One problem with meta-analyses, which may not be confined to the field of neuroimaging (Ioannidis, 2005), is the potential overrepresentation of positive findings in the published literature. A recent evaluation of meta-analyses of regional brain volume changes

in psychiatric disorders reported evidence for a considerable overreporting of significant group differences (except for the cerebral ventricles) (Ioannidis, 2011). Potential reasons include publication bias, selective reporting of brain regions showing group differences, and other arbitrary decisions that can be summarily termed “researcher degrees of freedom” (Simmons et al., 2011). However, the evaluation by Ioannidis (2011) did not include whole-brain volumetric studies that implement whole-brain correction for multiple comparisons, which this website should be less vulnerable to the selective reporting bias. Suggested improvements included the increase of power through multicenter studies, preregistration of clinical imaging studies, and definition of standardized analysis protocols (Ioannidis, 2011). Clinical

phenotypes in mental disorders may just be the endpoints of multiple converging pathophysiological pathways that are triggered by different combinations of genetic predisposition and environmental stress (Figure 2). As such, one way of improving the consistency of imaging findings in psychiatry may SRT1720 mw be to probe the biological pathways implicated these in specific mental disorders.

Current genetic models posit that multiple common variants with small effects or rare variants with larger effects confer the genetic vulnerability to psychotic disorders (Owen et al., 2010). Studies on patient samples that are not further differentiated by genotype may therefore obscure specific biological effects, whereas it may be possible to elucidate pathways to the disorder (Meyer-Lindenberg, 2010) through the effects of the risk variants on parameters of structural or functional neuroimaging, radioligand binding, or noninvasive neurophysiology. A particularly attractive aspect of this approach is that, in principle, it allows targeting any protein for which a functionally relevant genetic variant exists and would thus greatly expand the list of molecular mechanisms that can be investigated with neuroimaging (Table 2). It may also help overcome the lack of cellular resolution of current non-invasive neuroimaging techniques. Although high-resolution MRI at 7 Tesla can resolve the laminar structure of human cortex (Sánchez-Panchuelo et al., 2011), each layer contains a multitude of functionally and structurally diverse neurons that cannot be differentiated.