HEALTH: SECOND OPINION

IN AN ERA of computers and precision instruments, few people feel comfortable about accepting that even experts still make mistakes. Public attention is concentrated on those rare individuals who make very few mistakes and so achieve world-ranking success as Grand Prix racing drivers, tournament golfers or chess champions; but in the real world of non-extraordinary people, errors occur frequently and often predictably. The inevitability of error has been forced into doctors' consciousness as they have become subjected to medical audit - the process of checking the reliability of their diagnoses and the frequency of complications in patients they have operated on. One recent research study looked at mammography, the test used to screen women for breast cancer. The X-ray film is examined by a radiologist (a doctor with special training in the interpretation of X-ray images), who looks for early signs of cancer and, if these are seen, recommends further tests. The researchers took a series of mammograms done several years earlier (so that there could be no doubt about whether or not the women concerned had or had not gone on to develop breast cancer). They showed the films to 10 radiologists who regularly reported on mammograms and found substantial disagreements. Twenty-seven of the women had breast cancer, but the radiologists made the diagnosis in only three-quarters of these cases. Further tests were recommended for 36 per cent of the women without cancer. None of the radiologists achieved more than 85 per cent accuracy. (A recent survey in a Manchester hospital found that up to 30 per cent of breast cancers were "missed" by radiologists.)

Commenting on the results, the New England Journal of Medicine stressed that mammography (like many tests used in medicine) should not be expected to be consistently "right" - and we should not be surprised by the variability among radiologists in their interpretation of X-ray films. The more cautious ones will report many films as suspicious of cancer and so will refer many women without cancer for further tests - but will detect virtually all the women who do have early cancers. Others may only recommend further tests when they are virtually certain a cancer is present: they will refer far fewer healthy women for "unnecessary" tests but will miss some early cancers.

Variations in interpretation can have another cause: sometimes there is an abnormality on the X-ray film but a radiologist may not detect it. This applies to all sorts of X-rays - examinations to detect fractures after an accident, for example - and to other diagnostic tests such as electrocardiograms or ultrasound scans. Even experienced radio-logists miss abnormalities seen by their colleagues, though when the abnormality is pointed out the same observer sees it easily enough. This is in part a psychological pheno-menon, rather like being unable to see a bunch of keys on a desk cluttered with other objects, but other factors play a part, among them fatigue.

Mistakes in diagnostic tests such as mammography can be reduced in number. Radiologists' skills can be improved by further training - provided that the time and money are found to provide the training programmes. Research has shown that if two radiologists are asked to examine the films, they will between them detect more of the abnormalities; the rate of detection of cancers in Swedish mammo-graphy progammes went up 15 per cent when the films were seen by two doctors. But a high detection rate is always associated with a higher rate of false positives, each of which leads to a woman having to suffer the anxiety of having tests repeated and maybe a specimen of breast tissue removed before she is told that she doesn't have cancer.

The important message is that greater public awareness is needed of the fallibility of most medical tests. Deciding if a small shadow on an X-ray film is abnormal or not is a matter of judgement, and no doctor will always get it right. Those who get it wrong should not necessarily be condemned as lazy or incompetent - though they may, of course, be both.