Reductions in World Health Organization Risk Drinking Level are Associated with Reductions in Alcohol Use Disorder Diagnosis and Criteria: Evidence from an Alcohol Pharmacotherapy Trial

Abstract

Objectives:

To evaluate the validity of World Health Organization (WHO) risk drinking level reductions as meaningful endpoints for clinical practice and research, this study examined whether such reductions were associated with a lower likelihood of a current alcohol use disorder (AUD) diagnosis and fewer AUD criteria.

Methods:

We conducted a secondary data analysis to address these objectives using data from a multi-site randomized controlled trial of gabapentin enacarbil extended-release (GE-XR) in treating moderate to severe AUD among adults (N=346). Participants received GE-XR or placebo for 6 months. The Timeline Follow-back was used to assess WHO risk drinking level reductions and the Mini-International Neuropsychiatric Interview was used to assess DSM-5 AUD diagnosis and criteria at baseline (past year) and end of treatment (past month).

Results:

Most participants (80.1%) achieved at least a 1-level reduction in the WHO risk drinking levels from baseline to end of treatment and nearly half of participants (49.8%) achieved at least a 2-level reduction. At least a 1-level reduction or at least a 2-level reduction in WHO risk drinking level predicted lower odds of an active AUD diagnosis (1-level: OR=0.74, 95% CI [0.66, 0.84]; 2-level: OR=0.71, 95% CI [0.64, 0.79]) and fewer AUD criteria (1-level: B=−1.66, 95% CI [−2.35, −0.98]; 2-level: B=−1.76, 95% CI [−2.31, −1.21]) at end of treatment.

Conclusions:

WHO risk drinking level reductions correlate with DSM-5 AUD diagnosis and criteria, providing further evidence for their use as endpoints in alcohol intervention trials, which has potential implications for broadening the base of AUD treatment.

Introduction

Alcohol use disorder (AUD), one of the most common psychiatric disorders, has past-year and lifetime prevalences in the United States (U.S.) of 13.9% and 29.1%, respectively.¹ AUD is associated with substantial disability and comorbidity¹, poor psychosocial functioning², and high economic costs.³ Although many people diagnosed with AUD recover without treatment⁴, the modest efficacy⁵ and large gaps in the reach of existing treatments⁶ necessitate further treatment development. A critical need in advancing AUD treatment development is the validation of clinical trial endpoints that reflect meaningful reductions in drinking, short of total abstinence.^7,8 This reflects the fact that most people in AUD treatment neither seek nor achieve complete abstinence, yet are able to achieve meaningful reductions in drinking and improvements in functioning.⁹ Endpoints that are associated with reductions in AUD criteria are particularly needed as current research definitions of AUD recovery (e.g., that of the National Institute on Alcohol Abuse and Alcoholism) often include remission of AUD criteria as an essential condition for recovery.¹⁰

A historic misconception is that abstinence is necessary to recover from AUD, which has made abstinence the dominant endpoint used in clinical trials of AUD treatments.¹¹ This misconception, however, is refuted by decades of research¹² demonstrating that drinking reductions are associated with reductions in morbidity and mortality, lower healthcare costs, and improved psychosocial functioning¹³, and that there are no differences in the effects of treatment for AUD with abstinence versus non-abstinent goals.¹⁴ Importantly, many people who could benefit from AUD treatment prefer not to completely abstain from alcohol, thus reduced drinking as a viable treatment goal could broaden the base of those seeking AUD treatment.¹⁵ Consistent with this evidence, the U.S. Food and Drug Administration (FDA)¹⁶ and European Medicines Agency (EMA)¹⁷ now accept non-abstinent outcomes (low-risk [including no heavy drinking days] and reduced drinking, respectively) as endpoints in clinical trials of pharmacological treatments for AUD. At least a 2-level reduction in the World Health Organization (WHO) risk drinking levels is approved by the EMA as an intermediate harm reduction outcome and a growing body of research supports the clinical validity of the WHO measure¹⁸, including its comparative utility to abstinence and no heavy drinking days.¹⁹

