A total of 405 diabetic patients who first attended St Thomas’ Diabetes Clinic between 1982 and 1985 had a detailed standardized computerized first visit record, including a structured foot examination and toe vibration perception thresholds (VPT, Biothesiometer), were reviewed in 1995. None of the patients had a history of foot ulceration at first visit. Twenty‐five patients (6.2 %) developed foot ulcers (n = 11, 2.7 %) or had an amputation (n = 14, 3.5 %) over a mean 12‐year period. Twenty of these patients were then individually matched with 3 non‐ulcer patients. Statistically significant odds ratios (OR) were found for a baseline abnormal age‐adjusted toe VPT (OR 4.38, CI 1.11–17.26; p = 0.01); abnormal clinical examination (at least 1 abnormality out of: ankle jerks, tuning fork or cotton wool sensation; OR 2.3, CI 1.00–5.20; p < 0.01); and HbA1 (OR 1.30, CI 1.01–1.66; P < 0.02) in patients who subsequently developed lower extremity complications. The sensitivity of VPT (70 %) was better than that for clinical testing (55 %) in predicting long‐term complications, although all tests showed similar specificity (70–72 %). The risk of events also doubled for every 10 years of diabetes (OR 2.10, CI 1.11–4.30; p = 0.02). We conclude that age‐corrected VPT measurements, which are objective and simple to perform, are better predictors of future foot complications than semi‐quantitative tests in diabetes clinics. We encourage their use in the campaign to reduce the morbidity of diabetic peripheral neuropathy. © 1998 John Wiley & Sons, Ltd.