Introduction

More than half of US adults currently use alcohol1. Alcohol use has negative immediate and long-term health consequences. A single episode of excessive drinking can raise the risk of injuries, violence, and alcohol poisoning, among other harmful outcomes2. Alcohol is considered a Group 1 carcinogen, causally linked to seven types of cancer3. Indeed, in January 2025, the Surgeon General of the United States released a report raising awareness of the links between cancer and alcohol consumption4. Any increase in alcohol use is associated with an increase in risk of illness and mortality. Long-term consequences of chronic alcohol use also include other health-related illnesses, such as high blood pressure, heart and liver disease, and Alcohol Use Disorder (AUD)2. In the US, 178,000 deaths per year are attributable to alcohol use and increases are projected. Both alcohol use and alcohol related deaths increased in the US during the COVID-19 pandemic5,6. Given these trends and the risks associated with any alcohol use, treatments that decrease alcohol use even in people not meeting criteria for alcohol use disorder, have the potential to improve health.

Glucagon-like peptide-1 receptor agonists (GLP-1RAs) are effective anti-glycemic and weight-loss medications with a strong safety record. Preclinical studies with various types of GLP-1RAs have demonstrated reductions in alcohol intake7,8,9,10,11,12,13,14,15. In humans, retrospective cohort studies have found people taking GLP-1RAs are less likely to develop alcohol use disorder (AUD)16 and less likely to be hospitalized for acute alcohol intoxication17. Participants taking GLP-1RAs semaglutide or tirzepatide in a retrospective survey study, reported reduced drinking and binge drinking episodes18. GLP-1RA exenatide administered once a week decreased heavy drinking days and total alcohol consumption, but only for a subset of participants with body mass index (BMI) over 3019,20. A secondary analysis of a randomized controlled trial (RCT) for smoking cessation found GLP-1RA dulaglutide decreased alcohol intake by 36%21. Another GLP-1RA, semaglutide, once weekly, reduced alcohol consumption in a challenge task and drinks per drinking day22. Overall, there is substantial preclinical evidence and mounting retrospective and prospective RCT evidence that GLP-1RAs could be effective for reducing alcohol consumption.

The mechanism by which GLP-1RAs reduce alcohol intake is not well understood. However, like all drugs with addictive potential, alcohol must reach the brain to exert its effects. Alcohol enters the blood in the upper intestine and crosses the blood brain barrier where it acts in the central nervous system as a gamma-aminobutyric acid A (GABA-A) receptor agonist, among other targets23,24. Therefore, possible mechanisms for the effects of GLP-1RA involve central nervous system effects or peripheral effects on gastric emptying, as alcohol must reach the upper intestine on its path to the brain. Survey data demonstrate people taking GLP-1RAs semaglutide and tirzepatide report reduced sedative and stimulative effects of alcohol18. A scrape and analysis of the social media platform reddit revealed reduced craving and increased negative effects of drinking alcohol18. Once weekly GLP-1RA semaglutide reduced alcohol craving22. Similarly, GLP1-RA exenatide once weekly for 26 weeks reduced brain response, as measured by functional magnetic resonance imaging (fMRI), to alcohol cues in the ventral tegmental area19. Although GLP-1RAs are known to have action in the brain, it is possible one mechanism by which GLP-1RAs reduce consumption of ingested drugs such as alcohol is, at least in part, through peripheral mechanisms.

While medications that reduce alcohol intake such as naltrexone and acamprosate have central nervous system action, disulfiram reduces alcohol intake through peripheral mechanisms. Semaglutide, liraglutide and other GLP-1RAs have been shown to slow gastric emptying25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43. Slowed gastric emptying can lead to a slower rise in blood alcohol, as alcohol is not well absorbed in the stomach. This can alter alcohol’s addictive potential as the pharmacokinetics of a substance is known to influence intake44,45,46,47,48,49,50. Here, we sought to test these hypotheses by recruiting a cohort with obesity19 either taking GLP-1 RAs or not and test breath alcohol content (BrAc) and subjective feelings after an alcohol challenge dose. At the time of study start (July 2023), the only available RCT evidence demonstrated a reduction in alcohol intake in response to GLP-1RAs only in people with obesity, so we focused our study on that population19. We observed that despite both groups consuming a dose of alcohol calculated to raise blood alcohol to 0.1 g/dl (approximately 0.08 g/dl BrAC)51,52participants in the GLP-1RA group showed a slowed rise in BrAC and altered the subjective effects of alcohol.

