The results of a randomised, prospective, multicentre trial of the management of primary open angle glaucoma are presented at up to five years' follow up. Previously undiagnosed cases were selected with intraocular pressure of 26 mmHg or more on two occasions together with field loss characteristic of glaucoma. Analysis was performed on one eye selected at random from each of 99 patients. Conventional medical treatment followed in unsuccessful cases by trabeculectomy (group A) was compared with trabeculectomy at diagnosis followed when necessary by supplementary medical therapy (group B). The life expectancy of these glaucoma patients was found to be similar to that for the local population matched for age and sex. In group A after four years trabeculectomy had been performed in 53% of eyes because medical management had failed to control the disease. The rate of operation was lower in those patients with intraocular pressure less than 31 mmHg and mild relative field loss (17% at three years) than in those with intraocular pressure greater than 30 mmHg and dense scotomas (75% at three years). Early surgery provided much more stable control with fewer changes in treatment than in group A. The group mean intraocular pressure after trabeculectomy was 15.0 mmHg irrespective of the time of operation, and this was significantly lower than the intraocular pressure in those cases thought to be controlled on medical therapy alone at the end of the first year (20.8 mmHg). Early operation provided significantly better protection of visual field, and the extra loss of visual field with delayed operation occurred in the preoperative period. Changes in visual fields were not related to the use of miotics. There was no significant difference in the final visual acuity in the two groups, but six cases in group A lost central fixation because of progressive loss of visual field, and there were no such cases in group B. Cataract occurred in approximately 10% of cases in both groups, but in group A this happened with only half the number of operations and at a shorter postoperative follow-up than in group B. It appears that in cases of primary open angle glaucoma of this severity the risk of delaying operation are significantly greater than those of performing trabeculectomy as the primary treatment.