How to Streamline Drug Research?

We all know that information is valuable, and that more information is generally better than less.

But in the realm of pharmaceutical research (as in others, to be sure), there’s a troubling paradox: while successes are widely publicized, and while the results of clinical trials are usually published, the research from projects that fail before that stage is usually kept hidden. “A result,” writes Natasha Singer in The Times, “is that companies waste many millions going down experimental paths that their competitors have already found to be dead ends.”

You should read Singer’s article in full, but here are a few highlights:

Now big-picture thinkers, within the industry and outside it, are re-examining every stage of drug development — from molecule to market — in an effort to foster faster innovation. It’s a holistic approach, called “systems thinking,” that originated in methods that engineers used to streamline projects in the aeronautics and automotive industries.

The Massachusetts Institute of Technology, for one, started a pharmaceutical innovation program this year to help drug companies adapt some successful approaches now used in aeronautics, like lean management and information-sharing among rivals.


The M.I.T. project, called New Drug Development Paradigms, has gathered a powerful consortium of interested parties — including major drug makers and federal health authorities. One short-term goal is to identify, and rectify, the root causes of bottlenecks in the existing system. Longer term, the ambition is to create new prediction models, new ways to share information about the biology of diseases, and a new inclusiveness involving earlier participation of regulators, health insurers, health care providers, and patients.


In the absence of a public database on failed drug compounds, a small group from the Sloan School of Management at M.I.T. and the Harvard Business School has created Pharmer’s Market (not this one), an online prediction market that uses crowd-sourcing to forecast the likelihood of a drug’s success. Introduced last month, the market has invited biomedical researchers and other drug industry experts to place anonymous bets, using virtual money, on the likelihood that certain breast cancer drugs, currently in clinical trials, will succeed or fail.

After the clinical trials conclude, researchers can assess whether this kind of collective intelligence may be a useful predictive tool for drug companies, said Ragu Bharadwaj, who helped devise the project as an M.I.T. graduate student.

There is an obvious conflict between open-sourcing and a competitive marketplace, but it is exciting to see smart people pushing hard to change the shape of that marketplace for the public good.


Did someone think of the moral hazard? Researchers sabotaging their own or others projects to collect a huge windfall from their betting?

Joe Smith

As a general proposition in research it might be useful to have a way for researchers to document and publish approaches that failed so that others do not go down the same blind alley.


@ puneet - the betting is anonymous and with fake money, so I doubt it.


Of course, the sharing of past experiments will definitely benefit society in that, as explained, companies will not have to go through the same path as the other ones did, spending millions of dollars in the process. However, it is easy to understand why the companies DON'T do so. Why help the other company out if doing so would just give them an advantage? It doesn't make sense if you spend tons of money on something that didn't work and then give a piece of advice to the rival company that will only help them not make the same mistakes.

So what can be done so it's all fair and competitive? Bah, who knows. The only thing that comes to my mind is just one giant pharmaceutical company, but that's not fair. Or is it?

Water is a public good. Protection is a public good. Even a representative in court is public to anyone and everyone. So why can't life-saving cures be made public? It doesn't even have to be a government institution, just one big company with regulated prices so as to not rip the customers off. I'm not saying this is the answer...but could it be a possibility?


Jonathan Pryor

"There is an obvious conflict between open-sourcing and a competitive marketplace..."

Is there an obvious conflict? Similar things have happened and are happening in the "open source" world, in which large amounts of source code are shared, developed, maintained, and improved via market competitors. For example, the Linux kernel is co-developed between the market competitors IBM, Red Hat, Novell, and Ubuntu, and the result is an improved code base which is more widely used than if it were tied up with a single company.

The patent system is another example -- the information is intended to be shared (which is why patents expire), thus increasing the amount of shared knowledge usable competitors.

I would instead argue that this "open sourcing" -- sharing of information -- is REQUIRED for a competitive marketplace. (Think of the opposite -- if nothing were shared, then there would larger costs in entering new markets, resulting in fewer competitors, and a less efficient market. This is why we even have a patent system in the first place, to encourage sharing, improving the market!)


Michael F. Martin

Please say more about the obvious conflict between open-sourcing and a competitive market. How is open-sourcing anticompetitive? Doesn't open-sourcing certain information simply change the goal of the competitive game? Open-source companies are still competing in some arena.


Why not start a voluntary, but subscription-based database managed by a neutral third party whereby results of failed experiments are documented robustly but the contributors remain anonymous? (This is sort of the same as they way benchmarking works...everyone wants to know where they stand amongst their peers, but no one wants to see their name at the bottom of a category)
This way, Pharmas can see what's being done and make their decisions about whether to persue certain lines of research or not.


This is a problem in all of science. The push is to publish positive results, not negative results. The Journal of Negative Results in BioMedicine is one resource to combat this though.

David Chowes, New York City

Of course all clinical reseach trials and other results and other understandings gained through scientific investigation -- even if the treatment demonstates dangerous side effects and/or lack of efficacy should be made available to all.

This would benefit all: patients, pharaceutical companies...
Well, the entire society and the world.

