Research in the news: Even moderate alcohol intake may harm people with HIV - Risk of mortality and physiologic injury evident with lower alcohol exposure among HIV infected compared with uninfected men - VACS
Download the PDF here
Moderate alcohol consumption is more harmful to people with HIV than uninfected individuals, raising the risk of both mortality and other negative health effects, say Yale researchers. Their study is the first to demonstrate the increased harm among patients who have suppressed HIV with modern antiretroviral treatment (ART).
Research has shown that it takes fewer drinks for a person with HIV to feel the effects. However, most prior studies were done on HIV-positive individuals who had detectable virus. The Yale-led team set out to determine whether the risks associated with alcohol were higher among current patients who are more likely to have the infection under control with ART.
The researchers analyzed data on HIV-positive and uninfected patients from the Veterans Aging Cohort Study (VACS), a large population of individuals receiving care from the Veterans Health Administration, between 2008 and 2012. They examined the association between alcohol consumption and mortality and other signs of physiologic injury.
They found that HIV-positive individuals were more likely to die and experience physiological harm from alcohol consumption than uninfected individuals. Even consumption of one to two drinks per day was associated with increased risk for people with HIV. The finding was particularly notable because it held true for individuals with suppressed virus, said the researchers.
"It demonstrates that even among people on ART with suppressed viral load, who are much less sick in general, there is still an added effect of alcohol among those individuals than people without HIV," said Amy Justice, professor of general medicine and of public health. "It suggests the threshold for safe alcohol consumption is likely different for people with HIV."
"We used data from the Veterans Aging Cohort Study (VACS), a large electronic medical record (EMR) based cohort study of 47,805 HIV + and 99,061 uninfected patients receiving care in the Veterans Health Administration (VHA) from fiscal years 1997 to 2014 (Conigliaro et al., 2004; Justice et al., 2006a, 2006b)......
The Alcohol Use Disorders Identification Test - Consumption (AUDIT-C) is a commonly used tool to screen for alcohol use
Compared with uninfected individuals, mortality and physiologic injury associated with discrete levels of alcohol exposure were higher among HIV+ individuals. ......lower levels of alcohol exposure were associated with more harm among HIV+ individuals.......Among HIV+ individuals, AUDIT-C score ≥4 (hazard ratio [HR] 1.25, 95% CI 1.09-1.44) and ≥30 drinks per month (HR, 1.30, 95% CI 1.14-1.50) were associated with increased risk of mortality. Among uninfected individuals, AUDIT-C score ≥5 (HR, 1.19, 95% CI 1.07-1.32) and ≥70 drinks per month (HR 1.13, 95% CI 1.00-1.28) were associated with increased risk"
This is the most comprehensive study of the question of unhealthy alcohol consumption among HIV+ individuals to date....There are several, likely overlapping, reasons why HIV+ individuals might be more susceptible to physiologic harm from alcohol. First, they may experience higher blood alcohol levels given a unit exposure......even modest alcohol use is associated with poorer adherence to antiretroviral therapy which can in turn increase susceptibility to harm. Collectively, these data suggest that HIV+ individuals are more susceptible to physiologic harm from alcohol. Our results support this hypothesis"
Article in Press
Risk of mortality and physiologic injury evident with lower alcohol exposure among HIV infected compared with uninfected men
Drug and Alcohol Dependence Feb 4 2016
Amy C. Justice1,2# Amy.Justice2@va.gov, Kathleen A. McGinnis3, Janet P. Tate1,2, R. Scott Braithwaite4, Kendall J. Bryant5, Robert L. Cook6, E. Jennifer Edelman1,2, Lynn E. Fiellin2, Matthew S. Freiberg7,8, Adam J. Gordon3,9, Kevin L. Kraemer3,9, Brandon D.L. Marshall10, Emily C. Williams11, David A. Fiellin1,2
1Veterans Aging Cohort Study Coordinating Center, West Haven VA Healthcare System, 950 Campbell Ave, West Haven, CT, 06516, USA
2Department of Internal Medicine, and the Center for Interdisciplinary Research on AIDS (CIRA), Yale School of Medicine, Yale University 367 Cedar Street, New Haven, Connecticut, 06510, USA
