1 Introduction

Alcohol-associated liver disease (ALD) is a frequent complication of excessive alcohol intake that can encompass a spectrum of conditions from steatosis, acute inflammation and cirrhosis that may lead to hepatocellular carcinoma (HCC) [1, 2]. Over the past two decades, half a million individuals have died from alcohol-related causes with more than 60% of these deaths attributed to liver disease alone [3]. During the COVID-19 pandemic, online alcohol sales surged by more than 200%, accompanied by an approximately 40% rise in alcohol consumption, which further contributed to a spike in alcohol-related health disorders [4, 5]. The burden of liver disease is even greater among patients with diabetes with potentially higher risk of hepatic complications [6]. In fact, patients with diabetes and concurrent diseases such as metabolic dysfunction-associated steatohepatitis (MASH) and chronic hepatitis have a twofold higher risk of decompensation [7]. As such, there has been increased interest in examining the relationship of ALD with diabetes, as well as characterising optimal management approaches in clinical practice.

The metabolism of ethanol to acetaldehyde and eventually to acetate can cause a redox shift by impairing the NADH/NAD ratio [8]. This shift can disrupt metabolic pathways such as gluconeogenesis, glycogenolysis and the β-oxidation of fatty acids, potentially leading to ALD [9]. Owing to the shared pathogenesis of altered metabolism of glucose as well as lipids, a strong reciprocal relationship exists between diabetes and alcohol-related liver pathologies [10, 11]. Additionally, incretins such as glucagon-like peptide-1 are commonly implicated in food regulation with glycaemic control and may share neuronal pathways related to alcohol-mediated behaviours [12]. A relatively newer class of antiglycaemic drugs, including glucagon-like peptide-1 receptor agonists (GLP-1RAs), targets the incretin pathways and has become popular for managing diabetes [10, 13]. Previous studies have demonstrated that this class of medications can reduce aminotransferase levels among patients with metabolic-associated fatty liver disease, suggesting benefits beyond diabetes management [13]. These benefits may be due to associated weight reduction, decreased inflammatory markers and reduced lipotoxicity [10, 14]. Furthermore, Quddos et al. reported that semaglutide and tirzepatide are linked to decreased alcohol consumption among patients with obesity [12].

To date, the clinical evidence of the effectiveness of GLP-1RA medications among patients with ALD remains ill-defined. Patients with severe ALD such as cirrhosis are often excluded from randomised controlled trials due to safety concerns, underscoring a need for an observational study [15]. Therefore, the objective of the current study was to characterise the impact of GLP-1RAs on adverse liver outcomes (ALO) among patients with ALD and Type 2 diabetes mellitus (T2DM).

2 Methods

This retrospective cohort study utilised the IBM MarketScan database, a comprehensive national dataset containing longitudinal individual-level information on privately insured patients across inpatient, outpatient and prescription drug services [16]. The database has been previously validated to study epidemiological and clinical outcomes [17, 18]. The International Classification of Diseases, ninth and tenth editions (ICD-9/10) codes, identified patients under 65 years of age with T2DM, who had at least one inpatient and two outpatient diagnoses of ALD from 2013 to 2020 (Table S1) [19]. The index date of diagnosis was defined as the first record of ALD following 12 months of continuous enrolment in an insurance plan. Similarly, patients with prior exposure to GLP-1RA before the index diagnosis of ALD and individuals with other liver conditions were excluded (Figure 1). The national drug codes were employed to query the pharmaceutical records to identify patients who had at least one claim for GLP-1RAs therapy (semaglutide, albiglutide, liraglutide, dulaglutide, lixisenatide or exenatide) within a year after the index diagnosis of ALD. The remaining cohort undergoing routine care was characterised as the non-GLP-1RA group. The study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines. The current study was deemed exempt from approval by the Institutional Review Board at Ohio State University since the IBM Marketscan database meets the criteria for limited-use data.

