t echnolo g y r ep ort hta erythropoiesis-stimulating agents for anemia … · until april 2006,...

Supporting Informed Decisions

Erythropoiesis-Stimulating Agents for Anemia of Cancer or of Chemotherapy: Systematic Review and Economic Evaluation

t e c h n o l o g y r e p o r t

Canadian Agency forDrugs and Technologies

in Health

Agence canadienne des médicaments et des technologies de la santé

HTAIssue 119May 2009

Until April 2006, the Canadian Agency for Drugs and Technologies in Health (CADTH) was known as the Canadian Coordinating Office for Health Technology Assessment (CCOHTA).

Cite as: Tonelli M, Lloyd A, Lee H, Wiebe N, Hemmelgarn B, Reiman T, Manns B, Reaume MN, Klarenbach S. Erythropoiesis-stimulating agents for anemia of cancer or of chemotherapy: systematic review and economic evaluation [Technology report number 119]. Ottawa: Canadian Agency for Drugs and Technologies in Health; 2009. Production of this report is made possible by financial contributions from Health Canada and the governments of Alberta, British Columbia, Manitoba, New Brunswick, Newfoundland and Labrador, Northwest Territories, Nova Scotia, Nunavut, Ontario, Prince Edward Island, Saskatchewan, and Yukon. The Canadian Agency for Drugs and Technologies in Health takes sole responsibility for the final form and content of this report. The views expressed herein do not necessarily represent the views of Health Canada or any provincial or territorial government. Reproduction of this document for non-commercial purposes is permitted provided appropriate credit is given to CADTH. CADTH is funded by Canadian federal, provincial, and territorial governments. Legal Deposit – 2009 National Library of Canada ISBN: 978-1-897465-98-1 (print) ISBN: 978-1-897465-99-8 (online) H0468 – May 2009 PUBLICATIONS MAIL AGREEMENT NO. 40026386 RETURN UNDELIVERABLE CANADIAN ADDRESSES TO CANADIAN AGENCY FOR DRUGS AND TECHNOLOGIES IN HEALTH 600-865 CARLING AVENUE OTTAWA ON K1S 5S8

Publications can be requested from:

CADTH 600-865 Carling Avenue

Ottawa ON Canada K1S 5S8 Tel.: 613-226-2553 Fax: 613-226-5392

Email: [email protected]

or downloaded from CADTH’s website: http://www.cadth.ca

Canadian Agency for Drugs and Technologies in Health

Erythropoiesis-Stimulating Agents for Anemia of Cancer or of Chemotherapy: Systematic Review

and Economic Evaluation

Marcello Tonelli, MD SM FRCPC1 Anita Lloyd, MSc1

Helen Lee, MA PhD Candidate2 Natasha Wiebe, MMath PStat1

Brenda Hemmelgarn, PhD MD FRCPC2 Tony Reiman, BSc SM MD FRCPC3

Braden Manns, MD MSc FRCPC2 Martin Neil Reaume, BSc MD FRCPC MSc4

Scott Klarenbach, MD MSc FRCPC1

May 2009

______________________________________________________ 1 University of Alberta, Edmonton, Alberta 2 University of Calgary, Calgary, Alberta 3 Atlantic Health Sciences Corporation, Saint John, New Brunswick 4 The Ottawa Hospital Cancer Centre, Ottawa, Ontario


i

Reviewers These individuals kindly provided comments on this report.

External Reviewers Marion E. Jones, PhD (Economics) SOAS Associate Professor, Economics University of Regina Regina, SK

Keith O’Rourke, BA MBA DPhil Adjunct Professor University of Ottawa Ottawa, ON

Jason Hart, BSc MD Hematologist, Medical Oncologist BC Cancer Agency Vancouver Island Cancer Centre Vancouver, BC

Eric Nauenberg, PhD Associate Professor Department of Health Policy Management and Evaluation University of Toronto Toronto, ON

Douglas Stewart, BMSc MD FRCPC Professor University of Calgary Calgary, AB

CADTH Peer Review Group Reviewers Jocalyn Clark, BSc MSc PhD Adjunct Assistant Professor of Medicine University of Toronto Toronto, ON

Rick Audas, BBA MBA MA PhD Assistant Professor Faculty of Medicine Memorial University of Newfoundland St. John’s, NL

Industry: The following manufacturers were provided with an opportunity to comment on an earlier version of this report: Amgen Inc. and Janssen-Ortho Inc. All comments that were received were considered when preparing the final report.

This report is a review of existing public literature, studies, materials, and other information and documentation (collectively the “source documentation”) that are available to CADTH. The accuracy of the contents of the source documentation on which this report is based is not warranted, assured, or represented in any way by CADTH, and CADTH does not assume responsibility for the quality, propriety, inaccuracies, or reasonableness of any statements, information, or conclusions contained in the source documentation.

CADTH takes sole responsibility for the final form and content of this report. The statements and conclusions in this report are those of CADTH and not of its Panel members or reviewers.


ii

Authorship Marcello Tonelli, Scott Klarenbach, Anita Lloyd, Helen Lee, and Natasha Wiebe contributed to the conception and design, the acquisition of data, and the analysis and interpretation of data. Brenda Hemmelgarn and Braden Manns contributed to the conception and design and the acquisition of data. Marcello Tonelli, Scott Klarenbach, and Anita Lloyd drafted the report. All the authors contributed to the conception and design and the interpretation of data and critically revised the report for important intellectual content. All authors approved the final report. Acknowledgements Marcello Tonelli, Brenda Hemmelgarn, and Braden Manns were supported by New Investigator awards from the Canadian Institutes of Health Research. Marcello Tonelli, Scott Klarenbach, and Brenda Hemmelgarn were supported by Population Health Investigator awards from the Alberta Heritage Foundation for Medical Research, and Scott Klarenbach was supported by a Scholarship Award from the Kidney Foundation of Canada. Marcello Tonelli, Scott Klarenbach, Brenda Hemmelgarn, and Braden Manns were supported by an alternative funding plan from the Government of Alberta and the University of Alberta and the University of Calgary. Scott Klarenbach and Braden Manns were authors of the Canadian Society of Nephrology Clinical Practice Guidelines on management of anemia in chronic kidney disease, and Marcello Tonelli is chair of the Canadian Society of Nephrology Clinical Practice Guidelines committee. The authors of this report are grateful to Ellen Crumley for librarian support and Denise Adams, Mohammad Karkhaneh, Natasha Krahn, and Sophanny Tiv for additional reviewer support. We also thank Guiqing Lily Yao for providing details of the Birmingham economic model, which was used for reference. Conflicts of Interest Marcello Tonelli, Scott Klarenbach, Brenda Hemmelgarn, and Braden Manns received an unrestricted grant award to the Alberta Kidney Disease Network from Amgen Inc., and Amgen Inc. is not entitled to the results prior to publication and did not dictate the content or focus of the work. Tony Reiman has been an investigator on a number of clinical trials of erythropoietic agents for cancer patients and, although the centre did receive payment for work done by clinical trial nurses and study coordinators on a contractual basis, no direct payment was received for this work. All other authors declared no other conflicts of interest. Douglas Stewart has received honoraria for advisory boards and talks and research funding from Hoffmann-La Roche Ltd.


iii


iv

EXECUTIVE SUMMARY The Issue Available erythropoiesis-stimulating agents (ESAs) include epoetin alfa, epoetin beta, darbepoetin, and continuous erythropoietin receptor activator. These medications are mainly used in the treatment of anemia that is related to chronic kidney disease, but their use by people with anemia that is related to cancer (ARC) is increasing. ESAs are expensive, and there has been controversy about their safety in other populations. Therefore, a summary of the clinical benefits and cost-effectiveness of the use of ESAs in the treatment of ARC would help Canadian jurisdictions develop an evidence-based reimbursement policy for these drugs. Objectives The objectives were to assess the evidence on clinical efficacy, clinical harms, and the economic implications of ESA use by adult patients with anemia that is due to cancer or chemotherapy. The objectives were achieved by addressing four research questions. • What is the clinical-effectiveness (benefits and harms) of using ESAs for the treatment of

ARC, including the treatment of chemotherapy-induced anemia? • What is the cost-effectiveness of using ESAs in Canada for the treatment of ARC, including

the treatment of chemotherapy-induced anemia? • Are there differences in the clinical- and cost-effectiveness of ESAs based on the type of

cancer? • Are there specific groups of patients for whom red blood cell transfusions are not options for

the treatment of ARC and for whom the benefits of ESA therapy therefore differ from the general population?

