Association of under-reporting of alcohol use with HIV viral non-suppression and an all-cause mortality risk score among people with HIV on antiretroviral therapy

Abstract

Background.

Unhealthy alcohol use contributes to poor HIV-related outcomes. Alcohol under-reporting may be associated with HIV viral non-suppression and all-cause mortality risk, especially at the high end of the spectrum. We evaluated whether alcohol under-reporting was associated with these outcomes among PWH with unhealthy alcohol use who were receiving ART.

Methods.

We pooled data from six studies of PWH in Uganda (ADEPT-T, DIPT, Extend), United States (MWCCS), and Russia (Russia ARCH 1, St Peter). We included PWH on ART who engaged in unhealthy drinking as determined by the alcohol biomarker phosphatidylethanol (PEth)≥50ng/mL. Self-reported alcohol use was assessed using the Alcohol Use Disorders Identification Test–Consumption (AUDIT-C). We defined under-reporting of high-risk alcohol use as PEth≥200ng/mL and AUDIT-C score<6, and examined HIV viral non-suppression (≥200copies/mL or limit of assay detection) and an all-cause mortality risk score (Veterans Aging Cohort Study [VACS] Index 1.0) as outcomes. We explored level of -under-reporting of alcohol use using a continuous variable.

Results.

In our sample of 1,435 PWH with unhealthy drinking, 24% under-reported high-risk alcohol use (prevalence range 1–34% across studies). We observed positive, non-significant associations between under-reporting of high-risk alcohol use and viral non-suppression and all-cause mortality risk score. An exploratory continuous under-reporting variable was significantly positively associated with all-cause mortality risk.

Conclusions.

Our findings indicate that under-reporting of high-risk alcohol use is common among PWH and suggest that under-reporting of alcohol use may mask other health issues, as reflected by increased mortality index scores. Low-cost methods to improve alcohol use are needed.

Introduction

Unhealthy alcohol use (i.e., alcohol use that increases the risk for or has already caused health consequences, comprising the range from exceeding recommended limits to severe alcohol use disorder¹) is common among people with HIV (PWH) and plays an influential role in accelerating HIV disease progression and detrimentally impacting many steps of the HIV care continuum.^2–6 Unhealthy alcohol use contributes to a plethora of negative health outcomes including: reduced ART adherence,^7–9 HIV disease progression^10–12 (e.g., HIV viral non-suppression^4,5), HIV transmission through high-risk sexual behaviors,^13,14 various comorbidities (e.g., liver disease including viral hepatitis,¹⁵ cancer,^16,17 cardiovascular disease¹⁸), and mortality.^19,20

PWH may under-report their alcohol use due to various factors including HIV- and alcohol-related stigma, recall issues, and social desirability bias.²¹ Moreover, in clinical settings PWH may minimize their alcohol consumption if their HIV healthcare providers discourage it, given its known effects on HIV disease progression and HIV-related risk behaviors.²² Several studies have described under-reporting of alcohol use among PWH in global settings,^21,23–30 with a prevalence ranging from 12% to 45% among those self-reporting alcohol abstinence but having PEth values indicative of recent alcohol exposure.^{23–25,27–30}

Among PWH, under-reporting of alcohol use (especially high-risk alcohol use indicated by an Alcohol Use Disorder Identification Test – Consumption [AUDIT-C] score of 6 or higher¹) may increase vulnerability to adverse outcomes. Under-reporting of alcohol use by PWH at high risk can negatively impact receipt of appropriate medical care and linkage to alcohol use disorder treatment. A notable study among mostly male US veterans with and without HIV found that PWH under-reporting alcohol consumption (i.e., those self-reporting abstaining from alcohol but having detectable PEth levels) had a high 5-year mortality rate close to the mortality rate of those self-reporting the highest level of alcohol use (mortality rate of 5.69 per 100 person-years compared to 6.12 per 100 person-years).²⁸ We are unaware of any other studies examining the association of under-reporting of alcohol use with poor health outcomes in other populations or settings.³¹

