Paul Leaverton

Quotes

I think Bob Levy (Director of NHLBI) enjoyed visiting Congress every year saying, “CHD is the number-one killer disease of the industrialized world and we need more money to fight it. It keeps going up.” The National Center for Health Statistics, NCHS (where I was the associate director for research from 1974-80) was getting behind (budget issues) on tabulating national mortality data, four to five years behind. All of a sudden, four years came out at once, catching up with the backlog. About that time, I think Dr. Levy’s mouth, along with lots of peoples’ mouths, dropped wide open. All of a sudden there was a decline seen from the late Sixties presenting in the mid-Seventies – that’s how far behind they were – the age-adjusted CHD mortality rate wasn’t going up any more. It was actually coming down! Age-adjusted and age-specific and every way you wanted to look at it. About 2% a year, coronary deaths were declining. No one had anticipated this. (5)

Dr. Levy then convened a conference to address the phenomenon. He, and other scientists, wanted to know if the change was real; and, if real, what were the causes. The conclusion from that famous conference was that the decline probably was real. But the interesting political aspect was we didn’t know whether it was due to incidence (prevention) or better care after a heart attack. Most felt it was some of each but we didn’t know the proportions. I don’t think Dr. Levy’s budget approach to Congress missed a beat. Now, however, he probably said “We have an important positive turnaround (thanks to past research funding) but now we need to discover the underlying reasons so as to support them.” I think it worked and the funding was continually increased; perhaps even more so. Levy was sharp.

In my role as the new director of research at NCHS, I felt like we were (collectively) falling down on the job. Here was the number-one disease in the United States by every measure and we had a totally unexpected major change in mortality rates. Not only a major public health phenomenon, a major sociologic phenomenon, and we had the most highly-paid, well-funded statistical agency in the world to keep track of this [NCHS]- NHLBI were both pretty well funded – and we couldn’t even say what happened. . . We just didn’t have the data to know whether it was prevention or improved care after a heart attack. To me, that was embarrassing. (page 7) Of course, the real reason is that we didn’t (still don’t) have a genuine national health care system in America. So health records were far from standardized. No one could know the true national incidence nor mortality after a heart attack. There were no national health care records as many countries had.

Shortly after that conference, I was offered (in late 1979) a new position at the NHLBI and accepted to begin in January, 1980.

Dr. Don Fredrickson, the Director of NIH, resigned abruptly soon after my arrival. He called in people at fairly high management levels, to say he was leaving in two weeks. Unusual. His reason was that he was tired of explaining a new budget to a new person every go-around. He had had four bosses [at DHEW] in about 16 months. It was exceedingly difficult because most of these new politically appointed secretaries had no health background at all. It had become so politicized and so frustrating for him. We all felt sorry about this development; Dr. Fredrickson had been a great director.

Dr. Levy and Dr. Manning Feinleib wanted me to take on the responsibility for answering the main question which had emerged at the 1978 CHD decline conference. What were the reasons for the (now continuing) decline in coronary death rates? One of the issues was, can we look at hospital records in some sort of sampling way – around the country. They were going to allot a couple of million dollars to do it, or a million and a half, and Levy’s idea, or one of his parting shots (before he also resigned) was, “Make sure you do a pilot study first.” Because he wasn’t sure you could do it at all. That pilot effort became the Community Coronary Surveillance Study (CCHS) (9).

CCHS and ARIC

After two years our pilot study (CCHS) was completed and, sadly, the results were a bit ambiguous. Nevertheless, my NHLBI staff (including Havlik and Sharrett) decided that the majority of hospital records wouldn’t support an investment at the level being contemplated. Records were too different from place to place. One important hindrance; it was almost impossible to tell from routine hospital records if a bona-fide heart attack was the first one or not. We needed to know that to properly estimate incidence.

It was my decision, director of this project (and with the strong prodding of Sharrett), that the better way to go was to bite the bullet and not spend $5 million, but seek an estimated $50 million and do four new subject “Framinghams.” We needed new cohorts involving persons known to be free of CHD at the outset. Now the interesting thing is, the four dynamic and talented pilot study leaders, Drs. Kuller, LaBarthe, Borhani, and Tyroler did not agree with me and my staff. (I think Tyroler was a bit on the fence.) They strongly urged a meeting with my immediate boss, Bill Friedewald and NHLBI chief Claude Lenfant to try to overrule this decision. These were four of the most outstanding epidemiologists in the country and my good (they still are) friends! So a meeting was held with Lenfant, with the four of them, and they explained why they thought the pilot study hadn’t worked out. They admitted difficulties but believed that the problems had been identified and could be solved and that we ought to focus exclusively with community surveillance to address the issues of the decline in incidence or mortality.

