Patient-level interventions to reduce alcohol-related harms in low- and middle-income countries: A systematic review and meta-summary

Background Disease and disability from alcohol use disproportionately impact people in low- and middle-income countries (LMICs). While varied interventions have been shown to reduce alcohol use in high-income countries, their efficacy in LMICs has not been assessed. This systematic review describes current published literature on patient-level alcohol interventions in LMICs and specifically describes clinical trials evaluating interventions to reduce alcohol use in LMICs. Methods and findings In accordance with PRISMA, we performed a systematic review using an electronic search strategy from January 1, 1995 to December 1, 2020. Title, abstract, as well as full-text screening and extraction were performed in duplicate. A meta-summary was performed on randomized controlled trials (RCTs) that evaluated alcohol-related outcomes. We searched the following electronic databases: PubMed, EMBASE, Scopus, Web of Science, Cochrane, WHO Global Health Library, and PsycINFO. Articles that evaluated patient-level interventions targeting alcohol use and alcohol-related harm in LMICs were eligible for inclusion. No studies were excluded based on language. After screening 5,036 articles, 117 articles fit our inclusion criteria, 75 of which were RCTs. Of these RCTs, 93% were performed in 13 middle-income countries, while 7% were from 2 low-income countries. These RCTs evaluated brief interventions (24, defined as any intervention ranging from advice to counseling, lasting less than 1 hour per session up to 4 sessions), psychotherapy or counseling (15, defined as an interaction with a counselor longer than a brief intervention or that included a psychotherapeutic component), health promotion and education (20, defined as an intervention encouraged individuals’ agency of taking care of their health), or biologic treatments (19, defined as interventions where the biological function of alcohol use disorder (AUD) as the main nexus of intervention) with 3 mixing categories of intervention types. Due to high heterogeneity of intervention types, outcome measures, and follow-up times, we did not conduct meta-analysis to compare and contrast studies, but created a meta-summary of all 75 RCT studies. The most commonly evaluated intervention with the most consistent positive effect was a brief intervention; similarly, motivational interviewing (MI) techniques were most commonly utilized among the diverse array of interventions evaluated. Conclusions Our review demonstrated numerous patient-level interventions that have the potential to be effective in LMICs, but further research to standardize interventions, populations, and outcome measures is necessary to accurately assess their effectiveness. Brief interventions and MI techniques were the most commonly evaluated and had the most consistent positive effect on alcohol-related outcomes. Trial registration Protocol Registry: PROSPERO CRD42017055549


Introduction
Alcohol use is an important cause of chronic disease and injury. It is one of the top 5 risk factors for death and disability in the world [1][2][3]. The detrimental effects of alcohol use contribute to 3.3 million deaths and 139 million disability-adjusted life years (DALYs) lost globally each year [4]. Alcohol use has also been associated with risky behaviors, including crime, aggressive driving, interpersonal violence, and self-inflicted injury [5]. Such behaviors not only have harmful effects on the individual but also on the greater population [6]. Compared to high-income countries, low-and middle-income countries (LMICs) report higher rates of risky drinking behaviors, such as binge drinking and episodic drinking, as well as an earlier onset of alcohol consumption [4].
The World Health Organization (WHO) has placed an emphasis on the development and implementation of both policy-level and patient-level interventions to reduce harmful alcohol use in LMICs. While policy-level interventions are a crucial, cost-effective manner of reducing alcohol-related harms, context-appropriate, and effective patient-level interventions are also greatly needed to form multipronged alcohol harm reduction strategies [4]. A broad array of patient-level alcohol harm reduction interventions, such as brief interventions (for this paper, defined as any intervention ranging from advice to counseling, lasting less than 1 hour per session [7] up to 4 sessions [8], psychosocial interventions, and pharmacological treatments) have been found to be effective in high-resource settings [9,10]. Yet, alcohol use disorders (AUDs), characterized by moderate to severe alcohol abuse and dependence, remain a low priority of LMIC health systems [11]. Barriers, such as funding constraints, lack of policy, and low public awareness, often prevent access to psychosocial and pharmacological treatments that target AUDs [11]. Especially in some settings where alcohol use is culturally ingrained, adopting an alcohol harm reduction strategy, as opposed to focusing on abstinence, is crucial given the limited alcohol policy, health system treatments, and social support [12]. As such, WHO and The Lancet have recently issued calls to action to reduce hazardous alcohol use [4,13], yet the full scope of the evidence-based patient-level interventions to reduce harmful alcohol use in LMICs is missing from the literature. While narrative reviews of global alcohol-related harms have been published, we have found no systematic review conducted focusing on alcohol interventions specifically applicable to or evaluated in LMICs [1,11].
In order to address this gap, this paper aims to (1) review and describe the current published literature on patient-level alcohol interventions in LMICs; and (2) conduct a meta-summary of studies evaluating interventions to reduce alcohol use and harms in LMICs.

