[Skip to Navigation]
November 13, 1995

Evaluating Cholesterol Screening: The Importance of Controlling for Regression to the Mean

Author Affiliations

From the Division of General Medicine and Primary Care, Department of Medicine, Beth Israel Hospital and the Charles A. Dana Research Institute, Harvard-Thorndike Laboratory, Harvard Medical School (Drs Forrow, Calkins, and Delbanco, and Ms Allshouse); and the Department of Pathology, Beth Israel Hospital and Harvard Medical School (Dr Horowitz), Boston, Mass.

Arch Intern Med. 1995;155(20):2177-2184. doi:10.1001/archinte.1995.00430200056009

Objectives:  To determine the overall impact of a voluntary cholesterol screening program and to assess the importance of controlling for regression to the mean when evaluating the impact on higher-risk subgroups.

Design:  Longitudinal study comparing baseline survey and cholesterol results with those obtained 17 months later, with adjustment for regression to the mean.

Setting:  Participants were seen at three sites: an academic teaching hospital, an intermediate-care hospital, and a suburban community center.

Participants:  One thousand fifty-three participants were enrolled in a voluntary cholesterol screening program.

Measurements/Main Results:  Of 1053 study enrollees, 785 (75%) completed a follow-up questionnaire and 656 (62%) returned for follow-up cholesterol tests. Among all participants at high risk, measured cholesterol levels were 0.66 mmol/L (25.5 mg/dL) lower at follow-up, but 58% of this apparent change (0.38 mmol/L [14.7 mg/dL]) was attributable to the statistical phenomenon of regression to the mean. Among all participants at increased (both high and moderate) risk, after controlling for regression to the mean, reductions in cholesterol levels were significant for those younger than 60 years (—0.28 mmol/L [—10.8 mg/dL]; 95% confidence interval, —0.43 to —0.13; n=390) but not for those 60 years or older (+0.007 mmol/L [+0.3 mg/dL]; 95% confidence interval, —0.16 to +0.17; n=266).

Conclusions:  For subjects found at high risk in a cholesterol screening program, more than half of the apparent beneficial change in cholesterol level is attributable to regression to the mean. For participants older than 60 years, no clear benefit of cholesterol screening was documented.(Arch Intern Med. 1995;155:2177-2184)