PARIS, France--A treatment that even modestly improves survival may be important, but most cancer trials are too small to detect such differences, Professor Richard Peto, of Oxford's ICRF Clinical Trial Service Unit, said at the Fifth International Congress on Anti-Cancer Chemotherapy.
"Cancer trials need to be 10 times larger than their current size," Prof. Peto urged, noting that the average cancer trial includes no more than several hundred patients. Although the problem of false-positive results is well controlled by randomization, he said, it takes either a very large "megatrial" or a systematic overview of many different trials to address the problem of false-negatives.
Prof. Peto cautioned against undue emphasis on subgroups, particularly in small trials. "If you have a treatment that works, subgroup analysis is like a machine for producing false-negative results, and, if you have a treatment that doesn't work, subgroup analysis will likely produce false-positive results," he warned.
To illustrate, he reanalyzed the ISIS-2 trial (which proved that aspirin(Drug information on aspirin) saves lives in an acute heart attack) according to patients' zodiac signs. Even though the ISIS-2 trial is large and its overall results are highly significant, this subgroup analysis "revealed" that aspirin apparently failed to benefit patients born under Libra or Gemini.
Often, the ideal target for a cancer trial, Prof. Peto said, is several thousand patients. "This isn't going to be cutting-edge cancer research, but it is going to give useful information about widespread cancer treatments," he said. "First, ask a good, relevant question, and then radically simplify every aspect of the trial," he advised. "Minimize work, maximize size."
To get these large numbers, entry criteria cannot be excessively stringent. Prof. Peto suggested use of the "uncertainty principle"--if, and only if, the physician and patient are both substantially uncertain about the appropriateness of each of the trial treatments, then the patient should be considered eligible.
He described a formula that has proved successful in large-scale cardiovascular trials, which includes telephone randomization with no entry forms and simple 1-page forms completed just from clinical notes.