[1538] Specific Criteria Improve Interobserver Reproducibility in Myeloproliferative Neoplasm (MPN) Megakaryocyte Morphologic Assessment

Nikhil Sangle, Robert Schmidt, Rod Miles, Josef Prchal, Sherrie Perkins, Attilio Orazi, Mohamed Salama. University of Utah, Salt Lake City, UT; Weill Cornell Medical Center, New York City, NY

Background: The 2008 WHO Classification emphasizes the histotopography and morphology of megakaryocytes (MK) in the classification of MPN. Previous studies reported significant interobserver variability in the assessment of MPN morphologic features. In this study, we evaluated the interobserver reproducibility for morphologic characteristics of MK on bone marrow biopsies.
Design: 72 bone marrow core biopsies from patients diagnosed with MPN were revisited by 4 hematopathologists (SLP, AO, RRM, MES) with each case reviewed by at least 2. Criteria for morphologic features were predefined: normal MK number = 4/HPF in 10 HPFs; clusters = juxtaposed cell surfaces of ≥3 adjacent MK, with or without intervening marrow cells in loose or tight clustering, respectively. Other parameters including size, nuclear lobation, staghorn or cloudlike nuclei, and dysplasia (cytologic atypia) were not further defined. Interobserver agreement was assessed using kappa statistics for each of the morphologic features.
Results: Interobserver agreement was significantly better than would be expected by chance for majority of the parameters, including MK number, size, clustering, cluster type, lobation and recognition of staghorn nuclei (p≤0.002).

MK Morphology	kappa	z	p-value	aggrement
Number	0.47	4.42	0.000	0.85
Size	0.29	2.95	0.002	0.64
Clustering	0.63	5.06	0.000	0.86
Cluster type	0.34	4.11	0.000	0.56
Lobation	0.45	4.57	0.000	0.67
Staghorn	0.20	1.98	0.024	0.51
Cloudlike	0.11	1.13	0.1290	0.41
Dysplasia	-0.02	-0.23	0.5920	0.44

MK = megakaryocyte

Agreement was highest for the objectively predefined parameters: MK number (85%), clustering (86%), and clustering type (56%). Agreement was also significant for MK size (64%), but was poor for identification of staghorn, cloudlike & dysplastic nuclei (44-51%).
Conclusions: MK morphologic assessment showed statistically significant interobserver agreement for 6/8 parameters including number, clustering type, and size, but identification of other features important in classification of MPN (i.e. staghorn, cloudlike and dysplastic nuclei) showed poor agreement. Thus, interobserver agreement on the morphologic characteristics of MPNs may be improved by using specifically defined criteria.
Category: Hematopathology

Monday, March 19, 2012 1:15 PM

Platform Session: Section C, Monday Afternoon