3VA Pittsburgh Healthcare System, University Drive C, Pittsburgh, Pennsylvania, 15240, USA
4Department of Population Health New York University School of Medicine, 227 East 30th street, New York, NY 10016, USA
5National Institute on Alcohol Abuse and Alcoholism, 5635 Fishers Lane, MSC 9304, Bethesda, MD 20892-9304, USA
6Department of Epidemiology, University of Florida, PO Box 100231, Gainesville, FL, USA
7Division of Cardiovascular Medicine, Vanderbilt University School of Medicine, Nashville, TN, USA
8Geriatric Research, Education, and Clinical Center, Veterans Affairs Tennessee Valley Healthcare System, 2525 West End Avenue, Nashville, TN, USA
9Division of General Internal Medicine, University of Pittsburgh School of Medicine, Suite 600, 230 McKee Place, Pittsburgh, Pennsylvania, 15213, USA
10Department of Epidemiology, Brown University School of Public Health, 121 South Main Street, Providence, RI, USA, 02912
11University of Washington School of Public Health, 325 Ninth Avenue, Box 359762, Seattle, WA, USA
⋅ Individuals with HIV onantiretroviral treatment (ART) experience mortality at lower levels of alcohol use.
⋅ Individuals with HIV experience physiologic frailty at lower levels of alcohol use.
⋅ Alcohol consumption limits should be lower among HIV+ individuals.
HIV infected (HIV + ) individuals may be more susceptible to alcohol-related harm than uninfected individuals.
We analyzed data on HIV+ and uninfected individuals in the Veterans Aging Cohort Study (VACS) with an Alcohol Use Disorders Identification Test-Consumption AUDIT-C score from 2008 to 2012. We used Cox proportional hazards models to examine the association between alcohol exposure and mortality through July, 2014; and linear regression models to assess the association between alcohol exposure and physiologic injury based on VACS Index Scores. Models were adjusted for age, race/ethnicity, smoking, and Hepatitis C infection.
The sample included 18,145HIV+ and 42,228 uninfected individuals. Among HIV+ individuals, 76% had undetectable HIV-1 RNA (< 500 copies/ml). The threshold for an association of alcohol use with mortality and physiologic injury differed by HIV status. Among HIV+ individuals, AUDIT-C score ≥4 (hazard ratio [HR] 1.25, 95% CI 1.09-1.44) and ≥30 drinks per month (HR, 1.30, 95% CI 1.14-1.50) were associated with increased risk of mortality. Among uninfected individuals, AUDIT-C score ≥5 (HR, 1.19, 95% CI 1.07-1.32) and ≥70 drinks per month (HR 1.13, 95% CI 1.00-1.28) were associated with increased risk. Similarly, AUDIT-C threshold scores of 5-7 were associated with physiologic injury among HIV+ individuals (beta 0.47, 95% CI 0.22, 0.73) and a score of 8 or more was associated with injury in uninfected (beta 0.29, 95% CI 0.16, 0.42) individuals.
Despite antiretroviral therapy, HIV+ individuals experienced increased mortality and physiologic injury at lower levels of alcohol use compared with uninfected individuals. Alcohol consumption limits should be lower among HIV+ individuals.
Compared with uninfected individuals, mortality and physiologic injury associated with discrete levels of alcohol exposure were higher among HIV+ individuals. While we found evidence of a J-shaped association between alcohol use, mortality, and physiologic injury among uninfected subjects, there was little evidence suggesting a protective effect of alcohol at any level of use among HIV+ individuals. Further, lower levels of alcohol exposure were associated with more harm among HIV+ individuals. Thresholds for risk were also lower among HIV+ individuals when physiologic injury was estimated using the VACS Index. Based on our findings, HIV+ individuals consuming more than 30 drinks per month are at increased risk of all-cause mortality and physiologic frailty. This would translate to a recommended drinking limit for HIV+ individuals of no more than 1 drink containing alcohol per day. This is lower than the current limits by the National Institute on Alcohol Abuse and Alcoholism recommended for men, which is no more than 14 drinks per week (equivalent to 2 drinks per day), and similar to the recommendations for women and those over 65 years pf age (NIAAA, 2005). This is the first study to demonstrate an association between low levels of alcohol use and mortality or physiologic injury among HIV+ individuals in the modern antiretroviral treatment era.
