SEARCH

SEARCH BY CITATION

Keywords:

  • myelodysplastic syndromes;
  • MDS;
  • treatment outcome;
  • growth factors;
  • decision analysis

Abstract

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. REFERENCES

BACKGROUND.

The myelodysplastic syndromes (MDS) are divided into low-risk and high-risk diseases. Predictive models for response to growth factors (GF) have been developed based on red blood cell transfusion needs and erythropoietin levels. For low-risk MDS the optimal initial therapy (GF vs nongrowth factor [NGF] therapies, including differentiation and immunomodulatory agents) based on response rates to NGF and GF and survival, has not been defined.

METHODS.

A Markov decision analysis was performed on 799 low-risk MDS patients treated with either GF or NGF to determine the appropriate initial therapy. The treatment strategies analyzed included initial GF or NGF therapies, assuming 3 different states: Patients were either in the good GF predictive group (low transfusion needs and low erythropoietin levels), intermediate, or the poor GF predictive group (high transfusion needs and high erythropoietin levels).

RESULTS.

In the good GF predictive group, initial therapy with GF improved survival compared with NGF therapies at 3.38 years vs 2.57 years for a typical MDS patient. The advantage of GF to NGF was lost when NGF therapies produced a response in ≥46% of patients. In the intermediate or poor GF predictive groups, NGF maximized survival, provided response rates for NGF were >14% and 4%, respectively, for each predictive group. Quality of life adjustment did not alter the preferred strategy.

CONCLUSIONS.

Modeling estimates suggest that patients who fall into a good GF predictive group should almost always receive GF initially, whereas those in intermediate and poor predictive groups should almost always be treated with NGF. Cancer 2007 © 2007 American Cancer Society.

The myelodysplastic syndromes (MDS) represent a spectrum of heterogeneous hematopoietic disorders in which blood cells are derived from abnormal multipotent progenitor cells, with resultant peripheral blood cytopenias involving 1 or more lineages.1–3 Attention to cytogenetic features, clinical parameters (including the numbers and types of peripheral blood cytopenias and transfusion needs), and morphologic characteristics is crucial in determining prognosis, as has been demonstrated with the International Prognostic Scoring System (IPSS)4 and the World Health Organization Prognostic Scoring System.5

Whereas MDS can be subclassified into various subtypes, depending on the pathologic classification system applied,6, 7 in broad strokes these disorders are divided into “low-risk,” or indolent disease, in which pro-apoptotic forces predominate (frequently defined as “Low” or “Int-1” under the IPSS), and “high-risk,” or aggressive disease, in which pro-proliferative factors prevail (frequently defined as “Int-2” or “High” by the IPSS). In low-risk MDS, hematopoietic precursors have an impaired survival, due in part to the presence of inhibitory or inflammatory cytokines and enhanced angiogenesis.8–10 Therapeutic approaches capitalize on these distinctions. Thus, drugs that have shown more efficacy in high-risk MDS are pro-differentiating, promote the transcription of tumor suppressor genes, or are directly cytotoxic, whereas agents targeting low-risk MDS stimulate the remaining effective hematopoiesis and/or abrogate the effects of negative stimulatory cytokines. Recognizing this, and to standardize variable definitions of outcome, an International Working Group (IWG) created universal measures of response to therapy in MDS.11, 12

Once low-risk MDS patients require regular red blood cell transfusions, hematopoietic growth factor (GF) therapy (including recombinant erythropoietin alone or in combination with colony-stimulating factors) is often initiated. Predictive models for response to GF have been developed based on red blood cell (RBC) transfusion needs and erythropoietin (epo) levels.13, 14 Patients with low transfusion needs and low serum epo levels (ie, <2 units RBC per month and an epo level ≤500 U/L) have a good chance (74%) of responding to GF; patients with moderate transfusion needs and epo levels (ie, <2 units RBC per month with an epo level >500 U/L, or ≥2 units RBC per month with an epo level ≤500 U/L) have an intermediate chance (23%) of responding to GF; and patients with high transfusion needs and high epo levels (ie, ≥2 units RBC per month and an epo level >500 U/L) have a poor chance (7%) of responding to GF. There is no difference in survival among good, intermediate, and poor GF predictive groups.15 The optimal initial therapy (GF vs nongrowth factor [NGF] therapies, including differentiation agents and immunomodulatory drugs) in transfusion-dependent MDS patients based on response rates to NGF and GF and on survival has not been defined.

