Cytomorphology review of 100 newly diagnosed lower-risk MDS patients in the European LeukemiaNet MDS (EUMDS) registry reveals a high inter-observer concordance

Myelodysplastic syndromes (MDS) are a heterogeneous group of clonal diseases, characterized by cytopenias and dysplastic features in blood and bone marrow. Due to the heterogeneity of clinical presentation, degree of dysplastic features and a considerable list of differential diagnosis, the diagnosis of MDS is often a challenge, particularly in lower-risk MDS. The diagnosis of MDS is primarily based on cytomorphologic and histopathologic assessment of dysplasia, but cytogenetic evaluation is crucial for the classification and to determine the risk assessment using the revised International Prognostic Scoring System (IPSS-R) score [1]. Flow cytometry and molecular analysis provide additional valuable information to refine the diagnosis and prognosis of MDS, but the clinical and therapeutic impact of these methods is under investigation [2, 3]. To make a correct diagnosis of MDS, good cytomorphologic experience and adequate clinical information are required, including concomitant medication, concurrent infections and comorbidity, and other possible causes of cytopenia or dysplastic features. Therefore, a multidisciplinary approach is often required to make the definite diagnosis of MDS and to exclude the possibility of other disorders mimicking MDS [4, 5].

In the prospective European LeukemiaNet MDS (EUMDS) registry, patients with a newly diagnosed MDS belonging to the low or intermediate-1 IPSS risk group are included, within 100 days of making the definite diagnosis of MDS [6]. Patients with higher-risk MDS or therapy-related MDS are excluded. National and local ethics committees have approved the study, and all patients provided written informed consent for inclusion in the registry. The first 1000 patients were diagnosed with MDS between December 2007 and December 2010, and therefore, the diagnosis was based on the World Health Organization (WHO) 2001 classification, without central revision of diagnosis. Since a correct MDS diagnosis and classification are crucial in the EUMDS registry and may be difficult in the lower-risk group of MDS, a morphology sub-study was performed by reviewing 100 randomly selected cases by a panel group of 11 MDS cytomorphologists from 9 different countries. The cytomorphology sub-study was primarily aimed at a validation and reproducibility of the WHO 2001 criteria for MDS by an international panel and at the validation of original cytomorphologic accuracy and correct diagnostic inclusion into the registry. The aims of the study were (i) quality assessment of cytomorphologic diagnoses in the EUMDS registry and (ii) assessment of inter-observer variability in the review panel regarding diagnosis and degree of dysplasia.

The cytomorphology review was performed by a panel comprising 11 morphologists during a 3-day meeting at the Heinrich Heine University medical Centre in Düsseldorf, Germany. The aim was to review the original diagnosis at time of inclusion in 10% of the patients in the EUMDS registry, with five additional cases to replace inadequate material (only bone marrow trephine biopsy or slides with poor quality unable to examine).

Median age of the randomly selected patients at diagnosis was 72 years (range 21 to 95), there was a male predominance (67%) and the most common diagnosis recorded was RCMD (Table 1). In total, 774 slides were reviewed by the 11 raters, 45 slides (5.8%) were excluded from subsequent analyses as the panel members were unable to review the diagnosis due to the poor quality of the slide. One case was removed as only two panel members had reviewed it. Accordingly, in total, 98 cases were included in the review, and, on average, each case was reviewed by eight different panel members (range 3–9). In total, 727 diagnoses were made by the panel members.

Diagnosis of MDS is still based primarily on cytomorphology of blood and marrow, accompanied by histomorphology, cytogenetics, molecular findings and flow cytometry. The EUMDS registry is collecting data of newly diagnosed lower-risk MDS patients. Eligibility for entry into the registry is based on the local assessment of cytomorphology at the participating centres in 14 different countries and 118 sites. In order to assess the quality and accuracy of the MDS diagnoses in the registry, we organized a working group reviewing 10% of the first 1000 patients that were included into the registry. In our cytomorphologic review of 100 newly diagnosed lower-risk MDS patients, we could clearly demonstrate a high accuracy of the original diagnoses at entry into the registry and a high inter-observer concordance of the panel members with regard to diagnosis of MDS and assessment of the lineage involvement of dysplasia.

