Measuring the Alcohol in Alcohol-Associated Liver Disease: Choices and Challenges for Clinical Research

Alcohol-associated liver disease (ALD) has risen sharply in preceding decades to become a leading cause of morbidity and mortality in the United States and worldwide.^(1,2) In the United States, ALD is now the number-one reason for liver transplantation, with medical therapies offering limited effectiveness.⁽³⁾ Prednisolone is the most well-known and widely studied pharmacotherapy for alcohol-associated hepatitis (AAH), with modest efficacy in decreasing mortality in the short-term but lacking in long-term mortality reduction.^(4,5) In contrast, alcohol use, even as little as 1-2 drinks per day, increases long-term mortality and portal hypertensive complications, with heavier drinking worsening these outcomes.^(4,6) To address this rising tide of ALD, several ALD clinical trials are ongoing, many addressing AAH, and others focused on discovering and targeting novel compounds or repurposing older medications with the goal of reducing hepatic inflammation or boosting hepatocyte regeneration. Because of alcohol’s confounding effects on all liver-related outcomes, robust and accurate measurement of drinking is critical to all ALD clinical trials. To examine outcomes in ALD clinical research, clinical researchers must reckon with accurately assessing alcohol use (see Table 1).

TABLE 1.

Recommendations for Best Practices in Alcohol Use Clinical Drug Development in Patients With ALD

	Best Practices in ALD Clinical Trials
Outcomes	Primary alcohol-use outcome should be total abstinence
	Secondary alcohol-use outcome of alcohol-use reduction (PNHDD) can be considered
	If using PNHDD in outcomes, avoid group means and calculate the outcome at the responder level instead
	Consider non-alcohol-related outcomes, such as improvements in AUD insight, mental health, social functioning, interpersonal relationships, quality of life, as well as AUD treatment engagement and retention, reduction in polysubstance use, and regained sobriety following relapse
Measurement	Use validated questionnaires (AUDIT, TLFB) to assess alcohol use based on degree of granularity of detail desired
	Use alcohol biomarkers (PETH, urine ethyl glucuronide) as an adjunct to questionnaires in assessing abstinence
	Use validated psychometric instruments to detect changes in mental health domains
Trial design	At least 6 months of follow-up to ensure durable treatment response is accounted for, as opposed to shorter, more transient periods of abstinence that are known to occur without treatment
	Consider sex differences in measurement of alcohol and liver outcomes

Alcohol Use in Clinical Trials: What Endpoints Are Available?

Clinical research in alcohol use reduction, including medication development, has been ongoing for decades. Endpoints for alcohol use have been defined by the Food and Drug Administration (FDA) with respect to drug development for alcohol relapse. The draft FDA Guidance on drug development for alcohol use states: “Sponsors do not need to demonstrate a direct effect on the physical or psychosocial consequences of alcoholism in alcoholism clinical trials, but they should show modifications in drinking behavior ascribe to a particular treatment that are likely to translate to improvement in the physical and psychosocial consequences [emphasis added].”⁽⁷⁾ For the FDA, two endpoints have been deemed equally acceptable: total abstinence or percent of subjects with no heavy drinking days (PNHDD). A heavy drinking day is defined as consumption of more than four drinks per day for women and more than five drinks per day for men.

Why was PNHDD selected as an appropriate outcome? Using data from the National Epidemiologic Survey of Alcohol and Related Conditions, a large US-based survey, those subjects who had alcohol dependence in the past year but who had no heavy drinking days had a very low risk (< 5%) of experiencing recurrent alcohol dependence–telated symptoms compared to those who had any heavy drinking days.⁽⁸⁾ Furthermore, data from the same survey found that the recurrence of alcohol use disorder (AUD) symptoms was 14.6 times greater in those who had any heavy drinking days, even just one or two.⁽⁸⁾ Loosening the restrictions on abstinence as the only outcome also allows for a broader range of effects to be studied in the alcohol using population. Given that many patients either do not want to or perceive that they cannot completely abstain from all alcohol use, PNHDD also more accurately reflects patient goals and desires, and therefore may be more patient-centered.

