For the purpose of deriving a UL, three primary adverse effects of chronic vitamin A intake are discussed below: (1) reduced bone mineral density, (2) teratogenicity, and (3) liver abnormalities. High β-carotene intake has not been shown to cause hypervitaminosis A. Therefore, this review is limited to the adverse effects of preformed vitamin A or retinol. The terms vitamin A and retinol will be used interchangeably in the following sections. Because provitamin A carotenoids were not included in vitamin A supplements until the late 1980s, it is assumed that studies and case reports published before 1990 used preformed vitamin A in supplements. The UL derived here applies to chronic intake of preformed vitamin A from food, fortified food, and/or supplements.

Bone Mineral Density. Chronic, excessive vitamin A intake has been shown to lead to bone mineral loss in animals (Rohde et al., 1999), making such a consequence in humans biologically plausible. Most human case reports are not well described and epidemiological studies are inadequate in design. However, four studies provide interpretable evidence relating changes in bone mineral density (BMD) and risk of hip fracture with variation in dietary intake of preformed vitamin A (Freudenheim et al., 1986; Houtkooper et al., 1995; Melhus et al., 1998). The studies are distinguished by their well-described study designs and populations, adequate dietary intake estimates, and accurate methods for measuring BMD at multiple sites.

One two-part study (Melhus et al., 1998) suggests that a chronic intake of 1.5 mg/day of preformed vitamin A is associated with osteoporosis and increased risk of hip fracture. The first part, a cross-sectional multivariate regression analysis in 175 Swedish women 28 to 74 years of age, showed a consistent loss in BMD at four sites and in total BMD with increased preformed vitamin A intake. Numerous nutritional and non-nutritional exposures were concurrently assessed, allowing substantial control of potential confounders. With the use of stratified estimates of retinol intake in a univariate regression analysis, BMD was shown to increase with each 0.5 mg/day increment in intake above a reference intake of less than 0.5 mg/day, until intakes exceeded 1.5 mg/day. Above this level, mean BMD decreased markedly at each site. In a multivariate model, adjusting for effects of 14 other covariates, similar results were found. It is not clear whether the findings are equally applicable to pre- and postmenopausal women.