1 23

Annals of Hematology ISSN 0939-5555 Ann HematolDOI 10.1007/s00277-014-2252-4

Interobserver variance in myelodysplasticsyndromes with less than 5 % bone marrowblasts: unilineage vs. multilineage dysplasiaand reproducibility of the threshold of 2 %blastsPatricia Font, Javier Loscertales,Carlos Soto, Pilar Ricard, CarolinaMuñoz- Novas, Estela Martín-Clavero,Montserrat López-Rubio, et al.

1 23

Your article is protected by copyright and

all rights are held exclusively by Springer-

Verlag Berlin Heidelberg. This e-offprint is

for personal use only and shall not be self-

archived in electronic repositories. If you wish

to self-archive your article, please use the

accepted manuscript version for posting on

your own website. You may further deposit

the accepted manuscript version in any

repository, provided it is only made publicly

available 12 months after official publication

or later and provided acknowledgement is

given to the original source of publication

and a link is inserted to the published article

on Springer's website. The link must be

accompanied by the following text: "The final

publication is available at link.springer.com”.

ORIGINAL ARTICLE

Interobserver variance in myelodysplastic syndromes with lessthan 5 % bone marrow blasts: unilineage vs. multilineagedysplasia and reproducibility of the threshold of 2 % blasts

Patricia Font & Javier Loscertales & Carlos Soto & Pilar Ricard & Carolina Muñoz- Novas &

Estela Martín-Clavero & Montserrat López-Rubio & Luis Garcia-Alonso & Marta Callejas &Alfredo Bermejo & Celina Benavente & Mónica Ballesteros & Teresa Cedena &

María Calbacho & Raquel Urbina & Jesús Villarrubia & Santiago Gil & José María Bellón &

José Luis Diez-Martin & Ana Villegas

Received: 27 July 2014 /Accepted: 29 October 2014# Springer-Verlag Berlin Heidelberg 2014

Abstract Previous studies have shown the reproducibility ofthe 2008 World Health Organization (WHO) classification inmyelodysplastic syndromes (MDS), especially whenmultilineage dysplasia or excess of blasts are present. How-ever, there are few data regarding the reproducibility of MDSwith unilineage dysplasia. The revised International Prognos-tic Scoring System R-IPSS described two new morphologicalcategories, distinguishing bone marrow (BM) blast cell countbetween 0–2 % and >2–<5 %. This distinction is critical for

establishing prognosis, but the reproducibility of this thresh-old is still not demonstrated. The objectives of our study wereto explore the reliability of the 2008 WHO classification,regarding unilineage vs. multilineage dysplasia, by reviewing110 cases previously diagnosed with MDS, and to studywhether the threshold of ≤2 % BM blasts is reproducibleamong different observers. We used the same methodologyas in our previous paper [Font et al. (2013) Ann Hematol92:19–24], by encouraging investigators to include patients

P. Font (*) :M. Ballesteros : R. Urbina : J. L. Diez-MartinDepartment of Hematology, Hospital General Universitario GregorioMarañon, C/ Doctor Esquerdo 46, 28007 Madrid, Spaine-mail: pfontlopez@yahoo.es

P. Font : J. M. Bellón : J. L. Diez-MartinInstituto de Investigación Sanitaria Gregorio Marañón, Madrid,Spain

J. LoscertalesDepartment of Hematology, Hospital Universitario de La Princesa,Madrid, Spain

C. SotoDepartment of Hematology, Fundación Jiménez Díaz, Madrid, Spain

P. RicardDepartment of Hematology, Hospital Universitario FundaciónAlcorcón, Madrid, Spain

C. M.<. NovasDepartment of Hematology, Hospital Universitario Infanta Leonor,Madrid, Spain

E. Martín-Clavero : T. CedenaDepartment of Hematology, Hospital Universitario12 de Octubre,Madrid, Spain

M. López-Rubio :M. CallejasDepartment of Hematology, Hospital Universitario Príncipe deAsturias, Madrid, Spain

L. Garcia-AlonsoDepartment of Hematology, Hospital Universitario de Getafe,Madrid, Spain

A. BermejoDepartment of Hematology, Hospital Universitario de Fuenlabrada,Madrid, Spain

C. Benavente :A. VillegasDepartment of Hematology, Hospital Universitario Clínico SanCarlos, Madrid, Spain

M. Calbacho : J. VillarrubiaDepartment of Hematology, Hospital Universitario Ramón y Cajal,Madrid, Spain

S. GilDepartment of Hematology, Hospital Universitario Puerta de Hierro,Madrid, Spain

Ann HematolDOI 10.1007/s00277-014-2252-4

Author's personal copy

with <5 % BM blasts. Samples were collected from 11 hos-pitals and were evaluated by 11 morphologists. Each observerevaluated 20 samples, and each sample was analyzed inde-pendently by two morphologists. Discordance was observedin 36/108 suitable cases (33%, kappa test 0.503). Diagnosis ofMDS with unilineage dysplasia (refractory cytopenia withunilineage dysplasia (RCUD), refractory anemia with ringsideroblasts (RARS) or unclassifiable MDS) was assessed in33 patients, by either of the two observers. We combined thisseries with the cases with RCUD or RARS included in our2013 paper, thus obtaining 50 cases with unilineage dysplasiaby at least one of the observers. The whole series showed verylow agreement regarding RCUD (5/23, 21 %) and RARS(5/28, 18 %). Regarding BM blast count, the threshold of≤2 % was not reproducible (discordance rate 32/108 cases,kappa test 0.277). Our study shows that among MDS WHO2008 categories, interobserver discordance seems to be high incases with unilineage dysplasia. We also illustrate that thethreshold of ≤2 % BM blasts as settled by the R-IPSS maybe not easy to reproduce by morphologists in real practice.

