OUP user menu

The brain-derived neurotrophic factor gene in suicidal behaviour: a meta-analysis

Clement C. Zai, Mirko Manchia, Vincenzo De Luca, Arun K. Tiwari, Nabilah I. Chowdhury, Gwyneth C. Zai, Ryan P. Tong, Zeynep Yilmaz, Sajid A. Shaikh, John Strauss, James L. Kennedy
DOI: http://dx.doi.org/10.1017/S1461145711001313 1037-1042 First published online: 1 September 2012


Suicide is a prominent public health problem. Its aetiology is complex, and the brain-derived neurotrophic factor (BDNF) has been implicated. We performed the first meta-analysis of the functional BDNF marker Val66Met (rs6265, 196G>A) in suicidal behaviour using data from 11 previously published samples plus our present sample (total n=3352 subjects, 1202 with history of suicidal behaviour. The meta-analysis including all 12 studies showed a trend for the Met-carrying genotypes and Met allele conferring risk for suicide (random-effects model p=0.096; ORMet-carrier=1.13, 95% CI 0.98–1.30, and random-effects model p=0.032; ORMet=1.16, 95% CI 1.01–1.32, respectively). Furthermore, we found the Met allele and the Met allele-carrying genotypes to be associated with history of suicide attempt (eight studies; allelic meta-analysis – random-effects model: p=0.013; fixed-effects model: p=0.006; genotypic meta-analysis – random-effects model: p=0.017; fixed-effects model: p=0.008). Taken together, the results from our study suggest that BDNF Val66Met is involved in suicidality. Further studies are required to elucidate its role in suicidal behaviour.

Key words
  • Brain-derived neurotrophic factor (BDNF)
  • genetics
  • meta-analysis
  • schizophrenia
  • suicidal behaviour
  • suicide attempt


The World Health Organization estimates that almost one million deaths each year are attributable to suicide, and suicide attempt is close to 20 times more common than suicide completion. Thus, understanding the impact of the interplay between molecular and clinical factors behind suicidal attempt is a major public health concern. Approximately 90% of suicide victims suffer from at least one psychiatric disorder. Family and twin studies support a strong genetic basis of suicidal behaviour (Bondy et al. 2006). The concordant phenotype in twin studies includes both completed suicide and suicide attempts (Bondy et al. 2006), and their heritability appears to be at least partially independent of psychiatric disorders (Bondy et al. 2006). A meta-analysis of excess mortality rates in schizophrenia (SCZ) by Brown (1997) found suicides to be its leading cause, accounting for 28% of excess deaths in SCZ.

Brain-derived neurotrophic factor (BDNF) expression has been found altered in suicide attempters and victims (reviewed in Dwivedi, 2010). Recent genetic studies of the functional Val66Met (rs6265, 196G>A) polymorphism (Egan et al. 2003) and suicide have yielded mixed results (Hong et al. 2003; Huang & Lee, 2007; Spalletta et al. 2010; Zarrilli et al. 2009). A number of research groups found that patients with the Met allele were at increased risk for suicide in the context of various psychiatric disorders (Huang & Lee, 2007; Iga et al. 2007; Kim et al. 2008; Sarchiapone et al. 2008). In the current study, we aimed to perform a meta-analysis of the functional Val66Met polymorphism in suicidal behaviour, adding data from our sample of SCZ patients.

Patients and methods


The research subjects were recruited from the Centre for Addiction and Mental Health. We included only subjects of European ancestry in this study, as indicated by self-reported ethnicities of grandparents. The Structured Clinical Interview for DSM-IV (SCID-I) was administered as the primary diagnostic tool. A clinical summary of detailed information about sequence, context, and severity of symptoms was prepared for each patient. We recruited 196 SCZ probands with data on suicide attempt.

Genomic DNA was purified from whole-blood samples using a non-enzymatic method described previously (Zai et al. 2010a). Genotyping of the Val66Met polymorphism was done using TaqMan genotyping assay (Applied Biosystems, USA) as described previously (Zai et al. 2010a).

