The Search for Better Prognostic Models in Myelodysplastic Syndromes

Abstract

Myelodysplastic syndromes (MDS) are a group of heterogeneous bone marrow disorders characterized by a failure of hematopoiesis and an increased propensity for transformation to acute myeloid leukemia. Determining the prognosis of patients with MDS is essential for discerning the best therapy, which can vary from supportive care to allogeneic stem cell transplantation. The most widely used prognostic model in MDS is the International Prognostic Scoring System (IPSS), which estimates survival and risk of transformation to acute myeloid leukemia based on the percentage of blasts, karyotype, and number of cytopenias, but the IPSS has several limitations that preclude more widespread application. Over the past decade, several studies have reported on new prognostic factors for MDS, including transfusion dependency and DNA methylation abnormalities. More recently, two prognostic models for MDS that aim to overcome the limitations of the IPSS have been published. This review focuses on the most recent advances in this field, detailing current prognostic models and the more important risk factors in MDS.

Introduction

Myelodysplastic syndromes (MDS) are a group of clonal hematopoietic stem cell disorders characterized by hematopoietic cell dysplasia, ineffective hematopoiesis, peripheral blood cytopenias, and an increased risk of transformation to acute myeloid leukemia (AML) [1]. MDS is more common in the elderly population, with a median age at diagnosis of 71 years [2]. The overall annual incidence of MDS is 3.27 cases per 100,000 persons, increasing to 20.05 cases per 100,000 persons in the population 70 to 79 years of age [3]. Known risk factors for developing MDS include exposure to radiation therapy, chemotherapy, or benzene; tobacco use; and other hematopoietic stem cell disorders, including aplastic anemia and inherited bone marrow (BM) failure syndromes such as Fanconi anemia [4, 5]. Current minimal diagnostic criteria for MDS require the presence of peripheral blood cytopenia in association with at least one of the following: 1) dysplasia in at least 10% of cells from one or various hematopoietic cell lineages (erythroid, neutrophilic, megakaryocytic) and/or >15% ring sideroblasts; 2) 5% to 19% myeloid blasts; 3) characteristic karyotypic abnormality [6]. Additionally, all hematopoietic and nonhematopoietic disorders that can cause cytopenias must be excluded before a diagnosis can be firmly made.

MDS is a very heterogeneous group of entities: some are indolent with prolonged survival, whereas others have an aggressive course similar to AML, with a median survival of less than 6 months [1]. This heterogeneity is related to differences in the biology of the disease and has important clinical implications, as patients with low-risk MDS are usually managed with growth factors and supportive care measures, while patients with high-risk disease receive aggressive treatments such as chemotherapy and allogeneic stem cell transplantation (SCT) [1]. Thus, estimating the risk of death and transformation to AML of an individual patient with MDS is essential to indicate the proper treatment.

Ever since the French-American-British (FAB) classification of MDS in 1982 [7], several studies have proposed prognostic models in MDS with the objective of differentiating low-risk from high-risk patients [8, 9]. The International Prognostic Scoring System (IPSS) is the scoring model most often used for patients with MDS [10], but in more recent years, several studies have proposed alternative models that can overcome some of the limitations of the IPSS [11, 12••]. There is a continuous search for the ideal prognostic model in MDS. This review summarizes recent data in this field.

The International Prognostic Scoring System

The IPSS evaluated 816 patients with primary, untreated MDS and analyzed clinical and morphologic features that were associated with overall survival (OS) and risk of transformation to AML [10]. Variables significantly associated with these end points in the univariate analysis included FAB subtype, percentage of blasts, karyotype abnormalities, and number of cytopenias. The results of a Cox proportional hazards multivariate analysis confirmed that blast percentage, karyotype, and the number of cytopenias were associated with OS and risk of transformation to AML; a categoric prognostic model was developed, which could successfully predict outcome in these patients (Table 1) [10].

Table 1. International Prognostic Scoring System (IPSS) prognostic model for myelodysplastic syndromes.

