Coffee and Pancreatic Cancer
An Interview With Brian MacMahon (April-May 2001)

On March 12, 1981, the New England Journal of Medicine published a case-control study reporting a strong association between coffee consumption and pancreatic cancer. The principal investigator was Brian MacMahon and co-investigators included Stella Yen, Dimitrios Trichopoulos, Kenneth Warren and George Nardi, all from the Harvard School of Public Health. The association was not diminished by controlling for cigarette use, and for the sexes combined, there was a significant dose response relation. The relative risk associated with drinking up to two cups of coffee per day was 1.8 (95% C.L. 1.0 - 3.0) and that with three or more cups per day was 2.7 (1.6 - 4.7). The authors were guarded in their conclusion-- “...The positive association with coffee consumption that we observed must be evaluated with other data before serious consideration is given to the possibility of a causal relation.”

Their findings, if confirmed, would have considerable public health implications. Cancer of the pancreas is the fourth most common fatal malignant disease in the United States and accounts for approximately 20,000 deaths annually. The implications were made clear by the authors themselves when they estimated that as much as 50% of pancreatic cancer is potentially attributable to coffee consumption. An account of the study was published in the New York Times on the same day, and the report received wide spread publicity throughout the country. The study was criticized by the National Coffee Association and by epidemiologists in a subsequent story in Medical World News (MWN), much of it focusing on the appropriateness of the hospital control group. The New York Times editorial page pronounced it fit to believe because it had the Harvard University and New England Journal of Medicine seal of approval according to the MWN story.

The Epi Monitor contacted Dr. MacMahon and he willingly agreed to discuss the study and subsequent developments surrounding its publication. Following is the text of our interview.

Epi Monitor: In light of the comments that I have been made by other epidemiologists and the reactions you have received since publication of your report, what are your current feelings about the appropriateness of your control group?

MacMahon: I’m satisfied with it. I think that there was one point about it that we missed. We were concerned to show that the pancreatic cancer cases differed both from the cancer controls and the patients with non-malignant disease. What has been pointed out subsequently is the possibly heavy weighting of the control group with patients with gastrointestinal disorders. I think that if that point had occurred to me before publication I would have added information showing that the relationship holds whether you exclude gastrointestinal disorders or not, although it is true that the relative risks are reduced by excluding the patients with gastrointestinal disorders.
Moreover, you have to consider that cancer of the pancreas starts also with gastrointestinal symptoms, and it’s an open question as to whether or not you should exclude the patients with gastrointestinal disorders; maybe the control group with the GI diseases included is better than the one without them in the sense of being more similar to the pancreas cases. Also, don’t forget that the question we asked related to coffee drinking prior to onset of this illness, the illness for which the patient was hospitalized. I think there’s a question there and there’s a question in people’s mind, but I’m satisfied that we’ve taken care of that.

Epi Monitor: There were some questions which you raised yourself in the paper about the control group.

MacMahon: As you know in an observational study there is never a perfect control group, and while we do plan in a study which I may tell you about later to include a general population control group, I don’t think that is going to be perfect either. You will always have a high rate of refusal in a randomly sampled general population control group and the circumstances of the interview are not the same as when you are interviewing in hospitals. I think there are things to be said pro and con for a variety of controls.

Epi Monitor: What are your comments about the differing results that you’ve observed in males and females? Some people were skeptical because there was no dose response effect in the male group.

MacMahon: That could readily be due to chance. If one value in the male data were changed to another value well within the confidence limits of that rate ratio, we would have a nice dose relationship in males too.
Epi Monitor: In all the things that you’ve heard since the study was published, is there any other supporting evidence that people have brought to your attention or any other evidence which tends to make the association less likely?

MacMahon: The authors of three Letters to the Editor of the NEJM have kindly sent me copies. There are three sets of data--all relatively small, and I don’t think I should describe their findings until they are published, except to say perhaps of the three sets I have seen, one is supportive and two find no relationship.
Epi Monitor: These represent direct evidence on the association between coffee and pancreatic cancer. What about other indirect evidence --such as the low pancreatic cancer rates in Mormons--has any of that surfaced that tends to support or take away from the association?

MacMahon: No.

Epi Monitor: In light of fairly guarded statements in your paper about this association, what were your reactions to the criticisms? Were they justified?

MacMahon: I haven’t had much criticism. I’ve seen at least one abusive piece in the newspaper which was just taking off at us, and I have had letters saying--Why don’t you stay at home and treat people or help them get better instead of taking away our coffee--but I don’t consider that as criticism. I’ve had some questions such as the one on the control group put to me but again, not in any critical way.

Epi Monitor: Have you seen the piece in Medical World News?

MacMahon: No, I haven’t.

Epi Monitor: The headline reads “Storm Brews Over Study Tying Coffee to Pancreatic Cancer” and features criticisms from Alvin Feinstein, David Rush, and Irving Kessler.

MacMahon: I had a call from David Rush. I didn’t take the call, but I got the message. I didn’t know what he was referring to. He said that he felt his remarks had not been properly interpreted in the paper and I did not know which paper he was talking about. He must have been talking about that one.

Epi Monitor: If you have not received muchcriticism, your report certainly received much attention and publicity. What was it like?

