of Oncology

• Albert R. Jonsen, PhD

• University of Washington

• Department of Medical History and Ethics

• Peter A. Margolis, MD, PhD

• Assistant Professor, Department of Pediatrics

• Clinical Assistant Professor, Department of Epidemiology

• Children's Primary Care Research Group

• Barbara J. McNeil, MD, PhD

• Ridley-Watts Professor and Head of Health Care Policy

• Harvard Medical School

• Richard G. Roberts, MD, JD

• Department of Family Medicine

• University of Wisconsin School of Medicine

• Earl P. Steinberg, MD, MPP

• Covance Health Economics and Outcomes Services, Inc.

• Clifford Waldman, MD

• Medical Director

• Anthem BlueCross & BlueShield

• Paul J. Wallace, MD

• Kaiser Permanente Northwest

• Les Zendle, MD

• Associate Medical Director

• Kaiser Permanente Southern California

Patient Population Information

Age (descriptive statistics)
Age subgroup analysis in paper: yes/no
Gender subgroup analysis in paper: yes/no
Race subgroup analysis in paper: yes/no
Total number of patients in analysis
Type of dialysis population: #pre-dialysis, #dialysis (percentages will be automatically calculated cells)
Type of dialysis - # HD, #PD

Patient inclusion exclusion criteria - to be provided to TAG only for information

Treatment information

EPO yes/no
IV iron yes/no
Oral iron yes/no
Androgen yes/no
Transfusion yes/no
Other free text

Pediatric studies:
Nutritional supplementation yes/no [when defined as nutrition given by IV, NGT (nasogastric tube), NJT (nasojejunal tube), G-tube (gastrostomy tube), J-tube (jejunostomy tube)]
Growth hormone yes/no

Group Hematocrit description table

Summary hematocrit value
Hematocrit summary statistic (mean versus median)
SD hematocrit
Range
Number of patients in group

Outcome information

• Functional Status - issues of various outcome assessment tools used and appropriate
  intermediate outcomes

• Quality of life - KD QOL etc

• Survival

• Blood transfusions - various ways of reporting
  % pts requiring transfusion
  number of transfusions needed (units)

• Hospitalization - various dimensions of this outcome to consider
  Number of hospital days
  Number of overall hospitalizations
  Number of hospitalizations for a specific problem

• Treatment related morbidity - rate of new or worsened hypertension, vascular access thrombosis, cardiovascular events, any others? (seizures, etc...)

• Growth - height, weight, percentile, standard deviation score, head circumference

• Development - Stanford-Binet score, Bailey score, normal vs. delayed

• School Performance - within normal limits vs. special needs Regularly attend school - able vs. unable vs. part-time