Methods We conducted a systematic review of published and unpublished literature, using accepted methods for literature searches, article selection, data extraction, and quality assessment. For the systematic review of efficacy and harms, we included randomized controlled trials that included anemic adults with cancer who were receiving epoetin (alfa and beta), darbepoetin, continuous erythropoietin receptor activator, or no ESA. These trials compared clinical outcomes (all-cause mortality, cardiac event, hospitalization, quality of life (QoL), hypertension, red cell transfusions) and harms (adverse events) on the basis of hemoglobin targets (above or below 12 g/dL) and the route of ESA administration. We studied the following comparisons: ESA versus no ESA, early ESA intervention versus late ESA intervention, darbepoetin versus epoetin, different dosages, different schedules, and different routes of administration. For the economic analysis, we summarized the published studies that evaluated the incremental impact on relevant costs and health outcomes when the use of ESAs was compared with ESA non-use among patients with ARC. We also identified the parameters that significantly influenced the results. We performed a primary economic evaluation by creating a decision model to examine the costs and consequences of the use of ESAs by patients with anemia and cancer. We took the perspective of the Canadian public health care system. The model parameters and the time horizon for the analysis were based on summary data from studies that


v

were included in the systematic review. In addition to the base-case analysis, we performed one- way and scenario analyses, with specific interrogation of estimates for the effect of treatment on QoL, transfusion requirements, mortality, and certain patient subgroups. Finally, we estimated the budget impact of using ESAs to manage ARC, considering Canada as a whole and individual provincial payers. Results Clinical review We identified 52 trials (a total of 12,006 participants) that compared ESA with no ESA (42 trials compared epoetin alfa or beta with no ESA, 10 trials compared darbepoetin with no ESA). The median number of participants was 153, and the range was 60 to 989 participants. All-cause mortality during treatment was significantly different between groups (relative risk [RR]: 1.15 [95% confidence interval (CI) 1.03 to 1.29], P = 0.01), favouring no ESA. Treatment with ESAs prevented transfusions, compared with treatment with no ESA (RR: 0.64 [95% CI 0.56 to 0.73]), but led to an increased risk of thrombotic events (RR 1.69 [95% CI 1.27 to 2.24]) and serious adverse events (RR: 1.16 [95% CI 1.08 to 1.25]). There were no differences between groups in the risk of cardiovascular events (RR: 1.12 [95% CI 0.83 to 1.50]) or the incidence of hypertension (RR: 1.41 [95% CI 0.94 to 2.12]). ESAs led to clinically relevant improvements in QoL as indicated on the Linear Analog Scale Assessment, Functional Assessment of Cancer Therapy-Fatigue, and FACT-Anemia subscale scores (weighted mean difference [WMD]: 12.24 [95% CI 6.29 to 18.19]; WMD: 3.00 [95% CI 1.36 to 4.64]; WMD: 3.90 [95% CI 1.63 to 6.16]). Restricting analyses to studies in which ESAs were used in a fashion consistent with current practice guidelines (chemotherapy-induced anemia only; baseline hemoglobin < 10 g/dL; and target hemoglobin < 12 mg/dL) resulted in a much smaller evidence base: two studies with a total of 289 participants. There was no evidence that the risks or benefits of ESA treatment in this subgroup differed from those in the overall population of patients treated with ESAs for ARC. Specifically, P values from meta-regression comparing the RR of mortality (P = 0.25), serious adverse events (P = 0.48), transfusion (P = 0.40), or QoL (P > 0.40) in groups who did and did not meet these criteria were all non-significant. Economic Analysis For the base case, we considered all adult patients with ARC who participated in the studies that were included in the clinical review. This cohort had an average age of 62 years and initial hemoglobin of 10.4 g/dL, with 80% being treated with chemotherapy. Of the cohort, 23% had a hematological malignancy, and the remainder had a solid malignancy. With the ESA strategy, a mean of 42,111 units of epoetin per week was used for 15 weeks. The ESA and no ESA strategy permitted blood transfusions. Baseline and RR of death and transfusion from the clinical systematic review were used, and QoL was estimated based on the relationship between achieved hemoglobin and utility-based QoL scores. In the base-case analysis, when compared with no ESA and supportive transfusion, treatment with ESAs resulted in incremental costs of $8,643 and incremental benefits of 0.03 quality-adjusted life-years (QALYs) over a 15-week time frame, resulting in an incremental cost-utility ratio (ICUR) of $267,346. A one-year time frame resulted in the ESA strategy being dominated (more costly and with less QALYs) by the no ESA strategy. A sensitivity analysis that examined alternative estimates of QoL benefits and methods of incorporating reduced transfusion did not


vi

lead to significant reductions in the ICUR, although some resulted in ICURs up to $1.1 million/QALY. The ICUR was $125,668 in a scenario where multiple assumptions about QoL all favoured the ESA strategy. We performed 10 additional scenario analyses simulating the use of ESAs in accordance with current practice guidelines for ESAs. In three of these analyses, ESA therapy was dominated; in four the ICUR was > $100,000/QALY; and none had an ICUR < $70,000/QALY. Health Services Impact The number of Canadian patients who were treated with ESAs for ARC was estimated using two approaches. In the first approach, we considered ESA utilization data from Québec. The use of ESAs for ARC in Québec was estimated and extrapolated to other provinces and nationally. In the second approach, we used data from the Canadian Cancer Society and a European epidemiological study of anemia in cancer patients to estimate the number of incident patients who were undergoing treatment of cancer. Using these approaches, we estimated the annual cost of using ESAs to treat Canadian patients with ARC and hemoglobin levels of 10 g/dL or less to be $43 million to $73 million. Conclusions The use of ESAs by patients with cancer led to clinically meaningful improvements in QoL and decreased the risk of blood transfusions. However, ESA use led to an increased risk of all-cause mortality, a significant increase in the risk of serious adverse events, and cost-utility ratios that exceed commonly accepted standards for economic attractiveness. These considerations raise potential safety concerns about the use of ESAs to manage anemia in patients with cancer and suggest that further review by payers and regulatory authorities may be advisable.


vii

ABBREVIATIONS ARC anemia related to cancer ASCO American Society of Clinical Oncology ASH American Society of Hematology CERA continuous erythropoietin receptor activator CI confidence interval DCE discrete choice experiment ESA erythropoiesis-stimulating agent FACT Functional Assessment of Cancer Therapy ICUR incremental cost-utility ratio LASA Linear Analog Scale Assessment MI myocardial infarction QALY quality-adjusted life-year QoL quality of life RCT randomized controlled trial RR relative risk RRR relative risk ratio SD standard deviation TTO time trade-off WMD weighted mean difference WTP willingness to pay


viii

TABLE OF CONTENTS EXECUTIVE SUMMARY ............................................................................................................. iv ABBREVIATIONS ......................................................................................................................vii 1 INTRODUCTION...................................................................................................................1

1.1 Background ...................................................................................................................1 1.2 Overview of Technology................................................................................................1

2 ISSUE ...................................................................................................................................2 3 OBJECTIVES .......................................................................................................................2 4 CLINICAL REVIEW ..............................................................................................................2

4.1 Methods.........................................................................................................................2 4.1.1 Literature search................................................................................................3 4.1.2 Selection criteria and method ............................................................................3 4.1.3 Data extraction strategy.....................................................................................3 4.1.4 Strategy for quality assessment ........................................................................4 4.1.5 Data analysis methods ......................................................................................4

4.2 Results ..........................................................................................................................5 4.2.1 Quantity of research available ...........................................................................5 4.2.2 Trial characteristics............................................................................................5 4.2.3 Data analyses and synthesis.............................................................................8

5 ECONOMIC ANALYSIS .....................................................................................................17

5.1 Review of Economic Studies.......................................................................................17 5.1.1 Methods...........................................................................................................17 5.1.2 Results.............................................................................................................18 5.1.3 Discussion .......................................................................................................20

5.2 Primary Economic Evaluation .....................................................................................20 5.2.1 Methods...........................................................................................................20 5.2.2 Results.............................................................................................................25

6 HEALTH SERVICES IMPACT............................................................................................27

6.1 Population Impact and Budget Impact — Approach 1 ................................................28 6.2 Budget Impact — Approach 2 .....................................................................................28 6.3 Efficiency versus Equity ..............................................................................................29

7 DISCUSSION......................................................................................................................29

7.1 Results ........................................................................................................................29 7.1.1 Clinical review..................................................................................................29 7.1.2 Economic analysis...........................................................................................31

7.2 Study Limitations.........................................................................................................32 7.2.1 Clinical review..................................................................................................32 7.2.2 Economic analysis...........................................................................................34

7.3 Health Services Impact ...............................................................................................35 7.4 Knowledge Gaps.........................................................................................................35


ix

8 CONCLUSIONS..................................................................................................................36 9 REFERENCES....................................................................................................................37 APPENDICES — available from CADTH’s web site www.cadth.ca APPENDIX 1: Literature Search Strategies APPENDIX 2: Flowcharts of Selected Studies APPENDIX 3: Excluded Studies from Clinical Review APPENDIX 4: Tables for Clinical Review APPENDIX 5: Figures for Clinical Review APPENDIX 6: Tables for Economic Review APPENDIX 7: Tables for Economic Evaluation APPENDIX 8: Figures for Economic Evaluation APPENDIX 9: Tables for Health Services Impact APPENDIX 10: Forms