Results

Characteristics of participants

We recruited 24 individuals with obesity into this study (Table 1). The 24 individuals with obesity were taking a maintenance dose of GLP-1RA medication for weight loss (N = 14, GLP-1RA group) for at least 30 days, or not taking any medication for weight loss (N = 10, control group). Of the 14 in the GLP-1RA group, 1 individual was lost to follow up, and 3 were not invited based on being excluded for having gastric bypass surgery (Supplementary Fig. 1). The final sample (N = 20) consisted of majority white females, with a mean age of ~ 36 years and mean BMI of ~ 38. AUDIT scores were not different between controls with the GLP-1RA group (Range 1–6). Reported average drinking days were 1–2 in the past 30 days with an average of 1–2 drinks per use for both groups.

Table 1 Participant characteristics.
Full size table

The GLP-1RA group has reduced breath alcohol concentration (BrAC) and subjective feeling ratings during an alcohol administration session

To understand the effects of GLP-1RAs on alcohol absorption and its subjective effects, we compared BrAC and subjective feelings of intoxication between groups. Results of repeated measures linear mixed effect models indicated a significant interaction between group and time for BrAC (χ2(7) = 16.36, p = 0.022, f = 0.36). We observed an initial blunted response at time points 10 ( BrAC Control = 0.037, BrAC GLP-1RA = 0.021, t(113) = 3.21, p = 0.014), 15 ( BrAC Control = 0.028, BrAC GLP-1RA = 0.013, t(113) = 2.98, p = 0.028) and 20 min (BrAC Control = 0.037, BrAC GLP-1RA = 0.017, t(113) = 3.87, p = 0.001) in the GLP-1RA group as compared to the control group. From 35 min to 60 min, the GLP-1RA group’s BrAC was not significantly different than that of the control group. In addition, a main effect of sex (χ2(1) = 6.382, p = 0.012, f = 0.68) was observed with males having higher BrAC, overall. Differences between groups persisted across BrAC measurements taken in the recovery room (χ2(7) = 46.42, p < 0.001, f = 0.42; Supplementary Fig. 2). Additionally, cumulative BrAC calculated using area under the curve (AUC) was reduced in the GLP-1RA group as compared to the control group ((F(1,14) = 9.36, p = 0.008, f = 0.82); inset, Fig. 1A). The reduction indicates that the GLP-1RA medication delayed the initial rise in BrAC levels and resulted in more time with lower BrAC during the drinking session.

Fig. 1
figure 1

Objective and subjective effects of GLP-1RAs on alcohol intoxication during an alcohol administration session. (A) Breath alcohol concentration (BrAC) across drinking sessions. The x-axis displays timepoints (minutes), and the y-axis represents BrAC values (g/dL) during the drinking session. The inset shows the area under the curve reported during the 60 min. There is a decrease in BrAC at minutes 10–20 and decreased area under the curve in the GLP-1RA group. (B) Subjective feeling ratings -VAS (“How drunk do you feel?“) across timepoints in the drinking session. The inset shows the area under the curve reported during the 60 min. Mirroring the BrAC results, subjective feeling was also decreased in the GLP-1RA group. * p value < 0.05, # p = 0.06. Error bars represent standard errors. Group: Black – control, Gold – GLP-1RAs.

Full size image

The GLP-1RA group reported lower subjective feelings of intoxication during the early timepoints of the session, as indicated by reduced ratings on the Visual Analog Scale (VAS) for “How drunk do you feel?“. A significant interaction between group and time (χ2(7) = 18.86, p = 0.009, f = 0.39) was found in the linear mixed effects model, revealing that the GLP-1RA group consistently rated their intoxication lower across multiple timepoints compared to the control group (Fig. 1B). There were no statistically significant differences in post-hoc testing. Additionally, we calculated AUC for ratings across timepoints. The AUC was reduced in the GLP-1RA group as compared to the control group ((F(1,14) = 3.98 p = 0.06, f = 0.55); inset, Fig. 1B). This aligns with the objective BrAC measurements, demonstrating consistency between subjective experiences and physiological effects. BrAC was associated with subjective feeling ratings in all participants regardless of group (χ2=114.4, p < 0.001, f = 0.91). One participant appeared to be an outlier and rated their subjective drunkenness as 0 at all timepoints. All analyses on subjective feeling were run again and no changes in statistical significance were observed.

