CONTEXT: High-quality end-of-life (EOL) care depends on thoroughly assessing terminally ill patients' preferences for EOL care and tailoring care to individual needs. Studies on predictors of EOL-care preferences were primarily cross-sectional and assessed preferences for multiple life-sustaining treatments (LSTs), making clinical applications difficult. OBJECTIVE/METHODS: We examined factors predisposing cancer patients (N = 303) to specific LST-preference states (life-sustaining preferring, comfort preferring, uncertain, and nutrition preferring) derived from six LSTs (cardiopulmonary resuscitation, intensive care unit care, chest compression, intubation with mechanical ventilation, intravenous nutrition, and tube feeding) in patients' last six months by multilevel multinomial logistic regression. RESULTS: Participants with accurate prognostic awareness and physician-patient EOL-care discussions were less likely to be in life-sustaining-preferring, uncertain, and nutrition-preferring states than in the comfort-preferring state. Better quality of life (QOL) and more depressive symptoms predisposed participants to be less likely to be in the uncertain than in the comfort-preferring state. Membership in the nutrition-preferring rather than the comfort-preferring state was significantly higher for participants in the state of moderate symptom distress with severe functional impairment than in the state of mild symptom distress with high functioning. CONCLUSION: Accurate prognostic awareness, physician-patient EOL-care discussions, QOL, depressive symptoms, and symptom-functional states predisposed terminally ill cancer patients to distinct LST-preference states. Clinicians should cultivate patients' accurate prognostic awareness and facilitate EOL-care discussions to foster realistic expectations of LST efficacy at EOL. Clinicians should enhance patients' QOL to reduce uncertainty in EOL-care decision making and provide adequate psychological support to those with more depressive symptoms who prefer comfort care only.