Ethical considerations

The scientific work described in this article complies with the current laws of Canada and the USA, as well as the ethical standards established in the 1964 Declaration of Helsinki. Informed consent was obtained before patients' participation, and this study was approved by the Ethics Committee of the Centre for Addiction and Mental Health.


We performed PubMed searches using the key terms ‘BDNF’, ‘Val66Met’, ‘suicide’, ‘aggressive’, and ‘genetics’. We included papers examining Val66Met for association with suicide attempt or completion, with raw genotype counts available for subjects with suicidal behaviour and subjects without. We performed the meta-analysis using Stata version 8 (StataCorp, USA) as described previously (Zai et al. 2010b). Briefly, the pooled odds ratio was calculated using the inverse variance method under the fixed-effects model, and the DerSimonian and Laird method under the random-effects model within Stata. Meta-regression was performed to assess the effects of age, percentage males, and ethnicity on the results from the meta-analysis. Publication bias was estimated using Begg's and Egger's tests. The effect of each study was ascertained by sensitivity analysis.


Meta-analysis of BDNF Val66Met in suicidal behaviour

Our literature search using key terms ‘BDNF’, ‘suicide’, and ‘genetics’ yielded 22 studies. We included eight previously published studies that gave genotype counts for subjects with suicidal behaviour and subjects without suicidal behaviour. We found the article by Pregelj et al. (2011) using the search terms ‘Val66Met’ and ‘suicide’, and the article by Spalletta et al. (2010) using the search terms ‘Val66Met’ and ‘aggressive’. The genotype counts in suicide attempters and non-attempters for the Perroud et al. (2009) study were provided by the corresponding author. The genotype counts in suicide attempters and non-attempters for the de Luca et al. (2011) study were included as part of the present study. In total, we obtained genotypes for 3352 subjects, 1202 of which have a history of suicidal behaviour. The demographic information of each study is shown in Table 1. Two of the studies reported on suicide completion vs. healthy controls (Pregelj et al. 2011; Zarrilli et al. 2009) and one study did not specify on suicidal behaviour (Iga et al. 2007). Of the ten studies where suicide cases were matched with non-suicidal psychiatric controls, the primary diagnosis was SCZ in three samples, and either bipolar disorder or major depression in the remaining seven samples.

View this table:
Table 1

The meta-analysis including all 12 studies showed a trend for the Met-carrying genotypes and Met allele conferring risk for suicide (Fig. 1 a: random-effects model p=0.096; ORMet-carrier=1.13, 95% CI 0.98–1.30, and Fig. 1 b: random-effects model p=0.032; ORMet=1.16, 95% CI 1.01–1.32, respectively). There is significant heterogeneity among the studies (Table 2). Sensitivity analysis did not reveal a significant effect of removal of any single study on the pooled odds ratio (Table 2). Ethnicity and frequency appear to be contributing to the results of the genotypic meta-analysis (Table 1). Thus, we analysed studies on patients of East Asian and European ancestry separately (Table 2). We found the association was stronger within the East Asian (Chinese, Korean, Japanese) samples (n=755, of which 171 were suicide attempters: random-effects model p=0.035; ORMet-carrier=1.60, 95% CI 1.03–2.49, and random-effects model p=0.006; ORMet=1.49, 95% CI 1.12–1.97). When we removed one study where the genotype frequency was out of Hardy–Weinberg equilibrium (HWE) (Vincze et al. 2008), one study with unspecified suicidal behaviour (Iga et al. 2007), and two studies with a different case-control matching strategy in terms of suicide phenotype (suicide completion) and controls (healthy controls instead of psychiatric controls) (Pregelj et al. 2011; Zarrilli et al. 2009), we found that the between-study heterogeneity diminished [p(Met-carrier vs. Val/Val)=0.26, p(Met vs. Val)=0.21]. The association became more significant for the remaining eight studies of suicide attempt vs. non-attempt matched for psychiatric disorders, both under the random-effects model (n=1804, of which 433 were suicide attempters: p=0.017; ORMet-carrier=1.27, 95% CI 1.04–1.55, and p=0.013; ORMet=1.25, 95% CI 1.05–1.49) and under the fixed-effects model (p=0.008; ORMet-carrier=1.25, 95% CI 1.06–1.49, and p=0.006; ORMet=1.22, 95% CI 1.06–1.41) (Table 2). There is no evidence of significant publication bias associated with the genotypic and allelic analyses.

