Development of the Comprehensive Addiction Risk Evaluation System: Initial Participant Response to an Online Personalized Feedback Program Integrating Genomic, Behavioral, and Environmental Risk Information

Abstract

Introduction

We have made tremendous advances in understanding the etiology of substance use disorders (SUDs). Despite these advances, screening for SUDs has remained largely unchanged. In this paper, we describe an effort to build a program that integrates advances across genomics, developmental psychology, and epidemiology to provide individuals with personalized information about their addiction risk profile.

Methods

The program was developed based on foundational work from a NIDA-funded project that conducted multivariate analyses of externalizing phenotypes to advance gene identification for SUDs and then characterized how polygenic scores (PGS) and early life behavioral and environmental factors predicted SUDs in diverse longitudinal samples. Based on this work, we created PGS and a behavioral and environmental risk index to generate personalized risk profiles. We carefully considered ethical concerns when developing the program.

Results

We created a user-friendly, self-directed online platform that provides personalized risk information, including overall risk for developing an SUD based on an individual’s combination of genetic, behavioral, and environmental risk, and specific information about genetic risk, based on PGS, and behavioral/environmental risk. Data from the first 188 participants enrolled in an ongoing study to evaluate the platform indicate high satisfaction and low distress at receiving genetic information.

Conclusion

Provision of personalized feedback about addiction risk factors, including genetic information along with behavioral and environmental feedback, may be a viable way to promote earlier screening and intervention with the goal of preventing substance use problems before they start.

Introduction

We have made tremendous advances in understanding the etiology of substance use disorders (SUDs). Hundreds of genomic loci that impact risk for problematic substance use have been identified, in part by capitalizing on shared genetic influences that impact substance use outcomes and other behaviors and disorders characterized by behavioral disinhibition [1–4]. Polygenic scores (PGS) calculated from the results of these large, well-powered genome-wide association studies (GWAS) now accounting for up to 10% of the variance in substance use and related externalizing outcomes [1]. In parallel, many early environmental [5, 6] and behavioral [2, 7] factors have been identified that are robustly associated with elevated risk for SUDs across numerous studies. These include childhood behavior problems, including both externalizing behaviors such as conduct disorder and ADHD [8, 9] and internalizing behaviors such as depression and anxiety [10–12], as well as personality traits such as impulsivity and sensation-seeking [13, 14]. Early initiation of substance use and having more substance using peers in adolescence are also indicators of risk for developing problems [15–18]. When considered holistically, these etiological factors can powerfully differentiate levels of risk. For example, in analyses conducted in four large epidemiological samples, with data collected across three countries, among individuals who carried the lowest levels of genomic, behavioral, and environmental risk, only 4% went on to develop an SUD, as compared to 84% of those who had the highest levels of genomic, behavioral, and environmental risk [19].

Despite these advances, screening for SUDs has remained largely unchanged since the early 1980s, when concerted public health efforts at universal alcohol screening were initiated by the USA and World Health Organization [20]. Screening, Brief Intervention, and Referral to Treatment is the gold standard for substance use screening across a variety of settings [21] and is recommended by the US Preventive Services Task Force [22]. Many questionnaires are available online for substance use screening, both through SAMSHA [23] and the NIH, for both practitioners [24, 25] and the general public [26]. Many of these efforts to create publicly available online materials result from increasing awareness of the need to make research tools and results accessible to a variety of audiences, to support the translation of research advances into improvements in human health.

Although there is consistent evidence that Screening, Brief Intervention, and Referral to Treatment can be effective at reducing substance use and associated consequences [27], a problem with the screening measures that are currently in use is that they are nearly entirely focused on frequency of use and problems. In other words, they are designed to detect problems that have already started developing. Ideally, we want to prevent problems before they begin. Genomic and early behavioral risk factors can be informative to this end by providing information about an individual’s likelihood of developing substance use problems, potentially before they even initiate substance use. However, this screening information is not currently incorporated into clinical care or screening measures that are available to the public. Importantly, not only can we use genomic, behavioral, and environmental risk factors to identify at-risk individuals, but there is evidence that the individuals who are most at risk are also most likely to respond to intervention [28, 29]. Evidence-based strategies are available that address underlying risk factors and reduce the likelihood an individual will develop problems [30–32].