Because as many as two-thirds of patients with AUD will maintain short periods of abstinence (3 months or less) without treatment in early recovery,^9,10 the FDA recommends that clinical trial designs in alcohol use drug development have a minimum duration of 6 months to ensure that efficacy is attributable to the intervention and not to naturally occurring shorter periods of abstinence. In addition, the FDA recommends calculating PNHDD at the responder level and not as a group mean. Calculating a group mean for all patients in the study in which all days of non–heavy drinking are lumped together obscures meaningful effects in subgroups of patients. So, for example, an intervention that produces better effects in women might be obscured if heavy drinking days were calculated as a group mean rather than at the individual responder level and stratified by gender.

The choice of these two endpoints, abstinence or PNHDD, reflects two complementary views of addiction treatment. In a “total abstinence” model, complete abstention is the primary focus, and any alcohol use, even slips with quick return to full sobriety, are treated as breaks in abstinence in zero-sum fashion. This model has been the dominant approach in the addiction community and is the model used by Alcoholics Anonymous. In contrast, a “harm reduction” model focuses on reducing the harms of alcohol use, which may or may not include reduction in drinking or total abstinence. Harm reduction models align closely with patient desires as a guide for treatment goals and make frequent recourse to a motivational interviewing style of engagement. Both approaches to alcohol- use treatment have advantages and disadvantages and which approach is used should be tailored to the clinical scenario.

In line with harm-reduction modeling, use of the World Health Organization (WHO) drinking risk levels has been recently proposed⁽¹¹⁾ (see Table 2). Subjects are categorized in a range from abstinent to very high risk based on either grams of pure ethanol per day or mean number of standard drinks per day. Use of drinking metrics can be stymied in studies crossing international boundaries given that countries differ in how much alcohol is in a standard drink. In the United States, a standard drink contains 14 g of ethanol, while in India, for example, 8 g of ethanol is considered a standard drink. The WHO drinking risk measurement is expressed as either one-level or two-level reductions (e.g., someone moving from medium-risk drinking to low-risk would be a one-level reduction, while someone moving from very high risk to medium risk would be a two-level reduction). Meaningful improvement in functioning has been demonstrated for two-level reductions and shows larger effect sizes with less variable and lower numbers needed to treat.⁽¹¹⁾

TABLE 2.

Possible Alcohol Use Outcomes: Total Abstinence, Percent Non-heavy Drinking Days, and WHO Drinking Risk Levels

Definitions
Abstinence	Complete cessation of all alcohol use; any positive biomarker or slip is treated as a break in abstinence (preferred primary alcohol-use outcome in ALD clinical trials)
PNHDD	Percentage of days with no heavy drinking; heavy drinking is defined as four or more drinks for women or five or more drinks for men in 1 day
WHO drinking-risk levels	Risk level	Ethanol (g/day) (Men)	Ethanol (g/day) (Women)	Drinks/day (Men)	Drinks/day (Women)
	Low	1–40	1–20	0–3	0–1
	Medium	41–60	21–40	3–4	2–3
	High	60–1 00	41–60	4–7	3–4
	Very high	101 +	61+	7+	4+

Which Endpoint Should Be Used in ALD Trials?

For trials in liver disease, choosing an alcohol endpoint requires selecting the endpoint that would correspond to a level of alcohol use correlated with good liver-related outcomes. While alcohol reduction is beneficial, it is clear that, for patients with alcohol-associated cirrhosis or alcohol-associated hepatitis, any alcohol use produces worse outcomes. As little as one to two drinks per day nearly doubled the relative risk of death in women with alcohol-associated cirrhosis (relative risk 1.9, 95% confidence interval [CI] 1.1-3.1).⁽⁶⁾ For men, the use of alcohol was similarly deadly in those with cirrhosis, although women had twice the risk of mortality with the same level of alcohol use (e.g., at consumption levels of 36-48 g of ethanol per day, women had a relative risk of mortality of 10.1 [95% CI 7.5-13.5] compared with men at 5.6 [95% CI 4.5-7.0]).⁽⁶⁾ In alcohol-associated hepatitis, a similar story emerges. In those who survived the first 6 months following an episode of alcohol-associated hepatitis, consumption of 1-29 g of ethanol per day (approximately 1-2 drinks) resulted in an adjusted hazard ratio of death of 2.27, with mortality increasing as alcohol consumption increased (>100 g ethanol per day: hazard ratio 6.01 [95% CI: 3.64-9.91]).⁽⁴⁾ Alcohol consumption has also been clearly shown to increase rates of ascites, encephalopathy and variceal bleeding, with the odds ratio of variceal bleeding rising to 10.6 in those with active alcohol use.⁽¹²⁾

