Dr. Arthur Schatzkin
National Cancer Institute
National Institutes of Health
"Can Dietary Change Prevent Colorectal Cancer – Or Not?"
March 14, 2001

Dr. Arthur Schatzkin: This is the enemy – colorectal cancer – from a clinic or pathologic perspective. And this is the enemy from the population statistics point of view. In the United States alone, colorectal cancer will be diagnosed this year in approximately 130,000 men and women and will kill approximately 55,000 people and that’s just in the US. Now there has been a lot of interest in the genetics of colorectal cancer, the genes involved in the development of the disease over time, in potential genetic susceptibility to certain variants of colorectal cancer. But it’s pretty clear, as Dr. Childress said, that environmental factors, in the broadest sense, which could include behavior-related factors, are clearly important in the etiology in the causation of this disease. What you have here is you are simply looking at the mortality for colorectal cancer across a number of countries. You see the tremendous variation in the death rate. You can see the same thing for the incidence rate, the rate of occurrence in the disease. Something like a ten-fold difference in the rates. So clearly something is differing from country to country.

Along the same lines, in several countries around the world, these are data from China, from Shanghai. What you can see is starting at a relatively low rate of the disease in the 70s, there has been a fairly rapid increase in the incidence rate for colorectal cancer in Shanghai. You can see very similar data from Japan. Something is going on here over time and it is something that can’t be very easily explained by genetics alone.

In a third set of studies and data of this type are the migration data. What you see here for several different countries, these are migrant to Australia from the countries listed here. These are people who were there for a shorter period of time and these are people for a longer period of time. What you see is the fairly typical migration pattern for colorectal and many other cancers, which is a convergence from the rate of the country of origin to the rate of the country of destination, even to the point where we start with a higher rate and you see a somewhat lower rate after migration

Again these three sets of data, the international variation, the time trend data, the migration data certainly suggest that something in the environment is changing and that that change can profoundly influence the rate of the disease. Again just reminding us of the magnitude of colorectal cancer, here is some data from around the world: nearly 900,000 cases; 500,000 deaths. But diet, and this was the conclusion of a very recent survey and summary study from the American Institute of Cancer Research, "diet has long been regarded as the most important environmental influence on colorectal cancer." And diet makes sense as a key environmental factor that would go along with the kind of ecologic or aggregate data that I showed. Clearly diet varies from country to country. Diet is changing rapidly in Shanghai and Japan, and diet clearly changes with migration and acculturation. From a purely physiologic perspective, what we eat and the metabolites of what we eat comes into direct contact with the lining of the large bowel.

There are a whole variety of animal studies. Both older studies where potent carcinogens were used and newer studies using transgenic or genetically engineered rodent models show that diet can modulate the development of tumors in the large bowel in various animal models. A number of human metabolic studies have shown that if you alter what you eat, you can affect bile acid metabolism, transit rates, and a variety of potential intermediate markers that are punitively related to colorectal cancer.

And there have been a series of observational epidemiologic studies looking at individuals that have shown that a number of dietary factors seem to be related to one’s risk of colorectal cancer. And here you can see a list of several different hypotheses for nutritional factors that increase risk – dietary fat, red meat, processed meat, meat cooked at high temperature that produces substances called heterocyclic amines that are potent nutogens and carcinogens, sugar (and there is now an increasingly interesting hypothesis about the role of high glaucimic load and hyperinsulanemia and insulin resistance in the development of colorectal and possibly other cancers). Alcohol has been shown to be a risk factor for colorectal cancer, a weak one. And then several nutritional factors have been suggested to reduce risk, particularly vitamins E and D and folic acid. Calcium and selenium – there is some evidence that these may reduce risk. The whole dietary fiber hypothesis, which I will return to, and fruits and vegetables either in toto or specific classes or even individual fruits and vegetables. And finally I wanted to mention the notion of a multifactorial eating pattern. For example, people can think of the Asian cuisine or the Mediterranean cuisine that might involve several of these individual hypotheses that might put you at higher or lower risk.

Now having said all that, this says, "Let’s see. Who had the nutrition issues?" There are a lot of nutrition issues so with what really it is in diet and nutrition that may influence colorectal cancer. Now here we see this came out shortly after some of the big negative studies about diet fiber over the past year or two. [He shows an editorial cartoon.] "Sure. Now they tell me - after I’ve ruined a perfectly good set of patio furniture!" We have, of course, the enthusiasts who have no personal or fiscal interest whatsoever in the science here. But [showing cereal box] "Wheaties – may reduce the risk of some cancers in a low fat diet of whole grain foods like Wheaties". And there is clearly a certain amount of inconsistency, uncertainty, and angst not only among the scientists in the field but clearly among the lay population, as to what is it that we really should eat. Yes there was a strong "Fiber Hypothesis" but then a series of high profile articles come out and say, "Well it doesn’t look like fiber really does have anything to do with colorectal cancer."