Decision analysis has been used to answer similar therapeutic questions in other hematologic malignancies.16–18 It allows for testing multiple initial conditions while incorporating standardized risk factors (such as IPSS score), response rates, survival, and quality of life (QOL) data in outcomes. It is a statistical tool that allows the user to play out certain clinical scenarios, given already-established probabilities (such as the probability of responding to a drug or probability of living 1 year), to guide clinical decisions in future patients. In MDS this statistical technique has influenced practice in determining the optimal timing for allogeneic bone marrow transplantation (BMT).19 In that decision analysis, the probability of survival was calculated based on baseline IPSS risk scores and the risk of death or the potential benefit of performing a BMT at diagnosis, various years after diagnosis, or at transformation to acute myeloid leukemia (AML). The model concluded, for example, that patients with high-risk MDS gained the greatest survival by undergoing a BMT at diagnosis, rather than waiting. We performed a decision analysis to determine the most appropriate initial therapy for low-risk MDS patients falling into good, intermediate, or poor GF predictive groups.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. REFERENCES

Patients

Patients were identified from a MEDLINE search on the keywords “MDS,” “RA,” “RARS,” “treatment,” “GF,” and “chemo.” Original articles with individual patient characteristics, pathologically confirmed RA or RARS MDS subtypes, and documented effect of therapy were included. From 1985 through 2005, 164 articles representing 2604 patients with low-risk MDS were reviewed (Fig. 1). Individual patient data, for precisely quantifying expected outcomes, were available on 811 patients from 90 articles (grouped data cannot be used in decision analyses). In studies that included both low-risk and high-risk MDS patients, only data on low-risk MDS patients (MDS with <5% bone marrow blasts) were used in this decision analysis. When comparing baseline characteristics, response patterns, and overall survival between individual patient data and grouped patient data, no significant differences were detected; thus, individual data were determined to be a good representation of the grouped data. Patients with available individual data were then divided into those treated with NGF therapies (n = 405, from 79 articles comprising 1005 patients) and those treated with GF (n = 394, from 85 articles comprising 1587 patients). Another 12 patients received placebo and were not included in further analyses.

thumbnail image

Figure 1. Patient identification.

Download figure to PowerPoint

NGF therapies included differentiation agents, immunomodulators, and nonablative cytotoxic agents (thalidomide, lenalidomide, anti-thymocyte globulin, all-trans retinoic acid, amifostine, cyclosporine A, 5-azacytidine, cytarabine arabinoside, etanercept, infliximab, 9-cis retinoic acid, 13-cis retinoic acid, menatetrenone, danazol); GF therapy included erythropoietin, granulocyte colony-stimulating factor, and granulocyte-macrophage colony-stimulating factor (single therapy and combination studies). As interferon and interleukin therapy can be considered either NGF or GF therapies, patients treated with these agents were first included as NGF subjects, and then as GF subjects, with no change in results. Thus, for the purposes of these analyses, they were grouped and reported as NGF subjects. Patients receiving both single-agent GF and combined GF therapy were combined, as the 2 approaches yielded similar response rates. The number of patients with available cytogenetics data, baseline erythropoietin levels, and baseline transfusion amount did not differ for GF and NGF patients. IPSS scores were either recorded in a publication, or calculated when sufficient data were available, in 584 patients.

Statistical Methods and Decision Modeling

Responses were standardized and recalibrated (when necessary) according to the IWG criteria and included complete response, partial response, and hematologic improvement.11, 12 Median survival in GF patients was derived from the Jadersten et al. study,15 and was validated with data from GF patients in our database. Survival in GF and NGF patients was also determined directly from the database and was available for 607 patients. T-tests were performed for continuous variables, and chi-square for categorical values, comparing GF and NGF groups. Survival analysis was performed by assuming, for each article, a constant hazard for all patients.

Decision model

The primary decision examined in this study was to define the optimal initial treatment strategy in patients with low-risk MDS. Two possible strategies were explored: 1) treating patients initially with GF, and 2) treating patients initially with NGF therapies. These 2 strategies were examined assuming 3 different initial states: i) patients were in the good GF predictive group, ii) patients were in the intermediate GF predictive group, and iii) patients were in the poor GF predictive group. Analyses were performed using a decision-tree modeling approach. The final outcomes of interest include the survival time in years and the QOL-adjusted survival time expressed as the quality adjusted life years (QALYs).

