Background: Informed consent is a prerequisite for patients included in clinical trials. Trial design, inclusion criteria and legal requirements are increasingly complex. This complexity challenges design and delivery of written and oral trial information to ensure understandable information. To evaluate the level of concordance between patients' and informing physicians' assessments regarding patient understanding of trial information, we carried out a study based on paired questionnaire data from patients and their physicians. These assessments of patient understanding were further correlated with patients' factual knowledge of the information provided. Methods: This pilot study included patients and physicians immediately after the patients had received information on one of 23 ongoing phase III randomised cancer trials at two Swedish sites. In total, 46 patients and 17 physicians contributed data based on two new questionnaires with seven mirroring questions, where concordance was analysed with McNemar's test. These assessments of patients' self-estimated understanding were further correlated with the Patient Understanding of Research (Q-PUR) questionnaire that assesses factual knowledge of the information provided. Results: For each question, 47-61% of the patient-physician pairs were in concordance regarding their assessments of patients' 'fully understanding' or 'not fully understanding' various aspects of the trial information. For the discordant pairs, the physicians rated patient understanding lower than the patients themselves, for all seven questions. This difference was significant for five of the questions (P ≤ 0.017). The median Q-PUR knowledge score was 11 out of 12, but this score did not significantly correlate with the assessments, either from patients or from physicians. Conclusions: This study demonstrated a trend for physicians to rate the level of understanding of trial information among potential trial patients lower than the patients themselves. Application of Q-PUR revealed high knowledge scores, but without correlation to the assessments. These findings need validation in an independent setting, with an improved instrument with mirroring questions, and a better-matched measurement of patients' factual knowledge. These results suggest that physicians need to improve their ability to assess patient understanding of clinical trial information, in order to be able to tailor the patients' information individually.