[stuckindoors]

Dietary studies have been largely been poor since they rely on food frequency questionnaires.

The pertinent sentence from the methods section of this study is "Questionnaires have been completed biennially to update information on potential risk factors and occurrence of new diseases." Poor data in leads to poor data out.

I'm not sure I can quantify what I ate accurately in a twice yearly survey -- can you? Bias also plays a role in how these questions are answered.

The other factor is hazard ratio for red meat and CHD was 1.11. Thr hazard ratio for tabacco and lung disease which is a known association was between 14-17. The effect that they found is small in magnitude.

Epidemiology studies like these need to ranked low on the totem pole of types of evidence.

[admash]

Inaccuracies and bias in food frequency questionnaire reporting (collectively termed “measurement error”) are well known and can effectively be quantified via validation approaches (see https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5404419/ ) and remediated by well-established statistical techniques such as regression calibration (see https://dietassessmentprimer.cancer.gov/learn/regression.htm... ).

Furthermore, when regression calibration is performed, the magnitude of the calibrated effect is often shown to be greater than the uncalibrated result, so that 1.11 hazard ratio is likely to be larger in reality.

Finally, “small” is clearly relative. I’m certainly quite interested in knowing about things which increase or decrease my risk of having a heart attack or stroke by 10%.

Certainly a blinded randomized controlled feeding trial is preferable, but is not always feasible, particularly when it would entail controlling participants’ diets for long periods of time.