Protocol and registration
This systematic review is reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) Statement [14] (see S1 Table) and is registered in the PROSPERO database (International Prospective Register of Systematic Reviews) under the number CRD42017055549.

Eligibility criteria
Our primary criterion for article consideration was a patient-level alcohol or alcohol-related harm reduction intervention in a LMIC, as defined by our PICOS framework: LMIC Participants, patient-level Interventions, Compared to a control group, alcohol harm reduction Outcomes, all Study designs but focused on randomized clinical trials if there are enough. To be included, articles had to (1) evaluate a patient-level alcohol-related intervention's ability to reduce an (2) alcohol-related outcome in a (3) LMIC and be (4) peer-reviewed and published between January 1, 1995 and December 1, 2020. Study locations had to be classified as LMICs according to World Bank criteria at the time of the search [15]. The search strategy was inclusive of multiple study designs (randomized controlled trials [RCTs], prospective/retrospective cohort, quasi-experimental, or secondary data analyses with before and after intervention comparison) in case there was a dearth of literature from LMIC settings. Articles were excluded if they were abstracts only, literature or systematic reviews, meta-analyses, or commentaries. If 2 studies used the same data, then the most recent data were included in the review.

Information sources
We searched electronic databases (PubMed, EMBASE, Scopus, Web of Science, Cochrane, WHO Global Health Library, and PsycINFO) for articles that evaluated patient-level interventions aimed at reducing an alcohol-related outcome in LMICs. No studies were excluded for language. Additionally, we manually searched references and performed a citation analysis of the included articles using Web of Science and Google Scholar. Any citation that met the inclusion criteria based on the title and abstract was added.

Search
The initial search consisted of the MeSH terms "alcohol drinking," "low or middle income country," and "intervention." Search strategy demonstrates the search strategy used in PubMed, Embase, PsycINFO, and WHO Global Health Library databases (S1 Fig).

Study selection
Six pairs of reviewers from the specified individuals (KA, TC, SE, DE, SG, CP, LR, NS, AS, CY, and AP) independently reviewed the titles and abstracts, and any inconsistencies regarding inclusion were resolved by a third reviewer (DG or CS). Abstracts that did not provide enough information to determine eligibility were retrieved for full-text evaluation. Reviewers independently evaluated full-text articles and determined study eligibility. Disagreements were solved by consensus, and if disagreement persisted, a third reviewer's opinion was sought. After inclusion, we assessed each study for the study design. We reported all study designs in order to summarize the type and quality of study designs in the literature. Based on the large number of RCTs identified, we chose to narrow further analysis to RCTs.

Quality of studies
Since our systematic review included studies of different designs (RCTs, nonrandomized intervention, prospective/retrospective cohort, quasi-experimental, or secondary data/cross-sectional with before and after comparison), we opted to perform a data quality assessment according to study design using the following approaches. STrengthening the Reporting of OBservational studies in Epidemiology (STROBE) indicators were used for reporting observational studies. Two scales were used for nonrandomized studies: the A Cochrane Risk Of Bias Assessment Tool for Non-Randomized Studies (ACROBAT-NRS) [16] and Newcastle-Ottawa scale (NOS) [17]. Cochrane's revised risk-of-bias tool was used for randomized studies [18]. Finally, the Effective Practice and Organisation of Care (EPOC) suggested risk of bias indicators for interrupted time series studies (EPOC) [19]. We assigned risk of bias (low, moderate, and high risk) as suggested by the Cochrane Handbook [20] by study design. Studies were classified as (a) low risk of bias if all domains had low risk; (b) some concerns if at least 1 domain raised some concerns for bias; and (c) high risk of at least 1 domain was at high risk.

