Myelodysplastic syndromes: 2018 update on diagnosis, risk-stratification and management

Received: 2 October 2017 Accepted: 2 October 2017 DOI: 10.1002/ajh.24930 ANNUAL CLINICAL UPDATES IN HEMATOLOGICAL MALIGNANCIES Myelodysplastic syndromes: 2018 update on diagnosis, risk-stratification and management Guillermo Montalban-Bravo Guillermo Garcia-Manero Department of Leukemia, University of Texas MD Anderson Cancer Center, Houston, Texas Correspondence Guillermo Garcia-Manero, Department of Leukemia, MD Anderson Cancer Center, Box 428, 1515 Holcombe Blvd, Houston, TX 77030. Email: ggarciam@mdanderson.org Abstract Disease overview: The myelodysplastic syndromes (MDS) are a very heterogeneous group of myeloid disorders characterized by peripheral blood cytopenias and increased risk of transformation to acute myelogenous leukemia (AML). MDS occurs more frequently in older males and in individuals with prior exposure to cytotoxic therapy. Diagnosis: Diagnosis of MDS is based on morphological evidence of dysplasia upon visual examination of a bone marrow aspirate and biopsy. Information obtained from additional studies such as karyotype, flow cytometry or molecular genetics is usually complementary and may help refine diagnosis. Risk-stratification: Prognosis of patients with MDS can be calculated using a number of scoring systems. In general, all these scoring systems include analysis of peripheral cytopenias, percentage of blasts in the bone marrow and cytogenetic characteristics. The most commonly used system is probably the International Prognostic Scoring System (IPSS). IPSS is now replaced by the revised IPSS-R score. Although not systematically incorporated into new validated prognostic systems, somatic mutations can help define prognosis and should be considered as new prognostic factors. Risk-adapted therapy: Therapy is selected based on risk, transfusion needs, percent of bone marrow blasts and cytogenetic and mutational profiles. Goals of therapy are different in lower risk patients than in higher risk. In lower risk, the goal is to decrease transfusion needs and transformation to higher risk disease or AML, as well as to improve survival. In higher risk, the goal is to prolong survival. Current available therapies include growth factor support, lenalidomide, hypomethylating agents, intensive chemotherapy and allogeneic stem cell transplantation. The use of lenalidomide has significant clinical activity in patients with lower risk disease, anemia and a chromosome 5 alteration. 5-azacitidine and decitabine have activity in both lower and higher-risk MDS. 5-azacitidine has been shown to improve survival in higher risk MDS. A number of new molecular lesions have been described in MDS that may serve as new therapeutic targets or aid in the selection of currently available agents. Additional supportive care measures may include the use of prophylactic antibiotics and iron chelation. Management of progressive or refractory disease: At the present time there are no approved interventions for patients with progressive or refractory disease particularly after hypomethylating based therapy. Options include participation in a clinical trial or cytarabine based therapy and stem cell transplantation. Am J Hematol. 2018;93:129 147. wileyonlinelibrary.com/journal/ajh VC 2017 Wiley Periodicals, Inc. 129

130 MONTALBAN-BRAVO AND GARCIA-MANERO FIGURE 1 Cytogenetic classification of MDS. Adapted from Schanz et al 9 [Color figure can be viewed at wileyonlinelibrary.com] 1 DISEASE OVERVIEW MDS comprises a very heterogeneous group of myeloid malignancies with very distinct natural histories. 1 3 MDS occurs in 3 to 4 individuals per 10 5 in the US population. 4 Prevalence increases with age. For instance in individuals age 60 and above, prevalence is 7 to 35 per 10 5. 4 Other series have reported higher rates. 5 MDS affects more frequently males than females. 4 Exposure to prior chemo or radiation therapy is a risk for the development of MDS. MDS is usually suspected by the presence of cytopenia on a routine analysis of peripheral blood. This prompts evaluation of bone marrow cell morphology (aspirate) and cellularity (biopsy). A manual count of bone marrow blasts is fundamental for risk assessment. Cytogenetic analysis helps in predicting risk and in selecting therapy. Once this information is collected, the risk of the patient can be calculated. At the present time, the International Prognostic Scoring System (IPSS) 6 is still commonly used. Natural history and therapeutic decisions are different for patients with lower risk disease (low and INT-1) versus those with higher (INT-2 and high). In lower risk disease interventions have been traditionally developed to improve transfusion needs; whereas higher risk options were initially modeled following therapy of AML with remission induction being the goal. This concept was modified by the better understanding of the natural history of MDS and the development of new therapies, in particular the hypomethylating agents. Another important concept is that a large majority of patients with MDS die from causes intrinsic to the disease and not due to progression to AML. 7 This has important implications for the development of therapies in MDS. The revised IPSS score (IPSS-R) was published in 2012. 8 This score includes a new cytogenetic risk classification that divides patients into 5 cytogenetic categories (Figure 1). 8 10 The IPSS- R divides patients in 5 risk subgroups including an intermediate group. Because current therapies were approved using either FAB or IPSS criteria, we still use IPSS to decide therapy but IPSS-R to calculate prognosis. Although the use of next generation sequencing has not been incorporated into clinical practice in a validated manner, presence of known recurrent somatic mutations can be helpful and refine diagnosis and prognosis of patients with MDS 11 21 and, therefore, we recommend using this data in therapeutic decision making if available. 2 DIAGNOSIS The diagnosis of MDS is generally suspected based on the presence of an abnormal CBC. Diagnosis is confirmed by performing a bone marrow aspiration and biopsy. Both procedures provide different information. The bone marrow aspirate allows for detailed evaluation of cellular morphology and evaluation of percent of blasts. The bone marrow biopsy allows for determination of bone marrow cellularity and architecture. Diagnosis is established by the presence of dysplasia. A number of morphological classifications are in place to classify patients with MDS. The most recent one being the 2008 WHO version 22 which was recently revised in 2016. 23 There is some controversy regarding the utility of bone marrow biopsy. We believe that biopsy helps in diagnosis and potentially in selecting therapy. For instance, in our practice we only use immunotherapy (ATG, cyclosporine, steroids) in hypocellular MDS (marrow cellularity less than 20 to 30% depending on age). Cellularity is better assessed by bone marrow biopsy. A number of additional tests are needed to complete the laboratory evaluation of a patient with MDS. Most important of which is the analysis of bone marrow cytogenetics. It is well established that cytogenetic patterns are very heterogeneous in MDS. 24 Cytogenetics are