Considering that PWH who under-report high-risk alcohol use may have heightened risk for poor health outcomes above and beyond the risks of unhealthy alcohol use,⁶ we pooled data from six studies to examine whether under-reporting of high-risk alcohol use was associated with HIV viral non-suppression or all-cause mortality among PWH engaging in unhealthy alcohol use in various settings. We hypothesized that under-reporting of high-risk alcohol use would be positively associated with HIV viral non-suppression and an all-cause mortality risk score (i.e., the Veterans Aging Cohort Study [VACS] Index 1.0)

Methods

We pooled data from six studies conducted among PWH on ART, as described below. We restricted the samples to PWH on ART for at least 6 months, and visits at which they had evidence of unhealthy alcohol use, (i.e., PEth ≥ 50 ng/mL²³) which reduced the sample size from the parent studies.

Study Population and Design

The Alcohol Drinkers’ Exposure to Preventive Therapy for TB (ADEPT-T, NCT03302299) aimed to estimate the incidence of isoniazid-related toxicity among PWH who consume and abstain from alcohol in Uganda, and whether levels of alcohol use resulted in differing isoniazid-related toxicity (N=301).³² Individuals were recruited in an HIV clinic in Mbarara from 2017 to 2020. This analysis included data from the 131 participants with PEth ≥ 50 ng/mL.

The Drinker’s Intervention to Prevent Tuberculosis (DIPT) RCT assessed the efficacy of incentive-based approaches to decrease alcohol consumption and improve adherence to isoniazid medication among PWH and TB on ART reporting unhealthy drinking (N=680, NCT03492216).^33,34 Individuals were recruited at several HIV clinics in southwest Uganda from 2018 to 2021. Participants were randomized to one of four study arms: no incentives (Group 1, control), financial incentives for recent alcohol abstinence (Group 2), financial incentives for recent adherence to isoniazid medication (Group 3), and financial incentives for decreasing alcohol use and increasing isoniazid adherence (Group 4). This analysis included data from the 595 participants with PEth ≥ 50 ng/mL at one or more study visits.

The Extend RCT (NCT03928418) assessed the efficacy of counseling interventions to reduce alcohol use and increase HIV viral suppression among PWH on ART who self-reported unhealthy alcohol consumption in Uganda (N=272).³⁵ Individuals were recruited from 2019 to 2020 in the same HIV clinicl as the ADEPT-T study. Participants were randomized to one of three arms: brief advice (Arm 1, control), in-person counseling with boosters delivered by a counsellor over the phone (Arm 2), or in-person counseling with boosters delivered either through short messaging services (SMS) or interactive voice response (IVR, Arm 3). This analysis included data from the 232 participants with PEth ≥ 50 ng/mL.

The Multicenter AIDS Cohort Study (MACS) / Women’s Interagency HIV Study (WIHS) Combined Cohort Study (MWCCS) is an ongoing study that was established in 2019 by merging two longitudinal studies of men (MACS) and women (WIHS) with and at risk for HIV in the United States (N= 4,141).³⁶ Data collection occurs annually in 13 metropolitan cities. The MWCCS began collecting dried blood spots and testing PEth in the full sample of participants beginning in 2020; this analysis included data from the 352 participants with PEth ≥ 50 ng/mL.

The Russia Alcohol Research Collaboration on HIV/AIDS (Russia ARCH I) was conducted to assess the relationship between high-risk alcohol use and inflammatory markers among PWH in Russia (N=351).³⁷ Individuals were eligible to participate if they were PWH, had never received ART treatment, were aged 18–70 years, had a stable address within St. Petersburg or within 100km of St. Petersburg, could speak Russian, possessed a phone, could provide two contacts for follow-up, and were able to provide informed consent. Potential participants were recruited at clinical and non-clinical sites as well as through snowball sampling in St. Petersburg from 2012 to 2014. This analysis included data from the 14 participants on ART with PEth ≥ 50 ng/mL.