Four WHO risk drinking levels are defined by the number of grams of pure alcohol consumed per day: low, medium, high, and very high risk (Table 1).¹⁷ A growing body of research shows that at least a 1-level, but clearly at least a 2-level, reduction in WHO risk drinking levels during treatment for AUD and in the general U.S. population are associated with fewer negative alcohol-related consequences, fewer mental health symptoms, lower likelihood of psychological disorders (i.e., other substance use disorders, anxiety, and depression), reduced risk for cardiovascular disease, better liver function, and other indicators of greater mental and physical health.^18,20–24 Further, reductions in WHO risk drinking level during AUD treatment are maintained for one year posttreatment by most people, and the maintenance of these reductions are associated with continued improvements in functioning.^25,26 Although these findings strongly support the validity of 1- and 2-level reductions in WHO risk drinking levels as endpoints in clinical trials of AUD treatment reflective of psychosocial functioning, less research has examined their clinical validity in relation to AUD diagnosis and criteria. Further evidence for the criterion validity of 1- and 2-level reductions in WHO risk drinking levels with AUD diagnosis and criteria would support the adoption of these as meaningful endpoints in alcohol clinical trials that are inclusive of non-abstinence.

Table 1.

Definitions and frequencies of World Health Organization (WHO) risk drinking levels at baseline (N=346) and end of treatment (N=281) and reductions in WHO risk drinking levels from baseline to end of treatment among the observed sample.

WHO Risk Drinking Level (males/females)	Baseline: n (%)	End of treatment: n (%)
Abstinent (0 g)	0 (0%)	30 (11%)
Low risk (1–40 g/1–20 g)	0 (0%)	87 (31%)
Medium risk (41–60 g/21–40 g)	18 (5%)	61 (22%)
High risk (61–100 g/41–60 g)	99 (29%)	58 (21%)
Very high risk (101+ g/61+ g)	229 (66%)	45 (16%)
WHO Risk Drinking Level Reductions	Baseline to End of Treatment: n (%)
Increase	48 (17%)
No change	8 (3%)
1-level reduction	85 (30%)
2-level reduction	66 (23%)
3-level reduction	56 (20%)
4-level reduction	18 (6%)

To our knowledge, no studies have examined whether WHO risk drinking reductions are associated with AUD diagnosis and criteria in a clinical sample of individuals seeking treatment for AUD, and this gap may hinder treatment research and adoption in clinical practice. Reductions in AUD criteria represent meaningful outcomes that more directly reflect patient functioning than abstinence or other alcohol consumption-based outcomes.²⁷ Research examining the relationships between WHO risk drinking levels and AUD diagnosis/criteria have been limited to non-treatment seeking populations. For example, using data from the National Epidemiologic Survey on Alcohol and Related Conditions (NESARC), Shmulewitz and colleagues²⁴ found increasing prevalence of Diagnostic and Statistical Manual, 4^th edition (DSM-IV) diagnosis of alcohol dependence across increasing WHO risk drinking levels, and Hasin and colleagues²⁸ found that 1- to 3-level reductions in WHO risk drinking levels from 2001–2002 (Wave 1) to 2004–2005 (Wave 2) were associated with a lower likelihood of DSM-IV alcohol dependence diagnosis at Wave 2. However, these population-based findings may not generalize to treatment-seeking populations, who often drink more heavily and have more severe AUD than the general U.S. population. Replication and extension of general population findings using data from an AUD clinical trial are needed.

Overview of the present study

We conducted a secondary data analysis from a clinical trial of a novel pharmacotherapy for AUD to further test the validity of at least 1- and 2-level reductions in WHO risk drinking levels with AUD diagnosis and criteria. The focus on 1- and 2-level reductions is consistent with EMA guidelines, which define treatment responders as individuals who achieve at least a 2-level reduction (thus a non-responder would not achieve those reductions), and our prior research demonstrating that at least a 1-level reduction is related to clinically meaningful outcomes. More specifically, we tested our hypotheses that at least 1- and 2-level reductions in WHO risk drinking levels during treatment would be associated with a lower likelihood of an active AUD diagnosis and fewer AUD criteria at end of treatment.