MONTALBAN-BRAVO AND GARCIA-MANERO 131 of importance to calculate prognosis of patients and in some subsets of patients to select the most effective form of therapy. The most recent cytogenetic risk classification in MDS includes 5 different subgroups including 20 different alterations (Figure 1). 9 Cytogenetic patterns are not stable in MDS and a significant fraction of patients acquired additional cytogenetic changes. This phenomenon is associated with increased risk of transformation to AML and worse survival. 25 This data has implications for the follow up of patients at risk for clonal cytogenetic acquisition. A number of other assays can be used to help in the diagnosis of MDS. These include the use of flow cytometry, fluorescent in situ hybridization (FISH) and genomic sequencing techniques. Flow cytometry can help in the identification of abnormal phenotypic patterns and can be of help in cases of minimal dysplasia. Because the significant heterogeneity of cytogenetic alterations in MDS, there is no evidence that a panel of FISH probes could replace routine 20 metaphase cytogenetic analysis. Thus, in our opinion FISH and flow cytometry should not be considered part of the standard work up evaluation procedure of the patient with MDS and should be used in specific situations. More recently, and thanks to the discovery of multiple genetic mutations in MDS, we now have access to clinical tools for molecular annotation of patients with MDS. These can be of use to complement the other diagnostic tools to define and confirm diagnosis as, the vast majority of MDS patients, will carry 1 or more somatic mutations in genes involved in DNA methylation, chromatin regulation, RNA splicing, transcription regulation, DNA repair, cohesin function or signal transduction. 19,26 Although none of these mutations are part of diagnostic criteria of MDS, with the exception of SF3B1 mutations (which define MDS with ring sideroblasts in the presence of 5% ring sideroblasts 23 ), specific mutation patterns closely correlate to specific MDS subtypes and may define diagnosis in the future. 17 2.1 Diagnostic problems in MDS: CHIP, ICUS, and CCUS Because diagnosis of MDS is based on morphological assessment it can be subjective particularly in patients with early low risk disease. It is calculated that diagnostic discrepancy can occur at the time of initial presentation in close to 20% of patients. 27 This has obvious implications for therapeutic decision making and patient counseling. In general, diagnosis is obvious in patients with excess blasts. The problem is in patients without excess blasts were diagnosis is based on dysplasia. Clinical assessment is needed in patients with minimal or not diagnostic evidence of dysplasia. In these cases, it is recommended that other causes of cytopenia be excluded. Routine tests include the analysis of anemia and thrombocytopenia, and exclusion of cause of blood loss or inflammatory processes. When suspected, evaluation of GI tract needs to be considered. Once other potential causes of cytopenia are excluded, additional diagnostic tools including cytogenetic evaluation, flow cytometry and, more recently, DNA sequencing, can help define the diagnosis and predict patient outcomes. Patients with cytopenia but no dysplasia are considered in the subset of idiopathic cytopenia of undetermined significance (ICUS). A fraction of these patients may have cytogenetic abnormalities and, up to 36% may carry 1 or more somatic mutations in genes recurrently mutated in myeloid malignancies. 11,16 The presence of a somatic mutation in the context of cytopenias with no diagnostic criteria for MDS is now considered a clonal cytopenia of undetermined significance (CCUS). This distinction is supported by recent data which suggests that, although close to 25% of patients with ICUS may ultimately develop MDS or AML, this risk significantly increases in the presence of a clonal mutation from 9% to 82% at 5 years, particularly in the presence of highly predictive mutation patterns. 14,16 Further studies are required to refine and confirm these observations and determine the best therapeutic approach. Therefore, close observation of these patients is recommended at the present time. A detailed evaluation and careful differential diagnosis between ICUS, CCUS, and MDS is therefore essential (Table 1). Recently, the presence of clonal somatic mutations have been reported in hematopoietic cells from older individuals without evidence of hematological disorder 28,29 as part of age-related clonal hematopoiesis of undetermined potential (CHIP). In addition, clonal hematopoiesis has also been identified as a risk factor therapy-related MDS and AML. 30 Patients with evidence of clonal hematopoiesis are also at increased risk to develop MDS 29 but the clinical implications of these findings are less clearly understood at this point. 31 Finally, a subset of patients of importance are those with MDS/ MPN features. These are patients with evidence of a myeloproliferative component (with or without fibrosis). At the present time, with the exception of chronic myelomonocytic leukemia, we do not fully understand the natural history of patients with MDS/MPN. 32 In our center, they are currently treated as MDS but studies are ongoing to clarify this issue. Of note, specific MDS/MPN subtypes have particular mutational profiles and may benefit from specific therapeutic approaches. An example includes patients with MDS/MPN with ring sideroblasts and thrombocytosis who typically have mutations in SF3B1 and JAK2 23,33 and may have good response to therapy with lenalidomide. 34 3 RISK STRATIFICATION The prognosis of patients with MDS is very heterogeneous and thus the need to develop prognostic systems that allow risk stratification and help in the timing and choice of therapy. Apart from the intrinsic prognostic value of morphological classifications, 22 anumberofprognostic scores are currently in use in MDS. The IPSS 6 has been in place since 1997. The prognostic score includes percent of blasts, number of cytopenias and cytogenetics. IPSS is of fundamental importance as it has allowed the prognostication of patients for over 2 decades. This system is highly reproducible and very simple to use. The system has several limitations that have become evident over the years. The most important one is that it is not a very precise predictor of prognosis in patients with lower risk disease and that it attributes relatively little weight to cytogenetics. The IPSS-R includes different cut off points of cytopenias and incorporates the new cytogenetic MDS score. 8 IPSS-R should be the standard tool to assess risk although limitations still exist for its use as no drug therapy has been approved yet using IPSS-R. A

