•Identify three factors associated with increased platelet transfusion at the end of life.•Describe the relationship between number of visits by a palliative care consultant and the number of platelet transfusions received at end of life. For cancer patients near the end of life, transfusions decisions may be challenging because risks may sometimes outweigh benefits. To understand non-laboratory factors that are associated with platelet transfusions near time of death. Medical oncology inpatients dying at our cancer hospital between October 1, 2013 and September 30, 2015 were identified through an institutional database. Our primary outcome variable was the number of platelet units transfused within seven days before death. Gender, age, race, religion, state/country of origin, cancer diagnosis, oncologic treatment(surgery/chemotherapy/radiotherapy), intensive care unit(ICU) admission, resuscitation status, and Palliative Care consultation were collected as potential predictors. Associations between the predictors and number of platelet units received were evaluated using zero-hurdle negative binomial models. Of 1,248 decedents, 52% were male and most were Caucasian (73%), Christian (58%), local residents (92% from NY/NJ/CT). Palliative Care saw 49% of patients within seven days before death. Twenty-one percent received 1-6 platelet units and 9% received seven or more units (range 7-35) within seven days before death. In the multivariable model, leukemia, lymphoma, and myeloma diagnoses predicted both receipt of platelet transfusion and higher number of units received. Female gender (OR=1.7) and chemotherapy receipt (OR1.9) increased odds of receiving platelet transfusion as did younger age (OR=0.8 per 10-year increase in age). ICU admission predicted both receipt of platelet transfusion (OR=3.6) and higher number of units received (OR1.3). Compared to patients not seen by a Palliative Care consultant, patients with 3-7 visits (OR=0.5) and eight or more visits (OR=0.4) were significantly less likely to receive a transfusion. Demographic (e.g. gender, age) and health-related (hematologic malignancy, ICU admission) factors are associated with increased platelet transfusions for cancer inpatients approaching death. Palliative Care consultation may lower likelihood of platelet transfusion though prospective research is needed to demonstrate causality.

•Explain to all patients the nature and possibility of end-of-dose failure to patients when initiating therapy with fentanyl transdermal preparations.•Increase screening of patients on using transdermal fentanyl preparations for end-of-dose failure, with particular attention to patients with cancer cachexia. Transdermal fentanyl is widely used for patients with cancer pain requiring a long-acting opioid analgesic. The delivery of fentanyl in this form relies on a drug reservoir forming in subcutaneous fat. The manufacturer-recommended initial dosing frequency is every 72 hours. However, some patients experience “end-of-dose failure” (EDF) with worsening pain after a shorter interval. No studies have examined whether EDF correlates with body mass index (BMI). To assess whether low BMI is associated with prescription of transdermal fentanyl at q48h dosing frequency. All outpatient fentanyl patch prescriptions dispensed between January 1, 2014 and December 31, 2015 at our academic cancer center were extracted using an institutional database. For each prescription patient age, sex, race, patch dose, dosing frequency, and cancer type, as well as height and weight (used to calculate BMI) were collected. A generalized estimating equations (GEE) model was fit to account for correlation within patient and determine association between BMI and patch frequency. A total of 2,750 patients filled 6,569 fentanyl patch prescriptions over the study period. Patients were a median of 60 years old, 56.1% female, and 73.3% white (11.2% African American, 7.2% Asian). The median number of prescriptions per patient was 1 at a median dose of 37. Among 287 underweight patients (BMI<18.5), 34 (11.8%) received at least one q48h fentanyl patch prescription compared with 62 (10.2%) of obese (BMI>30) patients. Patients who received q48h patch had lower BMI (.58 unit decrease; p<0.01) than patients who received the q72h patch. Lower BMI is associated with q48h fentanyl patch dosing, suggesting that it may also be associated with end-of-dose failure in cancer patients. These patients should be closely monitored and alerted to this possibility.