Methods

This current secondary data analysis was based on a randomized, double-blind, placebo-controlled, multisite trial of gabapentin enacarbil extended-release (GE-XR) (HORIZANT^®; Arbor Pharmaceuticals, LLC, Atlanta, GA) for treating AUD.²⁹ In brief, the parent trial did not find effects of 600 mg of GE-XR twice daily for 6 months on alcohol use, craving, or negative alcohol-related consequences compared to placebo among patients with moderate or severe AUD at baseline. Participants in both conditions also received a computerized behavioral intervention.

Participants and procedures

Participants were 346 treatment-seeking individuals recruited from 10 U.S. academic sites. Participants met criteria for moderate or severe AUD (4+ criteria) in the past year as defined by the DSM-5. Eligibility criteria were: 1) ≥ 21 years of age; 2) ≥ 21/28 standard drinks per week for women/men and at least one heavy drinking occasion in the 28 days prior to consent; 3) 3 consecutive days of abstinence prior to randomization; 4) not diagnosed with a substance use disorder, other than alcohol or nicotine, or a major psychiatric disorder; 5) no medical condition contraindicated or exacerbated by gabapentin; 6) no use of psychiatric medications other than stable use of antidepressants. Research interviewers collected drinking measures and DSM-5 AUD criteria at baseline and end of treatment (28–29 week follow-up, or 1–2 weeks after the last in-clinic visit). Participants had a mean age of 49.75 (SD=10.99). The sample was 65% male, 72% White, 19% Black, 10% Hispanic or Latino, 5% Multiracial, 2% “unknown” or not reported, 1% American Indian/Alaska Native, and 1% Asian. Additionally, the sample had a mean of 15.25 years of education (SD=2.64) and 76% were employed. Most participants (75%) met criteria for severe AUD. Further details regarding the eligibility criteria, study procedures, and sample description are reported in the primary outcomes paper.²⁹

Measures

WHO risk drinking level.

WHO risk drinking levels were computed using standard procedures.^18–26,28 Average grams of pure alcohol consumed per day in the past 28 days at baseline (past month) and end of treatment (weeks 22 to 25) were calculated based on data from the Timeline Follow-back method.³⁰ Participants were classified into WHO risk drinking levels based on their average daily alcohol consumption at baseline and end of treatment (see Table 1). Changes (or lack thereof) in the WHO risk drinking levels from baseline to end of treatment were categorized to reflect WHO risk level increase, no change, 1-level reduction, 2-level reduction, 3-level reduction, or 4-level reduction. For regression analyses, the primary predictors were two binary variables that reflected at least a 1- or 2-level reduction in the WHO risk drinking levels from baseline to end of treatment.

AUD criteria and diagnosis.

The Adult Mini-International Neuropsychiatric Interview (MINI)³¹ 7.0—a short, valid, and reliable^32,33 structured interview—was used to assess DSM-5 AUD criteria and diagnosis (“active diagnosis” if 2+ criteria present). At baseline, AUD criteria and diagnosis were assessed over a past-year timeframe; at end of treatment, they were assessed over a past-month timeframe. For the current study, AUD criterion counts were used to operationally define “AUD severity,” consistent with current conceptualizations of AUD as unidimensional with a continuum severity that is reflected by the number of criteria present.²⁷