A Markov decision model20 was designed after a hypothetical cohort of patients transitioned among predefined health states. Possible health states considered in this model included “alive” with or without response to treatment, and “death” with or without response to treatment. The baseline starting age of this cohort was 60 years. One month was used as the transition cycle in the Markov model. The monthly transitional probabilities were derived from the median survival time in available publications.15, 21 The death state was considered absorbing by design, meaning no transitions occurred from this state to others.

To perform the decision analysis, certain assumptions had to be made, based on best practice standards for MDS. Patients initially treated with NGF continued on NGF therapies if they responded to the treatment. Nonresponders remained on therapy for only 2 months. Patients initially treated with GF continued on GF therapy if they responded to the treatment. Otherwise, an NGF treatment (including chemotherapy) was used to replace the initial therapy after an average of 2 months of failed GF therapy. In these patients, a similar scenario of NGF therapy was assumed for those who switched to NGF therapy after failing the GF therapy (Fig. 2). The response rates to NGF and GF therapies were derived from our database (Table 1).

thumbnail image

Figure 2. Markov decision tree. Notes of the decision tree model: This Markov tree starts with a cohort of patients with average age of 60. The Markov tree branches have transitional cycles of 1 month. Death from other causes have been considered based on age and from gender-specific mortality rates taken from life tables for the general population. rC, response rate of NGF; rG, response rate of GF; pC, monthly death probability for patients who responded to NGF; pG, monthly death probability for patients who responded to GF; pDie, monthly death probability for patients who do not respond to treatment.

Download figure to PowerPoint

Table 1. Variables in the Decision Model: Probabilities and Utilities
VariablesBase caseRange
Monthly transitional probabilities from survival to death
 If responds to nongrowth factor (pC)0.0090.007–0.011
 If responds to growth factor (pG)0.0160.012–0.020
 If no response to either (pDie)0.1090.082–0.136
Average response rate for nongrowth factor (rC)0.4150–1
Average response rate for growth factor (rG)0.4600–1
Overall quality of life utilities:
 With response to nongrowth factor0.630.54–0.73
 With response to growth factor0.630.54–0.73
 Without response0.530.38–0.59

The monthly death probabilities were assumed the same for patients who did not respond to NGF in both comparison arms. They were also assumed to be higher than the monthly death probabilities for patients who responded to NGF or GF therapy, which is clinically intuitive and supported by studies in low-risk MDS patients.15, 22, 23 Sensitivity analysis results were derived from the model for the cutpoint where either decision of GF or NGF treatment is indifferent.

There are limited data on QOL for MDS patients in the literature receiving either GF or NGF therapy.19, 24, 25 No utility measurements, the ideal outcome measures for decision analysis and cost-effectiveness analysis, were identified from the existing articles to assess the overall QOL in this population. Utilities are numerical values that range from 0–1.0 that represent the perceived value of a given health state. For example, a value of 0 may represent death, whereas 1.0 would represent perfect health. The best available source uses a European Organization for Research and Treatment of Cancer (EORTC) instrument and an overall QOL item administered to patients receiving either 5-azacytidine (an NGF therapy) or best supportive care.26 This study accompanied the phase 3 registration trial of MDS patients randomized to receive either 5-azacytidine or best supportive care.21 Scores from this item were transformed to develop utilities for our study. This EORTC instrument has been used in other studies to develop such utilities.27, 28 Values were incorporated for patients who did or did not respond to therapy, and a sensitivity analysis was performed in which assumptions varied within reasonable limits.

Morbidities from all other causes were considered based on age- and gender-specific mortality rates taken from life tables for the general population.29 The sensitivity analysis was conducted within the range of ±25% of the base case, or all possible values if needed.