132 MONTALBAN-BRAVO AND GARCIA-MANERO TABLE 1 Differential diagnosis of clonal or idiopathic cytopenias and myelodysplastic syndromes 11,16,21,26 Features CHIP IDUS ICUS CCUS MDS Cytopenias No No Yes (1 or more) Yes (1 or more) Yes (1 or more) Dysplasia No Yes in <10% bone marrow cellularity None or minimal (non diagnostic for MDS) None or minimal (non diagnostic for MDS) Yes (>10% of elements per lineage in at least 1 lineage) Somatic mutations Yes at a variant allele frequency 2%. Most commonly: DNMT3A, TET2, ASXL1, SRSF2, TP53. May appear associated with clonal hematopoiesis No. ICUS defined by absence of clonality Up to 36% overall with similar mutation VAF compared to MDS* 17% of ICUS without dysplasia 45% of ICUS with some dysplasia Up to 85% of patients Risk of progression Very low (0.5-1% per year) outside of therapy related setting. Unknown Up to 10% at 5 years Up to 80% at 5 years but determined by mutational patterns - CHIP: Clonal hematopoiesis of indeterminate potential. IDUS: Idiopathic dysplasia of undetermined significance. ICUS: Idiopathic cytopenia of undetermined significance. CCUS: Clonal cytopenia of undetermined significance. MDS: Myelodysplastic syndromes. new molecular IPSS system is expected soon. Several studies confirm the added value of mutational data in risk stratification when compared to the current prognostic models. 19,35,36 Both the IPSS and IPSS-R were developed in a very specific subset of patients: newly diagnosed patients at the time of initial presentation. Patients with proliferative features and CMML or that have received prior therapy were excluded. 6 To overcome these limitations, the global MDACC model was developed. 37 This model is summarized in Table 2. The importance of the global MDACC model is that it allows evaluation of all patients that are considered as MDS at any time during their course of their disease without needed WHO evaluation. It has also become apparent that the natural history of patients with lower risk disease is very heterogeneous. 38 We evaluated outcomes in a large series of patients with low or int-1 disease by IPSS. We found that prognosis varied significantly in patients with lower risk MDS and were able to develop a lower risk MDS specific prognostic score (Table 3). This has significant implications for the development of specific interventions for patients with lower risk disease. This model has been validated on several occasions and it is being used to identify patients with poor prognosis lower risk disease that could be candidates for early intervention. Bejar et al published data indicating that patients with poorer prognosis and lower risk disease accumulate a higher number of mutational events than the better risk counterpart. 39 This data provides a potential molecular basis for the identification of this group of patients and poorer prognosis. MDS occurs in older patients that suffer from comorbidities more frequently. None of the systems discussed above include impact of comorbidity to the calculation of the natural history of MDS patients. To study this issue, we used a comprehensive comorbidity score known as ACE-27 in a cohort of 500 patients with MDS. 40 Presence of comorbidity had a significant independent impact on survival and a prognostic score could be developed that included age, IPSS and ACE- 27 score. This data indicates the need to add comorbidity scores in MDS. Recently, other groups have confirmed the importance of comorbidity scores in MDS. 41 We recently confirmed that ACE-27 adds to the prognostic value of IPSS-R. 42 Additional prognostic scoring systems include systems for hypocellular MDS 43 and for therapy related MDS. 44 However, conventional risk stratification models such as IPSS-R, MDACC model or the WPSS seem equally valid in therapy related disease. 45 It should be noted that prognosis in t-mds is strongly associated with presence of cytogenetic alterations: diploid patients have prognosis not dissimilar to that of de novo patients. 44 TABLE 2 The global MDACC MDS prognostic model 37 Prognostic factor PS 2 2 Age 60 64 1 >64 2 Platelets 310 9 /L <30 3 30 49 2 50 199 1 Hemoglobin <12 g/dl 2 BM blast % 5to10 1 11 to 19 2 WBC > 20 310 9 /L 2 Alteration of chromosome 7 or 3 alterations 3 Prior transfusion 1 Points PS, performance status; BM, bone marrow; WBC, white blood cell count. Patients with 0 to 4 points had a median survival of 54 months and a 3 year 63% survival. Patients with 5 and 6 points had a median survival of 23 to 30 months and 3-year survival of 30 to 40%. Patients with 7 to 8 points had a median survival of 13 months and a 3-year survival rate of 13 to 19%. Patients with 9 or more points had a median survival of 5 to 10 months and a 2% 3-year survival. Adapted from ref 19. 37

MONTALBAN-BRAVO AND GARCIA-MANERO TABLE 3 MDACC MDS lower risk prognostic model 38 Characteristics Unfavorable cytogenetics 1 Age 60 years 2 Hemoglobin < 10 (g/dl) 1 Platelets < 50 3 10 9 /L 2 50 200 3 10 9 /L 1 Bone Marrow Blasts 4% 1 Points Score Median Survival 4-year OS (%) 0 NR 78 1 83 82 2 51 51 3 36 40 4 22 27 5 14 9 6 16 7 7 9 N/A Characteristics were selected from multivariate analysis model in patients with lower risk MDS. Each characteristic is associated with a number of points. Score is calculated by adding all points. Each score allows calculation of median survival (in months) and probability of survival at 4 years. Adapted from reference 20. 38 4 CYTOGENETIC AND MOLECULAR ALTERATIONS Over the last 3 years, a number of very important studies have been published describing comprehensive analysis of the incidence and 133 clinical impact of multiple genetic lesions in MDS. 46 Bejar et al first published an analysis of 18 genes using different techniques on 439 patients. 47 Subsequently, two major studies have described the mutational landscape of MDS in larger series analyzing more genes 19,48 (Figure 2). For instance, an European consortium has reported an analysis of mutations in 111 genes using next generation sequencing technology in a cohort of 738 patients. 