Systematic Review: Effect of Alcohol Intake on Adherence to Outpatient Medication Regimens for Chronic Diseases

Abstract

Objective:

Nonadherence to medications can lead to adverse health outcomes. Alcohol consumption has been shown to be associated with nonadherence to antiretroviral medications, but this relationship has not been examined at different drinking levels or with other chronic disease medications. We conducted a narrative synthesis of the association of alcohol consumption with nonadherence to medications for four chronic diseases.

Method:

We searched MEDLINE, PsycINFO, Cochrane Library, and Web of Science for relevant studies published through 2009. To be included in this analysis, studies had to be quantitative; have a sample size of 50 or greater; and examine the effect of alcohol consumption on medication adherence for diabetes, hypertension, depression, or HIV/AIDS. Study characteristics and results were abstracted according to pre-specified criteria, and study quality was assessed. Study heterogeneity prevented a systematic synthesis.

Results:

Sixty eligible studies addressed medication adherence for HIV in 47 (78%), diabetes in 6 (10%), hypertension in 2 (3%), both diabetes and hypertension in 1 (2%), depression in 2 (3%), and all medications in 2 (3%). Mean number of subjects was 245 (range: 57–61,511). Effect sizes for the association of alcohol use with nonadherence varied (0.76–4.76). Six of the seven highest quality studies reported significant effect sizes (p < .05), ranging from 1.43 to 3.6. Most (67%) studies reporting multivariate analyses, but only half of non-HIV medicine studies, reported significant associations.

Conclusions:

Most studies reported negative effects of alcohol consumption on adherence, but evidence among non-HIV studies was less consistent. These data suggest the relevance of addressing alcohol use in improving antiretroviral adherence and a need for further rigorous study in non-HIV chronic diseases.

Patients with chronic diseases commonly miss doses of their medications (Cantrell et al., 2011; DiMatteo, 2004). Poor medication adherence often leads to substantial unfavorable health outcomes (Kripalani et al., 2007; Osterberg and Blaschke, 2005). Recognizing common mutable factors that influence medication adherence, such as patient behaviors, has important clinical implications. Nationally, alcohol consumption is highly prevalent; in 2009, 52% of adults reported regular use (Centers for Disease Control and Prevention, 2009) and 15% reported heavy episodic drinking (Centers for Disease Control and Prevention, 2011). The exact mechanisms by which alcohol consumption could affect adherence are not fully known but may include forgetting to take doses because of intoxication and purposely skipping doses to avoid interactions of medications with alcohol (Cook et al., 2001; Kalichman et al., 2009). However, clinicians often do not detect, treat (McGlynn et al., 2003; Turner, 2009), or even consider problem drinking when helping their patients adhere to medication regimens (Hendershot et al., 2009).

In a meta-analysis of the association of alcohol consumption with adherence to antiretroviral therapy, alcohol users were 55% less likely to adhere to HIV medication regimens than abstainers or persons who drank relatively less (Hendershot et al., 2009). Another review found evidence to support the negative impact of alcohol use disorders on adherence to HIV therapy and HIV treatment outcomes (Azar et al., 2010). However, neither review examined the effects of different levels of alcohol consumption on HIV medication adherence, nor did they examine this relationship among populations taking medications for other chronic diseases. The near-perfect adherence required by HIV medication regimens as well as the higher prevalence of problem drinking among HIV-infected patients make it difficult to generalize from these findings to adherence to medications for other diseases, where medications are more forgiving and hence requirements for adherence are less strict.

In this systematic narrative review, we extend the evaluation of the effect of alcohol consumption on medication adherence beyond a focus on HIV to consider whether similar effects appear among persons taking medications for diabetes, hypertension, and depression. We focused on these four diseases because they represent common chronic conditions with the largest body of literature examining the relationship between alcohol use and medication adherence. We also examined studies in the HIV literature to determine whether different levels of alcohol consumption may demonstrate a dose-response association with medication adherence. Finally, we sought to determine whether higher quality studies conducted across diseases supported an association of alcohol consumption with medication adherence.

Method

Search strategy

To identify eligible peer-reviewed publications, we searched four databases (Medline, PsycINFO, the Cochrane Library, and Web of Science) for articles published through 2009 containing specific subject headings and text words (shown in Table 1 for Medline) with the assistance of a medical librarian. Publications were restricted to journal articles, clinical trials, meta-analyses, multicenter studies, and reviews. We reviewed bibliographies of each study included for full text review to identify additional articles.