RESULTS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. REFERENCES

Patients

Patient characteristics are displayed in Table 2. Patients receiving NGF therapies were younger, with a median age of 65 years (range, 21–83 years), compared with 69 years (range, 19–81 years) for GF patients (P < .05). The populations of GF and NGF patients were similar with respect to gender, cytogenetics, baseline transfusion needs (median of 2 pRBC transfusions per month [range, 0–8]), erythropoietin levels, and FAB and IPSS classifications. The IPSS score was 0–1.0 in 89.7% of the NGF patients and 91.2% of the GF patients. Patients receiving NGF therapies had a response rate (RR) of 41.5%, compared with 46% for GF patients (P < .05) and were more likely to have received prior therapies. Despite this, survival in the 2 groups did not differ, even after adjusting for age and baseline transfusion needs (hazard ratio 2.25 favoring GF patients, P = .14, Fig. 3).

thumbnail image

Figure 3. Survival of growth factor (GF) and nongrowth factor (NGF) patients (n = 607).

Download figure to PowerPoint

Table 2. Patient Characteristics
CharacteristicGrowth factors N = 394Nongrowth factors N = 405Significance
  1. MDS indicates myelodysplastic syndrome; RA, refractory anemia; RARS, refractory anemia with ringed sideroblasts; IPSS, International Prognostic Scoring System; RBC, red blood cells; NS, not significant.

MDS subtype (%)
 RA65.968.2NS
 RARS34.131.8
Men (%)68.964.4NS
Women (%)31.135.6
Age, y, median (range)69 (19–81)65 (21–83)P < .05
IPSS score (%)
 029.125.0NS
 0.529.532.9
 1.032.731.8
 1.55.28.1
 2.03.62.2
Cytogenetics (%)
 Normal49.652.9NS
 -Y3.22.1
 del (5q)9.110.4
 del (20q)5.16.6
 Trisomy 86.45.4
 Abnormal Chr 75.36.8
 Complex5.94.9
 Others15.410.9
Previous therapy (%)
 None96.979.1P < .001
 12.114.2
 1 or more1.06.7
Baseline RBC transfusions each mo (range)2 (0–7)2 (0–8)NS
Baseline erythropoietin level223 (4–6280)279 (17–4590)NS
Response rate46.0%41.5%P < .05
Median survival, mos4478NS

Decision Model

Life expectancies for the 3 GF predictive groups and the 2 treatment strategies are shown in Table 3. For an average patient with an RR of 30% for NGF therapies, NGF will be a preferred treatment if s/he is in the intermediate or poor GF predictive group. GF will be a preferred treatment if s/he is in the good GF predictive group. For a patient in the good GF predictive group, his/her survival would be 3.38 years for GF, versus 2.57 years for NGF. If the patient is in the intermediate or poor GF predictive group, his/her survival time for GF would decrease to 1.50 or 0.81 years, respectively, less than the 2.57 years for NGF therapies. The preferred treatment strategy for each GF predictive group did not change when using NGF RRs that ranged from 20% to 50%. This range included the 50% RR seen in patients treated with anti-thymocyte globulin (ATG) and cyclosporine A.

Table 3. Projected Outcomes for Low-Risk MDS Patients Based on Initial Treatment Strategy
Initial treatmentGF predictive group
GoodIntermediatePoor
  1. MDS indicates myelodysplastic syndrome; GF, growth factor; NGF, nongrowth factor; QALY, quality-adjusted life years.

  2. Assumes a 30% response rate to NGF therapy in a 60-year-old.

Years of survival if NGF2.572.572.57
Years of survival if GF3.381.500.91
QALY if NGF1.451.451.45
QALY if GF1.940.810.46

Using the decision model, with survival time as the outcome of interest, the base case results showed that patients in a good GF predictive group should receive GF unless NGF therapies have a chance of producing a response in ≥46% of patients (Fig. 4). Patients in the intermediate GF predictive group should receive NGF therapies if the therapies have a chance of producing a response in >14% of patients. Patients in the poor GF predictive group should receive NGF therapies in almost all cases, as long as the therapies have a chance of producing a response in >4% of patients. These results did not vary appreciably when analyses were limited to the 584 patients with an IPSS score, or the 90.5% of patients with an IPSS score ≤1.0. Results were also identical for previously untreated patients.

thumbnail image

Figure 4. Algorithm for choice of initial therapy in low-risk myelodysplastic syndromes (MDS) patients.

Download figure to PowerPoint

Sensitivity analyses indicated that the RRs for GF and NGF were the 2 most sensitive variables in the model. The change of their values was likely to change the final treatment choice between GF and NGF, which was demonstrated in the results above. Other variables, including risk of death, were not sensitive enough to change treatment decision-making. As an example, for an average patient with an RR of 30% for NGF, which is larger than the range between 11.2% and 18.0% (±25% of the threshold of the intermediate GF predictive group), NGF would still be the preferred treatment choice.

