Evaluation of human trial findings.

Final proof for the anticarcinogenic effect of any agent can best be obtained from a controlled trial in a sufficiently large population at risk of cancer. Interpretation of the results from such a trial may be difficult, particularly if the results are only marginally significant or negative, or if they are in conflict with epidemiological data or studies in laboratory animals. This article discusses difficulties in the evaluation of clinical trial findings, using the Finnish Alpha-Tocopherol, Beta-Carotene Cancer Prevention (ATBC) Study (the Alpha-Tocopherol, Beta-Carotene Cancer Prevention Group, 1994) as an example. Special attention is paid to problems associated with trial design, properties of the study population, time of intervention, and selection of the dose of the putative anticarcinogenic agent. Potential sources of bias are discussed. The advantages and disadvantages of a single large trial versus several small or medium-sized trials are examined.