1. Wayyyyy too much industry involvement, much of it not properly disclosed (OP here is a classical and representative example).

2. Lots of research from relatively small and "less known" institutes, done by researchers with hard to verify credentials.

3. Low scientific standards. That'd include:

  - Refusal to share data
  - Refusal to share data analysis methods
  - Not keeping the basics of a double-blinded test
  - No reproduction, and not even the possibility of independent reproducibility, as the methods are described so vaguely.