Print

26 March 2003

KEY GM CROP EXPERIMENT 'LACKS STATISTICAL POWER'

http://www.newscientist.com/news/news.jsp?id=ns99993547
Key GM crop experiment 'lacks statistical power'
19:00 26 March 03

Exclusive from New Scientist Print Edition

After four turbulent years, the largest experiment on genetically modified crops in the world is all but over, with the fieldwork done and research papers written.

However, the scientists may get little chance of a breather. Opponents of Britain's farm-scale trials have chosen the lull before the first findings are published in a few month's time to mount the most detailed attack on the science yet.

The trials were set up to address fears that the broad-spectrum herbicides used with many GM varieties would harm farmland wildlife. But a 47-page dossier of arguments and allegations, seen in advance by New Scientist, claims the multimillion-pound experiment cannot succeed in delivering a definitive answer.

The trials are likely to play a key role in determining whether – and on what basis – the British government sanctions commercial growing later in 2003. For now, the findings remain strictly under wraps. But based on the trials' methods and first-year pilot observations, which are already in print, campaigners claim the prospects for a clear verdict are bleak.

"That the trials look set to produce uncertain results is not a reflection on the scientists involved," says Pete Riley, whose team at Friends of the Earth compiled the report. "Rather it highlights the inherent problems of embarking on politically motivated science."

Field furore

Opposition to the trials is not new. From the start activists periodically ripped up trial crops while others claimed farmers were biasing the outcome by treating GM fields with less herbicide than would be used commercially – a charge the trial scientists rejected. Now, with the endgame in sight, campaigners are keen to shift the focus onto what even some neutral experts see as the experiment's potential Achilles' heel: its statistical power.

The trials involved farmers growing both conventional and GM varieties of sugar beet, maize or oilseed rape (canola) in neighbouring fields. There were up to 25 sites per crop per season, which researchers would regularly visit to count weeds, beetles and other biodiversity "indicators". The goal was to discover if the GM fields held significantly less, or more, wildlife than those with conventional crops. But what counts as "less" or "more"?

For weeds and insects, the scientists designed the trials to be sensitive enough to have an 80 per cent chance of detecting 1.5-fold differences between conventional and GM fields. However, the report claims this sensitivity target is unlikely to be met for every species because of "noise" in the data.

Based on the trials' own pilot observations, the report claims that levels of some key indicator organisms, including beetles and broad-leaved weeds, are likely to vary from field to field by far more than the 50 per cent margin that the trial allows for. If so, that could make detecting a 50 per cent difference between GM and non-GM fields impossible even if the difference is there.

Pure speculation

The report also takes issue with the 1.5-fold target difference itself, arguing that it is set too high. Previous research on the impact of herbicides on grey partridges found that much smaller differences in weed numbers – as slight as 13 per cent – were ecologically significant.

Les Firbank, who has coordinated the trials from the Institute of Terrestrial Ecology in Cumbria in north-west England, rejects the criticisms. "They're speculating on whether the experiment has been capable of delivering the stated power. That won't be answered until the data are published," he says.

Firbank says that the sensitivity target was never intended to be met for every species. "The interpretation comes not from looking at each species in isolation but from combining results from different species and looking for patterns."

Peter Green, president of Britain's Royal Statistical Society, says that while many of the report's points about statistical power are valid, such problems are not unique to these trials and there are well-established ways of handling them. "The danger with an issue that is so highly charged politically," he says, "is that some people will seek black and white answers when they are not attainable."

***

NEW ANALYSIS CASTS DOUBT ON GM FARM SCALE EVALUATIONS

 Mar 26

 http://www.foe.co.uk/resource/press_releases/new_analysis_casts_doubt_o0.html
 
 A new analysis by Friends of the Earth [1], published today and highlighted in this week's New Scientist magazine [2], suggests that the science and conduct of the Government-sponsored GM farm scale evaluations (FSE) will fail to provide any conclusive evidence on whether GM crops will do long-term harm to farmland wildlife.

 Friends of the Earth based their analysis on materials already published, including the tender documents, minutes and interim reports by the research consortium [3] who carried out the field work, and analysis of the results and the Scientific Steering Committee [4]. The FSE results are due to be published in the autumn.

 The main findings of Friends of the Earth's report are that [5]:

 Ecologically significant differences between GM and non GM crops may be missed because the experiment does not have sufficient statistical power.

 The scope of research was seriously limited by time and resource constraints.

 It may be impossible to detect any meaningful differences for some important indicator species.

 Monitoring of important soil organisms was dropped because of money and time constraints.

 Rare arable plants were excluded from the study because of time constraints.

 Modelling based on the results will be hampered by a lack of knowledge about interactions between different species, which food sources are preferred by which birds and mammals

 Poor geographical distribution of the trials undermines the relevance of the results (eg 45% of maize is grown in the SW region but only 8% of trials took place there).

 Advice on the use of weed killer on the GM crops was given by the companies who developed the technology, leading to concerns that the GM crops may have been managed to maximise biodiversity whilst ignoring the final yield.

Evidence that in the United States additional herbicides are used to achieve the required level of weed control in maize crops has been overlooked, meaning the maize results could be irrelevant.

 Friends of the Earth Real Food and Farming Campaigner Pete Riley said:

 "We have published this report because we think it is vital that the public, farmers and the Government realise the limitations of the Farm Scale Evaluation results. The se studies, due out in the autumn, are incapable of providing adequate evidence that GM crops have no impact on wildlife . T h is is not the fault of the researchers – their hands were tied .

 " The Government was not interested in properly investigat ing the long term impacts of GM crops , i t wanted to avoid the threat of a moratorium . But they cannot expect the British public to accept that the future commercialisation of GM crops poses no threat to wildlife without the hard evidence."

 ENDS

 Notes

 [1] Science as a Smokescreen, written by Emily Diamand, published by Friends of the Earth, March 2003. Full copies of the report are available at
 www.foe.co.uk/resource/reports/science_smokescreen.pdf[1] from 2pm, Wednesday, and from the press office at Friends of the Earth.

 [2] See New Scientist 29th March 2003, Vol. 177 No. 2388

 [3] The research contractors were the Institute of Hydrology and Ecology (formerly the Institute of Terrestrial Ecology), the Institute of Arable Crop Research and the Scottish Crop Research Institute.

 [4] The Scientific Steering Group chaired by Professor Chris Pollack was established before the contractors were appointed and set the parameters for the design of the research in the FSE.

 [5] A media briefing is available below/ from the press office at Friends of the Earth.

.. [1] http://www.foe.co.uk/resource/reports/science_smokescreen.pdf