2.1 Variables and Exposure

Variables of interest included patient age, sex, Charlson Comorbidity Index (CCI), claim-based frailty index, employment status, benefit plan type, US census region (northeast, northcentral, south and west) and baseline concurrent medical conditions (Table 1). The CCI was calculated after excluding mild and moderate-to-severe liver disease along with diabetes without chronic complications and was categorised as previously defined [20-22]. The claim-based frailty index is a measure of deficit accumulation frailty within the prior 12 months, estimated from the administrative dataset using Current Procedural Terminology Codes, Healthcare Common Procedure Coding System Codes and ICD codes [23]. Patients were grouped into three categories based on their employment status: retired, actively employed and ‘Other’. The ‘Other’ category included dependents, individuals with long-term disabilities and Consolidated Omnibus Budget Reconciliation Act (COBRA) continuers. Additionally, ICD-9/10 codes identified any concurrent baseline medical conditions related to the cardiovascular, central nervous and gastrointestinal systems (Table 2).

TABLE 1. Baseline characteristics of patients with alcoholic liver disease with type 2 diabetes by glucagon-like peptide-1 receptor agonists initiation status.

		Unweighted				Weighteda
Patient characteristics	Total N = 14 730	GLP-1RA N = 317 (2.2%)	Non-GLP-1RA N = 14 413 (97.8%)	Standard diff.b	p	GLP-1RA (%)c	Non-GLP-1RA (%)c	Standard diff.c	p
Age, years
Mean (SD)	55.3 ± 6.3	54.7 ± 6.6	55.8 ± 7.3	15.8	0.005	54.8 ± 6.1	54.7 ± 1.0	−0.022	0.100
Median (IQR)	57 (52–61)	56 (51–60)	58 (52–61)	−1.0	0.005	56 (51–60)	56 (51–60)	−0.022	0.100
Sex
Male	9752 (66.2%)	158 (49.8)	9594 (66.6)	34.6	< 0.001	49.8	50.8	0.020	0.810
Female	4978 (33.8%)	159 (50.2)	4819 (33.4)	34.6	< 0.001	50.2	49.2	−0.020	0.810
Region
Northeast	2839 (19.3%)	51 (16.1)	2788 (19.3)	8.4	0.001	16.1	16.7	−0.016	0.260
North Central	2757 (18.7%)	71 (22.4)	2686 (18.6)	9.4	0.001	22.4	18.9	0.087	0.260
South	6738 (45.7%)	166 (52.4)	6572 (45.6)	13.6	0.001	50.4	50.1	0.006	0.260
West	2294 (15.6%)	29 (9.1)	2265 (15.7)	20.1	0.001	11.1	14	−0.088	0.260
Type of health insurance
PPO	8224 (55.8%)	173 (54.6)	8051 (55.9)	2.6	0.038	54.6	53.6	0.020	0.470
HMO	1916 (13.0%)	30 (9.5)	1886 (13.1)	11.4	0.038	10.5	13.5	−0.092	0.470
Comprehensive	802 (5.4%)	14 (4.4)	788 (5.5)	5.1	0.038	4.4	5.1	−0.033	0.470
POS	969 (6.6%)	20 (6.3)	949 (6.6)	1.2	0.038	6.3	6.8	−0.020	0.470
Otherd	2819 (19.1%)	80 (25.2)	2739 (19.0)	15.0	0.038	24.2	21	0.077	0.470
Employment status
Other	7769 (52.7%)	199 (62.8)	7570 (52.5)	21.0	0.001	62.8	58.9	0.080	0.310
Full/part-time	2336 (15.9%)	45 (14.2)	2291 (15.9)	4.8	0.001	12.2	12.7	−0.015	0.310
Retired	4625 (31.4%)	73 (23.0)	4552 (31.6)	19.4	0.001	25.0	28.4	−0.077	0.310
CCI
0	8639 (58.6%)	191 (60.3%)	8448 (58.6%)	7.5	0.183	60.3	60	0.006	0.800
1–2	3595 (24.4%)	84 (26.5%)	3511 (24.4%)	10.5	0.183	26.5	24.8	0.033	0.800
3–4	1290 (8.8%)	27 (8.5%)	1263 (8.8)	2.5	0.183	8.5	8.3	0.015	0.800
5+	1206 (8.2%)	15 (4.7%)	1191 (8.3%)	14.6	0.183	4.7	6.8	−0.090	0.800
Claim-based frailty index
Nonfrail	6993 (47.5%)	134 (42.3)	6859 (47.6)	10.7	0.030	42.9	46.6	−0.074	0.310
Prefrail	7021 (47.7%)	173 (54.6)	6848 (47.5)	14.2	0.030	53.8	48.9	0.098	0.310
Frail	716 (4.9%)	10 (3.2)	706 (4.9)	8.6	0.030	3.3	4.5	−0.062	0.310

