Use of GLP‐1 receptor agonists and subsequent risk of alcohol‐related events. A nationwide register‐based cohort and self‐controlled case series study

Abstract

Animal studies have related glucagon‐like peptide 1 receptor agonists (GLP‐1) to lower alcohol intake. We examined whether GLP‐1 was associated with risk of alcohol‐related events in a nationwide cohort study and a self‐controlled case series analysis including all new users of GLP1 (n = 38 454) and dipeptidyl peptidase 4 inhibitors (DPP4) (n = 49 222) in Denmark 2009–2017. They were followed for hospital contacts with alcohol use disorder or purchase of drugs for treatment of alcohol dependence in nationwide registers from 2009 to 2018. Associations were examined using Cox proportional hazard and conditional Poisson regression. During follow‐up of median 4.1 years, 649 (0.7%) of participants were registered with an alcohol‐related event. Initiation of GLP‐1 treatment was associated with lower risk of an alcohol‐related event (Hazard ratio = 0.46 (95%CI: 0.24–0.86) compared with initiation of DPP4 during the first 3 months of follow‐up. Self‐controlled analysis showed the highest risk of alcohol‐related events in the 3‐month pretreatment period (incidence rate ratio [IRR] = 1.25 (1.00–1.58)), whereas the risk was lowest in the first 3‐month treatment period (IRR = 0.74 (0.56–0.97). In conclusion, compared with DPP4 users, individuals who start treatment with GLP‐1 had lower incidence of alcohol‐related events both in cohort and self‐controlled analyses. Thus, there might be a transient preventive effect of GLP1 on alcohol‐related events the first months after treatment initiation.

1. INTRODUCTION

Alcohol use disorders are some of the most prevalent mental disorders worldwide affecting 8.6% of men and 1.7% of women ¹ and with a lifetime prevalence of around 30%. ² , ³ Alcohol use disorders are associated with markedly reduced life expectancy, ⁴ a higher risk of concomitant chronic diseases and injuries as well as social deprivation including family disruption, unemployment and criminal convictions. ⁵

The treatment goal for most patients with alcohol use disorders is abstinence, and the way to achieve this often includes a combination of counselling (including motivational interviewing and encouraging patients to take responsibility for change) and medication. ⁶ To date, only few drugs such as disulfiram, acamprosate, nalmefene and naltrexone have been approved as treatment of alcohol use disorders. However, the effect of the drugs varies and a recent population‐based study found that disulfiram did not change the risk of hospitalization for alcohol use disorder, and patients using acamprosate actually have a higher risk of hospitalization for an alcohol use disorder. ⁷ Only naltrexone seems to be effective in reducing heavy drinking but less efficacious in promoting abstinence. ⁸

In recent years, evidence from animal studies have shown that intestinal hormones such as ghrelin and glucagon‐like‐peptide 1 (GLP‐1), which are well‐known as regulators of appetite and food intake, might also affect alcohol‐related behaviours. ⁹ GLP‐1 is a hormone secreted by the intestine in response to food intake and together with other gut hormones; it signals to the pancreas increasing insulin secretion and reducing glucagon secretion. ¹⁰ GLP‐1 receptor agonists have therefore been developed for the treatment of type 2 diabetes. Apart from the peripheral effects, GLP‐1 receptor agonists have also been found to function in the central nervous system. ¹¹ Here, GLP‐1 receptor agonists have been found to decrease consummatory behaviour by inhibition and modulation of dopaminergic neurons in the ventral tegmental area and the nucleus accumbens which are central areas involved in the mesolimbic reward system. ¹² , ¹³ , ¹⁴ To date, the GLP‐1 receptor agonists liraglutide and semaglutide are approved for treating obesity. ¹⁵

Studies on the effect of GLP‐1 receptor agonists on alcohol addiction have reported reduction in alcohol drinking behaviour in rodents and monkeys, ⁹ , ¹⁶ and polymorphisms in the GLP‐1 receptor gene have been associated with alcohol use disorders. ¹⁷ However, to date, no published studies on humans have examined the potential association between GLP‐1 receptor agonists and alcohol use disorders. Thus, the aim of this study was to examine the association between use of GLP‐1 receptor agonists and the risk of subsequent alcohol‐related events in Danish adults.