This is the most comprehensive study of the question of unhealthy alcohol consumption among HIV+ individuals to date. The VHA benefits from one of the largest and most comprehensive electronic health information systems in the world with some of the longest follow up (Affairs, 2006; Corrigan et al., 2003; McQueen et al., 2004). Our group has developed and validated a national multisite cohort of HIV+ individuals engaged in HIV treatment with demographically matched uninfected comparators. Drawing from the VHA EMR, we have clinical data supporting careful adjustment for important confounders including tobacco exposure and Hepatitis C infection. We were able to use national AUDIT-C data collected in the course of routine clinical care to assess varying levels and patterns of alcohol exposure.
There are several, likely overlapping, reasons why HIV+ individuals might be more susceptible to physiologic harm from alcohol. First, they may experience higher blood alcohol levels given a unit exposure. This has been demonstrated among those with untreated HIV infection (McCance-Katz et al., 2013, 2012). Recent analyses in VACS have demonstrated that HIV+ individuals with unsuppressed HIV-1 RNA report the lowest number of drinks required to feel intoxicated, HIV+ individuals with suppressed HIV-1 RNA report a higher number to experience intoxication and uninfected individuals report an even higher number of drinks to experience intoxication (McGinnis et al., 2015). Further, even modest alcohol use is associated with poorer adherence to antiretroviral therapy (Braithwaite and Bryant, 2010; Braithwaite et al., 2008, 2010) which can in turn increase susceptibility to harm. Collectively, these data suggest that HIV+ individuals are more susceptible to physiologic harm from alcohol. Our results support this hypothesis. The VACS Index, an integrated measure of physiologic injury, has been widely recognized as a means of estimating physiologic injury and risk of frailty related outcomes among HIV+ individuals in care (Akgun et al., 2014; Escota et al., 2014; Womack et al., 2013).
The VACS Index includes many biomarkers known to be directly or indirectly influenced by alcohol exposure. CD4 count and HIV VL are altered by non-adherence to antiretroviral therapy which is strongly associated with alcohol use (Braithwaite and Bryant, 2010; Braithwaite et al., 2008). Hemoglobin, platelets, aspartate transaminase and alanine transaminase are altered by direct toxicity of alcohol and by non-adherence to antiretroviral treatment (Anderson et al., 2014; Conigliaro et al., 2003; Lo et al., 2014; Schmitt et al., 1999; Sullivan et al., 2008). Given this dual effect of alcohol among HIV+ individuals, we might expect the VACS Index to be sensitive to adverse health effects from increasing alcohol exposure in this population and that is what we observed. Importantly, 76% of HIV+ individuals had achieved HIV virus suppression, suggesting that this susceptibility to alcohol does not disappear once suppressed. HIV+ individuals, even those on ART, are more susceptible to physiologic harm from alcohol.
An AUDIT-C score of 4 or more is often considered consistent with unhealthy alcohol use (Gordon et al., 2001) and 24% of HIV+ individuals in our study had scores at or above this threshold. However we found that HED and drinks per month were also independently associated with mortality suggesting that these criteria should be considered individually. When these were considered as separate criteria, 30%, or an additional 6%, of HIV+ individuals in VACS were identified with unhealthy alcohol use.
Accurate measurement of exposure to alcohol is challenging. While self-report has limitations, it remains the standard in clinical settings. We used the AUDIT-C from the VHA EMR. At most sites AUDIT-C is asked face to face by a healthcare provider or health technician who records the patient's response, an approach that may be subject to social desirability bias. There may also be quality issues related to how screening is conducted in routine clinical settings (non-verbatim screening, assumptions/inferences being made about patient responses, and inputting responses not reported; Williams et al., 2015). When compared with responses on a confidential survey, clinical AUDIT-C data tends to under report alcohol exposure equally by HIV status (Bradley et al., 2011). In addition, we did not separate out patients with alcohol use disorder in the current analysis. Because detection of alcohol use disorder in routine clinical settings is poor we were not confident that separating out those assigned a diagnosis of alcohol use disorder based on ICD-9 codes would help (Kim et al., 2012; Quan et al., 2008). While higher AUDIT-C scores have been proposed for this purpose, the AUDIT-C is not intended as a diagnostic tool for alcohol use disorder (Johnson et al., 2013; McGinnis et al., 2013; Rubinsky et al., 2010). These limitations would introduce misclassification which would dampen observed associations suggesting that the differences we were able to observe by HIV status may be under estimated. Finally, these data reflect information available to the clinician in the course of care.