I was always a little intimidated by Dr. Lenfant, as most people were, but anyway I stuck by my beliefs and recommendation. After the meeting, he called me back and said, “Well, before we take your advice (which I think I agree with), you’d better get some senior outside consultation. Create a blue ribbon panel to settle the matter.” So I got such a panel together. Not just my decision, fair enough, but people like Abraham Lilienfeld was on it. I think he chaired it; one of his last professional roles before he died. You [Henry Blackburn] were on it. Dick Carleton was on it. Millicent Higgins was a participant, even though she had already agreed to leave Ann Arbor to replace me shortly after that time. Maybe 10 or 12 pretty senior scientists. After hearing the arguments, all but two agreed with our staff decision.

Upon reading the panel’s report, Dr. Lenfant said, “Fine. Even though it is very expensive, I’ll try to find the money to do it, the $50 million, because you convinced me that we need four new cohort studies to answer the questions.” I was pleased and surprised by his attitude. That was a lot of money and it hadn’t been in the advanced budget plan (so important in the government.) Not easy for Claude. So even though he wasn’t a big supporter of epidemiology, at that time we kind of changed his mind a little bit. … And as you know, the RFP went out. We had a residual component of the CCSP, the community surveillance, but the bulk of the money, probably 80%, was to create four new Framingham-type cohorts of about 1500 volunteers each, I think. Then shortly after that was done, I did the site visits and I left Bethesda to return to academia. I set it up and then left NIH. Later, the four centers chosen became ARIC. (13)

Shortly after Dr. Lenfant took over and Manning was gone. I was named Acting Director of Epidemiology and Biometry and Bill Friedewald was Director of Clinical Trials. Barbara Packard, I think, was Head of Cardiology. There were four of us. I don’t remember the other person. Dr. Lenfant came over from main campus to have his first meeting with his branch chiefs in the Heart Division. We were all a little bit nervous because he’s our new boss. As he looked around the room, one of his comments was he wanted to put more emphasis on basic science. And you know, that conjures up to most of us the vision of bench science. I remember Bill Friedewald sort of looked at me and says, “Where does that leave us, Clinical Trials and Epidemiology?”

I tried to save the day a little bit. I remembered Milton Terris’ quote, but at the time I couldn’t remember who said it. But I raised my hand and said, “Dr. Lenfant, some people say epidemiology is the basic science of prevention.” That remark did not please him in the least. And I was never very popular with him after that. [ed. Of course, it wasn’t long before he learned his lesson about epidemiology.] (page 16) To his credit, he did later change his views somewhat. And, while our relationship was never warm, he and I did get along for the most part. (14)

Tracking:

I am happy to have worked with Ron Lauer when he and Bill Connor began the Muscatine Study in Iowa. Ron especially devoted much effort to the issue of childhood origins of CHD. Measuring risk factors in children and following them into adulthood was the goal. Tracking of CHD risk factors in individuals is a more complex issue than you might think; how best to handle it statistically. The most interesting question might be relative values versus absolute values. In other words, if you’re at the 90th percentile when you’re seven years old in, say systolic blood pressure, how well does that predict what you might be when you are 17 or 30? That’s one of the questions. And is percentile the way to do it or should you look at absolute values? There are some thorny statistical issues there, whether to track percentiles or whether to track the absolutes. Statisticians debate heatedly the best way to do it.

