The purpose of this study was to evaluate the usefulness of direct opthalmoscopy by non-opthalmologists in patients with hypertension. In a cross-sectional survey, we analysed the association between optic fundi abnormalities, individually and according to the criteria of Keith and Wagener (KW), with blood pressure and duration of known hypertension in 400 non-diabetic hypertensive patients. The optic fundi abnormalities were more frequent in patients with diastolic blood pressure (DBP) > 105 mm Hg (P = 0.002), SBP > 180 mm Hg (P < 0.0001) and with a duration of known hypertension > 3 years (P = 0.002). The severity of hypertension did not vary in parallel with the KW classes I and II: 34.5% of patients classified as KW I had a diastolic pressure of > 105 mm Hg compared with only 25.3% of those classified as KW II. Class III abnormalities were infrequent (2.5% of the whole cohort). In a logistic regression model, diffuse arteriolar narrowing was associated with DBP (P = 0.002) and age (P < 0.001). Abnormalities of the arteriovenous crossings were associated with SBP (P = 0.001) and duration of disease (P = 0.008). The positive predictive value of any fundoscopic abnormality to estimate the severity of hypertension was 59% and the negative value was 60%. The results of this study demonstrate that optic fundi examination by internists and cardiologists does not give an accurate assessment of the severity of hypertension in most patients, and that the Keith-Wagener classification of retinopathy has a limited applicability.(ABSTRACT TRUNCATED AT 250 WORDS)