Table 1.

Specific concepts, medical subject heading (MeSH) terms, and text words in Medline, used to identify articles on alcohol and adherence

Concept	MeSH terms	Text words
Alcohol consumption	Ethanol, alcohol drinking, alcoholism	Alcohol, alcohol drinking, alcoholism
Medication adherence	Patient compliance	Patient compliance, adherence, nonadherence
Diabetes	Diabetes mellitus	Diabetes
Depression	Depressive disorder, depression	Antidepressant
Hypertension	Hypertension	Hypertension
HIV/AIDS	AIDS	HIV

Inclusion criteria

Three reviewers developed inclusion criteria and guidelines for data coding and abstraction. Eligible studies met four criteria: the reports (a) were quantitative research studies; (b) had sample sizes of 50 or more adults; (c) addressed adherence to medications for diabetes, hypertension, depression, and/or HIV/AIDS; and (d) examined the association of alcohol consumption with medication adherence in either primary or secondary analyses. We excluded studies in acute care settings to focus on long-term medication adherence.

One reviewer examined all abstracts identified by the initial database search to evaluate eligibility. When inclusion criteria could not be assessed from the abstract, the full article was reviewed by one of four reviewers. Disagreements were resolved by group discussion. Overall, we identified 60 eligible articles for the review (Figure 1).

Figure 1.

Identified studies of alcohol consumption and adherence to medication for four chronic diseases. *A priori inclusion criteria included (a) quantitative research study; (b) sample size of 50 or more adults; (c) addressed adherence to medications for diabetes, hypertension, depression, or HIV/AIDS; and (d) examined the association of alcohol consumption with medication adherence in either primary or secondary analyses. We excluded studies in acute care settings including emergency room or inpatient hospital.

Data extraction

Eligible studies were independently coded for methods of measuring alcohol use and medication adherence and descriptive characteristics of the analytic sample, including sample size, chronic disease, proportion of women, study setting, and other unique characteristics. Data on the prevalence of alcohol consumption and poor adherence were extracted as well as findings regarding the association of alcohol consumption with adherence. Coding consistency was 95% based on three reviewers’ assessments of the same 10% random sample of identified articles.

Classification of alcohol and adherence measures

Reviewers categorized the definition of alcohol consumption within each article, influenced by guidelines set forth by the National Institute on Alcohol Abuse and Alcoholism (NIAAA), as heavy episodic drinking (defined as 5 or more drinks for men and 4 or more drinks for women within 2 hours), problem drinking (15 or more drinks per week or 5 or more drinks per day for men and 8 or more drinks per week or 4 or more drinks per day for women), or moderate drinking (up to 2 drinks per day for men and up to 1 drink per day for women) (NIAAA, 2005). When the amount or pattern of alcohol consumed was not specified, it was considered any use. Some studies reported specific quantities of alcohol consumption using either continuous or ordinal measures, and others reported scale scores; these we classified as continuous or ordinal as appropriate based on the definitions in the study. Measures were also classified according to the recall period used for alcohol use measurement.

The method that was used to measure medication adherence was considered objective if medication event monitoring system (MEMS) caps, pill counts, or pharmacy records were used or subjective if only self-report or provider report was used. Measures were also classified based on the period for which adherence was assessed.

Quality of evidence

Most evidence came from nonexperimental studies; therefore, we assessed methodological quality using appraisal criteria adapted from the Newcastle–Ottawa Scale as recommended by the Cochrane Non-Randomized Studies Methods Working Group (Reeves et al., 2008; Wells et al., 1999). We included six quality indices and assigned 1 or 2 points (maximum of 10 points possible) for each of the following: (a) representativeness of study sample (1 point), (b) alcohol consumption measured by a validated instrument (2 points), (c) medication adherence measured by a validated self-report instrument (1 point) or objective measures (2 points), (d) longitudinal study design and follow-up rate (2 points), (e) use of imputation or dummy variables to address missing data (1 point), and (f) adjustment for potential confounders (2 points). To develop consistency in rating quality, we compared our ratings for 15 (25%) studies and resolved differences. For reporting purposes, we defined the highest quality studies as those in the top 40th percentile (scoring at least 6 of 10 points).

