Search

59 Background: Historically, the prognosis of advanced melanoma has been poor. Recent landmark studies (KEYNOTE-006, CheckMate 066, CheckMate 037 and CheckMate 067) have shown significant improvement in outcomes of patients with advanced melanoma treated with ICIs. In a single institution retrospective study, we evaluated the efficacy and toxicities of ICIs in patients with advanced melanoma treated in real-world clinical practice. Methods: We included patients who received pembrolizumab (PEMBRO), nivolumab (NIVO), or ipilimumab plus nivolumab (IPI/NIVO) at the University of Rochester Medical Center from June 2015 to December 2018. Patient- and cancer-related characteristics were collected and compared between treatment groups. Outcomes including duration of treatment, response rates, and adverse events (AEs) were captured. Progression free survival (PFS) and overall survival (OS) were summarized using the Kaplan-Meier method. Results: We included 89 patients (55 received PEMBRO, 20 received NIVO, and 14 received IPI/NIVO); median age at ICI initiation was 68 years (range, 28-92) and 29% had Eastern Cooperative Oncology Group Performance Status (ECOG PS) of ≥2. Forty-five patients (51%) had ≥2 co-morbidities. Nine patients (10%) had autoimmune diseases prior to initiating ICIs and 34 patients (38%) had brain metastasis. ICIs were given as first line in 71 patients (80%). Compared to those who received PEMBRO or NIVO, patients who received IPI/NIVO were younger (median age was 61 vs.71 vs. 70, p=0.003) and had better ECOG PS (ECOG PS ≥2 was 0% vs. 40% vs. 33%, p=0.007). Table shows the outcomes for the three groups. Conclusions: Patients with advanced melanoma treated with ICIs derived similar response rates to those seen in published landmark studies. However, median OS was shorter (range was 38 months to not reached in published studies). [Table: see text]

e15580 Background: Trimodality treatment (tx) with neoadjuvant chemoradiation (CRT) followed by esophagectomy is standard tx for locally advanced EGJ cancer. Post-operatively, there is no strong consensus about role of routine surveillance imaging. At the University of Rochester, patients (pts) have surveillance CT scans every 4-6 months (mos) for the first 2 years post-esophagectomy and every 6-12 mos for the next 3 years. Methods: Pts were identified who underwent esophagectomy for T1-T3 EGJ cancer between January 2011 and December 2015 at our institution. Objectives were to describe the impact of timing and methods of recurrence detection (MoRD) on patient outcomes. Recurrence-free (RFS) and overall survival (OS) were graphed via the Kaplan-Meier method. Results: 138 pts underwent esophagectomy for EGJ cancer: 107 (77.5%) were male, median age was 64, and 116 patients (84.1%) had adenocarcinoma. 112 pts (81.2%) had neoadjuvant CRT. The entire cohort’s median OS was 38.4 mos. 68 pts (49.3%) relapsed with a median RFS of 20.0 mos. Recurrence was detected by routine imaging in 36 pts (52.9%), imaging triggered by symptoms in 27 pts (39.7%), and symptoms alone in 5 pts (7.4%). Post-relapse median OS was 2.3 mos when detected based on symptoms alone, 5.0 mos when detected by imaging triggered by symptoms, and 13.7 mos when detected by routine scans (log-rank p = 0.041). There was no significant association between baseline patient/tumor characteristics or pathologic response and MoRD . 53 patients (77.9%) received salvage/palliative tx with a median of 2 tx (IQR = 1). There was no significant association between MoRD and number of salvage/palliative tx. Conclusions: 49.3% of pts relapsed after esophagectomy for EGJ cancer, consistent with current literature. Almost half of relapses were detected based on symptoms despite routine imaging. Increased OS for pts with relapse detected by routine scans is likely related to lead time bias, but may also be related to increased tx intensity or less aggressive tumors. MoRD did not have a measurable impact on number of lines of post-relapse tx. Prospective randomized trials are needed to determine real benefit of regular surveillance scans among EGJ cancer survivors.

