Introduction

Pathologists use the descriptive term “dysplasia”, derived from the Greek word δυσπλασία (which means "malformation", from the roots δυσ- "bad-" and πλάθω "to create or to form"), to designate a particular set of recurrent cellular developmental abnormalities that can be visualized under the light microscope. Dysplasia can occur in tissues as diverse as the epithelia of the uterine cervix and the bronchial tree, but bone marrow dysplasia – referred to as “myelodysplasia” or “dysmyelopoiesis” – consists of some combination of intramedullary morphological alterations in the three major myeloid cell series: erythroid lineage dysplasia (dyserythropoiesis), granulocytic lineage dysplasia (dysgranulopoiesis), and megakaryocytic/platelet lineage dysplasia (dysmegakaryopoiesis).

Common examples of erythroid dysplasia include multinucleated red cell precursors, megaloblastoid maturation with nuclear-cytoplasmic developmental asynchrony, nuclear budding or internuclear bridging, karryorhexis, ring sideroblasts, cytoplasmic vacuolization, and periodic acid-Schiff-stain positivity. Hypogranular neutrophils, hypolobated (i.e., pseudo-Pelger-Huët) neutrophils, small granulocytes, pseudo-Chediak-Higashi granular inclusions, nuclear hypersegmentation, individual granulocytes containing both basophilic and eosinophilic granules (“eo-basos”), or dual esterase cytochemical staining characterize granulocyte dysplasia. Finally, megakaryocytic/platelet dysplasia is manifest as micromegakaryocytes, hypolobated or alobated megakaryocytic nuclei, as multiple widely separated nuclei, and as size and granulation abnormalities of more mature platelets.

Such dysplastic changes can be seen in several clinical settings: reactive conditions, due to the injurious effects of a drug or other toxin; nutritional deficiencies, such as lack of B12 or folate; or the clonal, neoplastic disorders that are collectively termed the myelodysplastic syndromes (MDS). For decades, confidently distinguishing MDS from other malignant and non-malignant entities has proven challenging for pathologists and clinicians alike [1–4]. Even the first formal disease classification mentioning MDS, the 1976 French-American-British (FAB) Co-operative Group classification of the acute leukemias, described MDS primarily in terms of what it is not – i.e., acute myeloid leukemia (AML) – rather than what it is, and the FAB investigators warned about the risk of confusing MDS and AML [1].

Diagnostic accuracy has practical consequences, and disagreements about the diagnosis are common. Among 915 patients with MDS who were referred to M. D. Anderson Cancer Center in Houston, Texas between 2005 and 2009, the diagnosis was revised from the diagnosis at the referring center in 12 % of cases [5]. Often the tertiary center diagnostic revision was only an assignment to a higher or lower MDS risk group resulting in a different therapeutic approach, but in some cases patients left Houston with an entirely new non-MDS diagnosis.

Minimal Diagnostic Criteria for MDS

The World Health Organization (WHO) updated the diagnostic classification of hematopoietic and lymphoid malignancies in 2001 (3rd edition [6]) and again in 2008 (4th edition [7]), building upon the FAB classifications of 1976 and 1982. While the WHO classifications state that MDS is defined by clonal hematopoiesis, dysplasia in at least one major myeloid lineage, ineffective hematopoiesis with resultant cytopenias, and a risk of development of acute myeloid leukemia (AML), minimal diagnostic criteria for MDS are not provided. The WHO classification does, however, state that at least 10 % of cells in a given lineage should be dysplastic to qualify as meaningful, and that the blast proportion should be based on assessment of at least 500 cells in the marrow and 200 cells in the peripheral blood [8].

In the author’s experience, morphologists in clinical practice do not actually count the number of dysplastic cells, especially since dysplasia has a qualitative component as well as quantitative; a cell may be only slightly abnormal (e.g., mildly megaloblastoid maturation) or may be dramatically deformed and frankly bizarre. Instead of laborious enumerating, most pathologists derive a general gestalt about mild, moderate, or extensive dysplastic changes in each lineage after reviewing numerous representative cells in a series of microscopic fields. The 10 % threshold is a guideline, meant to avoid having the occasional odd-looking cell seen in healthy elderly person result in saddling the patient with a diagnosis of MDS, and the 10 % proportion has not been prospectively validated.

