Citations

11012 Background: The proportion of women in the field of hematology and oncology (H&O) has increased over recent decades, but representation by women in leadership positions remains a challenge. Our aim was to examine representation of winners of recognition awards by gender and race from the major international hematology and oncology societies. Methods: Published award recipients from the seven major H&O societies were reviewed, with 26 years of data included (1994-2019). Awardee demographics and academic rank were collected and included in the analysis. Gender was determined based on awardee full name and verified by public data. Chi-square and Cochran-Armitage tests were used to analyze the data. Results: Over the past 26 years, 942 awards were presented at the 7 major H&O societies. We excluded 27 gender specific awards from the analysis. Of the 915 awardees included in analysis, award recipients were overwhelmingly men (77.9%) and non-Hispanic white (84.7%). Gender breakdown by society is described in table. ESMO and ASTRO represented the lowest distribution of women at 9.1% and 11.1%, respectively. Women awardees received 30.3% of the awards categorized as humanistic and education-related, while only receiving 16.0% of awards in the basic sciences category (p<0.01). The Cochran-Armitage test demonstrated an upward trend in the number of women awardees, from 10% between 1994-1998 to 25.6% between 2014-2019 (p=0.0004). Over the past five years, the average proportion of women medical oncologists was 35.6% per the AAMC. In this time period, women oncologists have received only 24.0% of all awards, suggesting the awardees included in the study period do not represent the proportion women in the field (p=0.00424). Black, Hispanic, and Asian awardees represented 3.7%, 3.3%, and 6.8% of the total awardees, respectively. Of the 64 Black and Hispanic awardees, 60.9% of the awards were for investigating healthcare disparities and only 4.6% were given for basic science research. Conclusions: During our study period, women physicians and investigators were less likely to receive recognition awards from the seven major H&O societies compared to men. We also observed a considerable low proportion of minority awardees in all oncology subspecialties. While the proportion of women awardees has increased over time, significant underrepresentation remains. [Table: see text]

e19538 Background: Chronic neutrophilic leukemia (CNL) is a rare and aggressive myeloproliferative neoplasm with poor survival. No standard of care exists. Case reports describe allogeneic stem cell transplantation (ASCT) as the only curative therapy. Using the National Cancer Database (NCDB), we studied prognostic factors and survival outcomes of CNL. Methods: Patients diagnosed in 2004-2015 were included. 9963/3 was the ICD-O-3 code. Overall survival (OS) was analyzed using the Kaplan-Meier method. Hazard ratios (HR) with confidence intervals (CI) were calculated using Cox proportional hazards model. Variables significant in univariate analysis were included in a multivariate analysis. Data on incidence was obtained from the Surveillance, Epidemiology, and End Results Program (SEER) 21 Registries (2004-2015). The analysis was done with SEER*Stat software. Incidence was age-adjusted to the U.S. 2000 standard population. Results: A total of 121 CNL patients were in NCDB. The median age at diagnosis was 70 years (range 21-90); 74 (61%) were males and 107 (88%) were ≥ 50 years. The median follow up time was 5.8 years (interquartile range [IQR] 2.9-7.4). 32 (26%) had a Charlson Deyo score (CDS) > 1. 80 (70%) had government insurance, while 34 (30%) had private insurance. Median OS was 2.2 years (IQR 0.8-6.7). OS was 70% at 1 year, 29% at 5 years, and 11% at 10 years. The year of diagnosis was not associated with improved OS (HR: 0.95 (95% CI: 0.88-1.02), p =0.14). 71 (59%) patients received chemotherapy. The median time to chemotherapy from the day of diagnosis was 15 days (IQR 7-35). The OS of those who received chemotherapy was 71% at 1 year and 27% at 5 years versus 69% and 32% at 1- and 5- years for those who did not ( p= 0.38). 2 patients received ASCT (age 52 and 61 years) as initial therapy and both were alive at 5 years. Factors predicting inferior OS on univariate analysis were age ≥ 50 years at diagnosis ( p= 0.005), male ( p< 0.001), CDS > 1 ( p= 0.038), and government insurance ( p= 0.004). The table shows the multivariate (MV) analysis. The overall incidence of CNL was 0.1 cases/1,000,000 individuals and did not change significantly from 2004-2015. The incidences by male, female, White, Black, and American Indian/Alaska Native were also 0.1 cases/1,000,000 individuals. Conclusions: OS and incidence of CNL have not changed over the years. Age ≥ 50 years, male gender, and use of government insurance predicted inferior OS. [Table: see text]