An important prerequisite for cytology is satisfactory quality-stained slides, in our case well-prepared blood and marrow films and ‘bone marrow aspirate squash’ slides without many preparation artefacts. Only 6% of the slides were of poor quality and could not be assessed properly, whereas 94% could be validated. Cytomorphology in MDS can be challenging, even for experienced morphologists, and sometimes a definite diagnosis is not possible or only based on genetic findings [11]. Minimal diagnostic criteria for the diagnosis of an MDS cannot be based on morphology only, as no single sign of dysplasia is MDS specific [12]. The slides that were evaluated during our microscope session were blinded, and the panel member did not have information on the clinical course of the patients and were not biased in their judgement. We assembled a very heterogeneous panel of cytologists in terms of the morphological experience of the members, and therefore we were prepared to be faced with discrepancies in the assessment of the diagnosis. Surprisingly, most disagreements were not significant. First of all, a small number of slides were detected as CMML and RAEB-2 and AML cases. This task was the easiest for the panel members as there was no disagreement. As a result, only 6% of the patients who entered into the registry were misdiagnosed by the centres. This confirms that the data quality of the registry is high and that subsequent analyses based on diagnosis by the local cytomorphologists are likely to be valid. The assessment of medullary blast counts was also extremely uniform, as only 6% of the cases were differently categorized (<5% blasts versus 5–9% blasts).

The assessment of lineage involvement of dysplasia was less concordant. Inter-observer variability with regard to description of dysplasia was reported by several groups [13‐15]. Our data show that our Panel team had substantial agreement on the assessment of dysmyelopoiesis and dysmegakaryopoiesis, whereas the assessment of dyserythropoiesis was less concordant. This finding is plausible, as some signs of dysplasia in myelo- and megakaryopoiesis are more specific for myeloid malignancies like micromegakaryocytes, mononuclear megakaryocytes, pseudo-pelger cells and degranulated myeloid precursors. These signs of dysplasia are usually not described in the marrow with reactive changes due to inflammation, toxic effects and others [12]. Morphological erythroid dysplasia can be evident in benign disorders like hemolysis, megaloblastic anemias and others. As a result, the panel members differed moderately in the categorization of uni- versus multilineage dysplasia. There is growing evidence that flow cytometry with regard to properly described dysplasia of erythropoiesis can be utilized if performed in a sophisticated way [3, 16, 17]. The separation of lower-risk MDS cases, namely distinguishing uni- and multilineage dysplasia, ring sideroblastic phenotype, and deletion of chromosome 5 is of great prognostic value, as shown by different groups [18‐21]. As our data show, knowledge of cytomorphology is necessary to properly describe dysplasia and to use the classification systems accordingly. Therefore, skills in cytomorphology are still mandatory in centres where patients with MDS are diagnosed and treated accordingly [22, 23].

The WHO 2001 proposals for the classification of MDS took into account the separation of uni- and multilineage dysplasia within the non-blastic MDS cases with ring sideroblasts. This feature was omitted by the WHO 2008 [24] proposals but was reintroduced in the WHO 2016 classification [21]. As a result, our data generally reflect the WHO 2016 classification with the exception of minor changes of the definition of MDS del(5q) and MDS unclassified and the introduction of SF3B1 mutations as classifying elements.

The slides that were evaluated during our microscope sessions were blinded and the panel members did not have information on the clinical course of the disease. The only bias was the information that slides have been diagnosed as lower-risk MDS by the local cytologists.

Limitations of the project were some missing iron stains, not primarily affecting the morphological results with regard to eligibility of the patients into the registry. An even better assessment of morphology could have been reached by adding histomorphology for all cases. Although the impact of histomorphology in lower-risk MDS is limited as compared with cytomorphology, it can be helpful in cases with poor cytologic material and can help to properly assess the cellularity and fibrosis [25]. Therefore, another panel review, including histopathologists, could be of added value.

In summary, there was a high inter-observer agreement after the elimination of poor quality slides within our very heterogeneous panel of 11 morphologists. As a result, the inclusion of the patients into the EUMDS registry could be shown to be correct in the vast majority of the patients. The 6 patients originally not diagnosed correctly by the local cytologists have not been withdrawn from the registry. The registry is assessing patients who were diagnosed in the real world as lower-risk MDS and managed as such. Therefore, a central morphology review is not mandatory for this particular registry.