2. MATERIAL AND METHODS

2.1. Study design

We conducted a new‐user, active‐comparator cohort study using information on the entire Danish population from nationwide registers. ¹⁸ The study population was identified in the Danish Civil Registration System using the unique personal identification number (CPR) which enables accurate linkage of recorded individual‐level information in all registers. ¹⁹ To limit confounding by indication—that is, the underlying condition for which GLP‐1 receptor agonists is prescribed may render the individual more susceptible to alcohol‐related events—independently of the effect of GLP‐1 receptor agonists—we used dipeptidyl peptidase‐4 (DPP‐4) inhibitors as comparator group, as this antidiabetic drug has been approved nearly simultaneously with GLP‐1 receptor agonists. To further account for time‐invariant confounders, we secondly included a self‐controlled case‐series design. ²⁰ In this design, the potential association between treatment with GLP‐1 receptor agonists or DPP‐4 inhibitors and an alcohol‐related event was investigated by comparing different time periods for individuals who experienced the outcome under study. The study was approved by the Regional Data Protection Agency. As the data only included register data, informed consent from participants was not required.

2.2. Study population

All refills of prescriptions for medication were identified in the Danish National Prescription Registry, which contains information on all prescribed drugs dispensed at pharmacies since 1995 including date of prescription redemption and anatomical therapeutic chemical (ATC) codes. ²¹

A cohort was created of new users of GLP‐1 receptor agonists and DPP‐4 inhibitors which were defined as a first‐time redeemed prescription of drugs with ATC codes A10BJ or A10BH, respectively. In Denmark, the first GLP‐1 receptor agonists and DPP4 inhibitors were approved for antidiabetic treatment in 2006 and 2007, respectively, but the number of users was not stable until 2009 (supporting information Table S1). Consequently, we did not include those who had initiated treatment before 2009 as they might represent a selected group of patients. ²² The date of first prescription of GLP‐1 receptor agonists or DPP‐4 inhibitors was used as the index date and we excluded the following: (1) individuals with prescriptions of both GLP‐1 receptor agonists and DPP‐4 inhibitors on the index date (n = 14), (2) individuals with prescriptions for GLP‐1 receptor agonists or DPP‐4 inhibitors within 365 days before the index date (n = 3525) and (3) individuals who immigrated or died before the index date or age< 15 years (n = 3556).

2.3. Outcomes: Alcohol‐related events

The outcome of interest was alcohol‐related events measured by the following: (1) Hospital contacts with a main diagnosis of alcohol use disorders (international classification of diseases [ICD]‐10 code DF10) in the Danish National Patient Registry, (2) registered treatments for alcoholism in the National Registry of Alcohol Treatment or (3) purchase of the benzodiazepine chlordiazepoxide (ATC code N05BA02), which is used for alcohol withdrawal syndrome or purchase of a medication against alcohol dependence (ATC code N07BB), registered in the Danish National Prescription Registry.

The Danish National Patient Registry includes information on physicians' diagnoses for all inpatient, outpatient and emergency contacts at Danish hospitals since 1995, whereas the National Registry of Alcohol Treatment includes data on public treatment for alcoholism treatment since 2006. ²³ , ²⁴ For individuals with more than one outpatient hospital contact, events recorded within 180 days of each other were categorized as the same, whereas emergency and inpatient events recorded within 15 days of each other were categorized as the same (e.g. due to discharge from psychiatric hospital to go to somatic hospital). If an individual had more than one purchase of an alcohol‐related medication, the following purchases within 180 days of each were assumed to belong to the same treatment event. For details, see supporting information Table S2 for included data and supporting information Table S3 for the distribution of alcohol‐related events on data source for the different populations.

2.4. Covariables

We selected several factors assumed to be associated with selection of patients for GLP‐1 receptor agonists or DPP‐4 inhibitors and risk of alcohol‐related events as covariates. From the Danish National Patient Registry and Danish National Prescription Registry, we obtained information on history of heart disease, stroke, diabetes, obesity, schizophrenia, major depression, alcohol and substance abuse. From the Danish Civil Registration System, we obtained information on sex, age, marital status, country of origin and the region of place of living. Data on highest achieved educational level were obtained from the Population's Education register. Educational level was categorized into four groups (low: primary school; middle: high school, vocational education [educations aimed towards manual work]; high: higher education, and higher advanced education and missing [patients born before 1920, older immigrants, or still receiving education]). Codes used for definitions of covariates are given in the supporting information Table S4.