Analysis

All unadjusted associations and, when available, effect sizes with confidence limits were examined among all studies and among those studies assigned the highest quality rating, conducting a dose-response analysis, and examining each chronic disease type. When necessary, we calculated the reciprocal of the odds ratios (OR) to reflect the association of alcohol with poorer rather than better adherence. We examined associations separately for studies conducting adjusted analyses. Because of the heterogeneity of study samples, measures, and analyses in this review, these results could not be pooled statistically.

Results

Study designs and quality

Of 60 eligible studies, most (75%) were conducted in the United States; international settings included Switzerland, France, Italy, Hungary, China, Brazil, and Ethiopia. The mean sample size was 245 and ranged from 57 to 61,511. Thirteen (22%) studies had more than 1,000 study subjects. Study settings included specialty (e.g., HIV; n = 29) and primary (n = 5) clinics, community-based practices (n = 14), multiple healthcare settings (n = 6), Veterans Affairs facilities (n = 6), substance use treatment centers (n = 3), a large health plan (n = 2), and a public health referral center (n = 1). The primary research question addressed the association of alcohol consumption with medication adherence in 21 (35%) studies, and 17 (28%) examined the association longitudinally.

Measures of adherence and alcohol consumption

Medication adherence.

Measures of medication adherence (Table 2) varied widely; 10 studies (17%) used at least one objective measure, and 16 (27%) used the self-reported Adherence Questionnaire developed by the AIDS Clinical Trial Group (Chesney et al., 2000). Most studies (n = 37) defined adherence as the proportion of prescribed pills or doses taken over a specified timeframe; more than half (n = 21) compared 100% adherence with other levels. Most studies (n = 38) assessed adherence over a 1-month timeframe, whereas a small subset (n = 11) examined a continuous measure.

Table 2.

Self-report and objective measures of alcohol use and medication adherence

Variable	n (% of reported measures)
Studies reporting alcohol measure	18(100)
AUDIT and AUDIT-C (Saunders et al., 1993)	6(33)
Addiction Severity Index (McLellan et al., 1980)	4(22)
Timeline Followback (Sobell and Sobell, 1992)	3(17)
DrInC (Miller et al., 1995)	1(6)
CAGE (Mayfield et al., 1974)	1(6)
DSM-IV (Millon et al., 1996)	1(6)
Alcohol Use Self-Report (Cox et al., 1996)	1(6)
Lifetime Drinking History (Skinner, 1979)	1(6)
Client Diagnostic Questionnaire (Aidala et al., 2004)	1(6)
Studies reporting self-report or objective adherence measure	36(100)
Objective measures	10(28)
Pharmacy records	4(11)
MEMS	7(19)
Composite of MEMS, pill count, and self-report	1(3)
Self-report measures	26(72)
ACTG (Chesney et al., 2000)	16(44)
Timeline Followback modified for adherence (Braithwaite et al., 2005)	4(11)
Swiss HIV Cohort Study (Conen et al., 2009)	1(3)
Adaptation of HCSUS and RAND measures (Friedman et al., 2009)	1(3)
Antiretroviral General Adherence Scale (McDonnell Holstad et al., 2006)	1(3)
Diabetes Self-Care Questionnaire (Johnson et al., 2000)	1(3)
Hill-Bone Adherence to High Blood Pressure Therapy Medication Subscale (Kim et al., 2003)	1(3)
Self-Care Inventory (Lerman et al., 2004)	1(3)

Notes: AUDIT = Alcohol Use Disorders Identification Test; AUDIT-C = AUDIT–consumption questions; DrInC = drinker inventory of consequences; DSM-IV = Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition; MEMS = medication event monitoring system; ACTG = AIDS Clinical Trials Group; HCSUS = HIV Cost and Services Utilization Study.

Alcohol consumption.

All studies measured alcohol consumption by self-report, and one also used a breath alcohol analyzer to assess truthfulness (Samet et al., 2004). As shown in Table 2, the most commonly used validated measures were the Alcohol Use Disorders Identification Test (AUDIT) and AUDIT–consumption questions (AUDIT-C) (n = 6) (Saunders et al., 1993). Most studies examined a 1-month timeframe to assess alcohol consumption (n = 17), although timeframes ranged from 2 weeks to lifetime use. Categorization of alcohol use patterns ranged from any use to dependence. Ten studies examined a continuous indicator of alcohol consumption.

