Research

Purpose: Although the inter-observer and intra-observer measurement variability of thoracic kyphosis has not been quantified, it is expected to be larger than in scoliosis. This variability is affected by radiograph technique, but more importantly by upper end vertebra (UEV) selection. Additionally, the variability is affected by radiographic quality and inherent difficulties in visualizing the upper thoracic spine.
Methods: Radiographs from 62 patients with Scheuermann's kyphosis were separated into two groups: Group A, UEV could be identified, Group B otherwise. Five spine surgeons identified the upper and lower end vertebrae and magnitude of the kyphosis. Each measured radiographs in Group A (N=41) using two methods, each performed twice. First method included no information; in the second the end vertebrae were provided. Group B (N=19) were measured with three methods, each performed twice. First, with no instruction; secondly with specific instructions on how to identify the UEV; lastly, the UEV was specified. Repeat measurements were separated by at least one week.
Results: The UEV, chosen by the senior author, occurred at T1 (29%), T2 (48%) or T3 (23%); the Cobb angle ranged from 44º to 118º (mean=71.2º). The intra-observer variability was 4.2° with no information and varied across observers (range 2.7° to 8.8º; p-value=0.001). Overall, 95% of the pairs of measures within observer were within 8° or less of each other. The variability in choice of UEV was low, 57% with no difference.
Specifying end vertebrae decreased the measurement variability within observer in Group A by 41 % from 4.4 to 2.6. For Group B films, the within observer variability decreased 35% from 4.8 with no instructions to 3.1 with the special measurement technique. The inter-observer variability was similar when the UEV was provided (average=3.1), indicating that the measurement technique was sufficient.
Within Group B, the chosen UEV with the special measurement technique was at a higher level (68% at T1 or T2) than when no information was provided (40% at T1 or T2; p=0.01). However, there was no difference in measured kyphosis.
Conclusion: A consistent technique in measurement of thoracic kyphosis improves precision and consistency within and across surgeons. The erroneous choice of UEV of the kyphotic curve underestimates the magnitude and length of the curve there by engendering superior junctional kyphosis.