But, I have read about this a number of times over many years. Why does the wheel have to be invented so many times -- and, people pay little or no attention to it?


What is the incentive to report a "failed combination"?

Could a "failed combination" report from a competitor be trusted?

What does it mean that a combination failed? Aren't there drugs on the market today which "failed" for one purpose, but we effective for another purpose?

If the US government sponsored pharm research, would the results be public knowledge?


Does it sometimes happen that people think they have conducted an experiment correctly, and for whatever reason, people later discover the results to have been incorrect?

I think so, right? So in addition to the issues of companies (justly!) not wanting to share this sort of info with their competitors, there are still benefits of repeating research.

Another benefit may be that upon repeating some type of research, researchers may think of a question, or notice something different that leads to some sort of discovery that was not initially sought after, but which could still benefit society and possibly in ways greater than the actual experiment was supposed to. And perhaps the original company or original scientists who had already studied the matter did not notice something important?


I'm all in for sharing the information regarding the failed experiments into a database after of course dealing with the various questions like credibility,possibility of future research,etc.Actually this initiative should be taken by the pharma companies for their own growth and profitability.It would be similar to the sharing of cell phone towers by cell phone operators or the ATM's by banks,just a bit more complex.

I've an Idea:
Why can't drug companies come together and create a unique database as explained below.
First of all, we will assume A, B, and C are THREE different DRUG companies. Ex will represent a experiment FAILED and Ex1 will represent an experiment which has already been conducted by A but FAILED, and is kept secret by A and, Ex1B will represent the duplicate experiment Ex1 which is just going to started by B as well. Lastly, we'll call the proposed central database as Q.

The problem with creating a Q is that A, B and C won't cooperate because each member(A,B and C) think the profit made out of its preexisting portfolio (medicine) is going to offset any loss occurred by a failed Ex. A solution would be to create a Q, where each member(A, B and C) will deposit some money as well as input any experiments which is going to be conducted as well as input Ex, and in this case, A inputed the result of Ex1 into the Q. Then, B is just going to start experimenting that same experiment (which is denoted as Ex1B) and has just registered that experiment to Q. But, Q notifies A (NOT B) that a failed experiment (previously conducted by them) is also going to start being conducted by an anonymous Drug company. it also says (to A) that, it (A) can make some money by warning or notifying that anonymous company the results of that failed Ex1. So, A creates a contract with Q and, Q adds some money inside A's deposit from B's deposit money. Then, Q notifies B about the impending failed experiment it has just started conducting, and it also says it(Q) has taken some of its(B's) deposit money for giving this warning.

One way for A, B and C to abuse Q is to input a lot of false information which haven't yet been attempted by the drug companies. A way to curb this behavior among A, B and C is to put a clause in the contract between Q and A, B,C that, each company(A,B and C) can at maximum, derive 25% of its profits from Q and the rest of the profit has to come from its medicine portfolio.




I really don't understand the controversy here. There is a database called where the results of every clinical trial are made completely public.

Diana c

Its is not a new information especially for those in the scientific community. As a former research chemist, I want to say that no one starts a project intending it to fail, hence the word research and most of the time the end usage is not know. Most synthesis do work and those result are published an put in searchable databases but not because a successful synthesis does not make a successful product. The steps from getting a compound from the lab to a bottle takes many steps, many people and several years. Knowing the amount of work and length of time it takes to produce pharmaceutical, I would not start such business.

Karl Siegemund

Jonathan (5):

The patent system does more: It forces the inventor to publish his invention long before the patent expires: As soon as it is granted, it has to be published.
Only in exchange for the publication the inventor gets the protection patent law offers. In other legislations the publication happens already with filing the patent.


On the database: Companies would only release failed compound data with a significant time lag, probably at least four or five years. Less time than that and they would effectively be giving their competitors a roadmap to their research efforts. If you and your competitor are trying to release the first drug against a given target, would you rather they repeat your mistakes, or pick up on your approach and release their drug before you get yours done?

Crowdsourcing: Ha. Has crowdsourcing ever worked in a case that doesn't involve estimating some physical measurement, gumballs in a jar, weight of a cow, etc? Try crowdsourcing this question: Is the value of the slope of the line Y=mX positive or negative? Oh yeah, don't tell people the value of m before making them answer. I'm sure the crowdsourcers will make some business consultant cash out of the project, but asking people who have even less information than those running the project won't help. If the needed insight is common in the crowd, then it should also be common in the 'crowd' of researchers actually working on the project (who share the same range of expertise as those in the crowd). If the needed insight is rare in the crowd, then the signal will be lost in the noise.

@#14: There is a website called Unfortunately, a lot of clinical trials never get registered, and of those that do a lot get fudged before or after registration (changing primary and secondary endpoints, etc).



Stephen, I just wanted to say first I find all of your articles and blogs stimulationg without being shrill or politically skewed. As for this blog, science is best served when failures are shared. Not because it stops others from repeating the mistakes or going down the same path. Evaluating the process and the decision making that went into them, teaches much more then the results themselves. What we think are major innovations are instead efficiencies and breakthroughs that are in truth small, sequential steps painstakingly put together over time. .