Statistical Analyses

Logistic and linear regression analyses were conducted to predict AUD diagnosis (binary outcome) and criteria (continuous outcome) based on WHO risk drinking reductions (focal predictors) and additional covariates among the entire study sample (i.e., participants assigned to both treatment arms). Separate models were conducted with 1- and 2-level WHO risk drinking reductions as focal predictors. The reference group for the 1-level reduction analyses comprised patients with no change or increases in WHO risk drinking levels from baseline to end of treatment; the reference group for the 2-level reduction analyses comprised patients with a 1-level reduction, no change, or an increase in WHO risk levels from baseline to end of treatment. As mentioned above, these reference groups were chosen to be consistent with prior work testing 1- and 2-level reductions, and EMA guidelines that define treatment responders as individuals who achieve at least a 2-level reduction (thus a non-responder is someone who achieves a 1-level reduction, no change, or an increase in drinking risk level). Additional supplemental models were tested with categories representing every 1-level change in WHO risk drinking levels entered as focal predictors to understand if higher risk-level reductions were associated with increasingly lower odds of AUD diagnosis and lower mean AUD criteria counts.

All analyses included the following covariates: treatment condition, baseline AUD criterion count, baseline WHO risk drinking level, sex, and age. Results are reported as adjusted odds ratios (for binary outcomes) or unstandardized regression coefficients (for continuous outcomes).

No participants were missing data at baseline; however, 18.8% had missing WHO risk drinking level and AUD diagnosis/criteria at end of treatment. Missing data were handled using one primary method and two alternative methods as sensitivity analyses. For our primary method, we report the results using multiple imputation of all missing data using chained equations with predictive mean matching.³⁴ For these analyses, 20 imputation datasets were generated, analyzed, and pooled. For our alternative methods, we report the results using multiple imputation of AUD diagnosis/criteria outcomes but assume that participants who were missing end of treatment drinking data had no changes in WHO risk drinking level from baseline. This method for handling missing data has generally been discouraged due to the substantial bias it introduces^35–37, but was included here as a sensitivity analysis. Also, as an alternative method and sensitivity analysis, shown in the Supplementary Materials, we estimated the models using multiple imputation of the clinical outcome and listwise deletion (i.e., complete case analysis) of individuals with missing data on the WHO risk level at the end of treatment. Finally, we also conducted sensitivity analyses to ascertain whether number of baseline AUD criteria moderated the effects. For the analyses examining baseline AUD criterion count as a moderator, baseline AUD criterion count was mean centered and its product with the binary 1- or 2-level reductions variables was added to the models.

All analyses were conducted using R 4.2.2³⁸ with the gtsummary³⁹ and mice packages.³⁴

Results

Descriptive statistics of WHO risk drinking levels

As reported in Table 1, most individuals reduced their drinking during treatment, with the majority achieving at least a 1-level reduction (80.1%) and many achieving at least a 2-level reduction (49.8%). The entire sample met AUD criteria at baseline with an average of 7.45 (SD=2.13) criteria. At end of treatment, of those assessed (N=281), 183 (65.1%) met criteria for AUD with an average of 3.11 (SD=2.56) criteria.

Table 2 provides the descriptive statistics for DSM-5 AUD diagnosis and criteria by the binary 1- and 2-level WHO risk drinking reduction variables. As shown, reductions in binary WHO risk levels were associated with lower rates of AUD diagnosis and lower average AUD criteria than in the reference groups. For AUD diagnosis, 61% of the sample with a 1-level or greater reduction met criteria for AUD (mean AUD criteria=2.79, SD=2.54) versus 91% of those with no change or an increase in WHO risk level (mean AUD criteria=4.33, SD=2.27). For a 2-level reduction, 51% of individuals with a 2-level or greater reduction met criteria for AUD (mean AUD criteria=2.30, SD=2.52) versus 83% of those with a 1-level reduction, no change or an increase in WHO risk level (mean AUD criteria=3.90, SD=2.36). Descriptive statistics for each WHO risk drinking level are reported in Supplementary Table 1.

Table 2.

At least 1- and 2-level reductions in World Health Organization (WHO) risk drinking level and alcohol use disorder (AUD) diagnosis and number of criteria positive at end of treatment with and without multiple imputation.

