We used survival analysis to evaluate 113 consecutive semiconstrained total elbow arthroplasties (TEAs) in 95 patients at a maximum follow-up of 99 months. Our criteria for failure were mechanical malfunction, revision for any reason, and deep infection. The primary diagnosis was inflammatory arthritis in 86 elbows, post-traumatic arthritis in 6, supracondylar nonunion or fracture in 12, osteoarthritis in 2 and other causes in 3. Seven failures were due to deep infection, and five of these had a primary diagnosis of inflammatory arthritis. Eight failures were revised or had revision recommended for aseptic loosening, and six of these were in patients with post-traumatic arthritis or supracondylar nonunion. The cumulative survival for TEAs performed for post-traumatic arthritis, fractures or supracondylar nonunion was 73% at three years and 53% at five years, significantly worse than the cumulative three- and five-year survivals of 92% and 90%, respectively, for patients with inflammatory arthritis. TEA with a semiconstrained prosthesis appears to have a satisfactory survival in selected patients with arthritic disorders. The incidence of deep infection was reduced by improvements in surgical technique and postoperative management, and the routine use of antibiotic-impregnated cement. The incidence of aseptic loosening was low, except in patients with supracondylar nonunion or post-traumatic arthritis.