To determine any significant differences in reproductive outcome from intracytoplasmic sperm injection (ICSI) with surgical sperm retrieval (SSR) between cycles using fresh and cryopreserved sperm, and between cycles using epididymal and testicular sperm.Retrospective national cohort study using data from the UK Human Fertilisation and Embryology Authority (HFEA), including all ICSI cycles performed in the UK over a 10-year period.All non-donor ICSI cycles between 2008 - 2017, categorised this by sperm source and cryopreservation status.SSR-ICSI using fresh or cryopreserved sperm, and using ejaculated, testicular and epididymal sperm.Live birth rate, pregnancy rate, implantation rate.We analysed data from all 214,649 ICSI cycles, including 199,818 cycles of ejaculated sperm, 5,646 of epididymal sperm and 9,185 of testicular sperm. Live births per ICSI cycle were 28.5% for ejaculated, 30.6% for epididymal, and 28.7% for testicular sperm cycles. Epididymal sperm cycles had a higher live birth rate than testicular sperm cycles (OR 1.067, 95% CI 1.014 - 1.123, p=0.013). This was despite a higher mean male age (42.5 vs 40.6 years, p0.001, 95%CI of difference 1.81-1.85 years) and female age (34.3 vs 34.0 years, p0.01, 95%CI of difference 0.32-0.34 years) in epididymal cycles compared to testicular cycles. Implantation rate (61.2% vs 58.0%, OR 1.086, 95%CI 1.041-1.133) and clinical pregnancy rate (34.3% vs 31.3%, OR 1.085, 95%CI 1.039-1.132) were also higher in epididymal compared to testicular cycles. There were no statistically significant differences in outcomes between cycles using fresh sperm compared to cryopreserved sperm for SSR-ICSI.Our study indicates that reproductive outcomes of SSR-ICSI are at least comparable to ICSI using ejaculated sperm, and does not support the preferential use of fresh sperm over cryopreserved sperm in SSR-ICSI. Births per SSR-ICSI cycle was higher for cycles using epididymal sperm compared to testicular sperm, however the differences were small, which may provide reassurance to patients undergoing these procedures. The results must be interpreted with caution as multivariable analysis was not possible due to aggregation of data.