Of course, Dr. Berenson (LSU) carried it one step further when he believed so wholeheartedly in tracking blood pressure that he started treating kids in the top percentiles – when their relative values were there and they were not at levels of concern as adult levels. He got in trouble actually treating these children because he was so convinced they were going to track into dangerous levels. I think they were mostly African-American children in Bogalusa. He was reprimanded for doing that even though he believed he sincerely believed he was helping them. Most of the pediatric community didn’t agree that it was a good thing to do. But it is an interesting issue; when to begin treating CHD risk factors. (25)

Clinical Trials

I love clinical trials as a concept and as a way to go in medical and public health science. I suspect that it forms the basis of what is now popularly called “evidenced-based medicine”. What better evidence could you have than a properly conducted clinical trial to, as what I like to say, “Sort out the competing hypotheses.” (27)

I still think and I teach this to this day that, after Galileo demonstrated that experimentation was the best way to sort out competing hypotheses, R.A. Fisher comes next because he showed us the power of randomization; random allocation. If you could analyze what tricks randomization could play in your observations, you could attribute the rest to causation if you did an experiment. This concept (which allowed precise statistical analyses, based on probability, of data) was, I believe, was a great contribution, as great as Galileo’s to experimental methodology. But we’ll see if history will bear this out.

The experimental design works beautifully in the area of physics, it works beautifully in the area of agronomy which is, of course, where Fisher and Kempthorne came from. I call that the ‘Fisherian model.” It works very well under many circumstances; in any field where there is a lot of variation among the responses of your experimental units. After all, the recognition of variation is why the whole field of statistics grew so much in the 20th century. I was so enamored with Fisher’s ideas still while I was at CDC that even after I got out of CDC… I was going to be a biostatistician but I didn’t want to mess with epidemiology because most epi studies are non-experimental. I wanted to do clinical trials and experiments where causation was clear.

But, anyway, I learned over the years that there’s a role for both non-experimental or observational data (as some people call it) as well as experimental. But if you can do it, you should really look to trials first. (29)

There is often a problem with surgical trials… I remember when I was sitting on the Heart Council, right next to a prominent heart surgeon one day, when the results of one of the first trials on coronary bypass surgery was reported. One aspect was that it was estimated that about a quarter of the bypasses were not needed. That if you only had one artery involved, you were probably better off not to have the bypass. The surgeon was not too pleased with the trial and stated “Clinical trials never convinced anybody of anything.” That’s a direct quote. Surgeons can often say that, because by the time one of their trials is done, they don’t do the procedure in exactly the same manner through time and I think that’s what he was implying. “That might have been a bad way, but we don’t do it that way anymore. The technology changes.” (33)

Despite special problems, surgery needs to emphasize trials more than they do, in my opinion. My favorite quote in all of science is one due to my late friend, Tom Chalmers, “One only has to review the graveyard of discarded therapies to realize how many patients would have benefited by being randomly assigned to a control group.” Perfect.

Ecological (Group) Studies

I have some troubles with ecological studies when people try to infer individual risk from group risk. They certainly have a role when the community is the unit, like for water fluoridation. If you’re exposing the whole community, it’s the perfect unit. Then there’s no issue of group risk versus individual risk. (34)

On Genetic Epidemiology

I do think the blend of genetics and epidemiology is going to be pretty interesting. I don’t think, in my opinion, that the new understanding of the human genome is going to do away with epidemiology. In fact, I have a feeling that, at least for a decade or two, it’s probably going to increase the need for epidemiology. Because on the surface, it looks like we’re going to have to redo everything by genetic subtypes. That would be the best way to look at the gene/environment interaction. [Of course] we couldn’t afford to do that. It’s going to be interesting to see how much we do of that and how much we do it through statistical/computer modeling.

For all chronic diseases, I think the same. I believe there’s genetic components, obviously, there’s environmental components. Sorting that out is going to be the challenge. I think epidemiology hasn’t done much about genetics in the past 50 years… But at least at the black-box level I see it this way: I see many people trying to duplicate some of the epidemiological studies of risk factors within certain phenotypes or genetic subtypes as a way to study the interaction between genes and environment. (45)

When asked what NHLBI has done right

Framingham and Honolulu. I think they’ve kept those places alive and they have funded the new ARICS and they funded the grandfather ARICS, an older group. I think those are going to pay dividends down the road just like it took ten or twenty years for Framingham to pay dividends. To have the foresight to realize – it’s not easy in the government to get something funded for 10 or 20 years. Government research must be very close to the politicians. They want to see results within three to five years like our grants are only three or four years. But for NHLBI to continue funding long-term studies, I think, is to their credit. I think that’s what they’ve done right. Maybe they should do more of it. We need those kinds of studies. I think the fact that they have collected and stored some material that allows going back and conducting genetic studies on them. I think Framingham’s in that game now, will be even more interesting in light of what we just said a while ago about genetics. (53)