Dietary Pattern Research
Martin Root, Ph.D
November 14, 2000

People do not eat isolated foods or individual nutrients. We tend to eat our food in meals and our meals tend to come from a dietary pattern that we grew up with. Only when we make dramatic changes in our circumstances do we make big changes in our diet. This might happen by getting married or emigrating to another country or being told by our doctor that our diet is killing us. These diets or cuisines are one of the reasons traveling is so much fun. People in different regions eat very differently than we are used to and it is an adventure to try out new foods that we have never seen before. What is rather common to the natives is novel to us.

Dietary research tends to be reductionist in the sense that scientists like to determine the cause and effects of single foods or single nutrients in the interplay of health and disease. In the golden age of nutrition, in the early 20^th century when the essential nutrients were being discovered, this approach was highly successful. In the current age when we are more interested in the causes of chronic diseases and not nutritional deficiencies, the reductionist approach has limitations. We observe that chronic disease rates vary dramatically around the world and that cuisines vary similarly. It would make sense that if we wanted to determine the dietary cause of disease that we should consider cuisines and dietary pattern holistically in addition to the more traditional one nutrient at a time approach.

As intuitively simple as this idea is, there are many difficulties in this approach of studying dietary patterns and disease. However, in the past decade nutritional scientists have been approaching this concept with a number of inventive and informative research tools. I will present here a summary of these techniques with some of their strengths and pitfalls. I will then summarize some of this powerful research as it pertains to the prevention of the diseases of most Western societies, the vascular diseases and the cancers.

The technique with the broadest view is the cross-cultural study. This type of study compares both the average diet and the average disease rates across a number of regional or cultural groups. An example is the China Study, which compared the diets and disease rates in 65 rural counties of China. Unlike the US where we eat essentially the same foods in Seattle and Atlanta, across rural China there are a great variety of cuisines. Even the most basic staple grains vary from rice to wheat to corn and millet across China. Diseases rates, particularly for the cancers, also vary from region to region. Other such studies have compared heart disease rates and the related risk factors, including diet, across several countries. Since both diet and disease rates vary dramatically between regions and countries many diet-disease connections can be made. This helps give clues that can be used in more focused studies on specific risk factors of a disease. The trouble with this sort of study is that it is too easy to make connections. Looking at all the countries around the world it can easily be argued that such factors as neon lights or even democracy cause colon cancer. Since diet is only one of many things that vary from country to country it is hard to pick the true causes of diseases from the dead ends.

Another way to look at dietary patterns and disease is to simply ask people what kind of diet they eat. A number of studies, in addition to using thorough food questionnaire, asked if the subjects considered themselves vegetarian or vegan. They then followed these folks for a number of years and to see what diseases they died from. An interesting article last year summarized the results from 5 large studies of this kind in which their subjects classified themselves by their usual dietary habits into 5 groups: Regular meat eaters, Occasional meat eaters, Fish eaters, Vegetarians, or Vegans. They then summarized what diseases these folks tended to die from. They found that self-declared vegetarians died of heart disease 34% less often than regular meat eaters. One trouble with this type of study is that what motivates people to identify themselves with a particular diet may be in large part whim or fashion. Another problem is that people who are trying to eat healthier, such as vegetarians, are also doing many other things that contribute to health, like losing weight, not smoking, and exercising regularly. Was it the diet or the exercise that reduced heart disease in the vegetarians?

There are already many government recommendations and published guidelines for healthy eating. Some researchers define what they consider to be a healthy diet. They then determine whether subjects meet those criteria and then determine what they got sick with or died from. A recent analysis of the results of the long-term Nurses Study defined a healthy lifestyle that included a low saturated fat diet with lots of fruits and vegetables and included higher exercise levels and excluded smokers and overweight subjects. They found that women who adhered to this healthy lifestyle had 83% less heart disease. Researchers have used numbers of serving of fruits and vegetables, adherence to the American Heart Association "prudent diet", and other definitions of good and bad diets. One trouble here is that people never eat the same thing from day to day and have rather poor memories of what they have eaten in the recent past. So which group a subject falls into may depend on whether they remember the apple they had for lunch yesterday when they are recalling their diet for the researcher. This type of research also makes it difficult to determine if there is a diet even more prudent than the diet they defined.

There is a common weakness in all of these dietary pattern studies that is most obvious with this guideline group. The lesson that we would like to take home from studies like the Nurses Study is that if we change our lifestyle today from a high risk one to a low risk one then our risk of disease immediately changes by the estimated amount, 83% in this case. This unfortunately is not necessarily true since the people in the low risk category have probably been living that way for years and years. A person who changes to this lifestyle similarly should only expect to assume that reduced risk after living that way for years and years. As an example, it takes about 3 years for people who quit smoking to acquire the heart disease risk of a non-smoker and 10-20 years to reduce their risk of lung cancer to non-smoker levels.

The most statistically elegant method of dietary pattern research is called factor analysis. Essentially a researcher puts all of the dietary information of the subjects into the computer and lets a statistical procedure identify the common dietary patterns and who is eating how much of which patterns. One of the immediate problems with this technique is that since the computer defines the diet, we, as mere humans, may not recognize the computer pattern as anything we can put a label on. It is up to the scientist to label the diets. Another problem is that what comes out depends on what goes in. Every researcher, it seems, uses a different type of dietary questionnaire, and so it is often difficult to compare results between studies. What patterns emerge also depend on what country the research is done in. Two such studies were done recently in China and Japan. They were nice enough studies on their own but it is difficult to transfer what is learned there to what is needed in the US. The lure of factor analysis is that it is objective, not dependent on what a researcher thinks is healthy or not. It also tends to be more "stable" statistically since it uses all of the information available and is not thrown off by variations in single foods. A recent factor analysis from the Health Professionals Follow-up Study, a long-term follow-up of 45,000 male health professionals, found a dietary pattern among the subjects that the researcher recognized as "Western". It was high in meat and refined carbohydrates among other things. As the proportion of Western diet went up in subjects’ diets, the level of heart disease went up uniformly, reflecting both the stability of the dietary pattern and also its uniformly risky nature for heart disease.

