The varying distribution of intra- and inter-vertebral height ratios determines the prevalence of vertebral fractures

Credible inferences regarding the burden of vertebral fractures (VFs) cannot be made without a globally accepted quantitative definition of 'fracture'. Currently, differences in anterior, middle, or posterior vertebral heights (VHs) within a vertebra, or between adjacent vertebrae, are used to define 'fracture'. However, VH differences are essential for the construction of thoracolumbar curves, evolutionary adaptations that provide stability in bipedal stance and gait. As there is no reference standard to distinguish anatomical variation from fracture, approaches to defining a VF use a reference range of VH ratios derived in premenopausal women or derived by trimming, a method that iteratively removes the tails of a distribution of VH ratios to produce a normal distribution. From this, reference ranges of VH ratio means and standard deviations (SDs) are obtained and a nominal deviation of 15% or more, or 3 SD or more is regarded as a 'fracture'. We measured VHs by quantitative vertebral morphometry (QVM) and bone mineral density (BMD) by dual energy X-ray absorptiometry in 697 Lebanese women (age 20-89 years) to compare the prevalence of VF ascertained by published methods and a new method that uses the premenopausal range (without trimming) and requires two VH abnormalities. VF prevalence using published methods reached 60% to 70% in pre- and post-menopausal women, and in women with normal or high BMD because VH ratios were not normally distributed and cut-offs used to define VF fracture fell within the observed distribution of the data. The new method resulted in a VF prevalence of 3.3% in younger and 14% in older women, 7% (high), 10% (middle), and 20% (low) BMD tertiles consistent with the notion that the method detected VF due to bone fragility. We suggest that using a fixed trimming method to define reference range and cut-offs or applying fixed cut-offs to identify VFs in populations, where these ratios are not normally distributed, may result in the capture of anatomical variation, not structural failure. Thus, group differences in the VF prevalence may reflect differences in methodology, not bone fragility. Improved criteria to define VF are needed before credible inferences can be made regarding the burden of VFs in women and men, and between sexes, races, countries, decades, and placebo arms of clinical trials.