Fig. 1

Forest plots for meta-analysis of the Val66Met polymorphism in suicidal behaviour under the random-effects model. (a) Results for the Met allele-carrying genotypes relative to the Val/Val genotype. (b) Results for the Met allele relative to the Val allele.

View this table:
Table 2


To our knowledge, this is the first meta-analytical study of the BDNF gene with suicidal behaviour. The genotype data from Kohli et al. (2010) was not available for us to include in this meta-analysis. While data being not available does not represent a clear bias, bias from this kind of omission cannot be ruled out. Because the secondary analyses with different ethnicities and suicide attempt are related to the primary analysis in all 12 studies, we did not correct for these subgroup analyses. In marked contrast to the results from our SCZ sample, the meta-analysis of the 12 studies with available genotypes indicated that the BDNF rs6265 Val66Met marker has statistically significant effect on risk of suicide, with the Met allele and Met-allele carriers at risk for suicidal behaviour. The association may also be stronger in the East Asian population, although we may not have enough European samples to detect an association. The results became more robust when we removed four studies – one study in which the genotype frequency is out of HWE (Vincze et al. 2008), another study where the suicidal behaviour is not specified (Iga et al. 2007), and the third and fourth studies where suicide completers are matched to healthy controls (Pregelj et al. 2011; Zarrilli et al. 2009). Results from Pregelj et al. (2011) and Zarrilli et al. (2009), for example, have been criticized since they could have been influenced by an association between Val66Met and the psychiatric disorder as opposed to suicidal behaviour. Similarly, results from Iga et al. (2007) could have been influenced by the inclusion of suicide ideation in the classification of suicidal behaviour. The study by Vincze et al. (2008) may have sampling bias and other technical issues. The refined sample exhibits an overall statistically significant association between the Met allele and risk for suicide attempt in subjects with various psychiatric disorders. The pooled odds ratio indicates that the contribution of Val66Met in suicidal attempt is significant but small, suggesting that Val66Met may be a marker in linkage disequilibrium with a nearby variant in BDNF. Deep re-sequencing of the BDNF gene will help to clarify this possibility. Additional genetic factors are probably involved in risk for suicidal behaviour. Testing genes along BDNF signalling pathways could shed light on underlying biological epistases (Kohli et al. 2010; Perroud et al. 2009).

Val66Met may influence the risk diathesis for suicide-related phenotypes, including impulsive aggression. Rodents deficient in BDNF display aggressive behaviour (Lyons et al. 1999). Environmental stress has been shown to reduce BDNF levels (reviewed in Dwivedi, 2010). Epigenetic interrogation of the BDNF gene may provide insight into the effect of the environment on BDNF expression and add to the genetic findings (De Luca et al. 2011; Keller et al. 2010). Given that the low-functioning BDNF Met allele appears to be a risk factor for suicidality, it could be hypothesized that a compound that increases BDNF function may be helpful for suicide prevention (Schmidt & Duman, 2010). Nonetheless, further replication studies and functional analyses are required to better understand how decreased BDNF signalling may be contributing to suicidal behaviour.


We thank the funding sources for the co-authors of this study: Canadian Institute of Health Research (J.L.K.), Sobey Fellowship (M.M.), C.R. Younger Foundation (C.C.Z.), Bebensee Foundation (C.C.Z.), Centre for Addiction and Mental Health Postdoctoral Fellowship (C.C.Z., A.K.T.), Eli Lilly (C.C.Z.), and the American Foundation on Suicide Prevention (C.C.Z., J.S., V.dL.). We also thank Dr N. Perroud for providing the genotype data for the Perroud et al. (2009) study, and Dr D. Brent for providing information on the Brent et al. (2010) study. We also thank all study participants.

Statement of Interest

J. L. Kennedy has received honorarium from Eli Lilly.


View Abstract