Discrepancy between self-reported intoxication and psychomotor vigilance task performance after alcohol intake: associations with BrAC, alcohol-metabolizing genotypes, and drinking tendency Authors: Midori Motoi, Nakyeong Shin, Momo Hama, Yukitaro Yasuda, Takayuki Nishimura, Eigo Nishimura, Hideo Toyoshima, Yuiko Araki, Takafumi Katsumura, Yali Xia, Tomoaki Ohashi, Nariaki Kuriyama, Yuna Miyauchi, Yuki Motomura, Shigekazu Higuchi Published: 2026 ABSTRACT AbstractIntroductionThe degrees of subjective intoxication (SI) and vigilance impairment following alcohol consumption vary among individuals. This variability may reflect both metabolic enzyme genotypes and drinking behavior. The purpose of this study was to examine the integrated associations among alcohol-metabolizing enzyme genotypes, drinking tendency, and acute alcohol-induced changes in breath alcohol concentration (BrAC), SI, and vigilance performance.MethodsAlcohol (0.5 g/kg body weight) was orally administered to 93 healthy adults. BrAC, SI, and the Psychomotor Vigilance Task (PVT) were measured at multiple time points between 0 and 100 min. Repeated-measures ANOVA was performed with time as the within-subject factor and alcohol dehydrogenase 1B (ADH1B) and aldehyde dehydrogenase 2 (ALDH2) genotypes as between-subject factors. Additionally, using structural equation modeling (SEM), we examined how BrAC, SI, and PVT lapse rates were jointly related to genotypes and the Alcohol Use Disorders Identification Test–Consumption (AUDIT-C) score.ResultsFor BrAC and SI, a significant ALDH2 × time interaction was observed. ALDH2*1/*1 participants had lower BrAC than ALDH2*1/*2 participants. Regarding PVT performance, the main effect of time was observed, but no significant main effects or interactions involving genotypes were found. Additionally, results from SEM showed that, in the 40-min model, BrAC positively predicted SI and lapse rate, and AUDIT-C negatively predicted SI but not lapse rate. ALDH2 genotype positively predicted AUDIT-C and negatively predicted BrAC; ALDH2 genotype also showed a significant direct negative path to SI but not on lapse rate. In the 100-min model, BrAC positively predicted SI but not lapse rate, and AUDIT-C negatively predicted SI but not lapse rate.ConclusionALDH2 genotype contributed to variability in BrAC and SI but not PVT performance. In the 40-min SEM model, higher AUDIT-C predicted lower SI despite no association with BrAC or PVT lapse rate. This pattern suggests a discrepancy between self-reported intoxication and PVT performance. These findings may inform risk communication and intervention strategies to reduce alcohol-related accidents and improve health education.Trial registrationUMIN Clinical Trials Registry (UMIN-CTR), UMIN000055964, registered on 2024–10-28.Supplementary InformationThe online version contains supplementary material available at 10.1186/s40101-026-00434-w.Keywords: Breath alcohol concentration, Subjective intoxication, ALDH2, AUDIT-C, PVT, Structural equation modeling Introduction Alcohol consumption poses significant public health challenges owing to its association with traffic accidents, violence, acute health damage, and chronic health consequences resulting from long-term alcohol use [1]. Acute alcohol intake rapidly impairs the psychological and behavioral functions essential for safe decision-making, including alertness, attention, and inhibitory control [2, 3]. These acute effects are well documented, and their magnitude varies markedly across individuals, even after the same alcohol dose. Such variability likely reflects influences operating at multiple levels, including biological factors (e.g., genetic background and sex) and behavioral factors. It remains unclear how these factors jointly shape subjective intoxication (SI) and vigilance impairment during acute alcohol exposure. This gap limits understanding of the mechanisms underlying risky drinking behaviors and alcohol-related harm. SI and the Psychomotor Vigilance Task (PVT) are useful behavioral indicators of drinking-related risk. SI is a key indicator of acute alcohol response. Differences in sensitivity to alcohol's effects predict future heavy drinking and the risk of developing Alcohol use disorder (AUD) [4]. PVT is an indicator that is acutely sensitive to sleep deprivation and alcohol-induced transient impairment, as evidenced by reduced reaction time distributions and an increased lapse rate [5], and predicts the risk of motor vehicle accidents [6]. Abbreviated PVT protocols, including a 3-min version, validly detect alcohol-induced cognitive impairment and lapses and are considered highly informative in practical and operational settings [7]. These indicators therefore provide complementary measurements of the acute functional consequences of alcohol intake. At the same time, breath alcohol concentration (BrAC) functions as an objective measure of systemic alcohol exposure following drinking. Alcohol is metabolized primarily via hepatic alcohol dehydrogenase (ADH) and aldehyde dehydrogenase (ALDH). Functional polymorphisms in alcohol dehydrogenase 1B (ADH1B; ADH1B*1 and ADH1B*2) and aldehyde dehydrogenase 2 (ALDH2; ALDH2*1 and ALDH2*2) substantially influence the rate of alcohol elimination [8]. Individuals carrying ADH1B*2 or ALDH2*2 alleles exhibit accelerated ethanol oxidation or impaired acetaldehyde clearance, respectively, resulting in [8] greater accumulation and stronger aversive physiological responses during alcohol consumption [9]. Many studies in healthy adult males have shown no difference between the genotypes, ALDH2*1/*1 and *1/*2, in peak blood alcohol concentration (BAC) or BrAC trajectories, implying that the BAC/BrAC time course is largely independent of ALDH2 genotype [9–11]. However, several studies—particularly in East Asian