In fact, dysplasia is worrisomely common in the bone marrow of healthy people [9]. In one study of 120 healthy bone marrow donors in which a “first-aspirate squash preparation” was assessed independently by four experienced morphologists, the WHO threshold of more than 10 % dysmyelopoiesis was detected in 46 % of patients, with bilineage dysplasia present in 26 % and trilineage dysplasia in 7 %, none of whom developed MDS during the study’s follow-up period [10]. The finding of erythroid dysplasia in most of 50 healthy volunteers in another series, including at least one or two dysplastic megakaryocytes in the marrow of 38 % of these volunteers [9], as well as relatively poor inter-rater reproducibility in dysplasia assessment except for specific obvious hallmarks such as ring sideroblasts [11], challenges the value of traditional morphology.

In 2006, a group of clinicians and pathologists participating in a workshop in Vienna attempted to define minimal diagnostic criteria for MDS (Table 1) [12•]. The workshop participants suggested that in order to diagnose MDS, two “prerequisite criteria” are necessary: first, a marked cytopenia (i.e., hemoglobin <11 g/dL, absolute neutrophil count <1.5 × 109/L, or platelets <100 × 109/L) must be present in at least one lineage and must endure for ≥6 months, unless cytogenetic studies confirm MDS earlier, and second, other clonal or non-clonal hematopoietic diseases or non-hematopoietic diseases must be excluded as the primary reason for cytopenia or dysplasia. In addition, the workshop recommended that at least one of three “decisive criteria” needs to be fulfilled to call a condition MDS: morphologic dysplasia in at least 10 % of all cells in one or more of the major cell lineages in the bone marrow aspirate, a typical MDS-associated cytogenetic abnormality, or a constant blast cell count of 5–19 %. Finally, the workshop participants also recommended “co-criteria” for patients who fulfill both prerequisite criteria and show typical clinical features of MDS, but do not demonstrate any of the three decisive criteria. These co-criteria include an abnormal marrow immunophenotype by flow cytometry, compatible with a diagnosis of MDS according to the European Leukemia Network criteria [13, 14], and evidence of a monoclonal cell population based on either a human androgen receptor assay, gene chip analysis, or mutation analysis. Workshop participants elected to err on the conservative side in terms of making an MDS diagnosis, arguing that because of the serious implications of the diagnosis of MDS, it is preferable to be exclusive rather than inclusive.

Table 1 Minimal Diagnostic Criteria for MDS as Proposed by 2006 Vienna Workshop [12•]
Full size table

These proposed criteria have several limitations. For example, a patient presenting with cytopenias, extensive multilineage dysplasia and 12 % blasts but a normal karyotype does not need to wait six months for diagnostic confirmation, even though this is mandated in the workshop criteria. A 6-month wait would only delay initiation of potentially helpful therapy and leave the patient at risk for cytopenia-related complications, and, unfortunately, spontaneous regression of MDS with excess blasts is exceptionally rare [15]. In addition, MDS can present with milder cytopenias than those enumerated in the prerequisite criteria. The proposed cytopenia degree and duration cutoffs are, therefore, arbitrary.

Furthermore, now that more than 25 MDS-associated somatic mutations are recognized [16•, 17•], it is increasingly possible to confirm clonal hematopoiesis in the presence of ambiguous morphology, but it is not yet clear that MDS diagnosed only by the finding of a typical somatic mutation (especially one present only at low allele burden) has the same natural history as MDS diagnosed by morphologic criteria – just as it is not clear that a clonal karyotypic abnormality detectable only by fluorescent in situ hybridization (FISH) has the same prognostic and therapeutic implications as a clone detected on routine G-banded metaphase karyotyping.

Disease classifications are only useful if they help predict the natural history of the classified condition and assist in choosing therapy. The modified FAB acute myeloid leukemia classification, M0-M7, is no longer widely used, because only M3 is approached differently; the FAB acute lymphoblastic leukemia classification (ALL L1-L3 subtypes) never found broad acceptance and has largely been replaced by immunophenotypic and molecular classification of ALL [18].

