How much is it safe to drink? For anyone who likes a couple of pints after work, a large G&T when they get home or a few bottles of Burgundy at the weekend, this is a question for our times. Research published last week, which received widespread press coverage, purported to give some answers. It was a study of more than 10,000 civil servants followed for 11 years in the famous Whitehall II study, led by one of Britain's foremost epidemiologists, Professor Sir Michael Marmot of University College, London.

But there was something strange about the study, published in the journal Addiction. The data appeared to contain much that was good news for drinkers yet the paper was couched in negative terms with warnings about the high risk of heart disease to women who drank heavily.

This was the line followed in most of the press reports. "Ladettes at high risk of heart disease" said the Daily Mail and others had similar headlines. It was a story that chimed with public concern about rising levels of drinking, especially among the young, and reinforced the Government's "sensible drinking" message.

But a glance at the tables that accompanied the paper suggested a different message, different even from the spin put on the data by the researchers themselves.

The paper said that men and women who drank heavily - more than 31 units a week (16 pints of beer) for men and 21 units (three and a half bottles of wine) for women - were at approximately twice the risk of dying prematurely.

But the tables showed that while the male heavy drinkers were at increased risk (though by 40 per cent not 100 per cent), the women who drank most - more than three and a half bottles of wine a week - were at lower risk than those who drank one bottle a week .

The paper also said the "optimal drinking frequency was once or twice a week". But that was only true for women. For men, those who drank "almost daily" had a lower risk of heart disease and premature death than those who drank less often.

Moreover, binge drinkers of both sexes, who liked to sink at least five pints of beer (men) or most of a bottle of wine (women) in a session, had a lower risk of dying and of developing heart disease than those who drank more moderately, according to the tables. Yet the paper said those who drank in binges "increased the risk of mortality and coronary heart disease".

I put these discrepancies to the lead author of the study, Annie Britton. She defended the paper and said some of the data were not statistically significant because the numbers involved were too small. In response to the detailed points, she said: "You haven't said anything that is factually incorrect but I would urge caution on that interpretation."

But I wonder whether the anxiety not to undermine the "sensible drinking" message did not cloud the researchers' judgement. As with Lord Hutton's suggestion that the Joint Intelligence Committee may have "subconsciously" got too close to Number 10 in drawing up the Iraq dossier, did the researchers subconsciously edit their paper to fit with the received wisdom on drinking?

We rely on scientists to tell the world as it is, not as it is meant to be. That is the basis of our trust in science. When we suspect scientists are operating to someone else's agenda, public confidence is undermined - witness the row over MMR vaccines.

But there is more than one way to skin a cat and the same set of data can be interpreted in different ways. When two or more medics are gathered together you can guarantee to have a score of opinions.

Last week three doctors raised doubts about Dr David Kelly's suicide, implying that he would not have lost sufficient blood from the severing of the ulnar artery in his wrist to cause his death. At least one senior accident and emergency specialist backed this claim, but it was attacked by three forensic pathologists who insisted the evidence showed Dr Kelly did indeed take his own life, a view confirmed by Lord Hutton.

There is less certainty in science than most people would like and that can hold dangers. We saw it in the Angela Cannings case, which has led to the biggest upset in legal history and the review of 258 cot death cases.

What could be worse for a bereaved mother than a mistaken charge of murder? What seems to have gone wrong is that scientists involved as expert witnesses, who were naturally eager to help the courts reach a decision, subconsciously worked to their lawyers' agendas, and reached decisions that were firmer than the facts could support.

The result has been not only a dreadful miscarriage of justice, but also a further undermining of public trust in science. And the real villain of this as of many similar stories is our inability to tolerate uncertainty.