48 Frequency of common mutations is shown in Table 4. Since then, multiple other studies describing the mutational landscape of MDS and its potential prognostic and therapeutic implications have been published. Despite the heterogeneity of some of these studies, mutations in genes such as RUNX1, 19,26,36 TP53, 19,20,26,35,36 or EZH2 19,36 have consistently been associated with adverse prognosis while mutations in the splicing factor SF3B1 are associated with very favorable outcomes and prolonged survival. 12,26,36 Although mutation burden and interaction are likely to further determine prognosis, further evidence is required to validate and apply this in the clinical setting. This emerging data will have a significant impact not only on our ability to prognosticate patients with MDS but also in potentially selecting therapy. For instance, it has been reported that patients with TET2 mutations may have higher response rates to azacitidine than those without mutation. 52,53 Similarly, presence of TP53 mutations was recently associated with high response rates to a 10- day schedule of decitabine. 18 However, previous reports suggest similar, but not superior, response rates to hypomethylating agents in patients with MDS and TP53 mutations compared to those with wildtype TP53. 53 56 In addition, the results of molecular mutations is aiding in the development of clinical trials for instance for patients with mutations in splicing genes (SF3B1, SRSF2, U2AF1, ZRSR2), IDH1, IDH2, FLT3, and RAS. In fact, some of these targeted therapies have been recently approved by the FDA for AML and similar approvals will likely follow for patients with MDS in the near future. Recently, two major studies 13,15 evaluating the prognostic value of somatic mutations in FIGURE 2 Distribution of mutational events in MDS. Adapted from Papaemmanuil et al 48 [Color figure can be viewed at wileyonlinelibrary.com]

134 MONTALBAN-BRAVO AND GARCIA-MANERO TABLE 4 Reported frequency of genetic lesions in MDS 47,49 51 Gene % Location Function SF3B1 28 2q33 Splicing factor TET2 21 4q24 Control of cytosine hydroxymethylation ASXL1 14 20q11 Epigenetic regulator SRSF2 12 17q25 Splicing factor RUNX1 9 21q22 Transcription factor TP53 8 17p13 Transcription factor U2AF1 7 21q22 Splicing factor EZH2 6 7q36 Polycomb group protein NRAS 4 1p13 Signal transduction JAK2 3 9p24 Tyrosine kinase ETV6 3 12p13 Transcription factor CBL 2 11q23 Signal transduction IDH2 2 15q26 Cell metabolism, epigenetic regulation NPM1 2 5q35 Phosphoprotein IDH1 1 2q33 As IDH1 KRAS <1 12p12 Signal transduction GNAS <1 20q13 G protein PTPN11 <1 12q24 Protein phosphatase BRAF <1 7q34 Raf kinase PTEN <1 10q23 Phosphatase CDKN2A <1 9q121 Cell cycle control 5.1.1 Erythroid growth factor support The use of erythroid stimulating agents (ESA) is common practice in community practice and in Europe. 58 It should be noted that no randomized study has ever proven that this intervention positively affects the survival of patients with MDS. A number of ESAs are available. Reported response rates range from 30 to 60% depending on study. 59 Data from the Swedish group indicated that addition of G-CSF to erythropoietin increases responses rates. In a retrospective observational study, early introduction of this combination in patients with low risk disease and minimally transfusion dependent patients had an impact on survival. 60 The Swedish group also developed an algorithm to predict response to ESA. 61 The French group also evaluated the impact of ESA on survival in a retrospective study of 284 patients and compared it to a group of patients that formed the IPSS cohort. 62 In this study patients treated with ESA had a better survival (HR for death was 0.43, 95% CI 0.25-0.72). 62 Results from a recently published randomized placebo-controlled phase 3 study of darbepoetin alfa for 147 patients with lower-risk MDS showed significantly higher erythroid responses (14.7% versus 0%, P 5.016) and reduction in transfusion incidence in weeks 5 24 of therapy (36% versus 59.2%, P 5.008) with darbepoetin compared to placebo, with no significant differences between groups in the incidence of thromboembolic events, solid malignancies or transformation to AML. 63 However, questions on potential tumorigenic effect of these drugs has resulted in increased scrutiny of their use. G-CSF is not approved by the FDA for patients with anemia of MDS in the US. The use of thrombopoietin agonists for patients with lower-risk MDS and thrombocytopenia has been explored in several studies. Initial results of a study with romiplostin 64 questioned its use in patients predicting outcomes after allogeneic stem cell transplantation concluded that mutations in TP53, RUNX1, ASXL1, JAK2, and RAS pathway genes are associated with significantly shorter overall survival or relapse free survival, 57 with TP53 mutations being particularly adverse. 5 RISK ADAPTED THERAPY At the present time, we still use IPSS to decide the choice of therapy for an individual patient. Below is a summary of options and recommendations for specific subsets of patients. 2 A treatment algorithm is shown in Figure 3. 5.1 Options for newly diagnosed patients with lower risk MDS Therapy in this subset of patients is based on the transfusion needs of the patients. Patients that are transfusion independent are usually observed until they become transfusion dependent. Below is a description of agents currently available for patients with lower risk MDS. A new important concept in lower risk MDS is the idea of early intervention in patients with poor prognosis lower risk MDS. We are currently testing this concept in prospective clinical trials using very low doses of hypomethylating agents. FIGURE 3 Proposed treatment algorithm of patients with MDS. Once diagnosis is confirmed, patients are divided into a lower and a higher risk category. Options for patients with lower risk disease include growth factors, lenalidomide and azanucleosides. Treatment in general is sequential: patients that do not respond to growth factors can be treated with lenalidomide or azanuclesoides, if appropriate. Patients that fail lenalidomide can subsequently be treated with azanucleosides. There is little experience in terms of outcomes with this approach. Patients that fail all three therapies should be considered for allosct and/or clinical trial. For patients with higher risk MDS options are allosct, AML-like therapy or azanucleoside. Prognosis of patients that fail any of these approaches is poor, particularly for those exposed to azanucleosides. In this setting allosct and clinical trial should be strongly considered [Color figure can be viewed at wileyonlinelibrary.com]