Data extraction
Five pairs of reviewers independently conducted the data extraction, and any disagreements were resolved by a third reviewer. General characteristics of the studies were recorded, such as year of publication, location where the study took place, inclusion and exclusion criteria, and participant characteristics. In addition, information on alcohol-related outcome measures, intervention type, and intervention impact or effectiveness measured as an effect size of outcome measures was extracted. The main outcome measures were Alcohol Use Disorders Identification Test (AUDIT) and Alcohol, Smoking and Substance Involvement Screening Test (ASSIST) scores, Rutgers Alcohol Problem Index (RAPI), number of drinking days, number of heavy drinking days, number of binge drinking days, drinks per drinking day, percent remaining abstinent from drinking alcohol, and percent relapsed back into drinking alcohol.

Data analysis
Initial evaluation of the papers indicated that a meta-analytical approach would result in high heterogeneity due to high methodological variability (e.g., outcome measures, study designs, and sample characteristics). Therefore, we conducted a meta-synthesis for all the included manuscripts, which qualitatively aggregated findings by grouping relevant findings into categories that represent the study's objectives (e.g., effectiveness of alcohol intervention). No manuscripts were excluded based on quality. The process involved summarizing main results of each included paper and performing a thematic analysis. Emerging themes on types of intervention and outcomes were presented. Interventions were grouped by similarity into 4 types: brief interventions, psychotherapy and counseling, health promotion and education, and biomedical treatments.
Using WHO and National Institute on Alcohol Abuse and Alcoholism (NIAAA) descriptions of brief interventions, we defined brief interventions as any intervention ranging from advice to counseling, lasting less than 1 hour per session [7] up to 4 sessions [8] independent of how the original study defined brief intervention. Interventions including a one-on-one interaction with a counselor that lasted longer than a brief intervention or that included a psychotherapeutic component were defined as psychotherapy and counseling. Motivational interviewing (MI) techniques could be included as either a brief intervention or psychotherapy and counseling, depending on how long and over how many sessions the intervention took place. A study was considered health promotion and education, independent of the study's definition, if an intervention encouraged individuals' agency of taking care of their health, such as risk reduction skills and health education [21]. Biomedical treatments were used as a taxonomy to group studies that had the biological function of AUD as the main nexus of intervention, including brain stimulation and medicines.

Study selection and description
In total, 5,036 abstracts were reviewed. From those, 500 articles were manually reviewed to identify 117 articles matching our inclusion and exclusion criteria (Fig 1). No studies were excluded based on language. Of these 117 studies, 75 were RCTs (Table 1) utilizing a vast array of interventions, which we categorized into 4 main categories of interventions, including brief interventions (24 studies) ( Table 2), psychotherapy or counseling (15) ( Table 3), health promotion and education (20) (Table 4), and biological treatments (19) ( Table 5). One study by Shin and colleagues had one arm in biomedical treatments and another arm in brief intervention [22]. Two other studies had one arm in psychotherapy or counseling and another arm in health promotion and education [23,24]. These 75 studies were performed in 15 countries, representing 8 upper middle-income countries (60% of studies), 5 lower middle-income countries, and 2 low-income countries (7% of studies) (S2 Fig). The majority of the studies came from Brazil (28%) and India (20%). Alcohol-related outcomes found included alcohol quantity or frequency measure, intention to use alcohol, use/abstinence/remission proportion or frequency, alcohol-related scores, alcohol cravings or cravings per day, or alcohol use during pregnancy or before sex.

