ALBERT

Description: Background: In the AZA-001 trial, azacitidine (AZA) altered the natural history of patients with higher-risk MDS (Fenaux et al., Lancet 2009) by significantly improving overall survival compared to conventional care regimens (24.5 months versus 15.0 months). Once approved for drug reimbursement by the provincial health ministry, Cancer Care Ontario (CCO) mandated that all eligible patients be enrolled in a prospective registry to ensure compliance with eligibility criteria and schedules of drug administration. Baseline characteristics were recorded and treatment response were to be submitted every 6 cycles of treatment. Our objectives were to audit this clinical program for results after 6 years and identify the prognostic markers for survival in a homogenous higher-risk MDS/low blast count AML patient population. Methods: Only higher-risk MDS patients (intermediate-2, high) as defined by the International Prognostic Scoring System (IPSS) and low blast count AML (20-30% blasts) treated with AZA in Ontario, Canada from June 1, 2010 to March 2, 2016 were eligible and included. Our primary outcome was overall survival (OS) from date of first AZA treatment. Our secondary outcomes were overall response rates and the predictors of OS including administration schedules, centre size or type (regional cancer versus community). Univariate and multivariable Cox proportional hazard model were used to determine the predictors of OS. Hazard ratios and generalized R2 (higher the R2, stronger association with OS) were also calculated. Results: 825 higher-risk MDS and 276 low blast count AML were included (n = 1101 total). Median age was 74 years (range 19 to 99), 65.2% were male and the IPSS scores were intermediate-2 (64.3%) and high-risk (35.7%). Sixty-six percent of patients were transfusion dependent (TD) at time of AZA initiation and 15.5% had received previous chemotherapy. By dosing schedule, 24.7% received AZA for 7 consecutive days (7d), 12.4% for 6 consecutive days (6d) and 62.9% by the 5-2-2 schedule. Overall, the median number of cycles received was 6 (range 1 to 67) and 8 (range 6 to 14) when restricted to the 692 (63%) patients who received at least 4 cycles of treatment. Dose reductions were seen in 33.3% of patients (mean 11.1 mg/m2 in those with reductions) and were more common over time (negative slope 0.178; p 〈 .0001). Of those with repeat bone marrow (n = 293) best response was complete response (CR) in 16.7% and partial response (PR) in 10.6%. Of those without CR/PR/progressive disease on bone marrow (n = 814), 20.4% experienced hematologic improvement including those with marrow CR and marrow stable disease. The actuarial median survival was 11.6 months (95% CI 10.7- 12.4) with a significantly longer OS for MDS compared with low blast count AML (12.4 months vs. 9.6 months; p = .0002) and 16.7 months (95% CI 15.2-18.1) for those receiving at least 4 cycles. There was no difference in OS between the 3 dosing schedules (11.7 months (7d) vs. 10.2 months (6d) vs. 12.0 months (5-2-2); p = .87; figure 1A), regional cancer centre vs. community (11.3 months vs. 11.7 months; p = .19; figure 1B) or by centre volume (11.4 months 〈 50 patients vs. 11.6 months 〉 50 patients; p = .38; figure 1C). On univariate analysis, the following were predictive of OS: blast percentage, number of cytopenias, karyotype, IPSS, WHO classification, TD, greater transfusion burden and secondary MDS. The multivariate model with the highest R2 (13.5%) included blast percentage (p 〈 .0001), karyotype (p 〈 .0001), cytopenias (p = .038), TD (p 〈 .0001) and secondary MDS (p = .0006) as summarized in the table below. Conclusions: In our large real world evaluation of AZA use in higher-risk MDS/low blast count AML, we validated the expected overall response rates to AZA but demonstrated a lower than expected OS compared to the AZA-001 trial. Reassuringly, survival did not differ by dosing schedules, centre volumes or center type (regional cancer vs. community). OS was higher in the 2/3 of patients who received at least 4 cycles of treatment, reinforcing the necessity of sustained administration until therapeutic benefits are realized. This represents the largest real world evaluation of AZA in higher-risk MDS/low blast count AML and additional analyses are underway. Table Table. Figure 1 Figure 1. Disclosures Mozessohn: Celgene: Honoraria. Buckstein:Celgene: Honoraria, Research Funding; Novartis: Honoraria.

