Concurrent use of alcohol interactive medications and alcohol in older adults: a systematic review of prevalence and associated adverse outcomes

A comprehensive systematic search was performed using MEDLINE (PubMed), Embase, Scopus and Web of Science. A combination of the following keywords and MeSH terms were used: “ethanol”, “alcohol”, “drug interactions”, “drug alcohol interaction” and “aged”. This search was supplemented by a search in Google Scholar and by hand searching references of retrieved articles. The search was restricted to English language articles and articles published since January 1990 to June 2016.

Studies were included if they met the following eligibility criteria: Observational studies reporting on the concurrent use of alcohol and alcohol interactive (AI) medicines in the same or overlapping recall periods in older adults. Studies also had to report on the quantity or frequency of alcohol consumption. We excluded studies which exclusively sampled patients with specific illnesses, or those seeking treatment for alcohol use disorders (AUD) or illicit drug use.

Title and abstracts of identified studies were reviewed by one reviewer (AH) to determine potential eligibility. Full text articles were then reviewed by two reviewers (AH/GC) for those studies considered eligible from title/abstract, or when it was unclear whether a study met the inclusion criteria. The following data were extracted by two reviewers (AH/GC): year of publication, country, study sample, study design, measurement and definition of alcohol interactive (AI) medications, measurement of alcohol consumption, prevalence of alcohol use and AI medication use and prevalence of concurrent alcohol and AI medication use. Adverse outcomes associated with concurrent use of alcohol and AI medications were also extracted if reported. Any uncertainty in relation to study eligibility and data extraction was resolved through discussion between two reviewers (GC/AH).

The risk of bias was evaluated, by two reviewers (AH/GC), using an adapted form of the Newcastle Ottawa cohort scale (NOS) [17]. This amended NOS scale allowed for the evaluation of cross-sectional studies, focussing on the risk of selection bias and information bias, specifically misclassification bias of both exposures (alcohol and AI medications) and the outcome (concurrent use of alcohol and AI medications) across included studies.

Ten of the included studies were conducted in Europe [14, 19, 21‐23, 25, 27‐29, 36], eight in North America [18, 20, 24, 30‐32, 34, 35] and two in Australia [26, 33]. (Table 1) All studies were cross-sectional [14, 18‐36]. Study settings varied across studies with; community-dwelling [14, 19, 21‐27, 31, 34‐36], both community dwelling or living independently in care facilities [32], general populations [20, 28, 33], hospital setting [29], retirement communities [18] and participants signed up to a pharmaceutical assistance contract for the elderly [30] reported.

Sample sizes varied from 311 to 83,321 participants [18, 30] . Men and women were included in all studies, with the exception of one Australian study which only included men [26]. Nine studies reported on a wide range of prescription and/or over the counter (OTC) medicines with potential to interact with alcohol [18‐21, 24, 27, 29‐31] . Three studies investigated any medication use during the recall period [23, 33, 35], and a further eight focused on psychotropic medications [14, 22, 25, 26, 28, 32, 34, 36]. Of the eight studies focusing on psychotropic medications, five investigated psychotropic medications alone [22, 25, 26, 34, 36] and three studies also included analgesics [14, 28, 32].

Of the nine studies focusing on a wide range of AI medications, all studies classified central nervous system (CNS) agents as AI medications (Table 1). Consistent with those studies investigating psychotropic medications [14, 22, 25, 26, 28, 32, 34, 36], the following drug classes were classified as AI medications, sedatives/hypnotics [18‐21, 24, 27, 30, 31], antidepressants [18‐21, 27, 30, 31], opioids/narcotics [18‐21, 27, 30, 31], anticonvulsants [19‐21, 24, 27, 30, 31] and anti-psychotics [19‐21, 27, 29, 30]. After CNS agents, cardiovascular medicines (CVS) were the most common AI medicines [18‐21, 24, 27, 29‐31], followed by antidiabetic drugs [18, 20, 21, 24, 27, 29‐31], warfarin [18‐21, 24, 27, 30, 31], gastrointestinal agents [18‐21, 24, 30, 31], non-steroidal anti-inflammatory drugs (NSAIDs) [18‐21, 29‐31], antibiotics/anti-infectives [21, 24, 27, 29‐31] and anti-histamines [19‐21, 30].

