Evidence-based nutrition. (Editorial).

Evidence-based medicine is one of the concepts that built up in the 1990s, along with problem-based learning in medical schools, government insistence on proved efficacy on new pharmaceuticals, and the growth of clinical epidemiology. The earliest medical practitioners could learn clinical medicine only from apprenticeship to an experienced master and their own experience. The practice of evidence-based medicine means integrating this clinical expertise from looking after individual patients with the best available external clinical evidence from systematic research... Good doctors use both individual clinical expertise and the best available external evidence, and neither alone is enough (1).

'Evidence-based nutrition' was bound to follow, though forms of it were sometimes used before 1990 (under different names) for difficult public health nutrition questions (2-4). In Australia the National Health and Medical Research Council (NHMRC) published a list for grading levels of evidence in 1998 (5). The National Heart Foundation followed with guidance in preparing its review of dietary fats and cardiovascular disease (6). It classifies different diet-cardiovascular disease relationships as 'good evidence', moderate evidence' and 'little evidence'. The Australian and New Zealand Food Authority included the NHMRC levels of evidence in the background papers for its committee on scientific substantiation of health claims. The committee, however, decided that, for nutrition, the levels should be revised and rearranged (Table 1) (7) because randomised controlled trials are relatively rare among studies on diet and disease (unlike pharmaceutical research) and food intake data in cohort studies are more reli able than in case-control studies.

The National Nutrition Committee of the Australian Academy of Science held a workshop on levels of evidence on 4 August 2000 at which the special features and problems of evidence-based nutrition were discussed (8,9). Then on 13 May 2001 FANO (Federation of Australian Nutrition Organisations) had a symposium on evidence-based nutrition in Adelaide the day after the annual Dietitians Association of Australia conference. The NHMRC's new dietary guidelines (to be published this year) 'are based on the best evidence available, although the working party recognises that in some cases the evidence is not complete' (10,11).

So evidence-based nutrition has arrived in Australia. It brings the potential to improve on past weaknesses in examining diet-disease relationships. It is hard now to believe that the original evidence for the notion that sucrose causes coronary heart disease was a single case-control study of estimated sugar intake of 20 men recovering in hospital from a myocardial infarction compared with 25 people the same age, some in hospital because of accidents, and some (non-randomised) healthy men in the same part of London, published as a full article in the Lancet! Another example of not evidence-based nutrition was an article on the front page of The Sydney Morning Herald in January 1995 headed 'Margarine linked to child asthma'.

The researchers say their theory is yet to be proved, but that it is backed by strong circumstantial evidence, including studies showing that asthma rates have risen in countries with increased polyunsaturated fat intakes (12).

The only data was a hypothesis! A group of American nutritionists ask, 'Do we facilitate the scientific process and the development of dietary guidance when findings from single studies are publicised?' (13)--as in the practice even with the eminent New England Journal of Medicine.

Exponents of evidence-based nutrition differ somewhat in where they put the emphasis but it seems to me we can expect these characteristics of evidence-based nutrition on the relationship of diet and the development of disease.

* All the evidence in the literature should be collected, both from search engines (Medline, etc) and from experience of specialists in the field. (Researchers are not always going to agree that a search is complete (14).)

* Disease outcomes are preferred evidence over intermediate, surrogate outcomes like blood pressure and blood cholesterol.

* The evidence must be assessed for quality. It is generally expected that randomised controlled trials would be more reliable, of better quality, than observational, such as cohort studies. But there are poor randomised controlled trials (e.g. too small, too short, wrong dose) and there are good cohort studies (large, good and repeated dietary assessment, confounding minimised). In nutrition at present we have several unresolved paradoxes that cohort studies show protective effects of vitamin E (against coronary heart disease), [beta]-carotene (against cancer) and cereal fibre (against coronary heart disease) while (smaller numbers of) randomised controlled trials have not confirmed these benefits.

* Evidence-based nutrition is very different from the evidence about therapeutic drugs because diets are much more complex than individual pharmaceuticals. There are multiple forms of vitamin E, different forms of [beta]-carotene and many carotenoids. In dietary trials other foods change when one food or component is added or removed. Food intake numbers are subjective and liable to error. Bio-availability may be critical.

* Evidence-based nutrition has to be more than a Cochrane-type review, which is typically a meta-analysis of randomised controlled trials, often excludes relevant data and does not translate comfortably into the field of nutrition (15-17).