Description: Introduction The use of surveillance CT imaging in patients with DLBCL in remission is neither effective to detect recurrence nor cost-effective. The ASH Choosing Wisely (CW) campaign, in particular, emphasizes the lack of benefit in imaging beyond 2-years of completion of therapy. We sought to describe the practice of surveillance imaging and predictors of this practice. Methods We used population-based health system administrative databases from Ontario, Canada. We studied a cohort of all adult patients ≥18 with diffuse large B-cell lymphoma who received R-CHOP therapy for curative intent between January 1, 2004 to June 30, 2011. Based on the CW campaign, we defined an index date of 2-years after the last dose of R-CHOP as the time-frame beyond which surveillance CT imaging would be inappropriate. The cumulative incidence of receiving CT scans within 3 years after the index date (i.e. from 2- to 5-years beyond the end of treatment) represented the primary outcome of interest (established within the Ontario Health Insurance Program database). To ensure that only surveillance scans in asymptomatic patients were captured, patients were censored 6 months prior to development of recurrent disease or a new cancer diagnosis, further chemotherapy/radiation, or if they died (censored at time of death). Predictor characteristics included baseline comorbidities (John Hopkins weighted Aggregated Diagnosis Groups (ADG) comorbidity score) and income quintile (linkage of the patient postal code to Statistics Canada Census data on average household income by postal code). Results The cohort consisted of 2,838 patients treated with R-CHOP during the study period. Median age at time of first R-CHOP dosing was 63 years (IQR 52-72) and the median number of cycles received was 6 (IQR 6-8). The cumulative incidence of receiving CT imaging from the index date (2-years from end of treatment) to 3-years beyond the end of treatment was 40.1% (95% CI 38.3%-41.9%). The cumulative incidence of imaging from the index date to 5-years beyond the end of treatment was 55.6 % (95% CI 53.7%-57.5%). During the follow-up period, patients ≥65 were more likely to receive imaging than those aged

Description: Background: The past 30 years has heralded significant improvements in the treatment and outcomes of patients with advanced-stage Hodgkin lymphoma, with the introduction of ABVD (doxorubicin, bleomycin, vinblastine, and dacarbazine) and various combinations of BEACOPPbaseline and BEACOPPescalated (bleomycin,etoposide, doxorubicin, cyclophosphamide, vincristine,procarbazine, and prednisone). Initially, BEACOPP-containing regimens demonstrated superior progression-free survival (PFS) and overall survival (OS) compared to ABVD, despite higher rates of hematologic toxicity, infections, and infertility. However, the superiority of this regimen was tempered by concerns of increased toxicity and long-term complications including secondary malignancy and infertility. Furthermore, quality-adjusted measures and patient preferences have not been factored into analyses of the evidence. We performed a decision analysis to explore the trade-off that occurs when initial improvements in progression-free survival and overall survival are balanced with increased morbidity and mortality associated with infections, secondary malignancies, and infertility in the BEACOPP-containing strategy. Methods: We developed a Markov decision-analytic model to compare ABVD versus BEACOPPbaseline and BEACOPPescalated (hereinafter referred to as BEACOPP) for a hypothetical cohort of transplant-eligible patients withnewly-diagnosed, advanced-stage Hodgkin lymphoma. The model simulates the clinical course of patients over a 20-year time horizon, with the end-points of life expectancy and quality-adjusted life expectancy. The baseline probabilities used in the model were derived from a systematic review