In 1994, Canadian researchers looked at 69 trials of anti-arthritis drugs funded by drug companies and published in prominent medical journals. In every case the drug made by the company was as good as the comparative treatment, and in a quarter of the trials it was better. Not once did a company fund a trial that proved unfavourable to it.

Yet the whole scientific point of doing such trials is to answer so far unanswered questions. Supposedly, researchers conduct trials when they are in a charmed state called "equipoise", which means they are genuinely uncertain which is the best treatment. If they think one treatment is better than another, then they shouldn't be conducting the trial.

A review published in 2003 found 30 studies that had compared the results of trials funded by drug companies with those funded from other sources. Trials funded by companies were four times more likely to have results favourable to them than those funded by others.

Yet the technical quality of the trials funded by drug companies was always as good and often better than the quality of those funded by other sources. This is not surprising, as drug company trials are tightly regulated. There are explicit high standards, and companies can afford to hire the best to conduct the trials.

How then do companies usually manage to fund research that is favourable to them? An answer is supplied in a recent issue of the BMJ by Dave Sackett and Andy Oxman, two tireless campaigners for the better use of scientific evidence in medicine. They have founded a spoof company called Harlot - which stands for How to Achieve positive Results without actually Lying to Overcome the Truth. They created the company after it finally dawned on them that "being good and being poor are causally related: being good doesn't pay".

Harlot plc promises to give drug companies and others the results they want. Your drug may be wholly ineffective, Sackett and Oxman promise, but as long as it isn't a lot worse than a sip of triple distilled water, then Harlot can produce positive results from a trial. Importantly, these results are not usually achieved by doing poor quality trials. The trick is in the question asked and the design of the trial. Sackett and Oxman, both experts on the design and analysis of trials, describe 13 methods for getting the results you want.

One of the commonest methods is to test a new drug not against an effective treatment but against a placebo. Ironically, regulators often require companies to do this. But what matters to patients is not whether a company's drug is better than nothing, but whether it is better than established treatments. Companies are nervous about these "head-to-head" trials, particularly if many drugs are being tested - because there may be only one winner and many losers. A huge publicly funded head-to-head trial of treatments for high blood pressure was published recently and threw companies into a tizz because it showed that long-established drugs that are off patent were better than newer, much more expensive drugs.

A company gets huge benefit from showing that its drug is better than a competitor's. But the company needs to control the trial, and Harlot suggests that a company compares its product with an inadequate dose of a competitor's product. This may have been the reason why previous trials on drugs for high blood pressure suggested that newer drugs were better.

A variant on this technique is to compare the drug with an excessive dose of the competitor's product: it is then possible to show that the company's drug has far fewer side-effects (because side-effects are more common with higher doses of a drug). This may have been the method for showing that new and expensive drugs for schizophrenia have fewer side-effects than older drugs.

Perhaps the most common method to avoid unfavourable results is to make sure that a trial is not big enough to show that a competitor product is either better or worse. Such trials are very common, and Silvio Garattini, a leading Italian researcher and critic of the drug industry, has proposed a consent form for them: "I understand that this trial is worthless for science and medicine, but will be of great use to the marketing department of Shangri-la Pharmaceuticals."

All this matters greatly because 70% of trials in major medical journals are funded by the drug industry. Often companies will buy reprints of these articles to use in promoting their drug. Sometimes they may spend up to £750,000.

Virtually all research on drugs is funded by the industry, because governments have taken the view that public money can be better spent elsewhere. The end result is that information on drugs (on which Britain spends £7bn a year) is distorted.

The Harlot article was written to amuse, but is as deadly serious as anything else published in the BMJ in the past 10 years. The public is being regularly deceived and exploited.

· Richard Smith is the editor of the British Medical Journal