Effect of GLP-1RAs on appetite and alcohol craving before and after an alcohol administration session

GLP-1RAs are known to reduce appetite and have been shown to reduce alcohol craving. Drinking alcohol increased appetite score in the control group, but not in the GLP-1RA group (Group x Time interaction (χ2(1) = 5.9106, p = 0.015, f = 0.57; Fig. 2A; Supplemental Table 2), indicating that the pattern of change in appetite before and after the drinking session differed between groups. Post hoc comparisons revealed a significant increase in appetite scores in the control group from pre- to post-alcohol administration (B = 23.27, t(18) = 5.0, p = 0.0005), while the GLP-1RA group showed no statistically significant change (B = 7.30, t(18) = 1.57, p = 0.41), however the GLP-1 group started with a slightly, but not significantly, higher baseline.

Fig. 2
figure 2

Appetite score and craving score before and after an alcohol administration session. (A) Average appetite score between groups across the drinking session. The x-axis represents timepoint and the y-axis represents appetite scores. Appetite score increased on average for both groups. (B) Average AUQ score between groups across timepoints. The x-axis represents time point, and the y-axis displays the craving score. Error bars represent observed standard errors. Craving score increased on average for both groups. Group: Black – control, Gold – GLP-1RAs.​ AUQ: Alcohol Urges Questionnaire. *p < 0.05.

Full size image

Next, we investigated the effects of GLP-1RAs on alcohol craving, measured before and after alcohol administration using AUQ. Craving scores were significantly lower in the GLP-1RA group compared to the control group, as evidenced by a significant main effect of group (χ2(1) = 4.71, p = 0.03, f = 0.58; Fig. 2B, Supplemental Table 2). Additionally, craving scores increased after alcohol consumption across both groups, as shown by a significant main effect of time (χ2(1) = 6.15, p = 0.013, f = 0.57, Fig. 2B). However, no statistically significant interaction of group by time was observed, indicating a lack of evidence that GLP-1RAs differentially modulate the increase in craving across the drinking session.

Effect of GLP1-RAs on blood glucose and nausea ratings before and after an alcohol administration session

Given GLP-1RAs are effective anti-glycemic medications, we examined whether alcohol consumption resulted in differential effects on blood sugar between groups. Blood glucose levels were slightly but not significantly higher in the control group (χ2(1) = 2.84, p = 0.09, f = 0.45; Fig. 3A, Supplemental Table 2). Blood glucose for both groups changed over time, with a significant main effect of time point (χ2(1) = 15.09, p = 0.0005, f = 0.63; Supplemental Table 2, Fig. 3A), but a significant interaction between time and group was not observed.

Fig. 3
figure 3

Blood glucose levels and nausea ratings before and after an alcohol administration session. (A) Average blood glucose levels between groups across the drinking session. The x-axis displays pre- and post-drinking timepoints, and the y-axis represents blood glucose levels. Blood glucose did not differ across groups. (B) Average nausea ratings between groups across the drinking session. The x-axis represents pre-, during, and post-drinking timepoints and the y-axis displays nausea scores. Nausea did not differ across groups. Error bars represent standard errors. Group: Black – control, Gold – GLP-1RA.

Full size image

To rule out the possibility that the reduced subjective feelings of intoxication in the GLP-1RA group were due to nausea, we assessed nausea levels using a VAS (“How nauseous do you feel at this time?”) right after each drink during the alcohol administration session. For nausea, a significant main effect of time was observed (χ2(3) = 14.79, p = 0.002, f = 0.54; Fig. 3B, Supplemental Table 2), suggesting that nausea levels increased over the drinking session in both groups independently of group. Neither the Group effect nor an interaction between group and time reached statistical significance, indicating that the GLP-1RA group did not experience significantly different levels of nausea compared to the control group.