7547 Background: Outcomes remain poor for patients with relapsed refractory DLBCL and HGBCL, especially those ineligible for a stem cell transplant or CAR T-cell therapy. Lenalidomide has efficacy in these groups, most notably in the activated B-cell like (ABC) subtype when defined by gene expression profiling. We analyzed the outcomes of consecutive patients with DLBCL treated with lenalidomide, focusing on characteristics such as transformed histology and MYC translocation status. Methods: We performed a retrospective review of consecutive patients with transformed indolent NHL, DLBCL, and HGBCL treated with lenalidomide at the University of Rochester between 2011-2018. Cell of origin was determined by Hans algorithm and FISH was performed to detect MYC, BCL2, and BCL6 translocations. Kaplan Meier estimates and descriptive statistics were utilized for analysis. Results: 62 patients were identified with a median age of 76 years, the majority with > 2 prior therapies. ORR was 43.5%, including 14 patients with a CR. Median PFS was 4.6 months with 18 patients achieving a PFS > 1 year. Median OS was 14 months. No difference in PFS, OS, or ORR was observed between the de novo germinal center b-cell (GCB) and non-GCB populations (PFS 4 vs. 5 months, p = 0.87). 16 patients with transformed FL had a median PFS of 24 months and a median OS of 46.7 months (vs. de novo GCB OS of 7.8 months, p = 0.02). Notably, 6/7 MYC+ patients achieved an objective response, including 3 with a CR. All patients with double and triple hit disease had an objective response, including two PRs and one CR. Conclusions: Our experience confirms the clinical activity of lenalidomide in heavily pretreated older adults with DLBCL. Outcomes did not differ by cell of origin using the Hans immunohistochemistry algorithm and durable responses were observed, particularly in transformed FL. Nearly all of the MYC+ patients responded, including those with double/triple hit genetics, and almost half achieved a CR. Based on our data, lenalidomide may be an optimal bridge to consolidative cellular therapy in high-risk patients irrespective of genotype, particularly in transformed indolent histologies.

e19058 Background: Diffuse large B-cell lymphoma (DLBCL) is the most common subtype of non-Hodgkin lymphoma. In this study, we explore reasons for non-enrollment in clinical trials for DLBCL and implications on trial design and interpretation. Methods: This is a retrospective analysis of patients (pts) with a pathological diagnosis of DLBCL or high grade B-cell lymphoma (HGBL) at University of Rochester (4/14-6/16) and New York-Presbyterian Hospital/Weill Cornell Medicine (NYP/WCM) (4/14-4/17).Ten clinical trials were opened during this time. Participants were divided into 3 groups: those treated in trial, those not enrolled in trial because of need for urgent treatment, and those not enrolled in trial for any other reason. We used a center-stratified Cox proportional hazards model to estimate association of trial enrollment with progression-free survival (PFS; time from start of treatment until progression/death or the last date the pt was known to be progression free) and overall survival (OS). Results: We identified 263 pts; 17% (n = 45) enrolled in a trial. Reasons for non-enrollment included not meeting eligibility criteria (n = 98), physician choice (n = 50), and pt choice (n = 38). For 32 pts, reasons were unclear. Of the 50 pts who were not enrolled because of physician choice, the primary reason for non-enrollment was the need for urgent treatment (n = 46). Pts who needed urgent treatment had higher risk clinical features compared with pts in trial (Table). Compared with those treated in trial and those not enrolled in trial for any other reason, those not enrolled in trial due to need for urgent treatment had an inferior PFS (HR 2.61, 95% CI 1.23–5.16) and OS (HR 2.27, 95% CI 1.21–4.06). Conclusions: At 2 academic institutions, 52% of patients with DLBCL or HGBL required urgent chemotherapy and failed to enroll on trials. Exclusion of such patients limits the applicability and generalizability of clinical trials in DLBCL. This barrier must be overcome so clinical trials may better reflect true DLBCL demographics. [Table: see text]

