Objective: To evaluate the short- and long-term effectiveness of exercise training in relation to pain, function and direct costs to health care systems attributable to hip osteoarthritis. Design: Prospective, two-year randomized controlled trial. Setting: An outpatient primary health care setting. Subjects: One hundred and twenty men and women aged from 55 to 80, with radiologically diagnosed hip osteoarthritis with associated clinical symptoms. Interventions: The combined exercise and general practitioner (GP) care group received 12 supervised (once per week) exercise sessions at the baseline and four additional booster sessions one year later. Both groups received standard GP care. Main measures: The WOMAC Osteoarthritis Index, physical functioning score of RAND-36 (SF-36), the use and health care system costs of doctor visits and physiotherapy associated with hip osteoarthritis, number of total hip replacements, the use of analgesic and non-steroidal anti-inflammatory drugs (NSAIDs), performance-based outcome scores and body mass index (BMI). Results: There were no differences between the groups with respect to WOMAC hip pain, physical functioning score of RAND-36, performance-based outcome scores or BMI. The effect of the exercise intervention on WOMAC function was statistically significant at 6 months (mean = −7.5; 95% confidence interval (CI) −13.9 to −1.0; P = 0.02) and 18 months (mean = −7.9; 95% CI −15.3 to −0.4; P = 0.04). There were no statistically significant differences in the total health care system costs between the groups. Conclusion: The mostly home-based exercise training programme provided in this study did not result in reduced hip pain over the two-year follow-up period.