Prevalence and factors associated with Alcohol Use Disorder among People Living with HIV/AIDS in Africa: A systematic review and Meta-analysis.

Background: Alcohol use disorder (AUD) in HIV/AIDS patient’s decreases adherence and effectiveness of medications and help-seeking to HIV/AIDS care and treatment. This study, therefore, assessed the average prevalence and associated factors of alcohol use disorder in HIV/AIDS patients. Methods: We did an electronic data search on PubMed, Scopus, EMBASE, Psych-INFO libraries, African index Medicus and African Journals Online (AJOL). Google scholar was also investigated for non-published articles. The reference lists of published articles were also reviewed. The stata-11meta-prop package was employed. Subgroup and sensitivity analyses were done. Cochran's Q-statistics and the Higgs I 2 test were used to check heterogeneity. Publication bias was evaluated with Egger's test and funnel plots. Results: Of 1362 articles identied using the search strategies; only 22 studies were included in the nal analysis. The average prevalence of AUD was 22.03% (95% CI: 17.18, 28.67). The average prevalence of AUD in South Africa (28.77%) was higher than in Uganda (16.61 %) and Nigeria (22.8%). The prevalence of AUD in studies published before 2011, 2011-2015, and after 2015 was found to be 13.47%, 24.93%, and 22.88% respectively. The average prevalence of AUD among studies with a sample size > 450 was 16.71% whereas it was 26.46% among studies with a sample size < 450. Furthermore, the average prevalence of hazardous, harmful, and dependent drinking was 10.87%, 8.1%, and 3.12 % respectively. Our narrative analysis showed that male sex, cigarette smoking, family history of alcohol use, missing ART medication, mental distress, khat chewing, low CD4 count, and low income were among the associated factors for AUD in people with HIV AIDS. On quantitative meta-analysis for associated factors of AUD, the AOR of being male, Cigarette smoking and khat chewing were 5.5, 3.95, and 3.34 respectively.

The prevalence of AUD in patients with retroviral infection has been evidenced higher than people without retroviral infection or the general public (6). A systematic review and meta-analysis study by Duko et al.2019 (7) showed that the average worldwide prevalence of AUD among patients with HIV/AIDS was found to be 29.8%. Other earlier studies in Africa reported that the prevalence of AUD ranges from 17% to 39.4% in Nigeria (8-10) ,1.4% to 33% in Uganda (11)(12)(13)(14), 6.6% to 48.5% in South Africa (15)(16)(17)(18) ,5.4% to 33% in Kenya (19)(20)(21)(22), and 14.8% in Zambia (23). There exists a scarcity of average data on alcohol use disorders in people with HIV in other African countries despite the high disease burden with people living with HIV(PLWHA)(24).
In Ethiopia studies conducted so far implied that AUD among HIV/AIDS patients was documented to be 32.6% in Jimma university specialized hospital (25), and 31.8% in Hawassa university comprehensive hospital(26), 14.2% in Bishoftu general hospital (27), 24.8% in Assela teaching hospital (28). Moreover, a systematic review and meta-analysis study of alcohol consumption in Ethiopia by Ayano et al.2019 (29) reported that the prevalence of hazardous alcohol consumption was 8.96%.
Factors found to be associated with AUDs in HIV/AIDS patients include male gender, psychological morbidity, smoking cigarettes, the Christian religion, lower education, peer pressure, parental modeling, and social pressures (9,10,(30)(31)(32). A study from Jimma, Ethiopia also reported that being male, smoking cigarettes, and psychological distress as the associated factors for AUDs (25). Similarly, another study from Hawassa, Ethiopia found factors such as being male, poor social support, medication nonadherence, khat chewing, and cigarette smoking as having a signi cant association with AUD (26).
The impacts of AUDs among PLHIV include decreased help-seeking to HIV/AIDS care and treatment facilities (33), adherence problem to antiretroviral treatment (ART) drugs (3,(34)(35)(36)(37), a decline of CD4 cells, increased load of the virus (36,38), rapid HIV/AIDS disease progression and rising of opportunistic diseases (38)(39)(40)(41)(42), development of drug-resistant HIV strains (3,36,38), and nally contributes to premature death (43,44). A study in the USA approved that if an individual with HIV consumes alcohol once/week his life span will be shortened by two years and by 6.5 years if alcohol is consumed daily (45). Besides, patients with AUD might afraid of interaction between ART drugs and alcohol so that, skipping ART drugs during alcohol intake (9,34). Risky sexual and non-sexual behaviors (44,(46)(47)(48) are also common in people with AUD.
Even though multiple studies have been performed and showed that retrovirally infected patients have a higher prevalence of AUD than the general population in Africa (7-23, 27, 28, 49-52), to date no published study in Africa per investigators knowledge reported the average prevalence and associated factors of alcohol use disorders in human immune-de ciency virus-infected patients. Knowledge of the average prevalence of AUD and detecting its associated factors would assist policy-makers and program implementers in deciding evidence-driven prevention and promotion and treatment activities in this area. So, this systematic review and meta-analysis study is warranted to (1) review the existing pieces of evidence on alcohol use disorder and its associated factors and (2) to determine the average prevalence of AUD and associated factors among PLWHA in Africa.