The Studying Partial agonists for Ethanol and Tobacco Elimination in Russians with HIV (St PETER) RCT (NCT02797587) was designed to assess the effects of varenicline, cytisine, and nicotine replacement therapy on alcohol use, smoking, inflammation, coronary heart disease risk, and mortality among PWH reporting unhealthy drinking (≥5 heavy drinking days in the past 30 days) and smoking ≥5 cigarettes per day on average (N=400).³⁸ Recruitment occurred at clinical and non-clinical sites as well as through snowball sampling in St. Petersburg from 2017 to 2020. Participants were randomized to one of four study arms: active varenicline and nicotine replacement therapy placebo (Arm 1), varenicline placebo and nicotine replacement therapy (Arm 2), active cytisine and nicotine replacement therapy placebo (Arm 3), and cytisine placebo and active nicotine replacement therapy (Arm 4). This analysis included data from the 111 participants with PEth ≥ 50 ng/mL.

Study procedures in each of these studies were approved by appropriate institutional review boards (IRB) and participants provided written informed consent.^36,38–43 Our secondary data analysis was considered exempt from additional IRB review.

Data Collection and Lab Testing

Participants enrolled in each of the studies completed interviewer-administered surveys in their local languages. These surveys collected data pertaining to sociodemographics, alcohol use, HIV-related behaviors, and HIV-related health outcomes.

Self-reported alcohol use was assessed using the 3-item AUDIT-C.⁴⁴ For ADEPT-T, DIPT, and Extend, the AUDIT-C recall window was the past 3 months. For MWCCS, the AUDIT-C recall window was the past 12 months; participants were given a score of zero if they reported no alcohol use in the past 3 months. For Russia ARCH I and St PETER, the AUDIT-C recall window was the past 12 months at the baseline visit and the past 6 months at the follow-up visits.

Biomarker-measured alcohol use was measured using PEth (16:0/18:1 homologue), a metabolite formed in the body only in the presence of ethanol that can be used to assess alcohol exposure in the past month.^45–47 In all studies, PEth testing was conducted using dried blood spots from venous blood draws. In MWCCS, PEth testing occurred at the University of Colorado (with a 5ng/mL limit of quantification) as described elsewhere.⁴⁸ For the other studies, PEth testing was conducted at the United States Drug Testing Laboratories (with a 8 ng/mL limit of quantification) as described elsewhere.⁴⁹

HIV viral load testing was conducted in all studies. In the studies in Uganda (i.e., ADEPT-T, Extend, and DIPT), the GeneXpert Dx System was used to test blood specimens for HIV viral load (HIV-1 RNA assay; lower limit of detection of ≤40 copies/mL). In MWCCS, the Roche Cobas HIV-1 PCR test was used to assess HIV viral load from plasma (Quest Diagnostics; lower limit of detection of ≤20 copies/mL). In Russia ARCH I, the AmpliSens^® Soft Monitor FRT 2.3 was used to quantify HIV viral load (lower limit of detection of <500 copies/mL). For St PETER, the STRATEGENE MX 3000 (USA) was used to quantify HIV viral load (lower limit of detection <300 copies/mL).

Statistical Analyses

The analytic plan for this study was preregistered online (https://osf.io/q4j8v). As we had approximately 5% missingness across the covariates in our analyses, we did not conduct any imputation and proceeded with a complete case analysis.

Main Exposure: Under-reporting of Alcohol Use.

To our knowledge, there are no validated assessment tools to quantify under-reporting of alcohol use. We limited our sample to PWH with PEth-confirmed unhealthy drinking (PEth ≥ 50 ng/mL) to examine the incremental effects of under-report beyond unhealthy drinking itself. We classified participants as under-reporting their high-risk alcohol use if PEth was ≥ 200 ng/mL (indicative of “chronic excessive” alcohol use⁴⁵) but their self-reported AUDIT-C score was less than 6 (indicative of no, low-, or medium- risk drinking⁴). Additionally, we explored a continuous variable to represent the degree of difference between PEth and self-report on a continuous scale, to overcome the drawbacks of dichotomous approaches.⁵⁰ To do so, we standardized biomarker-measured PEth and self-reported AUDIT-C scores in our sample into z-scores (subtracting the mean and dividing by the standard deviation); we then created the continuous alcohol under-reporting variable by subtracting the AUDIT-C z-score from the PEth z-score (with positive values indicating under-reporting of alcohol use). We similarly assessed the main exposure for subgroup analyses described below.

