Nutrition 331: Nutrition for Health
Study Guide: Unit 2
General Principles of Research in Nutrition
Nutrition is a rapidly expanding science. The last 40 years have seen an explosion of knowledge in the area. Often, though, today’s “discovery” is tomorrow’s discredited idea. For example, the role of dietary factors affecting colon cancer (fat), the common cold (vitamin C), and hyperactivity (food additives) have been both “proven” and “disproven.” Why is there such confusion? The perennial problem of nutrition is that it is not an exact science: rarely does it have the precision of physics or chemistry. Nutrition science is often like a mystery, where clues are gathered from many sources.
In this unit, we look at how nutrition research is conducted to enable you to make your own assessment of claims and to sort the wheat from the chaff.
The main part of this unit explains different nutrition research methods, including a discussion of the advantages and disadvantages of each method.
After completing this unit, you should be able to
- describe types of studies used in establishing the role of diet in disease and their advantages and disadvantages. These studies include epidemiological studies (including population, historical, case-control, and cohort studies), intervention studies, animal studies, biochemical and physiological studies, as well as anecdotal evidence.
- determine whether each type of study shows causality or association, and describe the potential for incorporating study conclusions into nutrition recommendations.
- discuss at least five features of a well-designed intervention study.
- assess nutrition information for reliability.
Section 1 Intervention Studies
Read pages 12–15 of the section titled “The Science of Nutrition” in Chapter 1: “Food Choices and Human Health.”
Intervention studies are familiar to most people. They are based on an experimental design used in all areas of science. The key features of intervention studies are shown in Figure 1.3 of the textbook (p. 13) with additional details on p. 14. When used in areas related to health and medicine, they are commonly known as randomized controlled trials (RCTs). They are widely used for testing new drugs.
In intervention studies, volunteers are instructed on changes they must make to their diets. Typically, there are two groups: an intervention group and a control group. The intervention group is instructed to make a single dietary change. The members of the control group continue with their regular diet. Because of the inclusion of a control group, this type of study is called a controlled trial. After an appropriate period of time, the two groups are compared. Any differences between the groups should be the result of the single intervention. To reduce the possibility of changes being attributable to random chance, researchers apply the following safeguards:
1) Test groups must be large enough to reduce the effect of random chance influences (usually 30 or more subjects per group); 2) to help prevent bias, subjects are randomly placed into the groups; and 3) the two groups are compared at the start to make sure that they are similar with respect to factors relevant to the study. For example, if the effect of diet on weight loss is being tested, then it is important that the members of the two groups are of similar weight. Generally, the two groups would also be matched for such factors as sex, age, and ethnicity.
The phenomenon of the placebo effect has been demonstrated repeatedly. If people with a particular illness are given an inert substance (a placebo) and are told that it is a medicine, many will report that they feel better. To overcome this phenomenon, the control group in an intervention study is normally given a treatment which, as nearly as possible, resembles the active treatment. To test the effect of sugar on energy intake, for example, the intervention group might be given drinks with sugar, while the control group will be given drinks of similar taste and appearance containing synthetic sweetener.
To overcome the bias of investigators, it is important that, whenever possible, both the patient and the investigator are “blind” as to whether subjects are receiving an active treatment or a placebo. Such a study is described as double-blind. If the investigator expects that a particular treatment will reduce pain, the investigator may “observe” a reduction in pain, even if the treatment is ineffective. One obvious way this bias can happen is through subjective observations; for example, the investigator might easily downgrade the severity of headaches in migraine patients being given a new treatment. Furthermore, the investigator may unintentionally convey his or her own biases to the patient. As a result, the patient apparently improves with the active treatment, not because it was more effective than the placebo, but because the investigator was more enthusiastic about it. In effect, the active treatment was given an extra dose of placebo effect. Double-blind studies are standard—the gold standard—in drug trials or when nutritional supplements are given.
In experiments involving major dietary changes, “blinding” of the subjects isn’t always possible. However, even then the subjects can be assessed in a blinded fashion. For example, when testing the effect of a low salt diet on blood pressure, the patient will know what diet he or she has been consuming, but the investigator who records the blood pressure should not, as measuring blood pressure is subjective to some extent. Likewise, in case-control studies, when a patient’s diet is assessed, the investigator should be unaware of the patient’s diagnosis.
Often, intervention studies are done without the inclusion of a control group. For example, a group of patients may be told to try a new treatment to see whether it is of benefit. Such studies are termed “uncontrolled” and are much less reliable than controlled studies, because we can never be sure whether the subjects would have responded in the observed way had they been given no treatment. One obvious source of such an error is the fact that many conditions clear up without treatment. Another source of error is in the assessment of subjects. For example, if people have their blood pressure measured on successive days, it will often appear to be going down. The stress of being in a physician’s office to have blood pressure measured often causes blood pressure to go up, an effect that subsides after a few such measurements. Hence, a poorly planned, uncontrolled trial could easily demonstrate that almost anything lowers blood pressure.