MONTALBAN-BRAVO AND GARCIA-MANERO 135 with lower risk MDS, particularly due to complications related to disease transformation and marrow fibrosis. Results from an extension study evaluating the use of romiplostin monotherapy in 60 patients with Low or Int-1 IPSS and thrombocytopenia (defined as platelet count of 50 3 10 9 /L) reported 57% platelet responses with a median response duration of 33 weeks and only 2 patients with progression to AML. 65 A number of studies are evaluating eltrombopag in MDS, another TPO agonist. Results from the phase 1 single-arm, randomized portion of the phase 2 superiority EQoL-MDS study 66 evaluating the efficacy and safety of eltrombopag in patients with lower-risk MDS where recently published and reported significantly higher platelet responses (47% versus 3%, P 5.0017) and fewer bleeding events (14% versus 42%, P 5.0025) in the eltrombopag arm compared to placebo. No differences in leukemic transformation where observed (12% versus 16%, P 5.81). Although promising, the use of TPO agonists should be restricted to clinical trials until additional data is available. Recommendation We believe that a course of ESA with or without G-CSF is not contraindicated in most patients with low risk MDS with significant anemia without other cytopenia. Data indicates that early incorporation of these agents is more effective. We maintain therapy for at least 3 months to judge efficacy. In responding patients, we continue therapy until transfusion effect is lost. 5.1.2 Lenalidomide Lenalidomide is approved in for patients with lower risk MDS, anemia and alteration of chromosome 5. 67 Phase I results 68 were confirmed in a subsequent phase II study of lenalidomide in patients with anemia and alteration of chromosome 5. 68 In that study 148 patients received 10 mg of lenalidomide for 21 days every 4 weeks or daily. Of those, 112 had decrease need for transfusions (76%; 95% CI 68 82) and 99 patients (67%; 95% CI, 59 to 74) became transfusion independent. Response was fast: median time 4.6 weeks. The median rise of hemoglobin was 5.4 g/dl. Of interest, cytogenetic responses were observed in close to 50% informative patients. Predictors of response included presence of a platelet count of 100 3 10 9 /L and less than 4 prior units of red cells transfused. It should be noted that in this study patients with a platelet count of less than 50 3 10 9 /L were excluded. Subsequently, these results were confirmed in a phase III known as AZA-004 comparing two different doses of lenalidomide (5 and 10 mg orally daily) versus placebo. This study was also designed to clarify the issue of transformation to AML. 10 mg daily was the superior arm and that there was no increased incidence of transformation to AML in patients treated with lenalidomide. 69 Of importance, a report from the initial MDS-003 trial of lenalidomide indicated a longer survival for patients responding to therapy. 70 This is additional evidence that lenalidomide can change the natural history of patients with 5q- MDS. Although current data supports initiating therapy at the time of transfusion dependence, data from a subset analysis from the multicenter RevMDS study including 381 patients treated with 10 mg daily of lenalidomide suggests early intervention in transfusion independent patients with Hgb <10 g/dl may be associated with erythroid responses. In this study, among 12 TI patients with Hgb < 10 g/dl therapy with lenalidomide was associated with universal erythroid responses and improved QoL scores compared to pretreatment values (112.5; P 5.02). 71 However, additional evidence is required to change the current criteria for indication of therapy. Identifying patients at risk of treatment failure or transformation remains an essential aspect of management of 5q- MDS. Age <65 years, 72 bone marrow blast count >5% 73,74 and transfusion dependence have been associated with AML transformation. 73 More recently, several studies have suggested the role of TP53 mutations and karyotypic complexity in disease progression and outcome. 74 79 Mutations in TP53 can be detected at early stages prior to therapy in 12 17% of these patients, 75 77,79 and TP53-mutant clones can emerge and expand through the course of lenalidomide therapy and at the time of disease progression. 76,78 Various studies have reported lower response rates and lower likelihood of complete cytogenetic response in TP53- mutated compared with wildtype patients when treated with lenalidomide. 75,77,79 However, an analysis from a multicenter study including 67 patients with 5q-MDS treated with lenalidomide did not confirm differences in TI rates or complete cytogenetic response based on TP53 mutational status. 76 A trend to shorter response duration (6 versus 16 cycles, P 5 NS) as well as higher risk of transformation and shorter event-free and overall survival were observed among patients with TP53-mutated 5q- MDS. In parallel, lenalidomide has been investigated in patients without chromosome 5 alterations. 80 In an initial study, 214 patients received lenalidomide 10 mg orally daily or 10 mg on days 1 to 21 of a 28-day cycle. Fifty six (26%) patients achieved transfusion independence after a median of 4.8 weeks of therapy. Median response duration was 41 weeks. An international phase III, randomized, placebo-controlled study including 239 transfusion-dependent lower-risk non-del5q MDS patients confirmed the efficacy of lenalidomide in this setting. 81 Patients receiving lenalidomide 10 mg daily for 21 days where significantly more likely to achieve RBC transfusion independence for 8 weeks compared to placebo (26.9% versus 2.5%, P <.001). Median duration of TI was 30.9 weeks and TI was associated with significant improvement in HRQoL scores. In addition, higher response rates were observed in patients with baseline endogenous EPO 500 mu/ml. Of note, a recent study not only reported the negative impact of TP53 mutations in response outcomes to lenalidomide in del5q MDS, but also suggested U2AF1 mutations may be likewise associated with a lower likelihood of response both in del5q and non-del5q MDS and that mutations in DEAD box RNA helicase genes (DDX41, DDX54, and DHX29) may be enriched among responders in non-del5q MDS. 82 However, these findings have to be confirmed on larger studies. Recommendation The degree of response with lenalidomide in patients with lower risk MDS, anemia, good platelets and del5q makes it the standard of care for this subset of patients. This is further reinforced by the data on survival in responding patients. 70 We do not consider this agent in patients with thrombocytopenia. Based on the results of the recent phase III randomized study, lenalidomide can be considered an option