Discussion

Here, we report that the acute physiological and subjective effects of alcohol differ in people with obesity taking a GLP-1RA from people with obesity not on a GLP-1RA medication. Specifically, after a challenge dose of alcohol, participants in the GLP-1RA group have a delayed rise in BrAC and a similar delayed rise in the subjective effects of alcohol. This difference in subjective effects was not explained by an increase in nausea in the GLP-1RA group. Participants in the GLP-1RA group reported lower craving of alcohol overall, although craving did rise over the course of the drinking session, as in the control group. This is consistent with previous studies reporting reduced alcohol craving after GLP-1RA semaglutide administration22.

GLP1-RAs have been shown to have action in the central nervous system that mediates their weight loss and appetite suppressant effects53,54. Accordingly, studies in rodents have demonstrated the ventral tegmental area and nucleus accumbens mediate the effects of GLP-1RAs on alcohol intake11,12,55. Fluorescently labeled semaglutide, one of the GLP1-RAs used in our study population, has been shown to bind in the nucleus accumbens10. In humans, 26 weeks of exenatide reduced brain response to alcohol cues in the ventral tegmental area19. However, while central pharmacodynamics is an important factor influencing drug intake, peripheral pharmacokinetics can also have a profound effect on intake and even neural response to the same drug56,57. This is most dramatically illustrated through differences in the intake patterns of the same drug with different routes of administration45,46,47,48,49,50,58known as the “rate hypothesis” of addiction49,59,60.

One mechanism by which the rate of absorption of alcohol could be altered is through peripheral effects on gastric emptying in people taking GLP-1RAs. At least seventeen GLP-1RA clinical trials documented a delay in gastric emptying25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41. Furthermore, there have been multiple case reports of retained stomach contents or regurgitation following approved fasting protocols for patients on GLP-1RAs61,62,63,64,65,66,67,68,69,70,71. Together, these data indicate gastric emptying is slowed in people taking GLP1-RAs29,42,72,73. Alcohol is not readily absorbed in the stomach, but rather in the upper intestine74 and slowed gastric emptying would therefore slow the effects of alcohol. Here, we find support for this hypothesis, in that our participants taking GLP1-RAs had a delayed rise in BrAC up to 20 min post alcohol consumption. This corresponded with a delayed increase in subjective feeling after alcohol consumption. We did not observe a difference in blood glucose changes in response to our alcohol challenge drink, which contained sugars in both the vodka and mixers. This could be taken as evidence that gastric emptying was similar for both groups, however, we obtained blood glucose levels 1 and 3 h post drinking. BrAC levels in the GLP-RA group were not different from controls by 1 h post drinking. Later studies should sample blood sugars more frequently to determine if the blood sugars and BrAC effects mirror each other.

Limitations

While participants were matched between the GLP-1RA and control groups, our sample size was small. We have therefore included a table of all our statistical analyses with effect sizes as a reference for power analyses for future studies. While most of our participants were taking semaglutide, some were on liraglutide (n = 2) and tirzepatide (n = 2), and some received their medications through compounding pharmacies. We were not able to compare differences between drug type or source, though we have previously reported large reductions in subjective effects of alcohol in people taking tirzepatide18. This was not a randomized control trial, rather we recruited participants already taking these medications, additionally we did not use a placebo for alcohol administration. Finally, while this study cannot definitively differentiate between central nervous system and peripheral effects of GLP-1RAs on alcohol intake, it does provide evidence that there is a role for peripheral mechanisms, which should be examined in future studies.

Conclusions

With recent recommendations towards reductions in drinks as effective goals for individuals who use alcohol75,76,77these findings facilitate the consideration of GLP-1RAs for reducing alcohol use. While this study is small, we provide essential preliminary data (e.g., effect sizes) for the design and development of larger randomized control trials testing the effectiveness of GLP-1RAs on reductions in alcohol use.

Methods

Participants

Data were collected between July 2023 and May 2024. Participants (N = 20) were recruited from Roanoke, VA, and surrounding areas using flyers, the internet, and word-of-mouth referrals. To be eligible, participants had to occasionally drink alcohol, be 21 years of age or older, and have a BMI of 30 or greater. For the experimental group, participants had to be on a maintenance dose (at least 4 weeks, Supplemental Table 1) of a GLP-1RA medication for weight loss. Exclusion criteria were a history of bariatric surgery, alcohol use disorder identification test (AUDIT) greater than 16, hemoglobin A1c measurement greater than 9, a history of seizure disorder or traumatic brain injury, on specific medications that interfere with alcohol metabolism such as certain antibiotics (reviewed on a case-by-case basis by study physician), or currently pregnant or lactating. All procedures were reviewed and approved by the Virginia Tech Institutional Review Board (#23–460).