1

1 INTRODUCTION 1.1 Background It is expected that cancer will be diagnosed in more than 160,000 Canadians in 2008,1 approximately 30% of whom will develop anemia.2, 3 Anemia related to cancer (ARC) may be due to the direct effects of the cancer or due to comorbidity such as bleeding, or it may be a direct complication of cancer treatment. Because anemia is associated with adverse clinical outcomes in people with cancer (impaired quality of life (QoL), decreased survival), treatment of anemia has been considered as a potential method of improving outcomes in this population. Erythropoiesis-stimulating agents (ESAs) are medications that can be used to manage anemia in people with cancer or other diseases. Clinical practice guidelines on the use of ESAs by patients with cancer have been published.4 ESAs are costly (the expected cost per week of therapy for ARC is $350 to $450), and the reimbursement policies for their use vary across Canadian jurisdictions. Given that ESAs are commonly used by Canadian patients with cancer, ESAs can be a major cost driver in this population. Because the risk of transmitting infectious illnesses through blood transfusion (a potential alternative to ESA treatment) is thought to be less than it was previously,5 there is interest in evaluating the effectiveness and cost-effectiveness of ESAs when they are used in this population. Studies suggest that in patients with cancer, ESAs may be associated with an increased risk of adverse events such as thromboembolism.6 In addition, there has been controversy about the potential adverse effects that are associated with ESA use among patients with chronic kidney disease.7, 8 Therefore, an assessment of the clinical- and cost-effectiveness of ESAs in ARC, and their potential for harm, would be useful to Canadian jurisdictions that seek to develop an evidence-based reimbursement policy for these drugs. 1.2 Overview of Technology The medication class of ESAs includes epoetin alfa, epoetin beta, and darbepoetin. Of these, epoetin alfa (Eprex, Janssen-Ortho Inc.) and darbepoetin (Aranesp, Amgen Canada Inc.) are available in Canada. A third product, epoetin beta (NeoRecormon, Hoffman-La Roche Ltd.) is similar to epoetin alfa and widely used in Europe, but it is unavailable in Canada. A fourth agent, continuous erythropoietin receptor activator (CERA), is being studied in cancer populations, but it is unavailable in Canada. In this report, the term “epoetin” refers to epoetin alfa or epoetin beta. The term “ESAs” refers to epoetins (alfa or beta) and darbepoetin. Epoetin and darbepoetin are biosynthetic forms of the glycoprotein hormone erythropoietin, a hematopoietic agent that mainly affects erythropoiesis. Both agents elevate or maintain red blood cell levels but are not intended for the treatment of severe anemia requiring immediate correction. Epoetin and darbepoetin can be administered by intravenous or subcutaneous injection. Because of its longer plasma half-life, darbepoetin can be administered less often than epoetin and is typically administered once every three weeks in patients with cancer. Epoetin and darbepoetin are indicated for the treatment of anemia in patients with non-myeloid malignancies where anemia is due to concomitant chemotherapy. In addition, epoetin and


2

darbepoetin are indicated for the treatment of anemia that is associated with chronic renal failure, and epoetin is indicated for the treatment of transfusion-dependent anemia related to therapy with zidovudine in HIV-infected patients, and the treatment of patients who are undergoing major elective surgery to reduce allogenic blood transfusions. Alternative or complementary treatments to ESAs include no treatment, iron therapy (if iron deficiency is documented or suspected), and red blood cell transfusion. Practice guidelines indicate when ESAs should be used in preference to blood transfusion and vice versa.4 2 ISSUE Several Canadian jurisdictions reimburse for the cost of ESAs in the treatment of ARC. If data do not indicate that ESAs are associated with effective and cost-effective improvements in clinical outcomes among patients with cancer (compared with available alternatives), then this would support a more restrictive approach to reimbursement. In contrast, evidence of clinical and economic benefits would support current practices and allow Canadian decision makers to budget appropriately for the care of cancer patients. 3 OBJECTIVES Our objectives were to perform a systematic review of the clinical efficacy and harms of ESAs and to conduct an economic evaluation and budget impact analysis assessing ESA use in adult patients with ARC. The objectives were achieved by addressing four research questions. • What is the clinical-effectiveness (benefits and harms) of using ESAs for the treatment of

ARC, including the treatment of chemotherapy-induced anemia? • What is the cost-effectiveness of using ESAs in Canada for the treatment of ARC, including

the treatment of chemotherapy-induced anemia? • Are there differences in the clinical- and cost-effectiveness of ESAs based on the type of

cancer? • Are there specific groups of patients for whom red blood cell transfusions are not options for

the treatment of ARC and for whom the benefits of ESA therapy therefore differ from the general population?

4 CLINICAL REVIEW 4.1 Methods The systematic review was conducted and reported in accordance with guidelines.9, 10 A protocol was written a priori and followed throughout the review process. The primary objective was to assess the efficacy and harms of using ESAs in the treatment of adult patients with anemia due to cancer or chemotherapy. We compared the use of ESA with no ESA, early versus late intervention, and darbepoetin with epoetin. Studies in which patients received ESAs


3

perioperatively or after surgery were included if the patients were anemic before surgery (to exclude trials that focused on anemia that was due to intraoperative blood loss). 4.1.1 Literature search

The following bibliographic databases were searched: MEDLINE (1950 to October 22, 2007), EMBASE (1988 to October 22, 2007), and all EBM Reviews (October 22, 2007). Searches were not restricted by language (where a translator could be found), regardless of the publication source. Grey literature searches included cancer registries and trial registries. Appendix 1 Table 1 shows the sources searched and the strategies used. The searches were supplemented by hand searches of the reference lists of 46 relevant reviews. Canadian manufacturers of ESAs (Amgen Canada Inc. and Janssen-Ortho Inc.) were contacted, as were the authors of included studies. 4.1.2 Selection criteria and method

a) Selection criteria Studies were eligible for inclusion in the systematic review if they satisfied the following criteria: • Study design: Parallel randomized controlled trials (RCTs) with 30 participants or more in

each treatment group (small RCTs and crossover trials were excluded to improve efficiency without a loss of power and because long-term clinical outcomes are not generally reported)

• Population: Anemic adults (18 years of age or older) with cancer • Intervention: Epoetin (alfa and beta), darbepoetin, or CERA • Comparators: A different agent, no ESA (for example, placebo), or a different method of

delivery (dose, schedule, route of administration [subcutaneous or intravenous], fixed or weighted dose)

• Outcomes: All-cause mortality, cardiac event (myocardial infarction [MI], stroke, heart failure, or revascularization), hospitalization, QoL (for example, as indicated on the Functional Assessment of Cancer Therapy [FACT] total and subscales, Linear Analog Scale Assessment [LASA]), hypertension, red cell transfusions, and adverse events.

b) Selection method Each citation or abstract was screened by a clinician (MT or BH) and another reviewer (AL or DA). Any trial that was considered to be relevant by any reviewer was retrieved for review. The full text of each potentially relevant article was independently assessed for inclusion by two reviewers (AL, MK or DA), who used the selection criteria and a preprinted form (Appendix 10 Form 1). An initial set of 10 eligible trials was used to calibrate the relevance assessment between reviewers. Disagreements were resolved by a third party (NW) through consensus. 4.1.3 Data extraction strategy

One reviewer (AL, NW, or MO) extracted data from selected trials, and a second reviewer (MK or DA) checked for accuracy. A statistical reviewer (AL) checked the numerical results for accuracy. We used a standard data extraction method to record the following properties of each RCT: trial characteristics (country, design, sample size, duration of treatment and follow-up, hemoglobin inclusion criteria [the level of hemoglobin that defines a patient’s eligibility for inclusion in a trial]); participants (age, gender, previous ESA use, type of cancer, type of


4

chemotherapy); illness severity (hemoglobin levels); therapeutic regimens (type, dose, schedule, route of administration, target hemoglobin); control regimens and co-interventions, and outcomes (timing of outcome, results). We captured trial-level data on the following outcomes: mortality (all-cause), cardiovascular events (MI, stroke, congestive heart failure, revascularization), hospitalization (all-cause, cardiovascular), QoL (anemia, fatigue, general domains from the FACT), red cell transfusions, hypertension, and adverse events. We captured intention-to-treat analyses where presented. We classified adverse events as serious (if defined as such by the primary authors or if their severity was unspecified but they led to withdrawal from therapy or study) and sought the incidence of seizures and thrombotic events. In addition, given that various QoL measures were reported across studies, only QoL measures that were used in more than one study were considered in each comparison. 4.1.4 Strategy for quality assessment

We assessed study quality using a condensed version of the Chalmers Index.11 We also considered characteristics that were known to be associated with study quality (for example, method of allocation concealment,12 randomization technique, double-blinding, and description of withdrawals and dropouts13) (Appendix 10 Form 2). Finally, we extracted data on funding sources, given the potential to introduce bias.14 The quality of all included studies is based on trial design (participant selection, participant allocation, description of therapy, blinding, and withdrawals), statistical analysis (sample size calculation, intention-to-treat, preliminary analysis), and presentation of results (accrual dates, non-eligible patients, confidence intervals [CIs], and adverse event reports). Two reviewers (AL, MK or DA) independently assessed each study. An initial set of five eligible trials was used to calibrate quality assessment between reviewers. Disagreements were resolved with a third party (NW) through consensus. 4.1.5 Data analysis methods

We analyzed data using Review Manager 4.2.10 (Oxford, England), Stata 10 (College Station, Texas), and SAS 9.1 (Cary, North Carolina). The results were pooled using the method of Dersimonian and Laird15 for trials that compared ESA with no ESA (placebo or none), early ESA intervention with late ESA intervention protocols, and darbepoetin with epoetin, separately. We used the relative risk (RR) and the weighted mean difference (WMD) to summarize dichotomous and continuous results respectively. If there were multiple time points that were reported per outcome, we presented the latest available time point.16 Single-value imputations were substituted for missing standard deviations (SDs).17 The following imputation methods were used: using the SD from baseline, using the maximum SD from the available trial results, dividing the width of the interquartile range by 1.35, dividing the width of the range using an empirical table provided by Pearson,18 and assuming a correlation of 0.5 for change-from-baseline SDs. Based on the data from each study, the most appropriate method of imputation was used. Other continuous outcomes (for example, length of hospital stay) were associated with right-tail skewed distributions and were not pooled. Measures of central tendency and spread, tests, and corresponding P values were extracted for each intervention and reported. Because of


