Background and purpose — We hypothesized that an ultra-short stem would load the proximal femur in a more physiological way and could therefore reduce the adaptive periprosthetic bone loss known as stress shielding.Patients and methods — 51 patients with primary hip osteoarthritis were randomized to total hip arthroplasty (THA) with either an ultra-short stem or a conventional tapered stem. The primary endpoint was change in periprosthetic bone mineral density (BMD), measured with dual-energy x-ray absorptiometry (DXA), in Gruen zones 1 and 7, two years after surgery. Secondary endpoints were change in periprosthetic BMD in the entire periprosthetic region, i.e. Gruen zones 1 through 7, stem migration measured with radiostereometric analysis (RSA), and function measured with self-administered functional scores.Results — The periprosthetic decrease in BMD was statistically significantly lower with the ultra-short stem. In Gruen zone 1, the mean difference was 18% (95% CI: −27% to −10%). In zone 7, the difference was 5% (CI: −12% to −3%) and for Gruen zones 1–7 the difference was also 5% (CI: −9% to −2%). During the first 6 weeks postoperatively, the ultra-short stems migrated 0.77 mm more on average than the conventional stems. 3 months after surgery, no further migration was seen. The functional scores improved during the study and were similar in the 2 groups.Interpretation — Up to 2 years after total hip arthroplasty, compared to the conventional tapered stem the ultra-short uncemented anatomical stem induced lower periprosthetic bone loss and had equally excellent stem fixation and clinical outcome.