I'll intrude into the quasi-medical discussion with two points:
- Confirmation Bias
- Correlation vs. Causality
Confirmation bias, common to both researchers and ordinary folks, is the strong tendency each of us has to acknowledge reports that confirm our beliefs, and disparage reports that disagree with our beliefs.
Correlation is demonstrated by a rooster always crowing when the sun rises. If you train a rooster to crow at a different time, it has no effect on the rising of the sun.
Causality is the ax that beheads the rooster. The ax causes the beheading.
The preponderance of "scientific" studies published in mass media (as well as a depressingly large fraction of those published in academic journals) clearly demonstrate only correlations, while leaping to the implication of causality. That's why today's "study" often contradicts yesterday's "study."
When I have a serious medical problem, I consult a physician, rather than a tobacconist. But that's just me.
Bob