“ICUS” and “IDUS”

Perhaps the most practical proposal that emerged from the 2006 workshop was the category of “idiopathic cytopenias of undetermined significance” to describe patients in whom MDS is possible but not yet certain [19, 20]. In the 1980s and 1990s, such patients whose diagnosis was still murky were sometimes told they might have “not quite MDS”, “not yet MDS”, or “pre-MDS”. Unlike other hematologic “undetermined significance” conditions such as monoclonal gammopathy of undetermined significance (MGUS) and monoclonal B cell lymphocytosis of undetermined significance (MBUS), ICUS is, by definition, not known to be a clonal disorder.

This proposal was later expanded to include “idiopathic dysplasia of undetermined significance” (IDUS), a term that describes individuals who lack meaningful cytopenias, but have dysplastic cellular features in the blood and bone marrow. For example, modern automated cytometers may flag hypogranular neutrophils on a patient with normal blood counts who is scheduled to undergo a routine hip arthroplasty for degenerative joint disease, prompting hematology consultation and leading to a bone marrow biopsy.

Prospective series are needed to more clearly define the long-term outcomes of ICUS and IDUS patients. IDUS appears to be relatively rare, but there is likely to be a major referral bias, as many patients with incidentally discovered dysplasia disappear into the great noisy sea of abnormal yet non-alarming laboratory reports generated in busy general clinical practices, and are never referred to a specialty center. Only ten patients with IDUS were identified in an Austrian series of 1,363 patients with suspected MDS or mild cytopenias seen between 1997 and 2010 [21]. But IDUS may augur trouble to come: four of those ten Austrian patients went on to develop an overt myeloid neoplasm.

Similarly, while the frequency of ICUS is also unknown, long-term follow-up is indicated. In one study from the Mayo Clinic in Minnesota, which has thus far only been presented in abstract form [22], among 2,899 marrow exams performed over a 13 year period to evaluate cytopenias where MDS was a diagnostic possibility, 579 (20 %) were non-diagnostic and could have met the workshop’s proposed criteria for ICUS. Of these possible ICUS cases, 387 (69 %) had minimal dysplasia but did not meet diagnostic criteria for MDS, while 182 patients had entirely normal marrow and normal karyotype. Of those 182 patients, 92 patients were lost to follow-up and 80 eventually were shown to have a non-myeloid disorder (Table 2) causing the cytopenias (e.g., marginal zone lymphoma, lupus, hepatic cirrhosis), but six of the remaining ten patients developed MDS or AML – 0.2 % of the initial marrow population.

Molecular techniques can enhance diagnostic assessment. In the past, proof of clonally-restricted hematopoiesis has rested on relatively insensitive techniques such as G6PD isoform analysis or human androgen receptor gene-based assay (HUMARA). In a German study of 67 patients with ICUS, 23 were eligible for HUMARA and 17 of those had non-clonal X-chromosome inactivation patterns. Two of the six patients with clonal patterns developed AML, while no AML cases were observed among the 17 patients with non-clonal patterns [23]. The advent of high-throughput re-sequencing techniques offers great promise for better defining the clonal architecture of MDS, and post-Sanger-era sequencing has recently revealed that even in cases without excess blasts, most marrow cells are already clonal [24•].

De Novo Versus Secondary, Exposure-Related MDS

In some cases, clinicians struggle to distinguish de novo, idiopathic MDS from secondary, therapy-related MDS. Sometimes patients with MDS have a history of treatment for another malignancy (e.g., breast cancer with doxorubicin and external-beam irradiation), yet have a karyotype at diagnosis such as isolated del(20q) that suggests their MDS is a second, unrelated neoplasm. Conversely, some patients with no known DNA toxin exposure and no family history of MDS present at a younger age than is typical for MDS (<55 years; the median age at MDS diagnosis is about 70) and with a complex karyotype, suggesting the possibility of an unrecognized toxic exposure. Investigators continue to work to define genetic pathways for leukemogenesis that may distinguish between therapy-related MDS and de novo MDS, and the advent of next-generation sequencing techniques should accelerate the pace of this effort [25, 26].