populations—have reported genotype-related differences in BrAC duration and elimination slope [12, 13]. Together, these findings suggest that the extent to which ALDH2 genotype influences BrAC/BAC kinetics warrants re-examination. While biological factors influence alcohol metabolism, repeated alcohol exposure can also lead to metabolic tolerance (enhanced metabolic capacity) and behavioral tolerance (reduced susceptibility to alcohol-induced impairment). The Alcohol Use Disorders Identification Test (AUDIT), which assesses the severity and risk of problem drinking, and the Alcohol Use Disorders Identification Test-Consumption (AUDIT-C), which evaluates alcohol consumption levels using three items from AUDIT, are widely used as indicators of drinking patterns [3, 14, 15]. A high AUDIT-C score is considered a threshold indicating potentially harmful drinking or alcohol use disorders [15]. Chronic heavy drinking may induce metabolic tolerance, partly via upregulation of hepatic alcohol-metabolizing enzymes, including cytochrome P450 family 2 subfamily E member 1 (CYP2E1) [16, 17]. Accordingly, BrAC/BAC profiles may vary across AUD/heavy drinkers (HD) and light drinkers (LD) groups due to combined effects of metabolic tolerance and absorption-related factors [16, 18]. In terms of behavioral tolerance, SI decreases with higher drinking tendency, though evidence is mixed regarding the association between habitual drinking and cognitive performance. A recent large-scale study suggests that HD and patients with AUD may exhibit less cognitive impairment and greater behavioral tolerance than LD [19]. This pattern raises the possibility of a mismatch between subjective intoxication and objective impairment. In this context, biological factors such as genetic background and sex are known to influence drinking behavior. Individuals with an inactive ALDH2 enzyme encoded by the *2 allele tend to limit alcohol consumption because of unpleasant reactions (acetaldehyde accumulation), which consequently acts as a protective factor against alcohol dependence [8, 11]. In addition, female participants generally consume less alcohol and drink less frequently than male participants. Sex also influences BrAC levels; female participants typically have a lower body water content and reduced gastric first-pass metabolism, leading to higher BrAC levels after consuming the same amount of alcohol as male participants [16, 20]. The optimal AUDIT-C cutoff is also set lower for female participants (e.g., ≥ 3), and their score distributions tend to skew lower [21]. Together, these findings suggest that both genotype and sex shape drinking behavior and alcohol pharmacokinetics, and that drinking patterns may contribute to behavioral tolerance. Clarifying how biological and behavioral factors jointly shape BrAC, subjective intoxication, and functional impairment is essential for understanding alcohol-related risk. To date, few studies have integrated genetic factors, drinking tendency, and their acute consequences (BrAC, SI, and PVT) within a single experimental framework. This study aimed to comprehensively evaluate variation in acute functional impairment and subjective perception during alcohol intoxication associated with ADH1B/ALDH2 genotype and sex, considering two pathways: alcohol metabolism and drinking tendency. We quantitatively measured changes in SI and vigilance performance (PVT) over time following alcohol consumption and used structural equation modeling (SEM) to evaluate the contributions of BrAC, AUDIT-C, and genetic polymorphisms (ADH1B/ALDH2) to these measures. Based on prior evidence, we tested the following hypotheses: Hypothesis 1: ADH1B polymorphisms are associated with BrAC, which in turn predicts SI and vigilance performance. Hypothesis 1: ADH1B polymorphisms are associated with BrAC, which in turn predicts SI and vigilance performance. Hypothesis 2: ALDH2 polymorphisms are associated with drinking tendency (AUDIT-C), which in turn predicts SI and PVT lapse rates. Hypothesis 2: ALDH2 polymorphisms are associated with drinking tendency (AUDIT-C), which in turn predicts SI and PVT lapse rates. Materials and methods Ethics statement This study was approved by the Kyushu University Hospital Institutional Review Board (approval no. 24013). After the experimental procedures had been explained, written informed consent was obtained from all participants before enrollment. The study was registered in the UMIN Clinical Trials Registry (UMIN000055964). Participants Healthy participants were recruited from Kyushu University and other nearby universities. A total of 104 participants (sex: male, n = 61; female, n = 43) participated in the screening process. Eligibility criteria were as follows: 1) Age ≥ 21 years, 2) BMI 18.0–30.0 kg/m2, 3) AUDIT score ≤ 14 points, 4) A history of consuming ≥ 50 g of pure alcohol in a day without clinically significant adverse health effects. Exclusion criteria included history of cardiovascular, renal, or hepatic disease; use of medications affecting sleep–wake cycles (e.g., first-generation antihistamines); shift work including night shifts; time zone travel (≥ 4 h) within the past month; sleep disorders; excessive caffeine intake (≥ 5 cups of coffee/day); and other factors determined by the physician. Participants showing a positive reaction on the alcohol patch test (Alcohol Constitution Test Patch, Life Care Giken Co., Ltd., Toyama, Japan) were also excluded. Accordingly, no participants with the ALDH2*2/*2 genotype were observed. This left a total of 98 eligible participants (sex: male, n = 58; female, n = 40). Study timeline Pre-test instructions: Participants were instructed to avoid staying up late for three days prior to the experiment, refrain from drinking alcohol the day before, and get eight hours of