Background: Late relapses (LR; relapse occurring after disease-free interval of two years) in testicular cancer are relatively uncommon. The limited data published is hampered by selection bias and incomplete data regarding follow-up. We aimed to investigate the frequency of LR and of very late relapses (VLR; relapse occurring after disease-free interval of five years), and the survival of patients with VLR in a cohort treated after 1995, compared to a cohort before 1995. Finally, we aimed to describe the number of missed relapses beyond a 5-year follow-up scheme. Methods: A total of 5712 patients, 2978 seminoma and 2734 nonseminoma, were diagnosed with testicular cancer in Norway, with 2207 patients diagnosed 1980-1995 and 5712 patients diagnosed 1995-2009. Data are complete, due to identification by the Cancer Registry of Norway and Norwegian Cause of Death Registry. Details regarding diagnosis, stage, treatment, and follow-up were obtained from medical records. Relapse rates have been estimated using Kaplan-Meyer. Results: A total of 472 patients experienced relapse, 186 seminoma and 286 nonseminoma. Of these 109 were LR (51 seminomas, 58 nonseminomas), 50 VLR (22 seminomas, 28 nonseminomas with 17 relapses (4 seminomas, 13 nonseminomas) beyond ten years. The median time to LR was 4.7 years (range 2.0-21.6). In clinical stage I patients, there were 306 relapses (7.9%); with overall 1.9% LR, 1.0% VLR, and 0.5% relapse beyond ten years. Patients followed with surveillance had a higher rate of LR compared to patients receiving adjuvant therapy (4.0% vs 1.0%). Three patients with LR died of testicular cancer, all three had a VLR and initial nonseminoma. In patients with metastatic disease, 166 patients experienced relapse (10.5%); with overall 3.6% LR 1.6% VLR, and 0,8% relapse beyond ten years. In nonseminoma diagnosed after 1995, the rate of VLR was 0.8% compared to 2.3% in the earlier cohort. Eight patients with VLR died of testicular cancer, all with initial nonseminoma, and seven of these were diagnosed in the earlier cohort. Outside a five-year follow-up scheme, 50 patients would be diagnosed with a VLR. Prolonging follow-up to ten years would potentially identify only 33 of these relapses. Conclusions: In this first population-based series investigating late relapses of testicular cancer with complete data regarding treatment and follow-up, we find a low rate of VLR in patients treated according to modern guidelines. We believe centralization of treatment, adherence to guidelines, prospective registration of patients, and subsequent reporting of results are key to these improved results. Patients with CS I followed by surveillance will have higher rate of VLR compared to patients receiving adjuvant treatment, resulting in a higher VLR rate in the modern cohort, without affecting survival. In metastatic disease, the VLR rate was drastically reduced in the modern cohort, with a subsequent improved survival.