The methodological quality of the included studies is detailed in Table 2. The external validity was high in 4 studies, as they reported on random population samples of community dwelling older adults aged ≥60 or ≥65 years [21, 24, 27] or in the case of Breslow et al., a random population sample with oversampling of older adults aged ≥60 years [20]. Ten studies were considered to have moderate external validity; five studies sampled random community dwelling older adults but with age restrictions which may have introduced selection bias [14, 19, 25, 26, 31] such as including adults aged between 53 and 75 years [25] or only including adults ≥75 years [19]. A further five studies reported on random population samples of community dwelling adults with subgroup analysis of older adults, however they did not report oversampling of older adults [22, 23, 28, 33, 34]. The six remaining studies were considered to have poor external validity, as the risk of selection bias is considered to be high [18, 29, 30, 32, 35, 36]. For example, Pringle et al. recruited enrollees in the Pennsylvania pharmaceutical assistance contract for the elderly, which may not be representative of all older adults as the members are older (mean age 78.8 years) and more likely to be female and white with multiple chronic conditions [30].

Internal validity was assessed by evaluating the potential risk of misclassification bias for both exposure and outcome across studies. Ten studies were considered to have a low risk of misclassifying exposure to medications which have the potential to interact with alcohol, as they used prescription claims data or in house inventories where they recorded details from the labels of medication containers or prescriptions and provided references supporting the inclusion of medicines as potentially alcohol interactive (Table 2) [14, 20, 21, 25, 26, 28, 30‐32, 34]. The potential for misclassification bias was considered high in the remaining studies as they relied on self-report for medication exposure [18, 22, 23, 27, 29, 33, 35, 36] and/or did not provide references supporting the inclusion of medicines as potentially alcohol interactive [19, 23, 24, 29, 33, 35].

While all studies relied on self-reported alcohol consumption, thus introducing potential biases in recall and reporting, 11 studies were considered to have a lower risk of bias as they reported on both quantity and frequency of alcohol consumption within a specified recall period, ranging from 1 week to 12 months (Table 2) [14, 18‐23, 25, 26, 31, 34]. The risk of misclassification bias, specifically underestimating exposure to alcohol, was considered to be higher in the remaining studies as they used quantity or frequency measures alone [24, 32, 35], did not specify the recall period [24, 27, 30, 35], restricted measurements to wine consumption [29, 36] and used a very narrow recall period “last working day or last weekend” [28]. Ascertainment of exposure to alcohol was unclear in one study [33].

In relation to outcome assessment, no study directly measured the concurrent use of alcohol and AI medications, rather all studies inferred concurrent use. Although possible for all studies, the risk of misclassifying the outcome of concurrent use was considered lowest in those studies identified as having a low risk of misclassifying exposure, who used the same recall period for both exposures or inferred concurrent use based on alcohol consumption within a specific recall period, ranging between 1 week and 12 months, and current or regular medication use (Table 2) [14, 20, 21, 25, 26, 31, 34].

As noted in Table 1, all estimates relate to studies from North America, Europe and Australia.

The prevalence of alcohol consumption ranged between 57 and 63% in studies reporting on nationally representative samples of community dwelling older adults (Table 3) [21, 24, 27]. Prevalence estimates of alcohol consumption were generally lower in studies sampling restricted age-groups, for example estimates ranged between 33.7 and 44% for studies restricted to older adults aged >70 years [26] and >75 years [19]. Similarly, prevalence estimates of alcohol consumption were generally lower in those studies considered to have a high risk of selection bias [18, 29, 30, 32, 35]. For example, 20% of older adults registered on the Pennsylvania pharmaceutical assistance contract for the elderly reported alcohol consumption [30]. All studies reporting on gender differences identified a higher prevalence of alcohol consumption in men [19‐21, 25, 29, 31, 36].

Exposure to alcohol interactive medications varied across studies; nationally representative studies of community dwelling older adults using objective measures of AI exposure estimate exposure at between 72 and 79% among the total study samples (Table 3) [20, 21]. Studies with objective measures of exposure to psychotropic medications in community dwelling older adults reported prevalence estimates of between 11.5–20% among the total study samples [14, 25, 26]. In contrast to alcohol consumption, use of psychotropic medications was significantly higher in women than men [14, 22, 25, 28, 32].