Keywords Myelodysplastic syndromes (MDS) . RCUD .

RARS .WHO classification . R-IPSS

Introduction

Myelodysplastic syndromes (MDS) are a heterogeneousgroup of diseases, characterized by cytopenia and dysplasticchanges of peripheral blood (PB) and bone marrow (BM) andwith an enhanced risk of transforming into acute myeloidleukemia (AML) [1]. The diagnosis of MDS is based onmorphology and cytogenetics. The French–American–British(FAB) classification [2], after being the gold standard for twodecades, was refined by the World Health Organization(WHO) in 2001 [3]. This classification was clearly improvedin the 2008 revision [4]. The authors of the revision describedthe refractory cytopenia with unilineage dysplasia (RCUD),including refractory anemia (RCUD-A), refractory neutrope-nia (RCUD-N) and refractory thrombocytopenia (RCUD-T).The definition of the unclassifiable MDS (MDS-U) was alsorefined, and the entity refractory cytopenia with multilineagedysplasia and ring sideroblasts (RCMD-RS) was merged torefractory cytopenia with multilineage dysplasia (RCMD),since a different prognosis was not demonstrated betweenboth entities.

Morphology is essential for establishing prognosis ofMDS. The percentage of BM blasts was one of the basis ofthe International Prognostic Scoring System (IPSS) [5]. TheWHO Prognostic Scoring System (WPSS) demonstrated dif-ferent prognosis regarding the WHO categories, showing alower risk for the patients with unilineage dysplasia [6]. Therevised IPSS (R-IPSS) in 2012 included several changes to

refine the prognostic value of the cytogenetics [7], cytopenia,and percentage of blasts [8]. The category of <5 % BM blastswas divided in two levels: 0–2 % and >2 and <5 %. There aresignificant differences between these two categories in overallsurvival and risk of AML evolution.

Recently, some authors have been exploring the reproduc-ibility of the 2008 WHO classification. Naqvi et al. highlight-ed the discrepancies in the MDACC after reviewing the diag-nosis from the referring centers [9]. Sennent et al. describedmoderate agreement among four observers reviewing 50 casesof MDS, with the lowest concordance in cases with refractoryanemia with excess of blasts type-1 (RAEB1) and refractoryanemia with RS (RARS) [10]. Our group, in 2013, foundgood concordance regarding cases with multilineage dyspla-sia and cases with excess of blasts, but we could not reproducethe diagnosis of most of the cases with unilineage dysplasia[11]. This finding seemed troublesome, but our number ofpatients with unilineage dysplasia was very small, and furtherstudies are required to confirm this finding.

Regarding the importance of the accuracy of the blastcount, the R-IPSS is the first model that considers the categorywith ≤2 % BM blasts as being associated with a better out-come in MDS. All of the studies mentioned above weredeveloped previously for the implementation of the R-IPSS;therefore, the reliability of the threshold of 2 % has not yetbeen demonstrated.

We, therefore, conducted a new study, the subject of thispaper, using the same methodology as in our previous paperfocusing on the cases with less than 5 % BM blasts [11]. Ourobjectives were the following:

1. To explore the reliability of the 2008 WHO classificationin cases with MDS and <5 % BM blasts, distinguishingunilineage vs. multilineage dysplasia

2. To study whether the threshold of 2 % BM blasts isreproducible among different observers

Methods

A total of 110 BM samples from patients previously diag-nosed with MDS, following the 2008 WHO classification,were evaluated in a clinical setting. We developed identicalmethodology that is described in our previous paper [11].

The study was performed in 11 hospitals, and each institu-tion provided ten cases. The cases were selected for the studyfollowing two instructions:

First, the diagnosis of MDS should have been previouslyestablished. This means that our investigators confirmed thediagnosis, by carefully reviewing the samples specifically forthe study. With this review, we avoided selecting borderline orconfusing cases that could not fulfill the criteria of MDS.

Ann Hematol

Author's personal copy

Second, we encouraged investigators to include patients with<5 % BM blasts, so we could focus on the low-risk MDSpatients. After observing these conditions, the investigatorscould follow their own criteria to select the samples, since theproportion of samples for each category was not predefined.

After investigators selected and reviewed their own sam-ples, these samples were sent to a central coordinator, whoexchanged the samples for the second review. There were anodd number (n=11) of observers, so we made four pairs ofinstitutions that evaluated 80 samples, and the remaining 30samples were evaluated by three observers organized in threepairs (see Fig. 1). This means that each observer evaluated 20samples. Each sample that was previously diagnosed withMDS was again analyzed independently by two morpholo-gists specifically for the study. The second observer wasblinded to the clinical and laboratory data, except for the PBcounts. Most of the observers had participated in our previousstudy, so in order to preserve the blind criteria, we establisheddifferent pairs of investigators.

Morphological analysis

The MDS samples consisted of BM smears from aspirate,with their correspondent PB smears. Specimens were stainedwith Wright in 86 cases and May–Grünwald–Giemsa in 24cases. Smears were stained with Prussian blue for evaluationof RS. The degree of dysplasia in the three hematopoieticlineages and the percentage of blasts and RS were reportedaccording to the criteria of the 2008 WHO classification [8].