Part of the rationale for substantial NIH investment in genomic and epidemiological research is that enhanced understanding of the complex interplay across genetic, environmental, and lifestyle factors could provide the foundation for precision medicine [33], with the idea being that the provision of personalized risk information will help prevent the development of problems and/or allow for earlier intervention before problems have become severe. However, for psychiatric and SUDs, there has been limited progress toward translating etiological advances into clinically relevant tools. We argue that SUDs are a model test case in which to explore important translational questions about the application of genetic epidemiological research to prevention/intervention, as individuals who are at elevated risk can proactively prevent the development of problems by terminating or reducing substance use and/or learning strategies to manage behavioral and emotional dysregulation patterns that elevate risk.

In this paper, we describe efforts to build a program that integrates advances across genomics, developmental psychology, and epidemiology to provide individuals with personalized information about their risk profile and then connect them with tailored evidence-based resources to reduce risk. This online program was developed through an administrative supplement submitted in response to NOT-DA-24-003 “Rapid Translation of Substance Use and Addiction Epidemiology and Prevention Intervention Research” and aims to translate the findings emerging from the parent project “Using the genetic architecture of SUDs to advance gene identification and understanding of pathways of risk” to study the application of genetic epidemiological findings as a novel prevention/intervention program to reduce risky patterns of substance use. We currently have a randomized controlled trial (RCT) underway to study the impact of the provision of personalized addiction risk information on substance use outcomes, and whether different types of resources provided alongside personalized feedback affects substance use outcomes. Here, we describe the development of the program and initial results on satisfaction and response to receiving personalized addiction risk information from the first wave of participants to receive personalized addiction profiles integrating genomic, behavioral, and environmental information.

Methods

Background to Program Development

This work stems from a NIDA-funded project designed to advance gene identification for SUDs and map the pathways by which genetic risk unfolds in conjunction with the environment in diverse longitudinal samples. Under that project, we conducted multivariate genomic analyses to boost power to identify genes involved in substance use outcomes by incorporating information about genetically correlated traits related to behavioral under-control, referred to as the externalizing spectrum [3, 14, 34, 35]. This strategy was rooted in the extensive twin literature demonstrating that different forms of substance use and problems, as well as other externalizing behaviors and disorders, share underlying genetic influences [2, 3, 34, 36]. We identified 579 independent loci that broadly influence risk for substance use and related externalizing outcomes [1]. PGS created from these results predict nearly 10% of the variance in externalizing behavior in independent samples. Notably, the number of loci identified using this strategy exceeds the number of loci identified for any SUD studied individually to date [1, 4, 37–40] and outperforms another multivariate analysis that focused more narrowly on SUDs (the addiction risk factor) [4]. Follow-up analyses demonstrate that our externalizing factor correlates >0.87 with the addiction risk factor, indicating that the shared genetic influences underlying all SUDs largely reflects this broad disposition toward behavioral under-control captured by externalizing [41]. Our approach illustrates the power gained by using additional phenotypes genetically correlated with SUDs to boost power to identify SUD-relevant genetic variation. Accordingly, the PGS resulting from our externalizing GWAS was selected for inclusion in our risk feedback platform. Additionally, we used the platform to raise awareness about the fact that most of the genetic variation influencing substance use problems operates via behavioral regulation.

In subsequent work, we integrated the PGS with epidemiological information on behavioral and environmental risk factors, using data from four longitudinal cohorts with extensive phenotypic and genotypic information: the National Longitudinal Study of Adolescent to Adult Health (Add Health [42, 43]), the Avon Longitudinal Study of Parents and Children (ALSPAC [44–46]), the Collaborative Study on the Genetics of Alcoholism (COGA [47, 48]), and the Finnish Twin Cohort Study [49]. These samples include population-based cohorts from three countries along with a high-risk sample enriched for SUD. We created a behavioral/environmental risk index (BERI¹) by tallying information about epidemiological risk factors that have been robustly associated with substance use outcomes in the literature, including childhood SES, trauma, peer substance use, family history, externalizing and internalizing problems, early initiation of substance use, and adolescent alcohol, tobacco, and cannabis use. Our integrative data analysis demonstrated that both the BERI and PGS were significantly associated with the development of SUDs by young adulthood, with an AUC of 0.78 for any substance dependence [19]. Table 1 shows the % of individuals meeting diagnostic criteria for any SUD in the combined samples data [19], for individuals with low/high PGS (defined as lower/upper quintile) across varying levels of the BERI. These numbers illustrate how the combination of epidemiological risk information and genetic data both contribute to the prediction of SUDs in young adulthood. It is notable that the PGS provides additional information to the BERI, since several of the variables included in the BERI capture behaviors that are intermediary factors reflecting genetic risk (internalizing/externalizing behaviors, early substance use, family history). This work was the foundation for selection of the behavioral/environmental and genomic feedback components included in the personalized addiction risk profiles feedback program.