Primary Outcome: HIV Viral Non-suppression.

We categorized viral non-suppression as ≥200 copies/mL or the limit of detection (i.e., ≥300 copies/mL and ≥500 copies/mL in St PETER and Russia ARCH I, respectively) at least 6 months after ART initiation.

Secondary Outcome: All-Cause Mortality Risk Score.

We modified the VACS Index 1.0, a widely implemented all-cause mortality risk score, to assess all-cause mortality risk in our pooled sample.^51,52 The VACS Index 1.0 is calculated using multiple components including: age, CD4 cell count, HIV viral load, biomarkers of anemia, liver function, and renal function (i.e., hemoglobin, aspartate aminotransferase, alanine aminotransferase, platelets, creatinine), as well as HCV serostatus. Since data regarding HCV serostatus were not collected in the studies in Russia or Uganda, we removed HCV from the VACS scoring in our sample.

Covariates.

We considered age (years, only in HIV viral non-suppression analyses), gender (men or women), body mass index (BMI, which has been shown to be related to PEth sensitivity⁵³), any other substance use (baseline), any current tobacco use (baseline), integrase strand transfer inhibitor (INSTI)-based ART regimen (yes or no), and study setting (low- or medium-resource setting corresponding to studies in Uganda and Russia, respectively, or high-resource setting for the study in the United States) as a priori covariates in our analyses.

Sample Characteristics.

We generated descriptive statistics to characterize our pooled sample overall and by study using medians and means for continuous variables and frequencies for categorical variables. We also calculated descriptive statistics by under-reporting of alcohol use status (yes versus no).

Under-reporting of Alcohol Use and HIV Viral Non-suppression.

To examine the association of under-reporting of high-risk alcohol use and HIV viral non-suppression among unhealthy drinkers with HIV, we implemented a mixed effects logistic regression accounting for within-individual and between-study correlations and adjusting for age, gender, BMI, any other substance use (baseline), INSTI-based ART regimen, and study setting. We excluded participants who did not have enough time on ART to achieve viral suppression (i.e., those with an HIV viral load blood draw that was within 6 months of ART initiation). We assessed effect heterogeneity by age, gender, and study country, as we theorized that the relationship between under-reporting of alcohol use and HIV viral non-suppression may vary in these particular subgroups. We also conducted sensitivity analyses by different AUDIT-C recall periods and study design (RCT versus observational). We similarly explored under-reporting of alcohol use as a continuous variable as described above.

Under-reporting of Alcohol Use and an All-Cause Mortality Risk Score.

To assess the association between under-reporting of high-risk alcohol use and an all-cause mortality risk score (i.e., VACS Index 1.0) among PWH with unhealthy drinking, we implemented a mixed effects linear regression model accounting for between-study correlations and adjusting for gender, BMI, any current tobacco use (baseline), any other substance use (baseline), INSTI-based ART regimen, and study setting. We did not adjust for age as it is one of the components used to calculate the VACS Index 1.0 (i.e., the outcome). We excluded participants from the Extend RCT and Russia ARCH I as these studies did not collect or had very limited serum creatinine data, which is needed to calculate the VACS Index 1.0. We examined effect heterogeneity by age, gender, and study country. We also conducted sensitivity analyses by AUDIT-C recall periods, study design, and HIV viral non-suppression, and explored under-reporting of alcohol use as a continuous variable as described above.

Results

Sampled Characteristics.

Overall, 1,435 participants (2,595 observations) were included in this analysis (Table 1). Participants were a median of 42 years of age (interquartile range [IQR]= 35 to 52 years), 65% were men, and 67% were from a low-resource setting. The median BMI was 22.5 kg/m² (IQR=19.8 to 26.1 kg/m²), 37% reported current tobacco use at the baseline visit, and 22% reported other drug use at the baseline visit. Median PEth was 246 ng/mL (IQR: 113 – 542 ng/mL), more than half (58%) had PEth values ≥ 200 ng/mL (i.e., “chronic excessive” alcohol use), and 51% self-reported an AUDIT-C score ≥ 6 (i.e., high to very high-level alcohol use). Eight percent of the sample had HIV viral non-suppression, and the median VACS Index 1.0 score was 12 (IQR= 6 to 24).