2.5. Statistical analysis

Firstly, we used Cox proportional hazard regression to calculate hazard ratio (HR) and 95% confidence intervals (CI) for the association between use of GLP‐1 receptor agonists and alcohol‐related events, compared with use of DPP‐4 inhibitors. In an intention‐to‐treat (ITT) analysis, cohort members were followed from the index date until first registration of an alcohol‐related event, emigration, death, switching to the other study drug or end of follow‐up (31 December 2018), whichever came first. Because it cannot necessarily be assumed that the potential effect of GLP‐1 receptor agonists on alcohol consumption may be long‐lasting beyond treatment cessation, we included an “as‐treated” analysis focusing the exposure only on time of treatment. In this analysis, we examined the association between continuous treatment and the outcome and followed individuals from the index date until first registration with an alcohol‐related event, emigration, death, switching to the other study drug or reaching the end of their first treatment period. Treatment duration was calculated based on the number of daily defined doses filled, and we added 90 days of additional follow‐up to prescriptions to account for irregular use. A gap between prescription fills exceeding 90 days was considered as end of the current treatment period. To avoid immortal time bias, we added 90 days of follow‐up to the duration of the last treatment episode. ²⁵ To explore if the risk varied during follow‐up, we examined the incidence rates for 0–90 and 90–365 days and later after the index date by splitting the data on follow‐up time. We accounted for potential confounders by multiple adjustment for age, sex, marital status, education, country of origin, somatic and psychiatric comorbidity, diabetes and time since diabetes diagnosis/treatment at baseline. Since the risk of alcohol‐related events vary with sex, age and comorbid obesity, we explored potential interactions with these variables in stratified analyses and, if applicable, tested by including interaction terms using likelihood ratio tests. This showed no significant interactions when analyses were stratified by these covariates.

In the self‐controlled case series design, individuals who initiated GLP‐1 receptor agonist were followed for outcomes in five different time periods: (1) time when unexposed to GLP‐1 receptor agonists (non‐treatment period), (2) a 3‐month pre‐exposure period up to and including the date of first prescription of GLP‐1 receptor agonists, (3) the first 3 months following GLP‐1 agonist initiation (initial exposed time), (4) 4–12 months following GLP‐1 agonist initiating (late exposed time) and (5) the following 13–24 months (consolidated exposed time) after initiation. The pre‐exposure period was defined because of concerns about confounding by indication. We used Poisson regression conditional on the individual person to calculate the incidence of alcohol‐related events in the different study periods and generated incidence rate ratios (IRRs), with 95% confidence intervals, using the non‐treatment period as reference. Individuals who discontinued or died were censored 1 month after the last prescription of GLP‐1 receptor agonists or at time of death. Self‐controlled time series remove time constant confounders but are still sensitive towards time‐varying factors such as age which we adjusted for. ²⁶ The above analysis was repeated for individuals who initiated DPP‐4 inhibitors. All analyses were performed in STATA‐16.

3. RESULTS

We identified 41 737 new users of GLP‐1 receptor agonists and 50 108 new users of DPP4 inhibitors in the Danish National Prescription Registry between 2009 and 2017. From this population, 38 544 GLP‐1 receptor agonist users and 49 222 users of DPP4 inhibitors complied with the inclusion criteria. At the index date, GLP‐1 agonist users generally had lower mean age, were less often women and were higher educated and were more often obese compared with users of DPP4 inhibitors (Table 1). During follow‐up of a median of 4.1 years (interquartile range [IQR] 2.1–6.9), a total of 649 (2.3%) participants experienced an alcohol‐related event corresponding to an incidence rate (IR) of 16.3 (95% CI 15.1–17.7) cases per 10 000 person‐years. In the “as‐treated” analysis, the median follow‐up was 2.5 years (IQR: 1.6–4.9) for users of GLP‐1 receptor agonists and 1.4 (IQR:0.7–2.5) for DPP‐4 inhibitor use. The median number of prescriptions was 16 (IQR: 5–38) for GLP‐1 receptor agonists and 7 (IQR: 3–14) for DPP‐4 inhibitors.