Table 1.

Percent of individuals meeting criteria for any SUD based on number of environmental risk factors and PGS quintiles

​	PGS
	bottom 20%	top 20%
BERI
0–2	4%	8%
3–4	18%	24%
5–6	32%	31%
7+	74%	84%

Feedback Components Included in the Program

Overview

The program provides feedback on (1) genetic risk and (2) behavioral and environmental risk. This information is combined into an overall risk section. In addition, the program assesses current substance use in order to provide feedback about whether the individual is currently using in ways that could be considered risky or harmful to their health. Additional details about each of these parts of the program are as follows.

Genetic Risk

Genetic risk information is based on PGS derived from the externalizing GWAS [1]. PGS were calculated for each individual using PRS-CS [50] and Plink2 [51]. We then pooled our PGS with the external epidemiological reference samples described under background to assess which quintile the individual’s PGS resided within, with quintile 5 corresponding to the top 20% of genetic risk, and quintile 1 corresponding to the bottom 20% of genetic risk. Individuals whose PGS were in quintile 1 were considered at “lower-than-average” genetic risk. Individuals in quintiles 2–4 were considered to be at “average” genetic risk, and individuals in quintile 5 were considered “higher-than-average.”

Behavioral and Environmental Risk

Eight dimensions that capture behavioral and environmental risk factors predictive of the development of SUDs were included in the final version of the platform based on the original BERI factors [19]. These eight risk dimensions were assessed via a survey consisting of 30 items created from the items included in the epidemiological samples used in the background analyses. Each of the eight risk factors was tallied as present/absent based on endorsement of any of the related criteria, and then the total number of risk factors was summed to create BERI scores ranging from 0 to 8. The final factors and their descriptions are listed in Table 2.

Table 2.

Final BERI risk factors

Risk factors	Description
Early initiation of substance use	Alcohol, cannabis, or tobacco initiation before the age of 15
Frequent adolescent substance use	Frequent (e.g., use 3 days a week or more often) alcohol, cannabis, or tobacco use before the age of 18
Adolescent peer substance use	A few to many friends in high school that used alcohol/cannabis at least once a month and/or daily tobacco use
Childhood externalizing	School suspension, conduct problems, ADHD diagnosis
Lifetime internalizing	History of anxiety or depressive symptoms/diagnosis
Exposure to stress or trauma	Experiencing at least one stressful or traumatic event before the age of 18 (e.g., death of a family member or close friend, chronic bulling/harassment, etc.)
Low childhood socioeconomic status	Parental education (e.g., no high school degree) and income (e.g., receiving public assistance and/or not enough money to meet family needs) during childhood
Parental history of substance use problems	Biological parent with a history of alcohol or other drug use problems

Overall Risk

Overall risk was evaluated using a look-up table that calculated the lifetime prevalence of any SUD (inclusive of alcohol, nicotine, opioid, cannabis, cocaine, stimulant, sedative, or other drug use disorders) based on the combined data from the longitudinal samples (N = 17,175) described under background, for each combination of PGS quintile (1–5) and BERI score (0–8). As a comparison point, the population prevalence of moderate or severe DSM-5 SUD was calculated using the publicly available NESARC III data (N = 36,309), a large US population-based sample [52], as 0.34. Cutoffs to present relative overall risk were determined based on discussions by the research team as follows: individuals whose risk profile (i.e., their combination of genetic and behavioral/environmental risk) was associated with an SUD prevalence rate of ≤0.24 were considered to have “lower-than-average” risk, combinations with prevalence rates between 0.25 and 0.44 were classified as “average,” and profiles associated with prevalence rates ≥0.45 were considered at “higher-than-average” risk.

Current Substance Use

Although the platform is designed to provide information about underlying risk factors with the goal of preventing problems before they begin, we also thought it important to assess current substance use in order to provide feedback about whether individuals were currently using in harmful ways. Individuals that reported yes to past 30-day alcohol or other drug use were administered the Alcohol Use Disorder Identification Test (AUDIT [53]) and/or Drug Use Disorder Identification Test (DUDIT [54]). Individuals with a score of 8 or more on the AUDIT and/or 6 or higher on the DUDIT received a pop-up warning (described in further detail in the Results).

