American Journal of Clinical Nutrition, Vol. 69, No. 6, 1365S-1367S,
June 1999
© 1999 American Society for Clinical Nutrition
Summary statement 1
Tim Byers,
Barbara Lyle and
Workshop Participants
 |
ABSTRACT
|
|---|
This statement summarizes the key points of discussion among a group of nutritional epidemiologists who met in Washington, DC, for 2 d in October of 1997 to reflect on the role of nutritional epidemiology in the development of dietary recommendations for the public. Although imprecision in the measurement of diet places limits on nutritional epidemiology, no other field of nutritional science can provide direct information on relations between nutrition and health in free-living human populations. Among the nutritional sciences, therefore, epidemiology was regarded as being critically important. Nutritional epidemiology can be improved in the future by the development of more precise measures of long-term dietary exposures, both by improved methods of self-reporting of diet and by the development of more useful biomarkers of long-term nutritional status. There is a need as well to reconsider the applicability of causal criteria as applied to nutritional epidemiology, because many of the important associations between dietary behaviors and chronic diseases cannot necessarily be expected to be either strong or to manifest linear dose-response relations. In the future, scientific evidence from the rapidly growing field of nutritional epidemiology will likely play an increasingly important role in developing nutrition policy and advice for the public.
Key Words: Nutrition • epidemiology • methodology • diet-disease relations • causal criteria • dose-response relations
|
RATIONALE FOR THE WORKSHOP
|
|---|
The public is increasingly interested in how food choices can influence long-term health. As scientific evidence showing that diet influences health continues to mount, demand for rational and consistent advice about foods and nutrients that best promote health is increasing. To provide such advice to the public, we must determine whether the scientific evidence is sufficient to support specific recommendations. The difference between advocacy and scientific objectivity in this task can be subtle. It is therefore important that we understand how to use and apply nutrition information obtained from all methods of inquiry, especially from nutritional epidemiology, which attracts particular attention from the media.
Nutritional scientists from the field of nutritional epidemiology and from agencies and companies that use data from epidemiologic studies met for 2 d in Washington, DC, in October of 1997 to consider the role of epidemiology in evaluating relations between nutrition and health. Workshop participants discussed past lessons, current challenges, and future possibilities for the role of epidemiology in contributing to the body of scientific evidence needed to make sound nutrition recommendations, and these discussions are included in this supplement
(
1
1
–
9
)
. In this summary statement, intended to complement the supplement articles, the major issues, questions, and conclusions discussed during the workshop are summarized.
|
ROLE OF NUTRITIONAL EPIDEMIOLOGY
|
|---|
Nutritional epidemiology is critically important to our understanding of relations between nutrition and health; it is the only method of nutritional science that provides direct information on relations between nutrition and health in human populations consuming usual amounts of food and nutrients. This information is particularly important for our understanding of chronic and degenerative diseases, which cannot often be appropriately modeled by animal studies or short-term human clinical trials. However, knowledge gained from epidemiologic studies in human populations must be integrated with that gained in other disciplines of scientific investigation, such as animal, in vitro, and clinical studies. Epidemiologic studies thus provide a critical link in the causal chain of inference between diet and human disease. Although epidemiologic findings should not be disregarded in the absence of supportive data from the laboratory, uncorroborated findings do suggest that additional information may be needed, particularly when findings differ between observational and experimental studies.
|
STUDY QUALITY: CHALLENGES AND OPPORTUNITIES
|
|---|
Epidemiologic research can meet high scientific standards and rigor when individual investigators use the best available estimates of dietary intakes, study designs appropriate for the relation being studied, and sound statistical techniques. Whether properly validated measures of dietary intakes and appropriate statistical methods were used to rigorously evaluate relations with specific diseases should be considered when evaluating the quality of findings from individual epidemiologic studies. No single problem has been more formidable to nutritional epidemiology than the measurement of dietary intakes. Quantitative biochemical indicators of nutritional status can provide useful information regarding nutrition and health, but they cannot replace the assessment of usual dietary intake. Therefore, we must continue to improve diet assessment methods and food composition databases and to use discipline to avoid the temptation to overinterpret findings based on subjects' reported dietary intakes. Moreover, accurate assessment of the relation between diet and disease requires a consideration of changes in both dietary habits and food composition that occur over time. This is true both for retrospective studies, in which dietary habits earlier in life might be relevant to disease risk, and for prospective studies, in which repeated measurements in diet might be needed.
Another critical need is to develop better biological indicators of nutritional status and dietary intakes as a means of obtaining more powerful and unbiased estimates of the relations between nutrition and health. In addition to using biomarkers of exposure, intermediate endpoints of disease can provide useful information about how dietary factors might play a role in the biochemical or physiologic changes that initiate disease processes. At this time, however, caution is needed when using intermediate endpoints because of our incomplete understanding of the mechanisms of etiology of many chronic diseases.