Now there are some particular challenges in this whole field that I wanted to mention. One is that, and I am not going into any detail here, people can discuss it afterwards, but there are a whole lot of problems with how we measure what people eat. If you try and think about how well you can remember what you ate five days ago or whether you can accurately summarize what you ate over the past year. We know that there is a lot of error in the way that we assess diet. And the problem with dietary assessment error is that it makes it harder to see true relationships between dietary factors and colorectal cancer and other diseases. It dilutes or tenuates associations. Another problem has been that studies are often set up...for example, within the United States, when you actually don’t have that much of a range of intake, and you may not be able to compare really high fat or really high fiber intake against people consuming really low fat and low fiber intake. Often a study in particular countries say the U.S. will have relatively homogenous fat and fiber intake.

And finally there is this problem of, what we call in the epidemiology business, confounding. Confounding is simply the notion that people who eat differently from one another also differ on a variety of other behavioral, lifestyle, and biologic characteristics. And it may be these other characteristics that are really the causal factors for the disease but since they are associated with and vary with the dietary factors it might seem to be that it’s just the dietary factors. This is a nagging concern with some of the nutritional epidemiology. Clinical trials, and these are becoming increasingly used in the prevention realm (there’s a term called experimental as well as observational epidemiology)…clinical trials can be particularly valuable in dealing with the confounding problem. That’s because a randomized trial, by assigning people through a statistical randomization procedure to one or another treatment group, really essentially rules out the possibility of confounding because the groups are virtually alike on a whole variety of characteristics that you can measure, as well as characteristics that we may not know about or may not measure. So a randomized trial, by minimizing confounding, can qualitatively raise the level of evidence that dietary change can reduce cancer risk.

And I am going to spend much of the rest of the talk talking about a study that got a lot of play over the past year. The polyp prevention trial, which we initiated about a decade ago (every time I say that, I’ve been around for longer than I had imagined)…but we actually began the study over ten years ago and we began it because we knew that there were strong diet and colorectal cancer hypotheses, but we wanted to see if we could up the level of evidence, if we could rule out the possibility of confounding, for example, by demonstrating that dietary change could affect the colorectal cancer endpoint…and I will talk about polyps in a second…whether you could definitively show that dietary change worked.

Now this Polyp Prevention Trial, the primary hypothesis was really a simple one. It was whether adoption of a low-fat, high-fiber, fruit- and vegetable-enriched eating plan would reduce the recurrence of colorectal adenomas, or colorectal adenomenas polyps, which are precursor lesions for most invasive colorectal cancers. Now these were the targets, these were the goals of our dietary intervention. We wanted to get people down to 20% of calories from fat, 20% of total energy. For total dietary fiber, we wanted to get people in our intervention group up to 18 grams per 1000 kilo-calories. So for someone consuming 2000 kilo-calories, it would be about 35 grams per day. And for fruits and vegetables, we wanted to decrease servings, about 3 and a half servings per day per 1000 kilo-calories, which is roughly 5-8 servings per day of fruits and vegetables.

Now why did we focus on fat, fiber, and fruits and vegetables. And I’ll try to give just a little bit of data on each of these. Now these were from an epidemiologic study that appeared over 10 years ago from the well-known Harvard sponsored Nurse’s Health Study. And they looked at a number of factors and here you see the relationship between animal fat, different amounts, these are the different quintiles of animal fat consumption and grams per day, and the relative risk of colorectal cancer. Here is the reference, shows the relative risk of one. It’s almost a double risk of colorectal cancer in the upper category, the upper quintile, of animal fat.

There are a whole series of studies from that time that showed some similar results. Is it plausible that dietary fat could affect carcinogenesis in a large amount? And the answer is yes, that dietary fat increases the concentration in the intestinal lumin of bioacids by changes in gut bacteria. And the bioacids themselves can be genotoxic and mutogenic, particularly the secondary bioacids, or directly citotoxic. Citotoxicity can lead to hyperperliferation. There can be through increased dietary fat intake, increased concentration of free fatty acids The point here is that there is plausible biology why dietary fat could increase the risk of colorectal cancer.

Now this says, "Wellness update: Thirty-year -old man starting on the 25,000 pound bran muffin he must consume over 40 years in order to reduce significantly his risk of death from large bowel cancer." Now dietary fiber and gender is a lot of cartoons and jokes but there is actually real biology and real science behind the dietary fiber hypothesis. In particular there is the observation by Dennis Burket some decades ago where he saw that in Africa diets with very high fiber content had very low incidence of colorectal cancer. And he surmised that fiber provided protection against colorectal carcinogenesis.