Ethical Considerations

Concerns have been raised that the provision of genomic information, especially for complex psychiatric traits, may be poorly understood and/or cause feelings of distress [55]. To address these possibilities, all individuals answered three true or false questions designed to assess their understanding of the results after viewing their personalized risk profile. These true/false items were as follows: The results I received from the program “definitively tell me whether or not I will develop problems with addiction,” “include only some of the genetic factors that can contribute to risk for addiction,” and “show that my lifestyle and environment play no role in whether I develop problems with addiction.” Additionally, individuals completed a 10-item questionnaire about affective response and distress (modified from the Feelings About genomic Testing [FACTOR] [56]). Individuals who were characterized as having low understanding by answering two out of the three true/false items incorrectly were contacted by the project team’s genetic counselor. Similarly, individuals who reported high levels of distress as indicated by an answer of “a great deal” to at least one or “a good deal” to at least two of a selected eight items from the scale, upon receiving their risk estimates were also contacted by the genetic counselor. All individuals will be given the option to talk to a genetic counselor at the termination of the study. Satisfaction with the program was measured with 12 items immediately after receiving personalized risk estimates. These items were designed by the study team and modified from previous prevention studies [30, 57].

Results

Development of Program Text

Our interdisciplinary team, with expertise in clinical and developmental psychology, genetics, and psychiatric genetic counseling, created program text to accompany the personalized feedback on genetic and behavioral/environmental factors. Our objective in developing the program text was to ensure it would be understandable to the general population (e.g., avoid scientific jargon, appropriate language for 18+ individuals) and be empowering and non-stigmatizing. We followed best practices when presenting information regarding risk (e.g., nondeterministic language). When presenting risk information, we used clear labels (e.g., “higher-than-average”) rather than percentiles which can lead to confusion and the misperception of risk [58]. We also paired risk results with graphics (e.g., charts and bell curves) to help with the understanding of results.

Platform Development

The program was built as an online platform in collaboration with a professional website development company with the goal of being visually appealing and easy to navigate. Following initial development of the website, our research team shared access with colleagues, psychiatric genetic counselors, and young adults (the target population for the RCT) to get preliminary feedback on elements of the program such as design and accessibility. Minor updates to the program text and graphics were made as a result.

Program Components

Participants begin by creating an account, which allows them to log in and return to the program at any time. After account creation, participants complete the behavioral and environmental risk index (BERI) survey and the AUDIT/DUDIT, if applicable. Participants that endorse hazardous levels of alcohol and/or other drug use based on their AUDIT/DUDIT scores receive a notification at the completion of their survey informing them that their current levels of substance use may be harmful and impact their health. They are also provided with links to websites that provide information about finding help for substance use problems (see Figure 1 for an example).

Fig. 1.

Harmful substance use warning (Copyright @2024 Rutgers, The State University of New Jersey. Content may not be reproduced without permission).

Upon survey completion, participants provide a mailing address for their DNA kit. Participants are mailed a saliva collection kit with a pre-paid return envelope. After the sample is returned to the study team, samples are sent in batches to an external laboratory for genotyping. It takes approximately 6–8 weeks for the laboratory to extract DNA and genotype the samples. Once the genotypic information is ready (×0.5 low-pass sequencing is being conducted for this project), the study team creates the PGS and calculates genetic ancestry from principal components. Genetic similarity group is calculated using 1000 Genomes Project (1KGP) phase 3 as reference. EIGENSOFT/SmartPCA [59] was used to perform principal components analysis using the 1KGP phase 3 reference panel to determine SNP weights for each eigenvector. This solution was then projected onto the participant data to generate the genetic similarity groupings. The PGS quintile and genetic similarity group is uploaded to the study platform backend to generate the risk profiles.

Participants are notified via email that their results are ready and are prompted to log into their account to view their results. The results homepage presents overall risk at the top of the page using icon arrays. The first icon array illustrates the population rate (3 out of 10 shaded figures). The second icon array shows the prevalence of SUD among individuals who have their same PGS quintile and BERI score. For example, a population prevalence of 0.7 (calculated as detailed in Methods) would be illustrated with 7 out of the 10 figures shaded. Underneath the overall risk, the summary of their genetic risk and behavioral and environmental risk is included. Participants then click on each of these sections to obtain more information about each area (Figure 2).

Fig. 2.

Overall results homepage (Copyright @2024 Rutgers, The State University of New Jersey. Content may not be reproduced without permission).