Table 1.

Participant characteristics overall and by study, at the first visit with PEth ≥ 50 ng/mL, n(%)

	Overall (N=1,435)	Uganda			United States	Russia
		ADEPT-T	DIPT	Extend	MWCCS	Russia ARCH	St Peter
		(n=131)	(n=595)	(n=232)	(n=352)	(n=14)	(n=111)
Age, median [IQR]	42 [35 – 52]	40 [33 – 47]	40 [32 – 47]	40 [34 – 47]	55 [48 – 60]	34 [32 – 37]	39 [36 – 43]
Gender
Women	498 (35)	38 (29)	148 (25)	62 (27)	208 (59)	5 (36)	37 (33)
Men	937 (65)	93 (71)	447 (75)	170 (73)	144 (41)	9 (64)	74 (67)
Race/ethnicity
Hispanic (any race)	41 (12)	-	-	-	41 (12)	-	-
Non-Hispanic Black	233 (66)	-	-	-	233 (66)	-	-
Non-Hispanic White	68 (19)	-	-	-	68 (19)	-	-
Non-Hispanic Other Race	10 (3)	-	-	-	10 (3)	-	-
BMI, median [IQR]	22.5 [19.8 – 26.1]	21.8 [19.6 – 24.6]	20.8 [18.8 – 23.5]	22.6 [20.3 – 25.8]	27.6 [24.4 – 33.1]	22.1 [19.8 – 23.9]	21.6 [19.8 – 24.2]
Any other substance use (baseline)
No	1,119 (78)	125 (95)	543 (91)	220 (95)	147 (42)	8 (57)	76 (69)
Yes	316 (22)	6 (5)	52 (9)	12 (5)	205 (58)	6 (43)	35 (32)
Any current tobacco use (baseline)
No	905 (63)	103 (79)	417 (70)	195 (84)	187 (53)	3 (21)	-
Yes	530 (37)	28 (21)	178 (30)	37 (16)	165 (47)	11 (79)	111 (100)
Study setting (country)
Low-resource (Uganda)	958 (67)	131 (100)	595 (100)	232 (100)	-	-	-
Medium-resource (Russia)	125 (9)	-	-	-	-	14 (100)	111 (100)
High-resource (United States)	352 (25)	-	-	-	352 (100)	-	-
INSTI-based ART regimen
No	690 (48)	97 (74)	294 (49)	64 (28)	110 (31)	14 (100)	111 (100)
Yes	745 (52)	34 (26)	301 (51)	168 (72)	242 (69)	-	-
≥95% ART Adherence
No	434 (30)	33 (25)	226 (38)	105 (45)	47 (14)	2 (14)	21 (19)
Yes	994 (70)	98 (75)	369 (62)	127 (55)	298 (86)	12 (86)	90 (81)
PEth (ng/mL)
Median [IQR]	246 [113 – 542]	244 [124 – 542]	313 [145 – 654]	294 [156 – 588]	170 [100 – 363]	194 [119 – 241]	153 [90 – 274]
50 to <200	608 (42)	57 (44)	209 (35)	79 (34)	190 (54)	8 (57)	65 (59)
≥200	827 (58)	74 (57)	386 (65)	153 (66)	162 (46)	6 (43)	46 (41)
AUDIT-C, median [IQR]	6 [4 – 8]	4 [3 – 6]	6 [4 – 8]	7 [5 – 10]	4 [2 – 6]	7 [5 – 12]	10 [8 – 10]
0–3 in men, 0–2 in women	255 (18)	40 (31)	75 (13)	17 (7)	119 (34)	3 (21)	1 (1)
4–5 in men, 3–5 in women	446 (31)	49 (37)	191 (32)	77 (33)	117 (33)	3 (21)	9 (8)
≥ 6	734 (51)	42 (32)	329 (55)	138 (60)	116 (33)	8 (57)	101 (91)
Under-reporting alcohol use
Categorical (high-risk use)a
No	1,093 (76)	86 (66)	440 (74)	182 (78)	264 (75)	11 (79)	110 (99)
Yes	342 (24)	45 (34)	155 (26)	50 (22)	88 (25)	3 (21)	1 (1)
Continuousb, median [IQR]	−0.4 [−1.1 – 0.4]	−0.1 [−0.8 – 0.5]	−0.4 [−1.0 – 0.4]	−0.5 [−1.3 – 0.1]	−0.1 [−0.7 – 0.6]	−0.5 [−2.3 – 0.2]	−1.8 [−2.3 – −1.1]
HIV viral load
Viral suppression	1,322 (92)	124 (95)	564 (95)	225 (97)	319 (91)	12 (86)	78 (70)
Viral non-suppression	113 (8)	7 (5)	31 (5)	7 (3)	33 (9)	2 (14)	33 (30)
VACS Index 1.0,cmedian [IQR]	12 [6 – 24]	6 [0 – 16]	10 [0 – 18]	-	22 [12 – 34]	-	17 [12 – 34]

