Two problems impede the study of age dependency of tooth loss rate using dental radiographs. One is that the exact time of tooth loss is often unknown. The other is that major causes of tooth loss are assumed to result from tooth extraction by dentists. Recently dentists have avoided performing tooth extraction as summarily as in years past due to developments in dental technique and increased concern on oral health. The effect of these changes in dental practice cannot be ignored on the estimation of tooth loss probability. We devised a modification of survival analysis taking into account these two problems by using data consisting of repeated observations for each individual. A flexible statistical model allowing for individual variation in tooth loss hazard was adopted. The proposed method was shown to be valid by a Monte Carlo simulation study in which the true parameters describing age-dependency of tooth loss probability were given. We illustrated our method by applying it to actual tooth loss data from 1,284 panoramic radiographs from 642 patients obtained in a daily clinical practice where each patient underwent panoramic examination at least twice. The age-dependency of tooth loss probability for all kinds of teeth were estimated. The smaller variance of tooth loss hazard between individuals is observed on the molars compared with those of other kinds of teeth.
1999 The Biometric Society of Japan