Twenty-two elderly patients (average age, seventy-two years) who had an atrophic, unstable, ununited fracture of the humeral diaphysis were managed with plate-and-screw fixation and application of an autogenous bone graft from the iliac crest. Fifteen of the patients had had at least one previous operation in an attempt to obtain union of the fracture. One patient had an active infection and two had a quiescent infection, all with Staphylococcus epidermidis. The average duration of nonunion before the patients were first seen by us was two years and four months (range, five months to sixteen years). Fifteen of the nonunions were synovial. In each patient, at least one modification of the standard technique of plate-and-screw fixation was needed as a result of osteopenia. In order to enhance fixation, the standard protocol incorporated the use of a long plate (with an average of eleven holes and an average length that was 76 percent of that of the bone), a plate with a blade (used in thirteen patients), and replacement of loose, 4.5-millimeter cortical-bone screws with 6.5-millimeter cancellous-bone screws (twelve patients). Spiked nuts (Schuhli nut; Synthes, Paoli, Pennsylvania) that lock the screws to the plate, creating a solid point of fixation analogous to a blade, were incorporated into the protocol when they became available (used in six patients). In five limbs, the nonunion was associated with an osseous defect that could not be addressed by shortening of the bone alone. Three of these limbs were stabilized with a bridge plate that had been contoured to stand away from the bone at the site of nonunion (so-called wave-plate osteosynthesis), and the remaining two limbs were stabilized with a combination of intramedullary and extramedullary plates. In one of these two limbs, the extramedullary plate was contoured (that is, a wave plate).The fracture united in twenty (91 percent) of the patients. There was no progressive loosening or breakage of a fixation device, even in two patients who had radiographs that were suggestive of an incomplete union. Five of the patients were followed for a limited duration (average, one year and six months) as a result of death or illness. They had two excellent results, two good results, and one poor result according to a modification of the rating system of Constant and Murley. The remaining seventeen patients, including the two who had a persistent nonunion, were followed for an average of three years and one month (range, two years to five years and ten months). They had significant improvements in all of the functional scores at the most recent follow-up evaluation: the average score according to the modified system of Constant and Murley increased from 9 to 72 points (p < 0.001), the average score according to the Enforced Social Dependency Scale decreased from 39 to 9 points (p < 0.001), and the average score based on the Disabilities of the Arm, Shoulder, and Hand Questionnaire decreased from 77 to 24 points (p < 0.001). According to the scores based on the Disabilities of the Arm, Shoulder, and Hand Questionnaire, nine of the seventeen patients who had been followed for more than two years had an excellent result, four had a good result, two had a fair result, and the two who had a persistent nonunion had a poor result. Complications included postoperative delirium, a stitch abscess, transient radial nerve palsy, a fracture distal to the plate, and the need for a blood transfusion, in one patient each. Two patients had a fibrous union. There were no major medical complications.An unstable, united fracture of the humeral diaphysis can be extremely disabling and may threaten the ability of an elderly patient to function independently. Operative treatment can be very successful when the techniques of plate-and-screw fixation are modified to address osteopenia and relative or absolute loss of bone. Healing of the fracture substantially improves function and the degree of independence.