Variable	Points
	0	0.5	1.0	1.5	2.0
Bone Marrow Blasts (%)	< 5	5-10	-	11-20	21-30
Cytopeniasa	0/1	2/3	-	-	-
Karyotypeb	Good	Intermediate	Poor	-	-
Risk Group	IPSS Score	25% AML Evolution (yr)		Overall Survival (yr)
Low	0		9.4	5.7
Intermediate-1	0.5-1.0		3.3	3.5
Intermediate-2	1.5-2.0		1.1	1.2
High	2.5-3.5		0.2	0.4

Legends

^aHb < 10 g/dL; Absolute Neutrophil Count < 1.8×10⁹/L; Platelet Count < 100×10⁹/L;

^bGood: Normal, del(5q), del(20q), -Y; Intermediate: all others; Poor: complex (> 3 abnormalities), chromosome 7 abnormalities

The IPSS is by far the best-known and most-used scoring model in MDS and has served to guide treatment decisions and the design of clinical trials for patients with this disease [13, 14]. However, the IPSS has several limitations that preclude its more widespread application. First, it excluded from the analysis patients with secondary MDS (i.e., therapy-related myeloid disorder [t-MDS]), who comprise 20% to 30% of patients with MDS and are usually associated with worse survival [5, 12••]. Patients with chronic myelomonocytic leukemia (CMML) who presented with a white blood cell (WBC) count of 12×10⁹/L or higher were also excluded, as their disease was considered to be closer to myeloproliferative neoplasms than to MDS. However, several clinical studies in patients with MDS also include cases of CMML regardless of WBC count, and drugs that are approved for MDS (e.g., decitabine) may also be highly active in CMML [15]. Patients who were previously treated were not included in the IPSS classification, and the model was valid only at the time of diagnosis. MDS is dynamic, however, so a prognostic model that is also dynamic and can predict outcome when the features of a patient's disease change during the clinical course is highly desirable.

Risk Factors in Myelodysplastic Syndromes

In the 13 years since the publication of the IPSS, several studies of prognostic factors in MDS have been published. These reports either refined well-known prognostic criteria such as morphology and karyotype or described new factors like DNA methylation patterns or transfusion dependency. A comprehensive review of all known risk factors in MDS is beyond the scope of this article, and we will focus only on the most important ones.

Bone Marrow Morphology

Morphologic evaluation of marrow findings is an important tool for classifying and predicting prognosis in patients with MDS [16]. Classically, the most important morphologic feature with prognostic relevance has been the percentage of blasts in a 500-cell manual count of a BM aspirate smear [10]. The importance of blasts was shown in the FAB classification, which roughly divides MDS into two subtypes: low-risk with less than 5% blasts (refractory anemia [RA], refractory anemia with ring sideroblasts [RARS]) versus high-risk with 5% or more of blasts (refractory anemia with excess blasts [RAEB], refractory anemia with excess blasts in transformation [RAEB-t]) [7]. The IPSS model also uses blast percentage as the sole morphologic marker for prognosis [10].

Neither the FAB classification nor the IPSS model recognizes that the presence of multilineage dysplasia in the absence of increased blasts defines a distinct subtype of MDS with worse survival. Before publication of the World Health Organization (WHO) classification, Rosati et al. [17] reported on 18 patients with refractory cytopenias and multilineage dysplasia (RCMD). These patients had more profound cytopenias than patients having RA or RARS and a higher percentage of abnormal karyotype (49% vs. 7%) [17]. Median survival with RCMD (24 months) was intermediate between survival with RA/RARS (108 months) and survival with RAEB (18 months). Other reports have confirmed the impact of severe dysplasia on the outcome of patients with MDS [18, 19].

The 3rd edition of the WHO classification of hematopoietic neoplasms added the category of RCMD to the roster of MDS subtypes [16]. In a recent publication, Malcovati et al. [20] compared the outcomes of 467 patients with MDS reclassified according to the WHO classification. Patients with RCMD with or without ring sideroblasts (RS) had significantly inferior survival compared with patients having RA/RARS (median OS, 49 months vs. 108 months; P<0.001) [20], confirming that multilineage dysplasia is associated with worse survival than unilineage involvement.

Other morphologic features that indicate a worse prognosis include BM fibrosis and aggregates or clusters of CD34+ cells [21–23]. Increased marrow fibrosis has long been reported in BM biopsies from patients with MDS. More recently, one large series confirmed that patients with hyperfibrotic MDS (grade 2–3 BM fibrosis) have shorter leukemia-free survival and OS [21]. The effect was independent from other variables and from the IPSS and WHO classification-based Prognostic Scoring System (WPSS) criteria [21]. In practical terms, the finding of grade 2 to 3 BM fibrosis had the equivalent effect of shifting patients one category up in the IPSS and WPSS risk scores. Other series have confirmed the poor prognosis of these patients [24].

