To date, the incorporation of genetic feedback in substance use interventions is almost exclusive to studies of tobacco use. Two randomized trials evaluated the efficacy of genetic feedback for GSTM1, which encodes an enzyme involved in detoxifying environmental carcinogens and shows an association with lung cancer risk [13, 14]. In one study, the addition of GSTM1 feedback to a multicomponent intervention resulted in greater abstinence rates 6 months (but not 12 months) later [15]. Another found that GSTM1 feedback led to decreases in smoking and greater motivation to quit smoking compared to a control group [16]. In one nonrandomized study, participants who received GSTM1 feedback showed high utilization of smoking cessation services, but risk perception did not differ across higher- and lower-risk genotype groups [17]. Other studies have focused on CYP2D6, which also has functional significance for the metabolism of environmental toxins. In one trial, the addition of CYP2D6 feedback to a multicomponent intervention predicted perceived health risks and perceived benefits of smoking cessation, as well as subsequent quit attempts, but did not predict cessation [18, 19]. A study evaluating feedback about the L-myc EcoR1 polymorphism found no overall effect on smoking rates [20]. Finally, providing smokers with genotype-specific feedback about risk for alpha-1 antitrypsin deficiency, a genetic condition that increases risk for emphysema, was associated with greater cessation rates among those at higher genetic risk [21]. Overall, the use of genetic feedback in smoking interventions has yielded modest and inconsistent effects on cognitive outcomes (e.g., risk perception) and no evidence of consistent or sustained effects on behavior.

The efficacy of genetic feedback interventions for reducing alcohol-related health risks has not been studied. Heavy alcohol consumption and alcohol dependence are associated with disease burden globally [22]; in particular, the risk for upper aerodigestive tract cancers increases significantly with cumulative alcohol exposure [23, 24]. Upper aerodigestive tract cancers are attributed largely to exposure to acetaldehyde, a metabolic byproduct of ethanol and an established animal carcinogen [23, 24]. The primary pathway of alcohol metabolism includes oxidation of ethanol to acetaldehyde by alcohol dehydrogenase enzymes, followed by oxidation of acetaldehyde, which is catalyzed primarily by the mitochondrial aldehyde dehydrogenase (ALDH) enzyme [25]. Genetic variations are demonstrated to influence the catalytic properties of these enzymes, leading to differences in rates of acetaldehyde production or elimination [25].

Variations in alcohol metabolizing genes are demonstrated to moderate risk for alcohol-related cancers [24, 26, 27]. ALDH2, which encodes the mitochondrial aldehyde dehydrogenase enzyme, shows the strongest association with cancer risk. The ALDH2*2 allele, which is almost exclusive to individuals of northeast Asian descent, encodes a functionally inactive enzyme subunit that leads to impaired acetaldehyde metabolism. Individuals with ALDH2*2 show increased levels of blood acetaldehyde and increased physiological responses to alcohol (e.g., skin flushing, tachycardia) following alcohol consumption [28]. For individuals homozygous for ALDH2*2 (ALDH2*2/*2 genotype), mitochondrial aldehyde dehydrogenase is completely inactive. As a result, these individuals show strong physiological reactions to alcohol, low drinking rates, and virtually no risk for alcohol dependence [28, 29]. ALDH2*2 heterozygotes (ALDH2*1/*2 genotype) also show elevated blood acetaldehyde during alcohol consumption and have lower rates of alcohol use and dependence than those with the common ALDH2*1/*1 genotype. However, protection against alcohol dependence in heterozygotes is incomplete; a sizable proportion report moderate or heavy drinking and some develop alcohol dependence [29].

It is widely established that ALDH2*2 heterozygotes who drink alcohol are at significantly increased risk for alcohol-related cancers, in particular squamous cell esophageal cancer (e.g., [26, 30–32]); this association shows a dose–response pattern [26, 33, 34]. Odds ratios for esophageal cancer risk for ALDH2*1/*2 individuals (compared to ALDH2*1/*1) have been estimated to range from four to 13 across Japanese studies [26], but odds ratios as high as 30–95 have been reported among heavy drinkers [26, 30, 34]. One meta-analysis, which included seven studies, reported summary odds ratios of 7.07 (95% confidence interval 3.67–13.6) for heavy drinkers and 2.49 (95% confidence interval 1.29–4.79) for moderate drinkers with the ALDH2*1/*2 genotype [35]. Notably, there is molecular evidence to support that elevated cancer risk among heterozygotes is attributable to acetaldehyde exposure during alcohol consumption [33]. Because ALDH2*1/*2 individuals are estimated to number 540 million worldwide—comprising 8% of the global population—experts have called for large-scale prevention efforts in this group [33].

One empirically supported approach for reducing alcohol-related risks is brief feedback and motivational enhancement interventions [36, 37]. Informed by motivational [38] and social psychological theories [39], these interventions typically incorporate personalized feedback about an individual’s drinking behavior relative to a given population or reference group. Theoretically, such information can enhance awareness of alcohol-related risks and highlight discrepancies between current behaviors and future goals [37, 39]. Whereas these interventions often provide normative feedback about the target behavior (alcohol use) in comparison to a reference group (e.g., college students), genetic feedback interventions provide individualized information about possible health risks based on one’s genotype relative to individuals with a different genotype [15–21]. While these two approaches appear compatible, personalized feedback interventions for alcohol use have not incorporated risk information specific to genetic variants. Additionally, whereas alcohol interventions have increasingly used web-based approaches to promote wide dissemination of personalized interventions [40, 41], few studies of genetic risk information have tested internet-delivered interventions [17].

