PURPOSE Evidence suggests that patient-reported outcomes (PROs) from randomized trials in oncology may not influence clinical decision making and patient choice. Reasons for this are currently unclear and little is known about patients' interpretation of PROs. This study assessed patients' understanding of multidimensional PROs in a graphical format. PATIENTS AND METHODS Semistructured interviews in which patients interpreted a series of graphs depicting simple, then multiple different hypothetical PROs associated with two treatments with identical chances of survival were audio recorded. The interviewer and a blinded observer (listening to audio recordings) scored patients' understanding of the graphs. Logistic regression examined the associations between patient understanding of the graphs and clinical and sociodemographic details. Results One hundred thirty-two patients with esophageal and gastric cancer were interviewed and 115 understood the first two graphs depicting different PROs of two treatments (87%; 95% CI,81 to 93). Simultaneous interpretation of adverse and beneficial treatment effects was achieved by 74 (66%; 95% CI, 57 to 75). Graphs showing complex, longitudinal data were correctly interpreted by 97 (73%; 95% CI, 66 to 81) and 108 (81%; 95% CI, 75 to 88), respectively. Univariable analyses demonstrated associations between patient understanding and patient age, educational level, and cancer site (P < or = .02 for all); however, in a multivariable model each of these associations was attenuated. CONCLUSION Most patients understand graphical multidimensional PROs, although a smaller majority were able to interpret more complex, or simultaneous, presentations. Additional work is needed to define methods for communicating clinical and PRO data from trials to allow patients to make informed treatment choices.