TABLE 1.

Comparison of baseline characteristics between individuals treated and untreated with GLP1 receptor agonists (GLP‐1) and dipeptidyl peptidase‐4 inhibitors (DDP‐4)

	GLP‐1	DPP‐4
Total	38 454	49 222
Women	17 073 (42.0)	20 660 (44.4)
Mean age (SD)	57.8 (12.1)	65.1 (12.5)
Basic education	14 154 (36.8)	19 919 (40.5)
Highest education	5322 (13.8)	5659 (11.4)
Unmarried	7051 (18.3)	6636 (13.5)
Region
Capital	11 072 (30.9)	13 781 (28.0)
Zealand	7381 (19.2)	7505 (15.3)
Southern Denmark	7478 (19.5)	10 757 (21.9)
Mid Denmark	7967 (20.7)	10 555 (21.4)
Northern Denmark	3402 (8.9)	6059 (12.3)
Comorbidity
Heart disease	7034 (18.3)	9893 (20.1)
Stroke history	2376 (6.2)	4772 (9.6)
Obesity	15 484 (40.3)	11 446 (23.2)
Diabetes diagnosis	24 799 (64.5)	22 187 (45.1)
Time since diabetes onset (years)	7.4 (6.5)	7.1 (6.4)
Oral antidiabetics	35 851 (93.2)	47 752 (97.1)
Time since initiation antidiabetics (years)	7.8 (5.1)	6.7 (5.2)
Alcohol or substance abuse	2040 (5.3)	2602 (5.3)
Schizophrenia	896 (2.3)	1156 (2.3)
Major depression	1725 (4.5)	1821 (3.7)

Abbreviations: DPP‐4, dipeptidyl peptidase 4 inhibitors; GLP‐1, glucagon‐like peptide 1 receptor agonists.

3.1. Association of GLP‐1 receptor agonists with alcohol‐related events

In the multiple adjusted Cox proportional hazard regression model, GLP‐1 receptor agonist use was associated with a lower risk of a subsequent alcohol‐related event compared with DPP‐4 inhibitor use both in the time period closest to initiation (HR_{90 days after initiation} 0.46 [95% CI 0.24–0.86]) and after 1 year of follow‐up (HR_{365 + days after initiation} 0.62 [95% CI 0.45–0.85]) (Table 2). In the “as‐treated” analysis, the risk estimates were rather similar but less precise.

TABLE 2.

The association between use of GLP‐1‐receptor agonists (GLP‐1) and alcohol‐related episodes compared with use of DPP‐4 inhibitors (DPP‐4) in Danish adults aged 15 years or above in 2009 through 2018

Intention‐to‐treat analysis
Follow‐up time	0–90 days	90–365 days	0–365 days	365–4619 days
Number of outcomes
GLP‐1/DPP‐4	15/34	46/48	61/82	245/261
Crude Hazard Ratio (95% Confidence intervals)
GLP‐1 versus DPP‐4	0.56 (0.30–1.03)	1.16 (0.77–1.73)	0.91 (0.66–1.27)	0.89 (0.75–1.06)
Adjusted * Hazard Ratio (95% Confidence intervals)
GLP‐1 versus DPP‐4	0.46 (0.24–0.86)	0.98 (0.64–1.49)	0.76 (0.53–1.07)	0.72 (0.60–0.86)
As‐treated analysis
Follow‐up time	0–90 days	90–365 days	0–365 days	365–4484 days
Number of outcomes
GLP‐1/DPP‐4	15/34	38/37	53/71	86/84
Crude Hazard Ratio (95% Confidence intervals)
GLP‐1 versus DPP‐4	0.56 (0.30–1.03)	1.27 (0.85–2.00)	1.02 (0.74–1.47)	0.78 (0.57–1.06)
Adjusted * Hazard Ratio (95% Confidence intervals)
GLP‐1 versus DPP‐4	0.46 (0.24–0.86)	1.04 (0.64–1.07)	0.84 (0.57–1.24)	0.62 (0.45–0.85)

Abbreviations: DPP‐4, dipeptidyl peptidase 4 inhibitors; GLP‐1, glucagon‐like peptide 1 receptor agonists.