At least 500 nucleated BM cells were counted. In cases withthe presence of ≥50 % erythroid precursors in the entirenucleated population, the myeloblasts were calculated as apercent of the non-erythroid cells.

Cytogenetics

Cytogenetics was available in 102 cases. Cytogenetic aberra-tions were grouped according to the original IPSS and the R-IPSS [5, 7, 8].

Statistics

Qualitative variables are expressed as frequency and percent-age. The generalized kappa statistic for multiple rates (κ) wasapplied to calculate the concordance of qualitative variablesamong observers. The generalized κ statistic was interpretedas follows: 0–0.20 poor/slight; 0.21–0.40 fair; 0.41–0.60moderate/good; 0.61–0.80 substantial/very good; >0.80 al-most perfect [12]. The SPSS 21 statistical software packageand Epidat 3.1 were used for all calculations.

The study was approved by the central Ethical Committeeof Clinical Research in accordance with the HelsinkiDeclaration.

Results

Interobserver agreement considering the whole series

Finally, 108 samples were considered suitable by the secondobservers. As expected, most of the cases corresponded tomultilineage dysplasia. Table 1 shows the diagnosis regarding2008WHO classification in the referring centers. Seven caseswere classified with RCUD (RCUD-A n=3, RCUD-N n=1,RCUD-T n=3), 15 with RARS, 73 with RCMD, 9 with MDSwith del(5q), 2 with RAEB1, and 2 with MDS-U. The MDS-U included one case with unilineage dysplasia and pancyto-penia and the only case of the series with 1 % blasts in the PB.There were 49 cases with RS: 33 RCMD, 15 RARS, and 1with del5q.

Discordance was observed in 36/108 cases (33 %). Theinterobserver agreement by all the series was moderate, 67 %(kappa test 0.503). The 36 cases with interobserver discrep-ancy are shown in Table 2.

The rate of discordance by pairs ranged from 20 to 52 %(Table 1). The pair 4 showed the highest rate of discrepancy;these observers included the highest number of cases with RSand the discordance was due to the absence of concordancebetween RARS and RCMD with RS.

The interobserver concordance was very good regardingMDS with del(5q) (8/9, 88 %) and also good regarding

10

samples

7

10

Samples

6

10

Samples

5

10

Samples

4b

10

Samples

3b

10

Samples

2b

10

Samples

1b

10

Samples

4a

10

Samples

3a

10

Samples

2a

10

Samples

1a

110

samples

Fig. 1 One hundred ten cases previously diagnosed with MDS wereevaluated by 11 hematologists. Each observer evaluated 20 samples, andeach sample was analyzed independently by two morphologists. Thespecimens were exchanged in reciprocal manner by a central coordinator

Ann Hematol

Author's personal copy

RCMD (56/73 cases, 77%), as we noted in our previous study[11]. Again, there was a very good correlation in identifyingthe threshold of 15 % RS (kappa test=0.755). There was anagreement in one of the two cases with RAEB-1. The onlycase with 1 % PB blasts, classified as MDS-U, was alsoidentified. Despite the bias of selecting cases with <5 % BMblasts, there were five cases of RCMD classified as RAEB-1by the second observer.

Interobserver agreement in unilineage dysplasia

Dysplasia in only one lineage was assessed at least once in 33patients, by either of the two observers. However, there wasagreement in only 3/15 cases originally diagnosed with RARS(20 %) and in 3/7 with RCUD (42 %). In our previous study,we found 21 patients with diagnosis of unilineage dysplasia(RCUD or RARS) by at least one observer, with concordancein only four cases (two RCUD and two RARS) [11]. In order

to explore the characteristics of the patients with unilineagedysplasia, we combined the cases from both series because wewere using identical methodology. We excluded four patientsfrom our 2013 study (three RARS and one RCUD-T) becausethey were also included in the current series. Therefore, weconsidered 50 patients with the diagnosis of unilineage dys-plasia proposed by either of the two observers. Characteristicsof patients are described in Table 3. There were 23 cases withRCUD and 28 cases with RARS by at least one observer,including two cases classified as RCUD-A by the secondobserver. One case with MDS-U due to isolateddysgranulopoiesis and pancytopenia was considered asRCMDby the second observer. The whole series showed verylow agreement regarding RCUD (5/23, 21 %) and RARS(5/28, 18 %).

Regarding the samples diagnosed with RCUD, a discrep-ancy was seen in 18 cases; in most of cases (14/18), thediscrepancy was due to the detection of multilineage dysplasiaand in two cases, there was no agreement in the isolateddysplastic lineage observed. Two cases with RARS wereconsidered RCUD-A by the second observer. There wereeight cases classified at least once as RCUD-T, eight casesas RCUD-N, and nine cases as RCUD-A (Table 3).

Five of our 23 cases with RCUD showed two cytopenias,according to the threshold established by the IPSS. If weconsider normal ranges for hemoglobin (12 g/dL in femalesand 13 g/dL in males), neutrophils (2×109/L), and platelets(150×109/L), there were 12/23 cases with RCUD and withbicytopenia or pancytopenia. Using the normal PB ranges,3/11 cases with isolated cytopenia were not correlated with thesame dysplastic lineage. Indeed, only two of the five caseswith morphological agreement showed isolated cytopenia inconcordance with the type of dysplasia.

