Prevalence of alcohol-associated liver disease: a systematic review and meta-analysis

Background:

Alcohol-associated liver disease (ALD) is a common cause of morbidity and premature mortality. To date, there has been no systematic synthesis of the prevalence of ALD. This systematic review was done with the aim of reporting the prevalence of ALD across different health care settings.

Methods:

PubMed and EMBASE were searched for studies reporting the prevalence of ALD in populations subjected to a universal testing process. Single-proportion meta-analysis was performed to estimate the prevalence of all ALD, alcohol-associated fatty liver, and alcohol-associated cirrhosis, in unselected populations, primary care, and among patients with alcohol-use disorder (AUD).

Results:

Thirty-five studies were included reporting on 513,278 persons, including 5968 cases of ALD, 18,844 cases of alcohol-associated fatty liver, and 502 cases of alcohol-associated cirrhosis. In unselected populations, the prevalence of ALD was 3.5% (95% CI, 2.0%–6.0%), the prevalence in primary care was 2.6% (0.5%–11.7%), and the prevalence in groups with AUD was 51.0% (11.1%–89.3%). The prevalence of alcohol-associated cirrhosis was 0.3% (0.2%–0.4%) in general populations, 1.7% (0.3%–10.2%) in primary care, and 12.9% (4.3%–33.2%) in groups with AUD.

Conclusions:

Liver disease or cirrhosis due to alcohol is not common in general populations and primary care but very common among patients with coexisting AUD. Targeted interventions for liver disease such as case finding will be more effective in at-risk populations.

Alcohol-associated liver disease (ALD) is a common cause of morbidity and premature mortality worldwide,1 with a significant burden on health care systems.2 ALD is considered as a spectrum ranging from simple steatosis (alcohol-associated fatty liver) through increasing inflammation and scarring to cirrhosis (alcohol-associated cirrhosis), with the risk of liver-related outcomes varying across this spectrum.3 Acute alcohol-associated hepatitis is an acute and dangerous manifestation of ALD that can occur at any stage of underlying disease.

Advanced ALD is relatively easy to identify, but ALD is clinically silent for a long time and early disease may go unrecognized before symptoms occur. In consequence, reliable estimates of the prevalence of early disease require active identification of cases. The clinically silent period of disease is also an opportunity for intervention using noninvasive tests to identify people at risk of progression to advanced disease. The use of these tests needs to be guided by data regarding the epidemiology of ALD as the performance of noninvasive tests will vary depending on the prevalence of disease.

The cause of ALD is excessive alcohol consumption, which is also a common global issue.4 The public health policies of local, regional, and national governments regarding alcohol therefore have a critical role to play in preventing ALD; these policies need to be informed by the best available information on the prevalence of ALD.

To date, there has been no systematic synthesis of the prevalence of ALD. In this manuscript, we present the results of a systematic review of the prevalence of ALD across different health care settings.

METHODS

A systematic literature search was done using the terms “alcoholic liver disease or ALD or alcohol-related liver disease” and “epidemiology or prevalence” to search by title in PubMed and EMBASE. The results were limited to clinical studies published as full papers in English, published between 1948 and November 2022 when the final search was done. Citing literature and reference lists of included papers were reviewed for other relevant papers not identified in the initial search. The titles of papers identified by the literature search were reviewed for relevance and those that appeared suitable were reviewed in more detail. To be included, studies had to report the prevalence of ALD in a specified population tested with a specified screening test, with information about how ALD was defined and the population studied. Studies that used coding from medical records were not included as this did not represent systematic screening for detection of ALD. Searches and data extraction were done by Sachin Amonker and Aryo Houshmand, both medical students at the University of Leeds under the supervision of Richard Parker.