Opportunities to evaluate associations between diet and disease in specific population subgroups, such as persons who smoke, have diabetes, or have specific genotypes, will increase over the next decade, because larger study populations and analyses of pooled prospective data are being used. In addition, dietary data can be presented and analyzed in several different ways at the discretion of the investigator (eg, analyses by foods, food groups, or nutrients, with or without adjustments for other dietary factors). These multiple methods and subgroup analyses will create the need for authors to describe findings from future epidemiologic studies in ways that fully define all possible analyses that could have been presented.
|
USEFULNESS OF CURRENT CRITERIA FOR INFERRING CAUSALITY IN NUTRITIONAL EPIDEMIOLOGY
|
|---|
Observational studies illuminate correlations between nutritional status and disease but do not prove that associations are causally related. The likelihood that an observed association is causally related requires a higher level of inference. Criteria used by epidemiologists to infer causality are most frequently adapted from those established in the 1960s by Bradford Hill (
10
), ie, consistency, strength, temporality, dose response, and biologic plausibility. Workshop participants expressed concern that many of the current criteria are problematic when applied to the evaluation of nutritional epidemiologic findings, and that work on the design of more useful guidelines for inferring causality and evaluating associations between nutritional status and health is needed. The following are examples of concerns about current criteria for causation.
Consistency of evidence
Although some single studies will carry more weight than others, it is rare that any single study is considered definitive. The power of the epidemiologic approach lies in observing similar findings consistently from a large set of diverse studies that address specific relations. It is difficult to make a case for a causal association without the results of most studies being in agreement. Some inconsistency can be expected, however, within a large set of studies that are influenced to various degrees by sampling and measurement errors of the types known to be present in nutritional epidemiology. Although inconsistencies may indicate that an association is not causal, they also provide an opportunity to learn about diet-disease relations by inferring reasons for discrepancies between studies that differ in known ways, such as in the populations, the various exposures to dietary factors, or the differing times frames in which the studies were conducted. More attention must be given to the critical analysis of inconsistencies between studies when scientific reviews and pooled analyses are undertaken.
Strength of association
The criterion of strength of association is problematic in nutritional studies because of the frequent occurrence of measurement errors. A 20% increase or decrease in risk related to a food or nutrient could be both biologically plausible and have a meaningful effect on public health as a result of a high degree of exposure, yet may well have resulted from a subtle bias or poor control of confounding factors. The challenge, then, is to distinguish the moderate effects that are real and have important public health consequences from those that are artifactual. A change of 50% in risk (eg, a relative risk of 1.5) may be considered strong in nutritional epidemiology, so the commonly cited criterion of relative risks <2.0 being suspect does not necessarily apply to nutritional epidemiology.
Dose response
The presence of a statistically significant linear or otherwise regularly increasing trend generally reinforces the evidence in favor of a causal association, but such a relation may not be seen in dietary data. The relation between nutrient intake and disease risk may have distinct thresholds, in which case the relation between disease risk and diet consumed in populations would not necessarily be expected to be linear.
Biological plausibility
In situations in which an a priori hypothesis of a nutrient-disease association is linked with an established biological mechanism, evidence of the association in an epidemiologic study can be called biologically plausible. However, unanticipated associations often emerge from epidemiologic studies. Because biological relations are complex, biological plausibility can be invoked in support of virtually any observation. Biological plausibility is therefore often too easy to argue as part of the post hoc explanations and should not be overly regarded in drawing causal inference.
Foods versus nutrients
The strength of epidemiologic data from observational studies lies in providing information on associations between health and foods as actually consumed. The inference of causal relations between health and nutrients requires more evidence of the specificity of the association. One of the problems is that the databases used currently for many nutrients and constituents of foods are inadequate. Epidemiologists need up-to-date information on the composition of food products, particularly those new to the marketplace. It would be helpful if food manufacturers would routinely make more detailed information available about food composition in addition to the information on the product label.
Associations between specific nutrients and health can be provided from clinical trials when the health outcomes are influenced over a short period of time. However, because of the limited generalizability of information provided by clinical trials (eg, limited dose, specific form of the nutrient, and specific subpopulation), we must often make recommendations based principally (and sometimes solely) on observational studies. The lesson of ß-carotene and lung cancer—in which an inferential leap was made from studies observing associations with foods to drawing conclusions about nutrient-specific causation and prevention—needs to be remembered. There is a need for caution in formulating recommendations that apply to specific nutrients on the basis of results from observational studies of foods.
|
NEED FOR METHODOLOGIC DEVELOPMENT
|
|---|
Methodologic advances are needed if our ability to accurately assess the foods we eat and evaluate the effects of measurement error on relations between intake estimates and disease outcomes are to improve. Investigators must be encouraged to conduct such studies, and funding agencies must be encouraged to recognize the importance of the advances that will come from well-designed methodologic studies. Greater resources must be appropriated for the conduct of methodologic studies that will advance methods for assessing and interpreting relations between diet and disease. Investigators should conduct analyses to evaluate the actual, rather than the theoretic, effects of measurement errors in assessing diet on the estimation of risks for disease attributed to foods, food groups, and components of foods. Finally, more work is needed in developing and evaluating analytic options to control or adjust for measurement error (eg, using energy adjustment techniques) and for disentangling complex nutrient-nutrient correlations.