Again, plausible biology…dietary fiber and the fermentation of fiber in the gut can lead to increased weight, increased frequency of defecation, decreased transit time, and ultimately dilution of the colonic constituents, including potential mutagens and carcinogens. In addition, fiber is thought to change gut chloric and influence bioacid material metabolism, may have a direct effect on energy metabolism, can bind bioacids, produces short chain fatty acids like buterate which have been shown to be protective against colorectal tumorgenesis in animal models. Finally, fruits and vegetables – the third factor. And just a quick…this is a summary from that American Institute of Cancer research book where they reviewed many epidemiologic studies of fruits and vegetable intake, vis a vis colorectal cancer risk. And that they found that for physical activity, but for vegetables in particular, the evidence was "convincing". They rated the evidence in their survey of the studies. So what this reflects is many different epidemiologic studies showing that by and large vegetable consumption is protective against colorectal cancer.

Again, biology, the whole antioxidant hypothesis. Antioxidants are found in fruits and vegetables. The notion that there can be constituents in vegetables that an modulate metabolizing enzymes, that can be important in the detoxification of potential carcinogens, and direct effects on immune system or self-proliferation and apoptosis through constituents in fruits and vegetables.

What we did in the polyp prevention trial was not focus on a single nutritional factor, but instead to intervene with a comprehensive nutritional plan that involved those three hypotheses – fat, fiber, and fruits and vegetables. And the rationale for this was that there can be joint effects of nutrients in foods that would be missed if you were just focusing on a single nutrient or a single food. Second, there can be unknown nutrients, or non-nutrative constituents of food that we simply might be missing if we are again focusing on one known nutrient or food. And finally, trials like the one that we did are somewhat infrequent, can be very expensive, and we wanted to maximize the likelihood of demonstrating a positive effect in our trial. In essence we wanted to compare an overall dietary pattern like this one [picture of fatty foods] versus something like this one [picture of balanced mix of healthy foods].

Now these are polyps [photo] and this was a trial of polyps. Nobody really cares about polyps except for the fact that they turn into cancer. The rationale for why we looked at adenomas as our end point was that adenoma recurrence is substantially greater than the incidence of cancer. By recurrence I mean once one or more adenomenas polyps, these are neoplastic lesions, are removed from the large bowel, if another one develops, that is considered a recurrent lesion. And the rate of this recurrence is nearly two orders of magnitude, almost one hundred times greater than the incidence rate of colorectal cancer. Therefore when investigators go to set up a prevention trial with recurring adenomas as an end point, we can do it with a study that is substantially smaller than a study that would go to invasive colorectal cancer, in which case we would need a trial in the neighborhood of 50-100,000 people.

Second, because the standard of colodoscopic surveillance was such that at the time people were still one year and four years after the original polyp was removed, we could piggyback a trial onto the standard surveillance practice and have a way of evaluating the end points. Did people develop new adenomas or did they not? And finally, the underlying biologic reason for looking at adenomenas influence is the so-called adenoma-carcinoma sequence, or polyp cancer sequence. And again the notion that the overwhelming majority of invasive colorectal cancers are thought to develop from adenomas, from polyps.

This was the design of the polyp prevention trial. Relatively simple design. We randomized a little over 2000 people and for a polyp for an adenoma trial this was by far the largest one that has been conducted. And they were randomized into essentially two groups. One was the intervention group – low fat, high fiber, high fruit and vegetable eating plan or diet. And the other group received just very simple, general nutritional information and maintained their usual diet. We started recruiting people in 1991 and went for about two and a half years through 1994. This is the basic baseline procedure. Everyone who got into the trial had to have at least one or more adenomas removed within six months of randomization. People underwent colonoscopy at one year, and this at the time was part of the standard surveillance practice. The one-year procedure was very advantageous to us because it was an opportunity for the bowel to be pretty completely cleared. One of the things that gastrointerologists have found is that with one colonoscopy there are certain percentages of missed lesions. We reasoned that after this "clearing" procedure in one year, we could be pretty certain that the bowels were free of most adenomas. The problem with having missed adenomas is that it reduces our power to detect an effect. The main analytic period for the study was between one year and four years and at four years patients come back to their gastrointerologist, their endoscopist, or another colonoscopic procedure and at that point we can ascertain whether or not they had a recurrence of adenomas.

When you alter factors like fiber, fruits and vegetables, and fat, other parts of the diet are going to change as well. We were hoping that red meat intake would change because around the time when we were getting this trial going, the red meat and colon cancer hypothesis was emerging even stronger. And what you can see is that we did get in the intervention group, even though it was not an explicit goal, we did get a reduction of about 20% in red meat. We got an increase in whole grains of about 40% in the intervention group compared to the control group. And for cruciferous vegetables, we got almost a doubling of intake in the intervention group compared to the control group. So a variety of things changed.