If QOL-adjusted survival time is the outcome of interest, the preferred treatment strategy has almost no change. Patients in a good GF predictive group should receive GF unless NGF therapies have a chance of producing a response in >46% of patients. Patients in the intermediate GF predictive group should receive NGF therapies if the therapies have a chance of producing a response in >14% of patients. Patients in the poor GF predictive group should receive NGF therapies in almost all cases, as long as the therapies have a chance of producing a response in >4% of patients.

Quality-adjusted life expectancies echo nonquality-adjusted findings (Table 3). If a patient is in the good GF predictive group, his/her QOL-adjusted survival time would be 1.94 QALYs for GF, vs 1.45 QALYs for NGF. If the patient is in the intermediate or poor GF predictive group, his/her QOL-adjusted survival time for GF would decrease to 0.91 or 0.46 years, respectively, less than the 1.45 QALYs for NGF.

DISCUSSION

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. REFERENCES

Treatment decision-making in patients with low-risk MDS used to be simple, when the standard of care was considered to be supportive and patients either responded to growth factors or relied on long-term blood product transfusions. The past decade has seen numerous clinical trials with NGF approaches and the Food and Drug Administration (FDA) approval of 3 drugs specifically for the treatment of MDS. Due to the limited efficacy of GF, we are experiencing a renaissance in disease-specific, targeted NGF therapies. With these newer classes of drugs has come the need to define when their use should supersede that of GF as initial therapy, particularly given their costs, variable response rates and definitions of response, and limited survival data.

Our treatment algorithm suggests that low-risk MDS patients falling into a good GF predictive group (low RBC transfusion needs and low epo levels) should initially be treated with GF unless NGF therapy response rates are ≥46%. Other low-risk MDS patients, once they reach a point at which therapy is being considered (frequently, once blood transfusions are being initiated or the hemoglobin has dropped below 9–10 g/dL), should probably be treated with NGF therapies initially.

The advantage of performing a decision analysis is that we were able to use a large amount of individual patient data from published trials of therapies for low-risk MDS. In doing so, we standardized response criteria and prognostic variables and pooled available survival data to develop a systematic approach to this patient population. We used an already-published predictive model for low-risk MDS patients because it is useful clinically, and expanded it to use in patients being considered for NGF therapies.

Naturally, the conclusions reached here apply to a large database of patients, and application to an individual patient must take into account factors difficult to assess in already-published data. For example, the cost of NGF therapy may be prohibitive to a patient with a high baseline erythropoietin level who is covered by Medicare for growth factor therapy received in a clinic setting, but who does not also have prescription drug coverage. Nonetheless, all things being equal, the results of this analysis suggest that this patient should be treated initially with NGF therapy.

This study has several potential limitations. First, there is a natural tendency toward a selection bias in patients who have been enrolled in a clinical trial of GF or NGF therapies, and 1 of the baseline characteristics reflects this: NGF patients were more likely to have received prior therapy, and thus likely to have had MDS for a longer period of time. Insufficient data were available from studies to actually report time since diagnosis for each cohort. However, this cohort's survival was no different from that of GF patients when baseline variables were taken into account, and the results and conclusions remained unaltered when only untreated patients were included in the analysis. The raw survival time was also incorporated into the decision model to yield the results presented herein. The counterargument would hold that patients enrolled onto clinical trials of NGF therapies were actually healthier, and that GF patients may have been deemed by their physicians to be too ill to receive NGF therapies. Similar performance statuses, when available, belie this claim, as do other similar baseline characteristics, including IPSS scores.

Survival data, as well as certain other baseline characteristics (such as cytogenetics and baseline transfusion needs and erythropoietin levels), were not available on all subjects. As these were not differentially absent in both GF and NGF groups, however, their absence would tend to result in random misclassification bias, and a tendency toward the null hypothesis of no difference. This was not the case. Analyses of the data including only patients with IPSS scores and survival did not alter the conclusions.