5

the differences expected between trials, we decided a priori to combine results using a random effects model.15 Statistical heterogeneity was quantified using I2,19, 20 which approximates the per cent of total variation (in- and between-study) due to between-study variation. We planned to use meta-regression21 a priori to examine whether certain variables (agent used, hemoglobin achieved, dose, age, gender, type of cancer, use of chemotherapy, duration of follow-up, and study quality) influenced the association between therapy and clinical outcome. Publication bias was assessed using weighted regression.22 In sensitivity analyses, we examined a cumulative plot to explore the findings when studies were ordered by publication year. When examining weekly initial ESA dose, darbepoetin alfa was converted to epoetin using a conversion factor of 1:169.23 In response to changes in the Health Canada label for darbepoetin, the protocol was amended in early 2009 to include analysis of subgroups defined by (a) the most recent guidelines issued by the American Society of Hematology and the American Society of Clinical Oncology (ASH/ASCO)4 (recommending the initiation of ESAs in patients with chemotherapy-associated anemia when hemoglobin is < 10 g/dL and targeting hemoglobin of ≤ 12 g/dL); and (b) the most recent Health Canada label for epoetin and darbepoetin (date of issue: January 16, 2009). 4.2 Results 4.2.1 Quantity of research available

In the literature search, we identified 1,948 citations in addition to 77 trials found outside the electronic database searches: 26 from manufacturers, 15 from trial and cancer registries, and 36 from references of systematic reviews. From these, 565 trials (560 articles) were retrieved for scrutiny (Appendix 2 Figure 1). A total of 489 trials were excluded (Appendix 2 Figure 1 and Appendix 3), resulting in 76 primary trials that met the selection criteria. In the 59 trials that were excluded because of a small sample size, mortality was reported for a total of 263 participants. Disagreements about eligibility occurred regarding 13 articles (kappa = 0.92) on aspects of intervention (3), no relevant control (3), outcome (2), sample size (1), study design (1), no anemia (1), and multiple publication (2). None of the 13 studies on which there was disagreement was ultimately included in the review. 4.2.2 Trial characteristics

Of the 76 included trials, 52 compared ESA with no ESA,6, 24-62,63-72 two compared the timing of ESA initiation (early versus late),73, 74 six compared darbepoetin with epoetin,75-80 and 16 focused on supplemental issues. Of these, eight trials studied different ESA doses,81-88 four studied different ESA dosing schedules,89-92 three studied different ESA dosing protocols (fixed versus weight-based dosing),93-95 and one studied the route of ESA administration.96 We identified 46 articles as multiple publications (Appendix 3). For the ESA versus no ESA comparison, one article63 reported on three trials. Several articles included results from multiple comparisons: six included ESA versus no ESA and a comparison of dosing protocols;42, 45, 54, 66, 67, 71 one included a darbepoetin versus epoetin comparison and a comparison of dosing protocols;75 one included a darbepoetin versus epoetin comparison, a comparison of dosing protocols, and a comparison of administration schedules;76 and one included comparisons of dosing protocols and administration schedules.87


6

a) Erythropoiesis-stimulating agents versus no erythropoiesis-stimulating agents Fifty-two trials (12,006 participants) compared ESA with no ESA (Appendix 4 Table 1A). Forty-two trials (7,356 participants)6, 24-33, 35, 36, 38-41, 43-47, 49-52, 55, 57, 58, 61-71 compared epoetin alfa or epoetin beta with no ESA, and 10 trials (4,650 participants)34, 37, 42, 48, 53, 54, 56, 59, 60, 72 compared darbepoetin alfa with no ESA. Four trials included patients who were undergoing surgery.27, 41, 45, 69 The median year of publication was 2003 (range 1993 to 2008). For the ESA versus no-ESA trials in patients who were not undergoing surgery, the median duration of treatment was 12 weeks (range eight to 28 weeks). The median duration of follow-up (reported by 19 studies) was 12 months (range one to 37 months). For the ESA versus no-ESA trials with surgery, the median duration of treatment was two weeks (range one to 3.6 weeks). The median duration of follow-up (reported by three studies) was three months (range one to 12 months). In all the trials that included patients who were undergoing surgery, ESAs were administered perioperatively. The dose of ESA was administered in fixed units (U for epoetin, µg for darbepoetin) or weight-based units (U/kg for epoetin, µg/kg for darbepoetin). For trials with epoetin, the median weekly starting doses were 35,000 U (range 7,000 U to 100,000 U) and 450 U/kg (range 300 U/kg to 2,100 U/kg). For trials that compared darbepoetin alfa with no ESA, the median weekly starting doses were 150 µg (range 100 µg to 300 µg) and 2.25 µg/kg (range 1.5 µg/kg to 5 µg/kg). All studies reported a subcutaneous route of administration. The median upper limit of hemoglobin as inclusion criteria was 11 g/dL (range 9 g/dL to 14 g/dL). Four trials reported separate inclusion criteria for males and females. The median upper limits of hemoglobin were 12 g/dL (range 11.5 g/dL to 13 g/dL) and 11 g/dL (range 10.5 g/dL to 12 g/dL) for males and females respectively. Approximately 27% of trials reported that the study participants had not previously received ESA therapy. Study participants with solid tumours were included in 30 trials.6, 27, 30, 32, 34, 35, 38-42, 44-47, 49, 50, 52, 55-58, 61, 62, 64, 68-72 Ten trials25, 26, 28, 31, 37, 43, 51, 65-67 included only study participants with hematological malignancies, and 11 trials24, 29, 36, 48, 53, 54, 59, 60, 63 included participants with solid tumours and hematological malignancies. One trial did not report the type of cancer,33 and another included patients with myelodysplastic syndrome.26 Seven trials26, 45, 46, 53, 59, 60, 63 reported that no chemotherapy was administered during the study. The median age of study participants was 62 years (range 48 years to 71 years), and the proportion of males ranged from 0% to 87% (median 44%). Few trials reported comorbidities among study participants. b) Early versus late initiation of erythropoiesis-stimulating agents Two trials73, 74 (473 participants) compared early initiation of ESA therapy with late initiation of ESA therapy (Appendix 4 Table 1B). Epoetin alfa was used in one trial,73 and darbepoetin alfa was used in one trial.74 The studies were published in 2006 and 2007. The median duration of treatment was 19 weeks (range 16 weeks to 22 weeks). A follow-up duration of one month was reported for one trial.74 All individuals who were included in these studies had hemoglobin levels between 10 g/dL and 12 g/dL. There are no accepted criteria for the severity of anemia, but one scale defines mild, moderate, and severe anemia as hemoglobin 10 to 11 g/dL, hemoglobin 8 to 10 g/dL, and hemoglobin < 8 g/dL, respectively.97 In the early ESA groups, treatment was started at the beginning of each trial at doses of 40,000 U weekly for the epoetin alfa trial and 300 µg every three weeks for the darbepoetin trial. Treatment was started in the late ESA groups when hemoglobin levels fell below 9 g/dL for the epoetin alfa trial and 10 g/dL for the darbepoetin trial. The ESAs were administered subcutaneously.


7

None of the study participants had previously received ESA therapy. The epoetin alfa trial included only participants with hematological malignancies. The darbepoetin trial included only individuals with solid tumours and hematological malignancies. Chemotherapy was administered during each trial. The mean age of participants ranged from 60 years to 63 years, and the mean proportion of males ranged from 35% to 52%. Darbepoetin versus epoetin Six trials75-80 (2,325 participants) compared darbepoetin with epoetin alfa (Appendix 4 Table 1C). The median year of publication was 2005 (range 2002 to 2006). The median duration of treatment was 16 weeks (range 12 to 16 weeks). The duration of follow-up was not reported for any trial. For each trial, the weekly starting dose of epoetin alfa was 40,000 U. Starting doses of darbepoetin were 200 µg every two weeks for four trials.77-80 For two trials,75, 76 there were multiple darbepoetin arms with weekly starting doses ranging from 1.5 µg/kg to 4.5 µg/kg. The ESAs were administered subcutaneously. The upper limit for hemoglobin inclusion was 11 g/dL for each study. All the trials reported that study participants had not previously received ESA therapy. Five75-79 of the six trials consisted of patients with solid tumours. The most common types were lung cancer and breast cancer. The sixth trial80 consisted of solid and hematological malignancies. Chemotherapy was administered to most patients during the trials. The mean age ranged from 56 years to 64 years. The proportion of males ranged from 0% to 36%. c) Supplemental issues Dosage Nine trials45, 66, 67, 71, 81-83, 85, 86 (1,435 participants) compared different dosages of epoetin, five trials42, 54, 75, 76, 84 (1,245 participants) compared different dosages of darbepoetin, and two trials87, 88 (311 participants) compared different dosages of CERA. Eight trials81-88 are included in Appendix 4 Table 1D. The remaining eight trials are included in Table 1A and Table 1C because they included comparisons for ESA versus no ESA and darbepoetin versus epoetin alfa respectively. The median year of publication was 2003 (range 1995 to 2007). The median duration of study was 12 weeks (range eight weeks to 24 weeks). One trial45 that included patients who were undergoing surgery had a duration of 25 days perioperatively. The mean age of participants ranged from 58 years to 72 years, and the median proportion of males was 47% (range 0% to 100%). Schedule Six trials76, 87, 89-92 (1,701 participants) compared eight schedule regimens. Four trials89-92 are included in Appendix 4 Table 1E. Two trials appear in Table 1C and Table 1D because they reported comparisons of darbepoetin versus epoetin and different dosing regimens respectively. The median year of publication was 2006 (range 2003 to 2007), and the median duration of the study was 13 weeks (range 12 weeks to 16 weeks). Epoetin alfa was used in one trial,91 epoetin beta in one,89 darbepoetin in three,76, 90, 92 and CERA in one.87 In one study,89 participants had hematological malignancies. In the remaining studies, study participants had solid tumours. The mean age of participants ranged from 62 to 66 years, and the proportion of males ranged from 25% to 64%.