e19088 Background: Significant disparities in the stage at diagnosis and survival outcomes exist between various groups diagnosed with hepatocellular carcinoma (HCC) based on sex, race, insurance coverage, and marital status. Previously published data shows disparities more in patients who are African-American, Asian and low income. Little data has been published on factors that influence timeliness to treatment initiation in patients diagnosed with HCC. Methods: Retrospective analysis was performed on 96,586 patients diagnosed with HCC from 2004-2014 using data in the National Cancer Database (NCDB). Time to treatment was divided into two categories: ≤40 days (early, n = 66322) and > 40 days (late, n = 30264). We carried out univariate and multivariate analyses to compare demographic, clinical, treatment, and facility-related factors influencing the timeliness of treatment initiation in HCC. Results: Univariate analysis revealed a significant difference in time to treatment initiation based on age, race, income, insurance status, type of area, geographic region, type of facility, and cancer stage (P < 0.001). Multivariate analysis showed that household income < $30,000/year, Pacific region, urban area of residence, black race, age 70-79 years old or ≥80 years old, academic centers, stage II disease and medicaid insurance were all factors with longer time to treatment initiation. Discussion: HCC is the sixth most common cancer and the second leading cause of cancer mortality worldwide. The 5-year relative survival for localized, regional, and distant stages are 32.6%, 10.8% and 2.4% respectively. In our results we noted significant disparities in time to treatment with respect to various socioeconomic factors. These results are comparable to that reported in other cancer types. Further subgroup analysis of our data shows 2581 patients who received their first treatment after 200 days. In a similar study done in Taiwan, those treated more than 181 days and 61–180 days after diagnosis had a 1.68 and a 1.39 increased risk of death respectively, which were statistically significant. Many factors contribute to delayed treatment, some of which are difficult to circumvent, however where possible efforts should be made to overcome these. Conclusions: Various socioeconomic factors were found to affect the time to treatment initiation in HCC patients. The next step would be to strategically implement policies and practices to address these factors.

Background: Disease progression prediction based on neuroimaging biomarkers is vital in Alzheimer’s disease (AD) research. Convolutional neural networks (CNN) have been proved to be powerful for various computer vision research by refining reliable and high-level feature maps from image patches. Objective: A key challenge in applying CNN to neuroimaging research is the limited labeled samples with high dimensional features. Another challenge is how to improve the prediction accuracy by joint analysis of multiple data sources (i.e., multiple time points or multiple biomarkers). To address these two challenges, we propose a novel multi-task learning framework based on CNN. Methods: First, we pre-trained CNN on the ImageNet dataset and transferred the knowledge from the pre-trained model to neuroimaging representation. We used this deep model as feature extractor to generate high-level feature maps of different tasks. Then a novel unsupervised learning method, termed Multi-task Stochastic Coordinate Coding (MSCC), was proposed for learning sparse features of multi-task feature maps by using shared and individual dictionaries. Finally, Lasso regression was performed on these multi-task sparse features to predict AD progression measured by the Mini-Mental State Examination (MMSE) and the Alzheimer’s Disease Assessment Scale cognitive subscale (ADAS-Cog). Results: We applied this novel CNN-MSCC system on the Alzheimer’s Disease Neuroimaging Initiative dataset to predict future MMSE/ADAS-Cog scales. We found our method achieved superior performances compared with seven other methods. Conclusion: Our work may add new insights into data augmentation and multi-task deep model research and facilitate the adoption of deep models in neuroimaging research.

