| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Ann Thorac Surg 2004;78:1131-1132
© 2004 The Society of Thoracic Surgeons
Department of Cardiac Surgery, St. George's Hospital and Medical School, London, SW17 0QT, UK
marjan.jahangiri{at}stgeorges.nhs.uk
To the Editor:
We read with interest the article by Lee and colleagues on the randomized trial of off-pump versus on-pump coronary artery bypass grafting (CABG) and neurologic morbidity [1]. One of the definitions the authors used to identify neurocognitive dysfunction was a 20% decline in 20% of the tests. This is an arbitrary and relatively insensitive definition. It is also affected by the statistical phenomenon of regression to the mean (RTM) [2]. Essentially, an individual with an extreme test score at one setting can obtain a score similar to the mean of the group, of which the individual is a member, at a subsequent assessment. This study, which categorizes patients as decline/not decline, is particularly prone to the effect of RTM [3]. Furthermore, it is not useful to label people as "decline" or "not decline" and then compare the proportions, as one would be discarding most of the information and turning a measurement into a yes or no.
The other method the authors have employed, the multivariate analysis of variance (MANOVA), compares the difference of the mean scores between each time point for each test category. There are two limitations to this analysis. First, the use of group means is insensitive as it does not evaluate individual performance and does not account for practice effects [4]. Comparing group means will show change in the net direction, regardless of the contribution each individual makes. Second, the use of difference scores does not control for imbalance in the preoperative scores (ie, individuals/groups with low scores would be expected to show a large change in score due to RTM) [5]. This could be the reason why the patients in their off-pump group showed a significant improvement in the Rey Auditory Verbal Learning Test (RAVLT). The authors attribute this change to practice effects, but why should practice effect be just confined to the RAVLT?
An alternative method of analysis would be to obtain for each patient a composite score, from all the individual test scores, by principal component analysis. The postoperative scores, adjusted for the preoperative scores, could then be compared between the two groups by using analysis of covariance (ANCOVA). ANCOVA takes RTM into account and is a powerful method of analysing test–retest data [5].
References
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| ANN THORAC SURG | ASIAN CARDIOVASC THORAC ANN | EUR J CARDIOTHORAC SURG |
| J THORAC CARDIOVASC SURG | ICVTS | ALL CTSNet JOURNALS |
