> Hasn't this been fixed with the requirement to pre-register trials?

If it was that simple...

Preregistration is a step in the right direction, but there are still plenty of ways to game the system, particularly as it's not properly enforced: http://compare-trials.org/

But it is the necessary step to highlight the problem so that actions can be taken.

Also, you can safely assume that the large majority of outcomes that were not published were:\

- negative

- worse than negative (i.e. safety issues?)

- not properly designed

So if you compare the number of trials disclosed/not disclosed for each drug/each company that gives you a pretty good picture of who is doing what already.

With that data, it's then a lot easier to approach regulators to do something about it.

My assumption is that most unpublished trials were "no effect".

Be a drug company. Develop a "drug" that is really a placebo (has no effect). Do 10 trials. 8 are "no effect", 1 is positive (by chance), 1 is negative (by cance). Publish the positive drug. Market the drug as having positive effect. Sell the drug, make billions. It would take decades for enough other research (mostly "no effect") to accumulate before it becomes obvious that your drug actually has no effect.

That’s the whole point of preregistration. The nine unpublished trials can be assumed to be negative, or at least neutral, and were aware that they were done because of preregistration. Therefore regulators have the ability to reject the notion of the effectiveness because so few of the studies reported positive outcomes. In addition, any cancelled studies now taint the authors and sponsors with a (small) amount of suspicion that they are refusing to publish non-positive effects which, cumulatively, will cause their results to be further discounted. That will hopefully provide a strong incentive to publish neutral and negative results.

A drug that genuinely has no effect is actually pretty good; most drugs have some sort of risk and many are seriously dangerous. Rofecoxib is believed to have caused around 100,000 heart attacks and 30,000 deaths. Market approval for rofecoxib was based on misleading data and its withdrawal from the market was substantially delayed by misleading data.

https://en.wikipedia.org/wiki/Rofecoxib

From the results there it is clear that they failed to game the system since the gaming was detected. Getting to the point where regulators can actively use the preregistration information is the next step, but the foundation is there.