ABSTRACT
STUDY DESIGN: Repeat clinical radiograph measurement of kyphosis was performed. OBJECTIVES: To evaluate precision in kyphosis measurement, the presence of intra- and interobserver error, and the effect of endplate selection and curve magnitude on error. SUMMARY OF BACKGROUND DATA: One study thus far has shown no interobserver difference in readings and an error interval of +/-11 degrees. METHODS: Four experienced examiners measured 30 radiographs of varying angles of kyphosis without preselected end vertebrae twice using the Cobb method. RESULTS: The mean intraobserver variance was 4.3 degrees (95% confidence interval, +/-9.6 degrees). One examiner had significantly better precision (P= 0.02) than the other examiners, who had no significant difference among them (P = 0.41). This examiner's mean intraobserver difference was 2.3 degrees (95% confidence limit, +/-6.2 degrees ). The 95% confidence limit for the interobserver difference was +/-8.7 degrees. The vertebral error index did not have a rank correlation with precision between readings. Magnitude of curve did not correlate with variance in measurement. CONCLUSIONS: The broad range in intra- and interobserver differences in the measurement of kyphosis should be taken into account in making management decisions or evaluating the success or failure of a treatment program. Careful technique in measurement may allow for improvement in individual precision.