TPS7574 Background: Pts with LBCL who have persistent disease assessed by dynamic PET after rituximab-based induction therapy have an increased risk of death (Casasnovas, et al. Blood. 2017). Axi-cel is an autologous anti-CD19 chimeric antigen receptor (CAR) T cell therapy approved for the treatment of pts with relapsed/refractory LBCL with ≥ 2 prior systemic therapies. In ZUMA-1, the registrational study of axi-cel in pts with refractory LBCL, the objective response rate (ORR) was 91% (70% complete response [CR] rate) in pts with double-expressor or high-grade LBCL with ongoing responses in 48% after a median follow-up of 27.1 mos (Locke FL, et al. Lancet Oncol 2019). Furthermore, pts with fewer prior lines of therapy and lower tumor burden had higher rates of ongoing responses and manageable safety (Locke et al. ASCO 2018. 3039). ZUMA-12 will investigate the efficacy and safety of axi-cel as a first-line therapy in newly diagnosed pts with high-risk LBCL who have PET-positive disease after 2 cycles of induction therapy. Methods: This Phase 2 study has a planned enrollment of ≈40 pts aged ≥ 18 y with high-risk LBCL, defined by the presence of MYC and BCL2 and/or BCL6 translocations by FISH or an IPI score ≥ 3 any time before enrollment, and an ECOG performance status of 0 – 1. Before enrollment, pts must have a Deauville score of 4 – 5 (Barrington SF et al. J Clin Oncol. 2014) after 2 cycles of chemoimmunotherapy that includes an anti-CD20 monoclonal antibody and anthracycline. After leukapheresis, pts with bulky or rapidly progressing disease may receive optional non-chemotherapy bridging therapy. Following conditioning therapy with cyclophosphamide (500 mg/m2) and fludarabine (30 mg/m2) for 3 days, pts will receive a single infusion of axi-cel at a target dose of 2 × 106 CAR T cells/kg. The primary endpoint is investigator-assessed CR rate per the Lugano classification (Cheson et al. J Clin Oncol. 2014). Key secondary endpoints include ORR, duration of response, event-free survival, progression-free survival, overall response, safety, relapse with CNS disease and levels of blood CAR T cells and serum cytokines over time. Accrual is ongoing. Clinical trial information: NCT03761056.

Summary Survival prediction in essential thrombocythaemia (ET) and polycythaemia vera (PV) is currently based on clinically‐derived variables; we examined the possibility of integrating genetic information for predicting survival. To this end, 906 molecularly‐annotated patients (416 Mayo Clinic; 490 University of Florence, Italy), including 502 ET and 404 PV, were recruited. Multivariable analysis identified spliceosome mutations to adversely affect overall ( SF3B1 , SRSF2 in ET and SRSF2 in PV) and myelofibrosis‐free ( U2AF1 , SF3B1 in ET) survival; TP53 mutations predicted leukaemic transformation in ET; “adverse” mutations occurred in 51 (10%) ET and 8 (2%) PV patients. We confirmed the independent survival effect of adverse mutations [hazard ratio (HR) 2·4, 95% CI 1·6–3·5], age >60 years (6·6, 4·6–9·7), male sex (1·8, 1·3–2·4) and leukocytosis ≥11 × 10 9 /l (1·6, 1·1–2·2), in ET, and adverse mutations (7·8, 3·1–17·0), age >67 years (5·4, 3·6–8·1), leukocytosis ≥15 × 10 9 /l (2·8, 1·8–4·2) and thrombosis history (2·0, 1·4–2·9), in PV. HR‐based risk point allocation allowed development of three‐tiered mutation‐enhanced international prognostic systems (MIPSS) which were validated in both cohorts and performance was shown to be superior to conventional scoring systems. Spliceosome mutations enhance survival prediction in ET and PV and identify patients at risk for fibrotic progression. TP53 mutations predict leukaemic transformation in ET.

AbstractDisease overviewChronic neutrophilic leukemia (CNL) is a rare, often aggressive myeloproliferative neoplasm (MPN) defined by persistent mature neutrophilic leukocytosis, bone marrow granulocyte hyperplasia, and frequent hepatosplenomegaly. The seminal discovery of oncogenic driver mutations in colony‐stimulating factor 3 receptor (CSF3R) in the majority of patients with CNL in 2013 anchored a new scientific framework, deepening our understanding of its molecular pathogenesis, providing a diagnostic biomarker, and rationalizing the use of pharmacological targeting.Diagnostic criteriaIn 2016, the World Health Organization (WHO) included the presence of activating CSF3R mutations as a central diagnostic feature of CNL. Other criteria include leukocytosis of ≥25 × 109/L comprising >80% neutrophils with <10% circulating precursors and rare blasts, and absence of dysplasia or monocytosis, while not fulfilling criteria for other MPN.Disease updatesIncreasingly comprehensive genetic profiling of CNL has disclosed a complex genomic landscape and additional prognostically relevant mutational combinations. Though prognostic determination and therapeutic decision‐making remain challenging, emerging data on prognostic markers and the use of newer therapeutic agents, such as JAK inhibitors, are helping to define state‐of‐the‐art management in CNL.

