The art of death reckoning

Studies Show

August 7, 1998

Medical research" brings images of brilliant, tireless, tieless doctors skilfully pipetting from bottles with brightly coloured stoppers into minute plastic tubes, responding with furrowed brow optimistically and immodestly to the eager, breathless respectful reporters' questions about their breakthrough. Of course breakthroughs rapidly turn to dust and in any case the brilliant doctor is one of philosophy not medicine and the reporter is responding to a scientifically illiterate PR person's press release.

The research that directly addresses medical problems of patients in the here and now is mostly very different from this kind of laboratory work. Explicit experiments on people are so difficult and dangerous that attention focuses on observation as the method of enquiry. So, medical researchers are like astronomers. The difference between them is that astronomers regularly make successful predictions to ten decimal places. Doctors have to admit that Sam Goldwyn's joke, "making predictions is very difficult, particularly about the future", is true. This is not surprising because doctors practise biology, a subject in which short-term prediction is very difficult, and long-term ones are impossible. Evolution explains the latter, and life's molecular and genetic complexities explain the former. Nevertheless, doctors have to prognose, and they do not do too bad a job - better than investment analysts or meteorologists - because over several hundred years they have built a superb database about the quantitative behaviour of disease - how long it takes a cancer to kill, for example. The problem comes when we ask about causes. For the cancer was it diet or genes, or living under power lines? Clearly these questions can only be answered by "studies", and John H. Fennick's book tells us how to interpret them.

The central and key technique is, of course, statistics. This is what the book is mostly about.

What a curious - and misleading - title: Studies Show: A Popular Guide to Understanding Scientific Studies. My initial impression was that it might have something to do with Broadway musicals or the exhibition of cattle and sheep, or that we might be hearing about the "strong programme" or other aspects of the sociology of science. I looked forward to yet another attack on the ridiculous antics of nuttier social constructivists. Not to be, and not surprising, considering the author is a former member of the technical staff at the Bell Telephone laboratories. What we get is a statistician's attempt to put his subject into a medical context, focusing on risks affecting life chances. He addresses methodology and how to judge the appropriateness of statistical approaches. Interpretation is the central theme. This is, of course, the real difficulty of observational, comparative and correlational work. It is one thing to make a finding that maths says is unlikely to be due to chance. It is entirely another thing to know how seriously to take it. Macaulay epitomised it nicely "Iwe are not inclined to ascribe much practical value to that analysis of the inductive method which Bacon has given I A plain man finds his stomach out of order. He never heard Bacon's name. But he proceeds in the strictest conformity with the rules laid down in the 2nd book of the Novum Organum I 'I ate mince pies on Monday and Wednesday and I was kept awake by indigestion all night'. This is the comparentia et intellectum instantiarum convenientium. 'I did not eat any on Tuesday and Friday and I was quite well.' This is the comparentia instantiarum in proximo quae natura data privantur. 'I ate very sparingly of them on Sunday, and was very slightly indisposed in the evening. But on Christmas day I almost dined on them and was so ill that I was in a great danger'; this is the comparentia instantiarum secundam magis et minus. 'It cannot have been the brandy I took with them, for I have drunk brandy daily for years without being the worse for it'; this is the rejecto naturarum. Our invalid then proceeded to what is termed by Bacon as the Vindemiatio and pronounces that mince pies do not agree with him."

How well does Fennick cope with this? The book has a racy, American style and is Socratic without the pupil's responses. It starts with pitfalls that attend studies that need statistics. There is heavy irony in the authors "rules": "Extol or ignore the obvious, promote ignorance as knowledge, assign a risk (a nice essay on this difficult subject), sample conveniently (opinion polls here), and, never repeat the study". He defines terms, and looks in detail at one example, death-by-car. More methodology, including comments on the presentation of data as an art follows, and then there is a rather nice chapter on the problems faced by risk managers. I liked one example "Suppose that the Federal Aviation Administration, feeling very safety conscious, required that airlines print on each ticket, and post at the gate, the chance that your flight will crash: "Flight 942, Destination Las Vegas, Chance of Crash: 0.000015". Or, after protesting, the airlines won the option of posting: "Flight 942, Destination Las Vegas, Chance of Getting There: 0.999985". Well, you know how soon the unemployment lines would fill up with pilots and stewardesses I if you think you really believe in risk numbers, tell me, how small would the number posted at the gate have to be before you would board the plane?" The book finishes with more methodology, a detailed look at the statistics of heart transplantation, a bit more methodology, and a conclusion that one should look for small P-values, consistency with a "settling" of the result as work proceeds, and good agreement with experimental results. There is a glossary, good bibliographies, but no index - a seriously bad omission. Meta-analysis is not mentioned either.

Despite being annoyed by eruptions of an irritating style, I liked this book. Too flippant as a textbook and too long for the journalist, its main appeal should be to the reflective, questioning non-expert who suspects that Disraeli was right about "lies, damned lies, and statistics".

Hugh Pennington is professor of bacteriology, University of Aberdeen.

Studies Show: A Popular Guide to Understanding Scientific Studies

Author - John Fennick
ISBN - 1 57392 136 X
Publisher - Prometheus
Price - £14.99
Pages - 240

Register to continue

Why register?

  • Registration is free and only takes a moment
  • Once registered, you can read 3 articles a month
  • Sign up for our newsletter
Please Login or Register to read this article.


Featured jobs