University of Minnesota

Arthur L. Klatsky

Year: April 10th, 2010
Location: via Telephone
Interviewed by: Blackburn, Henry


Arthur Klatsky and Gary Friedman and colleagues at Kaiser-Permanente in Oakland, CA were first to show in healthy populations the J-shaped relationship of alcohol consumption to CHD events and have most thoroughly pursued the possible causal nature of the association through major population studies. They consistently find, since the late 1960s, lower CHD and ischemic stroke risk in moderate drinkers than in abstainers and heavy drinkers. They find sufficiently plausible mechanisms in HDL raising, anti-clotting, anti-inflammatory, insulin sensitivity effects of alcohol to bolster the causal hypothesis for alcohol effects on risk. They find the association highly significant, greater in men than women, in younger than older, and independent of but influenced by body mass, blood pressure, and particularly smoking.

They continue to make contributions to understanding of the role of drinking pattern and culture (the more moderate, steady, and with-meals consumption, the more the association with low risk), and the type of beverage. They have also pioneered in finding the apparently independent protective effect of modest alcohol intake for diabetes and the possible enhancing “effect” for breast cancer risk.

In this interview, Klatsky takes us back to the first-ever scientific challenge to the idea that alcohol has inevitably unfavorable effects on health, in the Baltimore Study of Raymond Pearl at Johns Hopkins, an idea that alcohol in moderation was not harmful to survival, which he propounded based on sound evidence and presented at the height of the Prohibition Era.

Klatsky brings us up to the generally insoluble questions today, short of the presumably infeasible experiment (RCT), about the importance of the type of alcoholic drink, and calls for more direct studies of alcohol effects on intermediary conditions and biomarkers (Henry Blackburn).


Thomas Turner, Dean of Johns Hopkins, became the first president of what’s now called the Alcoholic Beverage Medical Research Foundation. When Gary [Friedman] and I published the article about alcohol before myocardial infarction, in 1974, he and a couple of other people, Brian McMahon you probably knew, and a [Hopkins cardiologist, Richard Ross, came and visited us and said they were interested that we were willing to publish something that was not unfavorable to alcohol. They wanted to see what our data were like and wanted us to collect more detailed data. They suggested a major mortality study.

Turner, who was a very erudite man, said we should know about Raymond Pearl’s work. So I went over to the University of California library and they actually had a copy of his 1926 book, “Alcohol and Mortality,” Pearl discusses animal experiments that he and that others did that show that particular animals given a little bit of alcohol live longer… And then he discusses some population data from Baltimore, from the Tuberculosis Risk Factor Study, which is the first that showed the J-shaped curve of mortality and alcohol consumption. Actually he writes beautifully. For example, he concludes: “Whether any particular person chooses to be a teetotaler, a moderate drinker or a sot, is a matter to be decided between himself, his inherited constitution, and ‘whatever Gods may be.” That summarizes it all.

I also quote Pearl” “You can’t judge the effects of nutrition by the excess of starvation.” He’s the one who showed that you shouldn’t lump all drinkers together. He did it 50 years before anybody else did that.

I also used a quote of Sir William Osler and now that I’m getting older I kind of like to put it in that context. He said that “alcohol is the old man’s milk.” I don’t know when he said that but I kind of like that.


I think the relationship to diabetes looks real: light to moderate drinkers have lower risk of incident diabetes. I think the data are looking more and more suggestive that moderate drinking in some way improves insulin sensitivity, which is a key step in the development of diabetes. I’ve seen speculation that anti-oxidant effects of alcohol have something to do with this insulin sensitivity business. But it’s all very vague. But I think there’s enough consistency of the data about alcohol and diabetes that makes it fairly convincing. I’m curious, I’ll grant.

H: Tell me about Gary Friedman and you getting interested in this alcohol question:

K: I’ll tell you that story. I started practicing at Kaiser Permanente in 1961, coming on 50 years next year. And I was more or less a clinical cardiologist full-time for the first 10 years or so, wrote an article or two but nothing about alcohol. And then Gary came and visited me in the late ‘60s, I would guess about early 1969. He decided that there were real epidemiology opportunities at Kaiser Permanente. There was already a multi-phasic examination and there were plans to store the data on computer. It really wasn’t done yet until the mid-1960s. But anyway he thought he would cast his lot with Kaiser Permanente for a while. So, by this time when he visited me, Morris Cullen had gotten a large NIH grant to study the value of routine physical examination with lab tests. It was called the Multiphasic Evaluation Study; it was a big grant for those days. As part if it, they got a state-of-the-art computer. It filled a huge room. I was kind of afraid of the thing because it made funny noises. It probably had a fraction of the power of a little hand-held thing now. That was the computer and they were storing all these data from these physical examinations. So Gary got the idea of looking for new risk factors for heart attack by doing computer matching. And he wanted a clinical associate for his grant proposal for the NIH. I did find that it was interesting and I signed on.