Abstract Introduction: Relapse is the primary cause of death after reduced intensity conditioning (RIC) hematopoietic cell transplantation (HCT). Documented responses have been reported after combination chemotherapy, hypomethylating agents, donor lymphocyte infusions, and second transplants. For patients with early relapse after RIC HCT while still on immunosuppression (IS), IS taper alone may induce significant graft-versus-tumor (GVT) activity. We describe the clinical characteristics and subsequent course of 48 patients with disease recurrence that responsed to IS taper alone without any additional chemotherapy, radiation, or donor lymphocyte infusion. Methods: We reviewed medical records of all patients with frank histologic or radiographic evidence of relapse (n=477) and those with impending relapse defined by development of cytopenias with a fall in donor derived hematopoietic cell chimerism (n=87) within one year of allogeneic RIC HCT who were on IS at time of relapse between January 1, 2004 and December 31, 2013 at Dana Farber Cancer Institute/Brigham and Women's Hospital. Complete response after IS taper was defined as complete recovery of peripheral blood counts, bone marrow biopsy with absence of disease, negative radiologic imaging, or chimerism recovery to above 90 percent. Partial response was defined as improvement in disease burden in bone marrow or by imaging. Survival probabilities were calculated using the Kaplan Meier method. Results: Forty-eight patients (37 male, 11 female) who relapsed within one year of HCT responded to IS taper alone. Of these, 14 had impending relapse and 34 had frank relapse at time of IS taper. The median age of patients was 62 years at time of transplant. The most frequent indication for transplant was MDS (n=14) followed by AML (n=11), non-Hodgkin's lymphoma (n=8), CLL (n=6), Hodgkin's disease (n=4), CML (n=3) and multiple myeloma (n= 2). At time of transplant, 9 patients had standard risk disease and 39 had high risk disease. All patients except one underwent conditioning with busulfan (≤ 6.4 mg/kg) and fludarabine. Five patients had graft versus host disease (GVHD) prior to relapse. The median time to frank or impending relapse was 105 days (range 57-360) after HCT. The median interval between starting and completing IS taper was 30 days (range 0-251), but 13 patients could not come off IS completely due to GVHD. The median time to documented response after initiation of IS taper was 77 days (range 14-189). Thirty-five patients had a complete response and the remaining 13 patients had a partial response. Nine patients subsequently relapsed late after initial response to IS taper at a median time of 2.38 years (range 0.88-3.95). Forty-five of the 48 patients developed or had a flare of previous GVHD as a consequence of IS taper. Of these, 26 patients had grade II-IV acute GVHD and 19 had chronic GVHD. The median time to developing GVHD after starting IS taper was 39 days (range 7-261 days). Twenty-five patients died during follow-up (10 from GVHD, 6 from disease, 7 from infection, 1 from congestive heart failure and 1 unknown). The median follow-up time among survivors was 4 years (range 1.6, 9.4). The median overall survival (OS) time from IS taper was 4.78 years (95% CI 2.93-7.35, see figure). The 4 year overall survival was 56%. There was no difference in survival for those with frank or impending relapse. Conclusion: Immunosuppression taper alone in patients who relapse early after RIC HCT can produce durable remissions but is almost always associated with development of GVHD. Understanding the clinical, molecular and immunologic characteristics of underlying disease will allow us to better predict who will respond to this manipulation. When feasible, this strategy should be given sufficient time to allow for a GVT response to develop in order to possibly avoid the adverse effects of more aggressive therapies. Figure 1 Figure 1. Disclosures No relevant conflicts of interest to declare.