In the self‐controlled analysis, we compared time periods before and after initiation of GLP‐1 receptor agonist treatment. This showed that the IR of an alcohol‐related event was highest during the last 3 months before GLP‐1 receptor agonist initiation (pretreatment period IR 83.8 [95% CI 79.3–88.2] per 10 000 person‐years), while the lowest rate was found in the 3 months following GLP‐1 receptor agonist treatment initiation (initial treatment period, IR 46.2 [95%CI 40.6–52.9] per 10 000 person‐years) (supporting information Table S5). Thus, when compared with the non‐treatment period, the IRR was slightly higher (IRR 1.25 [95% CI 1.00–1.58]) in the pretreatment period and significantly lower (0.74 [95% CI 0.56–0.97]) in the initial treatment period. During the following years of continued treatment, the risk did not vary from the non‐treatment period (Figure 1). The self‐controlled analysis in users of DPP‐4 inhibitors showed that compared with the non‐treatment period, the IRR was highest in the 3 months preceding treatment initiation and lowest in the 3 months following initiation of treatment.

FIGURE 1.

Incidence rate ratios of alcohol‐related events in different periods of time within patients initiating treatment with glucagon‐like peptide 1 receptor agonists (GLP‐1) receptor agonists or dipeptidyl peptidase 4 inhibitors (DPP‐4) inhibitors

4. DISCUSSION

In this nationwide register‐based cohort study, we found that use of GLP‐1 receptor agonists was associated with a lower risk of a subsequent alcohol‐related event compared with use of DPP‐4 inhibitors after adjustment for covariates. In a self‐controlled analysis, initiation of GLP‐1 receptor agonist treatment was also associated with a lower risk of an alcohol‐related event compared with the non‐treatment period but only during the first 3 months after start of treatment.

4.1. Comparison with related studies

Available data on the effects of GLP‐1 receptor agonists on alcohol intake in patients with alcohol use disorder are extremely limited. ¹⁴ Data in abstract format reported that liraglutide treatment reduces alcohol intake in patients with diabetes. ²⁷ Thus, most evidence of an effect of GLP‐1 receptor agonists on alcohol consumption is based on rodent studies ¹⁶ and studies in velvet monkeys, ²⁸ all showing that GLP‐1 receptor agonists reduce intake of alcohol. The present study was based on the hypothesis that GLP‐1 receptor agonists attenuate the ability of alcohol to activate the mesolimbic dopamine system, and through this decreases the motivation for alcohol consumption. ²⁹ , ³⁰ At first, our findings seemed to support that initiation of GLP‐1 receptor agonist treatment is associated with fewer alcohol‐related events (which we used as proxy for a lowered alcohol intake) compared with initiation of DPP4 inhibitors. However, the self‐controlled design—which efficiently accounts for unmeasured between person confounding—showed that GLP‐1 agonist users had low rates of alcoholic events during the first 3 months after treatment start but similar rates the following months. Thus, it is possible that the effect that is either only observed in the initial treatment phase, that is, 3 months or is caused by GLP‐1 receptor agonist‐related adverse events is due to unspecified factors, for example, more frequent contact to doctors, healthy life style advice or other lifestyle choices during this specific period of time. Most previous experimental studies have examined potential acute effects of GLP‐1 receptor agonists, while evidence on chronic or subchronic use of GLP‐1 receptor agonists and alcohol use is sparse. ¹⁶ It is thus possible that only acute but not chronic use of GLP‐1 receptor agonists affect alcohol consumption. In human studies testing the effect of the GLP‐1 receptor agonist liraglutide on CNS activation in response to viewing food pictures or to chocolate milk in obese subject, liraglutide significantly reduced CNS activation compared with insulin during the first 12 weeks; however, the effect ceased to be significant after 12 weeks of treatment. ³¹ , ³² In our “as‐treated” analysis, risk estimates were similar to those in the main analysis, which could support that the effect of GLP‐1 receptor agonists is only seen when the treatment is initiated and not during continuous treatment. Common side effects of GLP‐1 receptor agonists after initiation include malaise and nausea, ³³ and it is also a possibility that these side effects could cause lower alcohol consumption during the first 3 months after initiation. In the self‐controlled analyses, results were similar for users of GLP‐1 receptor agonists and DPP‐4 inhibitors, suggesting that the effect of GLP‐1 may also be due to non‐drug‐related factors during the specific time period after initiating a new drug treatment.

