Anesth Analg 2002;95:1823
© 2002 International Anesthesia Research Society
LETTERS TO THE EDITOR
Another Example of Regression to the Mean (Not)
R. Peter Alston, MD FRCA,
Ian J. Deary, MB ChB, PhD, MRCPsych, FRCP,
Michael J. Robson, MB ChB, FRCA,
Peter J. Andrews, MD FRCA, and
Michael J. Souter, MD FRCA
Department of Anaesthesia, Critical Care and Pain Medicine, Royal Infirmary of Edinburgh, Edinburgh, UK
Department of Psychology, University of Edinburgh
Department of Anaesthesia, Western General Hospital, Edinburgh, UK
Departments of Anesthesiology and Neurointensive Care, University of Washington, Seattle, WA
To the Editor: Landow in his letter in the June 2002 issue implies that our study of cognition following coronary artery bypass grafting surgery is flawed by regression to the mean (1,2).
Dichotomization of cognitive change into those who do/don’t decline results in people with high preoperative test scores appearing spuriously to have excess cognitive decline (3). However, we modeled postoperative cognitive scores as continuous variables to prevent this problem.
In regression to the mean, change scores show spurious correlation with baseline scores in instances where the baseline scores have themselves been used in computing change. We used postoperative cognitive scores as the dependent variable and preoperative cognitive scores as the first-entered independent variables in our models. Thus, "cognitive change" variance lay in the residuals that followed this procedure, and these are uncorrelated with the preoperative scores!
The extent of regression to the mean is a function of measurement error in the predictors. By using multiple cognitive indicators, which are individually reliable and highly correlated, we achieved highly reliable composite cognitive measures.
Our abjuring cognitive change thresholds, using a residualized approach to cognitive change, and multiple cognitive measures makes Landow’s comment on our study incomprehensible. Could he indicate specifically how its results might be affected by regression to the mean?
References
- Landow L. Another example of regression to the mean. Anesth Analg 2002; 94: 1673.[Free Full Text]
- Robson MJA, Alston RP, Deary IJ, et al. Jugular bulb oxyhaemoglobin desaturation, S100ß and neurologic and cognitive outcome after coronary artery surgery. Anesth Analg 2001; 93: 839–45.[Abstract/Free Full Text]
- Browne SM, Halligan PW, Wade DT, Taggart DP. Cognitive performance after cardiac operation: implications of regression toward the mean. J Thorac Cardiovasc Surg 1999; 117: 481–5.[Abstract/Free Full Text]
|