Measurement details

Subjective feeling rating

Participants were asked to rate their feeling of intoxication on a scale of 0–10, with 0 rated as not being drunk at all and 10 as the most drunk they have ever felt, by asking “How drunk do you feel right now?”

Appetite score

Participants were asked to indicate their current appetite using visual analog scales (VAS) as described by Flint and colleagues78. The VAS is a visual scale with each end representing the most positive and most negative ratings of hunger, satiety, fullness, and prospective food consumption. Appetite score was calculated as the average of each parameter.

Alcohol use disorder identification test (AUDIT)

AUDIT is a 10-item Core questionnaire to identify hazardous drinkers (whose drinking increases their risk of alcohol-related problems, though alcohol-associated harm has not yet occurred); harmful drinkers (who have had recent physical or mental harm from their drinking, but who are not alcohol-dependent); and people with alcohol dependence79.

Alcohol urges questionnaire

Craving was assessed using the Alcohol Urge Questionnaire (AUQ)80. Participants rate their feelings and thoughts about drinking in an 8-statement questionnaire at the time the measure is administered. Each question is a 7-item Likert scale ranging from “strongly disagree” to “strongly agree.” Craving score was calculated by averaging each item’s rating, with higher scores indicating greater craving.

Nausea rating

Participants rated their nausea on VAS anchored at No Nausea and Unbearable Nausea and prompted with “At this moment, how nauseous do you feel?”

Timeline follow-back (TLFB)

This interview is the standard for the quantitative assessment of alcohol use. Research staff provide participants with calendar-based memory cues to help them construct chronological reports of their recent alcohol use81. We had missing participant data (n = 2 (GLP-1RA group) and n = 3 (Control group) for TLFB which is denoted by an asterisk in Table 1.

The biphasic alcohol effects scale (BAES)

BAES is a self-report, unipolar adjective rating scale that is designed to measure both stimulant and sedative effects of alcohol82. The scale consists of fourteen items that are rated on a scale of 0 (not at all) to 10 (extremely). Participants rated the extent to which drinking alcohol produced these feelings at present. Scores are calculated as the sum of the items.

Additional questionnaires

Participants reported on the good effects, the bad effects, taste and liking of alcohol during the alcohol administration session. Specifically, the VAS statements were: (a) good effects: At this moment, are you experiencing good effects from the drink? (0 = No Good Effect, 50 = Neutral, and 100 = Strong Good Effects) (b) bad effects: At this moment, are you experiencing bad effects from the drink? (0 = No Bad Effect, 50 = Neutral, and 100 = Strong Bad Effects) (c) At this moment, how does the drink taste? (0 = Bad, 50 = Neutral, and 100 = Good), and (d) At this moment, my liking for the drink served to me is? (0 = strong disliking, 50 = neither like nor dislike, and 100 = strong liking).

Blood glucose

First, the participant’s finger was cleansed with an alcohol pad, allowed to dry, then punctured with a disposable lancet. Blood was then collected and glucose assessed using a Contour Next EZ Blood Glucose Meter and Counter Test Strips distributed by Ascensia Diabetes Care US, Inc (New Jersey, USA).

Breath alcohol

First, participants rinsed their mouths with water. Next, they were instructed to take a deep breath and blow the whole breath out into the breath alcohol analyzer (ALCO-SENSOR III, Intoximeters, Missouri, USA).

Drink stimulus

The alcoholic beverage content was calculated using Formula 14 from Brick83 for each participant to reach 0.1 g/dl blood alcohol content (0.08 g/dl BrAC)51,52.

$${\text{g}}={\text{BA}}{{\text{C}}_{{\text{target}}}}+[{\beta _{1 - n}} \times ({t_{\text{s}}}+{t_{\text{p}}})] \times \sum {{V_{\text{d}}}/B{l_{{{\text{H}}_2}{\text{O}}}}.}$$

This formula allows for the calculation of the amount of alcohol in any beverage and estimates the blood alcohol concentration in a range of subjects with individual characteristics and drinking patterns83. This formula has been widely used for alcohol administration studies, the formula accounts for (1) time spent drinking, (2) time post-consumption, (3) total body water, (4) sex, (5) age, (6) weight, (7) height, and (8) estimated alcohol elimination rate. Each drink was mixed at a 1:3 ratio of Tito’s Vodka to Juice (cranberry or orange). A target of 0.1 g/dl was utilized to allow for a 20% variance between blood alcohol content (BAC) and BrAC51,52 .