8

Route of administration One trial96 (120 participants) compared the subcutaneous and intravenous administration of darbepoetin (Appendix 4 Table 1F). The duration of treatment was 18 weeks. The weekly starting dose of darbepoetin was 4.5 µg/kg. The upper limit of hemoglobin for inclusion was 11 g/dL. The mean age was 64 years, and the proportion of males was 50%. Fixed dose versus weighted dose Three trials93-95 (1,494 participants) compared the fixed dose to the weighted dose of ESAs (Appendix 4 Table 1G). The duration of treatment was 12 weeks, 15 weeks, and 16 weeks. Two trials94, 95 compared the fixed and weighted doses of darbepoetin (500 μg every three weeks versus 2.25 μg/kg once per week and 325 μg once per week versus 4.5 μg/kg once per week). The third trial93 compared epoetin doses (10,000 U versus 150 U/kg). All doses were administered subcutaneously. The mean age was 61 years, and the mean proportion of males was 44%. 4.2.3 Data analyses and synthesis

a) Erythropoiesis-stimulating agents versus no treatment The 52 trials were generally of poor to moderate quality (Appendix 4 Table 2A). Fourteen trials6, 34, 37, 38, 45, 47, 54, 55, 57-59, 64, 67, 71 reported adequate concealment of treatment assignment, and 26 trials24, 26-29, 31, 34, 37, 38, 41, 42, 46-49, 54, 57, 59, 60, 63-65, 69, 72 were double-blinded. Seventeen studies25, 26, 28, 29, 34, 35, 37, 42, 46, 47, 49, 54, 58-61, 69 adequately described losses to follow-up (frequencies and reasons by treatment group). The loss to follow-up was less than 10% in seven trials.6, 26, 29, 31, 39, 55, 66 Twenty-six trials6, 24-26, 28, 29, 31, 32, 36-38, 43, 45-47, 49-54, 57, 60, 61, 66, 71 used intention-to-treat analyses, and four trials53, 55-57 were reported as being preliminary analyses. No trial provided a description of non-eligible patients. Twenty-six trials6, 24, 27-29, 31, 37, 38, 42, 44, 48-54, 57-61, 63, 70 reported a private source of funding, three trials55, 56, 62 reported mixed funding, one trial66 reported a public source of funding, and 24 studies did not report funding. In privately funded studies, the manufacturer provided the ESA that was used during the study or a manufacturer of ESAs financially supported the study. Publication bias To investigate the potential for publication bias, we used results from analyses comparing all-cause mortality and assessed the results using a funnel plot. Our funnel plot seemed to be symmetrical (Appendix 5 Figure 1A), and the weighted regression test detected no statistical evidence of publication bias (bias = 0.02, P = 0.92). A cumulative meta-analysis plot98 (Appendix 5 Figure 1B) was used to examine how pooled results may have changed across time. The cumulative RR shifted to more than 1 to indicate an increased risk that was associated with ESA treatment in studies that were published in 2003 and beyond. Descriptive statistics were calculated for the studies in the two periods. There was no difference between studies published before or after 2003, except that in more recent studies, darbepoetin was more likely to be studied than epoetin. All-cause mortality Twenty-eight trials (n = 6,525) reported on-treatment all-cause mortality (Appendix 5 Figure 2A). Three trials had an additional ESA arm, resulting in 31 ESA versus no ESA comparisons.


9

The RR of death was significantly higher among ESA recipients (RR: 1.15 [95% CI 1.03 to 1.29]; I2 = 0%). The estimates of treatment effect were similar when stratified by agent (RR for epoetin: 1.12 [95% CI 0.97 to 1.29]; I2 = 0%; darbepoetin: 1.22 [95% CI 1.01 to 1.47]; I2 = 0%]). The estimates of treatment effect were also similar when analyses were stratified by type of cancer (solid tumour RR: 1.16 [95% CI 0.99 to 1.37]) versus hematological malignancy (RR: 1.11 [95% CI 0.77 to 1.61]), restricted to trials in which participants received chemotherapy (RR: 1.04, 95% CI 0.86 to 1.26), restricted to trials in which the target hemoglobin was 12 g/dL or less (RR: 1.15, 95% CI 0.94 to 1.40), or excluded trials that studied the perioperative administration of ESAs (RR: 1.15, 95% CI 1.02 to 1.28). We used univariable meta-regression to explore the factors that modified the association between ESA use and all-cause mortality (Appendix 4 Table 3A). The potential explanatory variables were duration of treatment, the upper limit of hemoglobin inclusion criteria, baseline hemoglobin level, use of ESAs according to ASCO criteria, achieved hemoglobin level, weekly initial dose of ESAs, agent, type of cancer, surgery, use of chemotherapy, mean participant age, percentage of male participants, and quality items. None of these variables significantly modified the association between ESAs and mortality in meta-regression. Sixteen trials (5,075 participants) reported end-of-study all-cause mortality for a median follow-up of 14 months (Appendix 5 Figure 2B). There was no difference between ESAs and control recipients in this sensitivity analysis (RR: 1.01 [95% CI 0.96 to 1.07]; I2 = 35%), which included follow-up for a median of 10.5 months after randomized treatment assignment had ended. Of these trials, six reported mortality at the end of the randomized treatment assignment and found an RR for mortality of 1.16 (95% CI 0.93 to 1.43) associated with ESA use. Cardiovascular events Fourteen trials (3,281 participants) compared the frequency of cardiovascular events (for example, MI, heart failure, stroke) between treatment groups (Appendix 5 Figure 2C). The overall RR was not statistically significant (RR: 1.12 [95% CI 0.83 to 1.50]; I2 = 0%). Hospitalization Although four trials (895 participants) reported the duration of hospitalization (Appendix 4, Table 4), the data as reported were not normally distributed and therefore could not be combined in meta-analysis. Of note, none of the individual studies34, 36, 53 reported significant differences in the duration of hospitalization between groups — except one41 conducted in patients undergoing surgery, which found a shorter duration among ESA recipients. However, this latter analysis used an inappropriate (parametric) test given the skewed nature of the data, and thus its conclusions should be viewed with caution. Health-related quality of life The change of overall QoL was measured in seven trials (1,326 participants) using the LASA scale (Appendix 5 Figure 2D). The WMD significantly favoured patients who received ESAs (WMD: 12.24 units [95% CI 6.29 to 18.19]). The heterogeneity between studies was large (I2 = 81%), probably because of the inclusion of one particular study.46 To determine the cause of heterogeneity, we examined the clinical and demographic characteristics of the patient population as well as the study quality. We were unable to determine the cause and performed a sensitivity analysis by removing the study. When this study was removed, statistical evidence of


10

heterogeneity was no longer present (I2 = 0%), and the overall pooled WMD remained significant at 9.80 (95% CI 6.95 to 12.64). Three trials (526 participants) reported the change in Total FACT-Anemia (WMD: 14.66 [95% CI −1.09 to 30.41]; I2 = 93%) (Appendix 5 Figure 2E). The heterogeneity was large and likely due to the inclusion of one particular study.46 Three trials (709 participants) reported the change using the FACT-General scale (Appendix 5 Figure 2F). The WMD was significant, favouring better QoL for the ESA group (WMD: 4.11 [95% CI 2.00 to 6.22]; I2 = 0%). Ten trials (n = 3,169) reported the change in the Fatigue subscore of the FACT-Anemia scale (Appendix 5 Figure 2G). The WMD favoured patients who were receiving ESAs (WMD: 3.00 [95% CI 1.36 to 4.64]), although large heterogeneity was observed (I2 = 73%). For the Anemia subscore (from the FACT-Anemia scale; Appendix 5 Figure 2H), seven trials (1,420 participants) were pooled and resulted in an overall change score favouring ESA therapy (WMD: 3.90 [95% CI 1.63 to 6.16]). The heterogeneity was large (I2 = 84%). For two studies,31, 36 changes in the Anemia subscore were minimal, although both favoured ESA treatment. On removing these studies, statistical evidence of heterogeneity disappeared, and the result remained statistically significant (RR: 6.38 [95% CI 4.68 to 8.08]). For the Anemia and Fatigue subscores, all results except two59, 60 favoured ESA treatment. All differences in LASA and FACT scores met or exceeded the threshold for minimal clinically important differences (LASA 8 mm to 20 mm, FACT-Fatigue subscale 3 to 4, FACT-Anemia subscale undefined and assumed to be the same as Fatigue subscale).99-103 Red cell transfusions Twenty-six trials (5,321 participants) reported the numbers of participants who were receiving red cell transfusions during the study (Appendix 5 Figure 2I). Three trials had more than one treatment arm in which ESAs were given, resulting in 31 ESA versus no-ESA comparisons. ESA treatment reduced the risk of receiving transfusions by 36%, which was statistically significant (RR: 0.64 [95% CI 0.56 to 0.73]), although there was heterogeneity (I2 = 55%). On stratification by type of cancer (Appendix 5 Figure 2J), it was revealed that heterogeneity for studies with solid tumours was 56%. The direction of effect was consistent, with two of 20 comparisons reporting fewer transfusions for the no-ESA group. The heterogeneity for studies with hematological malignancies or mixed cancers was 0%. Fifteen trials (21 comparisons) reported the number of participants who were receiving red cell transfusions from Week 5 to the end of treatment (RR: 0.57 [95% CI 0.52 to 0.63); I2 = 0%] (Appendix 5 Figure 2K). This result was also statistically significant. Fifteen trials compared transfusion volume between groups (Appendix 5 Figure 2L). The result was statistically significant (WMD: −0.80 units [95% CI −0.99 to −0.61]; I2 = 12%) indicating fewer red cell units transfused for the ESA group. Although there is no accepted criterion, this constitutes a clinically relevant difference. Similar to the on-treatment all-cause mortality analysis, we used univariable meta-regression (Appendix 4 Table 3B) for the same potential modifier variables. The duration of treatment was a statistically significant modifier (relative risk ratio [RRR]: 0.89 [95% CI 0.83 to 0.95]) suggesting that the effect of ESAs on preventing transfusions increased in parallel with duration of treatment. None of the other potential modifiers of treatment effect was statistically significant, including darbepoetin (versus epoetin) (RRR: 0.90 [95% CI 0.58 to 1.39]; P = 0.62) and achieved hemoglobin for the ESA arm (RRR: 1.02 [95% CI 0.91 to 1.14]; P = 0.70).