Meta-summary
Brief interventions. The brief interventions category had the greatest number of RCTs in our study with 24 RCTs, and these interventions were the most similar to each other. The types of interventions included most commonly were WHO-based brief interventions (which utilizes some MI techniques) [28,29,[79][80][81][82] or MI interventions [22,58,76,88,89,91,93]. Some studies focused more on the intervention delivery, specifically nurse or layperson [67,70,71,75] or computer-based interventions [30,34,35,41]. Outcomes were also varied including harmful alcohol use scores (AUDIT) or alcohol misuse (ASSIST), abstinence or remission (ASSIST), and percent or number of days of drinking or heavy drinking.
Overall, the majority of the studies evaluating brief interventions demonstrated evidence of efficacy in one or more of their alcohol-related outcomes, for both short-(up to 3 months) and

PLOS MEDICINE
Systematic Review of Alcohol Intervention in Low-and Middle-Income Countries At 3 and 6 months, Assanangkornchai and colleagues found similar significant reductions in the frequency of alcohol use and other substances in both the intervention and control groups at the primary care setting [28]. Penpgid and colleagues did not find evidence of efficacy of a mixed alcohol and tobacco brief intervention compared to an alcohol-only or tobacco-only session on past week alcohol use and Alcohol ASSIST score. All 3 arms did have a significant reduction in their alcohol consumption compared to baseline [82]. Babor and colleagues found at 9 months that males who received brief or simple advice reported a 17% lower average daily alcohol consumption compared to the control group, with a reduction in the intensity of drinking at about 10%. Females reduced their consumption in both groups without between-group differences [29]. Pengpid and colleagues found that at 12 months postintervention, university students had a significant reduction in AUDIT score compared to control [81]. As for outpatients, they found no significant differences in the reduction in relation to control [80]. Peltzer and colleagues evaluated the impact of a brief intervention versus a health leaflet for TB clinic patients and did not find evidence of efficacy between control and intervention at 6 months (79).
Face to face and computer based (1)  Mertens and colleagues found those who received a nurse practitioner-delivered brief intervention reduced patients' alcohol ASSIST scores at 3 months by 38% versus 21% in the control arm [67]. At 3 months, Noknoy and colleagues found a significant difference in number of binge drinking days between intervention (0.29) and control group (1.36), but at 6 weeks and 6 months, there was no significant difference. At 6 weeks, 3 months, and 6 months, there were significant differences in the average drinks per drinking day between intervention (3.00, 2.73, and 2.26) and the control group (4.85, 5.06, and 4.02), but no difference in the number of drinking days between baseline and follow-up [75]. Nadkarni and colleagues found 36% remission (AUDIT 12-19) of alcohol use in the intervention group compared to the 26% of the control group. At 3 months follow-up, abstinence was significantly higher in the intervention (42%) compared to control (18%) groups. No effect on mean daily alcohol consumption or percent days of heavy drinking differences was found [70]. Results at 12 months showed maintained and enhanced effects on alcohol-related outcomes [71]. A pilot study found that for men with AUDIT>20, the CAP intervention arm had nonsignificant favorable outcomes for remission, proportion of nondrinkers, and ethanol consumption at 3-and 12-month follow-up as compared to enhanced usual care [72].
MI (6) 1, 3, and 6 months Alcohol Consumption Questionnaire RAPI score ACRQ APRA Alcohol abstinence ASSIST Drinking days ASI Segatto and colleagues found significant reduction in alcohol-related problems and alcohol use in the brief intervention and alcohol educational brochure groups but no significant differences between the groups for days of use and amount of use, RAPI, ACRQ and APRA scores, at 3 months follow-up [88]. Signor and colleagues found a significant difference between groups in the reduction of participants consuming alcohol at 6 months follow-up (70% of individuals in the helpline-based brief intervention group and 41% in the control/ minimal intervention group) [89]. Ward and colleagues found that those who received a brief MI at the primary care setting and resource list were more likely to reduce alcohol misuse than control at 3 months [93]. Pal and colleagues found men who received a brief intervention had a decreased average amount of alcohol use in prior 30 days (24.7 to 10.1 versus 26.1 to 19.1) and decreased Addiction Severity Index (0.36 to 0.18 versus 0.42 to 0.33) at 3 months compared to those who received simple advice only [76]. A significant reduction in AUDIT scores at 3 months follow-up was observed by Kamal and colleagues for an on-campus, nurse-delivered brief alcohol screening, and intervention as compared to general advice. The intervention group also had a significant shift of participants from high-to low-risk AUDIT zone as compared to the control group [58]. Shin found no differences in the proportion of abstinent days between intervention and control in a TB clinic [ WHO-based brief interventions were found to be efficacious to reduce average daily alcohol consumption in males at the health setting [29] and AUDIT average scores in university students [81]. With brief interventions delivered from a motivational intervention framework, Signor and colleagues found at 6-month follow-up, 70% of individuals in the helpline-based brief intervention group and 41% in the control/minimal intervention group had remained abstinent [89]. Similarly, this mode of delivery showed evidence of efficacy at the primary care setting to reduce alcohol use [76,91,93] and with university students [58]. Simao found that college students had a significant reduction in the amount of alcohol use per occasion and AUDIT scores up to 24 months after the intervention [90]. Other modes of delivery revealed that lay counselor-delivered interventions had significant differences between intervention and control [67,71,75], and a computerized intervention reduced alcohol use as much as an inperson motivational intervention [41]. One study focused on the efficacy of a women-focused brief intervention demonstrated efficacy of the interventions in reducing heavy drinking behavior and heavy drinking days in women [94].