sleep. Caffeine and smoking were prohibited on the day of the experiment. Participants were instructed to eat their usual lunch no later than five hours before the start of the experiment, and to drink only non-caffeinated beverages thereafter. Two hours before the start of the experiment, upon arrival at the laboratory, participants’ height and weight were measured, and saliva samples for genotyping were collected. Genotype determination was performed only after completion of all BrAC, SI, and PVT measurements. Thus, the experimenters were blind to the participants’ genotypes throughout the data collection process, minimizing the possibility of genotype-related experimenter bias during these measurements. Immediately afterwards, participants consumed a standardized light meal provided by the experimenters. Practice trials for the Go/No-go task and PVT were conducted before the main task to confirm participants understood the procedures. Following a 30-min break, baseline measurements were taken before drinking the alcohol, including SI assessment, BrAC measurement, and the main PVT trial. After drinking began (0 min), a sequence of 11 BrAC/subjective measurements were taken, one every 10 min; participants were instructed to rinse their mouths and swallow a sip of water 3 min before each BrAC measurement. Data from 10 to 30 min were excluded from the analysis. In the lab, up to six participants performed the task side by side, separated by partitions. The PVT task was performed at 0, 40, 70, and 100 min (Fig. 1). After completing the measurements, participants were confirmed to have BrAC ≤ 0.10 mg/L. Following a nurse's health check, they were dismissed. Experimental timeline. Participants began drinking at 0 min (time = 0). Breath alcohol concentration (BrAC) and subjective intoxication (SI) were assessed every 10 min from 0 to 100 min. The Psychomotor Vigilance Task (PVT) was administered at 0, 40, 70, and 100 min DNA analysis Saliva samples were collected using an Oragene Discover kit (OGR-600, DNA Genotek, Ottawa, Canada). Genomic DNA was then extracted from the saliva using the manufacturer's protocol with the PT-L2P-5 purification solution (DNA Genotek, Ottawa, Canada). The TaqMan SNP Genotyping Assay (Applied Biosystems, Waltham, MA, USA) was used to determine the genotypes of rs671 in ALDH2 and rs1229984 in ADH1B. Polymerase chain reaction (PCR) was performed using the CFX96 real-time PCR system (Bio-Rad Laboratories, Hercules, CA, USA). The reactions were conducted using TaqMan Genotyping Master Mix (Thermo Fisher Scientific, Waltham, MA, USA) under the manufacturer-recommended conditions. The TaqMan probe AssayIDs were the following: C_11703892_10 (rs671 in ALDH2) and C_2688467_20 (rs1229984 in ADH1B). Genotype calling was performed automatically using CFX Manager, and the discrimination was also visually confirmed using the plot. Alcohol administration Vodka at 0.5 g/kg body weight was administered orally in three separate doses 10 min apart, a total period of 30 min. The beverage was prepared by diluting 40% vodka to 12.5% with water and adding 8 g of lemon juice. Measurement indicators BrAC was measured using the ST-3000 breath alcohol detector (Sanko Techno, Tokyo, Japan). At each time point, three consecutive measurements were taken and their mean value was used for analysis. The participants were instructed to rinse their mouths thoroughly before the measurements. SI was assessed using a single-item visual analog scale (VAS) ranging from 0 (“not at all intoxicated”) to 100 (“very intoxicated”). At each measurement time point, participants rated their current level of subjective intoxication by adjusting a slider on a laptop screen using a mouse. The assessment form was implemented using Google Apps Script. Psychomotor Vigilance Task (PVT) To assess vigilance, subjects performed a 5-min PVT on a computer. Although the experimental protocol also included other cognitive tasks, such as a Go/No-go task, the present study focused on the PVT because the primary aim was to evaluate the temporal dynamics of vigilance impairment in relation to genetic factors and subjective perception. Stimuli were presented using Presentation (Neurobehavioral Systems, Berkeley, CA, USA). Inter-stimulus intervals were randomly set between two and ten seconds, and a numerical counter appeared at the center of the screen. Participants were instructed to press the spacebar as quickly as possible upon stimulus presentation. After each response, the reaction time (ms) was displayed as feedback. PVT performance was assessed using two metrics: lapse rate (%) and reciprocal reaction time (rRT; 1/RT). Lapse rate (%) was defined as the proportion of responses with reaction times exceeding 355 ms among valid trials; the count of such lapses (k) was divided by the number of valid trials (n) to obtain k/n. To avoid nonlinearity at the boundary values (0 or 1), lapse rate was analyzed on the logit scale after applying a small offset of (k + 0.5)/(n + 1). rRT was used as an index of reaction speed and was expressed in s⁻1 units. Participants with missing values at any time point were excluded from the analysis. Statistical analysis A total of 98 participants were enrolled. Participants with the ADH1B*1/*1 genotype (n = 5) were not combined with ADH1B*1/*2 participants because the analytic plan treated ADH1B genotypes separately based on their functional differences. Because the ADH1B*1/*1 subgroup was very small, these participants were excluded from the primary analysis to avoid unstable estimates. Therefore, 93 participants (male, n = 55; female, n = 38) were included in the primary analysis. Comparison of demographic characteristics Table 1 shows the baseline characteristics of the participants. Sex (male/female) was compared among the four ADH1B genotype × ALDH2 