Richard Shekelle

Quotes

Richard Shekelle on Type A Behavior Pattern and Coronary Heart Disease

RS:  I was a research psychologist early in my career with an interest in the role of psychosocial factors in causing disease.  For example, one of my studies back then supported the idea that incongruities in social status  — e.g., having an educational status that was higher or lower than one’s occupational status — was associated with risk of CHD among middle-aged men employed by the Western Electric Company at the Hawthorne Works in Chicago.  I later came to feel, however, that this line of work was basically futile.  After all, I thought, even if social status incongruities were causally related in some manner to CHD, what could be done about it?  My attention shifted to diet, but I never lost the willingness to consider the possibility that psychosocial factors might be involved in the pathogenesis of CHD.    I mention this to explain why, back in the 1970’s, I became interested in the idea that something called “the type A behavior pattern” might be associated with risk of CHD.  I have been retired now for over two decades, and no longer have the files that contained correspondence, memos, reprints, and the like, about my involvement in the Type A controversy, so most of the following  material is based solely on memory.]

The Western Collaborative Group Study (WCGS), after 8.5 years of follow-up, found that the Type A behavior pattern (TABP) was associated with increased risk of CHD and that this association could not be explained by potentially confounding factors.  The paper, published in 1975 by JAMA, was followed by creation of the NHLBI Behavioral Medicine Branch, which pronounced TABP as an established CHD risk factor on a par with high blood pressure, high serum cholesterol and cigarette smoking.  However, a single cohort study was a slender reed on which to base such a sweeping conclusion.  It took over 20 years of research by several investigators working in a variety of populations before the “big three” risk factors reached that august state.

When the Multiple Risk Factor Intervention Trial (MRFIT) was being organized, some of my colleagues and I thought it presented an opportunity to give the TABP hypothesis another test.  Jerry Stamler may have been the first to suggest it — I no longer remember — but certainly the MRFIT Behavior Pattern Study (BPS) would never have gotten off the ground without his support.  Many Principal Investigators were antipathetic toward the topic.  (As an aside, I remember being amazed at Jerry Stamler’s powers of persuasion.  More than once, the Steering Committee would take a vote in Jerry’s absence only to reverse that action the next month when Jerry was present and spoke against the previous action.)  In the end, the BPS did get off the ground and accomplished its objectives.  The results were published in 1985.

I will return later to the BPS.  I introduce the topic now only to explain why I was interested in the WCGS in 1984, the year I first noticed problems with the study.  Specifically, I noticed that the final report, in comparison to earlier reports, showed several changes in the baseline data:  28 more men excluded from the  population at risk because of prior CHD at baseline (141 instead of 113), 28 fewer men in the population at risk (3154 instead of 3182), 5 more Type A men (1589 instead of 1584), and 33 fewer Type B men (1565 instead of 1598).   These changes were not mentioned in the report.  I thought that clerical or typographical error was not a reasonable explanation.  If changes actually had been made in the baseline database, then knowledge of baseline observations might have influenced the determination of outcomes and vice versa, and bias could not be excluded as an explanation for the WCGS  results.

When I asked Dave Jenkins whether he could shed any light on this, he said that he had seen Rosenman and Friedman reviewing the baseline TABP classifications during the follow-up period, and advised them strongly against making any changes. 

Later, Stony Stallones told me that he had helped Rosenman  design the WCGS.  The research protocol had specified that baseline classifications of TABP would be sent to Stony, who would keep those data secure until follow-up was finished.  However, when Stony moved to Houston in 1968, he returned the data directly to Ray Rosenman. (Parenthetically,  Stony acknowledged that he had been wrong to do this.)  So, contrary to protocol, Rosenman and Friedman did have control of these data from 1968 onward.

In short, observer bias cannot be ruled out as an explanation for the positive association between TABP and 8.5 year  incidence of CHD  observed in the WCGS.  I think that observer bias is the most reasonable explanation given the fact that Ragland and Brand subsequently found no association between TABP and risk of CHD death in the same cohort after 20-plus years of follow-up for mortality.

Returning now to the MRFIT BPS, one of the design problems we faced was the absence of an objective standard by which we could ensure that our procedure for measuring TABP was comparable to that used in the WCGS.   Since Ray Rosenman had been responsible for those measurements, we asked Ray whether he would assist us  by training our interviewers, monitoring their performance, listening to recordings of their interviews, and taking responsibility for the validity of the final classification of behavior pattern.  Ray agreed.  During the course of working out the details of his involvement, Ray said that he wanted not only to listen to the recorded interview but also know CHD status of each participant before he made the final classification.  He explained that determination of behavior pattern was similar to making a medical diagnosis in a patient, and said something like:  If you’ve learned more about this patient during follow-up, don’t you want to go back and re-evaluate your diagnosis?  I explained why the baseline measurements, even if erroneous, should not be changed in light of subsequent events.  In epidemiology, we know that errors occur but we must ensure that those errors are not influenced by knowledge of outcomes.  Ray accepted this stricture, but I’m unsure whether he understood the reason.  This is something that some  clinicians seemingly don’t understand.

The MRFIT BPS found no association between TAPB and risk of CHD.   We submitted a paper to the New England Journal, but it was rejected. We submitted the paper to JAMA and after an unusually long time received a rather apologetic letter from an associate editor, who regretted to say that they were not going to publish it. He assured me that this decision had nothing to do with scientific merit but was based solely on editorial policy.  This seemed strange to me, but I had been raised in the tradition that you don’t argue with the judge, so I didn’t protest the decision.  Some five or six weeks later, JAMA published a lengthy  interview with Mike Friedman, who extolled the importance of TAPB.  The MRFIT paper finally appeared in The American Journal of Epidemiology.  As an aside, Oley Paul was really upset with me for not publishing in a major medical journal.  Jerry Stamler, who had been isolated in Pioppi while all this was going on, said he would have argued with the JAMA editor. 

We did present the MRFIT Behavior Pattern Study results at a meeting of the AHA Council on Epidemiology.  I think that would have been in March of 1986.  The paper received a standing ovation, which greatly surprised me.  Bill Castelli, who had been sitting immediately in front of the podium, got up and danced a little jig.  I stood there with a silly grin on my face.  That information got back to Ray Rosenman, who called me a few days later.  I can’t recall his words, but the message was clear.  He thought I had been playing a double game, getting his cooperation by appearing to be sincerely interested in TABP but in fact intent on destroying it.  Nothing could have been further from the truth.  I did not set out to either support or undermine the hypothesis, but only to give it a fair test, and that we were able to do.