11

Tumour response Two trials (247 participants) reported the numbers of participants who had complete and partial tumour responses (the change in size of a tumour over time where a complete response was typically defined as “no detectable residual tumour by clinical or radiological means”) during the trials (Appendix 5 Figures 2M and 2N). The pooled RR of tumour response that was associated with ESAs was not significantly different from that of the control (RR for complete response: 0.88 [95% CI 0.69 to 1.12]; I2 = 0%; RR for partial response: 0.70 [95% CI 0.44 to 1.11]; I2 = 0%). Hypertension Seventeen trials (3,792 participants) reported numbers of participants with hypertension (Appendix 5 Figure 2O) during the study. The pooled effect of ESA therapy was not statistically significant (RR: 1.41 [95% CI 0.94 to 2.12]; I2 = 0%). Adverse events Twenty-one trials (5,891 participants) reported numbers of participants with events considered by the investigator to be “serious” (Appendix 5 Figure 2P). Not all studies specified the criteria that were used to define “serious” but several used definitions similar to “an adverse event that was fatal or acutely life threatening, required prolonged hospitalization, resulted in persistent or significant disability, or resulted in malignancy or congenital malformation/anomaly.”36 The pooled RR of serious adverse events was significantly higher in ESA recipients (RR: 1.16 (95% CI 1.08 to 1.25]; I2 = 0%). Thirteen trials with 14 comparisons (3,420 participants) reported the numbers of participants who experienced thrombotic events (Appendix 5 Figure 2Q). The pooled RR of thrombotic events was significantly higher in ESA recipients (RR: 1.69 [95% CI 1.27 to 2.24]; I2 = 0%). Four trials reported the numbers of participants who experienced a seizure (Appendix 5 Figure 2R) during the trial. The RR was not statistically significant (RR: 1.08 [95% CI 0.35 to 3.30]; I2 = 0%). Subgroup Analyses The RR of clinical outcomes was examined in groups stratified by baseline hemoglobin < 10 g/dL, 10 to 12 g/dL, > 12 g/dL, by whether participants did or did not receive chemotherapy and by the target hemoglobin. These strata were defined to correspond to the ASCO4 criteria for use of ESA in patients with cancer (see Tables 1 to 4 below). As the criteria used to define each subgroup became more specific, the quantity of evidence available declined substantially. For example, only two studies (three cohorts) that reported mortality used ESAs in a fashion that appeared to correspond to the ASCO criteria. In addition, there was little evidence that the clinical benefits or safety of ESAs were different in patients defined by the ASCO criteria, as compared with the total population of cancer patients treated with ESAs. Specifically, meta-regression comparing the RR of mortality (P = 0.25), serious adverse events (P = 0.48), transfusion (P = 0.40), or QoL (P > 0.40) in ASCO and non-ASCO subgroups were all non-significant. The significant increase in serious adverse events as well as the significant benefits for transfusion prevention and QoL were observed in the ASCO subgroup, as in the non-ASCO subgroup. Although the RR of mortality was non-significant in


12

the ASCO subgroup, CIs were wide, meaning that a clinically relevant increase in mortality (similar to that observed in the primary analysis) could not be excluded.

Table 1: Mortality Mean Hb at Baseline

Patient Population

No. of Cohorts / No. of Participants

RR of Mortality (95% CI) I2

All Hb All patients 31/6,525 1.15 (1.03 to 1.29) I2 = 0% Hb < 10 g/dL All patients 14/3,631 1.04 (0.81 to 1.32) I2 = 28% Hb ≥ 10 to 12 g/dL All patients 14/2,478 1.16 (0.99 to 1.36) I2 = 0% Hb > 12 g/dL All patients 1/94 3.00 (0.13 to 71.82) I2 = na Hb unclear All patients 2/322 2.20 (0.38 to 12.79) I2 = 34% All Hb No CIA 8/2,252 1.22 (1.06 to 1.40) I2 = 0% All Hb CIA 23/4,273 1.04 (0.86 to 1.26) I2 = 0% All Hb Target Hb < 12 g/dL 9/2,436 1.15 (0.94 to 1.40) I2 = 2% Hb < 10 g/dL CIA 13/2,646 0.96 (0.73 to 1.26) I2 = 18% Hb < 10 g/dL CIA, target Hb < 12 g/dL 3/289 0.77 (0.36 to 1.66) I2 = 41%

Table 2: Red Blood Cell Transfusion

Mean Hb at Baseline

Patient Population


RR of Transfusion (95% CI) I2

All Hb All patients 31/5,321 0.64 (0.56 to 0.73) I2 = 55% Hb < 10 g/dL All patients 16/1,765 0.72 (0.62 to 0.84) I2 = 22% Hb ≥ 10 to 12 g/dL All patients 12/3,272 0.57 (0.47 to 0.69) I2 = 56% Hb > 12 g/dL All patients 2/175 0.46 (0.11 to 1.88) I2 = 34% Hb unclear All patients 1/109 1.07 (0.78 to 1.48) I2 = na All Hb No CIA 7/786 0.80 (0.66 to 0.98) I2 = 5% All Hb CIA 24/4,535 0.60 (0.52 to 0.70) I2 = 59% All Hb Target Hb < 12 g/dL 5/1,315 0.55 (0.42 to 0.73) I2 = 47% Hb < 10 g/dL CIA 11/1,647 0.71 (0.60 to 0.83) I2 = 26% Hb < 10 g/dL CIA, target Hb < 12 g/dL 2/361 0.50 (0.29 to 0.87) I2 = 65%

Table 3: Serious Adverse Events

Mean Hb at Baseline

Patient Population


RR of SAE (95% CI) I2

All Hb All patients 23/5,891 1.16 (1.08 to 1.25) I2 = 0% Hb < 10 g/dL All patients 11/2,908 1.13 (1.03 to 1.25) I2 = 0% Hb ≥ 10 to 12 g/dL All patients 11/2,782 1.22 (1.09 to 1.37) I2 = 0% Hb > 12 g/dL All patients No studies No studies Hb unclear All patients 1/201 0.88 (0.55 to 1.41) I2 = na All Hb No CIA 5/1,948 1.30 (1.00 to 1.68) I2 = 32% All Hb CIA 18/3,943 1.14 (1.05 to 1.25) I2 = 0% All Hb Target Hb < 12 g/dL 9/2,560 1.18 (1.07 to 1.31) I2 = 0% Hb < 10 g/dL CIA 10/1,923 1.11 (0.97 to 1.26) I2 = 0% Hb < 10 g/dL CIA, target Hb < 12 g/dL 4/505 1.27 (1.00 to 1.60) I2 = 0%


13

Table 4: Quality of Life Mean Hb at Baseline

Patient Population

QoL LASA Overall FACT-Fatigue Subscale

FACT-Anemia Subscale

All Hb All patients 12.24 (6.29 to 18.19) I2 = 81%

3.00 (1.36 to 4.64) I2 = 73%

3.90 (1.63 to 6.16) I2 = 84%

Hb < 10 g/dL All patients 10.80 (5.02 to 16.58) I2 = na

2.78 (-0.25 to 5.80) I2 = 82%

2.26 (-0.22 to 4.73) I2 = 89%

Hb ≥ 10 to 12 g/dL

All patients 13.77 (5.95 to 21.59) I2 = 85%

3.21 (1.16 to 5.26) I2 = 66%

6.20 (3.80 to 8.60) I2 = 0%

Hb > 12 g/dL All patients 4.60 (-5.27 to 14.47) I2 = na

No studies 6.40 (0.83 to 11.97) I2 = na

Hb unclear All patients No studies No studies No studies All Hb No CIA 27.50 (21.75 to 33.25)

I2 = na 0.92 (-1.85 to 3.69)