Another important limitation was the need to exclude those who reported no exposure to alcohol in the past 12 months. By excluding non-drinkers, we are not able to determine whether there is a level of alcohol exposure measured by AUDIT-C that is "safe". All our analyses were compared with the lowest score on the AUDIT-C among those reporting current alcohol use (a score of 1). Non-drinkers were excluded because non-drinkers are often people who quit drinking because they experienced adverse outcomes related to alcohol (so called, "sick quitters"; Fillmore et al., 2007; Shaper, 1990; Shaper and Wannamethee, 1998). For instance, in a recent VACS survey, 29% of subjects reported that they stopped drinking because they had problems due to alcohol (data not otherwise shown). Our results may only apply to men and those receiving care in the Veterans Healthcare System.
In conclusion, among HIV+ individuals on ART, lower thresholds of alcohol use are associated with mortality and physiologic injury than among uninfected individuals. In both groups HED and drinks per month are independently associated with mortality and considering each as separate criteria identifies more individuals with unhealthy alcohol use.
Alcohol Mortality & Morbidity in HIV+/-
The health impact of alcohol use is well described (Dawson et al., 2008; Rehm et al., 2003; Saitz, 2005). While the impact of alcohol on cause-specific mortality can vary by condition, most large observational studies demonstrate a dose-dependent association between higher levels of alcohol consumption and all-cause mortality (Di Castelnuovo et al., 2006; Harris et al., 2010; Knott et al., 2015; Plunk et al., 2014; Rogers et al., 2015). However, alcohol use is common among those with HIV infection (HIV+ individuals) and may be uniquely harmful in this population (Braithwaite and Bryant, 2010; Braithwaite et al., 2008, 2005; Cook et al., 2001). Among other effects, alcohol use is associated with poor adherence to life preserving combination antiretroviral treatment (ART; Braithwaite and Bryant, 2010; Braithwaite et al., 2008, 2005; Cook et al., 2001; Samet et al., 2004) and increases risk for liver fibrosis (Lim et al., 2014), a leading cause of death. Further, compared with uninfected individuals, HIV+ individuals report fewer drinks required to experience intoxication (McGinnis et al., 2015). As a result, some have suggested that "safe limits" for alcohol use among those with HIV infection should be lower than among uninfected individuals (Samet et al., 2007, 2004).
Further, health effects of alcohol among HIV+ individuals appear to vary depending on whether viral suppression is attained. Given a unit dose of alcohol, HIV+ individuals experience higher blood alcohol concentrations before, compared to after, initiating ART (McCance-Katz et al., 2012). In addition, having a detectable HIV viral load is associated with a lower number of drinks to feel intoxicated among HIV+ individuals (McGinnis et al., 2015). Since many the prior studies were dominated by HIV+ individuals who had not achieved viral suppression, it is important to determine whether differential harms from alcohol use persist among a more
current sample of HIV+ individuals, most of whom have achieved HIV-1 RNA suppression on ART, compared with uninfected individuals.
