The study's purposes were to estimate dental implant survival in a statistically valid manner and to compare three models for estimating survival. We estimated survival using three different statistical models: (1) randomly selecting one implant per patient; (2) utilizing all implants, assuming independence among implants from the same subject; and (3) utilizing all implants, assuming dependence among implants from the same subject. The cohort was composed of 660 patients who had 2286 implants placed. Due to the high success rates of implants, the five-year survival point and standard error estimates varied little among the three models. Patients at high risk for implant failure (smokers) manifested greater variation in the standard error estimates among the three models, 8.2%, 4.0%, and 5.6%, respectively. To obtain statistically valid survival confidence intervals when performing Kaplan-Meier survival analyses, we recommend adjusting for dependence when there are multiple observations within the same subject.