•Articulate the role of Ketamine in the management of cancer-related pain in an opioid-tolerant patient.•Evaluate the effectiveness and safety of Ketamine as a co-analgesic in treating cancer-related pain. Ketamine can be used as an analgesic for severe cancer pain when opioid side effects preclude dose escalation and other analgesics are not effective or are contraindicated. The effectiveness and tolerability of Ketamine in cancer patients has not been consistently demonstrated in the literature. Therefore, we conducted a retrospective review of the use of Ketamine infusions at our institution over the past ten years to identify its effects on opioid requirements, pain control, and reported side effects. From 2004 to 2014, Ketamine was used in the inpatient population in 201 cases. Of these, a total of 157 patients had admissions where they received intravenous (IV) Ketamine infusions at sub-anesthetic doses as an adjunct to opioid pain management for acute and chronic cancer-related pain. Demographic information, clinical variables, and patient outcomes were abstracted from the records. The median starting dose of Ketamine was 2mg/hr (range 1-4mg/hr) and the median ending dose was 12mg/hr (range 2-24mg/hr). Patients were treated with ketamine for a median of seven days (range 1-18 days). The median opioid dose prior to initiating Ketamine was 532mg Morphine equivalent (IVME) (range 45-9420mg IVME) and after completion was 288mg IVME (range 10-3699mg IVME). The median numeric pain scores before and after Ketamine infusion were not significantly different (7 vs 6). Ketamine was largely well tolerated, and the most commonly reported side effect was hallucinations. Patients who received Ketamine in our cancer center had severe pain despite receiving high dose opioids (22 mg IVME/h). In this population of patients with intractable cancer pain, the addition of Ketamine was associated with 46% opioid dose reduction with some improvement in analgesia.

•Identify 3 factors that predict increased RBC transfusion at end of life in cancer patients.•Identify 3 factors that do not predict RBC transfusion at end of life in cancer patients. In many situations, transfusions clearly improve longevity and/or quality of life (QOL). In terminal cancer patients the risk-benefit balance may sometimes be unclear. We undertook this retrospective study to understand which non-laboratory factors may predict red blood cell (RBC) transfusion in the last 7 days of life. Between 10/1/2013 and 9/30/2015, 1248 in-patients with cancer died on medical oncology services at our institution. The primary endpoint was number of RBC units received within 7 days before death. Predictors included patient sex, age, race, religion, state/country of origin, cancer diagnosis, intensive care unit (ICU) death, length of stay (LOS), Do Not Resuscitate status, chaplaincy visits, and palliative medicine consultation. Associations between the predictors and number of RBC units received were evaluated using zero-hurdle negative binomial models to account for large numbers of patients who received zero RBC units. Patients were mostly Caucasian (73%), Christian (58%), local residents (92% from NY/NJ/CT), and male (52%). Just under half (49%) were seen by palliative medicine within 7 days before death. Thirty-eight percent received 1-6 RBC units and 3% received 7 or more units (range 7-23) within 7 days before death. In the multivariable model, patients with leukemia, lymphoma, myeloma, and genitourinary cancers were significantly more likely to receive RBC transfusion. ICU death predicted both receipt of RBC transfusion (OR=4.7) and higher number of units received (RR=1.5). Female sex was associated with receiving transfusion (OR=1.6). In unadjusted analysis, patients who had a palliative medicine consultation were less to likely to receive a transfusion (OR=0.72) and tended to receive fewer units (RR=0.72), but this was not significant in multivariable analysis. Sex, cancer type, and ICU death seem to influence end-of-life (EOL) transfusions whereas race, religion, and state/country of origin do not. Palliative consultation may reduce RBC transfusion at EOL for some patients but more research is needed to identify subgroups most impacted.