Given these findings, is it appropriate to use PNHDD as outcomes in liver disease research? For advanced ALD (cirrhosis and AAH), the answer is no. While it may be useful to capture alcohol reduction, primary alcohol endpoints need to include total abstinence for clinical research in advanced ALD; given the massive confounder that even a small amount of alcohol would be on mortality and liver-related outcomes in these patients.

But what about for those with less severe ALD, such as alcohol-associated steatohepatitis or steatosis? While previously thought to be benign, alcohol-associated steatosis and steatohepatitis have 5-year risks of progression to cirrhosis of 6.9% and 16.0%, respectively,⁽¹³⁾ and have associated increases in mortality. While safe limits of alcohol use have not been readily defined, the American Association for the Study of Liver Disease ALD guidelines recommend that no consumption of alcohol is safe in those with any degree of liver disease.⁽¹⁴⁾ Similarly, the European Association for the Study of Liver Disease also recommends complete abstinence for those with alcohol-associated cirrhosis.⁽¹⁵⁾ Earlier stages of ALD represent a grayer area. For those with alcohol-associated steatosis or steatohepatitis, without significant fibrosis or cirrhosis, would alcohol reduction be enough? Total abstinence can completely reverse alcohol-associated fatty liver and, perhaps, steatohepatitis; this, coupled with the higher likelihood of complications of cirrhosis and mortality in those with even lesser degrees of ALD, means we should be cautious. Thus, at present, complete abstinence remains the appropriate recommendation for primary alcohol outcomes in trials with non-advanced ALD, although alcohol reduction or PNHDD could be used as a secondary outcome.

How Should We Measure Alcohol Use in Clinical Trials?

Measuring what constitutes a drink is challenging. Simply asking patients how much they are drinking now or in the recent past can be inaccurate, with patients frequently underestimating drinking amounts.⁽¹⁶⁾ Trials of relapse-prevention medications or behavioral interventions in the general AUD population without liver disease require robust and precise measurement of alcohol use. The choice of alcohol measurement tool depends on the level of granularity required in the study. For most alcohol studies, the Timeline Follow-back (TLFB) questionnaire is the gold standard for measurement of alcohol use.⁽¹⁷⁾ TLFB can be a time-consuming process but produces highly granular detail about drinking patterns. A trained member of research staff asks the subject to recall each day’s level of drinking over a preset timeframe (typically 1 week or 1 month). The total number, volume, and type of alcoholic beverages are recorded. However, if total abstinence is the primary outcome and granularity of alcohol consumption is not desired, TLFB may not be necessary. FDA recommendations allow for simpler questionnaires to assess alcohol consumption if granular alcohol detail is not required, such as the Alcohol Use Disorders Identification Test (AUDIT), a brief 3 to 10 question survey assessing alcohol use.⁽¹⁸⁾ ALD researchers should determine how much granularity they would like for the alcohol outcome and choose a measure, such as TLFB or AUDIT, accordingly.