These were the ultimate results that appeared in the New England Journal this past year. Lack of effect of a low fat, high fiber diet on the recurrence of colorectal adenomas. It’s a little like, for those of you who used to watch Saturday Night Live with the old Gilda Radner routine goes through and it’s the never mind kind of thing. After $30 million in 10 years, this was a profoundly null study. This 1 here reflects the relative risk of developing adenomas in the intervention group as compared to the control group if that relative risk were say a .5 it would mean that you reduce by half the recurrence rate. The 1 means that the recurrence rates were virtually identical in the two groups.

We were interested in whether we saw an effect of the nutritional intervention on large adenomas, or so-called advance adenomas. These are polyps that are large, have bad pathology villas components, or high grade aphasia. They are thought to be further along the pathway to cancer. In fact there was a smidge of a reduction in risk but it was no where near being statistically significant. So the study did not show an effect of this dietary change on any adenomas or large and advanced adenomas.

[Showing a Slide] This can be interpreted in a number of ways. It is sort of how the investigators felt after the data started coming in but it also suggests that you shouldn’t let your mind be hardened and concrete, that you need to be flexible in your thinking. So how then if we are trying to be flexible in our thinking but objective, how would we interpret the results of the polyp prevention trial and what implications does it have for the whole field of diet and cancer? Well, the straightforward interpretation is that dietary change has no effect on colorectal neoplasia, that is on adenoma development, and by inference (and that is the whole reason we did the study), dietary change has no effect on colorectal cancer. But we need to consider certain limitations of the polyp trial model, just as there are limitations in virtually all of the modalities of study, everything from animal studies to large perspective cohort studies. These studies are not perfect. They have limits. They have weaknesses.

First of all, the polyp trial model involves intervention for a relatively short period of time, 3-4 years. If in fact you need to intervene for 10 years or longer to really get an effect on adenoma recurrence, our study would not allow us to evaluate that. There is a biological aspect to this. What we were looking at is a limited component, a limited segment of neoplasia or carcinogenic process. We were looking at people who already had one polyp and they went on to develop a recurrent polyp. Most recurrent polyps are small. What that means is that if diet operates before the development of the first adenoma, or if diet operates in the growth of a small adenoma to a large one, or a large on to invasive cancer, again the polyp trial model (our study) would not be able to tell us that, would not be able to evaluate those possibilities.

And finally, there were some, although we tried as much as possible to develop an eating plan and a dietary intervention that would embrace multiple hypotheses, the nutrition side also has its potential limitations. Possibly people have to go down to 15% of calories from fat or they have to increase fruit and vegetable intake even more than the participants did in the polyp prevention trial. Maybe other factors are critical. Maybe its not fat and fiber. Maybe red meat is the real actor. We had some reduction of red meat intake, but only about 20%. Perhaps if a trial where people reduce their red intake by 50% or went off red meat entirely would show an effect that we did not see.

And finally there is some uncertainty about adherence. We can’t be absolutely certain that people made the changes that they did. We think they did. There was very close contact between the nutritionists and the participants and we did see some hard biomarker changes. We did, for example, see that in the intervention group the level of blood carodnoids went up significantly. There was weight loss in the intervention group compared to the control group. But there is some uncertainty to keep in mind. Therefore due to the limitations of the polyp trial model, it is still possible that dietary change reduces the risk of colorectal cancer.

So I would like to say that we now know exactly from the perspective of diet and colorectal cancer what’s good for you, what’s bad for you, what should be put out on the shelves…I think it’s fair to say that the evidence, even though our trial was null, and even though some studies have not shown the same findings that other studies have shown, it’s probably reasonable to yourself, or to advise people who ask, to reduce intake of red meat and saturated fat, to increase the consumption of whole grains and vegetables, and some people would argue about taking a folic acid pill (that’s I think a plus minus). But I think that kind of general dietary advice may, in the long run, if we do our job right, if we are lucky as well as smart, we may show that it, in fact, affects in a causal way colorectal cancer. But there is also very good evidence that those kind of dietary changes can affect other chronic diseases (cardiovascular disease and diabetes, for example). So even if making those kind of changes doesn’t do anything for carcinogenesis in the large bowel, it could do other good things for you and for other people. So we hope that one day we will really be able to nail what’s good for you so that in the long run we can learn how to eat right, live long, and prosper.

But I do want to retain the proper note of scientific skepticism. This [slide] says, "Yesterday in this space I predicted that cancer would come to an end. It did not however. I regret any inconvenience this may have caused you."

Thank you very much.