Another potential source of bias is the assumptions we made regarding the transitional probabilities. We assume the transitional probabilities were the same for patients who do not respond to NGF and those who do not respond to both GF and NGF, which might not be the case in reality. There are no available data to suggest the difference, including no clear-cut survival advantage to the use of GF or NGF in MDS patients. Additionally, we made the assumption that NGF replaces GF after an average of 2 months of failed GF therapy. As neither GF nor NGF therapies have been demonstrated to improve survival in low-risk MDS patients, building into the model a “watch and wait” period before this transition would not have had an effect on the findings.

In conclusion, patients with low-risk MDS should be classified according to GF predictive models of response, with therapy tailored according to likelihood of responding to growth factors. Those who fall into a good predictive group should almost always receive GFs, unless NGF approaches yield a high response rate, as might be the case with, for example, lenalidomide for MDS associated with a del(5q) cytogenetic abnormality, where response rates (defined as transfusion independence) approaches two-thirds.30, 31 Those in intermediate and poor predictive groups should almost always be treated initially with NGF therapies. These recommendations should only serve as a guide for treatment in the absence of randomized studies, with therapy tailored to the individual patient.

REFERENCES

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. REFERENCES
  • 1
    Sekeres M,Fu A,Maciejewski J,Golshayan A,Kalaycio M,Kattan M. A decision analysis to determine the appropriate treatment for low-risk myelodysplastic syndromes. Blood. 2005; 106: 711a.
  • 2
    Silverman L. The Myelodysplastic Syndrome. In: HollandJ, FreiE, BastR, KufeD, MortonD, WeichselbaumR, eds. Cancer Medicine. Baltimore: Williams and Wilkins; 1996: 25932615.
  • 3
    List AF,Vardiman J,Issa JP,DeWitte TM. Myelodysplastic syndromes. Hematology (Am Soc Hematol Educ Program). 2004: 297317.
  • 4
    Greenberg P,Cox C,LeBeau MM, et al. International scoring system for evaluating prognosis in myelodysplastic syndromes. Blood. 1997; 89: 20792088.
  • 5
    Malcovati L,Germing U,Kuendgen A, et al. A WHO classification-based prognostic scoring system (WPSS) for predicting survival in myelodysplastic syndromes. Blood. 2005; 106: 232a.
  • 6
    Bennett JM,Catovsky D,Daniel MT, et al. Proposals for the classification of the myelodysplastic syndromes. Br J Haematol. 1982; 51: 189199.
  • 7
    Harris NL,Jaffe ES,Diebold J, et al. The World Health Organization classification of neoplastic diseases of the hematopoietic and lymphoid tissues. Report of the Clinical Advisory Committee meeting, Airlie House, Virginia, November, 1997. Ann Oncol. 1999; 10: 14191432.
  • 8
    Allampallam K,Shetty V,Mundle S, et al. Biological significance of proliferation, apoptosis, cytokines, and monocyte/macrophage cells in bone marrow biopsies of 145 patients with myelodysplastic syndrome. Int J Hematol. 2002; 75: 289297.
  • 9
    Bellamy WT,Richter L,Sirjani D, et al. Vascular endothelial cell growth factor is an autocrine promoter of abnormal localized immature myeloid precursors and leukemia progenitor formation in myelodysplastic syndromes. Blood. 2001; 97: 14271434.
  • 10
    Kitagawa M,Saito I,Kuwata T, et al. Overexpression of tumor necrosis factor (TNF)-alpha and interferon (IFN)-gamma by bone marrow cells from patients with myelodysplastic syndromes. Leukemia. Dec 1997; 11: 20492054.
  • 11
    Cheson BD,Bennett JM,Kantarjian H, et al. Report of an international working group to standardize response criteria for myelodysplastic syndromes. Blood. 2000; 96: 36713674.
  • 12
    Cheson BD,Bennett JM,Kantarjian H, et al. Myelodysplastic syndromes standardized response criteria: further definition. Blood. 2001; 98: 1985.
  • 13