In addition to identifying associations between diet and specific health conditions, epidemiologic studies can provide information about the overall importance of the associations from a public health perspective. Recommendations to increase or decrease the consumption of certain foods can be strengthened when epidemiologic evidence suggests protective influences on a wide variety of diseases. Furthermore, epidemiologic strategies can be used to evaluate the proportion of disease in the population that may be influenced by dietary changes. In making dietary recommendations, we need to consider the possibility that specific diet components or patterns may affect a wide range of possible health outcomes. Some food choices may decrease the risk for one disease while increasing the risk for another, in which case techniques of computing attributable risk can help evaluate the net overall benefits to the population.
|
A GROWING ROLE FOR NUTRITIONAL EPIDEMIOLOGISTS
|
|---|
When evaluating links between nutrition and health, well-conducted epidemiologic studies can add greatly to the scientific evidence garnered from animal, in vitro, and human clinical studies. The large number of epidemiologic studies currently underway suggests that scientific evidence from nutritional epidemiology may play an increasingly important role in developing nutrition advice for the public. Judgments about study quality and interpretation of data from epidemiologic studies are complex issues and require specialized expertise. As a consequence, nutritionists with expertise in epidemiology are needed on scientific review committees and committees formulating nutrition guidance [eg,
Dietary Guidelines for Americans
(
11
) and the upcoming dietary reference intakes from Institute of Medicine]. Within each of these committees, a collaborative process should take place to evaluate the totality of evidence, including that from epidemiology along with other types of studies, in determining when evidence is sufficient to support specific nutrition recommendations.
|
FOOTNOTES
|
|---|
1 Reprints not available. Address correspondence to T Byers, Department of Preventive Medicine and Biometrics, University of Colorado School of Medicine, Campus Box C-245, 4200 East Ninth Avenue, Denver, CO 80262. E-mail:
Tim.Byers{at}uchsc.edu.
|
REFERENCES
|
|---|
-
Byers T. The role of epidemiology in developing nutritional recommendations: past, present, and future. Am J Clin Nutr
1999
;69(suppl)
:1304S
–8S.[Abstract/Free Full Text]
-
Potischman N, Weed DL. Causal criteria in nutritional epidemiology. Am J Clin Nutr
1999
;69(suppl)
:1309S
–14S.[Abstract/Free Full Text]
-
Freudenheim JL. Study design and hypothesis testing: issues in the evaluation of evidence from nutritional epidemiology. Am J Clin Nutr
1999
;69(suppl)
:1315S
–21S.[Abstract/Free Full Text]
-
Kushi LH. Vitamin E and heart disease: a case study. Am J Clin Nutr
1999
;69(suppl)
:1322S
–9S.[Abstract/Free Full Text]
-
Sempos CT, Liu K, Ernst ND. Food and nutrient exposures: what to consider when evaluating epidemiologic evidence. Am J Clin Nutr
1999
;69(suppl)
:1330S
–8S.[Abstract/Free Full Text]
-
Flegal KM. Evaluating epidemiologic evidence of the effects of food and nutrient exposures. Am J Clin Nutr
1999
;69(suppl)
:1339S
–44S.[Abstract/Free Full Text]
-
Albanes D. ß-Carotene and lung cancer: a case study. Am J Clin Nutr
1999
;69(suppl)
:1345S
–50S.[Abstract/Free Full Text]
-
Marshall JR, Chen Z. Diet and health risk: risk patterns and disease-specific associations. Am J Clin Nutr
1999
;69(suppl)
:1351S
–6S.[Abstract/Free Full Text]
-
Lewis CJ, Yetley EA. Health claims and observational human data: relation between dietary fat and cancer. Am J Clin Nutr
1999
; 69(suppl)
:1357S
–64S.[Abstract/Free Full Text]
-
Hill A. The environment and disease: association or causal? Proc R Soc Med
1965
;58
:295
–300.[Medline]
-
US Department of Health and Human Services, US Department of Agriculture. Nutrition and your health: dietary guidelines for Americans. 4th ed. Washington, DC: US Government Printing Office, 1995.
This article has been cited by other articles:
|
|
|
|
 
J. B. Wolf
Is Breast Really Best? Risk and Total Motherhood in the National Breastfeeding Awareness Campaign
Journal of Health Politics Policy and Law,
August 1, 2007;
32(4):
595 - 636.
[Abstract]
[PDF]
|
|
|
|
|
|
|
V. Moore
Commentary: Biological Freudianism and the quest for understanding of the social origins of health
Int. J. Epidemiol.,
February 1, 2005;
34(1):
18 - 20.
[Full Text]
[PDF]
|
|
|
TOP
ABSTRACT
RATIONALE FOR THE WORKSHOP