genotype groups using a chi-square test. Age, weight, height, BMI, and AUDIT-C were examined using a two-factor analysis of variance (ANOVA) (ADH1B × ALDH2). Table 1Participant characteristics by ADH1B × ALDH2 genotypeALDH2*1/*1ALDH2*1/*2OverallADH1B*2/*2ADH1B *1/*2ADH1B*2/*2ADH1B *1/*2CharacteristicN = 93N = 50N = 15N = 18N = 10Age22.3 (1.6)22.3 (1.1)22.3 (1.0)22.7 (3.3)21.6 (0.7)Sex, Female n (%)38 (41%)20 (40%)9 (50%)5 (33%)4 (40%)Height166.9 (9.3)167.3 (9.8)164.1 (7.9)167.2 (8.3)169.3 (10.5)Weight58.1 (8.8)58.9 (9.2)55.4 (6.7)57.1 (9.4)60.3 (9.3)BMI20.8 (2.1)21.0 (2.2)20.5 (1.7)20.3 (2.0)21.0 (2.1)AUDIT-C*4.2 (1.9)4.5 (1.8)4.5 (2.1)3.3 (2.0)3.6 (1.6)Values are mean (SD) unless otherwise stated. n indicates the number of participantsSex was compared using Fisher’s exact test*indicates a significant main effect of ALDH2 (two-way ANOVA: ADH1B × ALDH2); for AUDIT-C, p = 0.0199. No significant main effects of ADH1B or interactions were observed Participant characteristics by ADH1B × ALDH2 genotype Values are mean (SD) unless otherwise stated. n indicates the number of participants Sex was compared using Fisher’s exact test *indicates a significant main effect of ALDH2 (two-way ANOVA: ADH1B × ALDH2); for AUDIT-C, p = 0.0199. No significant main effects of ADH1B or interactions were observed Comparison of BrAC, SI, and PVT performance BrAC and SI were measured at ten-minute intervals between 0 and 100 min after drinking. For inferential statistics, a repeated- measures ANOVA was performed with time (8 levels: 0, 40–100 min) as the within-subjects factor and ADH1B and ALDH2 genotype as between-subjects factors. PVT indices (lapse rate, rRT) were analyzed using time (4 levels: 0, 40, 70, 100 min) as a within-subject factor and ADH1B and ALDH2 genotypes as between-subject factors. To address heterogeneity of variance, we used heteroskedasticity-consistent standard errors (HC3). When sphericity was violated, degrees of freedom based on the Greenhouse–Geisser correction were reported. Post hoc comparisons and graphical representations utilized estimated marginal means (EMMs) and 95% confidence intervals (CIs) for each condition, and Holm's method was applied for multiple comparison correction. Structural equation modeling SEM was conducted to examine the relationships among genotypes, drinking tendency, BrAC, SI, and PVT lapse rate. The model was specified as a hypothesis-driven observed-variable path model using measured variables only, rather than as a latent-variable model. In this framework, BrAC and AUDIT-C were included as intermediate variables representing acute alcohol exposure and habitual drinking tendency, respectively, rather than as formal biological mediators. Sex was coded as Female = 1, Male = 0; ALDH2 genotype was coded as *1/*1 = 1, *1/*2 = 0; and ADH1B genotype was coded as *1/*2 = 1, *2/*2 = 0. All were entered as binary variables, with the reference category set to zero. Continuous variables (BrAC, SI, lapse, AUDIT-C) were z-standardized within each time point before analysis. As a hypothesis-based common model, we placed ADH1B, ALDH2, and sex upstream, AUDIT-C and BrAC as intermediate variables and SI and Δlapse rate (difference from the lapse rate at 0 min) downstream. The same model was applied separately at 40 min (near-peak) and 100 min (decline phase). Specifically, the following structural equation model was assumed: first, AUDIT-C was predicted by ALDH2, sex, and ADH1B, and BrAC was predicted by ALDH2, AUDIT-C, sex, and ADH1B. As outcomes, SI was regressed on BrAC, AUDIT-C, and ALDH2, and lapse rate was regressed on BrAC, AUDIT-C, and ALDH2. The residual covariance between SI and lapse was freely estimated. Estimation used maximum likelihood with robust standard errors and a Satorra–Bentler scaled test statistic. All covariances between exogenous variables were freely estimated, and missing data were handled using full-information maximum likelihood. Model fit was assessed using the robust Satorra–Bentler chi-square statistic, the comparative fit index (CFI), Tucker–Lewis index (TLI), root mean square error of approximation (RMSEA), and standardized root mean square residual (SRMR), and standardized path coefficients with 95% CIs were reported. Calculation of BrAC decline rate A linear mixed-effects model was fitted with BrAC as the dependent variable, and time since drinking (minutes), centered at 80 min, as a continuous fixed effect. Random intercepts and slopes for each participant were included. Estimation used restricted maximum likelihood (REML), specifying bobyqa as the optimizer. Statistical software Analyses were performed using R software (version 4.4.3) [22] with the following packages: afex [23] for repeated-measures ANOVA with robust standard errors, emmeans [24] for estimated marginal means and post hoc comparisons, lavaan [25] for SEM, lme4 [26] for linear mixed-effects models, and dplyr [27], ggplot2 [28], and semPlot [29] for data processing and visualization. All statistical tests were two-tailed, with the significance level set at p < 0.05. Results Participant characteristics by genotype group Analyses were conducted in 93 participants (male participants, n = 55; female participants, n = 38; mean age 22.3 ± 1.6 years) after excluding ADH1B*1/*1 genotype (n = 5), which was rare in the present sample. Participants were classified into four groups by the combination of ADH1B (*1/*2 vs *2/*2) and ALDH2 (*1/*1 vs *1/*2) genotypes. AUDIT-C scores were significantly higher in ALDH2*1/*1 than in ALDH2*1/*2 (Table 1). Repeated-measures ANOVA BrAC BrAC was measured in all 93 participants. There was a significant ALDH2 genotype × time interaction (F(1.73, 153.90) = 3.67, p = 0.034, partial η2 = 0.040; Fig. 2A). ALDH2*1/*1 showed lower BrAC than ALDH2*1/*2 at 40 (p = 0.040), 50 (p = 0.012), 60 (p = 