MI and PST
A number of the studies found a reduction in alcohol-related outcomes. Daengthoen's intensive inpatient rehabilitation combination therapy intervention had a reduction in alcohol consumption and drink cravings [45]. Nattala found that a significantly higher percentage of dyadic intervention patients (57%) were abstinent compared to individual treatment (27%) or treatment as usual (30%) patients [73]. Sorsdahl and colleagues found a reduction in the Rendall-Mkosi and colleagues found that compared to the control, a 5-session intervention reduced the proportion of women at risk for AEP (51% intervention and 28% control) at 3 and 12 months. There were declines for both groups in the proportion of women who met criteria for risky drinking at 3 and 12 months (intervention 14.75% versus 10.94%), but the difference between the 2 groups was not significant [83].  Jirapramukpitak and colleagues found that home-based CM did not improve continuous abstinence over the 12-week intervention period. The higher-magnitude CM intervention arm did have a significantly higher abstinence rate in the postintervention followup period [54]. Rotheram-Borus and colleagues studied a HV for prenatal and postnatal visits for pregnant women up to 36 months postdelivery and did not find a direct association between intervention and alcohol use [84]. Bolton and colleagues found no differences in AUDIT scores between a control and intervention in Burmese refugees in Thailand [37].

PLOS MEDICINE
School based (6) STEP for HIV/AIDS and alcohol use 10 weeks Intention to use Chhabra and colleagues found no differences in intention to use alcohol after implementation of a STEP program compared to control at 10-week outcome assessment [40].
School-based curriculum using communication competence theory to develop use resistance strategies 8 months Drinks per day Drinking days Marsiglia and colleagues found that after an implementation of a school-based curriculum, both intervention and control groups had an increase in the amount of use and frequency of use, yet the intervention group had significantly less increase in amount and frequency of use [65]. ASSIST scale for those who received the MI with problem solving intervention compared to the control, yet there was no difference between the MI alone and the control group [92]. L'Engle, Rendall-Mkosi, and Moraes all had significant findings for their MI interventions reducing binge drinking up to 12 months, reducing the proportion of women at risk for alcohol-exposed pregnancies and increasing the proportion of abstinent patients [23, 61,83]. Ng and colleagues used a body-mind-spirit multidimensional intervention and reported significantly less alcohol cravings, drinking days, drinks per drinking day, and relapse in the intervention group compared to treatment as usual at 3 months [74]. Randomization to receive CBT, in different modalities, was found to be associated with a higher reduction in drinking days, drinks per drinking days [77,78], and AUDIT score [63] in comparison to usual care, at 3 months for HIV-infected outpatients reduction in mean AUDIT score [68], and alcohol consumption [53] in participants positive for intimate violence.
A few of the studies in this subgroup had null effects or found no difference between the intervention and control arms. Marques and colleagues found a reduction in many of their alcohol-related outcomes for the group and individual intervention arms at 15 months, but the HIV-alcohol risk reduction intervention 1, 3, and 6 months Alcohol use in sexual context Anticipated outcome of alcohol use Kalichman and colleagues found that a behavioral risk reduction counseling intervention for sexually transmitted infection clinic patients had a reduction in alcohol use and expectancies that alcohol enhances sexual experiences at 3-month follow-up [57]. Kalichman and colleagues found that compared to a 1-hour HIValcohol education group, the 3-hour brief behavioral HIV-alcohol risk reduction intervention reduced alcohol use before sex at 3 and 6 months [56]. Ahmadi [78].
HIV SR reduction arm and MI+risk reduction 3 and 6 months AUDIT Witte and colleagues studied the efficacy of a relationship-based SR reduction intervention, SR reduction intervention with MI compared to a wellness control to reduce harmful alcohol use among female sex workers. All groups were effective in reducing the AUDIT score from baseline to 6 months (wellness promotion −30.98 to 18.30, risk reduction −28.42 to 18.12, and risk reduction and MI −32.64 to 21.72), but there was no significant difference between groups [95].
Multifaceted district level mental healthcare plan + brief intervention 12 months AUDIT Jordans and colleagues found no statistical significant difference between control and intervention for the reduction in AUDIT scores from baseline and follow-up (B = 12.16; CI 95% −6.10; 1.79) [55].
