Purpose: To report implant failure rates in a large group of partially edentulous patients over a long-term period and to analyze possible associations between available clinical variables and implant failure. Materials and Methods: Altogether, 2,453 patients were consecutively provided with 9,167 implants in 2,915 partially edentulous arches (Kennedy Class I or Class II) between 1986 and 2015 at one referral clinic. All implant failures recorded at the clinic during follow-up were identified and analyzed in relation to available clinical variables by means of a multivariable regression analysis. Results: Overall cumulative survival rates (CSRs) for treated arches without any implant failures after 15 and 25 years were 91.7% and 90.3%, respectively, and the corresponding CSRs for implants were 96.1% and 95.3%, respectively. Out of the 209 arches with reported implant failures, 112 were reported with failures during the first year (54%). The strongest associations with risk for overall implant failure were related to two different surgeons (hazard ratio 6.14 [2.77 to 13.60] and 1.51 [1.14-1.99]). Number of implants, time since surgery, calendar year of surgery, and age at surgery were also significantly associated with overall implant failure. Furthermore, early implant failures were associated with implant surface (HR 2.27 [1.43 to 3.57]), indicating higher risk for implants with a turned surface. This variable was not observed for overall or late implant failures. Conclusion: Individual surgeons seem to have an important impact on risk for implant failure in combination with implant surface and number of placed implants during the early period after surgery. Besides the surgeon and number of implants, the overall trend seems to be a decreasing risk for failure the longer the patient is followed up, the later the patient is included, and the older the patient is at treatment.