Association between alcohol consumption and adherence for all studies

A statistically significant negative association between alcohol use and medication adherence was reported by 36 (60%) of all 60 studies and by 33 (67%) of the 49 studies conducting multivariate analyses (Table 3). The reported effect sizes ranged from ORs of 0.76 to 4.76 for all 60 studies, from 1.07 to 4.76 for those 36 studies that found a significant association, and from 0.76 to 3.35 of those that found a null association. A descriptive table listing characteristics of all studies is available from the authors on request.

Table 3.

Overall and multivariate significant findings by disease

	Overall			Multivariate
Variable	Total studies, n	Studies with significant associations, n (%)	Reported effect sizes (OR)	Total studies, n	Studies with significant associations, n (%)	Reported effect sizes (aOR)
All studies	60	36 (60)	0.76–4.76	49	33 (67)	0.76–4.76
HIV	47	31 (62)	0.76–4.76	38	27 (71)	0.76–4.76
Diabetes	7	3 (43)	2.50	4	2 (50)	2.50
Hypertension	3	1 (33)	N-R	3	1 (33)	N-R
Depression	2	1 (50)	2.04	1	1 (100)	2.04
All medications	2	2 (100)	1.5, 4.1	2	2 (100)	1.5, 4.1

Notes: OR = odds ratio; aOR = adjusted odds ratio; N-R = not reported.

High-quality studies

Of the seven studies assigned the highest quality ratings (Applebaum et al., 2009; Bryson et al., 2008; Golin et al., 2002; Howard et al., 2002; Kilbourne et al., 2005; Peretti-Watel et al., 2006; Samet et al., 2004), six (86%) reported a statistically significant negative association between alcohol consumption and adherence. Five of the seven high-quality studies assessed adherence only to HIV medications, all of which reported significant effects. Peretti-Watel et al. (2006) examined heavy episodic drinking (more than five drinks in a row at least twice a month during the past year) and found a significant relationship between poor adherence and heavy episodic drinking (“binges”) that occurred two or more times per month versus less often (OR = 1.43, 95% CI [1.02, 2.01] (Peretti-Watel et al., 2006). Samet et al. (2004) similarly found that those engaged in “at-risk” drinking (>14 drinks per week or >4 per day for men; >7 drinks per week or >3 per day for women) were less likely to report 100% adherence than those who abstained from alcohol in the past 30 days (OR = 3.6, CI [2.1, 6.2]) (Samet et al., 2004). Three HIV studies did not report effect sizes but found significant associations for alcohol dependence versus no dependence (definition and timeframe not reported) (Applebaum et al., 2009); any versus no alcohol use in the past 30 days (Golin et al., 2002); and use more than 1 day per week in the past 6 months versus less (Howard et al., 2002).

In one study of 30-day adherence to diabetes medication that conducted adjusted analyses, the odds of nonadherence were higher for heavy episodic drinkers (having six or more drinks on one occasion) versus those not engaging in this type of drinking behavior (OR 2.5, CI [0.83, 5.00]) but did not achieve statistical significance (Kilbourne et al., 2005). Another study of adherence to both antihypertensive and diabetes medications in the past 90 days and 1 year reported significant negative associations of mild, moderate, and severe alcohol misuse (defined by AUDIT-C scores of 4–5, 6–7, and 8–12, respectively) versus abstainers (AUDIT-C score of 0) only for antihypertensive adherence but not for diabetes medication adherence (no effect sizes reported) (Bryson et al., 2008).