Finally, all-cause mortality is an important health outcome, but it is not the only one of interest (Kinder et al., 2009). HIV+ individuals in care are aging (Justice, 2010) and experiencing increasing physiologic injury resulting in cumulative frailty or a greater vulnerability to stressors (Clegg et al., 2013). The Veterans Aging Cohort Study (VACS) Index incorporates weighted values of routinely available clinical biomarkers including age, CD4 count, HIV-1 RNA, hemoglobin, platelets, aspartate transaminase, alanine transaminase, creatinine, and Hepatitis C sero-status (Justice et al., 2013; Tate et al., 2013). The Index provides reliable estimates of five-year mortality in HIV + individuals (Tate et al., 2013) and has been shown to predict hospitalization (Akgun et al., 2013a), physical and cognitive performance (Marquine et al., 2014; Oursler et al., 2013), and fragility fractures (Womack et al., 2013). The Index also predicts hospitalization and mortality among uninfected individuals (Akgun et al., 2013a). As a result, the VACS Index is a clinically applicable measure of overall physiologic injury and may be a meaningful indicator of frailty (Akgun et al., 2013a; Escota et al., 2014; Shaper, 1990; Shaper and Wannamethee, 1998; Womack et al., 2013). The Alcohol Use Disorders Identification Test - Consumption (AUDIT-C) is a commonly used tool to screen for alcohol use in large clinical populations (Bradley et al., 2007). We use mortality, the VACS Index, and self-reported alcohol data, to compare health outcomes associated with patterns of alcohol use among HIV+ and uninfected individuals. Our first aim is to determine whether level
of alcohol use is associated with all-cause mortality and physiologic injury as measured by the VACS Index. Our second aim is to determine whether AUDIT-C thresholds for physiologic injury differ by HIV status.
EMR that providers must complete on their patients on a regular basis (Bradley et al., 2006). For each of the 3 questions in the AUDIT-C, 0-4 points are assigned based on the response and summed for a total score ranging from 0-12. Item 1 is: "How often do you have a drink containing alcohol?" Choices (points) include: never (0), monthly or less (1), 2-4 times a month (2), 2-3 times a week (3); or 4 or more times a week (4). Item 2 is: "How many standard drinks containing alcohol do you have on a typical day?" Choices (points) include: 0 (0), 1 or 2 (0); 3 or 4 (1); 5 or 6 (2); 7 to 9 (3); 10 or more (4). The original AUDIT-C Item 2 had no option for 0 number of drinks; therefore, we combined the responses of 0 and 1 or 2. Item 3 is: "How often do you have six or more drinks on one occasion?" Choices (points) include: never (0); less than monthly (1); monthly (2); weekly (3); daily or almost daily (4).
AUDIT-C has been evaluated using multiple cutoffs including 3+, 4+, and 5+ (Bradley et al., 2007; Bush et al., 1998; Gordon et al., 2001; Gual et al., 2002). Heavy episodic drinking (HED) has been assessed using a cutoff based on any positive response to AUDIT-C item 3 (McGinnis et al., 2013). Further, recommended cutoffs for AUDIT-C vary based on the population (male vs. female, 65 years), the different modes in which the AUDIT-C is asked (face to face vs. confidential survey), and the purpose for using the AUDIT-C (initial screening prior to further assessment vs. the only assessment). Because of the diversity of cutoffs used in prior literature, we initially employed integer categories of AUDIT-C to evaluate multiple candidate cutoff(s) (1, 2, 3, 4, 5-7, and 8+) for predicting mortality and physiologic frailty. Alcohol use is often evaluated using quantity-frequency measures (i.e., total number of drinks over a certain timeframe) and an average of more than 2 drinks per day is considered potentially unhealthy alcohol use for men (Saitz, 2005; NIAAA, 2005). Therefore, we also created a variable for estimated total number of drinks per month using the first two items in the AUDIT-C. Categories included 1-2, 3-7, 8-29, 30-69, and 70+ drinks per month. We also considered HED (Smith et al., 2009) frequency separately.
3.1 Demographics and Characteristics
There were 18,145 HIV+ and 42,228 uninfected individuals with an AUDIT-C score greater than zero. Mean age was 52.5 years for HIV+ and 54.0 years for uninfected
individuals. The distribution of race/ethnicity and AUDIT-C scores was generally similar by HIV status. Hepatitis C was more prevalent among HIV+ (31%) than uninfected individuals (16%, p<.001). Among HIV+ and uninfected individuals, 16% and 13% had a drug dependence-related diagnosis code (p<.001), respectively. Most HIV+ individuals (76%) had achieved HIV-1 RNA suppression demonstrating effective ART. Mean VACS index was 30 for HIV+ and 21
for uninfected individuals (p<.001) (Table 1). Among HIV+ individuals, considering HED and drinks per month as separate criteria results in identifying 30% with unhealthy alcohol use compared to 24% using the overall AUDIT-C threshold of 4 or higher.