Table 2a. At least 1-level reduction and AUD diagnosis and criteria at end of treatment
	Without Imputation (N=281)		With Imputation* (N=346)
DSM-5 AUD	Increase/no change (n=56)	1-level reduction + (n=225)	Increase/no change (n=121)	1-level reduction + (n=225)
Diagnosis (n [%])	49 (91%)	134 (61%)	56 (89%)	134 (61%)
No. of Criteria (M [SD])	4.33 (2.27)	2.79 (2.54)	4.19 (2.31)	2.79 (2.54)
Table 2b. At least 2-revel reduction and AUD diagnosis and criteria at end of treatment
	Without Imputation (N=281)		With Imputation* (N=346)
DSM-5 AUD	Increase/no change/1-level reduction (n=141)	2-level reduction + (n=140)	Increase/no change/1-level reduction (n=206)	2-level reduction + (n=140)
Diagnosis (n [%])	113 (83%)	70 (51%)	120 (83%)	70 (51%)
Criteria (M [SD])	3.90 (2.36)	2.30 (2.52)	3.87 (2.36)	2.30 (2.52)

^*Note. “Worst case imputation” is applied to participants missing all drinking data during the last 4 weeks of the treatment period. These individuals missing all drinking data were included in the referent group (non-responder; e.g., no change/increase/1-level reduction) for the WHO Risk Drinking Level reduction.

Table 3 shows that binary WHO risk drinking level endpoints were significantly associated with end of treatment AUD diagnosis, with at least a 1-level reduction (OR=0.74; 95% CI: 0.66, 0.84; p < 0.001) and at least a 2-level reduction (OR=0.71; 95% CI: 0.64, 0.79; p < 0.001) associated with lower odds of meeting contemporaneous AUD diagnostic criterium. Individuals who achieved at least a 1-level reduction (compared to increase or no change) were 26% less likely to meet criteria for AUD at end of treatment, and those who achieved at least a 2-level reduction (compared to increase, no change, or a 1-level reduction) were 29% less likely to meet criteria for AUD at end of treatment, holding other covariates constant. Table 4 shows individuals who achieved at least a 1-level reduction (compared to an increase or no change) met 1.66 fewer AUD criteria at end of treatment (95% CI: −2.35, −0.98; p < 0.001), and those who achieved at least a 2-level reduction (compared to increase, no change, or 1-level reduction) met 1.76 fewer AUD criteria at end of treatment (95% CI: −2.31, −1.21; p < 0.001), holding other covariates constant. Supplementary Tables 2 and 3 show the results of models with categories representing every 1-level change in WHO risk drinking levels entered simulatenously as focal predictors.¹

Table 3.

At least 1- and 2-level reductions in World Health Organization (WHO) risk drinking levels predicting alcohol use disorder (AUD) diagnosis at end of treatment with multiple imputation (N=346).

Predictor	OR	95% CI	p-value	Predictor	OR	95% CI	p-value
1-level + reduction a	0.74	0.66, 0.84	<.001	2-level + reduction e	0.71	0.64, 0.79	<.001
Baseline covariates				Baseline covariates
High risk drinking levelb	1.04	0.82, 1.32	.74	High risk drinking levelb	1.15	0.90, 1.46	.26
Very high risk drinking levelb	0.99	0.78, 1.24	.90	Very high risk drinking levelb	1.11	0.87, 1.40	.40
AUD criteria	1.05	1.02, 1.07	<.001	AUD criteria	1.05	1.02, 1.07	<.001
HORIZANT conditionc	1.03	0.93, 1.15	.51	HORIZANT conditionc	1.02	0.92, 1.13	.73
Age	1.00	0.99, 1.00	.14	Age	1.00	0.99, 1.00	.11
Male d	0.88	0.79, 0.98	.02	Maled	0.92	0.83, 1.03	.14

Note. Boldface indicates statistical significance (p < .05). OR=Odds ratio, CI=Confidence interval.

^aReferent group is increase or no change in World Health Organization risk drinking level.

^bReferent group is World Health Organization moderate risk drinking level.

^cReferent group is placebo (control) condition.

