Kate Belgrave (NZ Herald; 22 May, 11 May) falls for a fallacy that often appears in public discussions of scientific issues that affect ordinary people in their day to day lives. While the issue Ms Belgrave writes about is the immunisation of infants, the same problems occur in the discussion of a number of other sensitive issues such as cancer chemotherapy (including recent Herald articles by Brian Rudman and Steven Price), global warming and the efficacy of raising interest rates to bring down inflation.

We should be suspicious of any argument that uses the words "prove" or "proof", as in "I found no reports proving links between the measles-mumps-rubella [MMR] vaccine and either Crohn's disease or autism, as claimed in the British media in 1997". The irony of Ms Belgrave's commentary is that she went on to accept that there is proof that vaccinations do not always work and that there is a concern among medical specialists that the level of risk associated with the MMR vaccine is unknown.

There is a double standard that expects alternative treatments or explanations to be proven before they should be taken seriously, while accepting that orthodox treatments or explanations do not require proof. I am not at all sure why the burden of proof should always be placed on those on the outside with the fewest resources to conduct experiments, and the least ability to get their work published in establishment journals. Furthermore, the ethics of conducting properly designed experiments in medicine, environmental and social sciences are all questionable.

In the real world, proof is often impossible to obtain, and not only because of the ethical problems of experimentation. In medicine, the minimum requirement is a double blind trial, in which some of the subjects of the experiment are taking the treatment being tested, and the rest are taking a pretend treatment, called a placebo. Neither the experimenter nor the subjects know which 'treatment' they are taking. Statistical 'proof' occurs if the outcomes are 95% or more likely to have been caused by the difference between the two treatments; eg between the vaccination and the placebo. (There is no such thing as scientific proof; scientific hypotheses can only be disproved.)

Unfortunately for medical science, even this level of rigour does not provide even statistical proof. There is much anecdotal evidence of a 'placebo effect', where the expectation that a treatment will work plays a large part in the final outcome. This idea that expectations help to determine outcomes is now well established in mainstream economics: hence the importance we give to surveys of 'business confidence' and 'inflationary expectations'.

It is likely that many alternative medicines (and some mainstream medicines) work at least in part through the placebo effect. If we run an experiment and tell all the subjects that they might not be receiving the treatment, then their faith in a positive outcome will be diminished. If we deceive the subjects and convince them that they are all receiving the experimental treatment, then better results may appear for both the treatment and the placebo. (There are of course ethical problems in lying to people you are performing experiments on!)

The important point is that, if there is a placebo effect that leads to positive outcomes, then we need to find ways of harnessing it rather than belittling it. It is an effect that may never be proven, but can probably never be disproven. And of course the placebo effect, inasmuch as it exists, will help some subjects more than others; I suspect that people with religious faith are more likely to benefit from alternative treatments than are cynics and sceptics.

The issue is not just about proof; it is about finely balanced probabilities. When a person is choosing a treatment (or avoidance of a treatment) for themselves or their children, they have to weigh up a number of pros and cons, as well as to factor in the unknown. In the real world, all decisions are made with limited information and in a context where different people, with different values, will weigh both known and unknown information differently.

Indeed, referring back to my own speciality of economics, different people will value the economic cost of inflation quite differently. So many will regard the orthodox treatment (and its possible long-term side-effects) as being worse than the disease. Others will not. It is the same with chemotherapy and measles vaccinations.

The immunisation issue is particularly pertinent today because of the suspicion of links to autism. It was only last year that the problem of autism received a major publicity boost, following the conviction of Janine Albury-Thomson for the manslaughter of her autistic daughter Casey. We have seen enough to suggest that the cost - to parent and child - of autism is much greater than the cost of measles or mumps.

Even if there is only a 10% chance - ie far short of proof - that there really is a link from the MMR vaccine to autism, the balance of probabilities would be enough for many rational parents to reject that vaccine. While better information in the future may lead to different decisions by parents, in the meantime many parents will choose to manage their risk conservatively by minimising exposure to unknown risk factors.

Kate Belgrave appeals to history. Yes, there were epidemics in the past. Likewise there will be epidemics in the future, given the present level of antibiotic resistance and concern over the deteriorating state of our general levels of immunity. One important lesson of history that she fails to acknowledge is that orthodox medical treatments and scientific explanations have generally been superseded by new treatments and new explanations.

Science, Medicine and Liam Williams-Holloway    24 February 1999

Authoritarian versus Alternative Thinking in Godzone    16 May 1999

Rankin File