Whole genome sequencing studies have now shown that throughout human life, but especially after age 50, clonally-restricted mutations accumulate serially in all tissues, resulting in somatic mosaicism [27, 28]. Most of these mutations are meaningless and of no consequence – so-called “passenger” mutations – but if by random chance a mutation alters expression of a gene or the function of its protein product in such a way that this change confers a growth or survival advantage to a clone, neoplasia can eventually result. Exposure to ionizing radiation, or to chemotherapy containing DNA alkylating agents or topoisomerase inhibitors, may accelerate the rate of accumulation of mutations generally and increase the chance of acquiring a neoplasia-inducing “driver” mutation. Perhaps to some extent all MDS cases are actually secondary, “exposure-related”, but the exposure in most patients is unrecognized – background irradiation from the sun or from naturally occurring radioisotopes, dangerous but occult environmental waste, an unrecalled week spent painting a poorly ventilated room accompanied by an open can of turpentine, or the byproducts of normal cell metabolism and dietary substances. Only certain types of exposures and associated karyotypes are clearly linked to the poorer outcomes described with therapy-related MDS [29].

Germline Disorders that Can Predispose to, or Be Mistaken for, MDS

Some patients who develop MDS at a relatively young age do not have a specific toxin exposure, but instead harbor an inborn defect in DNA stability or repair that allowed mutations to accumulate in marrow cells at an abnormally rapid rate [30]. These patients’ MDS may appear typical, but it is important to recognize the underlying defect, as failure to do so can result in considerable harm when the patients undergo SCT using common myeloablative approaches.

Fanconi anemia (FA) is the most common of the germline disorders that predispose to MDS and AML. MDS/AML arising in the context of FA is usually associated with monosomy [31]. While most patients with FA are diagnosed prior to adolescence, especially if major body dysmorphology is present, up to one-third of patients first present with FA in adulthood and may not have any physical findings other than a few café au lait spots or hypopigmented skin patches. The upper end of the age range at presentation with marrow failure where chromosome breakage analysis for FA needs to be considered is unclear; a cutoff of 40 years has been proposed [32]. Similarly, dyskeratosis congenital (DC) and related telomere disorders are also known to present with atypical features and absence of expected DC stigmata such as fingernail dystrophy or oral leukoplakia, and the only clue to a congenital syndrome may be a family history of pulmonary fibrosis or cryptogenic hepatic cirrhosis [33].

Congenital neutropenias, including cyclic neutropenia, due to mutations in the ELANE2 gene encoding neutrophil elastase or the CSF3R gene encoding the G-CSF receptor, can both predispose to MDS and be mistaken for MDS [34–38]. Evolution of karyotypic abnormalities in these patients increases the risk of subsequent MDS/AML [35].

Some patients with non-syndromic MDS have a family history of MDS/AML, and it seems likely that numerous germline MDS-predisposing alleles exist and have yet to be discovered. A prodrome of thrombocytopenia preceding development of MDS or AML characterizes germline mutations in the RUNX1 (CBAF2, AML1) transcription factor, but the newly described GATA2 germline mutations that also increase risk for subsequent MDS development have no prodrome [39, 40].

Several other congenital syndromes that may initially present in adulthood and at first blush can cause concern for MDS have specific associated blood and marrow morphological findings that usually present little diagnostic difficulty, such as the congenital dyserythropoietic anemias (CDA) or the gray platelet syndrome (GPS) [41]. The internuclear bridging that characterizes CDA type I can also be seen in MDS [42], but a test for the CDAN1 gene that causes CDA type I is available, as is a test for the NBEAL2 gene associated with GPS [43].

MDS Versus Aplastic Anemia and Large Granular Lymphocyte Disorders

While most patients with MDS present with a bone marrow that is hypercellular or normocellular in comparison to age-matched healthy controls, in approximately 10–15 % of MDS cases, the marrow is instead hypocellular [44]. Distinguishing so-called “hypoplastic MDS” (hMDS) with a hypocellular marrow from aplastic anemia (AA) can be exceptionally difficult [45].