Methods
This Preferred Reporting Items for Systematic Reviews and Meta-analysis Protocols (PRISMA-P) 2015 (53) had been used as an important guideline during the current study.

Data sources and search strategies
We conducted our search in the following libraries; EMBASE, Psych-Info and Scopus, PubMed, African Index Medicus, and AJOL with a systematic search approach. Our search was not delimited to a speci c year's limitation of the studies. We conducted our search in PubMed with the following key terms and words: (Prevalence OR epidemiology OR magnitude OR incidence) AND (Alcohol use disorder OR alcohol abuse OR alcohol use) AND (HIV OR human immunode ciency virus OR AIDS OR PLWHA OR ART) AND (factor OR risk OR risk factor OR determinant) AND (Southern Africa OR Central Africa OR East Africa OR North Africa OR Western Africa OR Sub-Saharan Africa). Google scholar was also searched for nonpublished articles. The reference lists of included studies were far investigated adequately to further include unaddressed eligible literature. For further clari cation of unclear ideas, we contacted the authors of the included articles.

Inclusion criteria
All cross-sectional, case-control, and cohort studies that had reported the prevalence of alcohol use disorders and/or it's associated with human immune-de ciency virus-infected patients in Africa were included. A research article was eligible for inclusion to the current study if it was an observational study (cross-sectional, cohort and case-control study), the outcome studied should be AUD and the factors associated with it, and must be conducted in people living with HIV/AIDS who are on ART. In addition to this, the article had to be published in the English language and the area of study must be in Africa.

Exclusion criteria
During the screening and analysis stage of the current study; we excluded letters, previously done systematic reviews and meta-analysis studies, interventional studies, commentaries, and editorials. Duplicated studies were also excluded to avoid being doubled in the analysis.

Selection of studies for inclusion in the review
Articles retrieved from the search databases were stored, managed, and used in an EndNote reference manager. Two review authors (MN, AB) reviewed each study's title and abstracts stored in an EndNote reference manager independently and any disagreements between them whether to include or exclude the study for analysis was solved through discussion with the third author (YG).

Data extraction and management techniques
Using a standardized data extraction form, three authors (MN, AB, and YG) extracted data from included studies independently. The following elements were components of the data extraction form. First author last name, year of publication, study setting, sample size, number of events, data on the prevalence of alcohol use disorder, the tool used for assessment, associated factors, and odds ratio (OR) with 95% con dence intervals (CI).

Quality assessment methods
Two review authors (MN and AB) assessed the quality of all included studies independently. Differences regarding the quality score of included studies between the above review authors were solved with evidence-based discussion and a third reviewer (MN). The modi ed version of the Newcastle-Ottawa scale (NOS) (54) was used as a guideline for quality appraisal of the included studies which are cohort in design (11,17,21,22). Representativeness and size of the sample, comparability between study subjects, ascertainment of alcohol use disorder symptoms, and statistical quality were the dimensions of NOS in assessing the quality of each study. The quality of cross-sectional studies included (8-10, 12-16, 18-20, 23, 27, 28, 49, 51, 52, 55) was also evaluated using the Johanna Briggs Institute (JBI) critical appraisal checklist for prevalence studies (56).
The dimensions of JBI checklist for quality appraisal of prevalence studies include; the appropriateness of sample frame, sampling procedure of participants, adequacy of the sample size, appropriateness in the description of study subjects and setting, appropriateness of data analysis, usage of valid methods used for the identi cation of the condition, reliability of measurement, appropriateness of statistical analysis, and adequacy of the response rate.

