Reliability of centroid, Cobb, and Harrison posterior tangent methods: which to choose for analysis of thoracic kyphosis

Thirty lateral thoracic radiographs were digitized twice by each of the three examiners.

To determine the reliability of the centroid, Cobb, and Harrison posterior tangent methods when applied to analysis of thoracic kyphosis.

Reliability studies on measurements of thoracic kyphosis are rare.

Blind, repeated-measures design was used. Thirty lateral thoracic radiographs were digitized twice by each of three examiners. To evaluate reliability of determining global and segmental alignment, vertebral bodies of T1-T12 were digitized. Centroids at the intersection of vertebral body diagonals and tangents to posterior vertebral bodies were constructed by computer. Also the computer constructed global and segmental centroid angles, Cobb angles (two-line method), and posterior tangent intersection angles from T1 to T12. Interclass and Intraclass correlation coefficients for these data were calculated and interpreted.

From the points selected by examiners, all three methods have similar high ICC values for the global angles (> 0.94). For the segmental angles, the interobserver and intraobserver reliability is also very similar for all three methods, with ICCs in the good and excellent ranges (0.59-0.75 and 0.75-1.0, respectively). The mean absolute differences of observers' measurements are low, similar, and in the range of 0.9 degrees to 2.5 degrees.

The centroid, two-line Cobb, and Harrison posterior tangent methods, when applied to measurements of kyphosis, are all reliable and have similar small error ranges. The centroid method does not give an accurate segmental analysis, uses more points and more time in clinical applications, and results in smaller angles of total kyphosis than the Cobb or posterior tangent methods. The posterior tangents are the slopes along the curve.