At the very heart of this advance has been the Oxford University Clinical Trial Service Unit (CTSU), headed by Richard Peto and Rory Collins. While both researchers have accrued citation totals in the last decade that place them among the top 50 of the 13,000+ names in the ISI Essential Science Indicators Web Product listing for clinical medicine, their work in the last two years is exerting particular impact at present. In this publication’s latest annual roundup of the world’s hottest scientists (Science Watch 15[2]: 1-2, March/April 2004), Peto scored with six recent Hot Papers, and Collins, even more impressively, wound up in the top tier with eight highly cited reports published since late 2001. These include the results of the Medical Research Council/British Heart Foundation Heart Protection Study, published in The Lancet in 2002. This paper, a fixture in the Medicine Top Ten for several issues now, garnered 67 citations during January-February 2004 and currently ranks at #5 (see What's Hot in Medicine). Collins also has seven papers cited over 500 times each, and his paper with Peto and his CTSU collaborators in The Lancet in May 1998, "Tamoxifen for early breast cancer. An overview of the randomised trials," has been cited well over 1,000 times in just six years (see table below).

Collins, 49, earned his bachelor’s degree in applied statistics from George Washington University in 1977 and his medical degree, an M.B./B.S., in 1981 from St. Thomas’s Hospital Medical School at the University of London. Following his intern training, he then joined the cardiology department at the John Radcliffe Hospital and Peto’s CTSU at Oxford University. He also obtained his Master of Science degree in applied statistics at Oxford, studying with Peter Armitage, in 1983. Since 1986, Collins has been co-director of the CTSU at the University of Oxford and, since 1996, a professor of medicine and epidemiology as well.

  The Clinical Trial Service Unit at Oxford seems to be a unique institution. Tell us about its history and program.

Collins: Well, I came in 1981, and at the time Richard Peto was director of the CTSU, and he had about a dozen or so people who were doing both trials and epidemiology. The unit was set up to do cancer research, particularly tobacco-related, and to help people do better clinical trials. Now there are about 150 people, about 10 times what it was 20 years ago. And we do about half clinical trials, half observational epidemiology, and about half cancer, including leukemia, and half vascular disease. What we’re trying to do is look at the major causes of death and disability among adults, not just in developed countries but in developing countries as well. What prevents people from getting beyond middle age and into old age? Even in developing countries, such as China, infant mortality is falling very rapidly but deaths in middle age are not. So the focus is on trying to reduce death and disability from the common causes in middle age.

  How do you approach that?

Collins: One way is to try to understand the important causes better. Over the last 30 to 40 years, for instance, Richard Doll and Richard Peto have been trying to understand the full effects of tobacco worldwide. From studies that Peto set up with colleagues in India, for instance, we’ve come to understand that tobacco is a major cause of death from tuberculosis in India. It’s a major cause of death from emphysema in China. Studying different populations about known risk factors can be very informative. We have a recent paper that demonstrates that high blood pressure is about 50% more important as a cause of vascular mortality than people had realized. If you have a key risk factor like blood pressure, and you find out that it’s 50% more important than people had known, that’s considerably more pressing than learning about some new risk factor that’s not very strongly related to disease. For example, contrast that finding with a new report saying that some new genetic polymorphism might slightly increase the risk of disease.

  What do you mean by blood pressure being 50% more important as a cause of death?

Collins: It means that the risk previously thought to be associated with a particular difference in blood pressure—say, 10 millimeters in systolic blood pressure—was a substantial underestimate. There’s another point that this paper makes pertaining to blood pressure, which also turns out to be true for cholesterol and is very important to understand. That’s the continuous nature of the relationship between these risk factors and disease.

