ABSTRACT
AIM: The aim of this preliminary study was to test the reliability of radiographic evaluation of features of endodontic interest using a newly devised data collection system. METHODOLOGY: Twelve endodontic MSc postgraduate students and one specialist endodontist examined sample radiographs derived from a random selection of 42 patients seen previously on an Endodontic New Patient Clinic (EDI). Each student examined a random selection of 8-9 roots on periapical radiographs of single- and multirooted teeth, with and without previous root canal therapy and 3-4 dental panoramic tomograms (DPTs). A total of 100 roots were examined. A proforma was used to record observations on 67 radiographic features using predefined criteria. Intra-observer agreement was tested by asking the students to re-examine the radiographs. The principle investigator and the specialist endodontist examined the same radiographs and devised a Gold Standard using the same criteria. This was compared with the student assessments to determine inter-observer variation. The postgraduates then attended a revision session on the use of the form. Each student subsequently examined 8-9 different roots from the pool of radiographs. A further assessment of inter-observer variation was made by comparing these observations with the Gold Standard. RESULTS: Of the 67 radiographic features, only 25 had sufficient response to allow statistical analysis. Kappa values for intra- and inter-observer variation were estimated. These varied depending on the particular radiographic feature being assessed. Fifteen out of 25 intra-observer recordings showed 'good' or 'very good' Kappa agreement, but only three out of 25 inter-observer observations achieved 'good' or 'very good' values. Inter-observer variation was improved following the revision session with 16 out of 25 observations achieving 'good' or 'very good' Kappa agreement. CONCLUSIONS: Modification to the proforma, the criteria used, and training for radiographic assessment were considered necessary to improve the accuracy and reproducibility of the observations entered.