Cytogenetics was performed in 19/23 cases with RCUD.Cytogenetic abnormalities (-Y, del(5q), del(20q), i(21q), +8)

Table 1 Distribution of samples among the five pairs of observers, regarding 2008 WHO classification

2008 WHO classification

RCUD n=7 RARS n=15 RCMD n=73(33 RS)

RAEB1 n=2 Del5q n=9 MDS-U n=2

Pair 1. 20 suitable samples rate of discordance 9/20 (45 %) 2 (2) 6 (5) 11 (1) 0 1 (1) 0

Pair 2. 20 suitable samples rate of discordance 4/20 (20 %) 3 (1) 1 13 (2) 0 1 2 (1)

Pair 3. 20 suitable samples rate of discordance 4/20 (20 %) 1 (1) 3 (2) 14 (1) 0 2 0

Pair 4. 19 suitable samples rate of discordance 10/19 (52 %) 0 2 (2) 13 (7 RS) (7a) 1 (1) 3 0

Pair 5. 10 suitable samples Rate of discordance 4/10 (40 %) 1 1 (1) 8 (3) 0 0 0

Pair 6. 10 suitable samples rate of discordance 2/10 (20 %) 0 2 (2) 7 1 0 0

Pair 7. 9 suitable samples rate of discordance 3/9 (33 %) 0 0 7 (3) 0 2 0

Rate of agreement by 2nd observers (%) 3/7 (42) 3/15 (20) 56/73 (77) 1/2 (50) 8/9 (88) 1/2 (50)

The cases with discrepancies are shown in bracketsa 4/7 cases with RCMD and RS were diagnosed with RARS by the 2nd observer

RCUD refractory cytopenia with unilineage dysplasia, RARS refractory anemia with ring sideroblasts, RCMD refractory cytopenia with multilineagedysplasia, RAEB1 refractory anemia with excess blasts type 1, MDS-U unclassified-MDS

Table 2 Thirty-six caseswith interobserver dis-crepancies regarding2008 WHOclassification

The number of cases ineach category are shownin brackets

Referring diagnosis(n cases)

Review (n cases)

RCUD-A (2)

RCUD-N (1)

RCUD T (1)

RCMD (3)

RCUD-N (1)

RARS (12) RCMD (10)

RCUD-A (2)

RCMD (17) RARS (5)

RCUD (5)

RAEB-1 (5)

MDS/MPN (2)

RAEB-1 (1) RCMD (1)

MDS-U (1) RCMD (1)

MDS del5q (1) RCMD (1)

Ann Hematol

Author's personal copy

Table 3 Description of 50 cases with unilineage dysplasia by either of the two observers

Number 2008WHO 1stobserver

2008WHO 2ndobserver

Gender(F/M)

Hemoglobing/dL

WBC×109/L

ANC×109/L

Platelets×109/L

Ring SA1st observer

BM blasts1st observer

Karyotype IPSS R-IPSS 1stobserver (2ndobserver)d

1 RCUD-T RCUD-T M 16 18.8 14.9 18 <15 2.5c -Y Low Very low (low)

2 RCUD-T RCUD-T M 12.1 3.5 1.7 244 <15 0.5 -Y Low Very low

3 RCUD-A RCUD-A M 11.2 5.1 3.3 153 <15 0 N/A N/A N/A

4a RCUD-N RCUD-N M 8.6 4.3 2.7 10 <15 3c Del(5q) Int-1 Int (low)

5a RCUD-N RCUD-N F 9.8 3.2 1.7 119 <15 3 N/A N/A N/A

6 RCUD-A RCUD-N F 8.8 4.2 2.2 148 <15 0 Normal Low Low

7a RCUD-N RCUD-A M 10.6 7.4 5.0 199 <15 2 +8 Int-1 Low

8 RCUD-T RCMD M 13 10.7 8.0 24 <15 1 Normal Low Low

9 RCUD-A RCMD F 11.4 3.1 1.0 23 <15 <1 Normal Int-1 Low

10 RCUD-N RCMD M 14.3 7.5 4 101 <15 1 Normal Low Very low

11a RCUD-A RCMD M 14.1 6.3 3.3 83 <15 <1c Normal Low Very low (low)

12 RCMD RCUD-A F 11.5 5.6 3.4 353 ≥15b 1.2c Normal Low Low (low)

13 RCMD RCUD-T M 12.4 3.8 1.9 150 <15 <1 Normal Low Low

14 RCMD RCUD-A F 11.7 6.6 4.5 265 <15 0.5 Normal Low Low

15 RCMD RCUD-T M 14.7 4.5 2 108 <15 1.5 Normal Low Low

16 RCMD RCUD-N F 10.7 3.2 1.7 117 <15 4c Normal Low Low (low)

17a RCMD RCUD-N M 11.5 2.9 1.6 43 <15 1c Normal Low Low (low)

18a RCMD RCUD-N M 10.1 2.8 0.9 195 <15 1 Normal Low Very low

19a RCMD RCUD-T M 12.2 6.3 3.1 119 <15 1 i21q10 Int-1 Low

20a RCMD RCUD-T M 13.5 2.6 1.0 55 <15 1c Normal Int-1 Very low (low)

21a RCMD RCUD-T M 12.4 5.6 3.5 72 <15 0c N/A N/A N/A

22 MDS-U RCMD F 4.8 3.3 1.1 11 <15 1 Normal Int-1 Int

23 RARS RARS F 11.3 5.0 4.3 292 ≥15 1 N/A N/A N/A

24 RARS RARS M 10.6 8.6 4.6 253 ≥15 1 Normal Low Very low

25 RARS RARS F 10.3 3.6 1.4 118 ≥15 0 Normal low Very low

26a RARS RARS M 8.4 3.4 1.7 293 ≥15 3.5c +8 Int-1 Int (low)