0.030), and 90 min (p = 0.037), but not at 70, 80, or 100 min (p ≥ 0.059). There were also main effects of time (F(1.73, 153.90) = 562.09, p < 0.001, partial η2 = 0.863) and ALDH2 (F(1, 89) = 5.17, p = 0.025, partial η2 = 0.055). No significant effects involving ADH1B were observed (Table 2). BrAC trajectories of the ADH1B*1/*1 participants are shown in Supplementary Fig. 1. Time courses of breath alcohol concentration (BrAC) and subjective intoxication (SI) after alcohol consumption. A BrAC (mg/L) and B SI (VAS, 0–100) are shown across 0–100 min by ADH1B × ALDH2 genotype group. Points indicate estimated marginal means (EMMs) and colored bands indicate 95% confidence intervals (CIs) from the repeated-measures ANOVA model (n = 93 for BrAC; n = 85 for SI). Asterisks indicate Holm-adjusted pairwise differences in ALDH2 marginal means at each time point (p < 0.05) Table 2Results of two-way ANOVA for the effects on BrAC and SIBrACSIEffectF (df1, df2)p-valueηp2F (df1, df2)p-valueηp2ADH1B0.26(1, 89)0.6130.0030.1(1, 81)0.7590.001ALDH25.17(1, 89)0.0250.0556.54(1, 81)0.0120.075ADH1B:ALDH20.80(1, 89)0.3730.0090.04(1, 81)0.846< 0.001time562.09(1.73,153.9)< 0.0010.863136.49(2.66, 215.08)< 0.0010.628ADH1B:time1.42(1.73,153.9)0.2440.0161.27(2.66, 215.08)0.2870.015ALDH2:time3.67(1.73,153.9)0.0340.0405.41(2.66, 215.08)0.0020.063ADH1B:ALDH2:time0.37(1.73,153.9)0.6570.0040.30(2.66, 215.08)0.8010.004BrAC breath alcohol concentration. SI: subjective intoxicationF(df) F statistic with (df1, df2); df are Greenhouse–Geisser corrected for within-subject effectsηp2 partial eta squaredTests: heteroskedasticity-consistent covariance estimator (HC3) Results of two-way ANOVA for the effects on BrAC and SI BrAC breath alcohol concentration. SI: subjective intoxication F(df) F statistic with (df1, df2); df are Greenhouse–Geisser corrected for within-subject effects ηp2 partial eta squared Tests: heteroskedasticity-consistent covariance estimator (HC3) Subjective intoxication SI was analyzed in 85 participants (eight excluded due to missing data). There was a significant ALDH2 genotype × time interaction (F(2.66, 215.08) = 5.41, p = 0.002, partial η2 = 0.063; Fig. 2B). ALDH2*1/*1 reported lower SI than ALDH2*1/*2 at 40 (p < 0.001), 50 (p = 0.002), and 60 min (p = 0.013), but not from 70 to 100 min (p ≥ 0.063). There were also significant main effects of time (F(2.66, 215.08) = 136.49, p < 0.001, partial η2 = 0.628), and ALDH2 (F(1, 81) = 6.54, p = 0.012, partial η2 = 0.075). No significant effects involving ADH1B were observed (Table 2). PVT performance PVT performance was analyzed in 92 participants (one excluded due to missing data). For lapse rate, there was a significant main effect of time (F(2.68, 235.78) = 9.26, p < 0.001, partial η2 = 0.095; Fig. 3A), with higher lapse rates at 40 (p < 0.001), 70 (p < 0.001), and 100 min (p = 0.010) than at 0 min. For rRT, there was also a significant main effect of time (F(2.58, 227.00) = 15.36, p < 0.001, partial η2 = 0.149; Fig. 3B), with lower rRT at 40 (p < 0.001), 70 (p < 0.001), and 100 min (p = 0.020) than at 0 min. No significant effects involving ADH1B or ALDH2 genotype were observed for either measure (Table 3). Time courses of PVT performance after alcohol consumption. A PVT lapse rate (%) and B reciprocal reaction time (rRT; 1/s) are shown across 0–100 min by ADH1B × ALDH2 genotype group (n = 92). Points indicate estimated marginal means (EMMs) and error bars indicate 95% confidence intervals (CIs). In panel A, lapse rate (proportion of lapses > 355 ms on responded trials) was analyzed on the logit scale with a small offset and back-transformed to percentages. In panel B, rRT was calculated from responded trials excluding lapses > 500 ms. Brackets and asterisks indicate Holm-adjusted comparisons versus 0 min (p <.05) Table 3Results of two-way ANOVA for the effects on PVT performanceLapse raterRTEffectF (df1, df2)p-valueηp2F (df1, df2)p-valueηp2ADH1B0.11 (1, 88)0.7450.0010.28 (1, 88)0.5990.003ALDH20.00 (1, 88)0.966< 0.0010.23 (1, 88)0.6330.003ADH1B:ALDH20.03 (1, 88)0.861< 0.0010.04 (1, 88)0.848< 0.001time9.26 (2.68, 235.78)< 0.0010.09515.36 (2.58, 227.00)< 0.0010.149ADH1B: time0.58 (2.68, 235.78)0.6070.0070.82 (2.58, 227.00)0.4700.009ALDH2: time0.12 (2.68, 235.78)0.9330.0010.44 (2.58, 227.00)0.6960.005ADH1B: ALDH2: time0.63 (2.68, 235.78)0.5770.0070.88 (2.58, 227.00)0.4400.010Lapse rate: proportion of lapses (>355 ms) analyzed on the logit scale with a small offset and back-transformed to percentagesrRT: reciprocal reaction time (1/s) excluding lapses >500 msF(df): F statistic with (df1, df2); df are Greenhouse–Geisser corrected for within-subject effectsηp²: partial eta-squared Results of two-way ANOVA for the effects on PVT performance Lapse rate: proportion of lapses (>355 ms) analyzed on the logit scale with a small offset and back-transformed to percentages rRT: reciprocal reaction time (1/s) excluding lapses >500 ms F(df): F statistic with (df1, df2); df are Greenhouse–Geisser corrected for within-subject effects ηp²: partial eta-squared SEM analysis The path models were used to examine structured associations among genotype, sex, BrAC, AUDIT-C, SI, and PVT lapse rate at 40 and 100 min. Detailed estimates for all paths, including unstandardized coefficients, standard errors, 95% confidence intervals, standardized coefficients, and p-values, are provided in Supplementary Table 1. 