A systematic review and secondary data analysis of the interactions between the serotonin transporter 5-HTTLPR polymorphism and environmental and psychological factors in eating disorders

Abstract

Objectives

To summarize and synthesize the growing gene x environment (GxE) research investigating the promoter region of the serotonin transporter gene (5-HTTLPR) in the eating disorders (ED) field, and overcome the common limitation of low sample size, by undertaking a systematic review followed by a secondary data meta-analysis of studies identified by the review.

Method

A systematic review of articles using PsycINFO, PubMed, and EMBASE was undertaken to identify studies investigating the interaction between 5-HTTLPR and an environmental or psychological factor, with an ED-related outcome variable. Seven studies were identified by the systematic review, with complete data sets of five community (n = 1750, 64.5% female) and two clinical (n = 426, 100% female) samples combined to perform four secondary-data analyses: 5-HTTLPR x Traumatic Life Events to predict ED status (n = 909), 5-HTTLPR x Sexual and Physical Abuse to predict bulimic symptoms (n = 1097), 5-HTTLPR x Depression to predict bulimic symptoms (n = 1256), and 5-HTTLPR x Impulsiveness to predict disordered eating (n = 1149).

Results

Under a multiplicative model, the low function (s) allele of 5-HTTLPR interacted with traumatic life events and experiencing both sexual and physical abuse (but not only one) to predict increased likelihood of an ED and bulimic symptoms, respectively. However, under an additive model there was also an interaction between sexual and physical abuse considered independently and 5-HTTLPR, and no interaction with traumatic life events. No other GxE interactions were significant.

Conclusion

Early promising results should be followed-up with continued cross-institutional collaboration in order to achieve the large sample sizes necessary for genetic research.

Introduction

Over the past decade, etiological models of eating disorders (EDs) have increasingly acknowledged the role of genetics, with twin studies estimating a notable heritable component (approximately 40–60%; Bulik et al., 2006, Bulik et al., 2010, Fairweather-Schmidt and Wade, 2015, Trace et al., 2013). Investigations so far have not consistently identified specific candidate genes associated with increased ED risk, suggesting that hereditary factors in EDs may not operate via simple genetic association (Trace et al., 2013). Hence, studies are now increasingly examining whether environmental factors moderate the influence of candidate genes on risk for pathological eating behavior. Gene x environment (GxE) interaction research in the ED field is still relatively novel, with early studies identifying potential candidate genes associated with ED risk under specific environmental conditions (e.g., history of abuse; Steiger et al., 2012). In anticipation of the increased popularity of this research focus, it is timely to evaluate the current state of evidence and to highlight existing limitations, in order to guide the direction and methods of future GxE studies in eating pathology.

Previous research examining genetic influences on eating pathology has primarily focused on genes in the serotonin and dopamine systems linked to functions relevant to EDs, including appetite, mood, and reward sensitivity (e.g., SLC6A4, HTR2A, DRD2, DRD4, DAT1, and COMT; see Culbert et al., 2015, and Trace et al., 2013, for a review). Direct genetic association studies have not provided a clear picture of the links between specific genes and EDs or disordered eating symptoms, with many initial significant findings failing to achieve consistent replication (see Calati et al., 2011, Culbert et al., 2015, Lee and Lin, 2010, Scherag et al., 2010, Trace et al., 2013).

One reason for a lack of direct association between allele frequency and ED risk may be that this relationship is moderated by environmental factors. Under the diathesis-stress model of GxE interactions, individuals carrying a ‘risk’ allele may be more susceptible to EDs when exposed to environmental stressors, but show no differences in outcome in the absence of challenging environmental circumstances, compared to those without the risky genotype (Caspi et al., 2003, Monroe and Simons, 1991). The role of GxE interactions in psychology has gained increasing attention since Caspi et al. (2003) found that stressful life events increased susceptibility to depression for those with one or two copies of the short (s) allele of the serotonin transporter gene (5-HTTLPR polymorphism).

Studies have since largely focussed on 5-HTTLPR due to its biological relevance to psychiatric disorders (with the s-allele reducing serotonin transporter transcription efficiency; Heils et al., 1996), and early significant findings in the depression literature (Karg et al., 2011). Despite substantial research investigating GxE interactions with 5-HTTLPR and other polymorphisms, many studies are limited by small sample size, and replicability remains a major issue (see Duncan et al., 2014 for a review; Risch et al., 2009). Furthermore, most studies to date failed to control for confounding influences on the GxE interaction by not including all required covariate x gene and covariate x environment contrast terms in the regression model (Keller, 2014). Studies examining case-control samples have also tended to evaluate the GxE effect using logistic regression and have thus tested departures from a multiplicative model of interaction, which is believed to be less biologically plausible than an additive model (Rothman, 1976, Rothman and Greenland, 1998).

GxE studies of candidate genes in eating pathology have been scarce. A recent review by Culbert et al. (2015) highlighted the heterogeneity of candidate GxE research in eating pathology. Their investigation identified five studies examining candidate GxE interactions with eating pathology outcome variables. Two studies reported a significant GxE interaction for 5-HTTLPR (Karwautz et al., 2011; parenting styles; Akkermann et al., 2012; traumatic life events), while one study investigating a psychological factor did not (Racine et al., 2009; impulsivity). The two remaining studies examined other genes (NR3C1 x childhood abuse, Steiger et al., 2011; BDNF x restricted food intake; Akkermann et al., 2011), finding significant interactions to predict bulimia nervosa (BN) spectrum pathology. This paper presents a good start in summarising candidate GxE literature in eating disorders (although it was not a systematic review and thus omitted several studies, e.g., Stoltenberg et al., 2012, van Strien et al., 2010), and reflects the growing focus on gene x environment interactions in the eating disorders field.

While candidate GxE research in eating pathology is still in its infancy, it is not premature to consider how to best utilise academic resources to avoid the pitfalls GxE research has faced in other fields, such as lack of consistent replication and small sample sizes (Dick et al., 2015). This will aid greater accuracy in GxE findings, which is a vital step in increasing understanding of how individual differences at the genetic level can influence susceptibility to eating pathology. In the depression field, a protocol for a collaborative meta-analysis to achieve these aims has been published (N = 33,761), with authors aiming to re-analyse their data using a standardised analysis script to increase consistency of analytic methods and phenotypic definitions (Culverhouse et al., 2013). Future collaborations could integrate complete datasets for combined re-analysis rather than relying on summary statistics. No such study has been undertaken in the ED field so far.

Thus, the present study aims to provide a systematic, detailed overview and re-analysis of current GxE studies investigating 5-HTTLPR in eating pathology, to clarify the current state of knowledge and to encourage future research to build upon this via continued focus on replication of published findings and multi-institute collaborations to achieve larger sample sizes. Specifically, it will examine, via a systematic review, existing studies that have analysed how the interaction between 5-HTTLPR and an environmental or psychological factor influences ED status or sub-threshold ED symptomatology. Secondary data meta-analyses to re-analyse GxE interactions using larger sample sizes with appropriate control of confounding variables as per Keller (2014) will then be performed by aggregating the results of three or more existing studies in a series of smaller analyses. Each analysis will be tested according to the multiplicative model of interaction, for consistency with prior research, and also according to the additive model of interaction, because of the possibility that this better represents and may be more sensitive to identifying gene x environment interactions. This study will be reported according to PRISMA guidelines where applicable (Moher et al., 2010).