    Hellstrom-Lindberg E,Negrin R,Stein R, et al. Erythroid response to treatment with G-CSF plus erythropoietin for the anaemia of patients with myelodysplastic syndromes: proposal for a predictive model. Br J Haematol. Nov 1997; 99: 344351.
  • 14
    Hellstrom-Lindberg E,Gulbrandsen N,Lindberg G, et al. A validated decision model for treating the anaemia of myelodysplastic syndromes with erythropoietin + granulocyte colony-stimulating factor: significant effects on quality of life. Br J Haematol. Mar 2003; 120: 10371046.
  • 15
    Jadersten M,Montgomery SM,Dybedal I,Porwit-MacDonald A,Hellstrom-Lindberg E. Long-term outcome of treatment of anemia in MDS with erythropoietin and G-CSF. Blood. 2005; 106: 803811.
  • 16
    Lee SJ,Kuntz KM,Horowitz MM, et al. Unrelated donor bone marrow transplantation for chronic myelogenous leukemia: a decision analysis. Ann Intern Med. 1997; 127: 10801088.
  • 17
    Ng AK,Weeks JC,Mauch PM,Kuntz KM. Decision analysis on alternative treatment strategies for favorable-prognosis, early-stage Hodgkin's disease. J Clin Oncol. 1999; 17: 35773585.
  • 18
    Ng AK,Weeks JC,Mauch PM,Kuntz KM. Laparotomy versus no laparotomy in the management of early-stage, favorable-prognosis Hodgkin's disease: a decision analysis. J Clin Oncol. 1999; 17: 241252.
  • 19
    Cutler CS,Lee SJ,Greenberg P, et al. A decision analysis of allogeneic bone marrow transplantation for the myelodysplastic syndromes: delayed transplantation for low-risk myelodysplasia is associated with improved outcome. Blood. 2004; 104: 579585.
  • 20
    Sonnenberg FA,Beck JR. Markov models in medical decision making: a practical guide. Med Decis Making. 1993; 13: 322338.
  • 21
    Silverman LR,Demakos EP,Peterson BL, et al. Randomized controlled trial of azacitidine in patients with the myelodysplastic syndrome: a study of the cancer and leukemia group B. J Clin Oncol. 2002; 20: 24292440.
  • 22
    Yamada T,Tsurumi H,Kasahara S,Hara T,Sawada M,Moriwaki H. Immunosuppressive therapy for myelodysplastic syndrome: efficacy of methylprednisolone pulse therapy with or without cyclosporin A. J Cancer Res Clin Oncol. 2003; 129: 485491.
  • 23
    Kantarjian H,Issa JP,Rosenfeld CS, et al. Decitabine improves patient outcomes in myelodysplastic syndromes: results of a phase III randomized study. Cancer. 2006; 106: 17941803.
  • 24
    Spiriti MA,Latagliata R,Niscola P, et al. Impact of a new dosing regimen of epoetin alfa on quality of life and anemia in patients with low-risk myelodysplastic syndrome. Ann Hematol. 2005; 84: 167176.
  • 25
    Sekeres MA,Stone RM,Zahrieh D, et al. Decision-making and quality of life in older adults with acute myeloid leukemia or advanced myelodysplastic syndrome. Leukemia. 2004; 18: 809816.
  • 26
    Kornblith AB,Herndon JE2nd,Silverman LR, et al. Impact of azacytidine on the quality of life of patients with myelodysplastic syndrome treated in a randomized phase III trial: a Cancer and Leukemia Group B study. J Clin Oncol. 2002; 20: 24412452.
  • 27
    Gulbrandsen N,Wisloff F,Nord E,Lenhoff S,Hjorth M,Westin J. Cost-utility analysis of high-dose melphalan with autologous blood stem cell support vs. melphalan plus prednisone in patients younger than 60 years with multiple myeloma. Eur J Haematol. 2001; 66: 328336.
  • 28
    Bloomfield DJ,Krahn MD,Neogi T, et al. Economic evaluation of chemotherapy with mitoxantrone plus prednisone for symptomatic hormone-resistant prostate cancer: based on a Canadian randomized trial with palliative end points. J Clin Oncol. 1998; 16: 22722279.
  • 29
    National Center for Health Statistics. Vital Statistics of the United States, 1999, Mortality. Hyattsville, MD: Public Health Service; 2004.
  • 30
    List A,Kurtin S,Roe D, et al. Efficacy of lenalidomide in myelodysplastic syndromes. N Engl J Med. 2005; 352.
  • 31
    List A,Dewald G,Bennett J, et al. Hematologic and cytogenetic (CTG) response to lenalidomide (CC-5013) in patients with transfusion-dependent (TD) myelodysplastic syndrome (MDS) and chromosome 5q31.1 deletion: results of the multicenter MDS-003 study. J Clin Oncol (Proc Am Soc Clin Oncol). 2005; 23: 2s.