Evaluation of risk factor variations in relation to their baseline values in a controlled preventive trial....

This paper proposes a method for separating intervention impact, regression to the mean and spontaneous evolution effects in a controlled preventive trial. Variation dependence on baseline values is estimated by means of corresponding regression lines. Regression line slopes are adjusted to eliminate bias due to the difference between observed and 'true' values. Intervention impact is then evaluated using the control group as reference. Possible differences in factor baseline values are adjusted for. The method is used to analyse variation patterns for body mass index, serum cholesterol, systolic blood pressure, diastolic blood pressure and cigarette smoking in four years in the Rome Project of Coronary Heart Disease Prevention cohorts. All regression lines of variations vs baseline values show a negative slope. A highly significant improvement due to treatment is observed for blood pressure and body mass index, a less substantial but significant improvement is observed for serum cholesterol, and no effect is observed for cigarette smoking.