IN THE red corner, big drug companies hell-bent on profit regardless of the consequences. In the blue corner, plucky safety regulators and scientists determined to uncover the truth about the drugs the companies are selling.
If you believe everything you read, this is probably your view of the controversy over the safety of painkilling drugs like Vioxx. Known as Cox-2 inhibitors, these drugs were designed to bring relief to millions of arthritis sufferers around the world. Now they stand accused of increasing the risk of heart disease and stroke. And since Merck, the maker of Vioxx, voluntarily withdrew it last September, medical journals have been falling over each other to publish evidence that the company and drug regulators could have taken action far earlier.
The debacle is strikingly reminiscent of what happened more than 20 years ago with Opren, another painkiller for people with arthritis. Fortunately for the makers of Cox-2 inhibitors, the parallels aren鈥檛 perfect. While Opren sank into a morass of litigation, a panel of experts advising the US Food and Drug Administration (FDA) has recommended that Vioxx and similar drugs can stay on the market, albeit with much stronger safety warnings.
Advertisement
Yet at the heart of the furore lies a disturbing tendency throughout the medical community to base big claims on meagre evidence. The plain fact is that when it comes to detecting side effects, most clinical trials are too small to pick up anything less serious than the decimation of patients. It is almost impossible to be sure that a drug that appeared safe in trials will not cause ill effects when released onto the market.
It all comes down to a simple rule from basic statistics. Standard statistical theory shows that to be sure of having a decent chance of detecting side effects that occur in 1 in n people, you need a clinical trial involving around 8 times n2 people. Thus to detect side effects affecting 1 in 1000 patients 鈥� enough to spark a major health alert with mass-market drugs like Vioxx 鈥� you would need to recruit several million people to take part in your trial. In reality, clinical trials are nowhere near that size: most involve fewer than 1000 patients. That is often big enough to show the effectiveness of drugs, but such trials can struggle to detect side effects in as many as 1 in 10 patients.
These statistical facts are in danger of being lost in the controversy over Vioxx. Instead, we are seeing an outbreak of 20-20 hindsight from some quarters of the medical establishment. In December, The Lancet published a major review of Vioxx clinical trial results that apparently showed that convincing evidence of the heart-attack risk could have been spotted four years before the drug was withdrawn, had anyone bothered to look. An accompanying editorial claimed the results revealed 鈥渓ethal weaknesses鈥� in the FDA鈥檚 monitoring procedures. An editorial in The New England Journal of Medicine cited an internal FDA report showing that the agency had evidence of a 2.8-fold greater risk of potentially lethal side effects as early as 1999.
What the authors of the editorial did not make clear is that this suggestion of a 2.8-fold increase was based on so few patients that it carried no statistical credibility. Likewise, The Lancet鈥榮 editorial overlooks the fact that none of the early trials of Vioxx was big enough to detect anything but the most egregious side-effect levels. Evidence of potential risk did emerge in 2000, with the publication of results from a large trial of more than 8000 patients, but this prompted no outraged response from The Lancet at the time.
Merck insists there was no convincing evidence of extra risk until September 2004, when it voluntarily withdrew Vioxx after a trial of more than 2500 patients again indicated the drug was significantly increasing the risk of serious side effects. Critics are equally insistent that the evidence was compelling well before then. What is beyond question is that clinical trials are typically far too small to detect worrying levels of side effects.
These statistical facts of life can no longer be waved away as mere theory. It took a huge study involving almost 17,000 women, abruptly terminated in 2002, to reveal convincing evidence of the risks to women from hormone replacement therapy. The dark side of the class of antidepressants known as SSRIs, long suspected of increasing the risk of suicide attempts, is also becoming clear following results from more than 87,000 patients.
鈥淲hen it comes to detecting side effects most trials are too small to find anything less than decimation鈥�
We face a choice. We can wait for years while the vast numbers of patients needed to gauge safety accrue from countless small trials. Or we can adopt a more grown-up attitude to risk, and accept that every drug has its downside, approve those whose risks we deem acceptable, then closely monitor the evidence of risk and benefit as it emerges. Neither option is cheap or easy.
But as the old saying goes, if acquiring knowledge seems expensive, try ignorance.
