A clinically useful measurement must be reproducible. Drs McNamara, Baccetti, and Franchi claim that the CVM method has high reproducibility. They claim interoperator (between each other) agreement of 98.6% and intraoperator (between themselves) agreement of 100%.
The results of our study demonstrate that the CVM method has poor reproducibility. On average, CVM staging agreement among seasoned practicing orthodontists was below 50%, and the orthodontists agreed with their own CVM staging only 62% of the time. In a study at Ohio State University, the residents agreed with themselves 63% of the time for the CVM method.
In a follow-up study, we discovered why the CVM method has such poor reproducibility. The weakness arises, in part, from difficulty in classifying the vertebral bodies of C3 and C4 as trapezoidal, rectangular horizontal, square, or rectangular vertical.
Drs McNamara, Baccetti, and Franchi criticized our teaching of the CVM method to orthodontists because their diagram “never was proposed by the original authors as a guideline for the implementation of the CVM method in a clinical setting.” We trained our orthodontists using their method in a research setting, not a clinical setting. Our observers were experienced orthodontists. They received a cover letter explaining the procedure, and the exact reference material, diagrams, and descriptors accompanying the diagrams. They were allowed to use these reference materials freely during the judging with no time limit.
Our orthodontists looked at exactly the same radiographic images twice and used exactly the same logic to stage the vertebrae both times. Plainly speaking, if the CVM method works, then our orthodontists should have easily staged the images the same both times. They did not.
Instead, we discovered a key point: a few of our randomly chosen subjects were easily staged repeatedly, but most were not. This implies that, if researchers dramatically reduce their subject sample size from a large sample to a much smaller sample, then they run the risk of selecting for the more easily staged radiographs. This could be a major source of error for Drs McNamara, Baccetti, and Franchi, who reduced their sample size dramatically.
Additionally, the dentitions could give an observer clues as to the CVM stage. For this reason, we masked the dentitions in our sample. Drs McNamara, Baccetti, and Franchi did not.
Does the CVM method have merit? Some morphologic changes do occur with maturation. However, reliable and usable assessments must be reproducible in all subjects in a random population. Our study demonstrates that this is not true for the CVM method.
For 40 years, various methods of radiographic interpretation has surfaced purporting to accurately predict individual jaw growth. None has survived.