Resource use was routinely collected as part of the study. Non-inpatient resource use data was collected using a questionnaire distributed between January 1996 and September 1997. The incremental costs reported in the analysis have the study protocol driven costs removed. These were replaced with a pattern of clinic visits reflecting general practitioner and specialist clinical opinion on the implementation of intensive policy. Where a patient was still alive at the end of the follow-up, a simulation model was used to estimate the time from end of follow-up to death. It was assumed that there would be no continuation of benefit of therapy beyond the trial period in.