^aparticipants classified as under-reporting their alcohol use if PEth ≥200ng/mL and self-reported AUDIT-C scores <6

^bmeasured as PEth z-score minus AUDIT-C z-score

^ccalculated excluding Hepatitis C

Under-reporting of High-risk Alcohol Use.

In our pooled sample of PWH with unhealthy alcohol use, 24% of the participants were classified as under-reporting high-risk alcohol use at the first study visit (Table 1). In the studies in Uganda (i.e., DIPT, Extend, ADEPT-T), under-reporting of high-risk alcohol use ranged from 22% to 34%. In the single study conducted in the United States (i.e., MWCCS), 25% of PWH under-reported their high-risk alcohol use. In the Russia-based studies (i.e., St Peter, Russia ARCH I), under-reporting of high-risk alcohol use ranged from 1% to 21%. Among PWH with unhealthy alcohol use, those who under-reported high-risk alcohol use were more likely to be older and less likely to report tobacco or other substance use (Table 2).

Table 2.

Participant characteristics by under-reporting of high-risk alcohol use at first included study visit (n_ind=1,435)

	No under-reporting (nind=1093)	Under-reporting (nind= 342)	p-value
Age, median [IQR]	41 [35 – 51]	45 [38 – 53]	<0.01
Gender, n(%)			0.23
Women	370 (74.3)	128 (25.7)
Men	723 (77.2)	214 (22.8)
BMI, median [IQR]	22.7 [19.9 – 26.2]	22.0 [19.6 – 25.8]	0.17
Any other substance use (baseline), n(%)			<0.01
No	824 (73.6)	295 (26.4)
Yes	269 (85.1)	47 (14.9)
Any current tobacco use (baseline), n(%)			<0.01
No	665 (73.5)	240 (26.5)
Yes	428 (80.8)	102 (19.3)
INSTI-based ART regimen, n(%)			0.06
No	541 (78.4)	149 (21.6)
Yes	552 (74.1)	193 (25.9)
Study setting (country), n(%)			0.55
Low- or medium-resource (Uganda/Russia)	829 (76.6)	254 (23.5)
High-resource (United States)	264 (75.0)	88 (25.0)

Under-reporting of Alcohol Use and HIV Viral Non-suppression.

Under-reporting of alcohol use and HIV viral non-suppression were positively associated as expected, although the effects were not statistically significantly (Table 3). We did not find any subgroup differences by age, gender, or study country. Sensitivity analyses stratifying adjusted models by AUDIT-C recall periods and study design yielded results comparable to our main findings.

Table 3.