136 MONTALBAN-BRAVO AND GARCIA-MANERO for selected red cell-transfusion-dependent patients with non-del5q MDS. Evaluation of TP53 mutation status could be advisable in patients with del5q MDS prior to starting therapy with lenalidomide. Although current evidence does not support withholding therapy with lenalidomide in patients with TP53 mutations, these patients should be monitored closely for signs of inadequate response, loss of response or progression. 5.1.3 Azanucleosides Two azanuclesoides are approved for MDS: 5-azacitidine 83 and 5-aza- 2 0 -deoxycitidine (decitabine). 84 5-azacitidine is approved for all subsets of MDS whereas decitabine for those with INT-1 disease and above. However, these agents have not shown to modify the natural history of patients with lower risk disease. Different schedules of 5-azacitidine have been explored in MDS. In a community study a 5 day schedule was compared to a 7 day 5-2-2 schedule (weekend off) or a 5-2-5 schedule of 10 days. 85 Fifty patients were assigned to each arm (except 5-2-2 was 51 patients). Most patients had lower risk disease. Hematologic improvement was achieved by 44% to 56% of patients in each arm. Transfusion independency was documented in 50 to 64% of patients. There was a trend for better response rates and less toxicity with the 5-day schedule of 5-azacitidine. Therefore it is reasonable to use a shorter (5 day schedule) of 5-azacitidine in lower risk MDS. The results of phase 1 trial of an oral derivative of azacitidine have been published. 86 Results from the subsequent phase 2 component of the study, evaluating two extended dosing schedules of 14 and 21 days, where recently published and reported overall responses of 36 and 41% and achievement of transfusion independence in 31 and 38% of treated patients, respectively. 87 A multicenter randomized phase 3 trial is ongoing (NCT01566695). Until recently, there was relatively little data with decitabine in patients with lower risk disease. In the initial randomized trial of decitabine, 84 31 patients with INT-1 disease were treated. Four of 28 patients achieved some type of response. A randomized phase II trial was conducted exploring a 3-day subcutaneous schedule versus a weekly 3 3 monthly schedule. This study did not find significant differences between both arms, although it declared the daily x 3 schedule as an initial winner. 88 A randomized phase II trial comparing a 3-day schedule of azacitidine with a 3-day schedule of decitabine for lowerrisk MDS patients 89 reported high response rates (49% and 70% respectively, P 5.03) and cytogenetic response rates (61% versus 25%, P 5.02) with 32% and 16% of patients achievement of transfusion independence, respectively. With a median follow up of 20 months, the median overall-survival was not reached and the event-free survival was 18 months (Figure 4). Therapy was well tolerated with no early deaths and no significant grade 3 adverse events. Of note, therapy was able to negate the effect of negative prognostic factors as defined by the validated lower risk prognostic model. 36,38 An ongoing phase II randomized study (NCT02269280) of lower doses of decitabine (3-day schedule) or azacitidine (3-day schedule) versus a 5-day schedule of azacitidine in transfusion-dependent patients or BSC in transfusionindependent patients will determine whether these agents should be considered as standard of care in all patients or specific subgroups of patients with lower-risk MDS. Recommendation Both 5-azacitidine and decitabine are used in the US in patients with lower risk disease that are transfusion dependent. Most patients treated with these agents have failed or were not candidates for growth factor support or lenalidomide. Although results from the ongoing randomized phase II study will be needed to determine whether these agents can modify the natural history of lower-risk disease and be considered standard of care in the frontline setting or after growth factor support, recent data supports the use of these agents at lower doses particularly in patients with more adverse features. On the basis of previous studies, 86,87 the ongoing phase 3 study of oral azacitidine will also determine the role of this oral formulation in lower-risk disease. 5.1.4 Immune therapy This is an area of controversy. It is accepted that a subset of patients with MDS are characterized by deregulation of both cellular and innate immunity. 90 92 Based on this it will be logical that the use of immunemodulatory agents could have therapeutic benefit in MDS. The NIH group pioneered these approaches. Agents studied include antithymocyte globulin (ATG), cyclosporine, steroids. These therapies have been modeled after therapy of aplastic anemia. 90 The NIH group also developed an algorithm to predict response to this classes of agents. 93 This model included younger age, HLA-DR15 positivity, and shorter duration of transfusion dependency. Using this algorithm, the NIH group reported that alemtuzumab, an antibody against CD52, has significant activity in patients with MDS predicted to respond to immune suppressive therapy. 94 The group at Moffitt Cancer Center has suggested that a CD4/CD8 ratio could also be used to predict response. 95 Our group has not been capable to reproduce some of the data discussed above. Response rates with ATG observed at MDACC are significantly lower than those of the NIH. 96 The most important predictor for response for us has been the presence of marrow hypocellularity. 97 This is consistent with the results of Mufti et al in London. 98 Also recently the impact of ATG based therapy on survival has been questioned. Data from a Swiss study comparing ATG versus supportive care indicated a higher response rate but no survival benefit. 99 In this study, patients were randomized to a combination of horse ATG with cyclosporine versus best supportive care (BSC). Forty-five patients received ATG 1 CSA and 43 patients received BSC. By month 6, 13 of 45 patients on ATG 1 CSA had a hematologic response compared with four of 43 patients on BSC (P 5.0156). Despite higher response rates no significant effect on survival or transformation was observed. The NIH group has also reported on the clinical activity of eltrombopag in patients with aplastic anemia as salvage therapy 100 and in addition to standard immunosuppressive therapy. 101 This data was of importance due to the fact that responses were multilineage and not just restricted to platelets and significantly higher than those reported in historical cohorts with immunosuppressive therapy alone. A number of studies are investigating this agent in MDS.

