Objective To investigate the effectiveness and surgical skills of microsurgical repair of radial nerve deep branch injury. Methods Between March 2001 and February 2011, 49 cases of radial nerve deep branch injury were treated by microsurgical technique. There were 40 males and 9 females with an average age of 32 years (range, 19-58 years), including 13 cases of knife-cut injury, 9 cases of electric-saw injury, 7 cases of dagger-stab injury, 6 cases of glass-cut injury, 5 cases of iatrogenic injury, 4 cases of Monteggia fracture, 3 cases of nailgun injury, and 2 cases of crush injury of the forearm complicated by fracture of the proximal radius. The disease duration ranged from 3 hours to 3 years and 8 months (mean, 4.9 months). The sites of injury were at front of supinator tube in 15 cases, in the supinator tube in 23 cases, and at back of supinator tube in 11 cases. One-stage repair was performed by end-to-end suture in 21 cases, including 9 cases of epineurial neurorrhaphy and 12 cases of perineurial neurorrhaphy; two-stage repair was performed in 28 cases, including 26 cases of sural nerve graft and 2 cases of neurolysis. Results Postoperative wounds primarily healed. All patients were followed up 21.5 months on average (range, 12-39 months). At last follow-up, in 21 cases of one-stage repair, the muscle strength of the extensor pollicis longus was level 5 in 13 cases, and level 4 in 8 cases; in 28 cases of two-stage repair, the muscle strength of the extensor pollicis longus was level 5 in 2 cases, level 4 in 21 cases, level 3 in 4 cases, and level 2 in 1 case; and significant difference was found (Z= — 5.340, P=0.000). In 9 cases undergoing epineurial neurorrhaphy at one-stage repair, the muscle strength of the extensor pollicis longus was level 5 in 3 cases, and level 4 in 6 cases; in 12 cases undergoing perineurial neurorrhaphy at one-stage repair, the muscle strength of the extensor pollicis longus was level 5 in 10 cases, and level 4 in 2 cases; and significant difference was found (Z= — 2.279, P=0.023). In 26 cases undergoing nerve graft at two-stage repair, the muscle strength of the extensor pollicis longus was level 5 in 2 cases, level 4 in 20 cases, level 3 in 3 cases, and level 2 in 1 case; in 2 cases undergoing neurolysis at two-stage repair, the muscle strength of the extensor pollicis longus was level 4 in 1 case and level 3 in 1 case; and no significant difference was found (Z= — 1.117, P=0.264). According to the upper arm function assessment criterion issued by Hand Surgery Association of Chinese Medicine Association, the results were excellent in 18 cases, good in 3 cases in one-stage repair patients; excellent in 2 cases, good in 21 cases, fair in 4 cases, and poor in 1 case in two-stage repair patients; and there was significant difference (Z= — 5.340, P=0.000). Conclusion Microsurgical one-stage repair of radial nerve deep branch injury can obtain better effectiveness than two-stage repair by nerve graft, and perineurial neurorrhaphy is significantly better than epineurial neurorrhaphy.