Computerisation can remove almost all the potential errors in cervical screening. And, says Mark Gould, it can increase productivity threefold

It's a repetitive, stressful, closely monitored and low-paid job. When you get it wrong, patients are exposed to unnecessary anxiety or a medical intervention. In the worst case, they get ill or die and grieving families sue.

Hospital cytology departments have been demoralised by a series of cervical smear scandals which have hit public confidence in the service.

But the good news is that the screening programme works. That is the message chief medical officer Sir Kenneth Calman wants every health authority and trust to bellow from the rooftops.

Figures released by Sir Kenneth last week show that death rates are falling - down 26 per cent from 1994 to 1997. But 15 per cent of the female population have never had a smear. How dangerous that is, is shown by the fact that of the 800 women who died of cervical cancer in 1996-97, half had never had a smear.

Fear of pain and humiliation drive many women away from having a check- up. High-profile, systemic failures such as those in Kent and Canterbury, Exeter and Merthyr Tydfil do not inspire confidence.

So it is no wonder that for sound, practical and emotional reasons, clinicians are keen to let biotechnology and automation take the pressure off those who screen cervical cell samples taken from four million women every year.

Their hopes that new technology will renew confidence in the service - and perhaps boost recruitment - will be helped by a study to be published on 16 September by Professor Dulcie Coleman, head of cytology, at St Mary's Hospital, west London.

Her results from long-term tests on 20,000 women show that computerisation can remove both the drudgery and almost all the potential for error in the painstaking examination of slides, each one of which contains some 400,000 cells.

Professor Coleman's study produced results as consistent and reliable as a good human screener and showed that the computer can increase productivity threefold.

She used a system called PAPNET, which uses a computer linked to a microscope on a motorised platform to scan all 400,000 cells in each sample and grade them on a sliding scale of abnormality.

When the scan is complete, the human cytoscreener is presented with two computer screens, each containing 64 tiles showing what the computer has determined are the most abnormal cells.

`The cytoscreener takes about two minutes to look at all 128 tiles, compared to about eight minutes for a conventional slide. So the time saving is substantial,' Professor Coleman says.

Usually, around 70 per cent of the tiles are negative, so the cytoscreener only needs to examine 30 per cent of the sample tiles under a microscope.

Professor Coleman has co-ordinated the trial of 20,000 smears gathered from five hospitals in Oxford and the North and South Thames regions, comparing PAPNET to manual screening.

`Results show convincingly that the system is as sensitive as manual screening,' Professor Coleman says.

Just as important, it is faster and easier to use allowing the screener to get through three times more work than the manual system. Economies of scale mean that one central laboratory could process the slides and send computer images of the 'abnormal' slides back to the local unit.

PAPNET costs 200,000. Manufacturers Neuro Scientific Instruments claim that a big laboratory handling 60,000 slides a year would have unit costs of 1.30 per slide.

The system is also being tested by a team at the Royal Liverpool Hospital where pathologist Peter Smith says: `If the trials are successful, we hope that it will be used as our primary screening tool. That will change the way we work.'

For the cytologist, the added advantage is that it is easier and more comfortable to look at a computer screen than to look down a microscope.

'It will cut down the laborious and repetitive nature of the job and the psychological habituality of looking at the same thing day after day,' Dr Smith says.

However, Autopap, one of several other systems under trial in the UK, is a 'black box' which does away with the need for a screener altogether.

Smear samples go in at one end and a set of readings come out at the other, flagging up any problem area

One of the main stumbling blocks has been trying to decide what would be an acceptable false negative rate for the computer to match those of human screeners.

Professor Coleman is satisfied that the PAPNET system's neural network technology produces acceptable false negative and false positive rates - an important factor given the emotional stress to a patient of undergoing further tests if they are diagnosed as having a potential cancer.

Another advance in cancer research involves establishing the relationship between cervical cancer and human papilloma virus, HPV, which in some groups of women has a 50 per cent prevalence.

Researchers hope the HPV test might be used to further investigate the cancer potential of borderline smears. Almost all invasive cancers of the cervix contain genetic sequences of this virus.

The big question for researchers is how much HPV can be relied on as a marker for cancer.

Royal Liverpool Hospital consultant pathologist Dr Simon Herrington says the Dutch are already running parallel HPV testing with smears.

'If all cancers contain the virus, then detection of the virus should be a sensitive method for the detection of cancer.

'However, we know that many women infected with the virus do not have cancer and are unlikely to develop it. The challenge is therefore to devise a test which is both sensitive and specific for the detection of women with cancer or of those with a high risk of developing cancer.'

The answer might lie in the potency or viral load of the sample. The greater the viral load, the greater the risk. A US company already markets a home-use vaginal swab which is used like a tampon to detect HPV. Dr Herrington thinks it is too early to tell whether it will be useful here.

A technique known as liquid based cytology (LBC) means that one smear could provide sufficient material for a good quality slide, free of contaminants such as blood and other cells, and a second sample for a HPV test.

The smear is taken and the spatula sent to a laboratory where a technician prepares a clean slide, free of blood or cells stacked on top of each other. The screener, human or computer, receives a 'monolayer' - a single layer of easy-to-read cervical cells.

Using the manual system means the person taking the smear merely scrapes the edge of the spatula against the slide and takes pot-luck that an adequate sample has been transferred.

Professor Euphemia McGoogan, clinical director of pathology at Edinburgh Royal Infirmary, says: 'Using conventional preparations, a large amount of the sample is thrown in the bin with the spatula. Using LBC, you improve the content and quality of the sample. One smear can be used for a variety of different tests.

`Most borderline categories are false positive. That means a woman might need a second or third smear, or even a colposcopy, even though they may well be perfectly normal. The HPV test might mean we could do without these.'

The bad news is that technology will not be able to replace what can be the painful and undignified process of obtaining a cell sample using a spatula. Nor will it produce a 100 per cent-faithful screening process. There will always be deaths due to cervical cancer.

However, technology may provide the psychological crutch for a profession demoralised by negative publicity.

`Public confidence has got to be restored. Lots of screeners have left, so recruitment is a problem. You can't just pick cytoscreeners off a tree. It might be that new tools which can be phased in at centres of excellence will improve the situation,' Professor Coleman says.