Background: Data on the outcomes of intensive care unit (ICU) admissions for patients with advanced incurable chemoresistant solid tumor malignancies, and the benefits of subsequent/post-ICU anticancer treatments are limited but have end-of-life and ethical implications. Methods: An institutional database was queried to identify patients of the gastrointestinal (GI) medical oncology service of Memorial Sloan Kettering Cancer Center with ≥1 ICU admission during 2014. Records were reviewed for evidence of cancer control from cancer treatment after the ICU admission. Results: Twenty-eight patients who had progressed beyond at least first-line chemotherapy for metastatic GI adenocarcinoma were admitted to the ICU for sequelae of progressive clinical deterioration. The most frequent reasons for ICU admission were sepsis (39%) and acute respiratory failure (29%). Ten patients died in the ICU, 3 died during the same hospitalization after ICU discharge, and 15 were discharged from the hospital. Of these 15, the median survival from hospital discharge was 2.2 months and 6 received further chemotherapy but with no evidence of clinical benefit. Of these 6, 3 lived over 5 months but the treatment of 5 entailed recycling of previously ineffective chemotherapy agents (3) or those originally used in the adjuvant setting (2). Two of these patients received liver-directed therapy without benefit. Conclusions: Admissions to the ICU in this cancer population were associated with high morbidity and mortality and did not result in benefit from subsequent cancer treatment. These data can be used to help establish realistic expectations and care goals in previously treated patients having metastatic GI cancer with clinical deterioration.

e18185 Background: Naloxone is sometimes required to reverse acute intoxication from opioid pain medications. Although clinical guidelines exist, naloxone dosing is variable in practice. Higher doses can cause pain crisis, acute withdrawal, and possibly increased patient risk. We developed a multi-faceted intervention to increase guideline-concordant naloxone dosing. Methods: We surveyed medical staff and retrospectively collected 6 months of naloxone prescribing data to assess knowledge and frequency of guideline-concordant dosing. We reviewed the electronic order for naloxone and interviewed experienced nurses to facilitate mapping of the process of naloxone administration. We then designed 4 interventions aimed at adult inpatients on medical oncology, general medical and ICU services: 1. Simplify wording of naloxone guidelines; 2. Create naloxone kit containing all necessary equipment and instructions; 3. Auto-populate dose into electronic naloxone order; 4. Administer 5-minute educational inservice to staff (RN, NP, PA, MD). To assess rate of guideline concordance, naloxone prescribing data was collected for the pre-intervention (12 months), intervention (6 weeks), and post-intervention (8.5 months) groups. The survey was repeated after the intervention. For analysis, a generalized linear regression with logit link function and a generalized estimating equation method were used. Results: Over the 22-month study period, 137 patients received 148 naloxone doses. Patients were mean age 63.1 y/o, 54.7% female, and 75.9% white. Of the 90 naloxone doses given pre-intervention, only 24.4% were guideline-concordant. The 10 doses of naloxone given during the intervention period were guideline-concordant in 40% of the cases. Of the 48 naloxone doses given post-intervention, 62.5% were guideline-concordant, a significant increase compared to the pre-intervention group (p < 0.001). The improvement was most dramatic in the non-ICU services (26.4% to 73.5%, p < 0.001). Increases were maintained at 8 months (median concordant rate 53.5%/month (range: 33%-100%). Conclusions: A simple, inexpensive intervention can be used to increase guideline-concordant naloxone dosing and potentially avoid adverse effects of naloxone.

36 Background: A major limitation of prognostic tools such as the Eastern Cooperative Oncology Group (ECOG), Karnofsky, and Palliative Performance Scale is a reliance on subjective clinical assessment. An objective tool, the Glasgow Prognostic Score (GPS) is derived from C-reactive Protein (CRP) and albumin and has been validated in patients with operable and inoperable malignancies. One disadvantage of this tool is that CRP is not routinely measured in the United States. We examined if the Neutrophil-Lymphocyte Ratios (NLR) (Ahn, H.K., et al., Neutrophil-Lymphocyte Ratio Predicts Survival in Terminal Cancer Patients. J Palliat Med, 2016) could be substituted for CRP in the GPS to predict survival in patients with advanced pancreatic adenocarcinoma. METHODS A retrospective chart review identified patients at MSKCC with pathology-confirmed stage IV pancreatic adenocarcinoma diagnosed between 2011 to 2014. Pre-treatment absolute neutrophil count, absolute lymphocyte count, and albumin were extracted. The NLR for each patient was calculated and assigned: NLR ≤ 4 g/dl = 0, NLR > 4 g/dl = 1; serum albumin > 4 g/dl = 0, and serum albumin < 4 g/dl = 1. Combining NLR and albumin scores resulted in a composite MPS score of 0-2, similar to GPS. We evaluated the association of the MPS with overall survival. RESULTS N = 833 patients were identified with median survivals summarized in the table below. A log-rank test showed statistically significant differences in survival between MPS groups (p<0.00005). The MPS on univariate analysis had a HR of 1.36 (95% CI 1.23 - 1.50, p<0.0005) associated with overall survival. CONCLUSIONS The MPS, a novel composite of NLR and albumin, is an objective prognostic tool that divided this sample of patients into three clinically distinct subgroups. Further interrogation will control for performance status, disease characteristics and anti-cancer therapy. [Table: see text].