An increase in the blast proportion certainly favors hMDS over AA, as does the presence of extensive cellular dysplasia or clonal cytogenetic abnormalities. However, the relative paucity of cells in hypocellular marrows (whether AA or hMDS) limits detailed morphologic assessment, and routine karyotyping in such cases also frequently fails due to a lack of metaphases for analysis [46•]. In addition, up to one-half of patients with MDS, including those with hMDS, have a normal karyotype, which is unhelpful. Complicating matters further is the fact that AA can evolve to MDS or another clonal disorder, and this may be apparent cytogenetically long before the change becomes clinically manifest [47].

While some investigators feel an abnormal karyotype strictly excludes the diagnosis of AA, other groups believe that patients with AA can have an abnormal karyotype as a result of oligoclonal hematopoiesis due to diminished stem cell reserves [48, 49]. There may be even more overlap between AA and hMDS than previously suspected. For example, in a Cleveland Clinic analysis of 93 AA and 24 hMDS cases distinguished by conventional morphologic criteria, combined metaphase and single nucleotide polymorphism array (SNP-A) karyotyping detected clonal abnormalities including copy-neutral loss of heterozygosity in 19 % of AA and 54 % of hMDS cases [46•]. To date, whole genome sequencing has only been applied in familial AA [50], but in the future, detailed molecular genetic typing may uncover patterns of somatic mutations that will better help distinguish AA from hMDS.

The clinical and pathological overlap between AA and hMDS, coupled with the increased representation of the same human leukocyte antigen (HLA) type (HLA DR15) in patients with both disorders, has prompted clinicians to use immunosuppressive therapy with anti-thymocyte globulin and other drugs directed at autoreactive T lymphocytes for both marrow failure conditions [51, 52]. While HLA DR15 status coupled with younger patient age and normal karyotype may predict greater likelihood of response to immunosuppressive therapy [53], marrow hypocellularity alone has not been a consistent predictor of a favorable response [54].

Clonal disorders of large granular lymphocytes (LGL leukemias) often present with neutropenia and anemia and occasionally exhibit dysplastic cellular changes [55], and can be mistaken for MDS – or may co-exist with MDS [56, 57]. Similarly, patients with MDS may have a detectable T-cell clone [58]. The recent discovery of somatic mutations of Signal Transducer and Activator of Transcription 3 (STAT3) in 40 % of 77 patients with LGL leukemia will help refine molecular evaluation of immune-mediated marrow failure syndromes [59].

MDS Versus Acute Leukemia, Including Erythroleukemia

Prior to the 1970s, the diagnosis of AML required at least 50 % marrow blasts; the 1976 and 1982 FAB classifications set 30 % as the threshold between AML and chronic myeloid disorders such as MDS [1, 2]. The 2001 WHO Classification revision, in contrast, reduced the AML-defining blast threshold to 20 % and eliminated the FAB-defined MDS subtype refractory anemia with excess blasts in transformation (RAEB-t, 20–29 % marrow blasts), since patients with RAEB-t did not clearly have a different prognosis from those with higher blast threshold. This proposal met with some dissension, including investigators who reported that the degree of apoptosis in marrow samples from patients with RAEB-t was more akin to that observed in MDS than AML [60, 61].

Indeed, the RAEB-t category refuses to die, as regulatory approval of azacitidine and decitabine in the United States was based on FAB-defined MDS, so the 30 % blast threshold continues to define drug reimbursement policy. In 2010, AZA-001 clinical trial investigators separately analyzed the 20–29 % blast cohort in order to demonstrate that the survival benefit with azacitidine over conventional care extends to “low-blast count AML” [62]. As with most biological parameters, MDS and AML represent a continuum rather than sharply defined distinct entities, so discussion of a blast threshold in isolation from other clinical and biological parameters is perilous.

One practical problem when dealing with the MDS versus AML blast threshold arises in the previously treated patient. An individual who had 40 % blasts at the time of diagnosis (AML), received induction chemotherapy, and now has 5–10 % blasts, might be said to have persistent AML – but might also be considered to have MDS, especially if the AML initially arose from MDS. This semantic distinction may alter clinical trial eligibility, depending on the wording of inclusion and exclusion criteria.