Once papers were identified as being suitable data were extracted into a preprepared spreadsheet. The data recorded for each study were first author, year of publication, location of study (by nation), the total studied population, the type of studied population [ie, unselected cohorts, alcohol-use disorder (AUD), primary care, or secondary care], the average age and alcohol consumption in the total population, the sex distribution in the studied population, the prevalence of hazardous alcohol use in the studied population and the definition of alcohol misuse, the type and the means of defining ALD, and the number of persons in the studied cohort with ALD, hepatic steatosis, advanced fibrosis, and cirrhosis. Alcohol-associated hepatitis was not considered as this is an acute condition where incidence is better considered rather than prevalence. For each analysis, we considered the overall prevalence of ALD and subtypes and the prevalence among hazardous drinkers in that population.

Data were analyzed with single-proportion meta-analysis and forest plots to illustrate the results. All analyses were done using the “meta” package in R. Preplanned subgroup analyses were done for differing methods of defining ALD and for differing geographical areas based on World Health Organization regions. These subgroup analyses were done for the prevalence of all ALD and for cirrhosis. The risk of bias of included studies was assessed using the tool described by Hoy et al.5 The study is reported according to the MOOSE checklist for meta-analysis of observational studies.6

As only published data in the public domain were used in this study, ethical review was not sought and individual patient consent was not relevant. Data were analyzed in R using the following packages: meta7 and ggplot2.8 This systematic review was registered with the PROSPERO database (reference CRD42022327429). No specific funding was received for this research.

RESULTS

The literature search identified a total of 7645 studies. One hundred and twenty three duplicate studies were removed before screening so that 7522 titles were screened. Of these, 7308 were excluded leaving 214 records that were sought for retrieval. Of these studies, a further 47 were removed as duplicates. Of the one hundred and sixty-seven reports that were assessed for eligibility, one hundred and one were ineligible, twenty-five were review articles and one was rejected for language (Figure 1). Four studies were rejected as they relied on coding data, i.e. did not report a systematic examination of study populations. Two studies, Younossi 20219 and Wong 201910 used data from the National Health and Nutrition Examination Survey (NHANES) program in the United States; only data from Wong 2019 was included as this reported the most recent data (up until 2016).

FIGURE 1.

Systematic Review Flow Diagram.

The 35 included studies reported on a total of 513,278 persons, including 50,686 with AUD. The included studies reported on 5968 cases of ALD, 18,844 cases of alcohol-associated fatty liver, 780 cases of noncirrhotic fibrosis, and 502 cases of cirrhosis. A total of 23 studies reported the prevalence of ALD in unselected general populations, 7 reported on populations drawn from primary care, and 5 reported on cohorts with AUD. The majority of included studies were drawn from Europe (18 studies) and Western Pacific (14 studies) with 2 studies from the Americas and 1 from South-East Asia. Four of the 5 studies examining persons in alcohol treatment were drawn from Europe, and the majority of studies examining general unselected populations were from Western Pacific (12 of 23 studies). The details of included studies are shown in Table 1.

TABLE 1.

