[michaeltoth]

Scientists are also incentivized by the publishing system to find noteworthy results, which often leads to finding evidence where there is none, especially in fields like nutrition where the signal to noise ratio is so low

[jacobr1]

Or just as common but perhaps more misleading: results in isolation. Even with solid protocols and methods plenty studies, say with rat models, have a very limited factor they are testing, and identifying if there is an impact compared to a control. Unfortunately the baseline model parameters might have a huge impact on any difference observed. Like say diet prior to the study. Or genetics other than what was modified for the study. Or any number of similar things. Changing a single variable in a complex, non-linear, non-closed system with doesn't necessarily generalize to other conditions. So even good studies can mislead until they've been replicated and many variants attempted and some underlying mechanism identified.