Michael Honore Criqui

Quotes

I knew something about the disease process (PAD) from a clinical focus and I was interested in some of the fundamental epidemiologic issues. For example, how sensitive and specific are our assessments of the disease, what’s the predictive value? And what I realized is that in epidemiologic studies today the usual endpoint had been either ascertainment of intermittent claudication or from palpation of pulses. And in the ‘50s, which is somewhat before my time, various investigators had developed non-invasive ways to measure flow, which had got better and better.

But what hadn’t been done is the widespread use of more accurate non-invasive measures of PAD to define disease in the population.. . . the first important thing if you’re going to study a disease epidemiologically is to be able to define it. And we were one of the first to use a non-invasive test in a population and show that there was more disease than people thought [PAD] and that neither leg pain nor pulses were a very good marker for it. (2)

We knew that claudicants had a higher death rate. I was still in high school probably when they discovered in Framingham that the claudicants actually had a higher heart attack and stroke rate. There were several European studies that had reported this. The Malmo Study [Men of 1914] had reported similar results. So it was known that peripheral arterial disease was a marker, that claudication was at least a marker for future problems. It was also known that obviously patients with severe presentation of the disease clinically could lose their limbs and have other very important clinical outcomes. So what I think my work added was taking these non-invasive tests to a population base and showing how much disease was asymptomatic as well as what the survival was….These were even asymptomatic persons. See, people had only previously looked at survival . . with leg pain. So we extended it to the asymptomatic group, which has major implications for screening for disease, treatment of disease, population risk and so on.

The first thing we actually tried to do was just show the sensitivity, specificity, and predictive value of what the clinicians were doing to find PAD. We showed they weren’t very good for what we called the traditional measures – the leg pain, and the pulses – and then we suggested that if you really want to know who had PAD and who didn’t, those were not sensitive or specific enough measures. So we were one of the first to say you really can’t make the diagnosis. You can’t rule it in or rule it out really with these measures. You have to have a more objective and reliable assessment.(4)

More on PAD

Well, you know, it has sort of had what I would call an exponential trajectory. . . And it’s only really in the last five years others had a lot of interest. The NIH now has an interest, the AHA has an interest, NHLBI has put out a special RFA for it, the AHA has a special working group for it, it’s now mentioned in the AHA statistics. . . . So I think all these things that are happening now are the product of gradually sort of catching on and being confirmed. Various studies have looked at the Ankle-Brachial Index in many populations and have confirmed it. The real problem still is actually at the patient level not population level. You know, ABI is largely measured only in research studies, really not measured by clinicians. Whereas, ultra CT scans are measured by all kinds of clinicians. The ABI is simple, it’s cheap, takes little training, but they don’t get reimbursed for it and so it doesn’t get done. But the interest in PAD and patients has increased dramatically and so we’re hoping this will translate into more population interest. (6)

Alcohol

. . I was interested in actually a paper in the New England Journal of Medicine by Gary Friedman and Art Klatsky from Kaiser saying that alcohol increases the risk of hypertension. Art Klatsky, himself, said in 1974, I think, that alcohol might prevent coronary disease because light to moderate drinkers in a case-control study had a reduced risk for coronary disease. Well, it doesn’t add up that it could increase blood pressure and yet prevent coronary disease. Something weird was going on here. So what I decided to do was look at the matter. . .

I suggested [to the LRC population study] that I would like to work with alcohol data because I was interested in hypertension versus coronary disease – why would they go two different ways. . .What I found was what other people had found, that light to moderate consumption was ‘protective’ for coronary disease, but that light to moderate consumption didn’t increase blood pressure much, while heavier consumption increased blood pressure substantially.

And this added up and helped stimulate the idea that there was this window of opportunity, that there was actually a U- shaped curve and many investigators argued, “Well, it wasn’t U-shaped, it was an artifact of various things including selective migration of ill drinkers to non-drinking.” But actually the biological data and epidemiologic data fit together very nicely. That if someone could limit themselves to one or two drinks a day, they would get a nice HDL increase, they wouldn’t get much blood pressure change, they’d not get increased coagulability, they’d get improved insulin resistance, and they’d probably get less coronary disease. But with heavier drinking basically everything went awry. The blood pressure increased with sharp rebound conditions in the morning. Even though the HDL was still up it couldn’t protect you from the other problems and so it was this light to moderate drinking or one to two drinks that actually is beneficial. . . .the triglyceride effect is very much like the blood pressure effect, so in lower doses alcohol might actually lower triglycerides a little bit. The big increases in triglycerides you see with high doses of alcohol. I don’t know if anybody has tried to tease that out, but …