I2 = 69% No studies

All Hb CIA 9.80 (6.95 to 12.64) I2 = 0%

3.87 (2.16 to 5.57) I2 = 60%

3.90 (1.63 to 6.16) I2 = 84%

All Hb Target Hb < 12 g/dL

15.59 (3.26 to 27.92) I2 = 91%

2.61 (-0.12 to 5.34) I2 = 82%

3.61 (-1.84 to 9.07) I2 = 92%

Hb < 10 g/dL CIA 10.80 (5.02 to 16.58) I2 = na

4.03 (2.09 to 5.97) I2 = 21%

2.26 (-0.22 to 4.73) I2 = 89%

Hb < 10 g/dL CIA, target Hb < 12 g/dL

No studies 4.98 (2.01 to 4.65) I2 = na 1.01 (-0.02 to 2.304) I2 = na

CI = confidence interval; CIA = chemotherapy-induced anemia; FACT = Functional Assessment of Cancer Therapy, Hb = hemoglobin; LASA = Linear Analogue Self Assessment; na = not applicable; QoL = quality of life; RR = relative risk; SAE = serious adverse events. b) Early erythropoiesis-stimulating agent intervention versus late

erythropoiesis-stimulating agent intervention The two trials73, 74 were of poor quality (Appendix 4 Table 2B). One trial74 reported adequate concealment of the treatment assignment, and neither was double-blinded. One trial74 adequately described losses-to-follow-up (frequencies and reasons by treatment group) with corresponding total percentage of loss-to-follow-up of 22%. Both trials reported a private source of funding. All-cause mortality No trials reported all-cause mortality during treatment. Both trials (470 participants) reported mortality at one and 18 months of follow-up respectively (Appendix 5 Figure 3A). For both trials, there were fewer deaths in the early ESA intervention group, although the results were not statistically significant. The pooled RR was 0.83 (95% CI 0.35 to 1.97) with I2 = 0%. Cardiovascular events Both trials (470 participants) reported the numbers of participants who experienced cardiovascular events (arrhythmia, congestive heart failure, cardiac arrest, MI) (Appendix 5 Figure 3B). The pooled RR was not significant (RR 1.49 [95% CI 0.57 to 3.93]; I2 = 0%). Health-related quality of life One trial73 (232 participants) reported change scores for the General, Fatigue, and Anemia subscores of the FACT-Anemia scale as well as for the Total FACT-Anemia score. The WMDs were General (WMD 4.11 [95% CI 1.46 to 6.76]), Fatigue (WMD 3.13 [95% CI 0.85 to 5.41]), Anemia (WMD 3.63 [95% CI 0.75 to 6.51]), and Total (WMD 8.21 [95% CI 3.08 to 13.34]). All the WMDs were statistically significant, indicating that early ESA intervention improved QoL.


14

All scores except Anemia subscores exceeded previously stated thresholds of clinically important differences.99 This trial also reported a change in overall QoL on the LASA scale (WMD: 4.06 [95% CI −1.29 to 9.41]). This did not exceed a clinically important difference.101 Red cell transfusions Both trials (470 participants) reported the numbers of participants receiving red cell transfusions during treatment (Appendix 5 Figure 3C). The risk of receiving transfusions was significantly reduced by 33% in the early ESA intervention group (RR: 0.67 [95% CI 0.47 to 0.95]; I2 = 0%), compared with the later intervention group. Tumour response One trial73 (269 participants) reported the number of participants who had a complete tumour response during the trial. The RR of tumour response that was associated with early ESA intervention was not statistically significantly different from that of late ESA intervention (RR: 0.99 [95% CI 0.59 to 1.68]). Hypertension One trial74 (201 participants) reported the number of individuals with hypertension during the study. Two cases of hypertension were reported for the early ESA intervention group. One case of hypertension was reported for the late ESA intervention group. The resulting RR was not significant (RR 2.06 [95% CI 0.19 to 22.36]). Adverse events Both trials (470 participants) reported numbers of participants who experienced serious adverse events (Appendix 5 Figure 3D) as classified by the investigators. Heterogeneity was moderate (I2 = 65%), and the direction of effect was inconsistent between these trials, so pooled results are not reported. Both trials (n = 470) reported the numbers of participants who had thrombotic events (Appendix 5 Figure 3E). The same two trials compared early epoetin treatment with late epoetin treatment. For this outcome, heterogeneity was absent (I2 = 0%), so pooled results are presented. The pooled risk of thrombotic events was significantly greater among patients receiving early intervention (RR 3.75 [95% CI 1.66 to 8.46]). c) Darbepoetin versus epoetin The six trials75-80 were of poor quality (Appendix 4 Table 2C). Three77, 78, 80 reported adequate concealment of treatment assignment, and none was double-blinded or adequately described losses-to-follow-up. Five trials76-80 reported a private source of funding. All-cause mortality Two trials (1,567 participants) reported on-treatment all-cause mortality (Appendix 5, Figure 4A). There was no significant difference in risk of death between agents in either trial. All-cause mortality post-treatment was not reported. Cardiovascular events One trial79 (352 participants) reported the incidence of cardiovascular events (MI, arrhythmia, atrial fibrillation, atrial flutter, cardiorespiratory arrest). The RR was not significantly different between agents (RR: 1.24 [95% CI 0.34 to 4.53]).


15

Health-related quality of life One trial80 (729 participants) reported the change score on the Anemia subscale (a domain of the FACT-Anemia scale). There was no difference between agents (WMD: 0.60 [95% CI −3.08 to 4.28]). Two trials (n = 830) reported the change score for the Fatigue subscale (a domain of the FACT-Anemia scale, Appendix 5 Figure 4B). The WMD was not significantly different between agents (WMD: 2.19 [95% CI −1.23 to 5.62]; I2 = 14%). Red cell transfusions One trial78 (141 participants) reported the number of participants receiving red cell transfusions during treatment. The risk of receiving transfusions was not significantly different between groups (RR: 0.35 [95% CI 0.12 to 1.04]). Three trials (1,676 participants) with six darbepoetin versus epoetin comparisons reported the number of transfusions that were experienced from the fifth week to the end of study follow-up (Appendix 5 Figure 4C). The risk of transfusion was not significantly different between agents (RR: 0.84 [95% CI 0.50 to 1.40]). The heterogeneity was moderate (I2 = 59%). One trial75 with four comparisons reported fewer transfusions for the darbepoetin groups, whereas the remaining two trials79, 80 reported the opposite. Adverse events One trial78 (141 participants) reported the number of participants who experienced a serious adverse event. More serious adverse events (as defined by the investigator) were observed in the epoetin alfa group, but the RR was not statistically significant (RR: 0.66 [95% CI 0.33 to 1.32]). Three trials (1,702 participants) reported the frequency of thrombotic events (Appendix 5 Figure 4D). The pooled RR favoured darbepoetin but was not statistically significant (RR: 0.79 [95% CI 0.55 to 1.13]; I2 = 0%). d) Supplemental issues Dose The 16 trials42, 45, 54, 66, 67, 71, 75, 76, 81-88 were of poor to moderate quality (Appendix 4 Table 2D). Nine trials45, 54, 67, 71, 81-83, 86, 88 reported concealment of treatment assignment. Four trials42, 54, 84, 85 were double-blind. Losses to follow-up were described adequately in less than half of the trials.42, 54, 81, 84, 87, 88 These trials reported losses to follow-up between 0% and 28%. Nine trials42, 54, 76, 82-84, 86-88 reported private sources of funding. Different doses were compared for CERA, epoetin, and darbepoetin. As a result, unpooled RRs are presented. None of the RRs was significant at the P = 0.05 level (Appendix 5 Figures 5A to 5J), although the CIs did not exclude clinically important differences in either direction. Schedule The six trials76, 87, 89-92 were generally of poor quality (Appendix 4 Table 2E). One trial91 adequately concealed treatment assignment. None of the trials was double-blind. Three trials87, 90, 91 described those individuals who withdrew. The withdrawal percentages were 11%, 36%, and 46% respectively. Five trials76, 87, 89-91 reported private sources of funding. One trial90 examined the darbepoetin dose of 6.75 μg/kg every three weeks synchronously (starting ESA therapy on day 1 of the chemotherapy cycle) or asynchronously (starting ESA therapy on day 15 of the chemotherapy cycle) with chemotherapy. For all-cause mortality, one death in each group was reported (RR: 0.88 [95% CI 0.06 to 13.65]). The incidence of


16

transfusions during the fifth week until the end of the study was reported (RR: 1.01 [95% CI 0.40 to 2.52]). For the remaining trials, the cumulative dose was equal across ESA arms, but the schedule of administration was varied (for example, once a week versus every two weeks versus every three weeks). Although a variety of comparisons were recorded (Appendix 5 Figures 6A to 6E), none of the results was statistically significant. Route of administration The one trial96 was of poor quality (Appendix 4 Table 2F). The concealment of treatment allocation was unclear, and the trial was not double-blind. The withdrawals and dropouts were adequately described (loss of 43%). This study was privately funded. The risk of death during the study was not significantly different between routes of administration (RR: 1.33 higher for intravenous [95% CI 0.31 to 5.70]). The risk of death up to 30 days after the study was completed was not significantly different between routes of administration [RR: 1.40 higher for intravenous [95% CI 0.47 to 4.16)]. This study also reported the frequency of transfusions from week 5 to the end of the study, but did not find differences in the number of blood transfusions or the number of units that were transfused. Fixed versus weight-based dose The three trials93-95 were of poor quality (Appendix 4 Table 2G). Two trials94, 95 reported adequate concealment of treatment allocation. One trial was described as double-blind.95 All trials reported a description of withdrawals (which ranged from 16% to 35%), and all were privately funded. The risk of death in the three trials (1,479 participants) during treatment was not significantly different between groups (RR: 1.01 [95% CI 0.68 to 1.52]); heterogeneity was I2 = 16% (Appendix 5 Figure 7A). Two trials (947 participants) reported mortality that included follow-up after the experimental treatment was complete (Appendix 5 Figure 7B). One trial95 (705 participants) reported incidence of cardiovascular events. The RR was not significant (RR: 0.83 [95% CI 0.52 to 1.32]) although fewer cardiovascular events (arrhythmia, congestive heart failure, MI) were reported for the fixed dose group. Two studies (1,168 participants) reported the number of transfusions that were experienced by individuals during the study (Appendix 5 Figure 7C). The pooled RR was not significantly different between groups (RR: 0.98 [95% CI 0.61 to 1.58); I2 = 74%). Two studies reported the number of transfusions that occurred during the fifth week until the end of treatment (Appendix 5 Figure 7D). The pooled RR was not significantly different between groups (RR: 0.89 [95% CI 0.59 to 1.34]; I2 = 52%). Other results that were not significant included the incidence of hypertension (Appendix 5 Figure 7E; two trials; 1,237 participants), thrombotic events (Appendix 5 Figure 7F; two trials; 1,237 participants) and number of seizures (one trial;95 705 participants).