Intervention studies can be carried out on healthy subjects with the aim of preventing disease or on sick people to test a treatment. The latter type of intervention study is known as a clinical study (or clinical trial).
Intervention studies are of great value in showing that there truly is a cause-and-effect relationship between a nutrition variable and a disease.
Section 2 Anecdotal Evidence
The collection of anecdotal evidence is both the simplest and the least reliable avenue for studying the role of diet in disease. In essence, the investigator makes inferences from the experiences of one or more individuals. The textbook refers to this type of study as a case study.
The following example illustrates where the method can go wrong. A woman smokes during a pregnancy and gives birth to a healthy baby who grows up normally. Another woman does not smoke and gives birth to a baby who has Down’s syndrome. These experiences could suggest that smoking prevents Down’s syndrome. The problem is that taking isolated cases can be dangerously misleading. To be considered reliable, the same observation should be made repeatedly.
Often, anecdotal evidence can lead to important findings, as the following example illustrates. Many people in the United States noticed that they got headaches with specific symptoms after eating in Chinese restaurants. The cause was found to be monosodium glutamate (MSG), a commonly used food additive. The condition is known as MSG symptom complex (see p. 546 of the textbook). On this occasion, anecdotal evidence led to a reliable conclusion because there were many similar reports describing an unusual condition.
Anecdotal evidence has specific advantages and disadvantages. Its outstanding advantage, as in the case of headaches caused by MSG, is that completely new and previously unsuspected relationships can be discovered. Side effects from new drugs, for example, are often uncovered when many reports based on individual patients are collated and a distinctive pattern emerges.
The disadvantage of anecdotal evidence is its inherent unreliability: all types of errors can occur. First, there is chance. For example, it sometimes happens that several workers exposed to a particular chemical get the same type of cancer. Chance is always a possible explanation in such cases. Second, there is bias in the observation. As a result of the power of suggestion, people are often unreliable in describing their symptoms, particularly in the case of such symptoms as headaches and vague pains.
Because of these problems, anecdotal evidence is only accepted with much caution. In short, it provides suggestions for further study, but is seldom the basis for firm conclusions.
Advertisers may use anecdotal information to promote products. Anecdotal promotional statements are often referred to as testimonials. Companies promoting weight loss diets, foods, or supplements often rely on personal stories. Be skeptical of advertising that relies on anecdotal evidence.
Section 3 Introduction to Epidemiology and Population Studies
Epidemiology is the study of the incidence and distribution of diseases and of their control and prevention. Researchers simply observe what people do and whether or not they develop diseases. The four types of epidemiological studies are
- case-control, and
The textbook (Fig. 1.4, p. 14) discusses epidemiological studies.
Population studies compare disease patterns with various factors, including food consumption, on a country-wide basis. We will use a hypothetical example to illustrate what a population study is and to describe some of the limitations of this study method.
Hypothetical example: We wish to know the relationship between potato consumption and cancer of the pancreas. We take the data available on the quantity of potatoes eaten in countries A and B and compare it with the incidence of cancer of the pancreas. We observe that people in country A eat twice as much weight of potatoes as those in country B, and also have double the incidence of pancreatic cancer. On the surface, these data suggest that potatoes cause cancer of the pancreas.
There are two problems, however, with concluding that potatoes cause pancreatic cancer. First, there could be errors in the data. The second, more common limitation, is the presence of confounding variables.
Errors in the data: A closer examination of the data reveals that people in country A are well-off and waste half of their potatoes, whereas people in country B are poor and fail to diagnose half of their cases of pancreatic cancer. In other words, people in both groups eat a similar amount of potatoes and have a similar incidence of pancreatic cancer.
Confounding variables: With a little more research, we learn that people in country A typically have a high prevalence of obesity, but in country B, obesity is very uncommon. The confounding variable (obesity) might be the factor (rather than the potatoes) leading to more pancreatic cancer. Confounding variables occur with the factor under study and may mask the real cause of the disease. Because many factors tend to appear together in a diet or a lifestyle, it is quite easy for some confounding variable to be the culprit instead of the identified factor.
With a population study, researchers can demonstrate association: as one factor changes (e.g., potato consumption increases), the disease variable may change (e.g., pancreatic cancer). We must stress, however, that the fact that two factors are associated does not mean that one caused the other. Only intervention trials demonstrate causality (or cause and effect).