But the observational evidence is fairly convincing and so I think the issue is solved. What I think is also clear is that there are a number of people – up to 20 or 25% of the population or so – who are genetically prone to abusing alcohol. So public health recommendations have to be extremely cautious. (7)

NIH

. . . they’ve been good to me. I think that they have done a very good job funding major epidemiologic projects. I think one of the concerns, one of the tensions is that in recent years things are getting more complicated. The epidemiology and the clinical trials didn’t always agree. For example, antioxidants and hormones in women and so on. So I think at the moment there is a little bit of a crisis in epidemiologic research, not sufficient research in general where people are distrusting of anything if it’s not a randomized clinical trial, which, of course, are very expensive and have their own problems.

So I see that as a bit of a crisis in the discipline. One can no longer take for granted the results of observational studies. You know, smoking was bad and that was pretty much proved. Cholesterol was bad and that was proved. Blood pressure was bad and that was proved with clinical trials. But the hormone therapy benefits and antioxidant benefits and some of the other things – homocysteine lowering with folic acid – have not been proved in clinical trials and so, I think, we’re in a mode of saying what can we learn from observational data and can’t, and what else do we have to do? Do we need new methods to analyze data? Can we get around biases without trials and so on? So I think there is a whole new potential for development there. (11)

The New Genetics and NIH

I have heard both sides of the story. Their side of the story [at NHLBI] is that epidemiologic studies are extremely expensive and take large numbers of people and they take up a disproportionate amount of the budget compared to what they contribute. Obviously, as epidemiologists we think they are the most important studies and so on. But I certainly can accept the reason they feel the way they do and their point of view. They certainly are continuing to fund large epidemiologic studies.

There’s going to be a new MESA. There’s been, you know, ARIC, CHS, there’s been a lot of major studies. I’ve actually served on a special emphasis panel to determine whether NIH/NHLBI should do future epidemiologic studies and the answer of the emphasis panel was a resounding “yes” that, in fact, we should do future studies and the reasons were elaborated why we should.

I do think one thing that is really an obsession – I shouldn’t use the word obsession – an overarching interest of the NHLBI at the moment is genetics in general and proteonomics as well as genomics. And that’s a very complicated and extremely sophisticated business. I think there is a demand for epidemiology looking at the real issues about the selection of genes in populations and whether or not what you are going to find in a population study will go beyond a familial study where you actually have more genetic control. I think there are a lot of unresolved issues. . . . I’m a little bit cynical about it, I guess, . . this enormous amount of expenditure and effort going into this thing and at the far end of the output there is not what it should be in terms of ability to translate into health changes in populations. Now that may all change overnight and everybody is saying now we can continue with the genome we can make anything of it. That’s fine. I hope it does, but I haven’t seen convincing evidence that the amount of expenditure is justified. (13)

Psychosocial Risk Factors

I did my MPH at Berkeley and Len Syme had just left for a sabbatical in London and Michael Marmot was actually taking his job and teaching an advanced epidemiology course. Lisa Berkman was there at the time and so I had a pretty good acquaintance with psycho-social epidemiology, social networks as predictors and so on. I thought it was all very interesting stuff and was quite excited about it since my major focus was cardiovascular disease.

When I got to San Diego the first thing I did in my new study, Peripheral Arterial Disease, was incorporated a Type A questionnaire, which had been built on Meyer Friedman and the development of Type A by Dave Jenkins and thought this would be a really exciting adjunct to expand the epidemiology of cardiovascular disease in general. I became somewhat disenchanted with the inconsistency of the Type A results and I know now that they have isolated out hostility as a component.

