Radiographic Assessment of Alveolar Bone Loss

Radiographic and clinical assessments remain the standard diagnostic parameters used to detect and monitor the progression of periodontal diseases (PD). Objectives: This study examined the variability among observers in the radiographic interpretation of alveolar bone loss (ABL) and compared radiographic evaluations to clinical diagnoses of PD and computer assessments of ABL. Methods: The study population (19 females, 11 males; aged 21-67) had at least 5 teeth in each sextant. Based on clinical examinations, patients were classified: healthy (n=4); mild periodontitis (n=6); moderate periodontitis (n=13); and severe periodontitis (n=7). Complete radiographic series (16 periapical and 4 interproximal) were evaluated at two week intervals by 4 radiologists who were blinded with respect to clinical findings. Radiologists classified ABL as follows: None; mild <12%; moderate 12-40%; severe >40%. The computer algorithm calculated ABL based on interactive identification of CEJ, apex and alveolar crest on the mesial and distal aspect of each tooth. Results: Using weighted Kappa, moderate intra-observer agreement (0.52 - 0.80) and weaker inter-observer agreement (0.23 - 0.70) were found for radiographic interpretations. Radiographic assessments of ABL were moderately correlated to clinical diagnoses (Spearman r=0.41 - 0.56). The strength of the association was determined using the Goodman-Kruskal gamma statistic (0.53 p=0.014; 0.72 p<0.000;0.73 p<0.000;0.70 p<0.000). A similar strong association of computerized assessments of ABL and clinical diagnoses were found, r=0.62 gamma=0.72 p<0.000. Spearman correlations between radiologists and computerized assessments of ABL were 0.52; 0.70; 0.73; 0.75. Conclusions: Although radiographic assessments of alveolar bone levels are highly correlated with clinical diagnoses, there is wide variation in inter-observer agreement. Computerized assessments offer promise for minimizing inter-observer variation in interpretation of alveolar bone levels.