MONTALBAN-BRAVO AND GARCIA-MANERO 137 FIGURE 4 Survival outcomes of patients with lower-risk MDS treated with azacitidine or decitabine. (adapted from ref. 89 ) (A) Kaplan-Meier estimates for event-free survival. (B) Kaplan-Meier estimates for overall survival [Color figure can be viewed at wileyonlinelibrary.com] Recommendation This is a particular difficult group of patients. A majority of older patients cannot tolerate ATG and most patients are treated with some form of supportive care that could include cyclosporine, growth factors and steroids. The impact of it is not known either. In younger patients with severe hypoplastic MDS, allogeneic stem cell transplantation (allosct) should be considered as soon as possible. For those that are not candidates, a combination with horse ATG is recommended. We cannot recommend the use of alemtuzumab at the present time until more data from other clinical trials is reported. The data on eltrombopag is of interest. 5.1.5 Allogeneic stem cell transplantation AlloSCT is usually not recommended in patients with lower risk disease even if they are young. This is based on data from Cutler et al using a Markov model. 102 The anticipated early mortality with allosct cannot be overcomed by the potential beneficial survival effect of allosct. This concept was confirmed by Koreth et al at recently (Figure 5). 103 Using another Markov model, the investigators analyzed the impact of reduced intensity transplant in older patients with MDS. Patients with lower risk disease did not benefited from this less toxic transplant approaches. 103 In a study describing the outcomes of 438 patients with lower-risk MDS after failure to hypomethylating agents, the overall survival of patients who underwent allosct was significantly longer to that observed in patients not receiving further therapy, those receiving conventional chemotherapy or investigational agents (median survival of 39 months versus 10 months, 28 months and 17 months, respectively). 104 Recommendation We generally do not recommend allosct in patients with lower risk disease at initial presentation. That said because of time required for donor identification, we refer all potential candidate patients for a

138 MONTALBAN-BRAVO AND GARCIA-MANERO FIGURE 5 Impact on survival of stem cell transplantation in MDS. (adapted from ref. 103 ). (A) Monte Carlo analysis for low/intermediate-1 International Prognostic Scoring System (IPSS) myelodysplastic syndromes (MDS). Simulated Kaplan-Meier survival plots (n 5 10,000; with log-rank P value) are indicated for the modeled 10-year time period, comparing a strategy of early reduced-intensity conditioning (RIC) transplantation (blue line) versus no early RIC transplantation (gold line). The results graphically indicate survival benefit of the nontransplantation strategy in low/intermediate-1 IPSS MDS quality-adjusted life expectancy (QALE): two-way sensitivity analysis. Two-way sensitivity analysis plot for the utilities of the Markov states alive after RIC transplantation and alive with MDS without RIC transplantation is shown. The gold area indicates the range in which nontransplantation therapy produces superior QALE. The blue area indicates the range in which RIC transplantation produces superior QALE. The red square indicates the plausible range of quality of life (QoL) for alive with low/ intermediate-1 IPSS MDS and for alive after RIC transplantation and does not cross the threshold line. This result is interpreted as insensitive, that is, the conclusion regarding benefit does not change within the plausible QoL range. (C) Monte Carlo analysis for intermediate-2/ high IPSS MDS. Simulated Kaplan-Meier survival plots (n 5 10,000; with log-rank P value) are indicated for the modeled 10-year time period, comparing a strategy of early RIC transplantation (blue line) versus no early RIC transplantation (gold line). The results graphically indicate survival benefit of the early RIC transplantation strategy in intermediate-2/high IPSS MDS. (D) Intermediate-2/high IPSS MDS QALE: twoway sensitivity analysis. Two-way sensitivity plot for the utilities of the Markov states alive after RIC transplantation and alive with MDS without early RIC transplantation is shown. The gold area indicates the range in which nontransplantation therapy produces superior QALE. The blue area indicates the range in which RIC transplantation produces superior QALE. The red square indicates the plausible range of QoL for alive with intermediate-2/high IPSS MDS and for alive after RIC transplantation and does not cross the threshold line. This result is interpreted as insensitive, that is, the conclusion regarding benefit does not change within the plausible QoL range. HCT, hematopoietic cell transplantation [Color figure can be viewed at wileyonlinelibrary.com] transplant consult in anticipation of future needs. Patients that are candidates for allosct and that had been exposed to multiple therapies (growth factors, lenalidomide, azanuclesoides, etc.) should be considered for transplantation. These patients are also candidates for clinical trials. Patients with hypoplastic MDS that are young should be considered for allosct up front. 5.1.6 Supportive care measures in MDS A number of interventions can be used in patients with MDS. These include the use of prophylactic antibiotics and iron chelation. No randomized data exists to make formal recommendation for any of these interventions. In our experience patients with isolated neutropenia and MDS not receiving therapy are not at significantly increased risk of infection to support recommendation of prophylactic antibiotics. Prophylactic antibiotics are commonly used in the context of active therapy for these patients. The role of iron chelation in MDS is more complicated. Data from thallasemias indicates that iron chelation has an important role in this setting. Iron accumulation is frequent in MDS. The consequences of this are not fully understood in MDS. A recent study evaluating the impact of iron chelation therapy on overall survival of transfusiondependent lower-risk MDS patients reported longer median overall survival from the time of transfusion-dependence in patients receiving chelation (5.2 years versus 2.1 years, P <.001). 105 This survival advantage remained after matched pair analysis adjusting for age, frailty, comorbidity and R-IPSS. However, there are no randomized

