Predicting cumulative risk in endosseous dental implant failure

Predicting cumulative risk in endosseous dental implant failure
August 2004
Donald Hui, DDS
J. Hodges
N. Sandler
Journal of Oral and Maxillofacial Surgery Online
Statement of the Problem: This paper attempts to answer the following problems: 1) Is dental endosseous implant failure cumulative? (ie, is a patient more likely to have an implant fail if they smoke AND are diabetic than if they were ONLY a smoker OR ONLY a diabetic, etc.) 2) Several studies have shown that a particular patient or device characteristic is a risk factor for implant failure; are these independent risk factors for implant failure? 3) Are the risk factors of equal importance in predicting implant failure? 4) How do the risk factors rank in relative importance?
Materials and Methods: This matched, retrospective case-control study began with a dataset including all implants placed at the University of Minnesota, School of Dentistry between July 1, 1993, and June 30, 2003. A total of 3083 root-form endosseous dental implants were placed in 1461 human patients. The major outcome variable in question was dental implant failure. Failure, as defined by the Dental Implant Clinical Research Group, 1997, was any implant requiring removal regardless of the reason. Implant failure patients whose charts a) were incomplete and missing data, b) represented implants that were placed by outside facilities, or c) were unavailable for review were excluded. Each patient with an implant failure was matched with up to three patients having no failures, according to these characteristics: 1) fiscal year of placement, 2) level of training of operator, 3) implant brand, and 4) anatomic position. When more than three control patients matched a given case, three were selected at random from the group meeting the matching criteria. Risk factors we chose to study were categorized as follows: demographic (age and gender), medical status (history of immunocompromise, compromised wound healing, medications, hormone replacement), habits (smoking, alcohol), anatomic variables (position, presence or absence of bone-grafting prior to implant placement), implant-specific variables (brand/make, diameter, coating/texture), failure mode, timing and level of training (faculty or resident). Collection of data is ongoing; this abstract presents results from analyzing a preliminary dataset consisting of 78 implant failures (cases) matched to a total of 156 controls. Twenty-three, 32, and 23 failures (cases) were matched with 1, 2, and 3 controls, respectively, for an average of exactly 2 controls per failure (case).
Method of Data Analysis: This preliminary analysis had three steps: 1) univariate analyses considering each potential risk factor separately; 2) multivariate analyses including potential risk factors with P < .10 in the first step; and 3) a final multivariate analysis including potential risk factors with P < .10 in the second step. At each step, the analysis used generalized estimating equations (GEE; SAS statistical package v.8, procedure GENMOD) with each cluster formed from a failure and its matches, using the independence working correlation and Type 3 score tests. Because this is a matched, case-control study, variation between clusters is not meaningful, so each predictor was entered as a 0/1 indicator variable corrected to capture only within-cluster variation, ie, variation from its cluster average.
Results: In the univariate analyses (Step 1), these potential risk factors had P < .10 for testing their association with case vs control: Oral surgeons vs periodontists (P = .0001, failure more likely with periodontists); age (P = .07, failure more likely with increasing age); maxilla vs mandible (P = .07, failure more likely in maxilla); posterior vs anterior (failure more likely in posterior, P = .002); single vs multiple sites (failure more likely if placed either in the anterior or posterior rather than both, P = .005); bone grafting (P = .06, failure more likely if any bone grafting done compared to none); manufacturer (Restore more than Sustain, P = .003; Sustain more than Micro-Vent/Screw-Vent, P = .016). In the univariate analyses, we considered two ways to characterize smoking: ever vs never, and currently vs not currently smoking. Neither was significantly related to case vs control (P = .88 and .69, respectively). After steps 2 and 3, the multivariate analyses, only these potential risk factors had P < .10: oral surgeons vs periodontists (P = .0498, failure more likely with periodontists); age (P = .063, failure more likely with increasing age); posterior vs anterior (failure more likely in posterior, P = .003); single vs multiple sites (failure more likely if placed either in the anterior or posterior rather than both, P = .062). It appears that more than one potential risk factors does, in fact, have independent power to predict implant failures, and that risk of failure is cumulative. Also, the final risk factors do differ in their apparent association with implant failure. Sorted in decreasing order of absolute log odds (given in parentheses), they are: oral surgeons vs periodontists (ÇƒÏ1.94); both anterior and posterior vs only one (ÇƒÏ1.61); anterior vs posterior (ÇƒÏ0.98); and age (0.029 per year increased age; 34 years added age is needed to match the risk of posterior vs anterior).
Conclusion: Many of our results are consistent with what has been reported regarding risk factors for implant failure. We have not yet been able to show a correlation between systemic factors such as diabetes, immunocompromise and postmenopausal osteoporosis and implant failure. It is surprising that we find no relationship between smoking and implant failure, especially in light of prospectively collected data such as NHANES data. Effective and reliable prediction of risk will greatly assist in proper case selection. This will directly impact the overall prognosis of implant survival. Additional prospective studies focusing on cumulative risk in dental endosseous implant failure will be necessary to further solidify our findings and to ultimately quantify cumulative risk.
References
Chuang SK, Wei LJ, Douglass CW, et al: Risk factors for dental implant failure: A strategy for the analysis of clustered failure-time observations. J Dent Res 81:572, 2002
Vehemente VA, Chuang SK, Daher S, et al: Risk factors affecting dental implant survival. J Oral Implantol 28:74, 2002
Ekfeldt A, et al: A retrospective analysis of factors associated with multiple implant failures in maxillae. Clin Oral Implants Res 12:462, 2001
Publishing and Reprint Information
515 Delaware Street SE, Suite 7, 174 Moos Tower, Minneapolis, MN 55455 USA
Copyright ¨© 2004 by Elsevier Inc.
doi: 10.1016/j.joms.2004.05.173