The final method of identifying overall patterns in diets may seem contradictory. Individual nutrients or food components are measured and used as indicators of diet pattern in study subjects. A recent study of diet and lung cancer is a case in point. Total dietary carotenoids were calculated from the dietary records of the subjects and they were followed for 10 years. They found that those with the highest level of total carotenoids in their diets had the lowest lung cancer. Before getting too excited about the miraculous power of carotenoids please remember that foods that have carotenoids also have many other healthy nutrients in addition to a multitude of other phytochemicals, both helpful and harmful. In fact since carotenoids come almost exclusively from fruits and vegetables, the total carotenoid intake is actually a fairly accurate and stable estimate of fruit and vegetable intakes. What the study results probably mean at their most basic is that people who tend to eat diets with lots of green, red, orange, and yellow fruits and vegetables tend to get less cancer. But can you think of other characteristics of people who eat this way? These "confounders" (like no smoking and low body weights) also tend to be a problem here as they were in the self-defined vegetarian studies.

We have defined five ways of getting at the basic research question of what are people’s customary diets and what are the health consequences of them. There are strengths and weaknesses to each approach, but as a group these types of studies can help inform us and give us some direction toward healthy diets and lifestyles. Generally when considering diet as a whole we see that healthy diets tend to give healthy results. The results are not without problems but they do tend to be more stable and unidirectional than studies examining single nutrients and foods components. Here are some examples.

In the lung cancer study, where smoking was carefully accounted for, the researchers actually measured the intake of 5 different carotenoids. Their relation in protecting against lung cancer was marginally significant in part because each carotenoid comes from a small subset of foods. But when they were all combined, reflecting the intake of a wide variety of fruits and vegetables, the resulting effect on lung cancer was uniform, stable, and more easily interpretable. The message is to eat a wide variety of fruits and vegetable and you can cut you lung cancer risk by 30%.

In the study that found a "Western" dietary pattern that increased risk of heart disease, another more common pattern appeared to be a "Prudent" diet with lots of legumes, fruits, cruciferous vegetables, fish, and poultry. When these guys were ranked in how prudent their diet was, the top fifth had 30% less heart disease than the lowest fifth. Again the message is more consistent and clear. Don’t just add broccoli or All-Bran to your diet and get some uncertain improvement in health. Go all the way from Western to Prudent and get all the benefits.

Another group of investigators used the Food Pyramid to help define a Recommended Food Score (RFS) that reflected how closely a person kept to the dietary recommendations to eat "5-a-day" of fruits and vegetables and eat low-fat dairy and meats. They followed 42,000 women for over 5 years and compared the RFS of the 2000 women who died during that time with those that were still alive. When compared across the range of RFSs, there was a uniform decrease in all-cause mortality among women at each of 3 groups that had increasing healthy diets. Another message here is that a person does not have to adopt some extreme lifestyle suddenly. Every incremental change in total diet (and lifestyle) is rewarded by increased health. The downside, especially when using the Food Pyramid as your ruler, is that you have no idea how much better it could get if you went beyond the very modest recommendations in these FDA guidelines.

The studies that are most dramatic in their results are, unfortunately, the most difficult to draw usable recommendations from. These are the cross-cultural studies. One of the first of these studies compared the breast cancer mortality rates and fat intake among a handful of nations. The results were astounding. Clearly the nations with the lowest fat intake like Japan and Nigeria had the lowest breast cancer rates and countries like the US and Sweden with high fat intakes had high breast cancer rates. The message seemed deceptively straightforward—Eat less fat and get less breast cancer. Unfortunately when looking at women within a given country where other cultural factors are more evenly distributed, dietary fat seems to make little difference. The message from international studies should be more exactly stated as, be like Nigerians or Japanese to reduce your cancer risk. However, what exactly these other cultures are doing that is so right is very hard to determine.

Does this research support a vegetarian or vegan dietary pattern? The answer is, not as strongly as some of us would like. Vegans in particular, do not make up a very large proportion of the study populations used in these studies. Therefore this diet is not singled out as being uniquely healthful compared to other vegetarian or low-meat diets. Also, vegetarians as a group do many other healthy things that make it difficult to isolate their diets has the key healthy component.

The good news is that there are healthy diets and cuisines out there. Even without consulting nutritional tomes or reading the latest health news, many folks are naturally eating good foods everyday and enjoying the health benefits. You don’t have to ask yourself how much one more serving of broccoli or 10 more minutes on the treadmill will improve your health. If the change constitutes a move from an unhealthy dietary of lifestyle pattern to a healthy pattern then the scientific evidence is behind you. You already know the basics, more fruits and veggies, less read meat and smoking, and more exercise. Its just a matter of putting it all together.

Discuss this aritcle on the VegScience board.

Marty Root has a PhD in Nutrition from Cornell University, and works as a Senior Research Scientist at BioSignia, Inc. in Chapel Hill, NC. His work involves making statistical models that predict the onset of chronic diseases such as heart disease and the cancers, and he is a frequent contributor to VegSource.