Surveys or questionnaires based on direct patient recall of alcohol use are widely used and reasonably accurate, although inaccurate estimates of drink size may call this into question.⁽¹⁶⁾ Another option to confirm or refute abstinence are alcohol biomarkers. Direct alcohol biomarkers are urine, blood, or hair tests that directly measure metabolites of alcohol. Their use is recommended by the American Society of Addiction Medicine in alcohol recovery efforts. The American Association for the Study of Liver Diseases ALD guidance recommends the use of alcohol biomarkers to aid in diagnosis of alcohol use, particularly given that there is frequently a high (30% or greater) discordance between biomarker positivity and clinical disclosure of alcohol use.⁽¹⁹⁾ Urine ethyl glucuronide and urine ethyl sulfate are breakdown products of alcohol consumption that are excreted in the urine and detectable 3-5 days after last use. They do not appear to be affected by the presence of liver disease, but are renally metabolized, meaning that they may be positive for a longer period after alcohol consumption in those with some degree of renal dysfunction. There are also false positives and false negatives, of which researchers and clinicians must be aware. Phosphatidylethanol is a lipid moiety produced by phospholipase D1 on red blood cells in the presence of alcohol.⁽²⁰⁾ It detects alcohol consumption 2-3 weeks prior and also does not appear to be affected by liver disease, although there have been fewer studies in patients with advanced liver disease compared to urinary ethyl glucuronide.⁽²⁰⁾ Alcohol biomarker use is recommended in clinical trials in patients with ALD as an aid to determine whether subjects are abstinent and is of particular benefit in ALD trials in which total abstinence is the preferred alcohol-related outcome. Any discordance of patient-reported consumption and empirically measured sobriety via biomarkers may constitute, in and of itself, a secondary outcome of interest to researchers, given the importance of insight, communication, and alliance between patients and clinicians. Finally, novel alcohol detection technology, such as wearable biosensors and smartphone tracking devices, are being developed and tested.^(21,22) Wrist-worn sensors work through detection of transdermal alcohol concentrations, taking advantage of the approximate 1% of alcohol consumed that is excreted through the skin via sweat. Although promising, these technologies are still in early stages of development and have not been adopted widely outside of law enforcement (for earlier models of skin alcohol sensors) and research. However, they offer hope for accurate, moment-to-moment monitoring of alcohol use, which may offer more opportunity for real-time, granular alcohol use data in clinical trials.

Final Considerations: Biological Sex and Other Endpoints in Addition to Alcohol Consumption

Critical to any clinical trial in ALD are issues of sex differences. Alcohol consumption patterns, alcohol use treatment access, abstinence and relapse rates, and alcohol-use treatment-response rates all vary between men and women.^(9,23,24) In addition, alcohol’s impact on liver function varies by sex, with women being more susceptible to alcohol’s effects compared with men. Women are, on average, more likely to get advanced ALD with the same amount of alcohol use as men, have more severe alcoholic hepatitis when this occurs, and are more likely to die of alcohol consumption when advanced ALD is present than men drinking at comparable levels.^(6,25) As such, evaluation of alcohol consumption must be stratified by sex whenever possible. This is particularly important in studies of early, noncirrhotic ALD, when seemingly insignificant variation in alcohol use between men and women (e.g., one to two drink variation) actually portends larger clinical differences in outcomes.⁽⁶⁾

Although alcohol consumption should remain the most important alcohol-related outcome in ALD studies, an AUD diagnosis involves much more than alcohol consumption. AUD diagnoses require an accumulation of symptoms across several domains, including parameters of physiology, impaired control, social impairment, and hazardous use.⁽²⁶⁾ Measuring outcomes in these domains may also be important targets for clinical trials. Outcomes measuring reduction in alcohol use, such as those measured by PNHDD or WHO risk-level reductions, may more accurately reflect patient goals and motivation and could provide important information about how less drinking, as opposed to total abstinence, may still benefit patients with ALD. Other secondary outcomes of interest include improvements in AUD insight, mental health as measured by psychometrics (such as the PHQ9, GAD7, Short Inventory of Problems, and others), social functioning, interpersonal relationships, quality of life as well as AUD treatment engagement and retention, reduction in polysubstance use, and regained sobriety following relapse.

Conclusions

Given the negative effects of any alcohol use on liver-related outcomes, total abstinence remains the recommended primary alcohol-related outcome in clinical trials involving patients with ALD. Secondary outcomes such as reduction in drinking levels or improved mental health and quality of life should be considered alongside total abstinence and will give meaningful information as well. Improved alcohol-use measurement and accurate outcome definitions in clinical trials for ALD promise to increase the validity of results in a much-needed but undervalued research field.

Abbreviations:

AAH

alcohol-associated hepatitis

ALD

alcohol-associated liver disease

AUD

alcohol use disorder

CI

confidence interval

FDA

Food and Drug Administration

PNHDD

percent nonheavy drinking days

TLFB

Timeline Follow-back

WHO

World Health Organization