LAUSR

The myelodysplastic syndromes (MDS) are heterogeneous clonal myeloid neoplasms characterized by ineffective haematopoiesis, dysplasia and risk of leukaemia transformation (Haider et al, 2017). Several prognostic models using clinical and laboratory variables, such as the International Prognostic Scoring System (IPSS) and Revised IPSS (IPSS-R) (Greenberg et al, 2012) are used to predict outcomes. However, prognostic models remain imperfect and fail to identify some lower-risk MDS patients who eventually have worse clinical outcomes than their lower-risk MDS classification. Red blood cell (RBC) distribution width (RDW) describes the size heterogeneity of RBCs and is useful in the differential diagnosis of anaemia. A recent study in healthy individuals reported TP53 and U2AF1 mutations and a high RDW correlated with risk of progression to acute myeloid leukaemia (AML) (Abelson et al, 2018). Some older persons progress from age-related clonal haematopoiesis (ARCH) to clonal cytopenia of undetermined significance (CCUS) and on to lowand high-risk MDS before developing AML, and then, we hypothesized that RDW might play a role in MDS. We analysed data from 321 consecutive new-diagnosed MDS subjects without RBC transfusions at our centre from December 2009 to December 2017. Diagnoses were reclassified according to 2016 World Health Organization (WHO) definitions (Arber et al, 2016). Clinical variables are displayed in Table SI. All subjects had normal or high serum iron and normal serum folate and vitamin B12 levels. Of the subjects with evaluable cytogenetics, 279 (86 9%) were classified using the IPSS-R criteria (Table SI; Figure S1) (Greenberg et al, 2012). In the IPSS-R classification, we combined subjects with scores <3 5 into a lower-risk cohort; the higher-risk cohort had scores ≥3 5. Therapeutic regimen adjustments were permitted based on IPSS or IPSS-R risk categories, patient age, performance status and presence of comorbidities. Treatments are shown in Table SI. Follow-up data were available for 280 (87 2%) subjects with a median follow-up of survivors of 14 months (range, 1–85 months). Subjects gave written informed consent compliant with the Declaration of Helsinki. Red blood cell distribution width was measured by Sysmex XN-9000 Blood Cell Analyser. We used a bio-informatics tool X-tile to define a RDW threshold of 14 5% to predict prognosis (Camp et al, 2004). Subjects were classified as RDW (<14 5%; N = 74) or RDW (≥14 5%; N = 247) cohorts. RDW subjects had lower haemoglobin concentrations (78 [range; 28–147] g/l vs. 112 [47–153] g/l, P < 0 001; Table SI), higher neutrophil counts (1 26 [0 01–9 14] 910/l vs. 0 915 [0 09–11 17] 910/l; P = 0 004, Table SI), and poorer IPSS-R karyotypes (P = 0 012, Table SI). Next, we divided subjects into cohorts with bone marrow blasts <5% (N = 189) and ≥5% (N = 132). RDW subjects were more likely to have lower haemoglobin concentrations and higher IPSS-R risks in the bone marrow blast < and ≥5% cohorts (both P < 0 001 for haemoglobin, P = 0 006 and 0 007 for IPSS-R respectively; Tables SII and SIII) while poorer IPSS-R karyotypes in the <5% cohort (P = 0 019; Table SII) and higher neutrophil counts and lower platelet counts were found in the ≥5% cohort (P = 0 029 and P = 0 009; Table SIII). DNA was derived from bone marrow mononuclear cells obtained at diagnosis. 112 genes were sequenced in all subjects (Li et al, 2018). Distributions of mutations with a variable allele frequency (VAF) > 2% in the RDW and RDW groups are shown in Figure S2. U2AF1 and NRAS were mutated more frequently in RDW subjects (P < 0 001 and P = 0 045; Figure S2). In subset analysis, RDW subjects had mutations in U2AF1 and SF3B1 more often in the <5% bone marrow blasts cohort (P = 0 007, P = 0 02; Fig 1A) while U2AF1 and NRAS were more often mutated in the ≥5% cohort (P = 0 023; P = 0 034; Figure S2). Increasing RDW was associated with worse survival in the total cohort in univariate analyses (P = 0 01, Fig 1B and Table SIV). Mutated ASXL1 and TET2 were also associated with worse survival (Table SIV). High RDW remained its significance in survival after adjusting for other clinical and laboratory variables and for mutation topography (Table SIV). In subset analyses, RDW subjects had worse survivals in the <5% bone marrow blast cohort (P = 0 008; Fig 1C and Table I), but not in the ≥5% cohort (P = 0 07; Fig 1D). In the <5% bone marrow blast cohort, adverse mutations were ASXL1, TET2 and PTPN11 in univariate analyses (Table I). High RDW was still an independent adverse variable after adjusting for clinical and laboratory variables and mutation topography (Table I). Based on the b-coefficients (Table SV) of the above-mentioned prognostic factors, a linear risk score for subjects with bone marrow blasts <5% was developed: age ≥60 years 9 1 + IPSS-R cytogenetics poor 9 2/very poor 9 2 5 + RDW 9 2 5 + ASXL1 9 1 5 + PTPN11 9 3. Subjects were divided into three risk cohorts: (i) low: score <1 5; (ii) Correspondence