The association of the BDNF Val66Met polymorphism and the hippocampal volumes in healthy humans: A joint meta-analysis of published and new data

BACKGROUND
The brain-derived neurotrophic factor (BDNF) Val66Met polymorphism (refSNP Cluster Report: rs6265) is a common and functionally relevant single nucleotide polymorphism (SNP). The gene itself, as well as the SNP rs6265, have been implicated in hippocampal learning and memory. However, imaging genetic studies have produced controversial results about the impact of this SNP on hippocampal volumes in healthy subjects.


METHODS
We examined the association between the rs6265 polymorphism and hippocampal volume in 643 healthy young subjects using automatic segmentation and subsequently included these data in a meta-analysis based on published studies with 5298 healthy subjects in total.


RESULTS
We found no significant association between SNP rs6265 and hippocampal volumes in our sample (g=0.05, p=0.58). The meta-analysis revealed a small, albeit significant difference in hippocampal volumes between genotype groups, such that Met-carriers had slightly smaller hippocampal volumes than Val/Val homozygotes (g=0.09, p=0.04), an association that was only evident when manual (g=0.22, p=0.01) but not automatic tracing approaches (g=0.04, p=0.38) were used. Studies using manual tracing showed evidence for publication bias and a significant decrease in effect size over the years with increasing sample sizes.


CONCLUSIONS
This study does not support the association between SNP rs6265 and hippocampal volume in healthy individuals. The weakly significant effect observed in the meta-analysis is mainly driven by studies with small sample sizes. In contrast, our original data and the meta-analysis of automatically segmented hippocampal volumes, which was based on studies with large samples sizes, revealed no significant genotype effect. Thus, meta-analyses of the association between rs6265 and hippocampal volumes should consider possible biases related to measuring technique and sample size.


Introduction
Brain-derived neurotrophic factor (BDNF) -a member of the nerve growth factor family -plays an important role in neurogenesis and is implicated in several molecular processes in the central nervous system (Barde et al., 1982;Lu and Gottschalk, 2000;Park and Poo, 2013). BDNF is highly expressed in the hippocampus, a key region for adult neurogenesis (De Quervain and Papassotiropoulos, 2006;Milner et al., 1998), and is thought to be involved in learning and memory (Cunha et al., 2010). Pro-BDNF can induce apoptosis, while mature BDNF predominantly mediates cell survival and neuronal differentiation (Pang et al., 2004;Korte et al., 1995;Pastalkova et al., 2006). The single nucleotide polymorphism (SNP) rs6265 at codon 66 of the BDNF gene predicts a valine (Val) to methionine (Met) substitution in the pro-region of the protein, which is important for proper BDNF sorting. The Val66Met substitution has been investigated in a transgenic mouse model of defective BDNF secretion in hippocampal neurons (Chen et al., 2004;Egan et al., 2003). BDNF Met/Met mice have smaller hippocampal volumes, less dendritic arbor complexity of hippocampal neurons and impaired synaptic plasticity, as indicated by a decrease in NMDA-receptor-dependent long-term depression and long-term potentiation (Chen et al., 2006;Ninan et al., 2010).
Defects in synaptic plasticity and long-term potentiation, core mechanisms of hippocampus-dependent learning and memory, are thought to underlie -at least in part -neurocognitive impairments in a broad spectrum of neuropsychiatric disorders (Fusar-Poli et al., 2012;Lu et al., 2013). Another characteristic of neuropsychiatric disorders, such as schizophrenia, bipolar disorder, depression, post-traumatic stress disorders and personality disorders, is the reduction in hippocampal volume (Geuze et al., 2005;Smieskova et al., 2010;Walter et al., 2012). It is still not clear to what extent these hippocampal volume abnormalities are driven by genetic liability (Sullivan et al., 2003). One putative genetic risk factor of these alterations might be the BDNF polymorphism described above (Boulle et al., 2012;Frielingsdorf et al., 2010). The effect of this polymorphism has often been studied in healthy subjects, because in a healthy population, changes in brain volumes are independent of effects of illness or medication, and of disease-related genetic risk factors (Fusar-Poli et al., 2013;Smieskova et al., 2009).
To date findings from structural magnetic resonance imaging (sMRI) studies investigating genotype-dependent association of rs6265 SNP on hippocampal volumes are inconsistent. While three recent meta-analyses report that Met-carriers have smaller hippocampal volumes than Val/Val homozygotes (Hajek et al., 2012;Kambeitz et al., 2012;Molendijk et al., 2012a), the relation between rs6265 and hippocampal volumes is confounded by several problems: Firstly, two of these studies (Kambeitz et al., 2012;Molendijk et al., 2012a) included a variety of neurocognitive disorders, suggesting that hippocampal volumes were probably affected by burden of illness, medication or comorbid conditions and were not necessarily related to the SNP per se. Secondly, all of these meta-analyses incorporated studies with children/adolescents and elderly subjects. This can be critical, as hippocampal volumes undergo age-related changes (Karnik et al., 2010;Walhovd et al., 2011;Goodro et al., 2012). Finally, although one of the previous meta-analyses focuses exclusively on healthy subjects (Hajek et al., 2012), the analysis in this study was restricted to manual tracing of hippocampal volumes without considering automatic measurement techniques.
The present study aimed to control for these confounding factors. First, we assessed the association between the BDNF rs6265 polymorphism and hippocampal volumes using the automated tracing technique in 643 healthy young volunteers. Because the effect size of this association is known to be small (Kambeitz et al., 2012;Molendijk et al., 2012a), we then increased statistical power by means of meta-analytic techniques (Kim-Cohen et al., 2006;Munafò et al., 2009;Brandys et al., 2011). We therefore performed a systematic review of the hippocampal volumes in healthy subjects genotyped for SNP rs6265 and combined these data with our original results in a meta-analysis. Additionally, we examined the effect of potential moderators such as measuring technique, MR magnetic field strength, age, gender, ethnicity, Val/Met ratio, sample size, quality rating, hippocampal volumes normalized to intracranial volume (ICV), and publication year.

