A recent meta-analysis of 16 publications suggested that bone mineral density (BMD) is not associated with vitamin D receptor (VDR) gene polymorphism (VDRGP) at the 0.05 significance level when a study with genotyping mistakes is excluded. We wished to determine whether 'positive' findings supporting the BMD-VDRGP association may be explained by chance, and what factors affect the outcomes of these studies. Seventy-five articles and abstracts on the association of VDRGP with BMD and related skeletal phenotypes published before January 1997 were identified. Twenty-three of 67 (34.3%) studies on spinal BMD and 22 of 51 (43.1%) on femoral neck BMD had found a BMD-VDRGP association at p < 0.05, significantly (p = 7 x 10(-14) for spinal BMD, p = 9 x 10(-16) for hip BMD) higher than the expected 5% false positive rate under the null hypothesis of 'no association'. 'Positive' results were more frequently observed in studies on females before the menopause than those on females after the menopause (p < 0.02) or on male and female subjects combined (p < 0.05) when skeletal phenotypes at any bone sites were considered. The 'positive rate' among studies was also influenced by the age range of subjects studied and by the inclusion of subjects with osteoporosis. It is concluded that: (1) BMD is associated with VDRGP with high levels of confidence and (2) non-genetic factors and genetic heterogeneity interfere with the detection of the effects of VDRGP on bone phenotypes.