27a RARS RARS M 7.9 3.2 1.5 270 ≥15 3c N/A N/A N/A

28 RARS RCMD M 10.4 5.7 3.2 279 ≥15 <1c Del(20q) Low Very low (low)

29 RARS RCMD M 9.8 7 1.2 449 ≥15 1 Normal Int-1 Low

30 RARS RCUD-A F 10.3 10.5 8.1 233 ≥15b 1 Del(20q) Low Very low

31 RARS RCUD-A M 10.6 4.9 2.3 154 ≥15b 1 NA NA NA

32 RARS RCMD M 9.7 6 3.7 321 ≥15 <1 Normal Low Low

33 RARS RCMD M 11.5 6.8 4.4 234 ≥15b 1c Normal Low Very low (low)

34 RARS RCMD M 12.7 4.9 3 161 ≥15 1 +8 Int-1 Low

35 RARS RCMD F 10.4 3.5 1.4 110 ≥15 1 Normal Low Very low

36 RARS RCMD F 10.7 6.5 3.2 252 ≥15 0 Normal Low Very low

37 RARS RCMD F 10.3 3.3 1.4 255 ≥15 0 +8,del(20q) Int-1 Low

38 RARS RCMD F 7.9 3.5 1.9 238 ≥15 0 Normal Low Low

39 RARS RCMD M 9.2 7.6 4.5 400 ≥15 0 Normal Low Low

40a RARS RCMD M 11.0 7.6 3.6 174 ≥15 3c N/A N/A N/A

41a RARS RCMD M 10.1 7 1.2 445 ≥15 <1 Normal low Very low

42a RARS RCMD F 7.5 4.0 2.1 213 ≥15 0 +8 Int-1 Int

43a RARS RCMD F 8.7 4.0 1.8 306 ≥15 0 Normal Low Low

44 RCMD RARS M 11.9 3.9 0.9 145 ≥15 2 Normal Low Low

45 RCMD RARS F 8.4 4.3 1.1 268 ≥15 0.3 +8,del(20q) Int-1 Low

46 RCMD RARS F 10.5 4 2 230 ≥15 0.8 Normal Low Low

47 RCMD RARS M 9.4 5.7 3.2 223 ≥15 0.2 Normal Low Low

Ann Hematol

Author's personal copy

were detected in six cases. Of note, the two cases with RCUD-Twith agreement by both observers showed -Y, although onlyone case showed thrombocytopenia. The IPSS and R-IPSSwere calculated in 19 cases, using the count of BM blastsreported by both observers. All cases except one were consid-ered as low or very low risk according R-IPSS. However,there were four cases with very low count of platelets, which isrecognized as adverse prognosis [13].

In cases with RARS, the main cause of discrepancy was theassessment of mutlilineage dysplasia instead of erythroiddyplasia (26/28 cases). Agreement in the percentage of RSwas obtained in 24/28 cases; there were only two patients withRARS that were considered as RCUD-A by the reviewer(Table 3). Patients diagnosed with RARS by at least one ob-server usually showed isolated anemia. Neutropenia with lessthan 1.5×109/L was present in seven cases but only one caseshowed absolute neutrophil count less than 1.0×109/L. Fourcases showed thrombocytopenia with less than 150×109/L, butthere were no cases with less than 100×109/L. Cytogeneticswas obtained in 23/28 cases. There were three cases with +8,two cases with del(20q), and in two cases, both abnormalitieswere found. The R-IPSS classified only 2/23 cases as interme-diate, and the remaining cases showed low or very low risk.

Interobserver agreement regarding percentage of blasts

Finally, we explored the concordance with the new morpho-logical categories using the threshold of 2 % BM blasts. Bothreferring observers and blinded reviewers found a higher

number of cases with ≤2 % BM blasts. But the categorieswith ≤2 % and >2–<5 % BM blasts were not reproducible:discordance rate 32/108 cases, kappa test = 0.277, p=0.004(Table 4).

Discussion

In our study, we explored the reliability of the 2008 WHOclassification to distinguish different categories of low-riskMDS, with less than 5 % BM blasts. A limitation of ourresults was the fact that each case was reviewed by only twoobservers for this paper, and therefore, it could imply that ourconclusions might be weak. Furthermore, especially becausewe are focusing on an infrequent category of MDS, that isMDS with unilineage dysplasia, the results might themselvesreflect these limitations in their accuracy. In order to reducethe bias of only two observers, we introduced a high numberof morphologists working in pairs. Most of the 108 casesselected by the 11 referring hematologists showedmultilineage dysplasia. Despite the limitation of having onlytwo observers, we found high concordance regarding RCMDand MDS with del5q categories, which was a similar obser-vation as in our previous study [11].