Associations between under-reporting of alcohol use and HIV viral non-suppression among PWH with unhealthy alcohol use (n_ind= 1,435, n_obs= 2,595)

	Unadjusted		Adjusted
	Odds Ratio (95% CI)	p-value	Odds Ratio (95% CI)	p-value
Under-reported high-risk alcohol use (categorical variable)	1.16 (0.67, 1.98)	0.60	1.30 (0.74, 2.27)	0.36
Under-reported alcohol use (continuous z-score)	1.07 (0.87, 1.32)	0.50	1.15 (0.93, 1.42)	0.19

Abbreviations: CI= confidence interval

^*Adjusted for gender, age, BMI, any other substance use (baseline), INSTI-based ART regimen, and study setting

Under-reporting of Alcohol Use and All-Cause Mortality Risk Score.

Under-reporting of alcohol use and an all-cause mortality risk score were positively associated (Table 4). The adjusted model with the categorical alcohol under-reporting variable indicated a positive association (adjusted β= 1.82, 95% CI= −0.18 to 3.82, p-value= 0.07). The adjusted model with the exploratory continuous under-reporting variable found a significant positive association with the all-cause mortality risk score (adjusted β= 0.99, 95% CI= 0.28 to 1.70, p<0.01).. We did not find any subgroup differenes by age, gender, or study country. Sensitivity analyses stratifying adjusted models with the categorical alcohol under-reporting variable by AUDIT-C recall periods, study design, and HIV viral non-suppression yielded results comparable to our main findings.

Table 4.

Associations between under-reporting of alcohol use and an all-cause mortality risk score (VACS Index 1.0) among PWH with unhealthy alcohol use (n_ind= 1,080, n_obs=1,080)

	Unadjusted β (95% CI)	p-value	Adjusted β (95% CI)	p-value
Under-reported high-risk alcohol use (categorical variable)	2.39 (0.35, 4.42)	0.02	1.82 (−0.18, 3.82)	0.07
Under-reported alcohol use (continuous z-score)	1.34 (0.62, 2.06)	<0.01	0.99 (0.28, 1.70)	<0.01

Abbreviations: CI= confidence interval, β= beta coefficient

^*Adjusted for gender, BMI, current tobacco use (baseline), any other substance use (baseline), INSTI-based ART regimen, and study setting

Discussion

Our study pooled data from various settings globally among PWH on ART with PEth-confirmed unhealthy alcohol use (PEth ≥ 50 ng/mL) to assess the relationship between under-reporting of high-risk alcohol use and HIV viral non-suppression and an all-cause mortality risk score. We found that PEth levels were high in our sample, with median PEth of 246 ng/mL and a 24% prevalence of under-reporting of high-risk alcohol use (ranging from 1%−34% within studies). This suggests that high-risk alcohol use is common but under-reported by a substantial proportion of participants in our research. In clinical care, this can have serious implications including under recognition of disease etiology and hindering of treatment for alcohol use disorders.

Causes of under-reporting of alcohol use may include stigma or shame about high-risk alcohol use, which paradoxically may occur whenhealth care providers counsel PWH to reduce or abstain from alcohol consumption. Perceived and enacted stigma have been shown to impact patients’ candidness in discussing unhealthy alcohol use in clinical settings and can delay appropriate care for alcohol use and affect medical care as well⁵⁴. Currently recommended steps to reduce perceived stigma include using standardized screening tools and normalizing screening as a standard part of health care. Patients can be made aware that all patients are asked about alcohol use, and measures for informing them can include posting signs noting that all patients are asked about alcohol use. In addition, self-administered screening questionnaires (e.g., a questionnaire delivered through patient portals) can be helpful in reducing the stigma of alcohol use reporting.⁵⁵ However, self-administered screening faces hurdles in populations with low levels of literacy to complete forms and low-resource settings such as Uganda often lack access to digital forms. Therefore, we recommend more research to understand how best to improve screening for unhealthy alcohol use for PWH across settings, including ways to improve self-report and deliver brief alcohol counselling. These may include using objective measures like alcohol biomarkers. Although alcohol biomarkers are not typically used in primary medical care in part due to concerns about cost and patient acceptability, emerging data suggest that PEth testing in medical care may be feasible and acceptable when delivered in a respectful and relevant manner.^56–58 Further, alcohol biomarkers may improve alcohol interventions by reinforcing progress towards drinking goals.⁵⁸ Therefore, research is also needed to develop low-cost alcohol biomarkers and protocols for incorporating them in care. Promising candidates that deserve more study include point-of-care urine ethyl glucuronide (EtG) tests,⁵⁹ combinations of standard laboratory tests derived from machine-learning algorithms,⁶⁰ or emerging rapid methods for measuring PEth.^61,62

