Can orthodontists reliably identify incidental findings on panoramic radiographs?

We routinely use panoramic radiographs for orthodontic diagnosis and treatment planning. These images may reveal incidental findings (IFs) that are clinically significant. Yet, how effective are we at detecting these incidental findings?

We define Incidental findings (IFs) as any discoveries unrelated to the scan’s original purpose. We must recognise these findings and take any necessary actions. Moreover, one reason for conducting a radiographic scan is to discover incidental findings before initiating orthodontic treatment. In this respect, this study is clinically meaningful.

This new study looked at the ability of orthodontists to recognize IFs. 

A team from Toronto, Canada did this study. The excellent Orthodontics and Craniofacial Research published the paper.

What did they ask?

They did this study to

“Evaluate the ability of orthodontists to detect, interpret and recommend appropriate referrals for the management of incidental findings compared to an Oral Maxillofacial radiologist”.

What did they do?

They did a cross-sectional study. This has the following stages.

One researcher carried out a retrospective review of 1756 panoramic radiographic reports at the Faculty of Dentistry, University of Toronto. These reports were from 2014 to 2022 and involved patients aged 7 to 21. From these reports, they identified incidental findings. The researcher categorised these findings based on the anatomical regions in which they were located. 

They then assessed the level of risk associated with each incidental finding, using a risk assessment tool that they developed. This classified the findings into three risk categories: high, moderate, and low or absent. 

From a total of 1,756 panoramic radiographs, the researcher selected 12 images for further analysis. These radiographs included two high-risk incidental findings, two moderate-risk incidental findings and two low-risk incidental findings. Additionally, two radiographs showed no incidental findings.

They then asked orthodontists licensed to practice in Canada to join in an online survey.

The orthodontists who agreed to take part were asked to select the location of any IF, describe their internal structure, show whether they would refer the patient before starting treatment, and state whether an urgent referral would be required. They also selected a diagnostic category and individual radiological interpretation.

Finally, the authors compared the results for the orthodontists with those of an experienced oral maxillofacial radiologist.

They analysed the data using chi-squared statistics to evaluate relationships between the patients’ ages and the number of IFs. They then used Cohen’s Kappa to evaluate the interobserver agreement between the orthodontists and the oromaxillofacial radiologist.

What did they find?

Approximately 30% of radiographs had at least one incidental finding, ranging from zero to four per patient. The most common findings were impacted teeth (17.7%), followed by hypodontia (14.9%) and dense bone Islands (14.7%).

35% of the incidental findings were high risk.

Thirty out of seventy-five orthodontists completed the survey. The mean years of professional experience was 19.8 years. 

When examining the agreement on all the questions between the orthodontists and the radiologist, the result was only fair (kappa = 0.32). The level of agreement for location, internal structures and diagnosis was moderate, for specialist referral and diagnostic category it was fair and for need for referral and urgency of referral it was slight. Importantly, none of these levels of agreement was high.

My interpretation of the conclusions was

“Incidental findings are quite common in panoramic radiographs taken for orthodontic purposes.  These results support the need to strengthen the curriculum on radiologic interpretation and management of IFs during dental and orthodontic speciality training programs”.

What did I think?

This study raises several questions. First, it is significant that the radiologist and orthodontists had limited agreement on incidental findings. This indicates a need to enhance our radiological interpretive skills. Improving these skills is especially important when justifying radiation exposure in the search for incidental findings.

The approach they used in the study was established and comparable to other research evaluating diagnostic skills and radiograph interpretation. The authors made an effort to assess the radiographs in a manner that closely resembled a clinical situation.

However, I did have some concerns about the study. Firstly, they state that the percentage of IF was 30%. However, 32% of the IFs were impactions (17.7%), hypodontia (14.9%). We can presume that the orthodontic radiographs were taken to detect these problems. I not feel that these are IFs. As a result, the authors have over reported the number of IFs.

I also wanted to see some information on the number and type of IFs that were not reported by the orthodontists compared to the radiologist. Unfortunately, this information was not available. I appreciate that the team reported the levels of agreement between the two sets of viewers. Unfortunately, this was not based on the simple question of whether an IF was detected (yes/no). It was defined by calculating agreement of two operators over 14 decisions. In this respect, I felt that the moderate level of agreement they reported represented a good performance.

My overall feeling was that there was a lot of “white noise” in the data and I struggled to come to firm conclusions.

The authors highlighted several shortcomings in their method. Firstly, the study was not carried out in a clinical setting, which may influence the applicability of the results to real-world clinical environments. Secondly, there is a potential sampling bias among the participants, as most were instructors at university clinics, which limits the study’s generalisability. Finally, there are concerns about the standardisation of the images used in the study, as different machines with varying calibration settings were employed.

Final thoughts

Finally, we need to consider how these limitations might affect the findings. I am still uncertain about my interpretation of this paper, unless I have missed something. The most significant limitations are the low response rate, the characteristics of the orthodontists involved, the structure of the survey and the interpretation of the data.