Material and methods
2.1. Original data of 643 healthy subjects 2.1.1. Participants We recruited 643 healthy young subjects (383 women; age range 18-35 years, mean age ± standard deviation (SD) 22.87 ± 3.22). Participants filled in a self-rating questionnaire concerning their health status, medication, and drug consumption. All included subjects were free of any physical, neurological or psychiatric illness, and were taking no medication. 87% of the subjects were students and 91% were right-handed (see Table 1). The ethics committee of the Canton of Basel approved the experiments. Written informed consent was obtained from all subjects prior to participation.

Genotyping
DNA was extracted from saliva samples collected with the Oragene DNA sample collection kit using standard procedures (DNA Genotek Inc., Ontario, Canada). DNA samples were processed on the Affymetrix ® Genome-Wide Human SNP Array 6.0. in one centralized microarray facility. rs6265 (refSNP Cluster Report: rs6265) is represented on the array (AFFY|SNP A-2038925). Generation of SNP calls and array quality control were performed using the Affymetrix Genotyping Console Software 3.0 (Affymetrix Inc.). According to the manufacturer's recommendation, contrast quality control (QC) was chosen as QC metric, using the default value of 0.4. All samples passing QC criteria were subsequently genotyped using the Birdseed (v2) algorithm. Genotypic outliers were identified using Bayesian clustering algorithm (Bellenguez et al., 2012) and excluded (for more details see supplementary material).

Image acquisition and extraction of hippocampal volumes
We acquired an anatomical sequence with a radio-frequency pulses and rapid gradient-echo (MPRAGE) sequence. For this sequence, we used the following acquisition parameters: TE (echo time) = 3.37 ms, FOV (field of view) = 25.6 cm, acquisition matrix = 256 × 256 × 176, voxel size: 1 mm × 1 mm × 1 mm. Using a midsaggital scout image, 176 contiguous axial slices were placed along the anterior-posterior commissure (AC-PC) plane covering the entire brain with a TR = 2000 ms (Â = 8 degrees).
Segmentations of cortical and subcortical structures were retrieved from FreeSurfer 4.5 and labeling was based on the Desikan-Killiany Atlas (Desikan et al., 2006). We extracted raw volumes for both hippocampi for n = 805 subjects. Left and right hippocampal volumes were corrected separately for ICV, age, sex and differences due to software and gradient updates by using the z-transformed residuals of a linear regression. Afterwards we did a separate outlier-control for both hippocampal sides (mean ± 3.5 SD). For all subjects with complete dataset, we then calculated the corrected mean value of both hippocampal volumes. For a subgroup of n = 643 subjects we had additional genetic information regarding BDNF genotype. The corrected volumetric data of these subjects were included in all further analyses.