Disulfiram (5) 9 and 12 months # of days of abstinence # days until relapse # drinks per week # drinks per occasion OCDS The groups receiving disulfiram showed higher frequency of days of abstinence, higher days to first relapse, less craving and less relapse events than topiramate in alcohol-dependent men [49]. A similar pattern of results were observed when comparing disulfiram to naltrexone, but had higher cravings. No differences were observed in the amount of days to the first alcohol used [46,48,50]. Compared to acamprosate, the group receiving disulfiram showed higher abstinent days, fewer relapse events, and a higher number of days until first alcohol use and to first relapse. No difference was observed in the total number of abstinent days and a higher craving was observed in the disulfiram group Escitalopram + electroacupuncture (1) 4 weeks PACS Zhao and colleagues found that after 4-week treatment, the global scores of PACS declined significantly in both the escitalopram with electroacupuncture and the escitalopram without electroacupuncture groups (both P < 0.05). Furthermore, the decline in the rea -electroacupuncture group was superior to that in the sham electroacupuncture group (P < 0.05) [96]. Klauss and colleagues found that the percentage of relapse at 6 months follow-up was higher in the sham group (88%) than the tDCS group (50%) with no difference in cravings between the groups [59]. However, an intensive tDCS scheme was associated with a larger reduction in alcohol cravings when compared to sham-based control, also associated with lower relapse up to 3 months postintervention [60].
intervention arms were not significantly different from each other [64]. Similarly, Satyanarayan found a reduction in Severity of Alcohol Dependence Questionnaire (SADQ) scores for CBT and usual care arm patients but no significant difference between intervention arms; authors believed that this was because both intervention arms received similar alcohol reduction strategy intervention components [87]. Alternatively, Shin and colleagues found that their intervention, which focused on inpatient tuberculosis patients with severe AUDs, caused no change in alcohol-related outcomes, likely because the study did not include alcohol treatment-seeking patients, but had patients with low readiness to change or poor intervention participation rates combined with relatively low enrollment numbers [22]. Health promotion and education. In total, we found 20 RCTs, which evaluated health promotion and education interventions. Of these, 2 were based in the workplace [26,38], 5 in the community [23,37,43,84], 6 in schools [33,40,65,85,86], and 7 in clinics [39, [55][56][57], of which 1 was focused on women sex workers [95]. The majority of programs addressed alcohol use in the context of HIV/AIDS prevention and risk reduction [38][39][40]43,56,57,95].
About half (8 of 17) of the health promotion and education interventions were found to have positive results [25,26,38,39,54,56,57,65,66,84]. Some of these studies also had mixed results. For example, Aira and colleagues found a reduction in drinks per day and an improvement in attitudes toward drinking, but not a reduction in the total amount of alcohol consumption [26]. Similarly, Rotheram-Borus found that home visits for pre-and postnatal women were associated with a reduction in the use of alcohol during pregnancy, but this drinking resumed postpartum [84].
Meanwhile, a majority of the studies that found no effect of their interventions either were not adequately powered to detect the alcohol-related outcome [23,43] or were compared to another intervention rather than a control, thus potentially obscuring some potential reduction in harm [95]. Cubbins and colleagues evaluated a community-level intervention in which popular community individuals relayed education through casual conversations and found significant alcohol reduction in both the intervention and control groups, but no difference between the groups [43]. Chhabra and colleagues looked at the effectiveness of a Severity of Alcohol Dependence Questionnaire (STEP) school-based program but found that students, and more specifically girls, had an immediate reduction in their intent to use alcohol, but there was no difference in the intention to use alcohol at the 10-week outcome assessment [40].
Biomedical treatments. The final group of RCTs evaluated biomedical treatments and included 19 RCTs evaluating medications and brain stimulation. The  and one evaluated baclofen with a brief intervention [52]. One RCT evaluated the efficacy of adding acupuncture to an escitalopram treatment regimen [96].
The RCTs evaluating naltrexone and ondansetron found limited impact on abstinence, number of heavy drinking days or number of abstinent days [22,24,32], and abstinence [43]. Mixed effects were found by RCTs for topiramate, where Baltieri found an increase in abstinence at 4 weeks, although Likhitsathian found no differences in any drinking quantity or frequency measures up to 12 weeks [32,62].
On the other hand, the RCTs evaluating gabapentin, acamprosate, and baclofen exhibit more positive results. Furieri found that gabapentin was associated with a significant decrease in quantity and frequency of drinking and higher mean abstinent days [51]. Baltieri found that acamprosate improved abstinence rates but only for those who participated in AA [31]. Moreover, Gupta found that patients who received baclofen compared to a nutritional supplement, with a brief motivational intervention, were more likely to remain abstinent, have lower heavy drinking days, and fewer alcohol cravings [52].
We identified 4 RCTs that studied transcranial direct current stimulation, and all of them occurred at 2 institutions in Brazil. While 3 of these studies found a decrease in alcohol cravings compared to sham stimulation [36,44,60], one study found a lower relapse rate in the brain stimulation group but with no difference in alcohol cravings at 6-month follow-up [59]. Ultimately, 3 of the 4 studies in this group found more relapses in the intervention group at 4-week, 6-month, and 12-month follow-up [44, 59,60].

