Objectives: In most retrospective studies, the clinical performance of restorations had not been considered in survival analysis. This study investigated the effect of including the clinically unacceptable cases according to modified United States Public Health Service (USPHS) criteria into the failed data on the survival analysis of direct restorations as to the longevity and prognostic variables. Materials and Methods: Nine hundred and sixty-seven direct restorations were evaluated. The data of 204 retreated restorations were collected from the records, and clinical performance of 763 restorations in function was evaluated according to modified USPHS criteria by two observers. The longevity and prognostic variables of the restorations were compared with a factor of involving clinically unacceptable cases into the failures using Kaplan-Meier survival analysis and Cox proportional hazard model. Results: The median survival times of amalgam, composite resin and glass ionomer were 11.8, 11.0 and 6.8 years, respectively. Glass ionomer showed significantly lower longevity than composite resin and amalgam. When clinically unacceptable restorations were included into the failure, the median survival times of them decreased to 8.9, 9.7 and 6.4 years, respectively. Conclusions: After considering the clinical performance, composite resin was the only material that showed a difference in the longevity (p < 0.05) and the significantly higher relative risk of student group than professor group disappeared in operator groups. Even in the design of retrospective study, clinical evaluation needs to be included.