Findings in several retrospective studies have supported the hypothesis that hyperhomocysteinaemia may be an independent risk factor for premature arteriosclerotic disease. This prompted us to investigate whether frozen plasma samples could be used to study the question prospectively. Total plasma homocysteine concentrations in 6-16-year-old (10.9 +/- 2.5, mean +/- SD) frozen (-20 degrees C) and fresh samples from the same 76 men were 11.6 +/- 4.9 and 14.1 +/- 4.3 mumol l-1 respectively, the values being significantly correlated (r = 0.58, p < 0.001). The difference was not correlated to storage time. After the first sampling, 13 subjects had survived a stroke, 16 a myocardial infarction, and 51 were still healthy, being used as matched controls. In stroke patients, values from old and fresh samples were 14.2 +/- 5.5 and 16.4 +/- 4.8 mumol l-1, respectively, and tended to be higher (p = 0.06) than in matched controls whose respective values were 11.4 +/- 2.8 and 13.8 +/- 3.8 mumol l-1 (n = 22). No such differences were seen between patients with myocardial infarction and their matched controls. We conclude that total plasma homocysteine can be measured in up to 10-year-old frozen plasma samples, indicating that such samples can be used for prospective studies on the relationship between plasma homocysteine and vascular disease.