Participants whose genetic similarity group is not European receive a pop-up message that overlaps the genetic risk section and must be read and closed before they can access their genetic risk information. This message explains that most genetic risk estimates have been developed using data from individuals whose genomes are most similar to people from Europe, and that their genetic risk score may have reduced accuracy based on their more diverse genomic background (Figure 3).

Fig. 3.

Pop-up for participants of non-European ancestry (Copyright @2024 Rutgers, The State University of New Jersey. Content may not be reproduced without permission).

On the genetic feedback page, participants are presented with a bell curve showing the regions of the distribution that correspond to each of the three risk levels (lower-than-average/average/higher-than-average). Below the bell curve is text that provides psychoeducation about how genetic influences impact addiction outcomes. A section with the heading “Remember – Genes are not destiny!” is designed to counter genetic determinism and empower participants by emphasizing that understanding their risk can help them take positive steps to reduce the chance of developing problems, even with higher-than-average risk. Lastly, a pie chart displays how genes included in the current risk score only comprise 10% of the factors that impact the overall likelihood of experiencing substance use problems. This chart is paired with a footnote that this estimate is based on data from individuals of European ancestry (Figure 4).

Fig. 4.

Genetic risk results (Copyright @2024 Rutgers, The State University of New Jersey. Content may not be reproduced without permission).

On the behavioral and environmental feedback page, participants are shown how many of the 8 total risk factors assessed they endorsed. Accompanying text explains that these risk factors are nondeterministic and, parallel to the genetic feedback page, participants are reminded that there are things they can do to reduce their risk, even if they have experienced many of the identified risk factors. For each risk factor, a brief description about the risk factor and how it is related to substance use problems is provided Figure 5.

Fig. 5.

Behavioral/environmental risk results (Copyright @2024 Rutgers, The State University of New Jersey. Content may not be reproduced without permission).

In the ongoing study, after viewing their risk estimates, participants complete a brief survey assessing satisfaction, understanding, and affective response to receiving genetic information. Then participants are given access to resources. In the current RCT, there are three types of resources being provided (individuals are randomly assigned to one condition, with all information being available at the end of the study): (1) psychoeducation, whereby participants receive information about addiction, ways to support their well-being, and substance specific harm reduction information; (2) online CBT modules, whereby participants are directed to an external site with evidence-based online content tailored to the individual; and (3) genetic counseling, whereby individuals have the opportunity to discuss their results with a psychiatric genetic counselor in an online session.

Preliminary Results

At the time of writing, N = 188 participants had completed the program and received their personalized risk estimates. These individuals are participating in the ongoing RCT, which recruited young adults ages 18–26 via social media, ResearchMatch, and community flyers. The sample analyzed here was 76% female, 61% cisgender woman, 72% white, and 86% non-Hispanic. The mean age was 21.87 (SD = 2.32). Past 30-day alcohol use was endorsed by 68% of participants; there were no sex differences (p = 0.57). Males had significantly higher AUDIT Consumption scores (mean 4.47, SD = 1.74) than females (mean 3.42, SD = 2.0), p = 0.003. However, there were no significant differences between males (mean 3.28, SD = 3.86) and females (mean 1.86, SD = 2.91) for the AUDIT Problems scores. Other drug use in the past 30 days was endorsed by 40% of the sample, with no sex differences (p = 0.08). Males had higher DUDIT scores reflecting other drug use problems (mean 8.59, SD = 10.45) as compared to females (mean 3.59, SD = 6.38), p = 0.015. We report initial feedback on the program based upon this first wave of participants to complete the program.

Understanding

Ninety-five % of participants passed the understanding check, meaning they correctly answered 2 out of 3 true/false questions correctly. The 9 participants that answered two or more of the questions incorrectly were contacted by email to alert them that their survey responses indicated there may be confusion regarding their understanding of what the personalized risk information means. They were offered the opportunity to meet with a genetic counselor to discuss their results or to reach out to the study team with questions or concerns. Two participants responded and opted to meet with the genetic counselor. One individual was already randomized to the genetic counseling; thus, the understanding was discussed as part of the typical session. The other participant was not assigned to the genetic counseling condition, so a separate appointment was scheduled. Upon meeting with the participants, the genetic counselor deemed they both had adequate understanding.

Participant Satisfaction

Overall, participants were highly satisfied with the program: 92% strongly agreed or agreed that they enjoyed receiving their personalized risk profile, 87% strongly agreed or agreed that they learned new information, and 76% strongly agreed/agreed that they would recommend the program to a friend (see Figure 6 for responses across all satisfaction items).