MONTALBAN-BRAVO AND GARCIA-MANERO 139 prospective studies demonstrating superior outcomes with iron chelation therapy and there is controversial data with regard to the serum ferritin threshold which should trigger chelation. 106 The NCCN guidelines recommend the use of chelation therapy in patients with ferritin levels above 2500 ng/ml. 107 In our practice, we do not see patients with MDS and iron deposition that develop liver cirrhosis or cardiomyopathy as frequently as reported by other groups. 108 Iron accumulation could have a role in transformation to AML 109 and in the increased risk of infectious complications known to occur on these patients. A large phase III (NCT00940602) study is evaluating the role of iron chelation in MDS. Recommendations We do not routinely recommend antibiotics in patients with isolated neutropenia and MDS that are not receiving some form of cytotoxic or immunosuppressive therapy. We use prophylactic antibiotics in patients receiving active therapy. We use iron chelation in patients with ferritin levels in excess of 2500 ng/ml but we consider all these patients for a clinical trial of iron chelation. A number of agents are being studied for lower risk MDS. These include oral azacitidine 86,87 and oral decitabine, agents that modulate TGF-B pathway such as luspatercept (ACE-536) and sotatercept (ACE- 011), 113,114 proteasome inhibitors and antagonists of Toll-like receptor signaling 92, 115 or immune checkpoint regulators (PD-1, PDL-1 or CTLA-4). Among these, luspatercept, a selective activin receptor ligand which traps GDF11 blocking TGF-b signaling and promoting late stage erythropoeitic maturation, has shown very promising activity. Although still preliminary, 116 results from the ongoing PACE-MDS phase 2 multicenter study evaluating the activity of different doses of luspatercept in patients with lower-risk MDS with anemia or transfusion dependence reveal 63% erythroid responses with 38% of patients achieving TI. Particularly high response rates were observed among patients with refractory anemia with ring sideroblasts (RARS) or SF3B1 mutation (HI- E: 77%, RBC-TI: 44%) and those with low transfusion burden, defined as <4 RBC units per 8 weeks, (HI-E: 65%, RBC-TI: 75%). These results have lead not only to continuation of the extension study but also to a randomized study of luspatercept for patients with RARS (NCT02631070). 5.2 Options for patients with relapsed or refractory lower risk MDS and investigational new agents in lower risk MDS Treatment of patients with lower risk MDS is sequential. A common practice is to start growth factor support and then consider lenalidomide or an azanucleoside. In a recent large, multicenter, retrospective study including 1698 patients with lower-risk non-del5q MDS, 110 early failure of ESAs was associated with a higher risk of AML progression. In a univariate and multivariable analysis, none of the second-line treatments, including HMAs, lenalidomide and investigational agents, seemed to improve the OS of these patients significantly. Although there are limitations to a retrospective analysis with high heterogeneity, this may suggest consideration of earlier intervention with disease modifying therapy (DMT) may be an option at least for some lower-risk patients. A retrospective study evaluating the impact of the timing of DMT on the likelihood of achieving TI among 508 TD lower-risk MDS patients suggested early intervention (3 months from start of TD) with lenalidomide or HMAs may be associated with higher rates of TI. 111 In a randomized phase III study comparing lenalidomide 10 mg daily for 21 days as single agent or in combination with ESAs in 131 RBC transfusion-dependent lower-risk ESA-refractory non-del5q MDS patients, 112 the erythroid response rates after 4 cycles of therapy where 23.1% an 39.4% respectively (P 5.044) with a median response duration of 18.1 and 15.1 months respectively (P 5.47). Although growth factor support should be considered frontline, this emerging data suggest early initiation of disease modifying agents may have to be considered as reasonable options. However, randomized clinical trials will be required to support this therapeutic approach. Patients that fail either lenalidomide or azanucleoside are candidates for clinical trials or allosct. There is no drug approved in the US for patients with MDS and HMA failure. Recently the survival of these patients was calculated to be between 14 to 17 months. 104 5.3 Options for newly diagnosed patients with higher risk MDS Options for patients with higher risk MDS have evolved significantly over the last decade. Before that period most patients were treated with some of cytarabine based therapy modeled after AML. The used of azanucleosides has modified this practice. 2 5.3.1 Azanucleosides Decitabine was studied in an initial randomized comparing it to BSC. 84 In this study the dose of decitabine was 15 mg/m 2 IV infused over 3 hours every 8 hours for 3 days (at a dose of 135 mg/m 2 per course) and repeated every 6 weeks. Although there was no clear benefit in terms of survival in this study, the use of decitabine was associated withacompleteresponserateof9%andoverallresponserateof17%. These results led to the approval of decitabine in the US. Based on the results of a phase I trial of decitabine performed at MDACC, a Bayesian randomized phase II trial of three different doses and schedules of decitabine was conducted. 117 In this study a 5-day schedule of decitabine administered daily at a dose of 20 mg/m 2 was shown to be superior to a 10-day or subcutaneous schedule. A multicenter phase II trial of decitabine (ADOPT) using the 5-day schedule confirmed the safety of this schedule although response rates were lower than those reported by the MDACC. 118 In the ADOPT study the median number of courses administered was 5, the CR rate was 17% and the median survival was 19.4 months. No randomized survival study of a 5-day schedule of decitabine has been conducted in MDS. In parallel with this work, European investigators developed a randomized study of decitabine using the initial 3-day schedule. The major objective of the study was survival. Unfortunately, use of decitabine at this dose and schedule was not associated with improved survival in patients with higher risk MDS. 119 Despite all this data, the final dose and schedule of decitabine is not fully understood. Blum et al have indicated that a 10-day