40-min model The 40-min path model showed good fit (χ2(4) = 0.42, p = 0.98, CFI = 1.00, TLI = 1.31, RMSEA = 0.00 [90% CI 0.00–0.00], SRMR = 0.01; Fig. 4A). For downstream paths, higher BrAC was associated with higher SI (β = 0.28, p = 0.002) and higher lapse rate (β = 0.27, p = 0.001). Higher AUDIT-C was associated with lower SI (β = − 0.24, p = 0.013) but not with lapse rate (β = − 0.14, p = 0.11). For upstream paths, ALDH2 genotype was positively associated with AUDIT-C (β = 0.26, p = 0.011) and negatively associated with BrAC (β = − 0.22, p = 0.033). ALDH2 also showed a direct negative association with SI (β = − 0.26, p = 0.001), whereas the direct path to lapse rate was not significant (β = − 0.02, p = 0.87). No significant paths involving sex or ADH1B were observed. Common structural equation model fitted at 40 and 100 min after alcohol consumption. The same observed-variable path model structure was fitted at A 40 min and B 100 min to jointly model subjective intoxication (SI) and PVT lapse rate. Sex was coded as Female = 1 and Male = 0; ALDH2 genotype was coded as 1 for *1/*1 and 0 for *1/*2; ADH1B genotype was coded as 1 for *1/*2 and 0 for *2/*2. Only statistically significant paths (two-sided α = 0.05) are shown. Arrow color indicates the sign of the path coefficient (positive = orange, negative = blue). Numbers on arrows indicate standardized path coefficients (β) with 95% confidence intervals in parentheses 100-min model The 100-min path model also showed good fit (χ2(4) = 2.93, p = 0.57, CFI = 1.00, TLI = 1.12, RMSEA = 0.00 [90% CI 0.00–0.14], SRMR = 0.02; Fig. 4B). For downstream paths, BrAC remained positively associated with SI (β = 0.33, p < 0.001), whereas its association with lapse rate was not significant (β = 0.08, p = 0.52). AUDIT-C was again negatively associated with SI (β = − 0.28, p = 0.002) but not with lapse rate (β = − 0.08, p = 0.53). For upstream paths, the positive association between ALDH2 genotype and AUDIT-C persisted (β = 0.26, p = 0.011), whereas ALDH2 was not significantly associated with BrAC (β = − 0.15, p = 0.15). Sex was positively associated with BrAC (β = 0.29, p = 0.002; Female = 1, Male = 0). Direct paths from ALDH2 to SI (β = − 0.07, p = 0.48) and lapse rate (β = − 0.03, p = 0.80) were not significant. No significant paths involving ADH1B were observed. Several alternative SEMs were explored, but none improved model fit over the theoretically constrained main model. BrAC decline rate BrAC decline rate was analyzed in all 93 participants. The mean decline rate was − 0.00152 mg/L/min (95% CI − 0.00173, − 0.00135), estimated from a linear mixed-effects model with BrAC as the dependent variable and time since drinking (minutes) as a continuous fixed effect, with random intercepts and slopes for each participant. Discussion Time course and genotype-related variation in BrAC and SI In the present study, we examined associations of ADH1B/ALDH2 genotypes and AUDIT-C with BrAC, SI, and PVT performance. Results from the repeated measures ANOVA showed that BrAC and SI declined over time after 40 min, whereas PVT performance deteriorated, consistent with prior vigilance task studies [2, 30]. Contrary to our expectation, ADH1B genotype was not associated with BrAC levels over time. The ADH1B*2 allele encodes an ADH enzyme with higher catalytic activity than ADH1B*1; however, because the difference in ethanol-to-acetaldehyde conversion activity between *1/*2 and *2/*2 is relatively small [11, 31, 32] and ADH1B*1/*1 was excluded, our analysis may have been underpowered to detect ADH1B-related BrAC differences. The BrAC trajectories of the excluded ADH1B*1/*1 participants are shown descriptively in Supplementary Fig. 1. Although the small sample size (n = 5) precludes formal inference, these participants tended to show higher BrAC than the ADH1B*1/*2 and *2/*2 groups, as expected from this functional difference, which leads to slower ethanol oxidation. This pattern indicated that ADH1B*1/*1 participants could not be appropriately combined with the *1/*2 group. We therefore excluded these participants from the formal analysis. By contrast, ALDH2 genotype was significantly associated with BrAC. The ALDH2*2 allele reduces acetaldehyde oxidation, leading to acetaldehyde accumulation that may contribute to slower ethanol clearance and higher BrAC [13, 33]. Consistent with our results, a previous study reported a 1.96-fold higher ethanol clearance in Chinese individuals with ALDH2*1/*1 than in those with the ALDH2*1/*2 [12]. Thus, Hypothesis 1 was not supported: ADH1B genotype was not associated with BrAC in the present sample. Vigilance impairment and subjective–objective dissociation Lapse rates and reaction times in the PVT test changed over time after alcohol administration, but no differences were observed between the genotypes. It should be noted that even at the 100-min mark, when the average BrAC fell below Japan's legal limit of 0.15 mg/L, performance was significantly impaired compared with baseline. This suggests that measurable vigilance impairment can persist into the decline phase after moderate alcohol intake. Genotype-related patterns may be more apparent for subjective experience than for simple performance measures at moderate doses. Moreover, although acetaldehyde may impair cognitive function, many studies report little evidence for ALDH2-related differences under simple tasks or low alcohol doses (0.3–0.5 g/kg), and a Japanese study using facial flushing as a proxy for ALDH2*2/*2 genotype has reported greater subjective intoxication without differences in driving-related performance [34]. Together with our findings, this supports a clinically relevant dissociation: perceived intoxication and performance impairment do not necessarily align during alcohol intoxication. SEM integrates genotype, BrAC, and AUDIT-C in relation to SI We used a hypothesis-driven observed-variable path model to integrate genotype, BrAC, and AUDIT-C in relation to SI and PVT lapse rates at 40 and 100 min. In this model, BrAC and AUDIT-C were treated as intermediate variables rather than formal biological mediators. This approach allowed us to examine conceptual links between genetic background, systemic alcohol exposure, habitual drinking tendency, and