Procedures

Prior to enrollment, all participants completed an online pre-screening survey, fingerstick test for HbA1c, urine pregnancy test, as well as height and weight measurements to ensure eligibility. The study was approved by the VT IRB, and informed consent was obtained from all participants prior to their participation in the study. All methods were performed in accordance with relevant guidelines.

Session 1

After informed consent was obtained, participants completed a Timeline Follow Back (TLFB) to measure drinking in the past 30 days, Patient Health Questionnaire (PHQ-9), and a medical interview which asked about current and past medication history and other medical history to ensure safe participation in the study, as reviewed by a study physician.

Session 2

A day prior to the drinking session, participants were reminded to come in without eating anything. All participant sessions started between 8 and 10 AM. Once participants arrived, they were given a snack (Clif-Bar, Indianapolis, USA) to standardize caloric intake and ensure a similar gastric content prior to alcohol administration. Before proceeding to the Research Bar, the participants’ blood pressure, pulse, breath alcohol concentration and blood glucose level were measured. The participants then completed questionnaires related to drinking behaviors as described in the questionnaire section above.

Ninety minutes after consuming the snack, the participant was moved to the Research Bar. The drinking period of the session lasted one hour and was divided into three phases. In each phase the participant was served the prepared alcoholic beverage to be consumed within 10 min. Following this, participants were given a 10-minute break, completed subjective questionnaires about their current state and experience, including Visual Analogue Scales (VAS) as described above. During the breaks, research personnel also measured BrAC. Before measuring BrAC, participants were informed to rinse their mouth with water by swishing thoroughly, to avoid artificially high readings due to alcohol vapors in the mouth. This was repeated for three drinks over one hour.

Upon completion of the drinking period, participants were moved to a recovery room where they remained for 4 h to allow the alcohol consumed to be metabolized to 0.02 g/dL BrAC, the criterion for release. Researchers measured BrAC every 30 min, fingerstick blood glucose levels were measured immediately following the drinking session and at hour 3. At hour 3 participants completed the subjective questionnaires. After 4 h in the recovery room and a BrAC < 0.02 g/dL, the participant was released with Study Physician approval.

Statistical methods

Demographic and anthropomorphic variables were characterized with means and standard deviations for continuous variables and counts and percentages for categorical variables. They were additionally compared across the GLP-1RA and control groups to check for any differences. Welch’s two-sample t-test and Fisher’s exact tests were used for the continuous and categorical variables respectively.

Linear mixed-effects models were estimated using the lme484,85 package to assess the relationship between the dependent variable and the main effects of Group (GLP1-RAs vs. control), Time (Timepoints during the alcohol drinking session), and their interaction. If the interaction term was not significant, it was excluded from the model and only the main effects were reported. Age, AUDIT-C score, body mass index (BMI), and sex were also included in the models. To account for repeated measures within participants, a random intercept for each participant was included. Six models were fit with BrAC, perceived drunkenness, alcohol craving, appetite, blood glucose, and nausea scores as dependent variables, respectively (Supplemental Table 2).

For both BrAC and subjective feelings of drunkenness, area under the curve (AUC) was calculated using the trapezoidal rule (with x as time in minutes and y as the outcome). These two values were compared across groups using Welch’s t-test. To control for additional confounders, linear regression models were fit with each AUC as the dependent variable and group as the predictor of interest along with age, AUDIT-C score, body mass index (BMI), and sex.

Type III tests of fixed effects were conducted using the car package86 to evaluate the statistical significance of main effects and interaction terms. Cohen’s f effect sizes were calculated via the effect size package87 to quantify the magnitude of observed effects. Post hoc pairwise comparisons were performed for significant interaction effects using estimated marginal means with Sidak correction to control for multiple comparisons. These analyses were implemented using the emmeans package88and adjusted p-values are reported accordingly. All analyses were conducted using R (v4.3.0), with statistical significance defined as p < 0.05.