• Abbreviations: CCI, Charlson Comorbidity Index; diff, difference; GLP-1RA, glucagon-like peptide-1 receptor agonist; IQR, interquartile range; SD, standard deviation.
• ^a Pseudosample was created using the overlap weighting technique, ensuring that GLP-1RA initiation is independent of other clinicodemographic characteristics of the patients.
• ^b Absolute difference in means/proportions divided by pooled SD. The absolute value greater than 0.10 represented an imbalance between two study groups; smaller values indicate better balance.
• ^c Overlap weighted proportions and standardised differences were calculated using all patient demographics and medical history.
• ^d Including, COBRA, long-term disability, surviving spouse/dependent or other/unknown.

TABLE 2. Additional clinical characteristics of patients with alcoholic liver disease with type 2 diabetes by glucagon-like peptide-1 receptor agonists initiation.

		Unweighted				Weighteda
Patient characteristics	Total N = 14 730	GLP-1RA N = 317 (2.2%)	Non-GLP-1RA N = 14 413 (97.8%)	Standard diff.b	p	GLP-1RA (%)c	Non-GLP-1RA (%)c	Standard diff.c	p
Myocardial infarction	328 (2.2%)	2 (0.6%)	326 (2.3%)	14.3	0.052	1.6	2.1	−0.037	0.110
Congestive heart failure	1220 (8.3%)	14 (4.4%)	1206 (8.4%)	16.4	0.012	5.4	7.7	−0.093	0.090
Peripheral vascular disease	628 (4.3%)	8 (2.5%)	620 (4.3%)	9.9	0.121	2.5	3.6	−0.064	0.430
Cerebrovascular disease	475 (3.2%)	8 (2.5%)	467 (3.2%)	4.2	0.475	2.5	3.2	−0.042	0.640
Dementia	36 (0.2%)	0 (0.0%)	36 (0.2%)	418.7	0.373	0	0.3	−0.078	0.350
Hemiplagia or paraplegia	73 (0.5%)	1 (0.3%)	72 (0.5%)	3.2	0.644	0.3	0.5	−0.032	0.740
Diabetes with complications	2585 (17.5%)	68 (21.5%)	2517 (17.5%)	10.1	0.065	21.5	18.8	0.067	0.420
Hypertension	9917 (67.3%)	219 (69.1%)	9698 (67.3%)	3.9	0.499	69.1	68.2	0.019	0.810
Obesity	1259 (8.5%)	52 (16.4%)	1207 (8.4%)	24.5	< 0.001	16.4	15.6	0.022	0.790
Dyslipidaemia	8435 (57.3%)	198 (62.5%)	8237 (57.1%)	11.0	0.059	62.5	59.1	0.070	0.400
Renal disease	1344 (9.1%)	18 (5.7%)	1326 (9.2%)	13.4	0.031	5.7	7.6	−0.076	0.350
Chronic pulmonary disease	1429 (9.7%)	34 (10.7%)	1395 (9.7%)	3.3	0.533	10.7	9.7	0.033	0.690
Peptic ulcer disease	217 (1.5%)	4 (1.3%)	213 (1.5%)	1.7	0.752	1.3	1.2	0.009	0.970
Rheumatic disease	287 (1.9%)	11 (3.5%)	276 (1.9%)	9.9	0.048	3.5	2.3	0.072	0.380
AIDS/HIV	107 (0.7%)	1 (0.3%)	106 (0.7%)	5.7	0.384	0.3	0.7	−0.057	0.550
No malignancy	13 163 (89.4%)	295 (93.1%)	12 868 (89.3%)	13.4	0.031	6.9	8.3	−0.053	0.520
Metastatic solid tumour	399 (2.7%)	4 (1.3%)	395 (2.7%)	10.0	0.109	1.3	2.0	−0.055	0.490