Fig. 6.

Participant responses on program satisfaction items.

Affective Response

Overall, most participants reported minimal feelings of distress after receiving their personalized risk estimates, e.g., 96% reported “not at all” to feeling regret about their decision to receive genetic information, 84% reported they were “not at all” upset about their genetic test results (see Figure 7 for responses across all affective response items). Eight participants were contacted by a genetic counselor via email because their survey answers to the affective response items were flagged by the criteria described in the Methods. No participants responded with a request to meet with the genetic counselor.

Fig. 7.

Participant responses on FACTOR items assessing affective response to results.

For items with sufficient variability after collapsing the positive and negative response options for interpretability, we performed chi-squared and Fisher’s exact tests to evaluate whether satisfaction and distress response items differed by genetic risk level. There were significant differences for three items (Fig. 8) which assessed feeling sad, happy, and anxious/nervous about results. Individuals with higher genetic risk were more likely to endorse feeling sad as well as anxious/nervous and less likely to endorse feeling happy about their results (ps <0.05). Despite these differences, the majority of individuals who received “higher-than-average” genetic feedback still reported feeling “not at all” or only “a little bit” sad (84%) or anxious (71%).

Fig. 8.

Participant responses on FACTOR for items with significant differences by genetic risk level. Figure displays satisfaction items that were significantly different by genetic risk level. Percent % and (n) of participants are included for each item.

Discussion

To our knowledge, this is the first effort to provide personalized addiction risk information directly to individuals that combines genomic, behavioral, and environmental risk factors. It builds upon early studies that provided information about candidate genes involved in alcohol metabolism [60] and smoking outcomes [61]. Our development of this program stems from the authors’ strong conviction that it is critically important, to the point of being an ethical imperative, to translate our research advances to support the promotion of health. We found high levels of satisfaction with the program, with the vast majority of individuals (≥80%) reporting that they learned useful information, that they learned new information about genetic risk, that they enjoyed receiving their personalized risk profile, that the online format was convenient and the information was understandable. More than 75% of individuals endorsed that they would recommend the program to a friend. We find these results encouraging for the prospect of developing precision psychiatry approaches that incorporate personalized feedback. Further, our findings suggest that complex information about genetic, behavioral, and environmental risk can be understandable provided directly to individuals, which may serve as a cost-effective manner of scaling and delivering this information that is not dependent on the involvement of a health care provider.

There is active ongoing debate about whether genomic information related to psychiatric and SUDs should be provided directly to individuals, and at what point this information is “ready” for clinical application [62–65]. Many valid ethical concerns have been raised about the provision of complex genetic information in the form of PGS [66–68]. These include that individuals will misinterpret genetic risk to mean they are destined to develop problems and adopt counter-productive deterministic mindsets, and/or that the receipt of elevated genetic risk information will cause distress. These are issues that the genetic counseling profession has been addressing for many years, with best practices available for presenting genetic information in ways that are empowering and health promoting [69, 70]. We attempted to incorporate these best practices in the language and visual presentation of the results. We also put in place safeguards to identify and address misunderstanding or distress. In practice we found that very few people showed misunderstanding or distress (<5% each). Follow-up with those individuals who were flagged by our safety checks resulted in very few individuals who indicated that they wanted to meet with a genetic counselor or follow-up with the study team. Future RCT follow-ups with those individuals will hopefully clarify whether there are any lingering misunderstandings or feelings of distress. For the small number who responded (N = 2), we largely found that our safety check items were not indicative of substantial misunderstanding or significant distress.