of published randomized controlled trials. Key variables included response, relapse and survival rates comparing ABVD versus BEACOPP, risk of developing complications such as infection, infertility, or secondary malignancy with each strategy, and the estimated survival once secondary malignancy develops. We also incorporated therapies for relapsed disease including autologous stem cell transplantation, and post-transplant strategies, based on available data. Efficacy was discounted by 3%. The model incorporated data on health state utilities, which were derived from a review of the literature. Sensitivity analyses were performed for key variables. Results: Based on a 20-year model with 100,000 trials, life expectancy was 10.9 years with ABVD and 12.3 years with BEACOPP, resulting in a net benefit of 1.4 years for the BEACOPP strategy. The quality-adjusted life expectancies for the two strategies, respectively, were 9.1 and 10.5 years, with an expected benefit of 1.4 QALYs with BEACOPP. Sensitivity analyses demonstrated that the model was robust to the key variables of probability of death from secondary malignancy, probability of relapse, and probability of infertility secondary to BEACOPP. A range of relapse probabilities post-ABVD and BEACOPP were tested in sensitivity analyses, and BEACOPP was consistently superior, with the lowest difference between QALYs found to be 0.5 QALYs. In sensitivity analysis of treatment-related mortality secondary to BEACOPP, the threshold value was found to be 8% mortality over the6 monthtreatment period, a value much greater than that reported in the literature (see Figure 1). The threshold utility of infertility was found to be 0.60 in sensitivity analysis (see Figure 2), a value lower than the utility derived from a systemic review of the literature (0.87). Onmicrosimulation(100,000 trials), 88% of the simulations showed that BEACOPP was the preferred strategy compared to ABVD. Conclusion: The preferred treatment strategy for patients with newly diagnosed advanced-stage Hodgkin lymphoma is a combination BEACOPP regimen. This strategy maximizes life expectancy and quality-adjusted life years, accounting for the increased rates of hematologic toxicity, secondary malignancy, and infertility in patients receiving the BEACOPP strategy. The model was robust to sensitivity analyses of key variables tested through plausible ranges obtained from the published literature. Disclosures Buckstein: Novartis: Honoraria; Celgene: Honoraria, Research Funding. Prica:Celgene: Honoraria; Janssen: Honoraria.

Description: Introduction The cost of cancer care is rising to unsustainable levels, predominantly driven by an increase in expenditures for novel therapies. In the era of biologic therapies, these excessive costs are disproportionately borne by patients with hematologic malignancies. Although the American Society of Clinical Oncology (ASCO) and European Society of Medical Oncology (ESMO) have both developed frameworks to determine the relative value associated with new solid tumor therapies (and specifically with ESMO, to the exclusion of therapies in hematology), it is unclear if they can be applied in the assessment of value of treatments for blood cancers. Methods We evaluated the value of new therapies for hematologic malignancies using the ASCO (version 1 from August 10, 2015 and version 2 from May 31, 2016) and ESMO (version 1 or v1) frameworks. All US Food and Drug Administration, European Medicinal Agency, or Health Canada approved parenteral therapies for hematologic malignancies from 2006-2015 were identified. A systematic review of randomized controlled trials (RCTs) for these therapies was conducted. Two reviewers independently scored the trials using the ASCO value framework v1 (range of scores -20 to 130), v2 (lower range undefined to 180) and ESMO Magnitude of Clinical Benefit Scale (range 1-5). Disagreements