Association analysis
For the genetic association analysis, we used the WG-Permer software (www.wg-permer.org), with analysis of variance for quantitative phenotypes. This software corrects nominal p-values for multiple testing on a permutation-based procedure according to Westfall and Young (Westfall, 1993).
One-way analyses of variance (ANOVA) and chi-square tests were used to test for differences between genotype groups of age, sex, profession and handedness. These statistical analyses were performed with SPSS (IBM SPSS Statistics, Version 20, 2011). Values are presented as mean ± SD (see Table 1).

Literature search and inclusion criteria
Electronic searches were conducted using PubMed and Embase, considering all publications until the end of December 2012 with the following search terms: "BDNF Val66Met" AND "MRI" and "rs6265" AND "MRI". Additionally, a retrospective search was carried out on the reference lists of the included articles. This resulted in 86 publications, for which the abstracts were screened (more information is presented in Fig. 1). In this meta-analysis, we included healthy groups only. Firstly, we extracted studies addressing the relation between hippocampal volumes and the SNP. Secondly, the papers were filtered according to the following criteria: (a) published in a peer-reviewed journal, (b) reporting a relation between the SNP rs6265 and sMRI, (c) showing hippocampal data. A total of 27 publications met these criteria, whereof from one recent genome-wide association study (GWAS) data of 5 cohorts were obtained (Stein et al., 2012). Altogether a total of 32 samples, 31 previously published and our own data, were included in this meta-analysis. Criteria for exclusion were: mean age of participants (<15 or >65 years), not clearly defined healthy control group, overlapping datasets, and only left or only right hippocampal volume reported. The authors were contacted when information essential for the calculation of effect sizes was missing. Both measuring techniques, i.e. automated and manual tracing, were included. We followed the 'Preferred Reporting Items for Systematic Reviews and Meta-Analyses' (PRISMA) guidelines (Moher et al., 2010).

Data extraction
The following variables were extracted: First author name, publication year, number of independent samples per study. For each independent sample, we extracted sample size of genotype subgroups, ethnicity, gender, mean age of sample, Hardy-Weinberg equilibrium (HWE; calculated, when not reported), genotyping method, structural MRI measurement technique, direction of effect, field strength of MR scanner, mean hippocampal volumes and standard deviation, t-statistic, F-statistic and p-values per genotype, and whether the hippocampal volumes were normalized to ICV. To sustain statistical independence, one single effect size per sample was used for this meta-analysis.

Quality assessment
Using an 11-item checklist adapted from (Karg et al., 2011) the quality of the included studies was evaluated. In detail, the criteria were: (1) Funding -role in analysis and interpretation of data, (2) Sample size, (3) Clear inclusion criteria for participants, (4) Reported allele distribution, (5) Ethnicity assessed, (6) If mixed ethnicity: discussion of problems, (7) IQ/educational level available, (8) Inter-and intrarater reliability, (9) Report of HWE, (10) Sample in HWE and (11) Additional descriptive data including age, gender, genotyping method, magnetic field strength of scanner. For each category 0, 1 or 2 points were given. Finally, the included studies were rated according to the sum of the points and characterized as high (above 80% of the maximal sum of points), moderately high (60-79%), moderate (40-59%), moderately low (20-39%), and low quality studies (below 19%) (for more details see supplementary tables S1 and S2).

