What you want is a bunch of statisticians looking at the research and patient outcomes.
(If a bunch of statisticians looked at the research and patient outcomes and disagreed, then that would be an even more sensible reason not to. But in lieu of that, we expect dentists to have a reasonable idea what they're talking about.)
What actually happens (or so I hear) is that dentists are asked "do you recommend brushing twice a day with toothpastes such as toothident", and they say yes, because even though they've never heard of toothident, brushing your teeth is a good idea.
So an ad that seys "4/5 dentists recommend toothident" doesn't actually mean that 4/5 dentists recommend toothident, and is not a sensible reason to buy toothident. But the way people interpret the ad is a sensible reason to buy toothident, it's just false.
No! Just because they think it works doesn't mean it does work. Toothident might be the first toothpaste that comes to mind but that doesn't mean it's better than any other toothpaste - and it might be worse than other toothpastes. Perhaps the toothident company just have much better pens and sticky fixit pads than the other toothpaste companies.
Toothident is more likely to be recommended by 4/5 dentists in a world where it is better than other toothpastes, than in a world where it is worse.
So by Bayes theorem, if we see that 4/5 dentists recommend toothident, this is evidence that toothident is better than other brands of toothpaste.
It's not necessarily strong evidence, but it is definitely evidence. (And I do actually think it's fairly strong, if you can get 4/5 dentists to specifically recommend that one without priming them.)
Edit- Also dentists are not necessarily particularly rational. I know a dentist who told me that she rejects evidence based medicine and thinks that we should use traditional Chinese medicine instead. She isn't Chinese, she is a British middle-class hippy and when I asked her about the effect on rare species, she said that rare species are never used in Chinese medicine. I then asked her if she supports evidence based medicine when her patients are anesthetized, or does she think the dosage should be guessed and she said she didn't know what that had to do with it. Also, 5 is a tiny sample. From a sample of 5 dentists, 4 might tell you that the second coming of Christ is imminent, but that would tell you very little about either Christ's travel plans, or the opinions of dentists.
This is false and it is dangerous.
Dentists are not experts in toothpaste; they're not experts in the latest toothpaste research.
A dentist recommending toothident will probably have no idea about whether toothident is better or worse than some other toothpaste.
For a real world example of this see knee arthroscopy for rheumatoid arthritis. Doctors would recommend this surgery over other treatments until placebo[1] controlled trials found little benefit.
"X out of Y experts recommend $THING" would be a terrible reason to chose $THING over $OTHERTHING unless you have imformation about why they recommend it and how they made that decision.
I can think of plausible mechanisms that would discriminatingly push an unusually good toothpaste to the top of dentists' recommendations. Offhand, I can't think of any that would discriminatingly push an unusually bad toothpaste up there.
By "discriminatingly", I mean that e.g. advertising would work just as well for a good toothpaste as a bad one. But "the evidence suggests that this toothpaste is actually really good" is more likely to show up for a good toothpaste than a bad one. And I can't think of anything which is more likely to recommend a bad toothpaste than a good one.
So if I think there's a 90% chance that toothident is only recommended because of advertising, then I have no reason to think that toothident is particularly good or bad; but if there's a 10% chance that dentists do actually have a reason to recommend toothident, then I should use toothident, because I have no reason to favour anything else.
Certainly there are cases where this goes wrong, but I never claimed it was infallible.
The more noise there is relative to signal, the less effective this method gets, but it's still better than any other strategy (in this toy model).
Take nine 0s and a single 1, then generate ten random numbers, identically distributed however you choose, and add one of them to each of your starting numbers. Ask someone to pick out where the original 1 was. However you generated the random numbers, their best strategy is to choose the largest resulting number.
Probably tapping out here.
If we assume for a moment that 4/5 doctors really do recommend X, and that X is something related to one's health, why is it not a rational cue to buy X?