4.2. Strengths and weaknesses

Our study benefits from a longitudinal design and the use of nationwide population‐based registers in Denmark, a country with free access to health care. Register linkage enabled us to minimize probable selection or recall bias as all exposures were recorded independently of the outcome. Information on drug exposures was relative complete and entered as time‐varying covariate in cohort analysis. Finally, we used different strategies to account for confounding by indication as suggested by others. ³⁴

This observational study has two main limitations. Firstly, we used patient registers to define alcohol‐related events as proxies for alcohol consumption. To test this approach, we performed a validation study comparing the register‐based definition with information on self‐reported data on alcohol consumption in a longitudinal study of 2725 middle‐aged men (Online‐only Validation study). This study showed that low self‐reported alcohol consumption was associated with fewer alcohol‐related events, which indicates that our outcome measure may reflect changes in heavy alcohol consumption. The other main limitation is confounding by indication. Although we used several approaches to address potential confounding, each method has limitations. In the cohort analysis, all data used for adjustment were register‐based and the information on education, comorbidities and at baseline might not fully reflect differences, for example, in relation to lifestyle and social relations between patients selected for GLP‐1 receptor agonist treatment or not. We brought DPP‐4 inhibitors into the analysis as active comparator ¹⁸ as we assumed that these individuals in many ways would be similar to those treated with GLP‐1 receptor agonists. However, this assumption was not fulfilled as users of this antidiabetic had another sociodemographic profile and health status than GLP‐1 receptor agonist users. As the mechanism of action of DPP‐4 inhibitors also is to increase levels of active GLP‐1, ³⁵ however, to a lesser extent than GLP‐1 receptor agonists, ³⁶ using DPP‐4 inhibitors as a comparison group may have attenuated our results.

The self‐controlled case‐series design, on the other hand, accounts for time‐invariant confounders, for example, personality traits and family or genetic disposition but time‐varying confounders may still influence the association. For example, the higher risk of outcomes during the pretreatment period might be due to excessive drinking which might contribute to metabolic dysregulation leading to antidiabetic treatment. Similar, the lower risk during the initial treatment period may partly have been because of the support from health care facilities.

5. CONCLUSION

In conclusion, the incidence of alcohol‐related events was lower in individuals who started treatment with GLP‐1 receptor agonists than in individuals who started DPP4 inhibitors, but self‐controlled analyses indicated that this might be explained by confounding by indication. However, the analysis suggested that there might be a transient preventive effect on alcohol‐related events the first months after treatment initiation. Overall, this study does not support that treatment with the presently available GLP‐1 receptor agonists is an effective alternative to the existing treatment of alcohol use disorders, but the present study considers changes in heavy alcohol consumption only and our results may not be applicable to moderate alcohol use.

ACKNOWLEDGEMENT

None.

CONFLICT OF INTEREST

JR reports receiving consulting fees from Astra Zeneca, NovoNordisk, Boehringer‐Ingelheim, Bayer. AFJ has received an unrestricted research grant from Novo Nordisk A/S to investigate the effects of GLP‐1 receptor stimulation on weight gain and metabolic disturbances in patients with schizophrenia treated with an antipsychotic. The other authors declare no conflicts of interests.

Supporting information

Data S1. Supporting Information

Wium‐Andersen IK, Wium‐Andersen MK, Fink‐Jensen A, Rungby J, Jørgensen MB, Osler M. Use of GLP‐1 receptor agonists and subsequent risk of alcohol‐related events. A nationwide register‐based cohort and self‐controlled case series study. Basic Clin Pharmacol Toxicol. 2022;131(5):372‐379. doi: 10.1111/bcpt.13776

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from Statistics Denmark. Restrictions apply to the availability of these data, which were used under licence for this study.