The semiquantitative assessment of vertebral deformities is based on visual evaluation. The quantitative approach is based on different morphometric criteria. This study is aimed at comparing the impact of different reference groups to define normal vertebral shape on the diagnosis of verterbral deformities. Reference normal values were obtained in three groups of women: French, mixed European, and Argentinian. All these women had normal lumbar spine bone mineral density and no vertebral deformities according to the semiquantitative assessment. In a group of 135 women having vertebral deformities according to Genant's semiquantitative assessment, three different morphometric criteria were applied. Morphometric diagnosis disclosed a good agreement with semiquantitative assessment. Agreement of diagnosis was higher for a given cutoff using thresholds obtained in different reference groups (kappa = 0.84-0.96) and lower when different criteria were compared using thresholds obtained in the same reference group (kappa = 0. 75-0.85). When fracture thresholds obtained in three different cohorts were compared separately for the three morphometric criteria, agreement was the highest when the cutoff was based only on the arithmetical mean of vertebral heights and was independent of its standard deviation (SD). Average vertebral height ratios did not differ between the three reference cohorts, whereas SDs of vertebral height ratios were the highest in the mixed European cohort and the lowest in the French cohort (F = 7.41, p < 0.001). In the three groups of women of different nationality, SDs of vertebral height ratios, but not the arithmetical means, were significantly higher in the radiographs of poor quality compared with those of good quality. Thus, the main source of difference of diagnosis was related to different SDs whereas average height ratios were not different. Differences in SDs between the three groups were found to be related, at least partly, to poor quality of radiographs. The impact of the differences between populations seems less important, however, only three countries were compared. These findings suggest that those techniques that take into account the SD of vertebral height ratios will provide different reference values for vertebral morphometry. Because differences in SDs depend mainly on the quality of radiographs, they can be reduced by improving the X-ray technique and by the use of standardized protocols. This variability will result in the identification of a variable number of vertebral deformities in osteoporotic women. These results may be of importance especially for multicentric studies.