Mixed-effects Frailty Failure Time Methods to Estimate Dental Implant Survival

Mixed-effects Frailty Failure Time Methods to Estimate Dental Implant Survival
S.-K. CHUANG1, L.-J. WEI2, and T.B. DODSON1, 1 Massachusetts General Hospital and Harvard School of Dental Medicine, Boston, USA, 2 Harvard School of Public Health, Boston, MA, USA
2003
IADR
Purpose: The purpose of this study was to identify covariate effects associated with implant failure by applying clustered semi-parametric Cox proportional hazards frailty survival methods. To our knowledge, this method has not been described or applied widely in the dental research literature. Material and Methods: To address the research purpose, we used a retrospective cohort study design. The cohort was composed of patients having one or more Bicon implant(s) placed. Covariates were categorized as demographic, health status, implant-specific, anatomic, prosthetic, perioperative, operative, and reconstructive variables. The outcome variable was implant failure (explantation). Covariates for implant failure were identified using the frailty survival methods adjusted for clustered failure-time observations. Results: The sample was composed of 677 patients having 2349 implants placed. Covariates associated with implant failure (p < 0.15) included operating surgeon, tobacco use, peri-operative antibiotic use, implant position, implant length, well size, coating of implant, proximity of the implant to adjacent teeth or other implants, immediate implant placement, abutment diameter, prosthetic type, usage of reconstruction procedures, and implant staging. Based on the adjusted multivariate frailty model, covariates associated with implant failure were tobacco use, implant length and staging, proximity of the implant to adjacent teeth or other implants, and well size. Conclusions: Datasets composed of clustered observations are commonly encountered in dental research. Survival analyses of such datasets are exceedingly challenging propositions. We presented an innovative Cox proportional hazards frailty survival methods with clinical applications to implants as an example. We identified five factors associated with implant failure. Three of these variables, smoking status, well size, and staging of implant placement are under the direct control of the practitioner. Supported by Oral and Maxillofacial Surgery Research Foundation Fellowship in Clinical Investigation (S-KC), NIH grant K24 DE000448 (TBD) and MGH Department of OMS Research Fund (S-KC, TBD). E-mail: schuang@hsph.harvard.edu