Erythroleukemias – acute erythroid leukemias, i.e., a group of leukemias dominated by an immature erythroid cell population – have caused diagnostic controversy for decades. The 2008 WHO classification, like the 2001 proposal and the FAB system, defined two subtypes of erythroleukemia, distinguished by the presence or absence of a granulocytic component. The first subtype, erythroid/myeloid leukemia (FAB subtype M6a), is diagnosed when the marrow is populated by erythroid precursors representing ≥50 % of nucleated cells, and there are also ≥20 % myeloblasts in the non-erythroid cell population; <20 % myeloblasts would be MDS. The other subtype, “pure” erythroleukemia (FAB subtype M6b), can be diagnosed when ≥80 % of marrow cells are undifferentiated or proerythroblastic cells apparently committed to the erythroid lineage and no significant myeloblastic component is present. Complex karyotypes and background dysplasia are common in the erythroleukemias. The WHO specified that if multilineage dysplasia is present in ≥50 % of cells in two or more lineages, a diagnosis of “AML with myelodysplasia-related changes” should instead be made.

The WHO proposals, like those of the FAB group before them, have been extensively criticized, due to the heterogeneity of these syndromes [63–66]. One group evaluated a series of immunophenotypically-defined pure erythroid leukemia cases and found that the WHO algorithm reclassified all of them into other categories, raising the question of whether pure erythroid leukemia exists [65].

In the days of noted hematologist William Dameshek (1900–1969), the first editor of Blood, various malignant erythroid-dominant proliferations were lumped together as “di Guglielmo syndrome” (named after a 1923 report by Naples hematologist Giovanni di Guglielmo) and although it was recognized that they were markedly heterogeneous, treatment approaches were similar and the prognosis generally grim [67]. It is not clear that genuine advances have been made in this area in the last half-century.

The overlap between MDS and myeloproliferative neoplasms (MPN) such as chronic myelomonocytic leukemia (CMML) – considered a form of MDS by the FAB – is even more complex, and the WHO has defined a separate MDS/MPN overlap category since 2001. The distinction between MDS, MPN and overlap syndromes has recently been reviewed elsewhere [68] and space limitations preclude visiting these issues here.

The Special Case of Sideroblastic Anemias

When patients have ring sideroblasts in their marrow – i.e., erythroid precursors with at least five mitochondrial ferritin-containing siderotic granules encircling ≥1/3 of the nucleus – but also have multilineage dysplasia, excess blasts, or an abnormal karyotype, the patient is easily assigned to the MDS category [69, 70]. However, “pure” sideroblastic anemia, where only the erythroid lineage is dysplastic and the karyotype and blast proportion are normal, raises the possibility of a late presentation of a congenital sideroblastic anemia (CSA). Late-onset CDA is best described with germline mutations of ALAS2, which is X-linked. Patients with ALAS CDA can present after age 70, due to greater attrition of clones with the wildtype ALAS2 active than clones in which the mutant ALAS2 Lyonized during embryogenesis [71].

One clue to the possibility of an ALAS2 mutation is red blood cell microcytosis; RARS, in contrast, usually is associated with normocytic or macrocytic anemia. However, there is a growing number of recognized molecular causes of sideroblastic anemias, and several other subtypes are not associated with microcytosis (though presentation of these other types after age 50 has not been described) [72]. An empiric trial of pyridoxine therapy (vitamin B6) can improve erythropoiesis in several forms of CSA but is rarely helpful in RARS, and prolonged administration of high doses of pyridoxine can induce peripheral neuropathy [73].

In addition, numerous drugs, including several agents used for tuberculosis therapy, can cause sideroblastic anemia, but this history is usually obvious [74]. Copper deficiency must also be considered whenever ring sideroblasts are present, and the cause of copper deficiency is not apparent in at least one-third of cases [75]. Additionally, the prevalence of copper deficiency appears to be increasing, as bariatric surgery for obesity is growing in frequency and such surgeries put patients at increased risk for micronutrient deficiencies including copper; also, patients who take zinc supplements in a belief that it will ward off the common cold or other ailments put themselves at risk for copper deficiency [75]. Clues to copper deficiency other than patient history include the presence of an associated neuropathy or myeloid cell vacuolization [75, 76].