322 Background: Germline and somatic HRR alterations are present in 12% and 20% of patients with mPC, respectively, and can have treatment implications. NCCN guidelines recommend offering genetic testing for patients with mPC. This study examines genetic testing patterns in patients with mPC at UR. Methods: We conducted a retrospective study of all patients (n = 240) with mPC who saw medical oncology at UR from 10/1/2017 to 4/1/2018. We abstracted the following data from medical records: patient demographics, mPC characteristics, types of genetic testing, and clinically significant (CS) HRR alterations as defined by Myriad, Ambry, and Foundation One. To compare characteristics of patients who had genetic testing to those who did not, we used non-parametric Wilcoxon rank-sum test for continuous variables and Fisher’s exact test for categorical variables. Results: The median age at mPC diagnosis was 71 years (range: 42-93 years), 86.3% were white, and 35% had mPC at initial diagnosis. 64 patients (26.7%) were offered genetic testing and 40 patients (16.7%) completed germline (n = 12), somatic (n = 21), or both germline and somatic genetic testing (n = 7). Median time from diagnosis of mPC to germline and somatic genetic testing were 9 months and 35 months, respectively. Among the 24 patients who were offered, but did not have testing, the reasons were: patients’ refusal (n = 10), death/hospice care (n = 7), insurance issue (n = 3), patient’s preference to have somatic testing at disease progression (n = 3), and missed genetic counseling visit (n = 1). Compared to patients who did not have genetic testing, those who had testing were younger at the time of mPC diagnosis (63.5 vs 73.0 years; p < 0.0001). CS HRR alterations were identified in 8 of 40 patients tested (20%): somatic HRR alterations include BRCA2 (n = 2), ATM (n = 1), CDK 12 (n = 1), and C HEK2 (n = 1); germline HRR alterations include BRCA2 (n = 2) and CHEK2 (n = 1). Conclusions: We report that approximately one-fourth of patients with mPC at UR had genetic testing offered, with a 20% prevalence of HRR alterations in those tested. Future research should develop strategies to address barriers to genetic testing.

32 Background: Trimodality treatment with neoadjuvant chemoradiation (CRT) followed by surgery is a standard treatment for esophageal/GEJ (E/GJ) cancers. Following esophagectomy, there is no strong consensus about optimal surveillance and routine imaging. At our institution, patients have surveillance CT scans every 4-6 months for the first 2 years post-surgery and every 6-12 months for the next 3 years. Methods: An IRB-approved chart review was performed identifying patients who underwent surgical resection for locally advanced E/GJ cancer between January 2011 and December 2015 at the University of Rochester. Study objectives were to describe timing of and methods used to detect recurrence as well as their impact on patient outcomes. Recurrence-free (RFS) and overall survival (OS) were graphed via the Kaplan-Meier method. Results: 138 patients underwent surgical resection for E/GJ cancer during the study period: 107 (77.5%) were male, median age was 64, and 116 patients (84.1%) had adenocarcinoma. 111 patients (80.4%) received neoadjuvant CRT. Median OS for entire cohort was 43.4 months. 65 patients (47.1%) relapsed with a median RFS of 19.8 months. Recurrence was detected by routine imaging in 34 patients (52.3%), imaging triggered by symptoms in 25 patients (38.5%), and symptoms alone in 6 patients (9.2%). Median OS post-relapse was 1.5 months when detected based on symptoms alone, 5.0 months when detected by imaging triggered by symptoms, and 13.5 months when detected by routine scans (Log-rank p = 0.046). There were no significant associations between baseline patient /tumor characteristics and subsequent method of recurrence detection. Conclusions: 47.1% of patients suffered relapse after trimodality therapy for E/GJ cancer, consistent with published literature. Almost half of these were detected based on symptoms despite routine imaging. Increased OS for patients with relapse detected by routine scans is likely related to lead time bias, but may be related to increased treatment intensity, or due to less aggressive tumors. Prospective randomized trials are needed to determine the true benefit of regular surveillance scans among esophageal cancer survivors.