Although we were intentionally looking for cases withunilineage dysplasia, only seven cases with RCUD wereselected. This reflects the fact that RCUD is an infrequentcategory. The 2008 WHO classification quantifies these cases

Table 3 (continued)

Number 2008WHO 1stobserver

2008WHO 2ndobserver

Gender(F/M)

Hemoglobing/dL

WBC×109/L

ANC×109/L

Platelets×109/L

Ring SA1st observer

BM blasts1st observer

Karyotype IPSS R-IPSS 1stobserver (2ndobserver)d

48 RCMD RARS F 9.9 6.2 2.7 430 ≥15 2 Normal Low Low

49a RCMD RARS M 9.4 5.7 3.2 223 ≥15 0 N/A N/A N/A

50a RCMD RARS M 8.9 4.1 3.0 139 <15b 2 Normal Low Low

aCases obtained from our 2013 paper [11]b Cases with discordance in the percentage of RS, regarding the threshold of <15 %cCases with discordance in the percentage of BM blasts, regarding the threshold of ≤2 %d In cases with discordance in the percentage of BM blasts, the R-IPSS calculated with the BM blast count of the 2nd observer is shown in brackets

F female, M male, N/A not applicable, Int intermediate, Int-1 intermediate-1

Table 4 Description of BM blastcount by first and second observ-er. There was no agreement re-garding the threshold of ≤2%BMblasts. (kappa test 0.277 p=0.004)

Observer 2

≤2 % BM blasts >2 and <5 ≥5 % BM blasts Total

Observer 1 ≤2 % BM blasts 63 12 3 78

>2 and<5 15 12 1 28

≥5 % 1 0 1 2

Total 79 24 5 108

Ann Hematol

Author's personal copy

around 10% [4]. Data from the Dusseldorf registry shows thatless than 10 % of 1936 patients were RCUD, mostly RCUD-A. Cases with RCUD-T and RCUD-N were extremely rare,less than 1 % [14].

Our series included a high number of cases with RS (49/108, 45 %) but only 15 were considered RARS by the refer-ring hematologist. RARS represented 11 % in the Dusseldorfregistry [14] and 10 % in the Spanish MDS Registry[Bonanad, Spanish Group ofMDS, personal communication].Our previous paper showed a worrisome discordance in thediagnosis of unilineage dysplasia, but the number of patientswas too small, and no clear cut conclusions could be drawn.We combined the cases with unilineage dysplasia from bothstudies, since the design and methodology were identical.Finally, we could analyze 50 cases with unilineage dysplasia.To the best of our knowledge, this is the largest series of caseswith unilineage dysplasia reviewed by two observers.

Regarding RCUD, the characteristics of patients were veryheterogeneous. As previously described by Maassen et al.,bicytopenia and even pancytopenia are common, especially incases with RCUD-N and RCUD-T [14]. In accordance withthese authors, we also detected several discrepancies betweenthe type of cytopenia and the dysplastic lineage. Furthermore,our study demonstrated very low agreement between theobservers, as most of the cases were considered RCUD byone observer and RCMD by the other. The arbitrary thresholdof 10 % of dysplasia in each cell line might turn out quite low,especially when the investigators knew that the specimenswere previously diagnosed with MDS. Maassen et al. pro-posed to merge the three subcategories into RCUD, becausethey showed that prognosis was not significantly different[14]. However, the absence of reproducibility, together withthe absence of unequivocal features in a rare entity is a matterof concern. In practice, it seems very difficult to identifyaccurately the cases with RCUD, despite being demonstratedthat prognosis is better for patients with mild unilineagedysplastic changes [6]. The knowledge of the low-risk MDSis still incomplete, and with the limitations of our study, ourfindings suggest that probably the threshold of the criteriacurrently recommended by the WHO may not be completelyprecise and the categories grouped as RCUD might not cor-respond to a homogeneous entity.

Agreement in the diagnosis of RARS was also very low,because most cases were considered RCMD by either of thetwo morphologists. Unlike RCUD, our samples with RARSby either of the observers represented a very homogeneousgroup: most of them showed anemia as the sole abnormality,the platelet counts were usually normal, and there were nocases of severe neutropenia. Although the series was small,however, we note that +8 and del20q were recurrent cytoge-netic abnormalities. R-IPSS were of low or very low risk anddid not discriminate between unequivocal “pure” RARS andcases classified as RCMD by one of the two observers. Low

reproducibility of RARS was also noted by Sennent et al., intheir review of 50 cases performed by four observers [10]. Inour series, the discrepancy between RARS and RCMD withRS in a group of patients with very similar clinical featureshighlights a critical point, since each category implies differ-ent prognosis.

The recognition of RARS as a different entity frommultilineage dyspasia with RS was accomplished in 2000 byGerming et al. that was later on included in the WHO classi-fication [15]. The retrospective and prospective studies per-formed by Germing et al. led to incorporate the RCMD-RScategory into RCMD, since RS are not associated with favor-able prognosis in the setting of multilineage dysplasia [16,17]. Recently, Bacher et al. studied a series of 1082 patientswith RARS, RCMD-RS, and RCMD, demonstrating thatthese categories showed a similar proportion of good karyo-types according IPSS, and also there were similar favorableclinical outcomes during a follow-up of 2 years [18]. Theauthors conclude that their study is in accordance with thecombination of RCMD-RS and RCMD by the 2008 WHOclassification. This study could not distinguish specific genet-ic backgrounds or different clinical impact for RARS asdefined by theWHO, although the follow-up was rather short.Recently, several molecular abnormalities have been de-scribed in MDS; some of them are being associated withprognostic value, even independently of the current classifi-cation and scoring systems [19, 20]. Mutations of SF3B1 havebeen overrepresented in MDS with RS, and this finding hasbeen associated with favorable prognosis [21–23]. Indeed,SF3B1 and JAK2 mutations are independent factors for goodprognosis in RARS with marked thrombocytosis (RARS-T)[24].

