Shaky meta-analysis of GM benefits

January 2015
Corn harvest. CC photo by United Soybean Board on Flickr
A heavily-promoted 'meta-analysis' of the performance of GM crops world-wide concluded it "revealed robust evidence of GM crop benefits for farmers in developed and developing countries" which "may help to gradually increase public trust in this technology".
Meta-analysis is a statistical technique which examines the combined effect of a number of studies on the same topic, to give an overall picture of what the total data might, or might not, be showing. The more studies included, and the more closely comparable the data sets, the more robust the results.
What the meta-analysis found was "On average, GM technology adoption has reduced chemical pesticide use by 37%, increased crop yields by 22%, and increased farmer profits by 68%. Yield gains and pesticide reductions are larger for insect-resistant crops than for herbicide-tolerant crops. Yield and profit gains are higher in developing countries than in developed countries.
German agri-economists, Klümpfer and Qaim, based their meta-analysis on 147 publications. Studies were selected for inclusion on the bases that they presented original data on impacts of the three most widely adopted GM crops (soya, maize and cotton) with the two most common traits, herbicide-tolerance and insect-resistance. The 'impacts' included effects on crop yield, applied pesticide quantity and cost, total production cost, and farmer profit. The authors also used the statistics to asses the impacts of factors which might bias the outcome of the analysis, such as geographic location, GM trait, field conditions, source of funding, and type of publication. They were satisfied that no major biases were at play.
However, doubts have been raised about how "robust" Klümpfer and Qaim's pro-GM evidence is.
For example, because any one of the 'impacts' qualified a study for inclusion, plus the fact that select data from one study might appear in more than one publication, the actual number of studies factored into the statistics was over-stated, inflating the power of the meta-analysis.
Many of the studies were farmer surveys lacking actual measurements and without any non-GM control crop. These data are shaky.
Despite the fact that the main and longest-grown GM crop is herbicide-tolerant soya for human consumption, nearly 80% of the included studies involved 'Bt' insecticidal GM crops, and over 50% involved cotton. 
(COMMENT It's not clear how stacked traits for Bt or herbicide-tolerance in the same crops were factored in, nor whether applied pesticides referred to all such chemicals or only those relevant to the GM trait.)
Because both the main commercialised GM traits confer pest-control, they both encourage the evolution of pest-resistance. This means that success in the early years tends to reduce with time. Most of the studies were based on data from before the rise of insect-resistance and superweeds, and involved a single growing season. This introduces a major bias into the meta-analysis.
It seems that at least two studies which were eligible for inclusion, but perhaps tellingly failed to identify GM crop benefits, were omitted. One of these was a four-year record of yield and profitability of GM cotton in one state in America (Jost). The other was a study of yields achieved over the last 50 years in comparable US (GM) and EU (non-GM) crops (see AGRICULTURAL REALITY SHOWS NO YIELD BENEFIT FOR GM - January 2014).
The idea of increasing trust in GM through information "agronomic and economic" benefits to farmers is certainly misplaced. Public distrust of GM stems from concerns about harm to health and to the environment from GM, and also from the corporate control over our food supply and agriculture which GM clearly facilitates. The authors' conclusion that their data might generate some trust in products of the technology suggests a disconnection from the GM debate and an attempt at pro-GM propaganda.
Is this study gradually increasing your trust or your distrust?

  • Wilhelm Klümpfer and Matin Qaim, 2014, A Meta-Analysis of the Impacts of Genetically Modified Crops, PLOS ONE, 9:11, November 2014
  • Jack Heinemann, Correlation is not causation, Right Biotech , 27.11.14
  • P. Jost, et al., 2008, Economic comparison of transgenic and nontransgenic cotton production systems in Georgia, Agronomy Journal 100

No comments:

Post a comment

Thanks for your comment. All comments are moderated before they are published.