High-Impact Papers by Rory Collins et al., Published Since 1994
Rank	Paper	Citations
1	M. Clarke, et al., "Tamoxifen for early breast cancer: An overview of the randomised trials," Lancet, 351(9114): 1451-67, 1998.	1,124
2	P. Appleby, et al., "Indications for fibrinolytic therapy in suspected acute myocardial infarction: Collaborative overview of early mortality and major morbidity results from all randomized trials of more than 1000 patients," Lancet, 343(8893): 311-22, 1994.	802
3	R. Collins, et al., "MRC/BHF heart protection: Study of cholesterol lowering with simvastatin in 20356 high-risk individuals: A randomised placebo-controlled trial," Lancet, 360(9326): 7-22, 2002.	634
4	J. Danesh, R. Collins, R. Peto, "Chronic infections and coronary heart disease: Is there a link?" Lancet, 350(9075): 430-6, 1997.	628
5	O. Abe, et al., "Polychemotherapy for early breast cancer: An overview of the randomised trials," Lancet, 352(9132): 930-42, 1998.	601
6	R. Collins, et al., "ISIS-4: A randomized factorial trial assessing early oral captopril, oral mononitrate, and intrevenous magnesium sulfate in 58,050 patients with suspected acute myocardial infarction," Lancet, 345(8951): 669-85, 18 March 1995.	580
SOURCE: Thomson ISI Web of Science

There have been certain problems in the way researchers have been looking at risk factors. One is that they have been plotting risk against the risk factor, with risk on a normal scale: 1, 2, 3, 4, 5, 6, 7, etc.... When you do that you often get a curve which suggests that the level of risk flattens off at lower levels of the risk factor. If you instead plot risk on a doubling scale—1, 2, 4, 8, 16—what you get with many risk factors—such as blood pressure or cholesterol—is a straight line. And you might think, "Well, so what?" What that straight line means is that the same absolute difference in the risk factor is associated with the same proportional difference in risk wherever you are on the line. Let’s take blood pressure, for example: going from a systolic of 160 to 150 is associated with a halving of risk. But going from a systolic of 120 to110 is also associated with a halving of risk. So in proportional terms there is no flattening off. The same absolute difference in the risk factor is associated with about the same difference in risk throughout the range we’ve studied. This has enormous implications for lowering risk in the population because it suggests that if you shift the risk-factor levels down, then you will reduce risk in everybody—reducing risk by the same proportion irrespective of the level of the risk factor.

  Is this why, in regard to the various statin studies, they show that you get the same decrease in risk from cholesterol-lowering regardless of the initial cholesterol level?

Collins: Correct. People look at the observational epidemiology and see this flattening off of risk when plotted against a normal scale, and they say it must be that there are pleiotropic effects of statins—that statins must work in ways other than just lowering LDL cholesterol, because, when you lower the LDL levels of people whose LDL doesn’t appear to be high by Western standards, you get the same proportional reduction in risk. But if you plot against a doubling scale in the observational epidemiology, that’s exactly what you would anticipate. The same absolute reduction in LDL cholesterol should produce the same proportionate reduction in risk across the range, and that’s exactly what the trials have found. So the evidence that people use to argue for the pleiotropic effects of statins is actually evidence against them.

  Deciding to plot against a doubling scale sounds somewhat arbitrary. What’s the justification for doing so?

Collins: Straight lines are easier to understand. What you want in statistics are straight lines. If you can transform data to get straight lines, then these are often easier. What the straight lines say for blood pressure, and for blood cholesterol, is that the same absolute reduction in these risk factors should produce the same proportional reductions in risk irrespective of the starting levels.

Now, the other important aspect of observational epidemiology is to recognize that these straight lines also appear to be continuous throughout the range of populations we’ve studied. For blood cholesterol, it’s continuous even below the range that can be studied in Western populations. If you go to China, which we’ve done, you find that the relationship between cholesterol and the risk of vascular disease continues down to very much lower levels of cholesterol. This tells you that none of us in the West really have normal cholesterol levels. That’s what influenced our thinking when we set up the Heart Protection Study—that in both Western populations and Asian populations, lowering cholesterol should lower the risk irrespective of the starting cholesterol, and that’s what the Heart Protection Study demonstrated.

  Many of your most highly cited papers are meta-analyses—assessing treatments for heart disease and breast cancer, and assessing risk factors as well. Meta-analysis has always been a highly controversial tool in epidemiology. How do you see its advantages and disadvantages?