Ideally ascertainment of participants in population based

Ideally, ascertainment of participants in population-based studies should be free of selection biases, thus creating conditions for generating data representative of the general population – “a representative brain” (Falk et al., 2013). As we pointed out elsewhere, previous imaging cohorts used different recruitment strategies (samples of convenience vs. census-based sampling) and exclusion criteria (MR contraindications only vs. screening out children with any personal and family-based risk factors); not surprisingly, some of these strategies yielded “supernormal” samples (Paus, 2010). In the Saguenay Youth Study, we have carried out recruitment in all public high schools in the region and excluded from participation only adolescents with MRI contraindications and serious conditions likely to affect the brain (e.g., epilepsy) or heart (e.g., heart defects) development. By design, the sample is enriched by individuals born to mothers smoking cigarettes during pregnancy (50% vs. ∼20% expected in general population). Other – more subtle – biases include the requirement of having siblings and being able to contact both biological parents. The latter conditions are, however, unlikely to reduce representativeness of the sample, as the mean number of children per family in the general population is 1.5 (Quebec, 2014), and the two-parent requirement did not demand cohabitation. Recent replications of the relationship between externalizing behavior and substance use during adolescence in two geographically and culturally distinct samples recruited using very different strategies, namely the Saguenay Youth Study and the Northern Finland Birth Cohort 1986, suggest that findings obtained in our sample are generalizable (Lotfipour et al., 2014). Brain imaging represents a unique tool allowing one to obtain a wide array of quantitative phenotypes (Table 8). A number of considerations are at play when choosing specific MR sequences. For example, studies of typically developing children and adolescents are more likely to include scans sensitive to changes in myelination (e.g., magnetization transfer ratio or myelin water fraction, Dean et al., 2014) rather than those sensitive to white-matter hyperintensities, which are more common in the aging brain (e.g., T2 fluid attenuated inversion recovery [FLAIR]). But perhaps the most relevant considerations relate to the trait (vs. state) qualities of a given measure; after all, we SBHA supplier manufacturer most of our developmental work on the assumption that genes and early environments shape brain function and structure in a stable and long-term manner. Therefore, test–retest reliability of imaging-derived measures is paramount. Not surprisingly, various metrics derived from (multi-modal) structural images show high test–retest reliability. For example, Wonderlick and colleagues (Wonderlick et al., 2009) have evaluated test–retest reliability (two sessions, 2 weeks apart), and the influence of several acquisition parameters (same 3T scanner), for a number of morphometric measures derived from T1-weighted images by FreeSurfer. They found that the reliability – estimated with intra-class correlation coefficients (ICCs) – was “excellent” for most measures; with the exception the globus pallidus, all ICCs values were above 0.95. The test-retest reliability of DTI-based measures appears to vary across the measures and fiber tracts. For example, Wang et al. (2012) found excellent reliability (ICCs > 0.75) for the mean length of the corpus callosum and the uncinate fasciculus, and fair reliabilities (ICCs between 0.4 and 0.75) for fractional anisotropy in most fiber tracts. On the other hand, test–retest reliability of data obtained with fMRI has been characterized as “fair” (ICC: 0.4–0.75) in adults and adolescents, and “poor” (ICC < 0.4) in children; it is lower in regions with weak “activation”, as revealed by group t-maps (Caceres et al., 2009; Koolschijn et al., 2011; Plichta et al., 2012). The relatively low test–retest reliability of functional data is likely related to a number of factors, including the fact that the fMRI signal is an indirect measure of brain activity, its measurement is affected by a number of noise-generating factors (e.g., head motion, physiological “noise” related to respiration and cardiac cycle) and, most importantly, by the state of the participant during scanning. The latter factors, such as inter-individual and session-by-session variations in task-related behavior (performance, attention) and general state of arousal (anxiety, sleepiness) are very difficult to assess and control, thereby adding significant error to the measurement of the functional phenotype. For the above reasons, we advocate imaging protocols that put emphasis on multi-modal imaging of brain structure. We consider brain structure “a window into the individual\'s life history”, a notion supported by the wealth of imaging data on experience-related brain plasticity (Lovden et al., 2013).