17

5 ECONOMIC ANALYSIS 5.1 Review of Economic Studies 5.1.1 Methods

A protocol for the systematic review was written a priori and followed throughout the process. a) Literature search strategy In addition to the results from the literature searches for the clinical systematic review, we also screened citations that were found using economic search strategies. These searches were conducted in MEDLINE, EMBASE, EconLit, and the NHS Economic Evaluation database (Appendix 1 Table 2). b) Selection criteria Studies were eligible for inclusion in the systematic review if they met the following criteria: • Evaluated the incremental impact of an ESA against a comparator group on relevant costs

and health outcomes • included one of the following in the comparator group: placebo, no therapy with ESAs,

different ESA or same ESA but varying hemoglobin target, dose, or schedule • included (in a cost-minimization analysis) comparisons of different ESAs or comparisons of

alternative route or schedule of administration of ESAs to achieve a similar hemoglobin target, only if based on RCT data for effectiveness

• Examined a cohort of adult patients with malignancy and anemia. c) Selection method Two reviewers (SK or BM, NW) applied the selection criteria to the title and abstract of each citation. Full text articles were obtained for citations that could not be excluded by the title and abstract alone, and the full text was assessed in detail (SK, BM or CC) using a pre-printed form (Appendix 10, Form 3). To be included in the review, each study had to satisfy all the selection criteria. Disagreements between reviewers were resolved by consensus. d) Data extraction strategy One reviewer (SK) used a standard data extraction form to independently extract and document relevant information, which was verified by a second reviewer (NW). The types of information that were captured were author, title, intervention, comparators, study population, study design, time horizon, perspective, data sources for effects, data sources for costs, health-related QoL, currency, year, base-case incremental cost-effectiveness ratio results or incremental net benefit, sensitivity analysis, and conclusions. e) Strategy for quality assessment of studies The quality of each included study was independently assessed by two reviewers (SK and CC) using a checklist (Appendix 10, Form 4) that was adapted from Neumann et al.104 In addition, information on industry funding was recorded.


18

f) Data analysis methods We anticipated a priori that a small number of articles would be identified. Therefore, a qualitative synthesis of included studies was planned. Studies were grouped by the treatment strategies that were compared in the base-case analysis and by methodological approach (cost-utility, cost-benefit, willingness to pay [WTP], and cost analyses). The qualitative synthesis was intended to highlight the limitations and relevance of existing studies, costs, and QoL, and to assess the applicability of the findings to a contemporary Canadian setting. 5.1.2 Results

A total of 1,134 citations were identified from the combined searches. Of these, 58 were identified for scrutiny (Appendix 2 Figure 2). A total of 47 articles were excluded (Appendix 2 Figure 2), resulting in 11 studies that met the selection criteria. Disagreements arose with 4% of the articles (kappa = 0.87). Two disagreements resulted in exclusion because the studies were not economic evaluations. The 11 included articles are detailed in Appendix 6 Tables 1A and 1B. The study performance on the quality criteria appears in Appendix 6 Tables 2A and 2B. Five articles105-109 reported industry funding. a) Narrative review Cost-utility analyses Five of the 11 studies were cost-utility analyses. Without adjusting for inflation, four105, 110-112 studies reported a base-case cost per quality-adjusted life-year (QALY) in excess of $100,000. Only Martin et al.’s study106 reported an attractive incremental cost-utility ratio (ICUR) of £8,851/QALY, based on a stated threshold of £30,000/QALY. In this industry-sponsored trial, a group of patients with stage IV breast cancer was selected from a larger RCT of patients with solid tumours and malignancy where the treatment of anemia with epoetin was compared with the use of a placebo. A trend for survival advantage was noted in the trial for patients who were treated with epoetin in the group of patients with stage IV breast cancer. This was incorporated into the model (although it did not reach statistical significance, P = 0.13). If the survival estimate for the RCT cohort is used, the ICUR exceeded the stated threshold at £39,322/QALY. The impact of differential survival for subjects who were treated with ESAs was explored in the NICE (National Institute for Health and Clinical Excellence ) HTA.112 The use of the pooled risk of mortality of 1.03 (95% CI 0.88 to 1.21) resulted in an ICUR of £150,342/QALY; but in a sensitivity analysis, the use of the lower bound of the 95% CI (RR 0.88) reduced the ICUR to £39,568/QALY. Fagnoni et al.111 calculated the ICUR using an observational cohort of patients and converted the hemoglobin levels that were observed over time into the expected QoL using data from observational studies (the QoL was based on the LASA, which is not a preference-based measure of QoL). Barosi et al.110 populated a decision model with data from community trials and transformed achieved hemoglobin into QoL (based on the LASA). These studies reported base-case results of €310,577/QALY and $US 189,652/QALY respectively.


19

Cost-effectiveness, cost-benefit analyses, and willingness to pay Borget et al.113 defined effectiveness as a change in hemoglobin compared with an anchor hemoglobin value of 11 g/dL and reported average (rather than incremental) cost-effectiveness. This makes interpretation of the results difficult. Ortega et al.107 performed a WTP analysis and ascertained how much cancer patients and the general public would be willing to pay to receive epoetin and prevent blood transfusions. It was assumed that equal efficacy would be achieved with respect to hemoglobin level and subsequent QoL with either strategy. This estimate was used to assess the amount that cancer patients would be willing to pay during a three-month period (during which they would receive epoetin) and to ascertain the amount that members of the general public would pay in increased health care taxation per year. This latter value was then extrapolated to the amount that the patient would be willing to pay over his or her lifetime. After incorporating health care costs that were associated with epoetin use and WTP estimates, the use of epoetin was found to result in additional net costs. This suggests that even after incorporating patient preferences through WTP (from the patient or general population perspective), the use of epoetin does not seem to be an attractive use of resources using a cost-benefit analysis framework. In a similar manner, Ossa et al.108 used time trade-off (TTO) methods to derive utility scores (referenced to no anemia in cancer patients) and noted that cancer-related anemia has a significant impact on health-related QoL. Second, a discrete choice experiment (DCE) incorporating aspects of symptomatology and specifics of treatment for each modality was performed. By including monthly cost to the patient as an attribute in the DCE, the monthly WTP for epoetin was calculated at £368 (95% CI 318 to 419). Although the authors state that “the public value favorably the attributes of treatment with recombinant erythropoietin, and indicates a likely patient preference for treatment with erythropoietin over blood transfusion,” the WTP was not directly compared with the actual incremental costs of epoetin versus standard treatment with transfusion. Cost-analysis Casadevall et al.114 reported the results of an RCT in patients with myelodysplastic syndrome comparing epoetin and recombinant human granulocyte-colony stimulating factor with supportive care. Although anemia was improved in the treatment arm (based on achieving hemoglobin thresholds), no differences in QoL (FACT-Anemia) between the two treatment groups were found. Treatment with epoetin and recombinant human granulocyte-colony stimulating factor resulted in incremental costs of €17,977. In a cost comparison, Sheffield et al.115 evaluated the cost of strategies of epoetin use, including dose escalation with transfusion alone, based on practices at an oncology centre in the United States. Although effectiveness with respect to probability of achieving a hemoglobin response is reported, this analysis does not incorporate these effectiveness data. The conclusions state that from a health care system perspective, transfusion strategy is less costly compared with any epoetin use strategy. A second cost analysis by Reed et al.109 compares costs that were incurred in an RCT of epoetin versus darbepoetin. Although a difference in cost was found, the 95% cost estimates crossed unity and were largely driven by hospitalization, which was not a consequence of anemia or its treatment. Although the clinical consequence of hemoglobin response was provided, the use of a non-standard metric of effectiveness limits interpretation.


20

5.1.3 Discussion

The included studies consistently reported large incremental costs with ESA strategies compared with standard care, largely driven by the acquisition costs of the ESA. This result did not seem to be significantly altered in a range of costs for blood transfusion. In addition, most short- or long-term adverse events that were related to ESA use or blood transfusion did not seem to alter the results. The approach to measuring, valuing, and incorporating the health benefits of ESAs varied between studies that were included in our report. Cost and cost-effectiveness analyses (where hemoglobin is used as a measure of effectiveness) indicated increased cost with ESA use, but do not provide insight into the relative value of the health benefits achieved. Among studies that convert health outcomes into a common metric, such as QALYs or cost (cost-utility and cost-benefit), most of the base-case analyses indicated that managing cancer-related anemia with ESAs exceeds what is considered to be reasonable value for the health resources consumed. A significant limitation of studies is the absence of RCTs using preference-based utility scores as an outcome. Although disease-specific QoL scores are commonly used, their relevance to the estimation of QALYs is unclear. Methods of assigning a utility score using transformation of disease-specific measures or using observational data (where utility is assigned based on achieved hemoglobin) have been used, but may lead to biases that favour ESA treatment.116 Specifi

t echnolo g y r ep ort hta erythropoiesis-stimulating agents for anemia … · until april 2006,...

Documents