Despite these sources of possible error, population studies can sometimes yield valuable findings. The probable benefit of the Mediterranean diet was discovered when it was realized that populations of countries that border the Mediterranean have much lower rates of heart disease than do populations in countries of Northern Europe.
Section 4 Historical Studies
A second type of epidemiological study, the history of a disease, can often reveal important clues about the cause of the disease. Historical studies are useful when there are accurate records available of the number of people who have died from or have a specific health problem. We can illustrate this by the relationship between cigarette smoking and lung cancer. Lung cancer was rare at the time of the First World War (1914–1918), but it became progressively more common among men, and is now the leading cause of death from cancer in men. This trend perfectly matches the increase in smoking. On the other hand, the massive increase in lung cancer experienced by women has occurred since the 1970s. This pattern is consistent with the fact that women adopted the smoking habit much later than men. A current example of a disease history is the increasing rate of obesity and type 2 diabetes with growing affluence in China.
With respect to their advantages and disadvantages, historical studies are best compared with population studies. Historical studies have the advantage of being based on large populations, and they often involve large or even massive changes in disease incidence. However, disease statistics collected 60 or more years ago can often be unreliable. A more important disadvantage is that confounding variables often give rise to spurious relationships. Thus, there is the ever present danger of guilt by association. For example, any disease that has increased in incidence in the 20th century will be correlated with any lifestyle factor that has also changed; hence, the huge rise in heart disease in the 20th century could be blamed on factors as varied as television, cars, sugar, and soft margarine. Accordingly, historical evidence shows association between factors, but it is not definitively conclusive. We can, however, use historical evidence as a valuable clue to support a theory.
Section 5 Case-control and Cohort Studies
Case-control and cohort studies are other types of epidemiological studies. In each case, we look at the individuals within a single population, rather than making comparisons of different populations. Case-control and cohort studies attempt to explain why particular individuals in a population suffer from certain diseases, on the basis of their exposure to factors thought to be related to the disease.
In a cohort study (also known as a prospective study), a large number of subjects are recruited. Normally, the subjects are all in good health, especially with respect to the disease under study. They are questioned or examined to determine each person’s exposure to the particular factors under study. With diet studies, exposure is usually determined either by a 24-hour recall or by a food frequency questionnaire. The investigators then keep track of the subjects over subsequent years until enough of them have developed the disease of interest. The investigators then compare those who developed the disease to those who did not, particularly with respect to exposure to the possible disease-causing factors.
One example of a cohort study is the Nurses’ Health Study in the United States. In 1980, 80,000 female nurses completed a survey on their health status and lifestyle, including a food frequency questionnaire. After 10 years of follow-up, they were surveyed again with similar questions. The researchers acquired a gold mine of data about the relationship between many diseases, such as cancer and heart disease, and exposure to lifestyle factors.
A case-control study is similar to a cohort study, except that recruitment of subjects and data collection are done after the disease has developed. Two groups of subjects are recruited, one consisting of individuals suffering from the disease of interest, and the other (the control group) free of the disease. As the data are collected after the disease has developed, case-control studies are sometimes known as retrospective studies.
An example of a case-control study is a study in which men diagnosed with colon cancer (the cases) are asked to complete a lifestyle survey, including a food frequency questionnaire. “Controls” are also recruited and asked to complete the same survey. Controls would be men of similar age and economic background who do not have colon cancer.
An important source of error in cohort and case-control studies is the accuracy of diet assessments. The methods used (both 24-hour food recall and food frequency questionnaire) are prone to significant error due to inaccuracy in people’s ability to accurately remember what they have been eating.
A major challenge in research on lifestyle and disease risk is that most dietary and lifestyle factors are associated with one another. So, for example, a person who consumes plenty of vegetables also consumes a lot of fibre and nutrients, which are found in vegetables. People who value their health often eat plenty of vegetables, and they exercise and choose not to smoke. This makes it difficult to confidently identify which factors are truly responsible for causing or preventing particular diseases. These are examples of confounding variables (see Section 3 of this unit).
One rule that applies to all types of epidemiological studies is that they show association, not causation. In epidemiology, we conclude by saying that diet factor A is associated with disease B. But we cannot with certainty say that diet factor A causes disease B. Intervention studies are carried out so that we can move from association to causation.
Section 6 Animal Experiments and Studies of Disease Mechanisms
Animals are of great importance in investigating the causes of disease. In a typical experiment, the researcher will investigate the effects of different treatments on disease occurrence or on some other parameter relevant to health.
In many ways, animal experiments resemble intervention studies. The researcher will divide the animals into groups. Each group will receive a different treatment, but otherwise the groups will be treated identically. A control group will receive no treatment. At the end of the experiment, the animals will be examined to determine the effect of the treatment.
