Interobserver reliability of methods for paleopathological diagnosis of dental caries

There is disagreement as to whether clinical methods of caries diagnosis will produce reliable results in skeletal material, and studies vary greatly in the extent to which such methods are employed. The purpose of this study is, therefore, to evaluate the interobserver reliability of visual, tactile, and radiographic methods of scoring dental caries in ancient populations. Thirty‐three individuals from a Mogollon skeletal sample were scored on three occasions by pairs of examiners. The first exam employed visual criteria alone. In the second exam, visual methods were supplemented with the dental explorer. For the third exam, mesial and distal surfaces were scored from “intraoral” radiographs. Examiners worked independently, but did calibrate on material from clinical populations prior to the second and third exams. Exams were conducted on a “blind” basis, and analysis was delayed until all exams were completed. Caries scores were expressed as a percentage of total surfaces present per individual. Repeat‐measures ANOVA and intraclass correlation coefficients (ICC) were used to estimate interobserver reliabilities. For visual scores, ICC = 0.97, and there is no significant difference between examiners (P = 0.897). For visual plus explorer scores, ICC drops to 0.90, and there is a significant difference between examiners (P = 0.003). For the radiographs, ICC = 0.86, and there is no significant difference between observers (P = 0.117). These results suggest that clinical methods are less reliable in skeletal populations. Probable sources of error include postmortem dehydration of carious dentin, confusing radiolucencies engendered by worn or fractured enamel, and inadequate radiologic standards for root caries. These problems can be avoided through reliance on visual criteria.