Median follow-up was 4.8 years (interquartile range (IQR) 3.6 to 5.3). Mortality per 100 person-years was 2.7 among HIV+ and 1.8 among uninfected individuals (p<.001, Table 1). Mortality rates increased with increasing AUDIT-C scores for both HIV+ and uninfected individuals (Figure 1a). However, the difference in mortality rates between HIV+ and uninfected individuals widened with higher AUDIT-C scores (interaction term from Cox PH model: p=.02). Similar patterns and significant interactions were found using number of drinks per month (Figure 1b) or HED (Figure 1c) as the measure of alcohol exposure.
In multivariable models, based on hazard ratios (HRs), the association between AUDITC scores of one, two, and three with the outcomes were similar, so the categories were combined so that the AUDIT-C referent group contains individuals with an AUDIT-C of 1-3.
Among HIV+ individuals, adjusted HRs for mortality were significantly higher for those with AUDIT-C of 4, 5-7, and 8-12 compared to those with AUDIT-C of 1-3. Among uninfected individuals, HRs were significantly higher for those with AUDIT-C scores of 5-12 (Table 2). In a combined model HIV infection was associated with mortality (HR 1.35: 95% CI 1.26-1.45) and the interaction term for HIV and AUDIT-C was significant (p=.02). Age, smoking, and hepatitis C infection were also significant predictors of mortality in all three models.
Number of drinks per month and HED in the past 12 months were also important predictors of mortality. Among HIV+ individuals, HRs for mortality were significantly higher for those drinking an estimated 30 or more drinks per month compared to those drinking 1-2 drinks per month (Table 3). Among uninfected individuals, HRs were significantly higher for those drinking an estimated 70 or more drinks per month compared to those drinking 1-2 drinks per month. Notably, a "J-shaped" relationship between drinks per month and risk of mortality was only observed for the uninfected, whereby 3-7 drinks per month (compared to 1- 2 drinks per month) was found to be protective (HR 0.86: 95% CI 0.78-0.95). In combined models, HIV infection was associated with higher risk of mortality (HR 1.28: 95% CI 1.18, 1.40). Both drinks per month and HED in the past 12 months were independently associated with mortality and the interaction term for HIV and drinks per month was significant (p<.001). Age, smoking, and Hepatitis C infection were also significant predictors of mortality in all three models.
3.3 Physiologic Injury
Among HIV+ individuals, physiologic injury, as measured by VACS Index, increased with increasing AUDIT-C scores (Figure 2a). Similarly, injury increased with increasing number of drinks per month and increasing frequency of HED (Figures 2b-c). Additionally, the difference in physiologic injury between HIV+ and uninfected individuals was greater as all three measures of alcohol use increased (Figures 2a-c).
In linear regression models adjusting for the same factors as models predicting all-cause mortality, the levels of alcohol use associated with injury were lower among HIV+ than among uninfected individuals (Tables 4 and 5). AUDIT-C scores of 5-7 and 8-12 were significantly associated with physiologic injury among HIV+ individuals compared to those with AUDIT-C 1-3 (beta 0.47: 95% CI 0.22, 0.73; and beta 1.05: 95% CI 0.77, 1.34, respectively). In contrast, only a score of 8-12 was associated with injury among uninfected individuals (beta 0.29: 95% CI 0.16, 0.42) (Table 4). Interestingly, after adjusting for HED (which was not independently associated with VACS Index), drinking between 3 and 29 drinks a month was protective against injury among uninfected individuals (beta -0.23: 95% CI -0.33- (-0.12) and beta -0.16: 95% CI -0.27-(-0.04), respectively) whereas no level of use was protective among HIV+ individuals. Drinking 70 or more drinks per month was associated with injury among HIV+ individuals (beta 0.99: 95% CI 0.68-1.29) and HIV uninfected individuals (beta 0.19: 95% CI 0.04-0.34). The J-shaped associated between alcohol and physiologic injury was only observed for the uninfected group.