Tobacco Documents Online

#AyENAL OF PSYCHOSOMATIC RESEAR~~ RE: MS# 96-274: Increased serum mitogenic activity fbr arterial smooth muscle cells a.ssociated wit relaxation and low educational level in human subjects with high but not low hostility traifs_ !mplicationafor atherogenesis Author(s): Gutstein et al. pu~ REVIEWER'S COMMENTS (for author) rF~~ Introduction J Though there is an established relationship between hostility as a personality trait and coronary heart disease (CHD), the mechanism of the relationship is unknown. Therefore, a paper which looks at the hostility - CHD mechanism is welcome. More especially welcome is the fact that the authors used an established scale to measure hostility traits, a series of different psychological situations and some outcome variables which have a proven association with atherogenesis. On page 2 it is stated that hostility is a"main" factor. This is not the case. The recent meta- analyses of hostility and heart disease (Miller et al, 1996, Psychological Bulletin) found that hostility, at best, accounts for about 1% - 2% of the variants in heart disease; moreover, self- rated hostility predicts even less of the variance. It should be made clear that we are not dealing with a variable that explains a great deal of the variance in heart disease. It became obvious at this stage that there are large sections of this paper that I am not competent to review. Therefore, for knowledge of areas other than psychology and statistical design and analyses there should be separate expert reviewers. Method I did not understand why the authors had such disparate numbers of subjects in each of the three psychological manipulations. The imbalance of subject numbers is not a problem per se. However, once one realises that hostility is to be used to split groups into high and low level, the frustration and relaxation groups will become very small, especially for analyses with so many co-variates. Again, within the method section it was clear that an expert biological referee was required to evaluate the validity of the measures and designs. Statistical Analyses and Results The design of the study and the inevitable statistical analyses were sophisticated. My initial worry is about the number of subjects in order to cope with such an elaborate analyses. I did not find the clarity of the statistical analyses section to be carried on into the results section. On page 12, paragraph 2, it is not clear what the curvilinear relationship is between. continued overleaf -(hc kourn:d ~,I P..ih,~.uinaiir Rc...n~h i. i~nhli.hrJ h~ I_Lr% irr S, irn;r lu,

On page 13, paragraph 2, 1 found the first sentence of the results section to be very inaccessible. The next two sentences served to confuse even further. The reader is given no / way in or "feel" for the results. My comment here is one of clarity, not of analytic method. / Indeed, I thought the authors justified rather well their choice of analyses. On page 14, paragraph 2, I was not clear what was being correlated. It was about here also that I noted that selected analyses were being performed. Some consideration, therefore, of type one error needs to be done. On page 15, I found that the clarity of the exposition improved when the groups were split into high and low hostility levels. However, although I may have missed it, I could not find out what the cell sizes were for the psychological manipulation by hostility split. However, it is obvious that there can have been only around twenty subjects in two of the groups when split by hostility level. These are very low cell sizes and power analyses would indicate that such cell sizes are capable of detecting only very large effect sizes. Discussion Much of this section is a post-hoc justification of the one main result; i.e., that people with high hostility in the relaxation condition seem to be suffering most "atherogenic" activity. With lots of analyses and ielatively small numbers of subjects, one has little confidence in this result. General Comments The overall aim, outlook and content of the study was good. The very uneven numbers of subjects in groups tended to make the key result based on a very small number of subjects. The psychological manipulations - especially those associated with frustration and harassment - were trivial. It seems a pity to have made the subjects undertake all the medical checks and measurements that the study required only to irritate them when playing Super Mario Brothers. The statistical analyses was elaborate and quite well justified but unclear in exposition and some key facts were left out. i j J