Nutrition science mustn't accept a lower level of evidence
Bibliographic record
Abstract
We read with interest the article “Evidence-based criteria in the nutritional context” by Blumberg et al.,1 who postulated that the level of confidence in nutrition science should be lower than that needed for other branches of medicine, particularly the level of certainty required for drugs. While we agree that drug evidence requires a very high level of rigor, we believe that nutrition science should not and need not be complacent with lower levels of evidence than is required for other branches of medical science. Certainly drug evidence has a high requirement for rigor, for at least three reasons: 1) the involvement of commercial interests, 2) the ability to easily design placebo-controlled randomized controlled drug trials, and 3) the potential for powerful negative effects and side effects. However, all areas of science and medicine (including nutrition science) will always have unknowns. Therefore, we must rely on incomplete knowledge to inform practice. It is of prime importance that we be aware of what knowledge we have and which hypotheses we have not had confirmed, while continuing to strive to improve our understanding of biological systems. Sometimes, relationships that are described in observational studies are not found to be cause-and-effect relationships when randomized control trials are used to test them. Two examples in nutritional science of observed relationships that were not found to be causal include beta-carotene and B vitamins. Both beta-carotene and B vitamins showed promise in observational studies for the control of chronic diseases, i.e., beta-carotene for cancer2 and B vitamins for lowering homocysteine and cardiovascular disease.3 However, large randomized trials did not uphold the results from observational studies. Randomized trials of beta-carotene showed it increased the risk of lung and gastric cancer among smokers,4 and while effective in reducing homocysteine, folic acid had no significant effects on vascular outcomes.5 These examples illustrate how observational study data should not be considered to indicate proof of a relationship, and randomized trials are needed to make causal conclusions. Cause-and-effect relationships cannot be determined from observational data, since the variable being examined as the cause is not isolated from other possible causes. Observational studies are useful for developing hypotheses, which can then be proven or disproven by randomized controlled trials (RCTs), in which the variables are carefully controlled so that causality may be determined. Until we have consistent RCT evidence, the best knowledge we have is the graded body of existing evidence. In other words: “Correlation does not imply causation.” We agree with the authors that “nutrient policy decisions will have to be made using the totality of the available evidence,” as this is the basis for decision making in evidence-informed medicine and epidemiology. However, we disagree that we should accept observational studies as definitive in nutrition science and thus become complacent. There are many examples of RCTs in nutrition science.3,–10 In the field of nutrition, all types of research are needed, i.e., observational studies to document prevalence and generate hypotheses, animal studies to understand biochemical mechanisms, qualitative studies to understand the human condition, and RCTs to establish causal relationships. All areas of science need to consider the whole body of literature through quality appraisal and consideration of the highest quality RCT evidence, when available. The Oxford Centre for Evidence Based Medicine summarizes this concept well: “What are we to do when the irresistible force of the need to offer clinical advice meets with the immovable object of flawed evidence? All we can do is our best: give the advice, but alert the advisees to the flaws in the evidence on which it is based.”11
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.268 | 0.589 |
| Meta-epidemiology (narrow) | 0.002 | 0.002 |
| Meta-epidemiology (broad) | 0.013 | 0.006 |
| Bibliometrics | 0.009 | 0.005 |
| Science and technology studies | 0.005 | 0.029 |
| Scholarly communication | 0.024 | 0.032 |
| Open science | 0.016 | 0.009 |
| Research integrity | 0.060 | 0.099 |
| Insufficient payload (model declined to judge) | 0.009 | 0.009 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; the direct Gemma label and the distilled Codex classifier agree on what is shown here.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".