Maize (Zea mays L.) has been grown on over 10 million ha in Europe in recent years (Czarnak and Rodríguez-Cerezo 2010). Among the insects affecting maize yield, the Lepidopteran borers [Ostrinia nubilalis (Hübner) (Lepidoptera: Crambidae) and Sesamia nonagrioides Lefèbvre (Lepidoptera: Noctuidae)] are the main pests (Meissle et al. 2010). In addition to biological control and cultural practices, genetically modified (GM) varieties with the insecticidal properties of the entomopathogenic bacterium Bacillus thuringiensis Berliner (Bt maize) may provide efficient control of maize borers and reduce chemical applications (Meissle et al. 2010). A survey conducted among growers in Spain estimated that Bt maize is economically very efficient in areas with a high pest pressure and may replace most of the insecticidal sprayings carried out to control these pests (Gómez-Barbero et al. 2008). Spain is the main grower of Bt maize in Europe; from its authorization for cultivation in 1998–2012 the area of Bt maize has increased to 116,030 ha, representing 30 % of the total area under maize in the country (MAGRAMA 2012). In areas where maize borers are particularly damaging, such as the study area in Aragon and Catalonia (NE Iberian Peninsula), the concentration of Bt maize may reach 70 % of the maize grown for grain.

Bt maize has proven to be effective for controlling borers and has a potential role in containing expansion of Diabrotica virgifera virgifera LeConte (Coleoptera: Chrsyomelidae), an important pest in the USA that was introduced into Europe in the early 1990s. Furthermore, Bt maize has environmental benefits because it reduces the need for chemical applications. However its deployment has prompted extensive debate over risks for the environment and particularly for nontarget organisms, including arthropods. Arthropods provide important ecological services in maize ecosystems such as biological control, pollination and decomposition, and they form an important part of the biodiversity, so direct or indirect effects of Bt maize on arthropods may interfere with these services. The mandate to conduct laboratory and field trials to assess risks of GM crops for the environment and particularly for nontarget organisms prior to their authorization has been in place in the regulation of the EU and member States for several years (EFSA 2010). In response to this mandate, governmental bodies and companies have promoted and sponsored many European Research Area (ERA) activities in Europe to measure potential effects of GM maize on nontarget arthropods (NTAs) in both the laboratory and the field. A tiered approach to testing potential side effects of GM crops on NTAs, from the laboratory to the field, was proposed by Romeis et al. (2008). Usually, it is necessary to conduct field trials or use models when a GM crop has proven harmful in the laboratory or when the main potential effects are the result of complex interactions among many factors that cannot be studied in simple laboratory conditions. However, this field testing is difficult to interpret due to the interaction of many factors that cannot be studied in simple laboratory conditions.

In the last 12 years field trials for ERA purposes have been conducted in several European countries, including Spain, where the authors have assessed risks of GM Bt and herbicide-tolerant (HT) maize varieties for NTAs (Comas et al. 2013). Only occasional effects of Bt maize varieties on herbivore, predatory or parasitoid insects have been reported (Lumbierres et al. 2004, 2011; Pons et al. 2005; de la Poza et al. 2005; Albajes et al. 2012), and many authors have concluded that the involvement of the Bt traits in these occasional effects has not been proven (see the review of Naranjo 2009). However, the lack of negative effects could be due to the insufficient statistical power of the tests. The power of a test is the probability of rejecting the null hypothesis, no effects, when it is false and the alternative hypothesis is true. Then it measures the probability that the test detects an effect of a known magnitude using a specified experimental design and varies according to the magnitude of the effect specified (Perry et al. 2003). Inversely, the magnitude of an effect (the effect size) that a test is able to detect may be calculated for a specified power. The effect size is usually obtained as the scaled difference between the density of the organism recorded on the GM variety and the density of the organism on a control non-GM variety called the comparator. This value can be expressed as a percentage of the control mean density. Once the power of the test has been specified, we can obtain the minimum difference between the density of the organism on the control non-GM and the GM variety, given as a percentage of the control mean density, that the test is able to detect. This minimum difference is the detectability of the test. A field test with a high detection capability, for a given test power, is a test that is able to detect small significant difference between the control non-GM and the GM variety. Therefore, an improvement in the detectability of a test implies better detection of small significant effects.

Increasing the statistical power of individual field tests to satisfactory values would involve increasing sample size by increasing the number of replications, treatments or years/sites of trials, a rather costly approach. Alternatively, if several trials are available, a meta-analysis may improve statistical power by combining them and assuming a common measure of effect size. In fact, this approach can be used to integrate several independent trials, whether published or unpublished, that were not initially defined to be combined and thus obtain new and more robust results (Borenstein et al. 2009). This approach was used by Marvier et al. (2007), Wolfenbarger et al. (2008) and Naranjo (2009) to study the nontarget effects of Bt crops reported in 42, 45 and 63 field studies, respectively.

This study aimed to determine whether the no-effect conclusions reached by ANOVA analysis of single field trials with Bt maize varieties conducted in Spain from 2000 to 2010 is confirmed by a meta-analysis of all these trials. Complimentarily, the effect detection capacity of taxa recorded in field trials is calculated with a meta-analysis approach and compared with values of single trial analysis. To this end, among the 20 field trials carried out by authors in the period to measure nontarget effects of GM maize (Bt, HT and stacked traits), we selected 13 trials in which Bt and near-isogenic non-Bt varieties were compared for population density of several taxa, including arthropod herbivores, predators, omnivores, parasitoids and decomposers.