were descriptively presented and resolved by consensus. The concurrent validity between the ASCO and ESMO scores was measured by the Spearman correlation coefficient. Results Twenty-three RCTs in malignant hematology were identified and scored. Seven of 23 studies reported primary outcomes unique to hematologic malignancies (for example, time-to-progression in myeloma, cytogenetic response in chronic myeloid leukemia, and symptomatic response in multicentric Castleman's), other than the main outcomes used to derive ASCO/ESMO scores (overall and progression-free survival). The median ASCO v1 score for the trials was 24 (IQR 22-40, min 6 and max 53). The median ASCO v2 score was 26.7 (IQR 17.4-37.6, min -33.3 and max 116.3). The median ESMO score was 2 (IQR 2-3, min 1 and max 4). Using the ASCO v1 framework, 10 studies resulted in disagreements in scoring, predominantly due to variable interpretations of the scoring system. Two studies could not be scored. One study did not report toxicity grades and another study of maintenance rituximab in lymphoma did not report on conventional oncology outcomes that could fit into the ASCO model. Using the ASCO v2 framework, 14 studies resulted in disagreements, predominantly due to differences in scoring toxicities. With the ESMO Scale, 12 studies resulted in disagreements, most occurring due to variable interpretations in scoring survival or progression-free survival outcomes when median values were not provided in the study publication. Two studies could not be scored with the ESMO Scale. One study did not report hazard ratios and another study's reported outcomes did not fit the ESMO scoring options. The correlation coefficient between ASCO v1 and ESMO scores was 0.10 (95% CI: -0.37 to 0.53), suggesting that the correlation was not significantly different from chance only. The coefficient between ASCO v2 and ESMO was -0.21 (95% CI: -0.59 to 0.24) and between ASCO v1 and ASCO v2 was 0.30 (95% CI: -0.15 to 0.65). Conclusions Current value frameworks are challenging to apply to therapies for hematologic malignancies. When studies could be scored, the correlations between ASCO (v1 and v2) and ESMO results were poor, suggesting that these frameworks may not reliably identify the value of therapies in hematology. Consideration for the unique outcomes and toxicities in this population is warranted. The ASCO and ESMO frameworks continue to evolve, and a partnership with societies representing hematologists and their patients would be fruitful. Disclosures No relevant conflicts of interest to declare.

Description: Background: High dose chemotherapy followed by autologous stem cell transplant (ASCT) is the standard curative option for patients with relapsed or refractory, chemosensitive, aggressive non-Hodgkin lymphoma (NHL). The optimal timing for ASCT following salvage chemotherapy is not known. Cancer Care Ontario (CCO)-the cancer agency for Ontario, Canada's largest province-treatment guidelines recommend that no more than 91 days should elapse from the first day of salvage chemotherapy to stem cell transplant. We evaluated the impact of time to stem cell transplant in the context of the international CCTG LY.12 phase 3 clinical trial. Methods: Patients with relapsed or refractory (R/R) aggressive NHL were randomly assigned to gemcitabine, cisplatin and dexamethasone (GDP) or dexamethasone, cytarabine, cisplatin (DHAP), with or without rituximab, followed by ASCT [Crump JCO 2014]. Time interval definitions were based on CCO guidelines: Total Wait Time (TWT) as the number of days from the first day of salvage chemotherapy to day of ASCT; Apheresis Wait Time (AWT) as the number of days from the first day of salvage to the first day of stem cell collection; Stem cell transplant Wait Time (SWT) as the number of days from the last day of stem cell collection to the day of ASCT. Patients were considered to have experienced