Details of included studies

References	Country	Population	n	n hazardous alcohol	Age	Sex (% male)	Definition of hazardous alcohol use	Method	ALD	AFL	Fibrosis	AC
AshaRani et al11	Singapore	AUD	101	101	—	81	DSM V criteria for AUD	Biochemistry	18	—	—	—
Bayerl et al12	Germany	General population	384	88	56	58	≥ 20 g/d for women 40 g/day for men	Imaging	—	44	—	—
Bellentani et al13	Italy	General population	6917	707	—	—	>60 g/d in men >30 g/d in women	Imaging	254	233	—	20
Burza et al14	Italy	AUD	384	384	46	76	≥ 24 g/d women ≥36 g/d for men	Biochemistry + imaging	—	—	—	84
Caballería et al15	Spain	Primary care	3014	275	54	43	21 standard drinking U/wk in men 14 standard drinking U/wk in women	Transient elastography	44	—	28	—
Chang et al16	South Korea	General population	218,030	37,856	39	57	30 g/d or greater for men 20 g/d or greater for women	Imaging	—	13,890	—	—
Chaudhary et al17	Nepal	Hospital inpatients	538	—	—	—	—	Biochemistry + imaging	42	—	—	—
Chen et al18	China	General population	6598	1708	39	62	—	Imaging	450	283	—	23
Dang et al19	USA	General population	44,631	—	—	—	>28 g/d in women >42 g/d men	Biochemistry	2708	—	89	—
Fabrellas et al20	Spain	Primary care	495	46	47	—	>21 or 14 U in men or women resp	Transient elastography	—	—	3	—
Fan et al21	China	General population	3175	242	52	38	Alcohol consumption >210 g/wk for >5 y	Imaging	—	61	—	—
Foschi et al22	Italy	General population	3933	—	—	—	≥2 alcohol U/d in women ≥ 3 alcohol U/d in men	Imaging	—	570	—	—
Harman et al23	UK	Primary care	174	174	62	68	>14 U/wk in women >21 U/wk in men alcohol AUDIT questionnaire score ≥8 Read codes related to alcohol misuse	Transient elastography	—	—	31	—
Kotronen et al24	Finland	General population	2766	503	60	51	>20 g (for men) >10 g (for women) per day	Biochemistry	—	195	—	—
Labenz et al25	Germany	Primary care	11,859	—	60	56	—	Biochemistry	—	—	—	4
Laskus et al26	Poland	AUD	70	70	39	93	—	Biopsy	56	18	—	—
Li et al27	China	General population	9094	—	44	52	>40 g ethanol/d for males >20 g ethanol/d for female	Imaging	—	564	—	—
Llop et al28	Spain	General population	11,440	1049	50.3	58.1	AUDIT ≥ 8	Transient elastography	85	—	45	28
Lorenzo et al29	Spain	AUD	220	85	53.8	0	CAGE questionnaire	Biopsy	58	—	—	30
Nagappa et al30	India	Primary care	7624	818	46	56	—	Transient elastography	—	—	327	—
Park et al31	Korea	General population	7517	451	—	49.6	40 g ethanol/d for men 20 g ethanol/d for women	Biochemistry	107	—	—	5
Pendino et al32	Italy	General population	1553	438	45	47%	>28 g >36 g ethanol/d	Biochemistry	93	—	—	—
Pose et al33	Spain	General population	3014	275	—	—	—	—	—	—	28	—
Roulot et al34	France	General population	1190	—	64.8	68.32%	>30 g ethanol/d	Transient elastography	27	—	27	5
Sheron et al35	UK	Primary care	—	393	—	—	AUDIT score ≥ 8	Biochemistry	47	—	45	—
Shi et al36	China	General population	6043	—	—	—	—	Imaging	—	226	—	—
Tajima et al37	Japan	General population	4579	—	—	49.9	Quartiles of intake expressed as percentage of energy intake—highest quartile taken as hazardous	Imaging	—	324	—	—
Thiele et al38	Denmark	AUD	128	128	53	95	>36 g ethanol/d for men > 24 g ethanol/d for women	Biopsy	—	39	88	—
Wang et al39	China	General population	7295	3119	—	0.5	>40 g ethanol/d for men >20 g ethanol/d in women	Imaging	624	125	—	—
Wang et al40	China	General population	74,988	—	—	—	>60 g ethanol/d	Transient elastography	974	—	—	—
Wong et al10	USA	General population	34,423	—	—	—	>28 g ethanol/d women >42 g ethanol/d men	Biochemistry	—	1480	69	—
Yao et al41	—	General population	1690	—	47.9	67.46	—	Imaging	84	—	—	—
Yan et al42	China	General population	3762	—	46.4	63	>40 g ethanol/d in males >20 g ethanol/d in females	Biochemistry + imaging	309	—	—	—
Yoshimura et al43	Japan	Primary care	29,709	1520	—	100	>420 g ethanol/wk	Imaging	—	258	—	—
Zhou et al44	China	General population	3543	—	—	—	>40 g ethanol/d in males >20 g ethanol/d in females	Imaging	—	79	—	—

Abbreviations: AC, alcohol-associated cirrhosis; AFL, alcohol-associated fatty liver; ALD, alcohol-associated liver disease; AUD, alcohol-use disorder.

