Morrissy and colleagues measured how much the Cobb angle varies when one examiner, or different examiners, measure the same scoliosis radiograph, and what causes the error.
Four examiners each measured 48 radiographs of 20° to 40° curves six times, in three sets that progressively fixed the end vertebrae and then the protractor.
Key Findings
Repeat readings by the same examiner differed by more than 4.9° only 5% of the time when he chose the end vertebrae. With end vertebrae fixed this fell to 3.8°, and with one protractor to 2.8°.
Between examiners, the 95% limit was 7.2° when each chose end vertebrae and 6.3° when they were fixed, with or without a shared protractor. The largest difference was 14°.
Choice of end vertebrae was the largest source of intraobserver error. For one examiner, repeat readings within 3° rose from 75% to 98% once end vertebrae were fixed.
Mean angles differed systematically between examiners (p < 0.0001), likely due mainly to protractors using the cut edge as baseline, in the authors' view. With one shared protractor the means were closer but still differed (p < 0.001).
Every pair of readings that differed by more than 6° between examiners was a thoracolumbar or lumbar curve with an indistinct, cupped or biconcave end plate.
Experience did not predict precision. Examiners saw their results between sets and competed, so the authors expect clinical practice to be less precise, and positioning error was not studied.
Board pearl
A small Cobb change may be measurement error alone. The usual 5° progression criterion may not be adequate. Even with end vertebrae fixed, 5% of readings by different examiners differed by more than 6.3°.
Authors' conclusion
All of these findings emphasize the need to note the end-vertebrae of the curve accurately, to measure previous radiographs again if the end-vertebrae appear to have changed, and to have one person measure all of a patient’s radiographs whenever possible.
Morrissy and colleagues measured how much the Cobb angle varies when one examiner, or different examiners, measure the same scoliosis radiograph, and what causes the error.
Four examiners each measured 48 radiographs of 20° to 40° curves six times, in three sets that progressively fixed the end vertebrae and then the protractor.
Key Findings
Repeat readings by the same examiner differed by more than 4.9° only 5% of the time when he chose the end vertebrae. With end vertebrae fixed this fell to 3.8°, and with one protractor to 2.8°.
Between examiners, the 95% limit was 7.2° when each chose end vertebrae and 6.3° when they were fixed, with or without a shared protractor. The largest difference was 14°.
Choice of end vertebrae was the largest source of intraobserver error. For one examiner, repeat readings within 3° rose from 75% to 98% once end vertebrae were fixed.
Mean angles differed systematically between examiners (p < 0.0001), likely due mainly to protractors using the cut edge as baseline, in the authors' view. With one shared protractor the means were closer but still differed (p < 0.001).
Every pair of readings that differed by more than 6° between examiners was a thoracolumbar or lumbar curve with an indistinct, cupped or biconcave end plate.
Experience did not predict precision. Examiners saw their results between sets and competed, so the authors expect clinical practice to be less precise, and positioning error was not studied.
Board pearl
A small Cobb change may be measurement error alone. The usual 5° progression criterion may not be adequate. Even with end vertebrae fixed, 5% of readings by different examiners differed by more than 6.3°.
Authors' conclusion
All of these findings emphasize the need to note the end-vertebrae of the curve accurately, to measure previous radiographs again if the end-vertebrae appear to have changed, and to have one person measure all of a patient’s radiographs whenever possible.