MDS Versus Non-neoplastic Disorders

Perhaps the most important distinction that clinicians need to make is between dangerous MDS and easily correctable nutritional deficiencies, such as vitamin B12 (cobalamin). Because B12 deficiency can produce severe pancytopenia and striking megaloblastoid maturational changes, and profound cytopenias may be associated with only minimal neurological symptoms, it can be easy to miss B12 deficiency unless a B12 level is measured. In the past, a few patients have undergone allogeneic stem cell transplantation for vitamin B12 deficiency; this simply should not occur.

In 2012, it emerged that automated analyzers using vitamin B12 assays that apply a method based on the competitive binding of serum vitamin B12 with reagent intrinsic factor often report false elevation of B12 levels, especially in pernicious anemia with high titers of intrinsic factor antibodies, resulting in lack of sensitivity to B12 deficiency [77•]. When in doubt, methylmalonic acid (MMA) levels in the serum and urine are consistently elevated in the presence of tissue-level B12 deficiency, and MMA is straightforward to measure. Some laboratories will reflex test for MMA with B12 results at the lower end of the normal range.

Folate deficiency can also present with a macrocytic anemia, cytopenias, and megaloblastoid maturation and should also be sought in suspected MDS, though the author can not recall seeing a patient in the last five years who was referred to his academic center without having had a folate level measured at least once. Other nutritional deficiencies such as selenium can also cause cytopenias. It is beyond the scope of this review to discuss individual micronutrient deficiency risks and clinical presentations in detail, but these deficiencies should always be considered when evaluating a patient with cytopenias, especially when a gastrointestinal disorder associated with malabsorption is present.

Various toxins can cause cytopenias and dysplastic changes that can be mistaken for MDS. Alcohol is the most ubiquitous of these; and dysplastic changes and cytopenias may resolve within just a few weeks if heavy drinkers cease alcohol consumption [78, 79]. Alcohol abusers are also at risk for concomitant nutritional deficiencies, especially folate and thiamine, and for internal hemorrhage. However, many regular drinkers are reluctant to quit for long enough to evaluate the cause of anemia and the alcoholic who ceases drinking abruptly may experience withdrawal symptoms. Arsenic intoxication can cause cytopenias and dysplastic changes that impersonate MDS, but dysplastic changes are not usually seen after the usual course of arsenic trioxide for acute promyelocytic leukemia [80]. Numerous drugs can be associated with dysplasia, erythrocyte macrocytosis, or both – methotrexate, azathioprine, and chemotherapeutic agents are well recognized causes of these findings.

The human immunodeficiency virus (HIV) can induce cytopenias via a number of different mechanisms, including immune dysregulation and direct infection of hematopoietic precursors [81–83]. If a marrow exam is performed in a patient with HIV, megaloblastoid maturation and dysplastic changes may mimic MDS. These alterations usually resolve after initiation of anti-retroviral therapy.

Finally, a variety of autoimmune conditions can mimic MDS. Immune thrombocytopenia can be present together with MDS, as can autoimmune hemolytic anemia; the possibility of the former may provide insurance justification for romiplostim therapy in patients with severe thrombocytopenia. Rheumatoid arthritis, systemic lupus erythematosus, and other rheumatologic disorders can present with anemia or other cytopenias. Felty syndrome – the triad of rheumatoid arthritis, neutropenia, and splenomegaly, often associated with an increase in LGLs or detectable T cell clone – bridges RA, T-LGL, and neutropenia, but is typically associated with substantive splenomegaly, which is rare in MDS generally and should prompt consideration of other diagnoses [84]. A form of benign chronic neutropenia that is non-progressive is most common in middle-aged women and may be autoimmune in origin [37]. Chromosome X inactivation patterns in chronic idiopathic neutropenia can be either oligoclonal or polyclonal [85].

Table 2 A diagnostic approach to suspected MDS
Full size table

Conclusion

Just as not all that glitters is gold, not all that is dysplastic is myelodysplastic syndrome. Hematologists are often called upon to evaluate patients with cytopenias or dysplastic morphology in whom MDS is a possible diagnosis, but many other diagnoses are commonly encountered in routine clinical practice (Table 3). Evolving molecular pathology techniques including the mutation assays described above will help improve diagnostic accuracy and augment clinicians’ confidence in an MDS diagnosis.

Table 3 An MDS Differential Diagnosis
Full size table