Combination drugs	238 (1.6%)	12 (3.8)	426 (3.0)	4.4	0.390	3.8	3.1	0.038	0.640
Metformin	422 (2.9%)	17 (5.4)	405 (2.8)	13.1	0.007	5.4	4.9	0.023	0.790
SGLT-2	251 (1.7%)	16 (5.0)	235 (1.6)	19.1	< 0.001	5.0	4.4	0.028	0.690
DPP-4	764 (5.2%)	48 (15.1)	716 (5.0)	34.1	< 0.001	15.1	14.0	0.031	0.690
Sulphonylureas	414 (2.8%)	11 (3.5)	403 (2.8)	4.0	0.473	3.5	2.1	0.085	0.300
Thiazolidinedione	108 (0.7%)	3 (0.9)	105 (0.7)	2.2	0.653	0.9	0.9	0.000d	0.920
Statins	324 (2.2%)	8 (2.5)	316 (2.2)	2.0	0.691	2.5	2.0	0.034	0.660
Hepatic decompensation	4717 (32.0%)	71 (22.4%)	4646 (32.2%)	22.1	< 0.001	22.4	22.9	−0.012	0.890
Ascites	2820 (19.1%)	27 (8.5%)	2793 (19.4%)	31.9	< 0.001	8.5	8.9	−0.014	0.850
Hepatic encephalopathy	2290 (15.5%)	40 (12.6%)	2250 (15.6%)	8.6	0.146	12.6	12.7	−0.003	0.960
Hepatorenal syndrome	150 (1.0%)	0 (0.0%)	150 (1.0%)	398.2	0.068	0	0.5	−0.100	0.190
SBP	366 (2.5%)	4 (1.3%)	362 (2.5%)	8.8	0.157	1.3	1.7	−0.033	0.670
Variceal bleeding	700 (4.8%)	12 (3.8%)	688 (4.8%)	4.9	0.413	3.8	3.9	−0.005	0.960
Hepatocellular carcinoma	563 (3.8%)	1 (0.3)	562 (3.9%)	25.3	0.001	0.3	0.3	0.000d	0.960
Liver transplant	374 (2.5%)	4 (1.3%)	370 (2.6%)	9.4	0.144	1.3	1.2	0.009	0.980
Portal Hypertension	2345 (15.9%)	46 (14.5%)	2299 (16.0%)	4.2	0.488	14.5	12.6	0.056	0.500

• Abbreviations: AIDS, acquired immunodeficiency syndrome; DPP-4, dipeptidyl peptidase-4 inhibitors; GLP-1RA, glucagon-like peptide-1 receptor agonist; SBP, spontaneous bacterial peritonitis; SGLT-2, sodium–glucose cotransporter-2.
• ^a Pseudosample was created using the overlap weighting technique, ensuring that GLP-1RA initiation is independent of other clinicodemographic characteristics of the patients.
• ^b Absolute difference in means/proportions divided by pooled SD. The absolute value greater than 0.10 represented an imbalance between two study groups; smaller values indicate better balance.
• ^c Overlap weighted proportions and standardised differences calculated by using all patient demographics and medical history.
• ^d Overlapping weighting resulted in an exact balance for this variable.

The primary exposure variable was the initiation of GLP-1RAs, treated as a time-varying factor to accommodate the duration between the index diagnosis of ALD and the start date of GLP-1RA therapy. Accordingly, patients were categorised into the exposure group if GLP-1RAs were initiated before (1) the occurrence of an ALO event, (2) the end of enrolment in the insurance plan or (3) the conclusion of the study period. Conversely, patients who commenced GLP-1RA therapy after said events were classified as unexposed (Figure 2).