That was a fateful day in my life. I’ve said that on many occasions. I didn’t realize how fateful it was going to be. So we got to know each other pretty well because we reviewed cases; there were almost 500 of these patients that had their first heart attacks and had these examinations beforehand. So it was kind of prospective in spirit although the analysis was a case-control analysis with the heart attack patients matched by computer. This was one of the first computer-matching studies I think in people who hadn’t had heart attacks. So anyway we had to review these records to confirm the diagnosis. We got to know each other, and we had fun doing this. When we started to look at the results, actually there were lots of interesting things and the first thing that I wrote had to do with coffee, not with alcohol. I don’t know if you recall a study by a man named Herschel Jick of Boston that said that coffee drinking was related to higher risk of heart attacks. We had a coffee question about heavy coffee drinking that was totally unrelated to CHD. So we said look, this is a hot thing, it got published and in the newspapers and everything. So I wrote this little article. It was the fastest acceptance I ever got. JAMA took it right on the first shot, didn’t even have a revision. That kind of excited me. So then when we were looking over the results, that was a quite productive study because that was the first study that really suggested the inverse relation with CHD. That study also, I don’t know if you remember this, but it was the first study that showed a risk factor in a marker of inflammation, leukocyte count was a risk factor for coronary disease. That was a NEJM article. I think that was the first one that showed that and Gary saw that as being potentially important, but I just thought it was a confounder related to smoking or something. But he thought, no, that really looked pretty good.

So one of these things that we were looking over in the data was that abstainers from alcohol, we used the term teetotalers in the first article. And they were at higher risk. And Gary said to me, he was very generous,” maybe you’d like to be the person to report that finding.” And I said sure because I was interested in clinical cardiology, essentially in alcoholic cardiomyopathy and used to visit problem cases and stuff like that and it was pretty interesting. So I wrote that article and that’s what got me and got us into the alcohol field. So that’s the story. When I wrote the article, it was before we knew about alcohol and HDL; there were no plausible mechanisms.

K: We did the mortality study which showed the J-curve. That was published in ’81. We did other cardiovascular disease studies. We did a study of stroke and then in recent years concentrated more on cancer partly because I think that as time goes on, I think some of the mechanisms having to do with alcohol and cardiovascular disease are short-term ones. But cancer is more a long-term exposure I think that we have now for studying cancer. Anyway, that’s what I’ve been doing in recent years, mostly.

H: Tell me a little bit about your approach to adjusting the alcohol-CHD relation for blood pressure.

K: Well, I think that heavy drinking is for sure related to higher blood pressure. I think that’s very consistent. There have been reasonable short-term clinical experiments that show that. I think it’s still unresolved whether there’s a threshold and where it is, but there probably is a threshold that may be different for different people. But there probably is a threshold. But in any case, it could be an intermediary causal factor. So if you adjust for it, you don’t see the alcohol effect, so that’s why. Now cholesterol is a little trickier because total cholesterol isn’t that much related to alcohol drinking, but it might be because HDL cholesterol certainly is related to alcohol drinking. And I don’t think LDL cholesterol is much related to alcohol. We’ll come to the triglyceride issue in a minute. But anyhow, cholesterol could be in the causal chain. And then you say, why not adjust for it; why is it ok to adjust for BMI and smoking. I think they are correlated, both of them, to the endpoint, coronary disease, and [smoking] is certainly correlated to alcohol drinking. Alcohol and smoking are associated, but I think not causally. I may be wrong, but that’s what I think. I think actually that smoking, I’m sure it’s the most important to control for in an alcohol study.

H: The under-reporting worked out very well.

K: I thought that was pretty interesting the way it came out for high blood pressure. Now we applied that principle to the breast cancer association, I don’t know if you knew that. And we sort of thought maybe it would turn out that the relationship of 1 to 2 drinks per day also was largely due to an implicit under-reporting but it was not so for breast cancer. It didn’t matter whether they seemed to be underreporting or not. So it looks as if light to moderate drinking, or what was reported as moderate drinking is really related to increased risk of breast cancer. I’m talking about 1-2 drinks a day, not less than 1, but 1-2 drinks which most people would say… I know the USDA says more than one a day for women. Anyway that’s the category. But you know, 1-2 is pretty broad. It goes all the way from a full glass of wine to two double martinis. I guess I don’t like having to tell women it looks as if you really can’t drink more than a few drinks a week without increasing risk of cancer, but that’s the way it looks.

H: Tell me a little bit more about the anti-inflammatory effects.

K: You know, I think the anti-inflammatory effects have more to do with non-alcohol ingredients in the beverages. Particularly a lot of people think red wine, but also I think to a lesser extent white wine and dark beer and so forth. As best I can tell, alcohol itself doesn’t have much anti-inflammatory activity. I think that brings up the issue of the whole question of the beverage choice and I don’t know what you think about that. But that’s an unresolved issue, the one that really interests people.

