We studied, ten patients (11 elbows) who had undergone 14 allograft-prosthesis composite reconstructions following failure of a previous total elbow replacement with massive structural bone loss. There were nine women and one man with a mean age of 64 years (40 to 84), who were reviewed at a mean of 75 months (24 to 213). One patient developed a deep infection after 26 months and had the allograft-prosthesis composite removed, and two patients had mild pain. The median flexion-extension arc was 100° (95% confidence interval (CI) 76° to 124°). With the exception of the patient who had the infected failure, all the patients could use their elbows comfortably without splints or braces for activities of daily living. The mean Mayo Elbow Performance Index improved from 9.5 (95% CI 4.4 to 14.7) pre-operatively to 74 (95% CI 62.4 to 84.9) at final review. Radiologically, the rate of partial resorption was similar in the humeral and ulnar allografts (three of six and four of eight, respectively; p > 0.999). The patterns of resorption, however, were different. Union at the host-bone-allograft junction was also different between the humeral and ulnar allografts (one of six and seven of eight showing union, respectively; p = 0.03). At medium-term follow-up, allograft-prosthesis composite reconstruction appears to be a useful salvage technique for failed elbow replacements with massive bone loss. The effects of allograft resorption and host-bone-allograft junctional union on the longevity of allograft-prosthesis composite reconstruction, however, remain unknown, and it is our view that these patients should remain under long-term regular review.