Relationship between alcohol intake and poor adherence by level of alcohol consumption

Multivariate effect sizes by level of alcohol consumption are shown in Figures 2a–2e for those articles that reported effect sizes. The relationship between heavy episodic drinking and medication adherence was examined in five studies using multivariate analyses (Figure 2a), four of which found a significant relationship with ORs ranging from 1.43 to 4.1. Problem alcohol use was examined in nine multivariate analyses (Figure 2b), seven of which had significant findings; ORs ranged from 1.52 (CI [1.11, 2.07]) to 3.7 (CI [2.0, 7.1]). All seven multivariate studies of moderate drinking found a significant relationship with medication adherence, with ORs ranging from 1.29 (CI [1.06, 1.57]) to 3.0 (CI [2.0, 4.5]) (Figure 2c). Studies using measures of any alcohol use had less consistent findings: six of eight studies of any alcohol use (OR = 1.3, CI [1.1, 1.5], to OR = 4.76, CI [1.62, 14.08]) (Figure 2d) had significant findings. Even less consistent were the nine studies measuring alcohol consumption as a continuous or count variable, of which five (OR = 1.07, CI [1.03, 1.11], to OR = 2.33, CI [1.35, 3.85]) (Figure 2e) had significant findings.

Figure 2A.

Association of heavy episodic drinking with medication nonadherence. ^aData only available by subgroup and not for entire study sample. CI = confidence interval. *p < .05.

Figure 2E.

Alcohol use measured as continuous, count, or ordinal variable and association with medication nonadherence. ^aData only available by subgroup and not for entire study sample; ^bfrequency of drinking scale scores: 1 = 0 times in past 3 months, 2 = ≤1/month, 3 = ≥1/month but <1/week, 4 = ≥1/week but <1/day, 5 = every day. CI = confidence interval; ex-IDUs = patients with a history of injection drug use but no use during the study period. *p < .05.

Figure 2B.

Association of problem alcohol use with medication nonadherence. ^aData only available by subgroup and not for entire study sample. CI = confidence interval. *p < .05.

Figure 2C.

Association of moderate alcohol use with medication nonadherence. Unless otherwise specified, moderate drinking is defined as drinking at levels less than hazardous drinking defined by the National Institute on Alcohol Abuse and Alcoholism. ^aData only available by subgroup and not for entire study sample. CI = confidence interval. *p < .05.

Figure 2D.

Association of any alcohol use with medication nonadherence. All analyses compared any alcohol consumption with no alcohol consumption. CI = confidence interval. *p < .05.

Relationship between alcohol intake and poor adherence by disease type

Adherence to HIV medications.

Of 47 studies addressing adherence to HIV medications (Applebaum et al., 2009; Arnsten et al., 2002; Berg et al., 2004; Beyene et al., 2009; Bonolo et al., 2005; Bouhnik et al., 2002; Campos et al., 2010; Chander et al., 2006; Chesney et al., 2000; Cohn et al., 2008; Cook et al., 2001; de Jong et al., 2005; Duran et al., 2001a, 2001b; Friedman et al., 2009; Golin et al., 2002; Halkitis et al., 2003, 2005; Heckman et al., 2004; Holmes et al., 2007; Howard et al., 2002; Ingersoll, 2004; Johnson et al., 2003; Lazo et al., 2007; Leserman et al., 2008; Liu et al., 2006; Martini et al., 2004; McDonnell Holstad et al., 2006; Mellins et al., 2009; Moatti et al., 2000; Mugavero et al., 2006; Murphy et al., 2004; Parsons et al., 2007; Peretti-Watel et al., 2006; Protopopescu et al., 2009; Rothlind et al., 2005; Roux et al., 2009; Safren et al., 2001; Samet et al., 2004; Silva et al., 2009; Spire et al., 2002; Sullivan et al., 2007; Theall et al., 2007; Waldrop-Valverde and Valverde, 2005; Waldrop-Valverde et al., 2006; Wang and Wu, 2007), 38 conducted multivariate analyses and, of these, 27 (71%) reported statistically significant associations of poor adherence with alcohol consumption (range in reported effect sizes: 0.76–4.76). Among studies of specific patterns of alcohol use, significantly poorer adherence (ORs) was reported by 9 of 12 studies of problem drinking (1.22–3.70); 2 of 3 studies of heavy episodic drinking (1.43–1.8), of which one found an effect only in women; 8 of 11 studies of moderate alcohol consumption (1.28–3.0); 8 of 14 studies of any alcohol consumption (1.3–4.76); and 6 of 11 studies using continuous or ordinal measures of alcohol consumption (1.03–2.33).

Adherence to diabetes medications.

