We are grateful to Dr Perinetti for his interest and thoughtful comments in reference to our article and are delighted that he credits us for having implemented “elegant statistical models” in an overall “well-done study” with “robust results.” However, Dr Perinetti expressed concern as to whether any conclusions on the classic cervical vertebral maturation (CVM) method could be derived from our results, owing to the different methodologic approach.

We consider his comment helpful, because it broadens the discussion on the proper statistical approach to be used to validate the CVM method. In recent times, the CVM method has been labeled as questionable in a series of articles reporting poor reproducibility and reliability. As early as 2010, the AJO-DO offered a forum for authors to reply and respond to various claims concerning the reproducibility and repeatability of the CVM method. As mentioned in our article, the overall poor result of the CVM method is commonly blamed on the difficulty to classify and stage morphologic changes of the cervical vertebrae.

In statistical terms, the CVM method merges continuous changes of the vertebrae into grouped data. Yet, it is a fact that converting continuous data into categorical variables causes loss of information, power, and efficiency, and, especially, categorizing inherently continuous data is highly contested.

Hence, we did not intend to mimic the CVM method but, rather, envisioned an alternative approach to circumvent a major drawback of the CVM method. The continuous changes in the vertebrae were recorded and analyzed as continuous values. Although the choice over which variables to include in the regression analysis might be subject to debate, our results indicate conclusively that age-dependent changes in vertebral morphology are inherently insufficient for accurate age estimation.

We concur with Dr Perinetti’s opinion that our results should only be applied to the classic CVM “with caution.” Yet, it would be contrary to statistical evidence to expect the data to be more accurate by staging the results with the CVM method.