Abnormal localization of immature precursors (ALIP) in the central portion of the BM has been described in core BM biopsies of patients with MDS, being more common in RAEB and associated with poor outcome [23]. Immunohistochemistry studies have revealed that these aggregates are composed of CD34+ cells and represent clusters of blast cells [22]. All BM biopsies of patients with MDS should be evaluated for the presence of CD34+ clusters, as these patients have an increased risk of transformation to AML [21].

Cytogenetics

Karyotype has emerged as one of the most important prognostic factors in MDS. Chromosomal abnormalities are found in approximately 50% of patients with de novo MDS, and in 80% to 92% of patients with therapy-related MDS [25]. Balanced chromosomal translocations commonly found in AML—t(8;21), t(15;17)—are uncommon in MDS, which is more frequently characterized by gains and losses of chromosomal material and unbalanced translocations. Some of the most common cytogenetic abnormalities in MDS include del(5q) (6%), -7/del(7q) (2%), +8 (4.5%), del(20q) (1.7%), and -Y (2%) [25, 26]. Additionally, the 2001 WHO classification defines one subtype of MDS based on the karyotype (5q- syndrome), and it is expected that future updates of this classification will include more cytogenetically defined entities [16].

Regarding prognosis, the IPSS classified karyotype into three risk groups: good [diploid, del(20q), del(5q), -Y], intermediate (neither good or poor), and poor (chromosome 7 abnormalities and complex karyotypes) [10]. However, more recent studies using larger databases of patients with MDS have revealed that some rare chromosomal abnormalities that would normally be classified as intermediate have distinct prognostic significance, such as del(12p) (median survival 108 months) and +19 (median survival 19.8 months) [25]. Another limitation of the IPSS is that it underestimates the prognostic impact of karyotype relative to the percentage of blast cells. In an analysis of 3169 patients with MDS from Europe and the United States, Cox proportional hazards regression revealed that the relative hazard ratio (HR) for OS was 3.88 for complex karyotype with chromosome 5/7 abnormalities and 2.09 for sole -7/del (7q), whereas the HR for elevated blasts was 1.5 for 5% to 10% blasts, 1.64 for 11% to 20% blasts, and 1.81 for greater than 20% blasts. However the IPSS gives only 1.0 point for a poor-risk karyotype but 1.5 points for 11% to 20% blasts and 2.0 points for 20% to 29% blasts [27]. Thus, a revision of the cytogenetic grouping of the IPSS is needed, with reclassification of the risk group of specific chromosomal abnormalities and an increase in the impact of karyotype relative to the blast percentage. Recent publications analyzing large datasets of patients with MDS have provided revised cytogenetic classifications that predict survival outcomes more effectively than the IPSS; these may serve as a framework for future prognostic models [25, 26].

Even though G-banding metaphase cytogenetics is the standard method for evaluating karyotype in MDS, recent reports using techniques with higher resolution (comparative genomic hybridization [CGH] array; single nucleotide polymorphism [SNP] array) have revealed a wealth of additional chromosomal abnormalities in patients with MDS, characterized by microscopic gains and losses of chromosomal material and copy number-neutral loss of heterozygosity (CNN-LOH) due to uniparental disomy (UPD), even in patients with diploid karyotype by metaphase cytogenetics [28]. In one study, the investigators analyzed the genome of CD34+ cells from 174 patients with myeloid malignancies (including 74 patients with MDS) with 250 K SNP arrays [28]. Chromosomal defects were detected in 78% of patients with MDS, compared with 59% of patients by standard metaphase cytogenetics. UPD was found in 20% of MDS cases. SNP arrays detected new lesions in 62% of patients with normal cytogenetics. Importantly, the detection of new chromosomal lesions by SNP arrays affected survival: among patients with IPSS Int-1 MDS, median survival was 9 months for those with SNParray abnormalities versus 28 months for those without abnormalities. Future prognostic models in MDS should incorporate new technologies for detecting chromosomal defects, as doing so may lead to increased recognition of poor-risk patients.