Background: Early data from anti-CD19 chimeric antigen receptor (CAR) T-cell trials suggest that concurrent infection can lead to poor outcomes. 1,2 The relationship between CAR T-cell mediated inflammatory responses and infections is not well-established. With CAR T-cell therapies more readily available, practitioners must identify which patient, disease and CAR T-cell specific parameters are associated with an increased risk of infection to optimize outcomes. We provide a comprehensive analysis of infection risk within the first 30 days after CAR T-cell infusion across multiple types of CAR T-cell trials targeting distinct antigens. Methods: This was a single-center, retrospective study conducted at the National Cancer Institute analyzing infectious complications in subjects who underwent CAR T-cell therapy on one of four phase I clinical trials, targeting CD19, CD22, B-cell maturation antigen (BCMA) or disialoganglioside (GD2) from 2012 though 2018. Baseline characteristics are in Table 1. The primary objective was to establish the incidence of infections between initiation of lymphodepleting (LD) chemotherapy through day 30 after CAR T-cell infusion. The secondary objective was to identify risk factors for infection. Patients were censored at relapse and/or initiation of alternative therapy, including treatment for relapse or consolidative hematopoietic stem cell transplantation (HSCT). Univariate screening methods were used to identify parameters associated with increased risk of infection Results: Amongst 144 subjects, 52 (36.1%) received anti-CD19 CAR T-cells (CAR-T); 53 (36.8%) received anti-CD22 CAR-T, 26 (18.1%) received anti-BCMA CAR-T; and 13 (9%) received anti-GD2 CAR-T. The median age was 18 years (range: 4 to 66). Sixty-one (42.4%) had undergone at least one prior allogeneic hematopoietic stem cell transplant (HSCT) and 24 (16.7%) had at least one prior autologous HSCT. Sixty-eight (47.2%) had a history of a recent infection within 100 days of initiation of LD chemotherapy, of which 17 were a major chronic infection and 9 were considered severe. Fifty-eight (40.3%) subjects experienced a total of 103 infections from initiation of LD chemotherapy through day 30 post CAR T-cell infusion, with the median time to first infection being 6 days post-CAR infusion. Twenty-eight (19.4%) subjects had more than 1 infection; 20 (13.9%) subjects had C. difficile infection. The 103 infections consisted of 47 (45.6%) distinct episodes of focal bacterial infections (e.g., sinusitis, pneumonia, urinary tract infection), 25 (24.2%) episodes of bacteremia, 22 (21.4%) viral infections, and 9 (8.7%) invasive fungal infections. Fourteen infections (13.6%) occurred between initiation of LD chemotherapy and day 0 (prior to CAR T-cell infusion). Fever and neutropenia, without a source, was documented as a distinct entity in 85 patients (59.0%). Using univariate statistical screening methods, we identified 4 parameters that were individually associated with an increased risk of infection: increasing number of prior therapies (p=0.0034), prior history of recent infection within the past 100 days (p=0.0064), age > 18 (p =0.028), or enrollment on the CD22 CAR trial (p=0.0028). Eliminating the trial, since that is not generalizable, and combining these into a multivariable logistic regression model identified age > 18, prior history of infection and prior lines of therapies jointly associated with an increased risk of infection. Using these parameters, a predictive model was developed, which, when applied to the data set used to develop the model, can correctly predict 69 of 86 patients without an infection (80.3%; CI: 70.3-88.0%) as well as 32 of 58 patients with an infection (55.2%; 95% CI: 41.5- 68.3%). Conclusion: In this retrospective analysis, we provide the largest experience to date analyzing infection in the setting of CAR T-cell therapy across multiple CAR constructs. Our study demonstrates that adult age, prior history of infection, increased number of prior therapies and enrollment on a particular CAR T-cell trial, in this case the CD22 CAR-T trial, may lead to a higher risk of infection than in those without these risk factors. Further investigations are underway to evaluate clinical outcomes with infection which occur in the peri CAR T-cell setting and potential strategies to minimize infection risk. Generalizability of this model will be tested in an independent data set. Disclosures Lee: Kite, a Gilead Company: Research Funding; Harpoon Therapeutics: Consultancy; Juno Therapeutics: Other: External Advisory Board; ACI Clinical on behalf of Celgene:: Other: Independent Central Quality Review Committee. Mackall:Obsidian: Research Funding; Lyell: Consultancy, Equity Ownership, Other: Founder, Research Funding; Nektar: Other: Scientific Advisory Board; PACT: Other: Scientific Advisory Board; Bryologyx: Other: Scientific Advisory Board; Vor: Other: Scientific Advisory Board; Roche: Other: Scientific Advisory Board; Adaptimmune LLC: Other: Scientific Advisory Board; Glaxo-Smith-Kline: Other: Scientific Advisory Board; Allogene: Equity Ownership, Membership on an entity's Board of Directors or advisory committees; Apricity Health: Equity Ownership, Membership on an entity's Board of Directors or advisory committees; Unum Therapeutics: Equity Ownership, Membership on an entity's Board of Directors or advisory committees. Kochenderfer:Kite and Celgene: Research Funding; Bluebird and CRISPR Therapeutics: Other: received royalties on licensing of his inventions. OffLabel Disclosure: Conditioning chemotherapy for CAR T-cell therapy; this is a retrospective study that used different lymphodepletion regimens.