acute functional outcomes while maintaining a parsimonious model structure (Supplementary Table 1). At 40 min, the model suggested three pathways linking the ALDH2 genotype to SI. First, ALDH2 genotype influenced SI indirectly through BrAC. Second, ALDH2 genotype showed a direct path to SI. Specifically, participants with the ALDH2*1/*1 genotype reported lower SI. Compared with the ALDH2*1/*1 group, the ALDH2*1/*2 group has only 12–20% ALDH2 enzymatic activity, leading to elevated acetaldehyde levels and acute unpleasant reactions such as flushing and palpitations [9], which likely explains the higher SI reported by this group. Third, ALDH2 genotype was associated with AUDIT-C, which in turn predicted SI. Individuals with ALDH2*1/*2 tend to restrict their chronic drinking because of the unpleasant reactions [31]. In this study, participants with ALDH2*1/*1 had significantly higher AUDIT-C scores than those with ALDH2*1/*2 (Table 1; Fig. 4A). However, AUDIT-C did not predict BrAC, contrary to our mediation hypothesis that the pathway from ALDH2 genotype to BrAC is transmitted via AUDIT-C. This pattern is consistent with reports that AUDIT-C is not associated with physiological alcohol indices in healthy individuals [4]. In partial support of Hypothesis 2, ALDH2 genotype was associated with AUDIT-C, and AUDIT-C predicted SI; however, AUDIT-C was not associated with BrAC or PVT outcomes. Higher AUDIT-C predicted lower SI despite no association with BrAC or PVT lapse rate. Overall, our results suggest a discrepancy between self-reported intoxication and PVT performance. A prior study suggests that, within the normal-to-moderate range, differences in drinking tendency do not clearly affect vigilance function [35]. Consistent with our eligibility criterion (AUDIT ≤ 14), our results likewise indicate that, within this range, drinking tendency influence SI more than vigilance. The discrepancy between subjective perception and performance provides insight into the mechanisms underlying risk perception and overconfidence during drinking. Contrary to our initial hypothesis, drinking tendency affected neither BrAC nor vigilance but only SI, implying that routine drinking within the normal range may be insufficient to modify metabolic rate or behavioral tolerance and instead primarily alters self-assessment. In the 40-min model, a 1-point higher AUDIT-C (i.e., a one-category shift in drinking frequency or quantity) was associated with a 4.5-point lower SI, comparable to the SI difference associated with approximately 0.031 mg/L lower BrAC. Based on the average BrAC decline rate during the decline phase (40–100 min; − 0.00152 mg/L/min), this corresponds to roughly 20 min of the average BrAC decline (95% CI 8.7–48.3). This comparison is descriptive rather than causal. Nevertheless, it suggests that even modest differences in drinking tendency may translate into meaningful differences in perceived sobriety, potentially affecting decisions such as continued drinking or returning home. Field studies in bar settings have consistently reported that individuals with higher AUDIT-C scores show higher BrAC and are more likely to consume driving-impairing amounts, even when they are designated drivers [36, 37]. The dose in this study was 0.5 g/kg, a moderate amount. To ensure safety, it was administered over a period of 30 min, with water intake every 10 min. Compared with the high doses of 0.6 g/kg or more frequently used in previous studies, the peak BrAC was lower. Some studies have suggested that significant differences in psychomotor function between the two genotype groups were detected only after high-dose alcohol administration exceeding 0.5 g/kg [10, 13, 35]. Although vigilance impairment was detected in the Matsuura et al. study, the moderating roles of genotype and drinking tendency may not have been sufficiently evident. Even considering these findings, our results indicate that BrAC, which varied by ALDH2 genotype, serves as a common basis for both SI and vigilance performance, whereas drinking tendency was related only to SI. This pattern may help explain discrepancies between psychological states and behavioral performance under the influence of alcohol. In this context, the SEM results indicated that the salience of predictors differed across outcomes and time points. The association of AUDIT-C with SI persisted at 100 min, and ALDH2 genotype remained indirectly linked to SI, whereas none of the hypothesized factors were significantly associated with vigilance at that time point. For BrAC, association with ALDH2 genotype association was stronger at 40 min but weakened by 100 min. Limitations This study has several limitations. Participants were limited to young, healthy adults of East Asian ancestry, so caution is needed when generalizing the results to populations that differ in age, ethnicity, or drinking patterns. Because the observation period was limited to 100 min, the present study could not evaluate the complete recovery phase as BrAC continued to decline toward zero. In addition, measurements obtained at later time points may be influenced not only by declining alcohol effects but also by fatigue or the repeated-testing effects inherent in the protocol. A longer observation period with an optimized measurement schedule may help clarify whether the dissociation between SI and vigilance performance persists or changes during the further decline in BrAC. SEM was conducted at only two representative time points and therefore did not capture the full dynamics of change. Because of power constraints, we did not model complex interactions or include additional covariates. Therefore, these path models should be interpreted as hypothesis-driven summaries of structured associations rather than as confirmatory evidence for causal