The mainstay of safety testing of GM food is, and has always been, compositional analysis. This involves breaking down the food into components roughly similar to the assumed products of digestion, and measuring all the main groups of substances formed. This technique works well for assessing the quality of animal feed because livestock have a restricted diet and any deficiency in a major nutrient will lead to ill-health. For human beings, normally on a much more varied diet, it's not entirely relevant.
Compositional analysis will also routinely extend to any toxins known to be associated with the food. Novel, unexpected or unknown toxins such as might be generated by a GM plant won't be included in these measurements. The novel ability of the novel life-form to support pathogens will also not be revealed.
Another routine quality-control test performed on GM-commodity crops, such as maize and soya, is livestock feed-conversion. Such trials are a legal necessity prior to marketing of feed to ascertain its nutritional value, and are often used as the main or only, 'proof' of safety of a GM crop. However, their relevance to man is very remote, and since modern farm animals fattened on GM grains are as a rule slaughtered in youth, they don't live long enough to reveal chronic effects from their feed.
The only way to be certain that harmful elements are absent from food is to feed it first to an animal and then to a person under controlled conditions and watch what happens.
Feeding studies on animals are extremely difficult to perform. Their scope is severely limited by the natural diet of the test animals, and the extrapolation to humans is always uncertain. Healthy animals will be least likely to demonstrate adverse effects, while stressed or aged animals will be most susceptible. Choice of the strain of laboratory animal used will affect the data: inbred animals give more uniform results but may be unusually insensitive to the toxin tested, outbred mice are always variable, so that low levels of reaction may be significant. The standard test is a 90-day feeding study on healthy young rats. This will only be certain of revealing acute problems.
In practice, the routine GM feeding studies rarely use whole foods but are restricted to tests using analogue proteins isolated from non-food substitutes, GM bacteria. In this form, the test material is conveniently plentiful, easy to feed in standard amounts to standard laboratory rodents at a standard dose for a standard period of time. They are, however, irrelevant as a test for the safety of a variable whole food eaten by variable humans in variable amounts over variable times. Animals can be crammed full of the pure GM protein substitute, and if it doesn't kill them, this becomes an excuse for no further safety-testing of any GM crop expressing a similar protein, even if the new crop is stacked with several varieties of novel protein.
Very occasionally, EU regulators will require a 90-day feeding experiment using whole GM food during which indicators of toxicity are checked. The 90-day trial was developed to assess the effects of chemicals such as drugs, and is excellent for picking up the first symptoms of acute toxicity. It is much less appropriate for whole foods, because these consist of a complex of interacting materials able to mask toxic effects in the short-term. It is entirely inappropriate for investigating chronic health impacts because the animals' ability to withstand or overcome physiological challenges (such as, infections or stress) is not any part of it.
What the EU regulators don't require are continuous breeding studies on populations of laboratory animals fed a GM diet for several generations. (Note. Multigenerational studies mentioned in biotech industry applications to market GM refer to separate feeding trials performed on successive generations of animals.) This may be a very serious omission, because the only continuous breeding study carried out using mice fed a Bt-toxin containing maize(1) and rats fed herbicide-tolerant soya (2) found seriously reduced survival in successive litters.
A 2007 review of the published papers on feeding studies which examined signs of toxicity found only 27 covering 18 different GM crops and was able to summarise them in 1½ pages. (4) Of those listed, only 9 examined crops which have been commercialised (most of these nine are described 'Is RR soya safe?). All of the studies were limited in the scope of parameters recorded, the numbers of animals and duration of feeding.
In summary, the safety testing of GM food is too brief, uses inappropriate substitute materials, is based on tests not designed to be applied to the human condition, and too few of them have been carried out to publishable standards.
How did we come to get it so wrong? Well, for starters, the test protocol was designed by Monsanto. (3)
1.A. Velimirov et al.(2008), Biological effects of transgenic maize NK603 x MON810 fed in long term reproduction studies in mice, Institut für Ernährung, Forschungsinstitut für bilogischen Landbau, Vienn
2.Ermakova I. V. (2009), Influence of soy with the gene EPSPS CP4 on the physiological state and reproductive functions of rats in the first two generations, Russian Academy of Natural Sciences: Modern problems of science and education, 5
3.Andrew Cockburn (Monsanto), Assuring the safety of genetically modified (GM) foods: the importance of an holistic, integrative approach, Journal of Biotechnology, Volume 98, Issue1, 11th September 2002
4.Domingo J. L. (2007), Toxicity Studies of Genetically Modified Plants: A Review of the Published Literature, Critical Reviews in Food Science and Nutrition 47