Lymphoedema affects 25-28% of patients who have undergone axillary surgery or radiotherapy in the treatment of cancer. The prevalence of primary lymphoedema or of lymphoedema secondary to non-cancer causes is unknown. In addition, basic demographic and disease characteristics of the lymphoedema population remain undocumented. We conducted a prospective survey to determine these characteristics. Lymphoedema treatment clinics in the UK were invited to participate at the British Lymphology Society 1996 Annual Conference. Twenty-seven clinics collected a standard data set for all new referrals in a 3-month period in 1997. The sample size was 603. The mean number of new referrals was 7.4 patients per unit per month. Eighty per cent of patients had cancer-related lymphoedema, and 86% of patients were female. Patients' mean age was 61.0 years. Thirty-seven per cent of the sample had suffered lymphoedema for less than 3 months, and 15% for 5 years or more. Patients with non-cancer related oedema had suffered the condition disproportionately longer than the cancer-related group (chi 2 = 102.2, P < 0.001). Sixty-six per cent of patients were reported as having lymphoedema at more than one site. Seventy-four per cent of all patients suffered unilateral limb oedema, and 23% bilateral limb oedema. Bilateral swelling was recorded for 67% of the non-cancer related group and 13% of the cancer related group (chi 2 = 153.4, P < 0.001). Median percentage excess volume (PCEV) for all patients with unilateral limb oedema was 16.4%. There was a positive relationship between PCEV and duration of the condition (H = 24.0, P < 0.001), and PCEV and non-cancer related lymphoedema (H = 10.5, P < 0.05). These results suggest that patients with non-cancer related lymphoedema and those with cancer-related lymphoedema form two distinct groups. The former present substantial management problems which are made more complex by late referral.