3 Results

3.3 OPSW

The median follow-up time for all patients was 42.2 (IQR: 17.8–84.0) months (GLP-1RA, 55.0 [IQR: 26.0–91.3] months vs. non-GLP-1RA, 42.0 [IQR: 17.5–83.8] months). On average, patients waited 5.5 months (standard deviation [SD]: ±3.8) after being diagnosed with ALD before starting GLP-1RA. Similarly, the median time from GLP-1RA exposure to the incidence of ALO was 14.0 (IQR: 6.6–27.8) months. After applying OPSW, all the clinicodemographic characteristics of patients who did and did not receive GLP-1RA were well balanced (Table 1). Of note, patients on GLP-1RA had a lower occurrence of composite ALO, with 46 events over 379 person-years versus 75 ALO events over 353 person-years in the unexposed group. The overall incidence of ALO was lower in the GLP-1RA cohort (GLP-1RA: 12.0%, 95%CI 9.0–16.0 vs. non-GLP-1RA: 21.0%, 95%CI 20.0–22.0) with an absolute incidence risk reduction of 9.0% (95%CI 3.0%–15.0%) associated with GLP-1RA (Figure 3a). On further stratification and cause-specific analyses, this association was primarily driven by a reduction in the incidence of hepatic decompensation among the different ALO components. Particularly, there was a 7.0% (95%CI 4.0–8.0) absolute reduction in the incidence of hepatic decompensation with an incidence rate of 8.0% (95%CI 6.0–12.0) in the GLP-1RA group versus 15.0% (95%CI 14.0–16.0) in the non-GLP-1RA group (Figure 3b). In contrast, there were no differences in incidence rates of portal hypertension, HCC and liver transplantation between the two cohorts. On Poisson regression, after accounting for other clinicodemographic characteristics, the adjusted incidence rate of ALO and hepatic decompensation among GLP-1RA initiation group remained lower at 0.57 (95%CI 0.39–0.82) and 0.56 (95%CI 0.36–0.86), respectively (Table 3).

TABLE 3. Associations between glucagon-like peptide-1 receptor agonists initiation^a and risk of adverse liver outcomes^b in adults with alcoholic liver disease and type 2 diabetes.

Outcomes	Number of events	Person-years	Incidence rates per person-years (95%CI)	Absolute incidence rate difference per person-years (95%CI)	Incidence rate ratio (95% CI)	Adjusted incidence rate (95%CI)c
Adverse liver outcomes
No GLP-1RA	75	353	0.21 (0.20–0.22)	—	—	Ref.
GLP-1RA	46	379	0.12 (0.09–0.16)	−0.09 (−0.15 to −0.03)	0.55 (0.38–0.80)	0.57 (0.39–0.82)d
Hepatic decompensation
No GLP-1RA	57	381	0.15 (0.14–0.16)	—	—	Ref.
GLP-1RA	33	400	0.08 (0.06–0.12)	−0.07 (−0.08 to −0.04)	0.55 (0.42–0.74)	0.56 (0.36–0.86)d
Hepatocellular carcinoma
No GLP-1RA	9	429	0.02 (0.02–0.03)	—	—	Ref.
GLP-1RA	2	427	0.01 (0.00–0.05)	−0.02 (−0.02 to 0.02)	0.21 (0.05–1.82)	0.32 (0.06–1.63)
Liver transplant
No GLP-1RA	4	433	0.01 (0.008–0.012)	—	—	Ref.
GLP-1RA	1	428	0.002 (0.001–0.009)	−0.007 (−0.008 to −0.004)	0.22 (0.11–0.71)	0.57 (0.60–5.49)
Portal Hypertension
No GLP-1RA	35	398	0.09 (0.08–0.10)	—	—	Ref.
GLP-1RA	29	400	0.07 (0.05–0.11)	−0.02 (−0.03 to 0.01)	0.81 (0.61–1.10)	0.76 (0.46–1.24)

