In this review we have examined epidemiological data from a variety of sources to assess the relation between vitamin A intake and cancer risk. The potential for recall bias in case-control studies makes their interpretation difficult, particularly if we are searching for modest associations. Prospective data are preferable, but sparse. Studies of blood levels of carotenoids may be informative if the blood is stored at ultra-low temperatures; however, studies of blood retinol levels are largely uninformative as an index of dietary intake because blood retinol is not well correlated with intake except in vitamin A-deficient populations. We have also reviewed the evidence for an influence of vitamin A intake on the incidence of cancer at the three major cancer sites accounting for a substantial portion of cancers in developed countries. The available data are compatible with a modest inverse association between intake of vitamin A and breast cancer, although it is not clear whether this effect may be due to preformed vitamin A, carotenoids, or both. The evidence that vitamin A protects against colon cancer is unconvincing. In the case of prostate cancer, early suggestions that vitamin A may increase incidence have not been confirmed by subsequent studies. Fortunately, prospective data from a number of large ongoing cohort studies in the United States and Europe should be available within the next 5 years or so. These data will permit further assessment of potential correlations between vitamin A and cancer at various sites by analysis of much larger numbers of cases than are presently available.(ABSTRACT TRUNCATED AT 250 WORDS)