If you are a research scientist you would have to say what you are doing before you do it, what would be the primary outcome that you want to find and you would say how you are going to do it so you would stop scientists P-hacking, which is dredging the data with lots of tests until you find something interesting and it would stop Harking, which would be creating the hypothesis after you have actually done a study….
Why do registration?
Some people argue to find the missing studies because the published animal data is overwhelmingly positive and suggests that the negative stuff never sees the light of day, so it is biasing the view such that drugs are better than they actually are. Why were the studies not published….presumably because they don’t work.
If this data was evident would this allow us to pick better candidates to try in MS. Furthermore having the information in the public domain would mean that the animals were not used in vain.
Some journals are offering to review the study before it is done and they will agree to accept it irrespective of whether the study works or not. This is designed to reduce the bias associated with positive studies. However, it seems that it is not necessarily the journals that are not publishing negative data ,but authors are not bothering to try and publish negative data, because of the time and cost and effort needed to publish papers.
Will registration of animal studies make us find drugs useful for humans, I am not sure this is where the problem with translation of animal studies lies and registration would require a lot of effort.
Maybe we need a depository online, where we can show the methods and results (& raw data) of experiments (without comment would be fine) that don’t work, or even that do work and reproduce the study. Possibly this can be linked to the DOI of the paper whose data can’t or can be reproduced maybe the “The Power to Reproduce” or “The negative experiment” housed on a reputable site e’g NIH).
Maybe this could be a way such that dogma will not build up as quickly and unreproducible work can be quickly refuted. It is often hard to get papers published that disagree with published work taking years to achieve….this was evident when we tried to publish data where we were disagreeing with the conclusions of our own work that we had made earlier.
Are these mad ideas? Are they practical?
I try and tell people who are reading science papers that they need to be aware of the CONSORT and ARRIVE Guidelines which are guidelines how to report human and animal studies. It helps you to assess quality of the experiment and believability of the data. When I teach about the ARRIVE guidelines, I try and use real examples to show why they are worthwhile to consider.
So a registry
Many years ago I did an experiment that I did not publish showing that I could not repeat the contents of a Nature paper. This lack of effect was reproducible, but rather than waste more time and money doing more work to try and publish a negative paper, I dropped the approach and moved on.
I subsequently heard that four other labs had done the same. If there was a place where we could have pinned up the data, the readers could have made up their mind and the neuros too, as the approach did not work in MS.
Had the studies being registered we would be aware that four other labs had done the study. Why no papers on this proposal. However if there was a site where we could have linked the data to DOI of the paper and included the data, people could see how many animals were used, the negative/replication data is in public domain and available for meta analysis or analysis by the original group.
P.S. The World wide web did not exist when the experiments were done
Registration….Would Pharma do it…I doubt it because they would not be willing to say what they are up to publically before they do it. Why because it would disclose the idea before the patent, so nothing would get developed….They, or your science competitor with more resource to do the experiment you have alerted them to, will laugh at dozey academics giving all their ideas away for free, years before your competitors would know…is this a good thing?
Some of the old data does eventually surface in publications