H: I think you were one of the ones who suggested that we probably would have to deal with that argument for quite a while due to inadequate data.

K: That one’s going to be hard to resolve. But I guess I think that at least the majority of the effects on coronary disease has to do with ethyl alcohol. Certainly the HDL part of it does, and some of the anti-thrombotic effects. I think the fibrinogen effect, the anti-platelet effect, that’s alcohol. And most of the insulin sensitivity issue is probably alcohol. So if there’s an additional benefit to the phenolics in red wine, it’s additional, it’s not the primary thing. And to what extent it counts, I think we really don’t know. The wine drinkers best, the beer drinkers next, and the distilled spirits drinkers have the least benefit. They all have benefit, but that’s the order. But in our California population the wine drinkers really have healthier habits in general, are better educated, have higher socio-economic status. It’s all clustered together.

H: Moderate wine drinkers are from a different culture. Ok. Resveritrol, you haven’t paid much attention to it?

K: Well I certainly pay attention, it’s a fascinating substance. And the longevity business [industry] is a matter of huge doses [jn pills]. You’d have to drink more wine than anybody could to get those doses. Did you know that it has antioxidant effects? Well I think the story is not fully told yet. There’s a whole cottage industry, it’s more like a “mansion industry” about phenolics in wine and benefit, but it’s mostly with surrogate endpoints, mostly it’s animal experiments and surrogates. But there’s virtually nothing in terms of prospective human data. So whether those studies will be done or whether it’s possible to do them I don’t know. But we need that sort of thing.

H: I’ll have to check. What did you think of my friend Ducimetiere’s talk last year on the French Paradox?

K: I thought it was really right on. I think the French Paradox is a catchy phrase so people like it. People like slogans. I thought that was a very good talk and I agreed entirely. Firstly, he just talked about the fact that there really wasn’t much of a paradox. I think the whole idea really is probably superannuated let’s say. However, it’s part of the folklore almost.

K: Let me make a couple of observations. First, the fact that cardiology has been, over my 50 years experience, simply an incredible journey. I mean clinical cardiology. If you had told me when I was in training what we would be doing in my lifetime I would have said, you’re talking science fiction. And here it’s still developing. So I’ve really enjoyed that. Personally, for me this second career, which I really think it’s become, probably has been a personally great thing. It probably saved me from a mid-life crisis and who knows what else.

H: Good to hear.

K: It’s been pretty exciting too, the alcohol legacy.

H: Would you like to speculate where we are now in CVD Epi after going to the San Francisco meetings? Are we just going in circles or looking for these novel risk factors ad infinitum?

K: You mean in terms of alcohol?

H: No, in terms of where cardiovascular epidemiology is going.

K: Oh, I think a lot of the stuff that’s presented at the meetings is arcane, I must say. Actually part of the problem is young investigators who don’t have a real concept of what’s likely to be meaningful and what’s not likely to be meaningful. I admit that sometimes things that don’t seem likely to be meaningful turn out to be so. But still, I think that’s part of what’s going on. I gather you sort of react the same way?

H: I was disappointed, at times horrified, that the presentations of young epidemiologists had only to do with genetics or “personalized medicine,” almost nothing to do with social causes of epidemics. That’s not our primary mission in epi, to serve the clinic, we exist for the public health.

K: Right, a lot of it has to do with genetics. And that’s important, but there’s a sameness to those reports isn’t there? You sort of know they’re going to find a couple of SNPs that are responsible for some small proportion of whatever it is they’re studying.

H: Right, but have nothing to do with epidemic coronary disease, obesity, diabetes, etc.

K: Not much, not even meaningful.

H: Well you were about to volunteer what is going to happen with alcohol research.

K: I think that the ideas are going to stay pretty much where they are. I think that the important area really is moderate drinking. I think that heavy drinking has so many adverse effects that it almost doesn’t matter whether it’s related to something else new. Moderate drinking is the important thing. And it would be very nice if we could have randomized controlled trials, but I’m sort of doubtful that we’re going to be satisfied with anything that’s done because it’s so tricky. There are so many potential disparities in terms of beverage choice and whether it’s even possible to get people to comply over a long enough time. So a satisfactory trial will be a big issue.

Even if we could get institutional review boards to agree, which we might be possible with properly selected subjects, I think compliance would to be a really big issue. So I don’t know where we’re going. We’ll probably have to be satisfied with intermediary markers as endpoints and that won’t satisfy many. So I think the debates will go on.


Klatsky, AL. 2002 Alcohol and Cardiovascular Disease. A Historical Overview.

Ann. N.Y. Acad.Sci. 957:7-15.

Pearl, R. 1926 Alcohol and Longevity. A. Knopf, New York

Full Transcript Access

Full transcripts of interviews may be made available to those engaged with original materials for scholarly studies by contacting us.