Regarding the percentage of blasts, both observers in ourstudy reported counting ≤2%BM blasts more frequently (78and 79 cases) in the whole series. However, the review of thecases showed a striking difference between the categories with≤2 % BM blasts and >2 and <5 % (Table 4). The observationthat percentage of ≤2 % BM blasts are associated with betterprognosis was made by Matsuda et al. in 1998, even beforethe publication of the WHO classification and based on theFAB criteria [25]. Later on, in 2012, Germing et al. confirmedthe prognostic value of the threshold of 2 % BM blasts withdata from the Dusseldorf MDS Registry [26]. Finally, the R-IPSS, with a cohort of 7012 patients, demonstrated a signifi-cant difference of survival in a multivariate analysis betweenpatients with <2% and 2–5%BMblasts [8]. The enumerationof blast cells might sometimes be difficult, as noted by Muftiet al. [27]. These authors, on behalf of the InternationalWorking Group of Morphology of MDS (IWGM-MDS),proposed a consensus for the definition and enumeration ofBM blasts, which has been widely implemented in clinicalpractice [27]. The threshold of 5 % BM blasts in MDS hasbeen used since the definition of FAB criteria in 1982 [2], and

Ann Hematol

Author's personal copy

it is one of the most robust criteria for morphologists, and thereproducibility of this threshold have been proven many times[9–11, 28]. Counting less than 5 % is even a higher challengefor morphologists, and it could be explained the low rate ofagreement seen in our study regarding the threshold of 2 %BM blasts. Indeed, the recent implementation of the thresholdof 2 % could well mean that morphologists are still not toofamiliar with this protocol. But it is extremely important toobtain an accurate definition of the blast population for R-IPSS, since a small variation in the blast count implies differ-ences in prognosis. Our results show these discrepancies andhighlight the need to review carefully the cases with border-line count of blasts, which is close to 2 %. Whether flowcytometry immunophenotyping that is useful in both diagno-sis and prognosis of MDS and can also help in obtaining amore precise quantification of BM blasts is a matter of interest[29].

Finally, we emphasize the low rate of cases (only twocases) classified as MDS-U in our series that were indeedheterogeneous. As we noted in our previous paper [11], thedefinition of MDS-U has been clearly improved by the 2008WHO classification and the unclassifiable cases have becomeless frequent. Maasen et al. recommended eliminating thecategory MDS-U and classify these cases as RCMD, becausemorphological, clinical, and cytogenetic parameters are simi-lar for both categories [14].

In summary, our study reveals a high rate of discrepancy inthe reproducibility of RCUD and RARS. Probably, furtherstudies incorporating more observers are needed to elucidatethe reasons why it is difficult not only to diagnose but also toreproduce MDS with unilineage dysplasia. The threshold of2 % BM blasts as settled by the R-IPSS may not be easy toreproduce by morphology in real practice. It is to be hopedthat in the near future it will be possible to incorporate mo-lecular characteristics to the morphologic criteria in order toobtain a more accurate definition of the MDS classification.

Acknowledgments The Asociación Madrileña de Hematología yHemoterapia (AMHH) supported the courier service. The authors wouldlike to thank Angel Cedillo (AMHH) for his help in the exchange ofsamples among the hospitals and Dr. Santiago Bonanad for providing thedata from the Spanish Registry of MDS. We thank Lawrence JC Baronfor providing language support.

Conflict of interest The authors declare that they have no conflict ofinterest.

References

1. Garcia-Manero G (2014) Myelodysplastic syndromes: 2014 updateon diagnosis, risk-stratification, and management. Am J Hematol 89:98–108

2. Bennett JM, Catovsky D, Daniel MT, Flandrin G, Galton DA,Gralnick HR, Sultan C (1982) French–American–British (FAB)

Cooperative Group. Proposals for the classification of themyelodysplastic syndromes. Br J Haematol 51:189–199

3. Brunning RD, Bennett JM, Flandrin G, Matutes E, Head D,Vardiman JW, Harris NL (2001) Myelodysplastic syndromes.Introduction. In: Jaffe ES, Harris NL, Stein H, Vardiman JW (eds)World Health Organization classification of tumours. Pathology andgenetics of tumours of haematopoietic and lymphoid tissues. IARC,Lyon, pp 63–67

4. Brunning RD, Orazi A, Germing U, Le Beau MM, Porwit A,Baumann I, Vardiman JW, Hellstrom-Lindberg E (2008)Myelodysplastic syndromes/neoplasms, overview. In: SwerdlowSH, Campo E, Harris NL et al (eds) World Health Organizationclassification of tumours of haematopoietic and lymphoid tissues,4th edn. IARC, Lyon, pp 88–93

5. Greenberg P, Cox C, LeBeau MM, Fenaux P, Morel P, Sanz G et al(1997) International scoring system for evaluating prognosis inmyelodysplastic syndromes. Blood 89:2079–2088

6. Malcovati L, Germing U, Kuendgen A, Della Porta MG, Pascutto C,Invernizzi R et al (2007) Time-dependent prognostic scoring systemfor predicting survival and leukemic evolution in myelodysplasticsyndromes. J Clin Oncol 25:3503–3510

7. Schanz J, Tüchler H, Solé F, Mallo M, Luño E, Cervera J et al (2012)New comprehensive cytogenetic scoring system for primarymyelodysplastic syndromes (MDS) and oligoblastic acute myeloidleukemia after MDS derived from an international database merge. JClin Oncol 30:820–829

8. Greenberg P, Tuechler H, Schanz J, SanzG, Garcia-ManeroG, Solé Fet al (2012) Revised international prognostic scoring system formyelodysplastic syndromes. Blood 120:2454–2465

