A series of failed clinical trials were examined by Dr Jonathan Kimmelman of McGill University, Montreal, and Dr Alex London, of Carnegie Mellon University, Pennsylvania.
They said that problems arose when forecasts about humans trials were made from animal studies.
Focusing on a new drug's performance in pre-clinical trials produces 'overly optimistic predictions', the researchers said. 'We also need to look at how similar interventions have fared in the past,' Dr London said.
'If drugs that work on the same principle have failed development, there may be good grounds for tempering expectations,' he added.
The research recommended that techniques used in clinical trials, such as randomisation and blind testing, should also be used in the design of pre-clinical trials.