a delay in TWT, AWT or SWT if the time intervals exceeded 91, 70 and 21 days respectively. Overall survival (OS) and event-free survival (EFS) from transplant date were compared between patients who met and exceeded TWT targets using a Cox proportional hazards model. A linear regression model was applied to analyze TWT as a continuous variable. Univariate and multivariate analyses were performed to estimate the adjusted hazard ratio (HR) for TWT for the following co-variables: age ≤60, performance status 0/1, disease stage (I/II), presence of ≤1 extranodal sites, and response after cycle 2 (complete response, CR; complete response, unconfirmed, CRu; partial response, PR). Results: Of 619 patients enrolled on LY.12, 307 (47%) had sufficient response to salvage chemotherapy and adequate stem cell collection to complete ASCT on protocol. Among these, median age was 54.6 years, 64% were male and 94% had a performance status of 0 or 1. International Prognostic Index (IPI) score at relapse was 0-1 in 45%, 2 in 31% and ≥3 in 24%. The majority of patients had poor risk disease at study entry; 58% had a best response of stable disease (SD) or progressive disease (PD) to primary therapy, or initial duration of response 〈 1 year. Following up to 2 cycles of salvage chemotherapy, 75/307 (24%) achieved CR/CRu, 142/307 (46%) achieved PR, 89/307 (29%) had SD. One patient had missing data. The median TWT for the total transplanted population was 91 days (range 50-217). Median AWT and SWT were 63 (range 0-151) and 26 (range 6-146) days, respectively. Fifty percent of patients exceeded TWT target of 91 days; 32% and 57% of patients exceeded AWT and SWT targets. There was no difference in median OS (HR 0.96, 95% CI 0.66-1.39, p=0.81) or EFS (HR 1.13, 95% CI 0.82-1.55, p=0.46) between patients who exceeded and met TWT targets. The 4-year OS for patients who met and exceeded TWT was 62% and 64%, respectively. The 4-year EFS for patients who met and exceeded TWT was 43% and 50%, respectively. When analyzed as a continuous variable, TWT did not affect OS (HR 0.99) or EFS (HR 0.99). Comparison of the quartiles with shortest and longest TWT demonstrated HR 0.72 (95% CI 0.42-1.26, p=0.25) for overall survival and 0.69 (95% CI 0.44-1.09, p=0.11) for EFS. Comparison of the 10th and 90th percentiles for TWT demonstrated HR 0.67 (95% CI 0.28-1.59, p=0.36) for overall survival and 0.71 (95% CI 0.35-1.44, p=0.34) for EFS. Only the presence of ≤1 extranodal sites of disease was found to be predictive of OS in the transplanted population on univariate and multivariate analysis (adjusted HR 0.51, p=0.005). The median TWT was longer for the 31 patients transplanted at Italian centers, compared to 266 transplanted at Canadian centers (median TWT 90 vs. 118 days, t 〈 0.0001). Conclusion: In this exploratory analysis, limited to patients who completed transplant on the LY.12 clinical trial, we did not find evidence that those meeting current CCO ASCT wait time targets had superior outcomes compared with those who did not. Table. Table. Figure 1. Figure 1. Figure 2. Figure 2. Disclosures Kuruvilla: BMS: Consultancy, Honoraria; Celgene: Consultancy, Honoraria; Amgen: Honoraria; Abbvie: Consultancy, Honoraria; Janssen: Consultancy, Honoraria; Gilead: Consultancy, Honoraria; Merck: Honoraria; Roche Canada: Consultancy, Honoraria, Research Funding; Seattle Genetics: Consultancy, Honoraria; Lundbeck: Honoraria. Luminari:Roche: Membership on an entity's Board of Directors or advisory committees; Celgene: Membership on an entity's Board of Directors or advisory committees, Other: Travel, Accomodations, Expenses; Takeda: Other: Travel, Accomodations, Expenses; Teva Pharmaceuticals: Membership on an entity's Board of Directors or advisory committees; Pfizer: Honoraria. Hay:Amgen: Research Funding; Novartis: Research Funding; Janssen: Research Funding; Kite Pharmaceuticals: Research Funding.

