The purpose of this study was to compare the outcome, complications and survival of the three most commonly used surgical reconstructions of the proximal humerus after transarticular tumour resection. Between 1985 and 2005, 38 consecutive proximal humeral reconstructions using allograft-prosthesis composite (n = 10), osteoarticular allograft (n = 13) or a modular tumour prosthesis (n = 14) were performed in our clinic. The mean follow-up was ten years (1-25). Of these, 27 were disease free at latest follow-up (mean 16.8 years) and ten had died of disease. The endoprosthetic group presented the smallest complication rate of 21% (n =1), compared to 40% (n = 4) in the allograft-prosthesis composite and 62% (n = 8) in the osteoarticular allograft group. Only one revision was performed in the endoprosthetic group, in a case of shoulder instability. Infection after revision (n = 3), pseudoarthrosis (n = 2), fracture of the allograft (n = 3) and shoulder instability (n = 4) were the major complications of allograft use in general. Kaplan-Meier analysis showed a significantly better implant survival for the endoprosthetic group (log-rank p = 0.002). At final follow-up the Musculoskeletal Tumour Society scores were an average of 72% for the allograft-prosthetic composite (n = 7, median follow-up 17 years), 76% for the osteoarticular allograft (n = 3, 19 years) and 77% for the endoprosthetic reconstruction (n = 10, 5 years) groups. An endoprosthetic reconstruction after transarticular proximal humeral resection resulted in the lowest complication rate, highest implant survival and comparable functional results when compared to allograft-prosthesis composite and osteoarticular allograft use. We believe that the surgical approach that best preserves the abductor mechanism and provides sufficient surgical exposure for tumour resection contributed to better functional results and glenohumeral stability in the endoprosthetic group.