Tuesday, June 22, 2010

Physician, heal thyself

I have an undistinguished B Sc degree. I am a very minor politician and I work in marketing and it is for the last of these that I apologise.

Last night I was fortunate enough to be invited to attend the annual lecture of Sense about Science. The key speaker was Dr Fiona Godlee, the editor of the British Medical Journal, who combined being interesting, thoughtful, thought-provoking and witty.

The essence of her presentation involved failures in science, particularly medicine. In many cases the culprit was business. For example, she showed how business-sponsored scientific reviews of medical products were almost always positive, while independent reviews were far more mixed. Is it likely that these sponsored reviews are unbiased? Statistically the answer is no: the probability is so remote that it cannot be taken seriously. Over fiften or twenty minutes she presented her case cogently and with eloquence.

I started to disagree with her when she looked at some ways to address the problem. I must say I did so with trepidation. As I have already said, I have an undistinguished BSc while she has a strong scientific resume. But since my view is based on the philosophy of science (if that is not too grand a term), I'm going to put it forward anyway.

In the Olden Days, when I was at school, there was a fairly simple approach to science. It went roughly as follows:

  1. A scientist would think of an idea - the hypothesis.
  2. She would devise an experiment to test it.
  3. She would run the experiment and collect the necessary and relevant data.
  4. If the data supported the idea, it would become a new theory.
  5. Other scientists could repeat the process and get the same results, so independently validating the theory.

Intellectually this works. Practically, in the real world, it sometimes breaks down. For example, developing new drugs is extraordinarily expensive. Pharmaceutical companies invest huge amounts of money in each new drug and their scientists invest years of their lives. The must be a great temptation, even in the most scrupulous scientist, to look for the good effects and play down the bad. Possibly that may be succumbed to unconsciously: perhaps dismissing a strange result as an outlier. The commercial pressures will be there too.

At present, science does too little to eliminate opportunities for these errors or for fraud. I suspect outright fraud is fairly rare, although cases such as

show that they happen often enough to reach the mainstream media.

If I understood her correctly, the core of Dr Godlee's proposals to address these problems was to reduce commercial involvement in science, especially medicine. I think that would be wrong.

Science needs to look more to itself. Some of the finest brains in the past few centuries have been involved in science: why then is the very core of science, the scientific method, so subject to abuse?

I suggest that a fundamental issue is that scientific research is not sufficiently open to a basic commercial technique: audit. If at key stages research was subject to audit, the results would be more credible. This does not mean making the data or area of research prematurely public but it would mean that when it came to submit a paper for publication or to launch a new product there would be confidence that the research had been adequately conducted.

Who could do this auditing? I think that is less important than defining how it is done. For example, in the UK perhaps the Russell Group of universities could define: what independent checks should be made in each key area of science as that research progressed; what records must be verifiably submitted at each stage; what are acceptable ways of discounting results ("the subject died in a car accident") and unacceptable ("this outlier was so different that I suspect a measurement error was made").

This would not replace peer review. It would give reviewers greater confidence in the underlying methods and might well give them access to the underlying data.

It would benefit science. It would also benefit commerce: whether in medical, physical, chemical or any other area of scientific endeavour, the results of audited research would rightly be more trusted.

This, I think comes back to Dr Godlee's fundamental point: unreliable research undermines good science. If we can improve standards in research it will be of benefit not just to the scientific community but also to the public at large. Trustworthy results will lead to better decision-making, whether it means buying a new pill or investing in technology.

And should that day happen, I won't have to apologise for being in marketing.


Sciencecarol said...


Came across your blog as a result of your posting on Bad Science, and was interested becasue I had some thoughts along this vein myself and blogged about them here http://sciencecarol.wordpress.com/2010/07/13/sense-about-science-annual-lecture-dr-fiona-godlee/ I hadn't really thought about it from the point of view you put but I did kind of agree with you. The only problem with your suggestion is that the checks would need to be done on an international basis and the Russell Group, if at all, would only be able to be responsible for the UK. Much harder to implement! It's an interesting discussion though.

Sean Haffey said...

Hello Carol

Well things have been a bit busy in my personal life recently and my attention to this blog has suffered, so my apologies for not replying earlier.

Bear in mind the first statement I made that I have a very undistinguished science degree. So I am not best placed to say how things could be improved, although that won't stop me from makiing suggestions! To me, "how" the method could be made more robust can be decided largely by scientists (and may vary across disciplines).