• Abbreviations: CI, confidence interval, GLP-1RA: glucagon-like peptide 1 receptor agonists.
• ^a Glucagon-like peptide-1 receptor agonists initiation status was modelled as a time-dependent covariate.
• ^b Adverse liver outcomes were defined as (1) hepatic decompensation (ascites, hepatic encephalopathy, hepatorenal syndrome, spontaneous bacterial peritonitis and variceal bleeding), (2) hepatocellular carcinoma, (3) liver transplant and (4) portal hypertension.
• ^c Adjusted using overlapping weights estimated using age, sex, region of residence, type of health insurance, year of diagnosis, employment status, history of smoking, obesity, dyslipidaemia, hypertension, Charlson Comorbidity Index, claims-based frailty index, metformin, SGLT-2, DPP-4, sulphonylureas, thiazolidinediones and statins.
• ^d p < 0.05.

4 Discussion

Over the past decade, the primary cause of death due to chronic liver disease has shifted from viral hepatitis to fatty liver disease [28]. Consequently, there is a growing emphasis on understanding fatty liver such as ALD and MASH to address the rising liver-related health burden [29, 30]. Despite efforts to regulate alcohol access and reduce alcohol abuse, ALD continues to be one of the leading causes of liver-related illness and mortality [29-31]. In fact, over half of deaths from alcohol-related causes are due to liver disease with the associated mortality nearly doubling each year [3, 28]. Traditionally, medications such as acamprosate and naltrexone are approved for managing alcohol use disorder yet these drugs do not offer any metabolic benefits for concurrent alcohol-induced dysfunction [12, 32]. Of note, GLP-1RAs—which are well recognised in the management of diabetes mellitus through improved glycaemic control—have been suggested to help create an aversion to alcohol consumption [12]. Nonetheless, the impact of GLP-1RA intake on liver outcomes among patients with ALD had not been investigated. Therefore, the current study was important as it defined the association of GLP-1RA initiation with the incidence of ALO among patients with T2DM who were newly diagnosed with ALD. Notably, patients with GLP-1RA initiation after an index diagnosis of ALD had a lower overall incidence of ALO, particularly hepatic decompensation. Interestingly, on a stratified analysis, the benefit of GLP-1RA on composite ALO remained consistent across patients with or without baseline liver compensation.

There is evidence that alcohol-associated fatty liver disease, particularly when accompanied by advanced fibrosis, is an independent predictor of both liver-specific and all-cause mortality [33]. Diabetes, which is disproportionately common among patients with ALD, is another known risk factor that can contribute to liver-associated complications [34]. To this end, antidiabetic drugs such as metformin, thiazolidinediones and GLP-1RA can reverse steatosis and improve liver function [35]. In fact, previous studies have demonstrated that GLP-1RAs, particularly semaglutide, were associated with the resolution of steatohepatitis in patients with T2DM [36, 37]. This effect may result from reduced insulin resistance and weight loss, which in turn can improve overall metabolic function and reduce the liver fat content [36]. These desirable effects may explain findings in the current study, which demonstrated that initiation of GLP-1RA treatment reduced the incidence rate of ALO per person-years by 9% among patients with ALD. Interestingly, after adjusting for confounding variables in multivariable analysis, GLP-1RA remained independently associated with a 43% reduced incidence of composite ALO.

Managing ALO may differ based on different baseline liver compensation or decompensation status [10]. Prior evidence has suggested that GLP-1RA may be superior to other antidiabetic medications in reducing the risk of HCC and hepatic decompensation [38]. Similarly, among patients with T2DM who have been diagnosed with MASH, GLP-1RA has been suggested to be associated with improved liver-related outcomes [10]. In the present study, the beneficial effects of GLP-1RA were consistent across patients with incidence of hepatic decompensation (22.4% vs. 32.2%) and HCC (0.3% vs. 3.9%) diagnosis in the unweighted models of patients newly diagnosed with ALD. Importantly, after OPSW-based analysis, the effects of GLP-1RA on ALO were demonstrated to be primarily mediated through a marked (44%) reduction in the incidence of hepatic decompensation. Given that GLP-1RAs are primarily excreted unchanged through the kidneys, the benefits of the drug may be mainly among patients with compensated liver disease [39]. To address this question, a separate analysis was conducted among patients with baseline decompensation to assess the impact of GLP-1RA initiation on ALO. This analysis demonstrated that patients with already diagnosed liver decompensation experienced the same benefits of taking GLP-1RA as individuals with compensated liver disease. These findings aligned with a previous study in which patients with baseline decompensation associated with MASH had equally favourable effects from initiation of GLP-1RA [10]. There is prior evidence that GLP-1RA has the potential to enhance and possibly reverse diabetic nephropathy by affecting the renin–angiotensin–aldosterone axis (RAAS) [40]. It is worth noting that there is a common underlying mechanism associated with hepatorenal syndrome, and enhanced renal function may be a reason for similar benefits of GLP-1RA among patients with decompensated liver disease [41].

