Javascript not enabled
Clinical lead, Dr David Hamilton
Sharks like ice-cream - and other statistical folly

Much is being written lauding Professor Peter Higgs as he collects his Nobel Prize for introducing the word ‘boson’ to everyday language and making ‘particle accelerator’ a perfectly reasonable term to drop into conversation with one’s parents. Among all the related coverage is a fascinating offering from the good people at godparticle4u.com; who sell ‘bearings from the CERN accelerator’ which were “exposed to the surge of energy which showed substantial evidence of having the God Particle”. These bearings are described as having been sent to “the world’s leading universities”, who have demonstrated “amazing effects on pain reduction, in treating skin disorders and in enhancing performance in those suffering sexual dysfunction disorders” - a bargain at $199.99.

If we (generously) assume that these claims are based on true reports, then this wonderful nonsense is the latest example of the dangers in equating correlation with causation. My favourite illustration of this is the well-known link between shark attacks and ice cream sales. The problem is of hidden variables that our initial analysis can miss. In this case, we can reasonably hypothesize that instead of sharks enjoying mint-choc-chip; the weather and/or temperature may be implicated.

Are these examples too trivial for a research commentary? I think they serve to highlight a bigger issue. Determining the efficacy of a treatment is a very real practical problem; most clinicians will have their own example of patients with newspaper clippings of the latest ‘remedy’ for incurable diseases. This may well be due the Daily Mail’s ongoing attempts to dichotomise every known food group and inanimate object into cures and causes of cancer – or is this just another correlation?

Regression to the mean is an important related concept. This is the statistical phenomenon of outlier data moving toward the mean in subsequent non-randomly selected tests; or put another, way if a variable is extreme on first measurement, it will tend to be closer to the average on second measurement.

It’s important to bear these concepts in mind when evaluating the latest fad intervention, or indeed any intervention; particularly those that show a treatment effect by reducing pain levels in those with extremely high pain scores prior to the application of said intervention. Randomisation and blinding of prospective comparative trials help address these issues, but we must always remember that even in well conducted studies, statistically significant efficacy does not necessarily equate to a clinical or therapeutic benefit.

Incidentally, my favourite attempt at explaining the link between increased shark attacks and greater ice cream sales comes form one of our medical students; “perhaps It is easier to swim away from a shark if you are not trying to hold on to an ice cream".

Dr David Hamilton, Research Fellow, Department of Orthopaedics,  University of Edinburgh, UK
January 2014


Missed previous commentaries?