Eleven studies reported on the concurrent use of alcohol and alcohol interactive medications among the total study samples (Table 3) [14, 18‐20, 23, 24, 31‐33, 35, 36]. Eight of the 11 studies reported on alcohol consumption and a wide range of medicines with potential to interact with alcohol, with the prevalence of concurrent use ranging between 18 and 39% among the total study samples [18‐20, 23, 24, 31, 33, 35]. Only two of these studies were considered to have a low risk of misclassification bias for concurrent use [20, 31] . Breslow et al. estimated the prevalence of concurrent use among older adults at 35% [20]. Concurrent use was highest for cardiovascular agents (28%) followed by metabolic agents (17%), CNS agents (10%) and coagulation modifiers (10%). Approximately 7.75% of all older adults were identified as concurrent users of alcohol and psychotropic medications [20]. Qato et al. reported a prevalence estimate of 21%, with potential alcohol-medication interactions in older adults aged between 57 and 84 years being significantly more likely among men, white respondents, wealthier respondents, and those with higher education levels and greater comorbidities [31]. The prevalence of concurrent use was lower in studies which focused on psychotropic medications, ranging between 2 and 15.7% [14, 32, 36]. Only one study was identified as having a low risk of misclassifying concurrent use in older adults [14], with an estimated prevalence of 7.4% among the total study sample and 2.4% for concurrent heavy alcohol consumption and psychotropic medications. Older age (70–79 years), residing in rural or small town areas, living alone, higher social status, polypharmacy and a poor social network were independently associated with concurrent use of alcohol and psychotropic medications [14]. While a higher proportion (15.7%) of older adults in a French study [36] were identified as concurrent users of wine and benzodiazepines, the methods of ascertaining exposure to benzodiazepines was unclear, as was the assessment of concurrent use. Similarly, the lower estimate of 2% for concurrent use of alcohol and psychotropic medications in Sheahan et al. [32] is likely an artefact of their sample including patients in congregate care facilities.

Five studies assessed the prevalence of alcohol consumption among users of a broad range of alcohol interactive medications (Table 3) [19‐21, 27, 30]. Breslow [20] and Cousins [21] were identified as having a low risk of misclassifying concurrent use, and reported prevalence estimates of 45 and 60% respectively. Cousins et al. [21] found that older adults using AI medications were significantly less likely to report alcohol consumption compared to those unexposed to AI medications. Younger age (60–64 years), men, urban dwelling, higher levels of education and a history of smoking were independently associated with concurrent use of alcohol and AI medications [21]. A further six studies focused on psychotropic medication users [14, 22, 25, 26, 34, 36], four of which were identified as having a low risk of misclassification bias [14, 25, 26, 34]. Du et al. estimated the prevalence of alcohol consumption among users of psychotropic medications at 37.5% [14]. Similarly, Ilomaki et al. [25] reported that 38.9% of male psychotropic medication users, and 14.7% of their female users, consumed alcohol, with 26% of male and 8.3% of female users identified as heavy drinkers. Male psychotropic medication users consumed greater quantities of alcohol, and more often, than non-users. This pattern was not observed among women [25]. Between one-third and a half of all sedative or hypnotic users (34–54%) have been shown to consume alcohol [20, 21, 26, 34] with between 5 and 13% drinking heavily [21, 34]. Consistent with these findings, Ilomaki et al. [15] found that 26% of men aged greater than 70 years who use sedatives or hypnotics drink heavily; heavy drinking and daily drinking was significantly higher among male sedative or hypnotic users compared to non-users [26]. Concurrent use of alcohol among older adults taking antidepressants ranged between 42 and 53%, [20, 21] with 27% of male users aged greater than 70 years reporting concurrent use [26].

Only four studies reported on adverse outcomes. Three studies reported on falls, all three studies were cross-sectional [27, 32, 35]. A study by Immonen et al. [27] of 2100 older adults in Finland found that falls and injuries when a person has consumed alcohol in the past 12 months were more common among at risk drinkers (>7 drinks/week, or ≥5 drinks drinking days or ≥3 drinks several times/week) using AI medications (13.8%) compared to AI medication-users who were not considered as at risk drinkers (4.1%) (p < 0.001) [27]. In contrast to these findings, Sheahan et al.’s [32] study of older adults in America, which included patients in congregate care facilities, found that although the number of psychotropic drugs was associated with an increased odds of falling, the concurrent use of alcohol and psychotropic drugs was not. Similar non-significant associations were reported in Wong et al.’s [35] convenience sample of older adults in the US. One study examined the association between moderate alcohol consumption (≤ 40 g of wine per day) and adverse drug reactions among older adults at the point of admission to one of 81 acute care hospitals in Italy [29]. Among 22,778 participants, 3.9% were identified as having one or more adverse drug reactions. Moderate alcohol consumption was associated with a 24% increase in the odds of having an adverse drug reaction [29].