We hypothesized that under-reporting of alcohol use might have a clinically significant impact above the risks of unhealthy alcohol use itself and we therefore limited the sample to those with biomarker confirmed unhealthy drinking (i.e., PEth ≥50 ng/mL). We observed positive associations between under-reporting of alcohol use and HIV viral non-suppression (as expected) but our findings were not statistically significant, possibly due to the high prevalence of HIV viral suppression in our pooled sample. We did find a significant positive association between a continuous score indicating level of under-reporting of alcohol use and an all-cause mortality risk score. This latter finding is consistent with results from a study among veterans in the United States which found that PWH who under-reported their alcohol use had a high 5-year mortality rate that approximated the mortality rate of persons who self-reported the highest level of alcohol use.²⁸ While ample research has been done on the harms of alcohol use itself, especially to PWH, to our knowledge these are the only two studies examining the association of poor health outcomes with under-reporting of alcohol use. These findings together suggest the need for further research to better understand reasons for and ways to prevent under-reporting of alcohol consumption. Although the PWH in our sample were engaged in care, our data suggests that they may have experienced more adverse outcomes compared to those who do not under-report their alcohol use.

Our study has several strengths, including addressing the sizeable gap in knowledge regarding the association between under-reporting of high-risk alcohol use and certain health outcomes. Our use of pooled data across various settings globally allowed for greater heterogeneity in the characteristics of participants in our study sample, the ability to conduct several subgroup analyses, and increased statistical power to examine our research questions. Additionally, PEth levels can vary between individuals (even among those with the same alcohol consumption) given certain biological factors (e.g., hemoglobin, BMI, liver fibrosis) associated with PEth sensitivity. This work was strengthened by restricting our study sample to individuals engaged in unhealthy alcohol use (PEth ≥50 ng/mL) to limit the sample to those with confirmed unhealthy alcohol use; PEth has shown high sensitivity and specificity (>88% and >90%, respectively) at this threshhold.⁶³

Our study had certain limitations. Our study sample was limited to PWH receiving HIV care for at least 6 months and, as such, our findings may not be generalizable to PWH not in care; associations between under-reporting of alcohol use and HIV-related outcomes may be even stronger among those not engaged in care. In addition, all study participants were informed that we would test their blood for an alcohol biomarker that measures past month consumption, which can improve alcohol reporting.⁶⁴ Thus, our estimate of under-reporting of high-risk alcohol use may be lower than in settings without biomarker testing, such as clinical care. On the other hand, persons on ART are likely to receive advice within clinical care to reduce alcohol use, which itself may increase under-report if participants are embarrassed that they had not reduced their alcohol consumption. We used a categorical variable to measure under-reporting of alcohol use which may have oversimplified our data and led to measurement bias. However, we also explored under-reporting of alcohol use as a continuous variable derived from standardized z-scores of the biomarker-measured and self-reported alcohol use data, and we found stronger associations with the mortality risk score using this continuous variable. Also, despite pooling data from six studies, there was a low prevalence of HIV viral non-suppression and hence we were likely underpowered to detect significant differences in HIV viral non-suppression in our analyses. Further, differences by study setting such as differences in AUDIT-C questions, standard drink size, lower limits of detection for viral load, and other characteristics (e.g., latent TB infection in the ADEPT-T and DIPT studies) that tended to vary by country may have affected our results. However, exploratory stratification by country showed reduced statistical significance but did not suggest that country differences largely affected our findings (data not shown).

In summary, our findings in a sample of PWH who were in HIV care and had unhealthy drinking suggests that under-reporting of high-risk alcohol use is common and that under-reporting of alcohol use may mask other health issues. Methods to better identify PWH who under-report unhealthy alcohol use are needed, as this subgroup of PWH may notably benefit from increased engagement with harm reduction strategies and other health interventions.