Data analysis
Data were entered into an electronic database and quantitative meta-analysis was performed using the R 2.15.2 software (R Core Team, 2012). The effect size was calculated using Hedge's g, which provides an unbiased standardized mean difference that incorporates a correction for small sample sizes (Lipsey and Wilson, 2000). Hedge's g values above 0.2, 0.5 and 0.8 correspond to small, medium and large effect sizes respectively. Hedge's g was calculated using data of mean hippocampal volumes, standard deviations and sample sizes. Where these data were not available, we employed the t-statistic, F-statistic or p-values, together with the corresponding sample sizes. A positive value of the effect size reflected larger hippocampal volumes in the Val/Val homozygotes than for the Metcarriers of the SNP rs6265. We employed a random-effects model with the DerSimonian-Laird estimator using the metafor package (DerSimonian and Laird, 1986;Wolfgang Viechtbauer, 2010). The random-effects model shows more flexibility with respect to effect size variability between studies and study populations , as it incorporates the between-study variance 2 . And in case of high between-study heterogeneity, the random-effects model compared to the fixed-effects model is the model of choice (Ioannidis et al., 2007).
Cochran's Q test was then used to calculate between-group heterogeneity; the magnitude of heterogeneity was assessed by I 2 (Higgins and Thompson, 2002). I 2 is an estimate of variability across studies based on heterogeneity rather than chance, ranging from 0 to 100%. Values above 25%, 50% and 75% corresponded to low, moderate and high heterogeneity respectively (Higgins and Thompson, 2002). Furthermore, potential publication bias was investigated by funnel plot asymmetry and Egger's regression test (Egger et al., 1997). In case of a bias, "the trim and fill" method was used subsequently to identify and correct for publication bias detected by an asymmetric funnel plot (Duval and Tweedie, 2000). A series of meta-regression analyses was carried out to assess the impact of possibly moderating study design characteristics such as publication year, age of participants, gender ratio, ethnicity, Val/Met ratio, sample size, quality rating, magnetic field strength, hippocampal volumes normalized to intracranial volume and applied hippocampal measuring techniques. Most studies used a dominant allele approach, but two studies reported an additive allele comparison (Agartz et al., 2006;Gruber et al., 2012). Nevertheless, these were treated equivalently in this analysis.

Association analysis of 643 healthy subjects
Of the 643 subjects, 413 were homozygous for the Val allele, 204 were heterozygous Val/Met, and 26 were homozygous for Met allele. Met-carriers were taken together in a single group. Genotype groups did not differ according to age, sex, profession and handedness (see Table 1). All 643 subjects had complete genotype information. The genotype distribution did not deviate from the Hardy-Weinberg equilibrium (p = 0.90).
As shown in Fig. 2, there were no significant genotypedependent differences in the z-transformed scores of the left (Val/Val homozygous 0.029 ± 0.97 (n = 413), Met-carriers 0.001 ± 0.98 (n = 230); p = 0.25, see The difference between genotypes in mean hippocampal volumes resulted in a non-significant g of 0.05 (p = 0.58). We did not observe a main effect of age or sex as well as no interaction effect of sex and rs6265 genotype groups on hippocampal volumes (see supplementary methods and supplementary table  S3).

Meta-analysis of one original and 31 previously published samples
Meta-analysis of all datasets (k = 32) showed evidence for significant, albeit weak association between hippocampal volumes and SNP rs6265 (g = 0.09, se = 0.04, 95% CI = [0.01-0.17], Z = 2.08, p = 0.0376, see Fig. 3A and table S4), with indications of significant between-study heterogeneity (I 2 = 38.24%, Q(df = 31) = 50.20, p = 0.02). The effect was in the direction of slightly smaller hippocampal volumes for Met-carriers than for Val/Val homozygotes. Visual inspection of the funnel plot indicated evidence for potential publication bias (Fig. 3B, table S4). This was quantitatively confirmed by significant regression intercept in Egger's regression test (p = 0.0075). The trim and fill procedure suggested 8 missing studies on the left side of the funnel plot and a cor-

Effect of moderators
To further disentangle the moderating effect of the measurement technique, samples were subsequently subdivided into manually and automatically segmented volumes of the hippocampi. One study using semi-automated analysis was excluded from further analysis (Sanchez et al., 2011), leaving 13 samples with manual tracing (n = 829 subjects) and 18 samples using automated segmentation (n = 4426 subjects). The detected small effect size estimate of manual tracing samples indicated significantly smaller hippocampal volumes for Met-carriers compared to Val/Val subjects (g = 0.22, se = 0.09, 95% CI = [0.05-0.39], Z = 2.51, p = 0.0121, I 2 = 38.12%, Q(df = 12) = 19.39, p = 0.08, Trim and fill: 5 missing studies on left side of the funnel plot and a corrected non-significant g of 0.08, see Fig. 4A and table S4). The meta-analysis of the manual tracing samples revealed significant publication bias (Egger's test: z = 3.24, p = 0.0012), significant between-study heterogeneity and a significant moderator effect only for the sample size (ˇ= −0.72, F(1,11) = 12.07, p = 0.01). Analysis of the relation between years of publication and effect size revealed a significant decrease in the effect sizes with increasing sample size over the years, but only for manual tracing samples (see Fig. 5). In contrast, the overall effect size of the samples using automatic measurement techniques showed no significant genotype effect (g = 0.04, se = 0.05, 95% CI = [−0.05-0.13], Z = 0.89, p = 0.3751, I 2 = 37.87%, Q(df = 17) = 27.36, p = 0.05, see Fig. 4B and table S4).