Discussion
This is the first review, to our knowledge, of alcohol harm reduction interventions evaluated in LMICs. Most studies we found took place in middle-income countries; there was a noticeable gap of studies in the Middle East, North Africa, Europe, Central Asia, and South Asia regions. Overall, we found that there was limited uniformity for interventions, outcomes, and followup times across studies, which limited our ability to compare results. The vast majority of evaluations were limited to middle-income settings, leading to feasibility and generalizability concerns for low-income settings. Of all the RCTs, brief interventions were the most commonly studied; similarly, MI techniques were the most prevalent behavior change technique common in both brief and psychotherapy and counseling interventions. Brief interventions and motivational interviewing techniques also had the most consistent positive results in our findings.

Lack of uniformity limits effective comparisons
The studies included in our meta-summary used a wide variety of metrics to measure alcoholrelated outcomes of alcohol interventions; these metrics included (i) AUDIT scores; (ii) ASSIST scores; (iii) # of drinking days; (iv) # heavy drinking days; (v) # drinks per drinking day; (vi) # abstinent days; (vii) # drinks per day; (viii) % of patients abstinent; and (ix) % of patients relapsing.
The time period over which these outcomes were measured also varied considerably, from 3 months [41] to 24 months [90]. This lack of uniformity compromised our ability to discern the effectiveness of interventions or to compare results across studies. The diversity of alcohol consumption outcomes measures is due in part to varying recall, reference period, and definition of a "standard drink" [97][98][99]. Future study studies may benefit from using consistent outcome measures and adopting uniform methods of intervention implementation or study designs.