100 Background: Use of checkpoint inhibitors has expanded over recent years, and the gastrointestinal (GI) oncologic indications for immunotherapy (IO) continue to evolve. IO use in GI malignancies is still in its infancy, and we sought to describe our clinical experience thus far. Methods: We retrospectively identified all patients (pts) with GI malignancies treated with IO at the Wilmot Cancer Institute from December, 2015 through July, 2017. We examined patient and tumor-related clinicopathologic features. Our main outcomes of interest were best response on IO, and frequency/severity of IO toxicity. Fisher’s exact test was used to compare categorical factors between groups of interest. Results: 21 pts received IO for a GI malignancy, all of whom had metastases (mets). 76% of pts were male, and the median age was 62 years (range, 25 – 87). Common diagnoses were colorectal (CRC) (n = 10), gastroesophageal (GE) (n = 5), and hepatocellular cancers (HCC) (n = 4). 76% of pts received pembrolizumab, and 24% of pts received nivolumab. 67% received IO as ≥ 3rd-line therapy. The median time from diagnosis of metastatic disease to initiation of IO was 17.8 mos (range, 3 – 140). The median time on therapy was 2.62 months (95% CI 1.48, 6.52 mos) (range, 0.5 – 11 mos), with 3 pts still responding at the time of the analysis. Pts with HCC had the longest time on IO (8.95 mos), compared to 2.62 mos for GE, 2.39 mos for CRC, and 1.26 mos for all other (p = 0.10). Pts who received IO as 3rd or greater line of therapy had a significantly increased likelihood of non-response or progressive disease/death compared to pts who received IO as 2nd-line (p = 0.02). There was a trend towards non-response to IO in pts with liver mets vs. extrahepatic mets (p = 0.09). 57% experienced grade 3/4 toxicities: the most common were anemia (24%), hyponatremia (14%), and fatigue (14%). No IO-specific grade 3/4 immune-related toxicities were observed. Conclusions: The majority of our pts received IO as 3rd line therapy or beyond, and were significantly less likely to respond to treatment compared to those who received IO earlier on. The presence of liver mets is potentially associated with non-response to IO, and needs further exploration. No unexpected or severe IO-related toxicities were observed.

144 Background: Neoadjuvant (NA) chemoradiation (CRT) followed by surgery is a standard treatment for locally advanced esophageal/GEJ tumors. The optimal management strategy is less clear for unresectable tumors or in patients (pts) who are medically inoperable. Methods: We retrospectively identified all pts with stage II/III esophageal/GEJ cancers treated without surgery at our institution from 2011-2015. Objectives were to describe treatment regimens used and assess outcomes based on clinicopathologic features. Overall survival (OS) and progression-free survival (PFS) were estimated using Kaplan-Meier curves. Results: 38 pts were identified. 66% were male; median age was 69 years (range 43-91). 19 pts had adenocarcinoma, 18 had squamous cell, and 1 had small cell. Primary sites were distal (n = 22), mid (n = 13), and cervical esophagus (n = 3). 31 pts were treated with curative intent, 7 with best supportive care. 30 pts underwent CRT: 16 in a NA setting, 14 in a definitive setting. Chemotherapy regimens used were Carboplatin/Paclitaxel (19 pts) and Cisplatin/5-FU (10 pts). Median NA radiation (RT) dose was 45 Gy vs 50.4 Gy in the definitive setting (p = 0.008). Best response included 6 pts with complete responses (CR), 12 with partial responses, and 12 with progressive disease/death (PD). In the NA group, resection was avoided due to PD (n = 10), comorbidities (n = 4), unresectability (n = 1), and CR (n = 1). Median PFS and OS for all 38 patients were 6.5 and 11.1 months (mos). Median PFS and OS for adenocarcinoma pts were 4.6 and 7.2 mos, compared to 9.1 and 12.5 mos for pts with squamous. Compared to NA CRT, pts treated with definitive CRT had significantly prolonged PFS (18.5 vs 2.5 mos, p = 0.001) and OS (27.1 vs 5.3 mos, p = 0.03). Conclusions: In the absence of resection, outcomes for locally advanced esophageal/GEJ tumors remain poor, especially for adenocarcinoma pts who progress during NA CRT. A PET-directed approach as suggested by CALGB 80803 may help to identify these pts earlier in their course, allowing for treatment change. Multidisciplinary assessment to identify non-surgical candidates at diagnosis is critical since RT doses used in the definitive setting are generally higher.

6536 Background: Contemporary precision medicine trials in DLBCL often require real-time central pathology review for enrollment. Central review may lead to treatment delays and prevent high risk patients (pts) with aggressive presentations from enrolling onto clinical trials. We explored reasons pts with DLBCL were not enrolled on trials and the implication of non-enrollment on trial design and interpretation. Methods: We retrospectively analyzed all pts with histologic diagnosis of DLBCL or HGBL from 4/14 to 6/16 at the University of Rochester. Therapeutic trials open during this time included 3 sponsored and 2 NCTN studies. The Kaplan-Meier method was used to estimate the distribution of progression-free survival (PFS; time from start of treatment until progression/death or until the last date the patient was known to be progression free) and overall survival (OS). Results: 140 pts were identified; 22% enrolled on a trial. Reasons for non-enrollment included: 1) Protocol ineligibility (n=58); (2) Physician choice (n=24) and; 3) Patient choice (n=20). Reasons were unclear in 8 pts. Of the 24 pts who were not enrolled due to physician choice, 21 required urgent treatment secondary to symptoms or rapid progression. Compared to pts treated on trial, pts with rapid progression had higher risk clinical features (table). There was a trend towards a lower 1-year PFS rate in pts who required urgent treatment compared to those on trial (72.1% vs. 56.1%; p=0.08). There was no statistical difference in OS. Conclusions: At our institution, for patients with DLBCL meeting trial eligibility criteria, 42% required urgent chemotherapy and failed to enroll. Exclusion of these high risk patients in precision medicine trials has important implications in the interpretation and generalizability of clinical trials in DLBCL. In this curable malignancy, excluding high risk patients from trials limits the event rate, and associated power to demonstrate impact of novel therapies. [Table: see text]