Transfusion Dependency

The development of transfusion-dependent anemia has been associated with worse survival in patients with MDS [12••, 20]. One Italian study analyzed 467 patients with de novo MDS and demonstrated that development of transfusion dependency, number of transfused units of red blood cells (RBCs), and rate of RBC transfusion were associated with worse OS and leukemia-free survival in both univariate and multivariate analysis [20]. The effect was more relevant in patients without an increase in the percentage of blasts (RA, RARS, RCMD, 5q- syndrome). Development of iron overload (defined by a ferritin serum level >1,000 ng/mL) was also associated with worse survival. Other studies have confirmed the impact that transfusion dependency has on the outcome of patients with MDS [12••, 29].

It is still unknown why transfusion dependency affects survival in these patients. It could be a marker for a more aggressive disease phenotype, it could be related to the deleterious effect of iron overload, or both. In patients with MDS, the clinical impact of iron overload appears to be more relevant in patients with low-risk disease, and transfusiondependent patients with iron overload (ferritin >1,000 ng/mL) have a higher rate of death due to cardiac complications [20, 30]. However, published studies of T2* MRI failed to demonstrate an increase in myocardial iron content in most transfusion-dependent patients with MDS [31]. It could be that the development of cardiac iron overload requires a greater extent of transfusion dependency, as in one report only patients with a history of transfusion of more than 100 units of RBCs had signs of cardiac iron overload on MRI [31]. Additionally, it is possible that iron has other subtle effects on the outcomes of patients with MDS. Some studies have suggested that increased iron body stores are associated with increased infections, particularly fungal infections [32]. Iron overload may also contribute to leukemic evolution. Sanz et al. [29] reported that iron overload was an independent factor for leukemia-free survival in patients with MDS (HR, 6.6), and the effect was independent of the presence of transfusion dependency. In one retrospective report, iron chelation in patients with MDS was associated with improvement in OS and in the rate of leukemic transformation [33]. Ongoing randomized trials evaluating iron chelators in this disease may shed further light on this issue.

Age and Comorbidities

Older age is generally associated with a worse outcome in several types of cancer, including MDS [12••]. The reasons why older patients fare worse are probably more related to the performance status of the patient and the presence of comorbidities rather than to chronologic age per se. It can be difficult to assess the impact of comorbidities on survival; to that end, standardized comorbidities scores have been developed, including the Charlson Comorbidity Index (CCI) and the Hematopoietic Stem Cell Transplantation Comorbidity Index (HCT-CI), based on the CCI [34]. MDS is a disease of the elderly, but the impact of comorbidities on survival has been systematically investigated only recently. Most prognostic models (e.g., IPSS, WPSS) do not account for age and the presence of other diseases. Wang et al. [35] reported on 1708 patients with MDS diagnosed in the United States in 2001 and 2002, of whom 51% had identified comorbidities. In a multivariable regression analysis, the CCI significantly reflected survival: the HR was 1.19 for a CCI of 1 to 2 and 1.77 for a CCI of 3 or higher. Two other studies retrospectively evaluated the impact of the CCI and the HCT-CI in patients with MDS and concluded that increased comorbidities, as assessed by the HCT-CI, were independently associated with poor survival [36, 37]. Recently, Della Porta et al. [30] developed a specific comorbidity score to be applied to patients with MDS (MDS-CI), based on data from 840 patients. Relevant comorbidities included cardiac disease (2 points), moderate to severe liver disease (1 point), severe pulmonary disease (1 point), renal disease (1 point), and the presence of solid tumors (1 point). The MDS-CI was able to predict survival in a validation cohort of 504 patients, with a 2-year rate of nonleukemic deaths of 24% for patients with a low MDS-CI score (0 points), 42% with an intermediate score (1–2 points), and 61% with a high score (> 2 points).

Gene Mutations

Besides chromosomal defects, somatic gene mutations are also responsible for the pathogenesis of MDS. Compared with AML, relatively few gene mutations have been described in patients with MDS. Some of the more commonly mutated genes in MDS include RUNX1 (also known as AML1), TP53, TET2, ASXL1, and JAK2.

The gene encoding the transcription factor RUNX1, located at chromosome 21q22, can harbor mutations in patients with MDS [38]. Mutations of RUNX1 in MDS can occur in any portion of the gene but seem to cluster in the DNA binding domain and have been described in 14% of patients with MDS [38]. They are more frequent in patients with therapy-related MDS, and frequently occur in conjunction with chromosome 7 abnormalities [38]. Clinically, the finding of RUNX1 mutations in patients with MDS is associated with survival much poorer than for patients without RUNX1 mutations, but their independent impact on prognosis has not yet been shown [38].