9. Naqvi K, Jabbour E, Bueso-Ramos C, Pierce S, Borthakur G, EstrovZ et al (2011) Implications of discrepancy in morphologic diagnosisof myelodysplastic syndrome between referral and tertiary care cen-ters. Blood 118:4690–4693

10. Senent L, Arenillas L, Luño E, Ruiz JC, Sanz G, Florensa L(2013) Reproducibility of the World Health Organization 2008criteria for myelodysplastic syndromes. Haematologica 98:568–575

11. Font P, Loscertales J, Benavente C, Bermejo A, Callejas M, Garcia-Alonso L et al (2013) Inter-observer variance with the diagnosis ofmyelodysplastic syndromes (MDS) following the 2008 WHO classi-fication. Ann Hematol 92:19–24

12. Landis JR, KochGG (1977) Themeasurement of observer agreementfor categorical data. Biometrics 33:159–174

13. Gonzalez-Porras JR, Cordoba I, Such E, Nomdedeu B, Vallespi T,Carbonell F, Spanish Myelodysplastic Syndrome Registry et al(2011) Prognostic impact of severe thrombocytopenia in low-riskmyelodysplastic syndrome. Cancer 117:5529–5537

14. Maassen A, Strupp C, Giagounidis A, Kuendgen A, Nachtkamp K,Hildebrandt B et al (2013) Validation and proposals for a refinementof the WHO 2008 classification of myelodysplastic syndromes with-out excess of blasts. Leuk Res 37:64–70

15. Germing U, Gattermann N, Aivado M, Hildebrandt B, Aul C (2000)Two types of acquired idiopathic sideroblastic anaemia (AISA): atime-tested distinction. Br J Haematol 108:724–728

16. Germing U, Gattermann N, Strupp C, Aivado M, Aul C (2000)Validation of the WHO proposals for a new classification of primarymyelodysplastic syndromes: a retrospective analysis of 1600 patients.Leuk Res 24:983–992

17. Germing U, Strupp C, Kuendgen A, Isa S, Knipp S, Hildebrandt Bet al (2006) Prospective validation of the WHO proposals for theclassification of myelodysplastic syndromes. Haematologica 91:1596–1604

18. Bacher U, Kern W, Alpermann T, Schnittger S, Haferlach C,Haferlach T (2012) Prognosis of MDS subtypes RARS, RCMDand RCMD-RS are comparable but cytogenetics separates a sub-group with inferior clinical course. Leuk Res 36:826–831

Ann Hematol

Author's personal copy

19. Bejar R, Stevenson K, Abdel-Wahab O, Galili N, Nilsson B, Garcia-Manero G et al (2011) Clinical effect of point mutations inmyelodysplastic syndromes. N Engl J Med 364:2496–2506

20. Bejar R, Stevenson K, Caughey B, Abdel-Wahab O, Steensma D,Galili N et al (2012) Validation of a prognostic model and the impactof mutations in patients with lower-risk myelodysplastic syndromes.J Clin Oncol 30:3376–3382

21. Papaemmanuil E, Cazzola M, Boultwood J, Malcovati L, Vyas P,Bowen D et al (2011) Somatic SF3B1 mutation in myelodysplasiawith ringed sideroblasts. N Engl J Med 365:1384–1395

22. YoshidaK, SanadaM, Shiraishi Y, NowakD, Nagata Y, Yamamoto Ret al (2011) Frequent pathway mutations of splicing machinery inmyelodysplasia. Nature 478:64–69

23. Patnaik MM, Hanson CA, Sulai NH, Hodnefield JM, Knudson RA,Ketterling RP et al (2012) Prognostic irrelevance of ring sideroblastpercentage in World Health Organization defined myelodysplasticsyndromes without excess blasts. Blood 119:5674–5677

24. Broséus J, Alpermann T, Wulfert M, Florensa Brichs L, Jeromin S,Lippert E et al (2013) Age, JAK2(V617F) and SF3B1 mutations arethe main predicting factors for survival in refractory anaemia with ringsideroblasts and marked thrombocytosis. Leukemia 27:1826–1831

25. Matsuda A, Jinnai I, Yagasaki F, Kusumoto S, Minamihisamatsu M,Honda S et al (1998) Refractory anemia with severe dysplasia:clinical significance of morphological features in refractory anemia.Leukemia 12:482–485

26. GermingU,KuendgenA (2012) Prognostic scoring systems inMDS.Leuk Res 36:1463–1469

27. Mufti GJ, Bennett JM, Goasguen J, Bain BJ, Baumann I, Brunning Ret al (2008) International Working Group on Morphology ofMyelodysplastic Syndrome. Diagnosis and classification ofmyelodysplastic syndrome: International Working Group onMorphology ofmyelodysplastic syndrome (IWGM-MDS) consensusproposals for the definition and enumeration of myeloblasts and ringsideroblasts. Haematologica 93:1712–1717

28. Howe RB, Porwit-MacDonald A, Wanat R, Tehranchi R, Hellström-Lindberg E (2004) The WHO classification of MDS does make adifference. Blood 103:3265–3270

29. Van de Loosdrecht AA, Ireland R, Kern W, Della Porta MG, AlhanC, Balleisen JS et al (2013) Rationale for the clinical application offlow cytometry in patients with myelodysplastic syndromes: positionpaper of an International Consortium and the European LeukemiaNetWorking Group. Leuk Lymphoma 54:472–475

Ann Hematol

Author's personal copy