Note: Research using animals is governed by strict ethical regulations.
From the point of view of the researchers, conducting experiments on animals has two major advantages over conducting experiments on humans.
- Animal experiments can be carefully controlled. Because all aspects of the environment and diet are carefully controlled, there is very little difference between the experimental groups other than the experimental variable under study. In contrast, even the most carefully conducted intervention study on humans has little control over such factors as genetics and subject history. For practical reasons, most experiments on humans must be done on persons who live at home. People may be given instructions as to the changes they should make in their diet or other aspects of lifestyle, but there is no guarantee they will follow these instructions.
- Animal experiments permit testing of chemicals at potentially toxic levels, allowing researchers to examine the effects on various tissues post mortem. One example has been to feed animals very high fat diets to determine the effect on their arteries.
Animal experiments do, of course, have inherent drawbacks. The principal drawback is that the results from experiments on animals may not apply to humans. For example, most chemicals that cause cancer in humans also do so in animals, but about 10% do not. Much caution must therefore be used when extrapolating from animals to humans.
Another approach to studying the causes of disease is to examine the biochemical and physiological mechanisms that lie behind the disease. There is a vast amount of research activity in this area. For example, most research on cancer consists of studies of this type.
Section 7 Assimilation of Nutrition Research
We must stress that no single study or type of study is completely reliable. Rather, establishing the “facts” in nutrition is based on two key criteria: reproducibility (different research groups report essentially the same findings), and consistency in the finding derived using different methods (e.g., findings from population studies can be confirmed in both cohort studies and animal experiments).
The role of smoking in lung cancer illustrates how several methods can be used together to produce a convincing conclusion. As we discussed earlier, historical evidence supports a role for smoking in lung cancer. On its own, however, such evidence is weak. The most powerful evidence has come from case-control and cohort studies, which have consistently shown that smokers have a far higher risk of lung cancer than non-smokers. Moreover, such studies also show a dose-response relationship (i.e., risk increases with the number of cigarettes smoked). Animal studies support these observations: chemicals extracted from cigarette smoke induce cancer in animals.
Another example is the relationship between salt intake and hypertension (high blood pressure). Both case-control and population studies show a clear association between salt intake and risk of hypertension. Intervention studies show that when people reduce their salt intake, blood pressure falls. Historical evidence supports this conclusion: the fall in the mortality rate from stroke since the 1950s occurred simultaneously with a large drop in salt intake. Lastly, we have animal studies. Some strains of rats develop hypertension when they are fed a high salt diet. Taken as a whole, this evidence leaves little doubt that salt plays an important role in hypertension.
In conclusion, nutrition research is like a mystery. Nutrition researchers must gather clues from a variety of sources, and to draw conclusions, all the evidence must point in one direction. Along the way, there may be a few “red herrings.” In the end, there should be enough evidence to support a nutrition recommendation.
Section 8 Appraisal of Nutrition Information
From Chapter 1, “Food Choices and Human Health”:
Page 16 “Can I Trust the Media to Deliver Nutrition News?”
Page 18 “Consumer Corner: Reading Nutrition News with an Educated Eye.”
Pages 21–27 “Controversy 1: Sorting the Imposters from the Real Nutrition Experts.”
The text readings discuss reliable and unreliable sources of information. The previous section discussed how research is carried out. The findings of such research are published in peer-reviewed journals of which there are thousands. Prominent journals include the New England Journal of Medicine, the American Journal of Clinical Nutrition, and the Canadian Medical Association Journal. Journals such as these maintain high academic standards and only publish papers after they have been carefully checked by experts in the area. This process is known as peer review. These journals also publish reviews and opinion articles. This Study Guide refers to several dozen reliable papers that were published in peer-reviewed journals as well as to various websites that were written by experts. These references are listed at the end of the Study Guide.
Everywhere we turn, we are bombarded with nutrition information. Magazines and newspapers regularly report the hottest new information. Bookstores sell books on nutrition, most of which make extravagant claims. Much of this information is accurate, but a large part is unsubstantiated or even blatantly untrue.
Health food stores are notorious for selling books and magazines full of misinformation (food faddism or nutrition quackery).
It is important to have a healthy level of skepticism with nutrition information found in the media, books, or websites. You are strongly encouraged to check the credentials of the nutrition expert.
As the text discusses under “Who are the True Nutrition Experts?” (pp. 24–25), dietitians are a good source of reliable nutrition information. In Canada, registered dietitians have different professional designations, depending on the province. Look for R. D., P. Dt., R. Dt., or in Quebec, Dt. P. Different provinces also have different protected names. For example, in Alberta registered dietitian, registered nutritionist, and dietitian may only be used by professionals registered with the College of Dietitians of Alberta.