But the newspaper reports didn't tell the whole story - not surprisingly, as the Sun's was only a single sentence. According to reviewers at the NHS centre for reviews and dissemination, the articles misleadingly reported findings of a research study which itself had weaknesses in methodology.

Some of the research conclusions were "based on poor quality evidence and should be interpreted with caution", the reviewers concluded.

Although the scientists used "a reasonably well-conducted randomised controlled trial their findings relied on participants filling in questionnaires about fatigue and digestive symptoms". Such outcomes are "very subjective and could be subject to bias", the reviewers said.

They also pointed out that the study was sponsored by the Kellogg company, maker of All Bran.

The breakfast cereal study is one of the latest pieces of research to come under the scrutiny of the centre for reviews and dissemination's 'hitting the headlines' service.

Unsurprisingly, hitting the headlines pulled out all the stops for the MMR debate. It ran a two-part critique of articles in the Guardian, the Daily Telegraph and the Times summarising research into the vaccine, autism and bowel disease.

The broadsheets get some credit for accuracy: the Guardian reported a large study showing no increased risk of autism after the MMR jab "briefly but accurately".

An earlier investigation, however, is critical of the Daily Mail's coverage of links between the measles virus and bowel disease. "The Daily Mail misleadingly reports that the children in the study were autistic and focuses the rest of the report on the MMR vaccine," it concludes.

The centre's reports appear at the rate of one or two a week on the web front page of the national electronic library for health, run by the NHS information authority.

This week, the authority approved funding of £67,548 for the service, which has been running as a pilot, to continue until the end of March 2003.

The team behind hitting the headlines does not see itself as part of the NHS spin machine. It investigates articles based on published research into the effectiveness of medical interventions or diagnostic tests; the sort of stories that appear under headlines suggesting either a miracle cure or a new health scare.

"We investigate the type of article that worried patients might take along to GP surgeries," said Alison Booth, dissemination officer at the centre, which is based at the University of York.

The centre concentrates on articles that appear in more than one newspaper and which are based on research that it can scrutinise.

"We identify a story that fits the remit and get hold of the published research," Ms Booth said. Independent reviewers then appraise the research and the way it is portrayed in the article.

The result of the investigation appears as a bullet-point summary and a more detailed critique within days of the original coverage. "We aim at a 48-hour turnaround, but that depends on getting hold of the journals," Ms Booth said.

This can be difficult in the case of more obscure publications. "We go to extraordinary lengths to get hold of the original research," she said.

Perhaps surprisingly, the fourth estate comes out reasonably well from the centre's scrutiny. "The article is usually reasonably accurate," Ms Booth said, "the problem is more often in the headline on top."

Monitoring the quality of health information in the mass media was one of the main reasons for setting up the national electronic library for health (NeLH), which was conceived in 1998 and has been running as a pilot project for just over a year.

However the NeLH is still waiting for the Treasury to approve the business case for a full-blown service. This is now not expected until 2003, an NHS information authority spokesperson said.

In a related initiative, the Centre for Health Information Quality, based at the Help for Health Trust in Winchester, is developing a seal of approval, called the TriangleMark, to show where information meets its standards. Its funding stream, however, is still uncertain: the centre is hoping that the TriangleMark will be self-funding.

However research published this week suggests that, if the NHS is seeking to rid the web of inaccurate health information, it faces an uphill task.

A study of 121 health websites reported in this week's British Medical Journal finds that only 49% displayed the date when the information was put on the site, and 82% did not show the relevance or validity of it.

The five topics looked at were chronic obstructive pulmonary disease, ankle sprain, emergency contraception, menorrhagia (abnormally heavy menstrual bleeding) and female contraception.

Lead author Khalid Khan, a consultant at Birmingham Women's Hospital's education resource centre, said: "Our study shows that features of website credibility have only slight or, at best, moderate correlation with accuracy of information in five common health topics."