When a trial produces results its sponsor dislikes, the sponsor can legally leave it unpublished, so positive findings appear in print about twice as often as negative ones. Doctors and patients then judge a treatment from a filtered record that shows its benefits and hides its failures. Each published study can be honest and the total picture still be false.
A trial can be built to favour a product without faking any data. The tricks are to test it on unusually ideal patients, compare it against a rival given at the wrong dose, keep the trial too short or too small, and stop it at whichever point flatters the result. This is why trials funded by the maker come out in the product's favour far more often than independent ones.
Academic papers are trusted because they look independent. Companies exploit that trust when their own staff secretly plan and write the papers, when the authors' ties go undisclosed, or when the company owns the journal outright. Doctors' training and patient groups can be funded the same way, so the sources people rely on to check a seller may be paid by that seller.
Regulators see most of the trial data, but mostly from early in a product's life. They do not pass it to doctors, to patients, or even to other parts of government. They also approve new drugs without requiring that they work better than existing ones, so approval says much less than people assume.
Research gets done where someone stands to profit from the answer. For several important, long-standing medical problems, nobody knows which treatment is best, because no one gains financially from running the comparison. The gaps in the evidence follow where the money is, not what patients need to know.