3 ms·
Indeed. The original article comes across as really quite biased too. Longitudinal cohort studies can never nail causality but they are quite good at providing
by azalemeth 4y ago
Indeed. The original article comes across as really quite biased too. Longitudinal cohort studies can never nail causality but they are quite good at providing measures of apparent risk.
In cardiology, an odds ratio of about 1.2, called tiny by the article, can actually be considered quite large in some circumstances. Very roughly speaking, it's about equivalent to your risk per ten years of age of developing heart failure, or per 20 mmHg of systolic blood pressure [1].
[1] https://link.springer.com/article/10.1007/s10557-022-07355-7 https://link.springer.com/article/10.1007/s10557-022-07355-7
- radford-neal 4y agoThe point isn't that an odds ratio of 1.2 is too small to be important. It is that if the estimated odds ratio from the study, with all its biases, is 1.2, then one can't be at all sure that the actual odds ratio is even greater than 1. (An odds ratio less than 1 would indicate that red meat was protective, rather than a risk factor.)