While the benefits may extend beyond managing T2DM alone, utilisation of GLP-1RA remains relatively low [42]. Previously reported usage data ranged from as low as 9% to 15% among patients with diabetes [10, 43]. In the current study, only about 3% of patients initiated GLP-1RA treatment, likely due to the inclusion of patients with T2DM and concurrent ALD. The underuse of GLP-1RA can be multifaceted and may include factors such as therapeutic inertia among clinicians, patients' concerns about adverse effects and the mode of delivery [44, 45]. Additionally, disparities in care have been noted with patients from higher socioeconomic backgrounds being more likely to receive modern antidiabetic treatments, including GLP-1RA [43]. Similarly, rising costs can be another contributing factor to low initiation rates of GLP-1RA [46]. According to a report by Sumarsono et al., GLP-1RAs are the second costliest antiglycaemic medication with a notable 135% cost increase over the last 5 years [46]. Additionally, the Patient Protection and Affordable Care Act is expected to increase the cost burden on patients and private insurers may also shift cost shares [47]. Consequently, rising treatment expenses can place patients at risk of ‘financial toxicity’, which in turn can reduce preference for GLP-1RA [47, 48]. Recently, following increased attention to the wide-ranging benefits of GLP-1RA, the FDA approved this class of drugs for additional indications, including weight management and the secondary prevention of cardiovascular events [42]. Data from the current study suggest that there may be additional benefits of GLP-1RA for patients with liver disease. Therefore, healthcare organisations and policymakers should focus on improving access to this class of drugs for patients with diabetes.

The findings of the current study should be interpreted considering certain limitations. The use of a retrospective administrative database may be prone to inaccurate data input and missing values. The IBM MarketSdcan database included only individuals with employer-sponsored health insurance in the United States, which may not represent other populations, such as uninsured individuals or those covered by government-sponsored health plans like Medicare or Medicaid [16]. Additionally, data collection relied on convenience sampling, which may limit the generalisability of the results. Since ICD and national drug codes were used to select the patient cohort, exposure of interest and outcomes, there was a risk of misclassification bias and residual confounding attributable to potential coding errors [49]. This was an observational study that only reported associations and could not ascertain causation. Nonetheless, the causal inference approach using overlap weighting alleviated confounding by indication [10]. Moreover, GLP-1RA was modelled as a time-varying variable to address immortal time bias. Approximately one-third of patients prescribed GLP-1RA fail to complete treatment [50]. To this point, the current study was unable to ascertain treatment completion rates, potentially limiting the study of the long-term benefits of GLP-1RA therapy. However, the focus of our investigation primarily centred on a 3-year follow-up period, which helped mitigate this limitation. Furthermore, due to the limited number of variables, granular assessment of patient-level factors (e.g., race/ethnicity) or severity of comorbidities that may drive the effects of GLP-1RA could not be assessed. Despite these limitations, the study provides valuable insight into the impact that GLP-1RA may have on patients with concurrent liver disease, which otherwise would be challenging to study in randomised controlled trials.

In conclusion, GLP-1RA may potentially mitigate the risk of ALO, particularly hepatic decompensation, among patients with ALD and concurrent T2DM. As such, these medications may be an effective therapeutic tool among these patients. Future trials are needed to further investigate the role of GLP-1RA among patients with liver disease.

Conflicts of Interest

The authors declare no conflicts of interest.

Disclaimer

The authors have nothing to report.