7543 Background: Certain non-Hodgkin lymphomas (NHL) such as primary mediastinal B-cell, Burkitt’s, and high-grade B-cell lymphoma with MYC and BCL2 rearrangements are often treated with infusional dose-adjusted rituximab, etoposide, prednisone, vincristine, cyclophosphamide and doxorubicin (DA-R-EPOCH), which requires a central line. We have observed meaningful line-associated complications (LAC) in patients (pts) treated with DA-R-EPOCH. With the ongoing use of this regimen, we sought to identify the rates and correlates of LAC in this population. Methods: We retrospectively identified all pts treated with DA-R-EPOCH at the Wilmot Cancer Institute between 3/2011 and 10/2015. Our primary endpoint was the rate of LAC, including venous thromboembolism (VTE), extravasation, and line-associated infection (LAI) diagnosed during DA-R-EPOCH therapy. Our secondary endpoint was the rate of VTE during therapy. Rates and 95% confidence intervals (95% CI) were calculated for all endpoints. Univariate logistic regression was used to calculate odds ratios to evaluate potential predictors. Results: 43 pts received DA-R-EPOCH during the study period. 70% of pts were male; median age was 52 years. 17 pts (39.5%, 95% CI 0.25 - 0.56) experienced at least 1 LAC: 15 pts (35%, 95% CI 0.21 - 0.51) had VTE; 3 pts had LAI; and 2 pts experienced extravasations. Grade 3 toxicity was seen in 41% (7/17): 4 pts with VTE, and 3 pts with LAI. Both extravasation events were grade 2, and both occurred with mediports. In univariate analysis, BMI ≥ 35 kg/m² and receiving therapy via peripherally inserted central catheter (PICC) line were significantly associated with an increased risk of VTE (p=0.04 and p=0.02, respectively). Conclusions: 40% of pts receiving DA-R-EPOCH therapy for treatment of NHL developed LAC, almost half of whom experienced grade 3 toxicities. Clinicians need to balance these risks when selecting therapy, particularly with the lack of randomized data to support the DA-R-EPOCH approach in many circumstances. Given observed extravasations, we avoid mediports in favor of PICC lines, however this approach carries a significant risk of VTE. Future studies are needed to evaluate the role of prophylactic anticoagulation in this population.

273 Background: Cholangiocarcinoma (CC) is a highly lethal malignancy with a poor 5 year overall survival in the advanced setting. Benefits of treating CC with chemotherapy are limited and no FDA approved targeted therapies currently exist. CC is associated with a desmoplastic reaction comprised of extracellular matrix, fibroblasts and inflammatory leukocytes. Myeloid-derived suppressor cells (MDSC) are the predominant tumor infiltrating leukocytes in several cancers, but little is known about their role in CC. MDSCs are derived from a monocytic (M-MDSC) or granulocytic (G-MDSC) origin and infiltrate the tumor microenvironment (TME) where they suppress anti-tumor immunity and play a crucial role in tumor growth, chemoresistance and metastasis. Here, we characterize the prevalence and prognostic significance of MDSC in CC. Methods: A de-identified human CC clinical database (n = 220) with linked tumor tissue was collected under an IRB approved protocol. Total white blood cell count and percent monocytes (a surrogate marker for M-MDSC prevalence in peripheral circulation) before starting therapy were collected and correlated with overall survival in patients with unresectable CC. Immunohistochemistry for myeloid cell markers (CD33, CD11b, CD14, CCR2, and CD15) was performed on formalin-fixed paraffin embedded human CC and normal liver sections. Staining was scored by a pathologist or quantified with digital imaging software. Results: Analysis of chemotherapy naive blood specimens revealed that an increase in monocyte percentage inversely correlated with survival in patients with unresectable CC (p = 0.0117). Statistical significance was more pronounced in patients with intrahepatic CC (p = 0.0091) compared to other disease sites. There was an increased level of M-MDSC and G-MDSC in CC compared to normal liver tissue and all 22 tested CC samples stained positively for MDSC markers within the tumor stroma. Conclusions: The prevalence of circulating monocytes correlates negatively with prognosis and MDSCs infiltrate the TME in unresectable CC. Therefore, targeting MDSC is an attractive therapeutic strategy and future studies aim to fully characterize the TME and block MDSC recruitment in a mouse model of CC.

