There is no consensus on how the difference between control and experimental outcome rates, the clinically important difference, should be estimated when designing a clinical trial. We sought to determine whether community and academic clinicians had different perceptions as to what would constitute a clinically important increase in survival, when asked to respond in absolute or relative terms, before a trial was started rather than when the results were already known. A telephone survey of 25 practicing Canadian oncologists was performed. Questions were asked as to the importance of acceptable and minimally acceptable improvements in survival for a hypothetical trial of pancreatic cancer where the baseline survival was expected to be between 2 and 8 months. Responses were sought for absolute (additional months) or relative gains (percent improvement) in survival. The mean absolute additional survival expectations corresponded to at least a doubling of baseline survival and tended to be greatest when the prognosis was poorest (p = 0.06). Relative expectations for improved survival varied with baseline survival (p