The current study evaluated the feasibility, acceptability, and short-term efficacy of brief, web-based intervention incorporating personalized feedback and risk information specific to ALDH2 genotype. The primary aim was to examine whether genetic feedback and information about alcohol-related cancer risk would influence drinking behavior among individuals with the ALDH2*1/*2 genotype. However, a focus on ALDH2 allowed the additional goal of evaluating feedback about genetic risk for alcohol dependence. Specifically, meta-analyses show that the risk for alcohol dependence among ALDH2*1/*1 individuals is roughly 4.5 times higher than for ALDH2*1/*2 individuals and 8.3 times higher than for ALDH2*2/*2 individuals [29]. Thus, we evaluated genetic risk information specific to (a) alcohol-related cancers (targeting ALDH2*1/*2 individuals) and (b) alcohol dependence (targeting ALDH2*1/*1 individuals, given increased risk relative to ALDH2*1/*2 individuals). Changes in drinking behavior over a 30-day period following the intervention served as the primary outcome. Secondary outcomes included cognitive and motivational correlates of behavior change that were conceptually relevant based on theoretical considerations and prior empirical findings.

Previous studies incorporating genetic feedback as an adjunct to smoking interventions have focused primarily on adults who were motivated for cessation. In contrast, this study targeted nontreatment-seeking young adults who generally reported little motivation to reduce their drinking, perhaps reflective of modest drinking rates in this sample. In this context, evidence for significant intervention effects among ALDH2*1/*2 individuals is notable for at least two reasons. First, given that even moderate drinking is associated with elevated cancer risk in this group, interventions for moderate drinkers could have a significant impact on overall disease burden in this subgroup [33]. Second, the results are consistent with the notion that personalized genetic feedback might be efficacious for addressing complex behaviors even when individuals are precontemplative of change. These results provide a basis for examining the efficacy of more comprehensive primary interventions addressing alcohol use among young adults with the ALDH2*1/*2 genotype [33]. To the extent that such interventions might influence drinking trajectories, primary interventions could play a significant role in reducing cancer risk attributable to cumulative alcohol exposure.

Our focus on ALDH2 allowed the secondary goal of evaluating health risk information specific to genetic risk for alcohol dependence. In contrast to the risk information delivered to ALDH2*1/*2 participants, providing ALDH2*1/*1 individuals with information about increased risk for alcohol dependence did not influence drinking. Because genetic feedback precludes random assignment to feedback conditions, the two feedback groups could not be compared directly. However, evaluating these two types of feedback in same study allows for some speculation about differential intervention effects. One possibility is that genetic risk information specific to a medical diagnosis is more salient than risk information for a substance use disorder. For example, consistent with existing theory on responses to genetic risk information [44, 52], to the extent that a cancer diagnosis is perceived as (a) more severe or (b) having a stronger genetic basis compared to alcohol dependence, cancer-related risk information would be perceived as more salient. Perceived disease severity and perceived genetic contribution to disease are therefore two constructs of theoretical relevance for future studies [44, 52]. An alternative explanation for the differential intervention effects is that preexisting differences across genotype groups could influence intervention response. For instance, because ALDH2*1/*2 individuals show heightened physiological responses to alcohol, genetic information could have been more salient for this group in that they could explicitly relate genetic feedback information to past drinking experiences. Differences in drinking history across genotypes could also contribute to differential intervention response.

Whereas associations of individual genetic variants with substance use behavior are usually small, focusing on ALDH2 allowed for the communication of relatively large effect sizes (i.e., odds ratios) with respect to alcohol-related health outcomes. Theoretically, the ability to explain increasing variance in genetic risk should increase the efficacy of genetic feedback interventions [53]. However, simulation studies suggest that explaining even moderate variance in genetic risk for complex disorders will require the consideration of vast numbers of susceptibility alleles [5]. Although such studies are underway, accurate prediction of genetic risk for complex diseases using genome-wide panels is not yet a reality [4]. In the interim, studies of genetic feedback that focus on established risk markers, as was the goal in this study, can be informative from a proof-of-concept standpoint. It is also noteworthy that genetic feedback appears capable of promoting health protective behaviors irrespective of whether participants have the higher-or lower-risk genotype [17, 54] or whether preventative measures exist for the disorder in question [55]. Thus, one possibility is that simply providing personalized genetic feedback could enhance intervention uptake or efficacy, irrespective of the specificity of test results.

Limitations of this study include a moderately sized sample, a relatively short follow-up period, and a focus on a specific population. Additionally, given the preliminary nature of this study and its goals to establish intervention feasibility and acceptability, we did not aim to test a comprehensive theoretical model or to conduct formal meditational analyses in evaluating intervention effects. Future studies could address additional theoretical constructs relevant for health risk perception (for example, self-efficacy, coping appraisals, perceived disease severity [43, 44, 52]) as mediators or moderators of intervention efficacy. Notably, experimental analog studies have proven useful for evaluating effects of genetic risk information on psychological and motivational responses in the context of established health behavior theories [43, 53] and would be useful for refining interventions such as the one described here. Another limitation concerns our assessment of risk perception, which relied on a single-item measure developed for the current study. Ideally, future studies could develop improved risk perception measures with established psychometric properties. Using self-report measures of drinking behavior is also a potential limitation, as the veracity of participants’ reports cannot be confirmed.