Skip navigation
Link to Back issues listing | Back Issue Listing with content Index | Subject Index

Making too much from too little (Editorial)

One of the most difficult areas in our world of evidence is when there is not much information, and we have to make the best of what we have. Our sin, all too often, is that we make too much of too little. It is hard to stop ourselves from doing it, and the trouble is that, if we are enthusiastic believers, we can convince others of our case when the evidence just isn't there.

The Scottish legal system has a verdict that is neither guilty, nor innocent, but rather "not proven". It is rarely used and indicates when there is insufficient evidence to convict, but rather too much to be certain of innocence. It is a verdict that we might use profitably when thinking about evidence in our own field.

Making too much of too little is a sin too frequently encountered in systematic reviews, and one where "not proven" is much more satisfactory. Bandolier will return to this theme, but this month offers two examples. One is counselling in primary care , where the evidence from trials is anything but overwhelming. Some may consider that evidence from practice is good, but where is it? Another, the use of finasteride for haematuria in BPH where other causes have been excluded, shows that limited evidence from trials accords with limited evidence from practice.

On diagnostic tests

Readers have asked why Bandolier seems to have withdrawn from examining evidence about diagnostic tests. The answer was simply that we were having trouble finding better ways to describe the results, and even likelihood ratios were insufficiently intuitive for some. Using natural frequencies, and probability of disease with a positive or negative test, is one possible way forward.

Two examples are offered. One comes from a systematic review of studies of the Ottawa ankle and foot tests, clinical paradigms designed to exclude fractures and save on radiographs. Over a decade we now have lots of evidence, and the results are interesting. Another involves the results of the quadruple test for Down's syndrome screening.

Bandolier would like to move this forward, so any comments would be helpful, as would news of any studies on diagnostic tests readers (including industrial readers) would like to bring to our attention.

next story in this issue