Uncertain feasibility of implementing interventions in low-income country setting
The vast majority of the studies in this review were conducted in middle-income countries. Thus, the feasibility of implementing these interventions and their effectiveness in low-income settings is uncertain. Low-income countries face greater barriers (such as scarcity of medical facilities, limited training available to medical staff, infrastructural barriers to healthcare access, and effective patient communication/follow-up) to implementing effective healthcare than either high-or middle-income countries. As a clear example, all 4 studies [36,44,59,60] that used brain stimulation as an intervention were conducted in Brazil, an upper middle-income country. In addition to its uncertain effectiveness, brain stimulation requires expensive equipment and specific facilities, and it is not likely to be feasible in some low-income country settings. In another example, although psychotherapy and counseling interventions are demonstrably effective [45, 61,73,83,92], none of these studies took place in a low-income country, so the feasibility of implementing this type of intervention is uncertain. Given that infrastructure in low-income settings is even more limited in mental health and substance abuse facilities and professionals, with a greater associated stigma, an alcohol use reduction intervention implementation of this kind is still potentially unfeasible. Similarly, medication shows some evidence of a positive effect on abstinence from alcohol, but reliable availability of medication is essential for this intervention to be effective; thus, medication may not be a feasible intervention in a low-income country [100,101].
Instead, the most studied and potentially most feasible intervention is a brief intervention. In our systematic review, 6 studies evaluated brief interventions in South Africa [67,[79][80][81]93,94]. Brief interventions have been studied to decrease alcohol use and alcohol-related consequences in a variety of settings and countries [102][103][104][105]. They have also been suggested to be cost-effective in high-income countries [106]. In low-and middle-income settings, brief interventions are likely to be feasible because they can be delivered by nonprofessionals requiring less training.

Limitations
There are a few limitations to our study. First, our search strategy did not exclude studies due to language, and, yet, we found no manuscripts in other languages. Thus, either there is no non-English language literature on this topic or the data sets we searched have limited non-English language articles. Second, our ability to conduct a thorough meta-analysis was restricted by nonuniform outcome measurements, a wide variety of outcome assessment times, and a wide variety of interventions, making it difficult to compare interventions and their effects. To compensate for this, we summarized results from RCTs qualitatively. Similarly, we conducted our database search for only LMICs at that time. This might limit our findings by excluding articles from countries that have become high income since the study occurred or erroneously including countries that were high income but then reduced their status at the time of the database search; in the former case, we cannot determine the number of potential studies, but for the latter case, we rechecked the World Bank status of all countries and their intervention time periods to ensure this was not occurring. Finally, we tried to group types of interventions based on standard definitions rather than study-specific descriptions that might limit the interpretation of effect size and differ from the original author's description.

Improving future research
Future alcohol harm reduction intervention studies should use uniform reporting. Studies ranged widely in their intervention type, framework, population, augmentation, or boosters, as well as outcome assessment frequency and timing. Adherence to 1 or 2 sets of standardized outcome reporting measures at a specified time period would greatly improve comparability across time and geographic location, allowing for a meta-analysis of intervention methods. Based on our review, brief interventions using the ASSIST or AUDIT scoring systems are the most widely used and appear to provide the best standardization among outcomes. Overall, future research should include both comparative effectiveness to determine best interventions for LMIC settings but also most effective implementation strategies including target populations.

Conclusions
In conclusion, alcohol harm reduction interventions in LMICs are nonuniform in nature, skewed in geographic regions where applied, and result in uncertain effectiveness over varying time horizons. Feasible options specific to low-income countries are most likely brief interventions and interventions that utilize motivational interviewing techniques. Identifying uniform methods of implementation and assessment of alcohol harm reduction interventions can be a first step toward establishing a set of evidence-based protocols for treatment for low-income settings. Current studies in brief interventions, psychotherapy, and brain stimulation show promise, but have been tested primarily or exclusively in middle-income settings. Feasibility testing in low-income settings, comparative effectiveness testing, and uniform reporting methods are needed to help determine the most effective alcohol harm reduction strategies for lowincome settings in order to address this global health crisis.