By Robert Matthews
Two huge industries affecting the lives of millions of people are currently subject to big health alerts. Concern over serious side-effects has cast a long shadow over promising new painkillers, known as cox-2 inhibitors, developed by the pharmaceutical industry. Evidence linking the drugs to an increased risk of heart attacks led the US giant Merck to withdraw its version, known as Vioxx, from the market last September, and an investigation by the -US Food and Drug Administration is currently under way.
More recently, it was the turn of the UK food industry, with the discovery of traces of a banned dye known as Sudan I in a sauce made by Premier Foods, a leading UK supplier. In the ensuing health scare, the UK Food Standards Agency found that hundreds of products had been inadvertently contaminated by the dye, which has been linked to cancer.
As the initial furore starts to fade, both these health alerts are being seen primarily as wake-up calls to business and regulators alike about the monitoring of product safety.
In the case of cox-2 inhibitors, the FDA looks set to allow their continued use - albeit with much sterner safety warnings to protect those most at risk from side-effects. Meanwhile, as shops and supermarkets in the UK hunt down produce contaminated with Sudan I, the FSA has continued to stress that the risks involved are "very small".
As well it might, for it is now clear that the scientific case against Sudan I is far from compelling. Laboratory safety tests involved feeding rodents with levels of Sudan I equivalent to human consumption of the sauce that triggered the scare at a rate of three tonnes a day for two years.
Even after such gargantuan exposure, the animals failed to produce consistent evidence of a cancer risk. Other tests hinted at links with bladder and liver tumours - but only after the dye was injected directly into the organs of laboratory animals.
While the scientific basis for both the Sudan I and cox-2 inhibitor health scares may be contentious, they have highlighted the need for close surveillance and prompt action if problems emerge. At the same time, however, an even more fundamental question has gone begging: just how reliable are animal tests of product safety?
In the case of food safety, the relevance to humans of animal tests involving colossal intakes or direct injection into organs is clearly questionable. The use of animals in drug safety testing raises altogether more complex issues, however - as the cox-2 painkillers controversy shows.
In line with standard practice, Vioxx and the other drugs were tested in at least two different types of animal before entering clinical trials with humans. One of the main aims of such "pre-clinical" testing is to detect signs of serious side-effects. In the case of the cox-2 drugs, the animal testing failed to warn of the cardiovascular effects that have prompted the current furore. Indeed, several animal studies suggested the drugs would actually reduce the risk of such side-effects.
So what went wrong? Anti-vivisectionists have been quick to voice their standard objection: animals are not humans.
For all its familiarity, it is an argument that does have relevance to the cox-2 inhibitors. In 2000, barely a year after the launch of Vioxx, a study of more than 8,000 patients suggested that those taking the drug faced a significantly increased risk of heart attack. Yet subsequent animal-based research continued to suggest such drugs could reduce the risk - prompting even Merck's experts to concede in The American Heart Journal that: "The relevance of these animal models in predicting effects in humans is uncertain."
It is becoming clear that such uncertainty extends far beyond one class of blockbuster drug.
Leading journal Nature Reviews Drug Discovery last year published a review of the evidence that animals are reliable predictors of toxic effects in humans. The authors found that the evidence was "fragmentary", with the few published studies pointing to "significant over- and under-prediction of adverse effects from animal studies that varies with the particular organ or system".
The review also highlighted the lack of basic data needed for a scientific assessment of animal testing, such as measures of predictive power and their statistical significance.
As it stands, the evidence suggests animal tests may be unduly sensitive, wrongly predicting toxicity in compounds that are in fact harmless to humans. If so, it would be an ironic twist to the widely held belief that tests of animal are crucial to the advancement of medicine, as they may in fact be blocking the development of many safe and effective new treatments.
Yet in the absence of large-scale studies comparing drug responses in animals and humans, it is impossible to know. Supporters and critics of animal testing continue to trade anecdotes of individual successes and failures, most published studies being so small they lack statistical credibility.
In another irony, the drive to minimise the use of animals has compelled researchers to draw conclusions from meagre evidence. For example, the studies designed to investigate the effect of cox-2 inhibitors on cardiovascular risk typically involved fewer than 20 mice.
The authors of last year's review called on regulatory bodies and drugs companies to publish data currently languishing in their files. Whether the outcome will confirm or confound the view that animals usefully predict human reactions remains to be seen.
What is clear is that, given the paucity of systematic evidence, it is not necessary to be a placard-waving protestor to harbour doubts about the validity of animal testing.
The writer is visiting reader in science at Aston University, Birmingham
* Detecting nasty side-effects is harder than it looks.
* The health scares over cox-2 drugs and the food dye Sudan 1 have highlighted the challenge of assessing health risks from limited data. While studies involving huge numbers of patients or laboratory animals are clearly better at detecting side-effects than small ones, they are also far more expensive and time-consuming. However, the ability of a study to detect risk does not increase pro-rata with size: to double the sensitivity, the required number of patients quadruples.
* Worst of all, estimating the required numbers demands some guessing at the likely level of risk - a bad guess raises the danger of the study being "underpowered", lacking the numbers needed to detect the true level of risk.
* One solution is to set up a trial so large that it is sure to have a reasonable chance of detecting serious side-effects in one patient out of every N taking the drug. Statistical theory then shows that a comparison of four times N-squared patients taking the drug with the same number taking a placebo will do the trick. But for blockbuster drugs such as Vioxx, side-effects affecting one in 1,000 patients constitute a health alert and detecting that level of risk needs a study involving millions.
* The only way of acquiring such vast numbers is for pharmaceutical companies and regulators to keep drugs under close surveillance long after approval.