36 Background: A major limitation of prognostic tools such as the Eastern Cooperative Oncology Group (ECOG), Karnofsky, and Palliative Performance Scale is a reliance on subjective clinical assessment. An objective tool, the Glasgow Prognostic Score (GPS) is derived from C-reactive Protein (CRP) and albumin and has been validated in patients with operable and inoperable malignancies. One disadvantage of this tool is that CRP is not routinely measured in the United States. We examined if the Neutrophil-Lymphocyte Ratios (NLR) (Ahn, H.K., et al., Neutrophil-Lymphocyte Ratio Predicts Survival in Terminal Cancer Patients. J Palliat Med, 2016) could be substituted for CRP in the GPS to predict survival in patients with advanced pancreatic adenocarcinoma. Methods: A retrospective chart review identified patients at MSKCC with pathology-confirmed stage IV pancreatic adenocarcinoma diagnosed between 2011 to 2014. Pre-treatment absolute neutrophil count, absolute lymphocyte count, and albumin were extracted. The NLR for each patient was calculated and assigned: NLR ≤ 4 g/dl = 0, NLR > 4 g/dl = 1; serum albumin > 4 g/dl = 0, and serum albumin < 4 g/dl = 1. Combining NLR and albumin scores resulted in a composite MPS score of 0-2, similar to GPS. We evaluated the association of the MPS with overall survival. Results: N = 833 patients were identified with median survivals summarized in the table below. A log-rank test showed statistically significant differences in survival between MPS groups (p<0.00005). The MPS on univariate analysis had a HR of 1.36 (95% CI 1.23 – 1.50, p<0.0005) associated with overall survival. Conclusions: The MPS, a novel composite of NLR and albumin, is an objective prognostic tool that divided this sample of patients into three clinically distinct subgroups. Further interrogation will control for performance status, disease characteristics and anti-cancer therapy. [Table: see text]

Background: Multiple myeloma (MM) is characterized by the neoplastic proliferation of plasma cells that most often produce a single monoclonal immunoglobulin (M-protein). In approximately 2% of cases however, two or more distinct M-proteins of different immunoglobulin isotype are noted at diagnosis. The International Myeloma Working Group (IMWG) does not provide specific guidelines for determination of treatment response for MM with multiple M-proteins (MMP) and evidence comparing response and treatment outcomes of patients (pts) with MMP to those with a single M-protein (SMP) is lacking. The current tacit convention for assessing treatment response is to sum all M-proteins to get an aggregate M-protein value to compare to subsequent aggregate M-proteins over time. This approach has not been formally validated; it is unknown whether pts with MMP have a different natural history of the disease or response to treatment. We therefore conducted a retrospective analysis comparing clinical outcomes of 1st-line therapy of patients with MMP vs SMP.

Widespread use of the novel agents bortezomib and lenalidomide has improved outcomes in multiple myeloma (MM). Despite remarkable progress, patients will eventually relapse after exhausting treatment with these drugs. Management of myeloma that is refractory to both bortezomib and lenalidomide (double-refractory MM, DRMM) is complicated due to disease, patient, and treatment-related factors and new therapies for these patients are required. A review of the unique challenges of treating DRMM, recently FDA-approved therapeutic agents, and selected novel drugs under active clinical investigation, is presented below.

this difference in access, patients whose first adjuvant therapy was systemic therapy experienced significantly shorter elapsed times from surgery to the start of adjuvant therapy than did patients whose first adjuvant therapy was radiation therapy.