mediation. In addition, we did not adjust for variables that may affect PVT performance, such as sleepiness or fatigue. Finally, the functional assessment in this study was primarily centered on vigilance using the PVT, because the primary aim was to repeatedly assess this specific domain during acute alcohol exposure using a brief and well-established task. Therefore, other cognitive functions relevant to real-life situations, such as inhibitory control, divided attention, or more complex task performance, were not evaluated in the present analysis. Although sex was included as a covariate in the path model, the present study was not specifically designed to examine sex differences. While our model identified a significant association between sex and BrAC during the decline phase, no significant direct paths from sex to SI or PVT lapse rate were observed. In addition, factors that may contribute to sex-related variation in alcohol responses, such as body composition, body water distribution, and menstrual cycle status, were not controlled. Therefore, these sex-related findings should be interpreted with caution and should not be taken as evidence for specific physiological mechanisms. Future studies should target broader age ranges, more diverse genotypes (particularly including ADH1B*1/*1), or clinical samples, and use multiple alcohol doses and more complex cognitive tasks to characterize gene–environment interactions more precisely. Conclusions This study comprehensively examined variation in BrAC, SI, and vigilance task performance during alcohol intoxication associated with ADH1B/ALDH2 genotypes and AUDIT-C score within a single modeling framework. The results indicate that BrAC was related to both SI and impaired vigilance function; conversely, drinking tendency was not associated with BrAC itself but only was related to self-reported SI and was not significantly associated with vigilance function. This suggests that while drinking tendency may reinforce the sensation of "not being drunk," the actual decline in cognitive performance may progress to the same degree regardless of drinking tendency. Furthermore, the ALDH2 genotype had a significant association with BrAC, revealing that differences in metabolic capacity may be indirectly linked to subjective and behavioral outcomes through BrAC. These findings underscore the importance of considering subjective and objective measures when interpreting the psychological and behavioral effects of alcohol. They also highlight how genetic factors and drinking tendency jointly shape these outcomes. Supplementary Information Supplementary Material 1. Acknowledgements We are grateful to all participants who took part in this study. We thank Nurse Yachiyo Higashi for her support in medical screening and safety management, and the members of the Higuchi Laboratory, Kyushu University, for their assistance during the study. English language editing was provided by Editage (www.editage.jp). Abbreviations Alcohol dehydrogenase 1B Aldehyde dehydrogenase 2 Analysis of variance Alcohol use disorder Alcohol Use Disorders Identification Test Alcohol Use Disorders Identification Test—Consumption Blood alcohol concentration Breath alcohol concentration Comparative fit index Confidence interval Heavy drinkers Light drinkers Partial eta squared Psychomotor Vigilance Task Root mean square error of approximation Reciprocal reaction time Standard error Structural equation modeling Subjective intoxication Standardized root mean square residual Tucker–Lewis index Visual analog scale Standardized path coefficient Authors’ contributions MM, YKM, and SH conceived the study and designed the methodology with NS, TN, YX, TO, NK, and YNM. MM, NS, YY, MH, and EN collected the data. Medical professionals (Dr. HT and Nurse YA) provided medical supervision, assisted with physiological measurements, and monitored participants throughout the experimental sessions. MM, NS, MH, and YY performed the formal analyses and prepared the figures. MM drafted the manuscript, and TN, SH, and TK critically revised it for important intellectual content. SH supervised the study and secured funding. All authors approved the final version and agreed to be accountable for all aspects of the work. Funding This study was funded by Honda R&D Co., Ltd. The sponsor reviewed and approved the study protocol but had no access to the raw data and did not influence the statistical analyses or the decision to submit this manuscript for publication. This work was also supported by the Tokuo Fujii Research Fund (Support for Article Processing Charge of International Academic Papers), Kyushu University. Data availability All datasets generated and/or analyzed in this study are available from the corresponding authors upon reasonable request. Declarations Ethics approval and consent to participate This study was approved by the Kyushu University Hospital Institutional Review Board (approval no. 24013). After the experimental procedures were explained, written informed consent was obtained from all participants before enrollment. The study was registered in the UMIN Clinical Trials Registry (UMIN000055964). Consent for publication Not applicable. Competing interests TN, YX, TO, and YNM are employees of Honda R&D Co., Ltd., and NK is an employee of Honda Research Institute Japan Co., Ltd. No other competing interests are declared. Honda R&D Co., Ltd. funded this study. The sponsor reviewed and approved the study protocol but had no access to the raw data and did not influence the statistical analyses or the decision to submit this manuscript for publication. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1. Data Availability Statement All datasets generated and/or analyzed in this study are available from the corresponding authors upon reasonable request.