Vertebral fractures are the most common complication associated with osteoporosis. The definition of prevalent or incident vertebral fractures is important in epidemiological evaluation and in clinical trials of osteoporosis treatment. There have been few reports regarding the morphometric cutoff criteria for incident fractures in Asians, and only a few reports of their occurrence in Caucasians. The aim of this study was to establish a definition for incident fractures in Japanese women. A total of 279 women, aged 64.8 +/- 9.9 (mean +/- SD) years, were recruited, and repeated thoracolumbar spinal radiographs were performed at a mean of 1.45 +/- 0.44 years after the baseline film. Vertebral fractures were graded from 0 to 3 in both baseline and follow-up films using a semiquantitative assessment by the consensus of three readers. Incident fractures were defined as an increase of grade, except grade 3, and a vertebral body height of less than 12 mm on the baseline film. As morphometric criteria, both absolute height reduction (AHr) and the percent height ratio (PHr) were calculated. Prevalent fractures were observed in 207 vertebrae (6.2%) and 85 cases (30.5%), and incident fractures in 42 vertebrae (1.3%) and 29 cases (10.4%). As morphometric criteria for incident fracture, both a PHr > or =15% and an AHr > or =3 mm or > or =4 mm indicated sensitivity of 49.17% and 50.83% and specificity of 99.92% and 99.90%, respectively. In conclusion, the morphometric criteria of incident fracture combined with a semiquantitative assessment will provide useful information in the study of clinical osteoporosis, especially for international comparisons.