BACKGROUND:Waiting times for cancer care continue to be an important issue for Canadians. We evaluated 2 cohorts of breast cancer patients to compare changes in elapsed times to care, to determine the proportion of patients who received their postoperative oncology consultation within the recommended time and to examine elapsed times between date of surgery and start of first adjuvant therapy.METHODS:We conducted a retrospective chart review of all women with surgically treated breast cancer who were referred to a provincial cancer centre for adjuvant therapy. The first cohort comprised women referred between Sept. 1, 1999, and Sept. 1, 2000 (n = 342), and the second cohort comprised women referred between Sept. 1, 2003, and Sept. 1, 2004 (n = 295). A general linear model with a stepwise selection was used to identify dominant factors that influenced elapsed times; covariates included cohort period, age at diagnosis, place of residence, disease stage, type of surgery, type of adjuvant therapy, distance to cancer centre, median household income and mean education level.RESULTS:The overall median time from disease detection to the start of first adjuvant therapy for the combined cohorts was 96 days (quartiles 76, 122); this interval was longer for patients in the second cohort (90 v. 102 days, p < 0.001). For the combined cohorts, significantly more patients saw a radiation oncologist within the recommended time from date of surgery than did patients referred to a medical oncologist (82.7% v. 51.7%; p < 0.001). Patients who received adjuvant radiation therapy as their first adjuvant treatment waited longer from the date of definitive surgery to the start of treatment than did patients who received chemotherapy or hormonal treatment (77 v. 48 or 42 days; p < 0.001).INTERPRETATION:The median elapsed time from the detection of breast cancer to the start of first adjuvant therapy was longer in the second cohort (referred in 2003/04) than in the first cohort (referred in 1999/2000). The proportion of patients whose first oncology consultation was within the recommended timeframe varied significantly according to type of oncology specialist, favouring radiation oncology. Despite this difference in access, patients whose first adjuvant therapy was systemic therapy experienced significantly shorter elapsed times from surgery to the start of adjuvant therapy than did patients whose first adjuvant therapy was radiation therapy.

6005 Background: Wait times for cancer treatment are of significant concern to the Canadian public. Previous studies of time to breast cancer care have examined single care events such as time to surgery. The spectrum of care however extends from disease detection to completion of adjuvant therapies. Our objective was to compare elapsed times from breast cancer detection to start of first adjuvant therapy in a Canadian province over 2 time periods; 1999–2000 and 2003–2004. Methods: A retrospective chart review was performed. All eligible women had pathologic confirmation of invasive disease and were referred to a provincial cancer center between Sept 1, 1999-Sept 1, 2000 (cohort 1) and Sept 1, 2003-Sept 1, 2004 (cohort 2). All dates were abstracted from patient charts and electronic records. The log-rank test was used to assess differences in time to events between cohorts. Results: Care intervals assessed (cohort 1 vs 2, median days); (I) detection to pathologic confirmation; 13d vs 14d (p=.12) (II) pathologic confirmation to definitive surgery; 21d vs 24d (p=.002) (III) definitive surgery to referral receipt at cancer center; 16d vs 19d (p=.005) (IV) referral receipt to patient contact; 10.5d vs 6d (p=.04) (V) patient contact to first med/rad onc appt; 5d vs 6d (p=.48) (VI) first appt to start of first adjuvant therapy 11d vs 18.5d (p=.03). Summary elapsed times are presented below with interquartile ranges (IQR). Conclusions: The majority of women experience long elapsed times for breast cancer care. Elapsed times have lengthened over the cohorts studied, although the overall difference did not reach statistical significance. Intervals prior to referral receipt at cancer centers account for prolongation in elapsed times between cohorts. Our data, spanning a sequence of care events, more fully elucidates waiting time burden and may provide a framework for the design and evaluation of programs aimed at reducing elapsed care times. No significant financial relationships to disclose.