Description: Background: Although the natural history of follicular lymphoma is indolent with a median overall survival of about 12-15 years, the disease is heterogeneous. The 5 and 10 year overall survival (OS) of low, intermediate and high risk FLIPI is 91%, 78% and 53% and 71%, 51% and 36% using standard rituximab-based treatment. 5-year progression-free survival (PFS) is 80%, 70% and 48% respectively. Methods: Based upon this we conducted an investigator-initiated single-centre Phase II trial of intensified therapy with CHOP-R followed by (90)Y ibritumomab tiuxetan consolidation and 24 months of rituximab maintenance as treatment for patients with intermediate and high-risk newly diagnosed symptomatic follicular lymphoma. 33 patients were enrolled. Results: The addition of (90)Y ibritumomab tiuxetan was well tolerated but resulted in asymptomatic grade 3 or 4 thrombocytopenia and neutropenia in11-36% and 10-24% of patients between weeks 2-8 post (90) Y. After 9 years of follow-up (median follow-up 61 months) the 0S for intermediate and high risk FLIPI was 95% and 78%. The 5 year PFS was 79% and 64% for intermediate and high risk FLIPI, respectively. Responses at three months post consolidation were as follows: 3/33 (9%) achieved CR, 25/33(76%), achieved CRU, 1/33(3%) had PR, and 1/33(3%) had PD. Three patients did not receive (90)Y ibritumomab tiuxetan due to disease progression 2/33(6%), or death 1/33(3%). Of 19 patients who had a molecular marker for their lymphoma, 18 (95%) achieved molecular remissions in peripheral blood with CHOP-R therapy. Nine (47%) of these patients have been recently assessed for MRD and remain in molecular remission. The therapy resulted in decreased levels of IgG, IgM and IgA below the lower normal level in 33%, 40% and 23% of patients respectively post therapy. These levels did not recover in most of these patients. B cells were depleted to undetectable levels during therapy including rituximab maintenance. In 18 evaluable patients only 11 recovered normal B cell counts post maintenance rituximab. There was no correlation between normal B-cell recovery and Ig levels. Many patients with low or no B cell counts had normal IgG levels, whereas some patients who regained normal B cell counts were still unable to reach normal Ig levels. No patient developed human anti-mouse antibody. Immunity to measles, mumps, or rubella was retained post therapy. Patients did not have significant infections or opportunistic infections (although 2 developed Grade 1 shingles post (90)Y ibritumomab tiuxetan) and none required IVIG. Conclusions: We conclude that this intensified regimen is highly active in cyto-reducing lymphoma in high and intermediate risk FLIPI follicular lymphoma patients. The toxicity is tolerable although a significant percentage of patients will end up with persistent asymptomatic reductions in B cells and serum Ig. Only randomized trials will determine whether this regimen enhances outcome over standard of care in this higher risk follicular lymphoma population. References: 1.Examination of the follicular lymphoma international prognostic index (FLIPI) in the National LymphoCare study (NLCS): a prospective US patient cohort treated predominantly in community practices. Nooka AK, Nabhan C, Zhou X, Taylor MD, Byrtek M, Miller TP, Friedberg JW, Zelenetz AD, Link BK, Cerhan JR, Dillon H, Sinha R, Shenoy PJ, Levy D, Dawson K, Hirata JH, Flowers CR. Ann Oncol. 2013 Feb;24(2):441-8. doi: 10.1093/annonc/mds429. Epub 2012 Oct 5 2.Validation, revision and extension of the Follicular Lymphoma International Prognostic Index (FLIPI) in a population-based setting. van de Schans SA, Steyerberg EW, Nijziel MR, Creemers GJ, Janssen-Heijnen ML, van Spronsen DJ. Ann Oncol. 2009 Oct;20(10):1697-702. doi: 10.1093/annonc/mdp053. Epub 2009 Jun 23. PMID: 19549712 Disclosures Buckstein: Novartis: Honoraria; Celgene: Honoraria, Research Funding.

Description: Research groups are increasingly utilizing value frameworks, but little is known of their reliability. To assess framework concordance and interrater reliability between two major value frameworks currently in use, we identified all previously published datasets containing both scores from the American Society of Clinical Oncology Value Framework (ASCO-VF) and grades from the European Society for Medical Oncology-Magnitude of Clinical Benefit Scale (ESMO-MCBS). The intraclass correlation coefficient (ICC) was used to assess interrater reliability. Four eligible studies contained drugs evaluated by both value frameworks, resulting in a dataset of 39 grades/scores for discrete drug indications. ICC was 0.82 (95% confidence interval = 0.70 to 0.90) for ASCO-VF and 0.88 (95% confidence interval = 0.80 to 0.93) for ESMO-MCBS. Absolute concordance was found to be 5% for ASCO-VF and 44% for ESMO-MCBS, increasing to 74% and 80% when deviations within 20 points and 1 grade were considered, respectively. Interrater reliability of ASCO-VF and ESMO-MCBS is, therefore, near perfect, while absolute concordance is poor. This has implications when considering framework outputs in drug funding or treatment decision making.