But I basically have had a couple concerns with the psycho-social field in terms of the operating definitions of the exposures and then the lack of consistency of the outcomes. Now I don’t claim to be an expert in the area, but some of the more consistent stuff I saw initially was Type A, which kind of went away and then, you know, the Berkman-Syme thing about social networks appeared to have some solid validity. The Marmot look on education, social class, all seemed to have validity and I think that in terms of interpreting any of our more micro results—blood pressure, cholesterol, etc. are more macro— how they play out, I think that these are major factors. I don’t know about psycho-social variables so much in themselves, but in terms of demographic variables, I think that race and social class and certain social variables such as social networks really are important in the way disease transmits. So I think it’s been interesting, with some very important observations in psycho-social epidemiology. (17)

Criqui-Shaper Debate on Alcohol

I remember this one well. This was 1993 in Oslo, Norway on the occasion of the . . International Meeting on Preventive Cardiology. It basically occurs every four years. . But, anyway, I received an invitation asking to debate Dr. Shaper on the question of whether alcohol was protective for coronary disease, . . . The room was full and it was a lot of fun. But what’s interesting is that since then I’ve worked with Gerry, published a couple of papers with him and he’s slowly come around to agree that moderate alcohol consumption probably is protective, that, in fact, it’s not all an artifact. But we both agree that it should not be used for public health purposes because of the hazards of the substance.

I would not want to speak for Gerry, but I think he has changed his opinion that, in fact, light to moderate alcohol consumption is probably protective. And even looked at in the British Regional Heart Study that he was PI on that if you look at it with the appropriate exclusions you can see there is a benefit…

One of the things that Gerry said …he really never believed. . . that HDL was a mediating factor for the benefits of alcohol, because the changes in HDL were so small. So what we did is we actually showed . . the mean changes and how big they were and [that[ . . changes that large were associated with substantial differences in CHD risk, at which point I think he hadn’t really appreciated how big the changes were and how much they meant. (20)

About History

I think it would be really interesting for everyone to know how cardiovascular epidemiology developed . . . It seems to me that it was the first of the chronic diseases to really develop a hardcore scientific discipline. The cancer epidemiology has really lagged behind. Neurological diseases has had a lot of interest. Dementia has now a huge interest. But CVD Epi sort of set the standards.

When Jerry Cornfield incorporated the logistic equation that was a boon for cardiovascular disease. The shift from infectious disease. . . So I think that’s real important to point out in the document . . the leaders in that because those leaders were not only leaders for cardiovascular disease, but they set the channel for everybody down the way.

. . .the computer revolution has changed everything. But also the fact that the teaching of epidemiology for whatever reason has somehow evolved in a lot of places to be . . really the heart and soul of evidence-based medicine or critical thinking. That you should be able to teach somebody critical thinking who is a doctor. . . what control groups really were and what bias really was and what was confounding. . . I think it’s really important to do clinical epidemiology and teach this to physicians so physicians are not necessarily swayed by drug company arguments, which don’t hold up, . .that physicians have a sense of individual patients versus the population at risk. I think physicians need these tools, these concepts. (22)

About the Tahoe 10-day Seminar in CVD Epidemiology

I remember one thing that was very formative in my career and that’s the seminar where we are sitting here today. And I was a fellow in 1978 and joined the faculty in 1980. But I can remember my excitement about receiving the preliminary notice…. I heard about the seminar, I applied, I was invited to participate as an observer, . .as overqualified. But I came and I thought, you know, well, the idea of being an observer when I already knew everything. Well, I got to Tahoe and found out that not only was there a ton of stuff I didn’t know, but the opportunity to talk in this kind of a . . privileged basis was a tremendous boon to me and I became very excited about it. I remember the preliminary reading materials …Zen and the Art of Motorcycle Maintenance and . . I did actually read it and that was a phenomenal insight into logic and the limits of logical thinking, the scientific method and applying that in the context of the intimacy of Tahoe. The opportunity to really get out of . . the grind for two weeks, I think was a tremendous experience. So much that I volunteered any day, any time, any place to be on the faculty and was taken up on it a year or so later and have been doing it ever since. (25)

Mistakes?

Well, I think, you know, probably at one time believed that I could learn as much from observational data as from randomized controlled trial data, so I think that was an error in my own judgment; that if I looked closely enough at the data and stratified closely enough and looked at the biases and inclusions in the study that I could actually tease out the effect of exposures in observational data versus control trial data. And it’s sometimes true, but not always true. So that’s just a more general thing. In terms of. . . fundamental errors, I don’t know. Once in a grant review where someone said, “These investigators didn’t even adjust for …” Even though I knew …about the issue, but I made the mistake of not doing it… So I’ve learned to trust my own judgment now…I tend to present a little more complicated picture, but it’s a little more accurate than a simpler picture that glosses over some of the details. (27)