There are limitations to the current program and areas that deserve further consideration. Many choices about the presentation of the results were subjective and represented collective decisions by the team at the time of development. The cut-offs for providing feedback of “lower-than-average,” “average,” and “higher-than-average” for the overall risk section and assigning labels of “lower-than-average”/“higher-than-average” to PGS quintiles corresponding to the bottom and top 20% were determined through discussions by the research team rather than being strictly guided by evidence-based thresholds. More recently, we have discussed whether the labels for overall risk are necessary and the relative risk information (icon arrays) should be presented absent any labels. It has also been suggested to us that everyone whose PGS does not indicate they are at elevated risk should be classified as “average” genetic risk, so as to avoid inadvertently promote risk-enhancing behaviors among individuals who are told they are at “lower-than-average” genetic risk. Previous work, in which we provided emerging adults with varying hypothetical levels of genetic risk and asked them how it might alter their substance use behavior, found no evidence that individuals who received lower-than-average genetic risk feedback intended to increase their substance use [71]. Nevertheless, these are active areas for discussion by the field. Other decisions that were made that may warrant further consideration include whether all individuals who answer one of the three understanding items incorrectly should be contacted for follow-up, rather than the 2/3 threshold we used, and whether the NESARC moderate/severe SUD prevalence is the best reference for population prevalence. We hope that presenting our program will inspire constructive conversations about how to further improve the provision of personalized risk information. The reality is that our program was developed within the limited funding structure and timeframe provided by an NIH administrative supplement, and we acknowledge that there are many aspects of the program that would benefit from additional refinement and study.

The decision to bin PGS into quintiles for provision of risk feedback was also a reflection of this being a beta version of the program. Though the implementation of arbitrary thresholds for risk stratification is common in psychiatry [72], as exemplified by diagnoses of mild, moderate, and severe SUDs, it necessarily results in loss of information in service of providing more easily interpretable categories of risk. We did internal checks to ensure that the individuals to whom we were providing personalized genetic feedback of “higher-than-average” risk were not changing classification as a function of methodological nuances in calculation of the PGS. However, future developments of personalized feedback programs would benefit from more sophisticated algorithms to create risk profiles. This will necessitate funding by NIH, commercial partners, or other funding agencies to develop and refine more advanced feedback platforms.

One of the known limitations of genomic research is that most discovery samples are comprised individuals of European genomic ancestry, making the resulting PGS less predictive when applied to individuals with differing genomic backgrounds [66, 67]. This creates challenges for the application of genomic advances to clinical practice. On the one hand, we do not want to provide individuals with inaccurate information about their genetic risk; on the other hand, we do not want to create a discriminatory system whereby genomic information is only made available to some individuals. Though efforts are underway to enhance representation in genomic research and ensure that genomic advances are applicable to all individuals [39, 73, 74], remedying the extant discrepancies in the research base will take time. The strategy we decided upon for our personalized feedback program was to provide information about this limitation to individuals for whom their genomic data suggested their risk scores may be less accurate and to alert them to this issue before their genetic risk information could be viewed. Though less than ideal, we decided this was the best way to proceed at this time. We felt comfortable with this course of action in part because the behavioral and environmental information is weighted more heavily in the calculation of overall risk and also assesses behaviors that are known to reflect genetic risk [75]. How to deal with the inequity that exists in genetic research as we translate this information for clinical application is obviously an area that merits careful consideration.

The goal of providing personalized risk information is ultimately to support individuals in promoting their health and well-being. Our ongoing RCT, in which we are studying substance use and mental health outcomes up to 3 months after receiving risk profiles, will provide important information about the extent to which the provision of personalized addiction risk information modifies behavior. It is encouraging that in the survey completed immediately after viewing their results, 85% of individuals endorsed that the genetic information was at least a little bit helpful in planning for the future (33% said it was a great or good deal helpful), and 49% of participants strongly agreed or agreed that “the information I received will influence the choices I make about alcohol and other drug use in the future.” That said, there are well-documented discrepancies between health intentions and actual behavior [76], so longitudinal follow-up will be critical.

In conclusion, approximately 80% of individuals state that they want access to their genetic risk information for psychiatric and SUDs [77]; however, few public-facing programs are available to provide this information, despite drastic advances in genomics with resulting PGS now predictive of behavioral outcomes at levels on par with socio-environmental variables currently used in prevention programs [1]. Interest in direct-to-consumer genetic testing has grown exponentially over the past decade [78–80] and genetic testing is a multibillion dollar industry. Accordingly, we believe that the research community should be actively involved in driving forward research on how to deliver genomic information in ways that are understandable, ethically conscience, and health promoting. Our development of “CARES” (Comprehensive Addiction Risk Evaluation System) is a first step in that direction. There are many important and outstanding questions to be addressed about how best to bring genomic advances to the public, and we hope that our efforts to move in this direction will inspire the research community to engage in constructive conversation about how best to translate our scientific advances to support health outcomes.

Acknowledgments

The Externalizing Consortium would like to thank the following groups for making the research possible: 23andMe, Add Health, Vanderbilt University Medical Center’s BioVU, Collaborative Study on the Genetics of Alcoholism (COGA), the Psychiatric Genomics Consortium’s Substance Use Disorders working group, UK10K Consortium, UK Biobank, and Philadelphia Neurodevelopmental Cohort.