The presence of absence of ALD was defined by imaging in 4 studies,13,18,39,41 transient elastography in 6 studies,15,23,28,33,34,40 biochemistry in 5 studies,11,19,25,31,32,35,45 and biopsy in 2 studies.26,29

Alcohol-associated fatty liver was mainly defined by imaging in 12 studies,12,13,16,18,21,22,27,36,37,39,42–44 also by biochemistry in 4 studies10,24,33,42 (including 1 study using imaging and biochemistry), and 2 studies used biopsy.26,38 Fibrosis was defined with significant variation between studies: only 1 study used biopsy;26 other studies used transient elastography with cutoffs of ≥6.0,30 ≥6.8,20 ≥8,15,33,34 and ≥10 kPa;28 2 studies used the Fibrosis-4 score with a cutoff of 2.6710,19; 1 study used the AST to platelet ratio index (APRI) score25; and 1 study used the Southampton traffic light test.35 Thus, there were only 2 sets of studies that used directly comparable methodology to define fibrosis. Alcohol-associated cirrhosis was defined with biochemistry and imaging in 1 study,14 imaging in 3 studies,13,18,46) transient elastography using cutoffs of 1528 and 13 kPa,30 a combination of transient elastography, imaging and histology in 1 study23 and 3 studies used biopsy.26,29,38 The methods for diagnosis and any cutoffs used are tabulated in Supplemental Table 1 (http://links.lww.com/HC9/A249).

The individual assessment of bias for each included study is shown in Supplemental Table 2 (http://links.lww.com/HC9/A250). Thirty studies were considered low risk and 5 studies were considered high risk. Publication bias was assessed with funnel plots for each of the primary outcomes (Supplemental Material, http://links.lww.com/HC9/A256)—asymmetry in the funnel plots might indicate a level of bias in published material but may also simply reflect consistency in the estimates of prevalence.

Prevalence of ALD

The prevalence of ALD varied across different settings: in unselected populations, the overall prevalence of any type of ALD was 3.5% (95% CI, 2.0%–6.0%), whereas in groups with AUD, the prevalence was 55.1% (17.8%–87.4%) (Table 2, Figure 2). The prevalence of ALD cirrhosis in general populations was 0.3% (0.2%–0.4%) and 12.9% (4.3%–33.2%) in groups with AUD. Forest plots for reported outcomes in each population are provided in the Supplemental Material, Supplemental Figures 2,4,6,8 (http://links.lww.com/HC9/A256).

TABLE 2.

Estimates for prevalence of ALD, alcohol-associated fatty liver, and alcohol-associated cirrhosis in unselected populations and in hazardous drinkers

	Prevalence in unselected population				Prevalence in hazardous drinkers
		Subgroup of ALD (%)				Subgroup of ALD (%)
	ALD	Fatty liver	Fibrosis	Cirrhosis	ALD	Fatty liver	Fibrosis	Cirrhosis
General population	3.5x (2.0–6.0)	5.0 (3.5–6.9)	0.5 (0.1–1.4)	0.3 (0.2–0.4)	26.0 (14–31)	28.1 (13–48)	6.6 (3.2–5.7)	2.2 (2–4)
Primary care	2.6 (0.5–11.7)	0.9 (0.8–1.0)	3.0 (1.0–8.6)	1.7 (0.3–10.2)	15.4 (11.6–20.2)	17 (15.2–19.2)	15.6 (7.9–28.5)	9.8 (0.3–78.1)
Alcohol-use disorder					55.1 (17.8–87.4)	28.8 (17.9–35.1)	68.8 (60.2–76.2)	12.9 (4.3–33.2)

Abbreviation: ALD, alcohol-associated liver disease.

FIGURE 2.

Prevalence of alcohol-associated liver disease: (A) presence of any liver disease, (B) prevalence of alcohol-associated fatty liver, (C) prevalence of hepatic fibrosis, and (D) prevalence of fibrosis.

