10 Background: Little is known about the individual and combined effect of early-adulthood obesity and cumulative smoking on the survival of esophageal adenocarcinoma (EAC) patients. Methods: We analyzed two independent cohorts of EAC patients: 235 patients from Toronto, Canada (TO, 2006-2011) and 329 patients from Boston, USA (BO,1999-2004). Associations between early adulthood body mass index (EA-BMI) and smoking with overall survival (OS) were assessed using Cox proportional hazard models, adjusted for stage, treatment, and other relevant covariates. Results: Median age (range) for TO dataset was 64(29-88)yrs; for BO dataset, 64(21-91)yrs. Males comprised 86% of TO and 89% of BO datasets. 90% of TO and 98% of BO patients were Caucasians. The Median (range) for packyears was 34 (0.2-118; TO) and 34 (0.2-212; BO). The Median (range) for EA-BMI was 24(15-44; TO) and 24(15-47; BO). Median BMI 1 yr prior to diagnosis was 25(16-43; TO) and 25(20-49; BO). 92% of TO and 88% of BO patients had ECOG 0 or 1. Disease stage distribution (early/locally-advanced/metastatic) was 11%/64%/25% (TO) and 30%/52%/18% (BO). For TO, the aHR for smoking was 1.03 (95%CI: 1.02-1.04; p=8E-08) per packyear, while for BO, smoking also independently conferred worse OS, with aHR of 1.007 (95%CI: 1.002-1.01; p=0.003) for each packyear increase. The aHRs for being underweight (EA-BMI<18.5), overweight (EA-BMI 25-30), and obese (EA-BMI>30) in early adulthood were 2.19 (95%CI: 1.0-4.6), 1.89 (95%CI:1.2-3.0), and 2.49 (95%CI:1.5-4.2), respectively for the TO dataset (global p=0.003 for EA-BMI). In BO, the corresponding values were 1.30 (95%CI: 0.8-2.2), 1.45 (95%CI: 1.0-2.5), and 2.39 (95%CI:1.5-3.8), respectively (global p=0.002). In contrast, BMI at one year prior to diagnosis had no association with OS in either study. Conclusions: Elevated BMI in early adulthood and heavy cumulative smoking history are independently associated with increased mortality risk in two North American EAC populations. These survival differences may reflect comorbidity differences, biological differences or both, and offer insight into how key modifiable behaviors in prevention can also affect cancer prognoses. AS, LC, DCC and GL contributed equally.