The publication bias would go in the other direction. A drug company showing the effectiveness of vitamin pills would publish that research to help them sell m

The publication bias would go in the other direction. A drug company showing the effectiveness of vitamin pills would publish that research to help them sell more vitamin pills; A drug company finding vitamin pills to be ineffective would be incentivised to not publish the research. Further more, we're not talking about just one study here against a sea of studies in favour of vitamin supplements. Free radical theory has been shown to be effective in a petridish, but it has a neutral or negative effect in most large scale human studies.