Statement of Ethics

This study protocol was reviewed and approved by the Rutgers University Institutional Review Board, Approval No. [Pro2024002407]. Written informed consent was obtained from all participants to participate in the study.

Conflict of Interest Statement

The CARES platform has been licensed by Rutgers University to Thrive Genetics, Inc. Danielle Dick is the Chief Scientific Officer of Thrive Genetics, Inc. Dr. Dick is also on the Advisory Board for the Seek Women’s Health Company. She owns stock in both companies. She received royalties from authoring the book, The Child Code: Understanding Your Child’s Unique Nature for Happier, More Effective Parenting, published by Avery, an imprint of the Penguin group. Dr. Danielle M. Dick was a member of the journal’s Editorial Board at the time of submission. Dr. Danielle Dick is supported by NIH R01 AA015416, P50 AA022537, R25 AA027402, R01 AA030996, and U10 AA008401 from the National Institute on Alcohol Abuse and Alcoholism (NIAAA); R01 DA050721, R25 DA061485, and R01 DA021421 from the National Institute on Drug Abuse (NIDA); and UM1 TR004789 from the National Center for Advancing Translational Sciences (NCATS). The content is solely the responsibility of the authors and does not necessarily represent the official views of the above funding bodies. Maia Choi has received support from the National Institute on Drug Abuse as a T32 fellow (T32 DA55569). The content is solely the responsibility of the authors and does not necessarily represent the official views of the above funding bodies. Fazil Aliev is supported by NIH R01 AA015416 (Finnish Twin Study), P50 AA022537 (Alcohol Research Center), R01 AA030996, and U10 AA008401 (COGA) from the National Institute on Alcohol Abuse and Alcoholism (NIAAA); R01 DA050721 (Externalizing Consortium) and R01 DA021421 from the National Institute on Drug Abuse (NIDA); and UM1 TR004789 from the National Center for Advancing Translational Sciences (NCATS). The content is solely the responsibility of the authors and does not necessarily represent the official views of the above funding bodies. Emily Balcke is supported by NIH R01 AA015416 and P50 AA022537 from the National Institute on Alcohol Abuse and Alcoholism (NIAAA), R01 DA050721 from the National Institute on Drug Abuse (NIDA), and UM1 TR004789 from the National Center for Advancing Translational Sciences (NCATS). The content is solely the responsibility of the authors and does not necessarily represent the official views of the above funding bodies. Diya Patel has no conflicts of interest to declare. Kennedy Borle is supported by CIHR Banting and Best Doctoral Fellowship, BC Children’s Hospital Research Brain, Behaviour, and Development Trainee Award, and the UBC Public Scholar Initiative. The content is solely the responsibility of the authors and does not necessarily represent the official views of the above funding bodies. Jehannine Austin receives research funding from Genome BC and Genome Canada. The content is solely the responsibility of the authors and does not necessarily represent the official views of the above funding bodies.

Funding Sources

This project was funded through an administrative supplement to R01 DA050721 from the National Institute on Drug Abuse (NIDA). The parent grant was funded by the National Institute on Alcohol Abuse and Alcoholism (R01 AA015416 – administrative supplement to DMD) and the National Institute on Drug Abuse (R01 DA050721 to DMD).

Author Contributions

Danielle Dick and Jehannine Austin: conceptualization, methodology, investigation, resources, writing – original draft, supervision, and funding acquisition. Maia Choi: conceptualization, methodology, formal analysis, investigation, data Curation, writing – original draft, and visualization. Emily Balcke: conceptualization, methodology, investigation, writing – original draft, and visualization. Fazil Aliev: conceptualization, methodology, formal analysis, investigation, data curation, and writing – reviewing and editing. Diya Patel: investigation and writing – reviewing and editing. Kennedy Borle: conceptualization, investigation, and writing – reviewing and editing.

Funding Statement

This project was funded through an administrative supplement to R01 DA050721 from the National Institute on Drug Abuse (NIDA). The parent grant was funded by the National Institute on Alcohol Abuse and Alcoholism (R01 AA015416 – administrative supplement to DMD) and the National Institute on Drug Abuse (R01 DA050721 to DMD).

Data Availability Statement

The participants of this study did not give written consent for their data to be shared publicly, so due to the sensitive nature of the research, supporting data are not available. Further inquiries can be directed to the corresponding author.