Subgroup analyses

Methods of diagnosis

As different methods of diagnosis will have varying accuracy for detection of disease, a subgroup analysis was done after categorizing studies according to the method of diagnosis of ALD. In general, there were too few studies that analyzed the same outcome to allow for robust comparison between diagnostic methods, with the exception of studies describing presence of ALD in general populations. In this case, the study from Dang et al19 using imaging to define liver disease tended to produce higher estimates of ALD prevalence, that is, 7.0% (5.6%–8.6%) versus 3.8% (1.5%–9.4%) for biochemistry and 1.3% (0.7%–2.4%) for transient elastography, although estimates for the prevalence of alcohol-associated cirrhosis were similar for imaging and transient elastography [0.3% (0.2%–0.4%) and 0.2% (0.1%–0.4%), respectively] (Table 3).

TABLE 3.

Prevalence depending on method of diagnosis

	Alcohol-associated liver disease					Alcohol-associated cirrhosis
	Overall estimate	Biochemistry	Imaging	Biopsy	Transient elastography	Overall estimate	Biochemistry	Imaging	Biopsy	Transient elastography
General population	3.5 (2.0–6.0)	3.8 (1.5–9.4)	7.0 (5.6–8.6)	—	1.3 (0.7–2.4)	0.3 (0.2–0.4)	—	0.3 (0.2–0.4)	—	0.2 (0.1–0.4)
Primary care	2.6 (0.5–11.7)	2.2 (0.6–4.4)	—	—	3.2 (0.7–13)	1.7 (0.3–10.2)	—	—	—	1.7 (0.3–10.2)
Alcohol-use disorder	74 (60.8–83.9)	—	—	74 (60.8 –83.9)	—	12.9 (4.3 –33.2)	—	—	12.9 (4.3 –33.2)	—

Geographic location

Included papers were categorized by the World Health Organization region where the study was performed to illustrate the prevalence of ALD geographically (Table 4). No data were available for many areas. Where comparable, one study from Dang et al19 produced higher estimates of the prevalence of ALD than in Europe or Western Pacific, and a higher estimate of the prevalence of ALD cirrhosis in primary care was seen in South-East Asia [3.8% (3.5%–4.2%)] compared with Europe [0.6% (0.2%–1.8%)]. This analysis also illustrated that all data regarding the prevalence of ALD in primary care and in AUD were derived from European cohorts.

TABLE 4.

Prevalence depending on geographical source of study

	Alcohol-associated liver disease					Alcohol-associated cirrhosis
	Overall estimate	Americas	Europe	South-East Asia	Western Pacific	Overall estimate	North America	Europe	South-East Asia	Western Pacific
General population	3.5 (2.0–6.0)	7 (5.6–8.6)	3.0 (1.0–8.3)	—	3.6 (1.6–7.7)	0.3 (0.2–0.4)	—	0.3 (0.2–0.4)	—	0.4 (0.2–0.5)
Primary care	2.6 (0.5–11.7)	—	2.6 (0.5–11.7)	—	—	1.7 (0.3–10.2)	—	0.6 (0.2–1.8)	3.8 (3.5–4.2)	—
Alcohol-use disorder	74 (60.8–83.9)	—	74 (60.8–83.9)	—	—	12.9 (4.3–33.2)	—	12.9 (4.3–33.2)	—	—

CONCLUSIONS

This systematic review and meta-analysis addressed the prevalence of ALD. In unselected populations, ~3 people in a hundred have some form of ALD, but ALD cirrhosis is rare, affecting 3 people in a thousand. As expected, the prevalence of ALD in persons with AUD is higher: three-quarters have some form of ALD and more than 1 person in 10 has cirrhosis. These estimates vary slightly depending on the method of diagnosis and the geographic location of included studies, although these subgroup analyses to compare diagnostic modalities and geographic regions were severely limited by a lack of data. A major finding of this review is the paucity of data from systematic surveys to explore the prevalence of ALD, and this reduces the accuracy of our estimates.