Abstract Background: Current prognostication in primary myelofibrosis (PMF) utilizes MIPSS70 (mutation-enhanced international prognostic scoring system for transplant-age patients), MIPSS70+ version 2.0 (karyotype-enhanced MIPSS70) and GIPSS (genetically-inspired prognostic scoring system, which is based on mutations and karyotype) (JCO 2018;36:310; JCO doi: 10.1200/JCO.2018.78.9867; Leukemia. 2018;doi:10.1038/s41375-018-0107). These contemporary models are inter-complimentary and highly efficient in predicting overall survival. However, specific risk factors for very long survival (e.g. 20+ years) and short-term mortality (i.e. death within 5 years of diagnosis) have not been systematically looked into. Methods: Study patients were recruited from the Mayo Clinic, Rochester, MN, USA. Diagnoses were according to the 2016 World Health Organization criteria (Blood. 2016;127:2391). Logistic regression statistics was used to identify predictors of 20-year survival and 5-year mortality. Variables evaluated in the logistic model included those that are currently listed in MIPSS70, MIPSS70+ version 2.0 and GIPSS, as well as age (≤70 vs >70 years) and sex. Receiver operating characteristic (ROC) curves and area under the curve (AUC) were used to estimate predictive accuracy. The JMP® Pro 13.0.0 software from SAS Institute, Cary, NC, USA, was used for all calculations. Results: A total of 1,282 consecutive patients with PMF (median age 65 years, range 19-92; 63% males) were used to identify 26 (2%) patients (median age 51 years, range 28-71; 38% males) who survived their disease for at least 20 years (very long-lived patients) and 626 (49%) patients (median age 68 years, range 27-92; 66% males) who died within 5 years of their diagnosis. Clinically-derived information, including the clinical variables used in MIPSS70 and MIPSS70+ version 2.0, with the exception of fibrosis grade (information available in 60%), was available in 99% of the study population; cytogenetic information was available in 94%, driver mutational status in 71% and the full profile of high molecular risk (HMR) mutations, including ASXL1, SRSF2, U2AF1 Q157, EZH2, IDH1 and IDH2, in 42%. MIPSS70+ version 2.0 risk distribution for the entire study population (n=1,282) was 4% very low risk, 16% low risk, 19% intermediate risk, 40% high risk and 21% very high risk; the corresponding percentages for the 26 very long-lived patients were 14%, 29%, 14%, 43% and 0% and for the 626 short-term survivors were 0%, 4%, 10%, 47% and 39% (p<0.001). Cytogenetic, driver mutation and HMR mutation information was available in 100%, 58% and 27% of the 26 long-term survivors, respectively, and 98%, 63% and 39% of the 626 short-term survivors. Multivariable logistic regression analysis of 20-year survival identified the following seven variables as being favorable: age ≤70 years (p=0.002); female sex (p=0.03); hemoglobin level ≥10 g/dl for women and ≥11 g/dl for men (p=0.03), leukocyte count ≤25 x 109/l (p=0.009), platelet count ≥100 x 109/l (p=0.002), circulating blasts ≤1% (p=0.03) and absence of constitutional symptoms (p=0.04). The particular phenotypic profile exhibited a high degree of predictive accuracy with an estimated AUC of 0.90 (Figure 1a); karyotype and mutations did not retain significance in this particular analysis; in fact, 5 (71%) of the 7 long-term survivors who were informative harbored at least one HMR mutation. A similar analysis for 5-year mortality identified the following risk factors: high molecular risk mutations (p<0.001); unfavorable or very high risk karyotype (p<0.001); absence of type 1/like CALR mutations (p<0.001); age >70 years (p<0.001); constitutional symptoms (p<0.001); hemoglobin level <10 g/dl for women and <11 g/dl for men (p<0.001); leukocyte count >25 x 109/l (p=0.004); and circulating blasts >1% (p=0.001); predictive accuracy was estimated at AUC 0.87 (Figure 1b). Conclusions: The phenotypic profile of long-lived patients in PMF includes young age (≤70 years old), female sex, asymptomatic disease (i.e. no constitutional symptoms) and absence of moderate to severe anemia, thrombocytopenia or leukocytosis and ≤1% circulating blasts; each one of these parameters, with the exception of female sex, were present in ≥80% of long-term survivors. Karyotype and mutations were more relevant in predicting short-term mortality. Disclosures No relevant conflicts of interest to declare.