How Reliable is Genetic Sensitivity Research for Application in Practice?

Background Information

Differences in sensitivity have a partial genetic basis with several studies reporting that specific regions of certain genes (described as candidate genes) are related to sensitivity. One example is the serotonin transporter linked polymorphic region (5-HTTLPR), a genetic variation found in the SLC6A4 gene.

The 5-HTTLPR region is typically characterized as either a short or a long version, referred to as allelic variants. The short allelic variant, compared to the long allelic variant, has been associated with increased Environmental Sensitivity, such as a stronger response to the positive effects of psychological interventions.

However, research that focuses on individual candidate genes has been criticized regarding its reliability. Replicating these research is therefore fundamental to determine the reliability of these kinds of studies that investigate genetic sensitivity.

Research Replication

Research replication is the bedrock of science. To be confident that research discoveries are real, and not the result of some quirk of the data or other unintended factors, published research studies should be conducted multiple times. If the same results are obtained again and, hopefully, several times over, there is confidence that the original findings are reliable.

In 2015, the Open Science Collaboration [2] published a study where they attempted to replicate 100 original studies and, unfortunately, found that less than half reached the criteria to be considered a successful replication, setting off the “replication crisis” in psychology.

Perhaps in no other area of psychological research have issues with replication been highlighted as research that utilizes measured genes (DNA derived genotypes), and in particular candidate gene-by-environment interaction (cGxE) research. These concerns have developed in parallel with intervention research that has increasingly incorporated DNA to help understand why interventions work better for some than others. Given that this genetically informed research has direct implications for intervention programming, assessing the reliability of this research is critical.

Study Aim

To help address replication issues in genetically informed intervention research, my colleagues and I conducted a study to replicate and extend one of the first gene-by-intervention interaction studies [3], which showed intervention efficacy was modified by adolescents’ genetic differences (i.e., short versus long gene variants) in the serotonin transporter gene (5-HTTLPR).

How the study was conducted

The replication we conducted was based on the genetic subsample data from the PROSPER project, a community-university evidence-based system designed to deliver prevention-intervention programs, in the USA. The PROSPER project included 28 rural and small town communities in Iowa and Pennsylvania randomized into 14 intervention and 14 control units.

Data were collected on students when they were approximately 11-12 years old and annually until the age of 17-18 years. Intervention programing, consisting of a classroom-based and family-group format component, were delivered when adolescents were approximately 11-13 years old, and were designed to enhance personal goal setting, social norms, decision making, and navigating peer contexts. Detailed information on the PROSPER project can be found here (http://helpingkidsprosper.org/).

In attempting to replicate the findings reported by Brody and colleagues [3], we sought to stay as close to the original study as possible. Brody and colleagues examined change in risk behavior initiation, which was defined as the onset of alcohol and marijuana use as well as sexual behavior. We had similar measures in PROSPER but not exact, so we created a substance misuse initiation measure that quantified onset of alcohol intoxication, marijuana use, and prescription drug use.

Brody and colleagues [3] found 1) that adolescents who participated in the intervention program (Strong African American Families; https://cfr.uga.edu/saaf-programs/saaf/) showed lower risk initiation, 2) that having at least one copy of the 5-HTTLPR short genetic variant was related to higher risk initiation, and 3) that adolescents in the control group with at least one copy of the 5-HTTLPR short genetic variant had the highest risk initiation compared to adolescents with the 5-HTTLPR long variant in either the intervention or control groups as well as 5-HTTLPR short variant adolescents in the intervention.

Key Findings

Using our substance misuse initiation measure and a similar analytic model, we found 1) that adolescents in the intervention showed lower substance misuse initiation, similar to the Brody study [3], 2) we did not find differences based on 5-HTTLPR genotype, inconsistent with the Brody findings., and 3) we found the same group differences reported by Brody and colleagues [3]: 5-HTTLPR short gene variant adolescents in the control group showed the highest substance misuse initiation but other adolescents with the same gene variant showed considerably lower substance misuse initiation. This pattern was consistent with a vulnerability model where adolescents were at higher risk for substance misuse if they had the 5-HTTLPR short variant, but the interventions delivered in the PROSPER project were able to mitigate that risk.

Implications

Based on the results of this and other research, arguments have been made that valuable resources could be saved if we could identify adolescents who would benefit most from one or another intervention program and provide the programming for them and not others. There are some merits to this argument. For example, it might not make sense to provide a writing intervention that was designed for children with autism to all children in a school.

However, when it comes to deciding who gets what resources based directly on genotypes, things get much more precarious. For example, issues with racism can become apparent pretty quickly. Research on 5-HTTLPR is a case-in-point given the short gene variant is much more frequent among a European ancestry populations compared to African ancestry. In addition, the results reported in the PROSPER study were the result of intervention programming that was delivered universally, that is all 11-13 years old students in each community were invited to participate (and 90% did so at wave 1).

Providing resources in the form of intervention programming to a selection of adolescents – based on genotype and presumed sensitivity – has the potential to generate stigma or other unforeseen undesirable outcomes that could undermine the efficacy of the program and potentially cause other problems. Last, even with reliability evidence as demonstrated by my colleagues and I and the more recent advances in how genotypes are being studied (e.g., polygenic scores), the science is nowhere near to where it would need to be to use such findings for intervention selection processes.