OBJECTIVE To assess the accuracy of preoperative localisation of abnormal parathyroid glands in patients with primary hyperparathyroidism by comparing results of echography and 201thallium and (99m)technetium subtraction scans with the operative findings. METHODS Retrospective study. METHODS Teaching hospital, Taiwan. METHODS 84 Patients with primary hyperparathyroidism, 83 of whom had bilateral exploration of the neck and thymus and one mediastinotomy. Seventy-three patients (69 with adenomas and 4 with hyperplasia) had high-resolution echography and 56 (52 with adenomas and 4 with hyperplasia) had 210Tl/99mTc subtraction scans. METHODS Results of follow-up. RESULTS 78 Patients had parathyroid adenomas and 6 had hyperplasia. Parathyroid echography had a sensitivity (> or = 0.5 cm) of 55%, a specificity (< 0.5 cm) of 100%, a positive predictive value of 90%, and an accuracy of 77%. 210Tl/99mTc subtraction scanning had a sensitivity (> or = 0.5 cm) of 70%, a specificity (< 0.5 cm) of 100%, a positive predictive value of 100%, and an accuracy of 86%. The double tracer scan was slightly but not significantly more accurate than echography (p = 0.09). There were few postoperative complications except for transient hypocalcaemia, which usually lasted less than two weeks. All but one of the operations was successful. That one developed permanent hypocalcaemia, but it might have been caused by previous thyroid surgery and parathyroid injury. CONCLUSIONS To increase the success rate of parathyroid surgery, we recommend preoperative localisation with 210Tl/99mTc subtraction scan instead of echography, and routine bilateral exploration of the neck and thymus.