Of seven studies of adherence to diabetes medications, just one reported significant findings in multivariate analysis, with significantly poorer adherence among those consuming 2–2.9 drinks per day versus abstaining (Ahmed et al., 2006). In two unadjusted studies, heavy drinkers were significantly less adherent than light drinkers and abstainers (Cox et al., 1996), whereas persons who drink any alcohol were less adherent to insulin than abstainers, but no effect was observed for oral agents (Johnson et al., 2000). In three studies performing adjusted analyses and one unadjusted study, none of the measures of alcohol use were associated with diabetes medication adherence (Bryson et al., 2008; Hankó et al., 2007; Kilbourne et al., 2005; Lerman et al., 2004).

Adherence to antihypertensive medications.

All three studies of adherence to antihypertensive medication performed multivariate analyses but did not report effect sizes. (Bryson et al., 2008; Kim et al., 2003; Wang et al., 2002). One reported a significant association between poor adherence and alcohol misuse (categorized as mild, moderate, or severe misuse) versus no use (Bryson et al., 2008).

Adherence to antidepressant medications.

Two studies examined adherence to antidepressant medications, and one reported a significant adjusted association with alcohol use disorder (OR = 2.04, CI [1.42, 2.77]) (Akincigil et al., 2007). The second study reported no association but conducted only an unadjusted analysis of a lifetime history of an alcohol disorder versus all others (Baca-Garcia et al., 2009).

Adherence to all medications.

Two studies conducted by the same research group reported statistically significant adjusted associations of alcohol consumption with poor adherence to all types of medications taken by more than 2,700 (HIV-negative and matched HIV-positive) subjects in the multisite Veterans Aging Cohort Study (Braithwaite et al., 2005, 2008). In one study, heavy episodic drinkers (those who consumed ≥5 drinks within 1 day during the past 30 days) were over four times more likely to be poorly adherent (OR = 4.1, CI [2.6, 5.6]) versus abstainers, whereas drinkers not engaging in heavy episodic use were somewhat more likely to be poorly adherent than abstainers (OR = 1.5, CI [1.1, 2.1]) (Braithwaite et al., 2005). Although the association between heavy episodic drinking and adherence was slightly stronger among HIV-infected individuals than HIV-negative people, the effect size was large in both groups (OR = 4.3 vs. 3.3, respectively) (Braithwaite et al., 2005). In the second study, adherence was significantly poorer on days when individuals drank: 6 or more versus 0.5 or fewer drinks, more than his or her average amount, to impairment, or hazardously (effect sizes not reported) (Braithwaite et al., 2008).

Dose-response studies

Three studies examined and demonstrated a temporal dose-response relationship of alcohol consumption with medication adherence (Braithwaite et al., 2005, 2008; Bryson et al., 2008). Braithwaite et al. used Timeline Followback methods to assess whether alcohol consumption on a particular day was associated with poor adherence to prescribed medications on the same day (Braithwaite et al., 2005). Missing doses of any medication on surveyed days showed a significant trend (p < .001) in which abstainers missed 2.4% of doses whereas drinkers not engaging in heavy episodic use missed 3.5% on drinking days, 3.1% on postdrinking days, and 2.1% on nondrinking days. Heavy episodic drinkers (defined previously) missed doses on 11.0% of drinking days, 7.0% of postdrinking days, and 4.1% of nondrinking days (p < .001 for trend). In the same cohort (Braithwaite et al., 2008), consideration of an individual’s threshold for cognitive impairment from alcohol consumption improved prediction of poor adherence.

Bryson et al. (2008) investigated a dose-response relationship of alcohol intake and poor adherence using pharmacy refill data for more than 13,000 patients taking antihypertensive and/or diabetes medication. The AUDIT-C score was linearly associated with poorer adherence to antihypertensive drugs but not with adherence to diabetes medications.

Discussion

Optimizing patients’ adherence to medications has great public health importance. Previous studies have reported that patients with chronic diseases often do not adhere to prescribed medications, leading to poor health outcomes (Kripalani et al., 2007; Osterberg and Blaschke, 2005). In one meta-analysis, only 60% of patients with chronic diseases took all their prescribed doses (Cantrell et al., 2011; DiMatteo, 2004). Because alcohol consumption is quite common, its potential negative effects on medication adherence may significantly compromise chronic disease management and outcomes. Most of the research on this topic has been conducted in HIV-infected patients, but it is quite plausible that patterns observed for HIV medication adherence are similar to those for adherence to medications used for much more common diseases.