^dReferent group is female.

^eReferent group is increase, no change, or 1-level reduction in World Health Organization risk drinking level.

Table 4.

At least 1- and 2-level reductions in World Health Organization (WHO) risk drinking levels predicting alcohol use disorder (AUD) criterion count at end of treatment with multiple imputation (N=346).

Predictor	B	95% CI	p-value	Predictor	B	95% CI	p-value
1-level + reduction a	−1.66	−2.35, −0.98	<.001	2-level + reduction e	−1.84	−2.39, −1.29	<.001
Baseline covariates				Baseline covariates
High risk drinking levelb	0.87	−0.39, 2.13	.18	High risk drinking level b	1.43	0.12, 2.73	.03
Very high risk drinking levelb	0.88	−0.36, 2.12	.16	Very high risk drinking level b	1.55	0.31, 2.79	.01
AUD criteria	0.40	0.26, 0.53	<.001	AUD criteria	0.40	0.27, 0.53	<.001
HORIZANT condition c	0.63	0.10, 1.15	.02	HORIZANT conditionc	0.55	−0.02, 1.12	.06
Age	−0.02	−0.05, 0.01	.14	Age	−0.02	−0.05, 0.00	.08
Maled	−0.32	−0.93, 0.28	.16	Maled	−0.19	−0.79, 0.42	.55

Note. Boldface indicates statistical significance (p < .05). B=unstandardized coefficient., CI=Confidence interval.

^aReferent group is increase or no change in World Health Organization risk drinking level.

^bReferent group is World Health Organization moderate risk drinking level.

^cReferent group is placebo (control) condition.

^dReferent group is female.

^eReferent group is increase, no change, or 1-level reduction in World Health Organization risk drinking level.

Results did not substantively differ using worst case scenario imputation or complete case analysis and also did not differ when adding AUD severity at baseline as a moderator (see Supplementary Tables 4–9).

Discussion

To establish the clinical validity of 1- and 2-level reductions in WHO risk drinking level as endpoints in clinical trials for AUD it is important to demonstrate thatachieving these endpoints will provide symptom relief from AUD. The current study found reductions in WHO risk drinking levels during treatment were significantly associated with a lower likelihood of meeting contemporaneous AUD diagnosis and fewer AUD criteria at end of treatment. At least 1- and 2-level reductions were associated with a 26% and 29% lower likelihood of meeting criteria for AUD diagnosis, respectively, and 1.66 and 1.76 fewer AUD criteria, respectively, at the end of treatment. Importantly, these results did not differ by the number of AUD criteria present at baseline, suggesting that participants with moderate or severe AUD were equally likely to benefit from reductions in WHO risk drinking levels. Sensitivity analyses that used different methods to address missing data indicated similar results, lending support to the robustness of these findings.

These findings are consistent with the growing body of evidence demonstrating the clinical validity of reductions in WHO risk drinking level.^18–26,28 More specifically, although a 2-level reduction had larger associations than a 1-level reduction, at least a 1-level or greater reduction is associated with clinically meaningful improvement, in this case, a lower likelihood of current AUD diagnosis and fewer AUD criteria. Two prior studies found that reductions in WHO risk drinking level were associated with a lower prevalence of alcohol dependence in the general U.S. population of drinkers.^24,28 An important contribution of the present study is that it extends these findings to a clinical context, specifically a pharmacotherapy trial for AUD. Reductions in WHO risk drinking level have been proposed as endpoints in AUD treatment trials, which has largely motivated the body of research examining their clinical validity. Reductions in WHO risk drinking levels are less stringent than the currently accepted endpoints guided by the FDA (i.e., abstinence or no heavy drinking days) in that they capture improvements that might otherwise be considered treatment “failures.” Thus, reductions in WHO risk drinking levels could enhance treatment delivery by capturing meaningful improvements in drinking in alcohol clinical trials that the current FDA-guided endpoints do not capture. Because most patients do not achieve absintence or no heavy drinking days, their treatment would appear to have failed despite meaningful reductions in drinking and improved behavioral and physical function. Alcohol treatment researchers could utilize validated outcome measures that reflect meaningful drinking reductions, short of complete abstinence, to help define patients’ goals and experiences more realistically and reduce the stigma associated with non-abstinence or intermittent heavy drinking. Our replication of the associations of reductions in WHO risk drinking level with a lower likelihood of meeting contemporaneous AUD diagnosis and fewer AUD criteria in an AUD pharmacotherapy trial provides support for the clinical validity of these endpoints. In addition, clinicians, third party payors, and interested other entities (e.g. forensic, employee assistance programs) look to AUD diagnostic criteria and coding to evaluate treatment necessity and efficacy. Therefore, extending the WHO risk drinking level metric into this dimension would have important practical implications.

