Let's improve medical science, for free

What will the election mean for British health, science and education? One thing everyone agrees on is that the next few years will be a time for belt-tightening, and the issue of funding and cuts has dominated debates so far. But there's one simple thing that the next government can do to improve our health, and ensure we get value for money from British medical research: a law requiring the registration of clinical trials. Randomised, placebo-controlled trials are the foundation of modern medical science. They are how we know which treatments work. A drug must pass a certain number of trials before it's approved to be sold, and the National Institute of Clinical Excellence (Nice) uses trials to make its often controversial decisions about which drugs are worth the money. Clinical trials determine almost every aspect of the treatment we receive. Yet there's a big problem with the current system of clinical trials: selective publication. Suppose a company runs a trial of a new drug it has just developed, but unfortunately, the results show that the drug doesn't work – people taking the drug did no better than people given placebo sugar pills. If people found out about that data, it would be bad news for the company: drugs that don't work don't sell. So there would be a strong incentive not to release the findings. Instead, it could file the inconvenient results away and run another trial, and if necessary another, until, just by chance, the drug seems to work – and then publish those positive results. Even for drugs that do work, publishing only the most positive results makes them seem to work better than they really do. This is called publication bias. It's bad for science, and bad for everyone's health. By making ineffective treatments seem to work, making expensive new drugs seem to be better than old, cheap ones, and concealing evidence of side effects, selective publication undermines the whole point of doing research in the first place. By its nature, publication bias is difficult to detect. We'll probably never know the full extent of the problem, but studies have shown that it has happened, and continues to happen, on a large scale. Contrary to popular belief, drug companies are not the only culprits. One recent study found evidence that it occurs in trials of Cognitive Behavioural Therapy for clinical depression, and publication bias in trials of complementary and alternative medicine has also been reported . It's everywhere. Yet it's legal, almost everywhere in the world. By contrast, if someone were to simply make up some positive results, or go through data fiddling the numbers to make them "better", that would be scientific fraud, an extremely serious matter. Doing several trials and only publishing the best ones is, in its effects, just as misleading – but it's allowed. Fortunately, there's a simple, powerful, and cheap solution. In 2007, the US passed a pioneering law called the Food and Drug Administration Amendment Act (FDAAA). This requires anyone running a trial of an investigational drug or medical device, that includes any American patients, to register the details of the study in a public database before it starts. Anyone can view the database online . No more disappearing trials. Trial sponsors are also required to release the results when they become available, whether they're positive or not. Failure to comply can be punished by fines, and/or withdrawal of funding. The FDAAA is an important piece of legislation. However, neither Britain, nor the EU, has an equivalent system in place. There's already an EU clinical trial database , but unlike the American one, it's not open to the public. Government health authorities have access to it, but doctors, scientists and patients don't. This largely defeats the point. There are also a number of international public databases, and since 2005 the editors of most of the world's top medical journals have refused to publish trials that weren't appropriately pre-registered. This was a good step, and a welcome sign that publication bias is being taken seriously. But it's not enough, because plenty of other journals still publish unregistered trials. Ultimately, only governments can stop publication bias. The next parliament should follow the American example and require public registration and data reporting for all clinical trials involving British patients. It would hardly cost anything. It would, at a stroke, improve British science, and everyone's health.