Abstract

BACKGROUND: MMM is a rare chronic myeloproliferative disease with heterogeneous clinical presentation, natural history and prognosis. Knowledge about MMM is burdened by limited sample size and selection bias of the available studies.

AIM: To get a clinical classification of MMM for biological and clinical studies.

METHODS: The database of the RIMM, a population-based prospective cohort started in 1999, was enquired and data of 861 patients at diagnosis were extracted. The EM-algorithm (WEKA software) was used for a non-supervised data-mining based on 11 variables: age, spleen size, constitutional symptoms, hemoglobin level, platelet and blood cell count, LDH and percentages of erythroblasts, immature myeloid cells, blasts, and basophils. Differences among clusters were first tested with ANOVA or n-way Chi-square and subsequently investigated with t-test or 2x2 Chi-square. Survival differences among clusters were tested through Cox univariate and multivariate regression analysis.

RESULTS: Five clusters (C) were identified: C1 identified younger patients with thrombocytosis and scarce myeloproliferation. C2 also identified patients with a high platelet count, but older and anemic. C3 identified patients with intense myeloproliferation (marked leukocytosis and splenomegaly), while C4 (5% of the patients) collected patients with an even more severe myeloproliferation, associated with a high number of circulating immature myeloid cells and blasts. C4 was also characterized by frequent chromosome abnormalities, anemia and thrombocythopenia. Finally, C5 identified patients with bilinear/trilinear cytopenias. The clusters also differed for variables not incorporated by the data-mining process, i.e. CD34+ cell count (p=0.028); i.e. the rate of homozygous V617F mutation of JAK2 gene, which was significantly higher in C3 and C4 (32.4% vs 12.5%, p=0.019). The clusters also significantly differed for overall survival (p<0.0001), independently of patients’ age. Finally, referral patterns differed among clusters, with C5 and C1 patients being mostly referred to Internal Medicine and Hematology units (p<0.01), respectively.

CONCLUSIONS: Thanks to a large and unbiased repository of cases, we identified 5 classes of MMM patients with specific clinical presentation. The biologic and prognostic value of the clusters deserves further investigations at longer follow-up periods.

Clinical parameters in the different clusters (means)

CLUSTERAGE (yrs) *HB (g/l) */PLT*10(9)*WBC *10(9) *SPLEEN (cm) *BLASTS*(%)CD34 %/μlSEVERITY SCORE§ *
* p<0.0001 § The score considers HB, WBC, PLT and SPLEEN (Barosi, Leuk Lymph 2002). 
56.7 13.3 /583 11.5 3.5 0.21 0.85 /63 1.25 
72.1 10.6 /482 18.3 2.7 0.82 1.12 /297 2.04 
65.1 10.8 /303 18.3 9.9 1.47 1.7 /313 2.54 
67.6 8.4 /110 27.5 13.4 5.66 2.72 /275 4.30 
68.2 8.4 /134 4.2 6.8 0.78 2.92 /52 3.36 
All 65.5 10.7 /362 13.5 6.5 1.18 1.78 /200 2.36 
CLUSTERAGE (yrs) *HB (g/l) */PLT*10(9)*WBC *10(9) *SPLEEN (cm) *BLASTS*(%)CD34 %/μlSEVERITY SCORE§ *
* p<0.0001 § The score considers HB, WBC, PLT and SPLEEN (Barosi, Leuk Lymph 2002). 
56.7 13.3 /583 11.5 3.5 0.21 0.85 /63 1.25 
72.1 10.6 /482 18.3 2.7 0.82 1.12 /297 2.04 
65.1 10.8 /303 18.3 9.9 1.47 1.7 /313 2.54 
67.6 8.4 /110 27.5 13.4 5.66 2.72 /275 4.30 
68.2 8.4 /134 4.2 6.8 0.78 2.92 /52 3.36 
All 65.5 10.7 /362 13.5 6.5 1.18 1.78 /200 2.36 

Author notes

Corresponding author