The overall estimated prevalence of ALD of 3.5% can be compared with estimates for the prevalence of NAFLD based on recent meta-analysis, where an overall estimate of 32.4% was reported,47 almost 10 times greater. However, it must be remembered that the harm caused by ALD is similar or greater than NAFLD: ALD remains the leading indication for transplantation in many programs.48 The overall prevalence is comparable to estimates of the global prevalence of hepatitis B,49 although much lower than the prevalence of hepatitis B in areas of endemic disease. Other liver diseases including hepatitis C50 have lower overall prevalence than ALD.

The limits of the search strategy were selected to only include studies that deliberately set out to find cases of liver disease. This meant that several studies that reported on the prevalence of ALD based on coding of medical records were excluded. Although this reduces the number of studies and patients that we can report on, this limitation was chosen to ensure that prevalence data were not influenced by selection biases that might be present in studies based on coding. Despite searching for relevant studies from 1948 onward, the vast majority of included studies were from after 2000: only the studies by Bellentani et al13 and Laskus et al26 were from before this date. This in part reflects the interest in population-based studies that has developed from the landmark Dionysius study that Bellentani and colleagues performed in Italy in the early 1990s. The lack of data from pre-2000 is not necessarily a weakness: the predominance of recent data gives our conclusions greater relevance to current practice.

The estimates that we present are clearly limited by the availability of suitable studies to include in meta-analysis. The lack of available data and the variation in definitions of disease significantly limit our findings. It is also noteworthy that our systematic review included data from slightly more than half a million individuals, whereas similar papers in NAFLD included over 1 million. The assessment of quality of the included papers was generally reassuring where the most common reason for risk of bias was the inclusion of participants in such a way that the study cohort represented the study population. Although the funnel plots might indicate publication bias, this could also simply indicate similar findings from surveys looking for prevalence. The findings of this meta-analysis may be generalizable to Europe or Western Pacific, but the lack of studies from other regions limits the applicability in the Americas, Africa, and South-East Asia.

The estimates for ALD prevalence generated from the meta-analyses have significant heterogeneity (Supplemental Figures 1, 3, 5, and 7, http://links.lww.com/HC9/A249, http://links.lww.com/HC9/A251, http://links.lww.com/HC9/A253, http://links.lww.com/HC9/A255). This may be due at least in part to the differing methods of diagnosis and differing geographic locations of the studies. Planned subgroup analyses to examine diagnostic methods and geographic location revealed the paucity of data once these additional limits were imposed on the data; we could not drill further down to combine both location and method of diagnosis to explore this further. An additional source of heterogeneity may be differing cutoffs employed by differing studies. For example, there were several different cutoffs for transient elastography to define fibrosis. In the absence of an accepted definition for cutoff values, this is not unexpected, but a consensus from the field to inform future studies would be a step forward, for example, using values for the most recent BAVENO consensus statement of 15 kPa to diagnose compensated advanced chronic liver disease would bring a degree of consistency.

There are likely numerous applications of the estimates of prevalence of ALD that we have generated; in hepatology, this can be applied to the early detection of liver disease. Given the current performance of noninvasive tests for fibrosis, early detection of ALD is only likely to be efficient in the context of groups in the treatment for AUD. In other settings, including people with AUD in primary care, the prevalence of ALD or fibrosis cirrhosis is likely too low to consider screening for asymptomatic ALD, although this approach is supported by some guidelines.

Future research should focus on large-scale systematic surveys of prevalence of liver disease, including ALD, in regions where these data are currently lacking, using widely accepted methods of diagnosis. It is notable that more recent studies have used transient elastography to evaluate the prevalence of disease. Given the availability of controlled attenuation parameter included with many devices and the general acceptability to clinicians, this method would represent a convenient method to gauge presence and risk of disease in large cohorts.

This systematic review and meta-analysis is the first comprehensive data synthesis of the epidemiology of ALD. ALD as a whole is not uncommon, but advanced disease is largely confined to groups with coexisting AUD. The prevalence estimates that we report can guide hepatology practice and wider health policy.