The cell cycle checkpoint gene TP53 has been found to be mutated in roughly 13% of patients with MDS and is associated with a very poor outcome [39]. The prognostic value of TP53 mutations is independent of the IPSS, and these mutations are usually seen in patients with advanced MDS with complex karyotypes [39, 40].

TET2 mutations are found in roughly 20% of patients with MDS [41]. TET2 and family members TET1 and TET3 encode enzymes responsible for catalyzing the conversion of 5-methylcytosine to 5-hydroxymethylcytosine, which may be one of the enzymes responsible for DNA demethylation [42]. Mutations that can involve the TET2 gene include missense, nonsense, and frameshift mutations, and they seem to occur all over the gene [41]. Kosmider et al. [41] analyzed 99 MDS patients and reported that the presence of TET2 mutations was associated with improved outcome: 5-year OS was 76.9% versus 18.3% for patients without mutations (P=0.005), and 3-year leukemia-free survival was 89.3% versus 69.7% for patients without mutations (P=0.03). However, in another series of 355 patients with MDS and CMML, the presence of TET2 mutations had no impact on survival [43]. More studies are needed to clarify the influence of TET2 in MDS evolution.

DNA Methylation Abnormalities

Epigenetic abnormalities, including aberrant DNA methylation and histone covalent modification, are found in cancer cells and contribute to the malignant phenotype by suppressing expression of tumor suppressor genes and genes related to cellular differentiation. In MDS, abnormalities of DNA methylation have been described, and whole genome studies revealed that MDS and secondary AML are characterized by greater DNA methylation than in normal cells and de novo AML [44]. Recently, Shen et al. [45••] reported on the methylation profile of 10 different genes in 317 patients with MDS, the majority of whom were treated with hypomethylating agents (decitabine, azacitidine). The methylation levels of individual genes were combined to obtain a methylation score. Patients characterized as having a high methylation score had worse OS (12.3 months vs. 17.5 months; P=0.04) and progression-free survival (6.4 months vs. 14.9 months; P=0.01). In multivariate analysis, the methylation score was independent of the IPSS score.

Recent Prognostic Models in MDS

WHO Classification-Based Prognostic Scoring System

The WPSS model is a dynamic prognostic model that can be applied both at diagnosis and at any time during disease progression [11]. The learning cohort for the model consisted of 426 patients with MDS treated at the University of Pavia in Italy. Patients with CMML and t-MDS were excluded. In a previous report, the same group demonstrated that the most important variables affecting survival were WHO subtype, karyotype, and transfusion dependency [20]. Accordingly, multivariable Cox regression analysis for OS and risk of AML was performed with these three variables included as time-dependent continuous variables. Cytogenetic abnormalities were grouped as per the IPSS criteria. WHO subtypes were grouped into RA/RARS/del(5q), RCMD/RCMD-RS, RAEB-I, and RAEB-II. These three variables were used to define the WPSS prognostic system (Table 2). Patients were effectively divided into five risk groups: very low (0 points), low (1 point), intermediate (2 points), high (3-4 points), and very high (5-6 points). The WPSS efficiently differentiated these five groups regarding survival and time to AML (Table 2). The score was validated in a cohort of 739 German patients. Furthermore, comparison with age-adjusted mortality of the Italian population revealed that patients in all risk categories of the WPSS except for very low risk had higher mortality than the general population. A later report confirmed that the WPSS predicted survival and probability of relapse in patients who underwent allogeneic SCT [46].

Table 2. WHO Classification–Based Prognostic Scoring System (WPSS) prognostic model for myelodysplastic syndromes.

