Background. The degree of maintenance of risk factor changes observed in post-trial follow-up of participants in the Multiple Risk Factor Intervention Trial (MRFIT) provides insights into long-term determinants of preventive behaviors, as well as post-trial mortality differentials. Methods. Nine hundred eighty-nine former MRFIT participants at four clinical centers were invited to be reexamined 2-3.5 years after the end of this 6- to 8-year trial. Seventy percent came to the clinics for measurements of systolic and diastolic blood pressure, plasma lipids and lipoproteins, serum thiocyanate, and dietary food frequency. With the addition of phone interviews, 82% provided self-report data on cigarette smoking and use of antihypertensive drugs. Results. Comparison of Special Intervention (SI) and Usual Care (UC) groups showed that differences persisted for total (6.5 mg/dl, P = 0.02) and low-density lipoprotein cholesterol (5.4 mg/dl, P = 0.04), and for diastolic blood pressure (1.6 mm Hg, P = 0.02). Differences were not significant for systolic blood pressure (1.8 mm Hg, P = 0.12), cigarette smoking, and serum thiocyanate. The plasma cholesterol difference represented 70% of the value at the end of intervention, and food frequency profiles were consistent with maintenance of an SI-UC difference. Loss of the SI-UC difference in smoking was attributable to both UC cessation and SI recidivism. The blood pressure difference was not attributable to different proportions of SI and UC men on medications, but possibly to differences in diuretic doses, use of multiple drugs, and adherence. There was a persisting contrast in use of specific diuretics, with more frequent use of chlorthalidone and less of hydrochlorothiazide in the SI group. Conclusion. These findings suggest that cholesterol-lowering dietary changes are self-sustaining, while smoking cessation is less so, and may require continued intervention. Finally, the SI-UC differences in use of specific diuretics remains one hypothesis for explaining a portion of post-trial mortality trends.