Our synthesis of HIV medication adherence research shows a relatively consistent negative association of alcohol consumption with medication adherence as reported in other systematic reviews of this topic (Azar et al., 2010; Hendershot et al., 2009). However, our study is the first to conduct an assessment of study quality, which finds that the most rigorously conducted studies show a significant inverse relationship of greater alcohol use with HIV medication adherence. This effect was stronger for greater levels of alcohol consumption but also appeared to exist for any use. Only two studies conducted by the same group evaluated the alcohol-adherence relationship in both HIV-infected and non-HIV-infected veterans (Braithwaite et al., 2005, 2008). This research showed a stronger dose-response relationship for HIV-infected patients, but heavy episodic drinking showed a strong negative association in both HIV-infected and HIV-negative subjects.

The studies of the association of alcohol consumption with adherence to medications for non-HIV-related chronic conditions generally suffered from poorer quality. Of the eight studies examining exclusively non-HIV medications that conducted adjusted analyses, only four reported a significant association. The largest number of non-HIV studies examined adherence to medications for diabetes. Of seven studies identified as being high quality in our review, six found a significant association between alcohol and adherence but only two examined adherence to non-HIV medications. One high-quality study of adherence to diabetes medications reported a significant association with heavy episodic drinking, but neither study examining antihypertensives showed significant findings.

The stronger association of alcohol use with medication adherence among HIV-positive patients could be attributed to the relative complexity of HIV medication regimens. Increased regimen complexity has been demonstrated to negatively affect adherence to medications for chronic diseases, particularly HIV/AIDS, diabetes, and hypertension (with insufficient literature conducted on this relationship in the area of hypertension) (Ingersoll and Cohen, 2008). Although currently many HIV regimens consist of only one daily dose, they have decreased in complexity relatively recently; in 2007, participants in HIV medication trials took more medications on average than those in other chronic disease medication trials (Gossec et al., 2007). It is possible that alcohol use may have a more detrimental effect on adherence to more complex regimens such as those for HIV.

Additionally, the stronger association between alcohol use and HIV medication adherence could be attributed to characteristics of the different patient populations. Heavy alcohol consumption is more prevalent among HIV-infected patients than the general population (Galvan et al., 2002); further, HIV-infected patients are generally younger than most patients with diabetes and hypertension, and the prevalence of heavy alcohol consumption decreases with older age (Eigenbrodt et al., 2001). It is also possible that the same level of alcohol use may lead to greater cognitive impairment among HIV-infected persons (Rothlind et al., 2005). Furthermore, not all studies adjust for mood disorders such as depression, another possible mechanism by which alcohol consumption could negatively affect adherence, especially in persons with HIV infection (Grenard et al., 2011).

Our findings also suggest that lower levels of alcohol use may influence adherence by a different mechanism than that of heavier use, such as heavy episodic or problem drinking. For example, in the national Nurses’ Health Study, consumption of one alcoholic beverage daily did not impair cognition and may have even improved it (Stampfer et al., 2005). People who drink may be more likely to simply forget to take medications or may worry about drug interactions. Alternatively, measurement bias could result in an overestimate of the association of lower levels of alcohol use as a result of the well-known tendency of individuals to underestimate their alcohol use (Klatsky el al., 2006). Studies examining “any drinking” versus all others have distorted results by including heavy drinkers with persons who drink very little. Further, the reference group may include former heavy drinkers with true abstainers.

This study reinforces the need for more rigorous research on the association of alcohol use with medication adherence among people who are not infected with HIV. Particularly, the near-perfect level of adherence required for HIV medications may mean that the negative effects of alcohol use are more evident for that condition than for others. Further research of adherence to medications for diabetes, hypertension, and depression needs to use validated alcohol measures that assess multiple categories of alcohol consumption (e.g., alcohol dependence as well as heavy episodic drinking, hazardous drinking, and any drinking, such as the AUDIT), as well as objectively measured adherence levels. Where possible, studies should control for potential confounders such as mental illness and conduct dose-response analyses.

Despite the limitations of the current literature, this review draws attention to a potentially important gap in the evidence about alcohol on medication adherence. The potential public health implications of poor adherence to chronic disease medications support the need for more rigorous studies to evaluate this association.