Background: Patients with advanced cancer and their caregivers often have different preferences regarding patients’ EOL care. Disagreement in a patient-caregiver dyad can increase stress and result in suboptimal care. Understanding factors that promote agreement, as well as the effect of agreement on care received at EOL, can inform interventions to improve communication and EOL decision-making for patients and caregivers. Methods: 205 patients (Stage III or IV cancer plus limited prognosis) and their caregivers were recruited to a randomized controlled trial of a communication intervention for patients, caregivers, and providers (Cancer Communication Study, PI: Epstein). Before intervention, patients completed the Preferences for Life-Extending Treatment questionnaire, which asked their preference regarding experimental treatment, life support, and palliative care; caregivers were asked about patients’ preferences. Binomial logistic regressions analyses modeled agreement in preferences as a function of patient and caregiver demographic characteristics and EOL care received as a function of patient-caregiver agreement. Results: The majority of patient-caregiver dyads agreed about experimental treatment (60.3%), life support (63.4%), and palliative care (70.7%). Dyads were more likely to agree about palliative care when patients were female (OR = 1.94, p = .03) and non-Hispanic white (OR = 2.10, p = .07) and when caregivers were college educated (OR = 2.04, p = .03). Of the 82 patients who died during study follow-up, 57 (69.5%) received EOL care congruent with their preferences. In 19 of the 38 (50%) cases where patient-caregiver dyads disagreed, caregivers’ preferences predicted EOL care received. Dyadic agreement about life support was associated with increased odds of patients receiving/not receiving life support congruent with their preference (OR = 3.02, p = .02). Conclusions: Facilitating agreement between patients and caregivers could improve receipt of patient-centered care. A communication intervention designed to increase dyadic agreement by helping patients and caregivers discuss challenging EOL decisions might improve EOL care delivery. Clinical trial information: NCT01485627