Description: Introduction:The prognosis of patients with hematologic malignancies (HM) admitted to intensive care units (ICU) is historically poor due to complications of treatment and disease progression with reported overall mortality rates of 24.3% to 84.1%. There is little known regarding predictive variables for ICU admission in adult patients with HM. Objective: Our primary objective was to audit the clinical outcomes including mortality of our HM patients admitted to any hospital ICU unit and compare their characteristics and outcomes with non-ICU hospitalized patients. A secondary objective was to identify the predictive factors for ICU admission and survival. Methods: In this single centre retrospective study, we audited 656/2141 consecutive patients with HM who were admitted to our hospital from 2009-2015 and compared the disease, patient characteristics, and clinical outcomes of HM patients who did or did not get admitted to any ICU. We excluded patients admitted for palliative care and in instances where the patient was admitted more than once, we included only the most recent admission. We enriched for patients who were admitted to the ICU over this 6-year period to improve the statistical power of comparing these two populations. The variables considered included: reason for admission, underlying diagnosis, the modified Charlson Comorbidity Index (CCI), selected laboratory parameters at admission to hospital, previous chemotherapy type, line of treatment, timing and intent, body mass index (BMI), GSCF use, age and sex. We also screened for electronic documentation of advanced directives preceding ICU or hospital admission. We compared the characteristics of ICU (n=179, 27%) and non ICU (n=477, 73%) admitted patients using the Fisher exact test and Wilcoxon rank-sum non parametric test for categorical and continuous variables, respectively. To search for the significant predictive factors for ICU admissions and mortality, univariate and multivariate logistic regression analyses were used. Results: Over the 6-year period, the admission rate to any ICU for HM patients was 9.4%. Median age of the 656 patients was 65.0 (IQR 55-74) and 57% were male with a median time from last chemotherapy and diagnosis of 1 month (IQR 0-3) and 7.8 months (IQR 2.3-44), respectively. Selected patient and disease characteristics comparing ICU and non- ICU admitted patients are in Table 1. There were no differences in median age, BMI, gender, time from last chemotherapy, reasons for admission, WBC, and line of chemotherapy. Patients admitted to any ICU were more likely to have received chemotherapy with curative intent (p=0.0323) and have myeloid cancers (p= 0.0008). They had shorter times from diagnosis, lower hemoglobin, platelet and albumin levels but higher creatinine levels, lactate dehydrogenase (LDH) and liver enzymes. ICU admitted patients had higher CCI scores and in particular history of cardio/cerebro-vascular disease. They also had lower rates of advanced directives (32% vs 49%, p=0.0001). 50% of ICU patients were mechanically ventilated and 25% received vasopressors. Median length of hospital stay was 19 days for ICU and 6 days for non-ICU patients and median ICU length of stay was 3 days (IQR 2-6 days). 34 patients (19.0%) died in ICU and but in hospital mortality was 33% (compared with 8% for non ICU). In the multivariate analysis, six covariates significantly related to ICU admission: chemotherapy intent (curative vs. palliative), history of myocardial infarction, lower platelets, creatinine and albumin levels and no advanced directives (Table 2). Median overall survival was 4.7 months vs. not yet reached for ICU and non ICU admitted patients respectively. 30-day and 6-month survivals were 67% and 47% (ICU) compared with 89%, and 77% in non-ICU admitted patients. The predictors of death in ICU admitted patients were male sex (OR 2.6, 95% CI 1-6.7, p =0.03) and mechanical ventilation (OR 6.55, 95% CI 2.6-18.5, p=0.0001). Conclusions:We validate the previously reported high rates of ICU mortality for patients with hematologic malignancies and have identified the risk factors for ICU admission. Patients without advanced directives have a high likelihood for admission to ICU and potentially represent a targetable group for interventions to avoid inappropriately aggressive care. A predictive score is in development to help identify the patients warranting closer scrutiny and goals of care discussions. Disclosures Buckstein: Celgene: Honoraria, Research Funding; Novartis: Honoraria.