Beyond further evidence that reductions in WHO risk drinking level are valid endpoints, these findings have other important clinical implications. Findings of a lower likelihood of meeting current criteria for AUD of 26% and 29% and 1.66 and 1.76 fewer AUD criteria for 1- and 2-level reductions, respectively, are substantial. Further support for the clinical validity of endpoints that are inclusive of non-abstinence is consistent with most people’s goals to reduce their drinking short of total abstinence during treatment¹⁵ and could thereby increase treatment seeking and receptivity. In other words, knowing that reductions in drinking, which reflect most people’s goals, are associated with reductions in AUD symptoms could help to motivate people to seek treatment. For the same reason, clinicians may be more likely to accept reductions in WHO risk drinking levels as meaningful outcomes that confer lower rates of AUD diagnosis, including using a 1-level reduction as partial success if even greater reductions are contemplated.

Limitations

There are several limitations of the present study. First, it was a secondary data analysis, and the original trial was not designed to test our hypotheses. The sample was required to meet at least 4 criteria of AUD at baseline which restricted the range of AUD criteria. Future research with larger samples representing a broader range of AUD criteria are needed to replicate our findings. Second, the causal inference (i.e., reductions in WHO risk drinking levels caused reductions in AUD criteria/diagnosis from AUD) is limited by the non-experimental, observational design of the present study. Third, the sample was predominantly male, employed, and non-Hispanic white, limiting generalizability of our findings. Research is needed to replicate these findings in more sociodemographically and economically diverse samples. Fourth, it is recognized that a one month period at the end of the clinical trial is not consistent with the DSM-5 standard of a 12 month period of AUD assessment. Future studies should assess the stability of the reduction of AUD criteria and diagnosis over longer posttreatment periods in relationship to WHO risk drinking level criteria at study end. Fifth, we did not separately investigate whether the AUD craving criterion was endorsed, which is important to consider when evaluating AUD remission (given that craving can be endorsed as a criterion under the DSM-5 definition of remission). Finally, while the parent study included negative blood phosphatidylethanol (PEth) as an exploratory outcome, we did not consider biochemical verification of alcohol use, and future studies that replicate these findings with more objective measures of alcohol use are needed.

Conclusions

The present study extends evidence for the clinical validity of reductions in WHO risk drinking level in relation to AUD diagnosis and criteria in a clinical trial of a novel pharmacotherapy for AUD. Our findings add to the growing evidence that reductions in WHO risk drinking levels of 1-level or greater reflect meaningful reductions in drinking that correspond to a host of clinical improvements, including reductions in AUD criteria and a greater likelihood of no longer meeting criteria for AUD. Adoption of at least 1- and 2-level reductions as endpoints in AUD treatment trials could broaden access to and use of AUD treatment by appealing to the majority of individuals with AUD who prefer not to abstain entirely from alcohol. Further, adoption of these endpoints by U.S. regulatory agencies and elsewhere could help to standardize outcomes across AUD treatment trials internationally. For instance, the EMA already acknowledges the use of these endpoints. Further international adoption could increase the cumulative knowledge regarding the treatment of AUD and, ultimately, may improve health and well-being globally.