This is a general problem in science (not just biomedical research). I'm a physicist, and we see the same sorts of issues.
It all comes down to how academic research is funded and judged: number of papers, number of students graduated, and amount of money raised. Inside granting agencies, this is how different research efforts are compared to determine which programs get (more) funding and who gets cut. The importance of the work, the correctness of the work, and the ethical behavior (or not) of the researchers are not considered. Scientists are not stupid, if those are the metrics used to determine funding, they optimize for those things.
If we want to fix science we need a different set of metrics.
I'd suggest replacing the three metrics above with: number of validated results, public interest, and amount of private investment in the work. This would apply specifically to government granting programs.
"Validated results" requires a third party to validate, that should be government labs validating academic/commercial work (we're talking about reviews of government grants) and the opposite for new work done at government labs.
"Public interest" is much easier to track now than it used to be. A simple metric would just be google search ranking (although I'm sure something better could be used).
Private investment may seem overly commercial to some people, but we have a big problem right now with a lack of development of scientific work. Last year was the first time since 2000 that private investment in startup companies exceeded government investment in basic research (in the US). Commercialization is much more expensive than basic research; we're still only passing on a fraction of the potential practical work. We need to motivate people doing basic research to work more with industry (where appropriate, right). In addition, you have several diseases (usually "orphans") where private donations for disease research are greater than government investment (i.e. Lyme disease). Maybe that's fine, but the granting folks need to take a look at why that is and whether they're really investing public dollars where they need to go.
Lastly, I would change the system every 10 years or so. The longer any set metric is used, the more likely it is that people are gaming the system rather than working in the public interest.