Discussion
In this paper, we present a joint analysis of the relation between the BDNF SNP rs6265 and the hippocampal volumes in healthy young subjects. Specifically, we first explored whether hippocampal volumes of 643 healthy individuals differed between Val/Val homozygotes and Met-carriers. These data were further incorporated into a meta-analysis of previously published studies subsuming a total of 5298 healthy subjects.
Several studies report BDNF-dependent volume differences in the hippocampus of patients with neuropsychiatric disorders such as bipolar disorder and schizophrenia (Chepenik et al., 2009;Szeszko et al., 2005) as well as between healthy controls and patients of the same genotype (Chepenik et al., 2009;Gonul et al., 2011;Koolschijn et al., 2010;Smith et al., 2012). Other studies in patient populations found no association of the rs6265 polymorphism and hippocampal volumes (Agartz et al., 2006;Cerasa et al., 2010;Cole et al., 2011;Dutt et al., 2009;Frodl et al., 2007;Gruber et al., 2012;Jessen et al., 2009;Molendijk et al., 2012b;Takahashi et al., 2008). Two recent meta-analyses did not find a significant association of SNP rs6265 and hippocampal structure in neuropsychiatric disorders, including schizophrenia, bipolar disorder, depressive and anxiety disorders (Kambeitz et al., 2012;Molendijk et al., 2012a). However, the meta-analyses were not conducted separately per psychiatric disease category and treatment  effects may have influenced the hippocampal volumes (Fusar-Poli et al., 2013). Inconsistent findings in studies of healthy subjects and psychiatric patients raise the question if BDNF-dependent structural hippocampal differences are specific for different developmental stages. Until now, only few studies have addressed this issue by investigating the relationship between BDNF and hippocampal volumes in neonates, children and adolescents and also elderly. Two studies have not observed BDNF-dependent differences in hippocampal volumes in children and adolescents (age range 8-19) (Mueller et al., 2013;Toro et al., 2009). In contrast, Knickmeyer and colleagues find rs6265-dependent differences in hippocampal volumes in neonates (Knickmeyer et al., 2013). However, in order to investigate the influence of developmental stages on BDNF-dependent effects, additional longitudinal studies will be necessary. For instance, Knickmeyer and colleagues will implement a follow-up design, collecting data over several time points (at age 1, 2, 4 and 6 years of age) (Knickmeyer et al., 2013). Moreover, several studies report hippocampal volume reductions in aging (Driscoll et al., 2003;Erickson et al., 2010;Malykhin et al., 2008;Raz et al., 2010). Erikson and colleagues investigated the relationship between serum BDNF levels, age, hippocampal volume and memory performance (Erickson et al., 2010). Age was associated with reduced hippocampal volumes as well as reduced BDNF serum levels and poorer memory performance. In his review, Von Bohlen und Halbach suggests a role of BDNF in age-dependent processes in the hippocampus (Von Bohlen und Halbach, 2010). However, studies investigating the association of rs6265 with hippocampal volumes in also aged populations report inconsistent results (Brooks et al., 2014;Karnik et al., 2010;Sanchez et al., 2011).
The importance of the hippocampus in learning and memory is well established (Squire and Wixted, 2011) and it has been suggested that BDNF plays a role in these processes (Baj et al., 2013;Cunha et al., 2010). Even though we did not find BDNF-dependent differences in hippocampal volumes, the absence of difference on the anatomical level does not rule out that BDNF modulates other processes in the hippocampus. Indeed, two studies included in this meta-analysis provide support for BDNF-dependent differences in hippocampal activation during memory paradigms in the absence of structural differences (Cerasa et al., 2010;Molendijk et al., 2012b), which is further supported by additional studies analyzing functional MRI data (Dennis et al., 2011;Egan et al., 2003;Hariri et al., 2003;Hashimoto et al., 2008). However, the meta-analysis by Kambeitz and colleagues did not find an association between rs6265 and hippocampus-mediated memory activation, which might be explained by the large variety of paradigms combining working and episodic memory processes (Kambeitz et al., 2012). Moreover, meta-analyses assessing an association between rs6265 and declarative memory performance revealed contradictory results (Kambeitz et al., 2012;Mandelman and Grigorenko, 2012).