350 Background: The NCCN defines “distress” as a “multifactorial…emotional experience…that may interfere with the ability to cope effectively with cancer, its physical symptoms and its treatment.” NCCN DT scores ≥ 4 on a scale of 1-10 are associated with increased biopsychosocial distress among oncology patients. We piloted the use of an electronic tool for distress screening in an academic GU medical oncology practice to determine feasibility and impact on workflow. Methods: At each clinic visit, patients completed a modified NCCN DT with distress score only and pain interference, depression, and anxiety PROMIS domains on an iPad before meeting the clinician. Results were immediately available in the electronic medical record. Results: Between July 1, and August 31, 2015, 108 assessments were completed by 73 patients; 12% were new patients and 88% were follow-ups. Patients ranged from 30 to 93 years of age. Mean DT score was 3.67 [standard deviation (SD) 2.79]. DT scores of ≥ 4 were reported in 41% of assessments. Mean distress scores based on cancer stage were: stages 1-3, 4.04 (SD 2.96; N = 23); and stage 4, 3.56 (SD 2.75; N = 85). Mean distress scores based on cancer diagnosis were: prostate, 3.20 (SD 2.38; N = 46); urothelial, 3.97 (SD 3.35; N = 33); kidney, 3.92 (SD 2.55; N = 24); and testicular, 4.80 (SD 3.42; N = 5). Mean T scores for each of the PROMIS domains were: pain interference, 54.16 (SD 10.28; N = 103); depression, 49.54 (SD 8.95; N = 101), and anxiety, 50.79 (SD 9.37; N = 101). Median time to assessment completion was 3.77 minutes (mean 6.15 minutes). Mean times by age groups were: 50-59, 7.65 minutes (N = 21); 60-69, 5.57 (N = 48); 70-79, 11.61 (N = 16); and ≥ 80, 10.21 (N = 12). Conclusions: Electronic biopsychosocial distress screening at each clinic visit is feasible and has minimal impact on workflow. Pain interference but neither anxiety nor depression scores were increased in this population. Future research is needed on the influence of cancer stage and type on distress and on the associations between NCCN DT and PROMIS scores.

443 Background: Local-regional radioembolization with Yitrium-90 (Y-90) has become standard practice for patients with hepatocellular carcinoma (HCC) either as a bridge to transplant, or for local disease control. Outcomes data in the United States are limited and here we review our institutional experience with Y-90 radioembolization. Methods: We retrospectively reviewed charts from 70 patients with HCC who were treated with Y-90 from May 2010- January 2014. Clinical variables including Child-Pugh class and CLIP score were extracted from patient records. The Cox proportional hazards model was used to determine prognostic factors, and Kaplan-Meier curves were used to determine PFS and OS. Results: Median age was 61 (range 43-82), 79% Caucasian, 84% male, and 79% Child-Pugh class A. Median progression free survival (PFS) was 8.4 months (95% CI 6-10.7) and overall survival (OS) was 14.2 months (95% CI 9.7-21). Overall survival significantly differed by Child -Pugh score (p= 0.009), CLIP score (p=0.003), and presence of portal vein thrombosis (PVT) (p=0.0384), based on the log-rank test comparing Kaplan-Meier curves. Using univariate Cox proportional hazards models, both elevated baseline AFP, measured on a log scale (HR 1.79, 95% CI 1.32-2.43, p=0.0002) and post Y-90 treatment with sorafenib (HR=2.30, 95% CI 1.07-4.95, p=0.03) were associated with worse mortality. Elevated AFP (HR 2.45, 95% CI 1.73-3.47, p