MacMahon: It was just impossible to take all the calls. It has died down now, but four or five calls were coming in at a time. It is an interesting experience. The first copies of the NEJM were received on a Monday or Tuesday. On Monday afternoon and Tuesday, I got calls from some of the better known journals. These people obviously read the article and they had intelligent questions to ask. It was no problem to deal with that. By Wednesday the word had gotten out in some fashion, maybe on a network, to lots of small radio stations and newspapers all over the country. They had not seen the article and they said please tell us what you did and what you found. By the time you’ve done that a 100 times in a day, you’re a bit sick of it. So there was the physical impossibility of answering all the calls, and next, I got to the point where I decided I just couldn’t live my life for that. I gave the thing five days and after that I quit.

Epi Monitor: Was the reaction and publicity unexpected?

MacMahon: I expected some reaction but not anything of this strength.

Epi Monitor: Why do you think your report received this much publicity?

MacMahon: I think it was a combination of them not having anything else to do that day, the fact that it is coffee--it’s almost like apple pie as Americans go--and it gives some of them an opportunity to have a go at the mad scientists who are always finding things that cause cancer.

Epi Monitor: What advice would you give to other epidemiologists who are confronted with the possibility of the same kind of publicity?

MacMahon: I would publish it on the day that the President gets shot or some other thing that keeps these people busy.

Epi Monitor: Do you think that in the end the publicity was a good or bad thing?

MacMahon: It depends on whether it’s right or not. If it turns out to be right--that the association is there--then I think it will be a good thing. We’re having all kinds of difficulty with the bureaucracy that has grown up around human subject committees. I think that a demonstration to the public that from time to time this kind of research does turn up a useful thing will all be to the good. If it turns out that the association is not present--that something is wrong with our data--then, of course, it will be a bad thing because it will tend to discredit the method.

Epi Monitor: There is a fair amount of cynicism in society now because it seems like whenever we read the newspapers we are learning about some other thing that we should or should not do. Do you think that this is a problem today in terms of the reputation of epidemiology? Do you think that in the end if we’re not careful about the validity of associations that we publish, that we will make our work more difficult in the long run?

MacMahon: Yes, I think we should be very careful about publishing data like that, especially data that are likely to be inflammatory.

Epi Monitor: Did you think about that at all before publishing your report?

MacMahon: I’ve been sitting on this for two years thinking about it. But it is 369 cases, the largest series by far that has been published. It is a strong association and you have an obligation to get out what appear to be meaningful associations. You have that obligation as well as the obligation to be careful of what you do. You have to find the line somewhere in between. I think that on a data set this size, with the findings that we had, there was an obligation to get it out.

Epi Monitor: Is the fact that you were concerned to some extent about the reputation or credibility of epidemiology in general one of the considerations that made you wait so long before publishing it?

MacMahon: Yes. And concerned about my own reputation also.

Epi Monitor: The New York Times reported that you have stopped drinking coffee, but that you would not presume to advise others. What were some of the factors that entered into your decision to quit?

MacMahon: Well, it was in part real in the sense that it was based on my assessment of the likelihood that this association was going to turn out to be real. It’s also in part to produce a public image that is consistent. I wouldn’t be concerned at all about my epidemiology colleagues who understand the nature of the data and the attending uncertainties. I wouldn’t be concerned at all about consistency with them. I can very well see an epidemiologist looking at this set of data, deciding for the moment that he or she is going to continue drinking coffee, but that isn’t how the general public and the press want it. They want black and white answers--is it or isn’t it? If you force yourself to a black and white answer, I’m still of the opinion that this is more likely black than white. If the press forces you to that kind of answer-- which they do when they ask you what do you do--you can’t say that I really believe this but on the other hand I am going to continue drinking coffee myself. That I think is an insult to the data and to the study.

Epi Monitor: One final point would be your own plans for the future. Do you have plans for additional studies and could you outline in some detail that epidemiologists would appreciate where you go from here?

MacMahon: We’re hoping to begin a new case control study with the same kind of methodology but with two additions. The question will be focused more on coffee, trying to get some estimate of lifetime consumption as well as what kind of coffee is most usually used, whether it’s dried or percolated, decaffeinated or not, and what additives are used. The second major change would be to incorporate a general population control group.
Epi Monitor: Are you going to do two case control studies?

MacMahon: No, one study with two control groups.

Epi Monitor: Would it be with a new case series?

MacMahon: Yes.

Epi Monitor: Also in the Boston area?

MacMahon: Yes, 10 hospitals, using hospitalized cases, hospitalized controls, and general population controls.

Epi Monitor: What length of time do you think this would involve?

MacMahon: Four years.

Epi Monitor: Are there any other things you would like to say or comment on that haven’t been raised or that you think other epidemiologists would find interesting?

MacMahon: I think that it’s important to say that the great majority of the letters I received were trying to be helpful and some of them were indeed very helpful. They included very detailed accounts from people in the coffee industry who really wanted me to be educated about what a complex substance coffee is, some very interesting case reports, and concerned letters from people with families and friends that they have been worrying about.

Epi Monitor: How many letters have you received?

MacMahon: I haven’t counted them, but it must be two or three hundred.

Epi Monitor: And these have come from all over the world?

MacMahon: Yes.

Epi Monitor: Have you ever had this kind of attention from anything else you’ve ever done before?

MacMahon: No.

Originally Published April/May 1981