In our meta-analysis we observed an effect of the applied measuring technique (manually traced vs. automatically measured hippocampal volumes) after we investigated the effect of several moderators due to significant between-study heterogeneity and publication bias. First, the overall meta-analysis showed a weakly (g = 0.09) significant association between hippocampal volumes and SNP rs6265. In particular, Val/Val homozygotes had significantly larger hippocampal volumes than Met-carriers. The direction of the effect is in accordance with recent meta-analyses of healthy subjects (Hajek et al., 2012;Kambeitz et al., 2012;Molendijk et al., 2012a), but the effect size in this study was considerably smaller. To further disentangle the dissociable effect of these two measurement approaches, subsequent analyses were conducted after separating the samples by the hippocampus measuring technique. We found that Met-carriers had smaller hippocampal volumes than Val/Val homozygotes (g = 0.22) when the hippocampi were manually segmented. In contrast, we did not find a significant genotype effect with automatic segmentation (g = 0.04). This latter result is consistent with the findings of our original sample in 643 healthy subjects, where we used the automatic segmentation technique from FreeSurfer and also with the results of a recent GWAS analysis in 5776 healthy subjects (Stein et al., 2012). Even though manual segmentation is generally considered as the gold standard due to the precise delineation of anatomical structures, the increasing sample size of imaging studies renders the process of manual segmentation less practicable, as it is both costly and time consuming. Several studies compared manual and different automatic segmentation methods and report comparable accuracy, sensitivity and reproducibility (Bergouignan et al., 2009 Specifically, automated segmentation of the hippocampus using FreeSurfer shows higher correlations with manual segmentation compared to FSL/First (Doring et al., 2011;Morey et al., 2009). Nonetheless, it has been shown that, compared to manual segmentation, FreeSurfer and FSL overestimate hippocampal volumes (Doring et al., 2011;Morey et al., 2009) while they are underestimated by SACHA (Bergouignan et al., 2009). However, our meta-analysis across studies using only manual tracing samples revealed a publication bias, between-study heterogeneity and a moderator effect for the sample size. These effects were further studied in detail to investigate the relation between sample size and publication year. We showed that effect sizes shrink as a function of publication year and sample size. In contrast to the findings of previous meta-analyses (Kambeitz et al., 2012;Molendijk et al., 2012a), this decrease in effect size could not be attributed to publication year alone, but was also linked to an increase in sample size.
Several limitations of our analyses need to be considered. In our meta-analysis, we could not address laterality differences or differences in specific hippocampal sub-regions as many of the included studies only report total hippocampal volumes. Furthermore, we explicitly focused on the impact of the rs6265 polymorphism on hippocampal volumes in healthy subjects, without considering the effect of other SNPs, gene-gene interactions (Honea et al., 2009) or gene-environment interactions Gerritsen et al., 2012). This is of particular relevance, as the impact of the BDNF SNP rs6265 on hippocampal volume could be modified by other SNPs that have already been shown to impact the volume of the hippocampus, such as the Val159Met polymorphism of catecholamine-O-methyltransferase (COMT) (Cerasa et al., 2008;Dutt et al., 2009;Ehrlich et al., 2010;Honea et al., 2009;Taylor et al., 2007), an SNP of ZNF804a (Donohoe et al., 2011;Wei et al., 2012) or the intergenic variant rs7294919 (Stein et al., 2012). Finally, we did not observe a main effect of sex and age on hippocampal volumes, nor did we observe an interaction effect of sex and genotype on hippocampal volumes. Other studies found sex- (Cahill, 2006;Goldstein et al., 2001;Liu et al., 2010;Ruigrok et al., 2013), and age-dependent differences in hippocampal volumes (Driscoll et al., 2003;Malykhin et al., 2008;Raz et al., 2010). Since the association of rs6265 and age-dependent hippocampal changes revealed controversial results (Brooks et al., 2014;Karnik et al., 2010;Sanchez et al., 2011) and the role of sex in this association is not well understood, it would be interesting if future studies would address these questions. Potential reasons for the absence of such effects in our original study are the applied correction for intracranial volume and the limited age-range of our sample.
In summary, the present study does not support the association between SNP rs6265 and hippocampal volumes in healthy individuals. The weak effect observed in the meta-analysis is mainly driven by studies with small sample sizes applying manual segmentation of hippocampi. Our findings confirm the results of previous results based on a large sample size. Moreover, our findings demonstrate an effect of measuring techniques, publication year and sample size.