* The evidence available to associate food and drink with disease is mostly observational epidemiology rather than randomised controlled trials. For some relationships between food and disease (e.g. does broccoli protect against cancer?) it will never be possible to accumulate sufficient randomised primary prevention trials. The challenge is to further improve the methods of observational epidemiology, particularly for usual food intake (better biomarkers) and to reduce the effects of confounding. Emerging conclusions have to be examined against criteria of strength, reproducibility and consistency with other findings from the big picture of disease history, from animal models and from molecular biology.

Evidence-based nutrition can only follow part way towards the mathematical precision of the multiple randomised controlled trials for testing pharmaceuticals in disease treatment.

Acceptance is growing at the scientific level that we should use all the objective evidence about diet and disease, wherever it comes from and consider its reliability. This will give us much better understanding than the bits of material so often fed to the general public by newspapers, women's magazines and electronic media. It is also to be hoped that rules of evidence will be able to minimise commercial influence on statements and beliefs about diet and disease.

A. Stewart Truswell

Emeritus Professor of Human Nutrition

The University of Sydney 

Table 1. ANZFA proposed levels of evidence for health claims on foods
(7)

Grade  Type of evidence

  A    Systematic review of all randomised controlled trials
  B    Properly designed randomised controlled trials or well-
       designed pseudo-randomised randomised controlled trials
  C    Cohort (prospective) studies
  D    Case-control studies or interrupted time-series with a
       parallel cohort group
  E    Comparative studies with a historical control
  F    Case series
  G    Other relevant information, such as reports of expert
       committees

References

(1.) Sackett DL, Richardson WS, Rosenberg W, Haynes RB. Evidence-based medicine. How to practice and teach EBM. New York & Edinburgh: Churchill Livingstone; 1997.

(2.) Willett W. Nutritional epidemiology. New York & Oxford: Oxford University Press; 1990.

(3.) Ahrens EH, Connor WE, editors, Symposium. Report of the task force on the evidence relating six dietary factors to the nation's health. Am J Gun Nutr 1979;32:2621-746.

(4.) Truswell AS. Ascorbic acid and colds. N Eng J Med 1986;315:709.

(5.) National Health and Medical Research Council. A guide to the development, implementation and evaluation of clinical practice guidelines, second edition. Canberra: Australian Government Publishing Service; 1998.

(6.) National Heart Foundation of Australia. A review of the relationship between dietary fat and cardiovascular disease. Aust J Nutr Diet 1999;56:(4 Suppl):S5-S22.

(7.) Australia and New Zealand Food Authority. Review of health and related claims full assessment report. Proposal P153 and pilot for management framework for health claims. Draft enquiry report. Proposal P170. Canberra & Wellington: Austalia and New Zealand Food Authority; 2000.

(8.) Truswell AS. Levels and kinds of evidence For public-health nutrition. Lancet 2001;357:1060-2.

(9.) Truswell AS. Levels of evidence for nutritional science. Food Australia 2001;53:272-6.

(10.) National Health and Medical Research Council. Dietary guidelines for Australians (draft for public consultation). Canberra: NHMRC; 2001.

(11.) Yudkin J, Roddy J. Levels of dietary sucrose in patients with occlusive atherosclerotic disease. Lancet 1964;ii:608.

(12.) Sweet M. Margarine linked to child asthma. The Sydney Morning Herald, 5 January 1995: p.1.

(13.) Wellman NS, Scarbrough FE, Ziegler RG, Lyle B. Do we facilitate the scientific process and the development of dietary guidance when findings from single studies are publicized? An American Society for Nutritional Sciences Controversy session report. Am J Clin Nutr 1999;70:802-5.

(14.) McLellan F. 1966 and all that-when is a literature search done? Lancet 2001;358:646.

(15.) Truswell AS. Cholesterol-lowering effects of dietary fiber: a meta-analysis [letter]. Am J Clin Nutr 1999;70:942-3.

(16.) Mann J, Skeaff M, Truswell AS. Dietary fats and prevention of cardiovascular disease. Conclusion may have been underplayed. BMJ 2001;323:1000-1.

(17.) Brunner E, Rayner M, Thorogood M, Margetts B, Hooper L, Summerbell C, et al. Making public health nutrition relevant to evidence-based action. Pub Health Nutr 2001;4:1297-9.