Variable	Points
	0	1	2	3
WHO Subtype	RA, RARS, del (5q)	RCMD, RCMD-RS	RAEB-I	RAEB-II
Karyotypea	Good	Intermediate	Poor	-
Transfusion Dependencyb	No	Yes
Risk Group	Points	Median Overall Survival (mo)c	% Evolution to AML at 2 yearsc
Very Low	0	103-141	0-3%
Low	1	66-72	6-11%
Intermediate	2	40-48	21-28%
High	3-4	21-26	38-52%
Very High	5-6	9-12	79-80%

Legends

^aGood: Normal, del(5q), del(20q), -Y; Intermediate: all others; Poor: complex (> 3 abnormalities), chromosome 7 abnormalities;

^bhaving at least one transfusion every 8 weeks over a period of 4 months;

^cvalues are given for both learning and validation cohort

M.D. Anderson Cancer Center Prognostic Model

Kantarjian et al. [12••] proposed a new prognostic model that accounts for variables not originally included in the IPSS. A total of 1915 patients with MDS were divided into a training cohort (958 patients) and validation cohort (957 patients). Patients with prior treatment for MDS, secondary MDS, and CMML were included. By Cox proportional hazards multivariable regression, the following covariates were associated with OS: older age, poor performance status, higher percentage of BM blasts, chromosome 7 abnormalities, complex karyotype, thrombocytopenia, leukocytosis, and prior transfusions. Based on these factors, the authors proposed a new risk model for MDS (Table 3).

Table 3. M.D. Anderson prognostic model for myelodysplastic syndromes.

Prognostic Factor	Points
Performance Status ™ 2	2
Age, years
•60-64	1
•™65	2
Platelets, ×109/L
•<30	3
•30-49	2
•50-199	1
Hemoglobin 12 g/dL	2
Bone Marrow Blasts, %
•5-10	1
•11-29	2
White Blood Cell Count > 20×109/L	2
Karyotype: Chromosome 7 abnormality or complex ™ 3 abnormalities	3
Prior Transfusion, yes	1
Score	Median OS, Months
Low (0-4 points)	54
Intermediate-1 (5-6 points)	25
Intermediate-2 (7-8 points)	14
High (™ 9 points)	6

OS overall survival (Data from Kantarjian et al. [12••].)

Patients were grouped into four risk categories, similar to the IPSS: low (0–4 points), intermediate-1 (5–6 points), intermediate-2 (7–8 points), and high (≥ 9 points). The M. D. Anderson model can be applied at any time during the disease course, and it includes all patients with MDS, regardless of previous treatment. The new model was able to predict survival within each subgroup of the IPSS, demonstrating that it added prognostic power to the IPSS. Conversely, the IPSS could not discriminate survival within the four risk groups of the new model. Subgroup analysis revealed that the model was valid in patients with less than 20% blasts and in patients with CMML. An independent validation of the M.D. Anderson model compared it against the WPSS and the IPSS. Among 1,074 patients who could be classified by all three prognostic groups, the M.D. Anderson model best identified the lowestrisk patients, and it could be applied to a broader group of patients [47].

Other Models

More recently, several reports have been published that describe prognostic models applied to specific subsets of patients with MDS. Garcia-Manero et al. [48•] analyzed the outcomes of 856 patients with low/Int-1 risk MDS by the IPSS. They identified five variables (platelet count, percentage of blasts, age, hemoglobin, and karyotype) that were used to built a model that predicted survival among patients with low-risk MDS. Survival ranged from 80.3 months in the lowest-risk group to 14.2 months in the high-risk group, similar to patients with Int-2 risk in the M.D. Anderson model. Similar prognostic models analyzing patients with t-MDS and hypoplastic MDS were recently presented [49, 50]. Overall, these studies indicate that the current version of the IPSS is suboptimal, as it is possible to identify patients with very different outcomes among each IPSS category. Ideally, there should be no need for independent prognostic models for each subgroup of patients with MDS, unless clinical and biologic evidence indicates that the entity being studied has a different biology and pathophysiology.

Conclusions

At the present time, there are two new prognostic models for MDS in addition to the IPSS. Both try to correct deficiencies of the IPSS and incorporate new risk factors that have been described in the past decade. However, they still have limitations: the WPSS excludes certain groups of patients, and the M.D. Anderson model did not include the WHO subtypes in the analysis. Currently, a revised version of the IPSS is being developed. Future prognostic models should ideally include most important variables in the analysis to truly discern which have independent significance. Clinically useful risk factors ideally should have consistent effect across different populations of patients with MDS, should be reproducible among different centers, and should be widely available (i.e., not limited to research centers). Additionally, it is important to remember that prognostic models can be affected by treatment, which may change the prognostic value of a previous known risk factor. As our knowledge of the pathophysiology of MDS and its treatment evolves, we can expect that our search for better prognostic models in this disease will continue.