Data synthesis and analysis
The qualitative data for the present study was done with a narrative description and systematic review.
The average prevalence of AUD and the average odds ratios (OR) of associated factors were calculated through random-effects (57) and quality-effects models (58). Heterogeneity between the included studies was investigated with both Cochrane's Q-statistic and the Higgs I 2 statistics. A Higgs-I 2 value greater than 50% was considered as an implication of heterogeneity between studies (59).
Since the average prevalence of AUD was found to have signi cant heterogeneity in doing the analysis, a sub-group analysis was performed to explore the sources of heterogeneity. All statistical analyses had been performed using Meta-XL version 5.3 (60) and the STATA11 Meta-prop package (61). Publication bias was assessed with funnel plot (62) and Eggers regression test.

Identi cation of studies
A total of 1357 articles were identi ed in our electronic search using the speci ed databases. and a manual search for the reference lists other articles resulted in 5 additional articles. This makes the total number of search results to be 1362 articles. Of these, 28 were duplicates and removed. A detailed screening of the remaining articles resulted in only 62 articles to be reviewed their full text for eligibility.
Finally, only twenty-two articles were included in the analysis by ful lling our pre-speci ed inclusion criteria ( Figure 1).

Quality of Included Studies
In general, the summary quality assessment result of cohort studies included in the current meta-analysis ranges from 7 to 10 according to the scoring system of Newcastle Ottawa scale; one with moderate and three with good quality. The quality assessment for the remaining 18 cross-sectional studies based on the JBI checklist for prevalence studies ranges from 6 to 10; implying an appropriate methodological quality of the included studies.
The average prevalence of alcohol use disorder among HIV/AIDS patients who were on antiretroviral therapy in Africa Twenty-two studies had been included in the nal meta-analysis to determine the average prevalence of AUD among patients on antiretroviral therapy in Africa (7-23, 27, 28, 49-52). The reported prevalence of AUD among included studies in this review and meta-analysis study ranges from as low as 1.4% in Uganda (12) to as high as 48.5% in South Africa (16). The average prevalence of AUD among patients on antiretroviral therapy in Africa using the random effect model was 22.03 % (95% CI: 17.18, 28.67). This average prevalence has been in uenced by a signi cant heterogeneity (I 2 =99.8%, p-value < 0.001) from the difference between the incorporated studies ( Figure 2).
The average prevalence of hazardous alcohol use among HIV/AIDS patients who are on antiretroviral therapy in Africa.

The average prevalence of harmful alcohol use among HIV/AIDS patients on antiretroviral therapy in Africa
Seven studies reported data on the prevalence of harmful drinking in HIV/AIDS patients (8,10,19,22,27,49,51). The average prevalence of harmful drinking among these studies was obtained to be 8.1% (95% CI: 1.04, 15.17) and was having a signi cant heterogeneity (I 2 =99.5%, p-value < 0.001) ( Figure 5).
The average prevalence of dependent drinking among HIV/AIDS patients who are on antiretroviral therapy in Africa Seven other studies (8,10,19,22,27,49,51) also reported dependent drinking in HIV/AIDS patients on antiretroviral therapy. The average prevalence of dependent drinking in these studies was 3.12 % (95% CI: 1.45, 6.70) and an obvious heterogeneity has also been detected in the result (I 2 =99.6%, p-value < 0.001) ( Figure 6). The average prevalence of dependent drinking among studies that utilized a sample of more than 400 (8,18,22,27,51)

Sensitivity analysis
To detect further the source of heterogeneity that in uences the average prevalence of AUD, we also did one study leave out at a time sensitivity analysis. The result from the sensitivity analysis revealed that the average estimated prevalence of AUD obtained when each study was left out from analysis was within the 95% con dence interval of the average prevalence of AUD when all studies were run together. Therefore, the result of the average prevalence of AUD in HIV patients was not in uenced by a single particular study. Moreover, the sensitivity analysis result revealed that the average AUD prevalence ranges between 20.77 (95% CI: 16.33, 25.31) and 22.98% (95% CI: 18.05, 27.91) when each study was excluded (Table 3).

Publication bias
We carried out an Egger's publication bias plot to detect the presence of publication bias but it is near the origin and the result of Eggers publication bias plot had insigni cant p-value(P=0.22), on condition that no substantial publication bias for the prevalence AUD in Africa. Moreover, a visual inspection from a funnel plot for a Logit event rate of prevalence of AUD in HIV AIDS patients against its standard error suggests additional evidence for the absence of a small study effect (Figure 7).
The association between male sex and alcohol use disorder in HIV/AIDS patients The association of being a male and higher risk of AUD in HIV/AIDS patients was reported in seven of the included studies (7,9,10,16,28,49,51). The average adjusted odds ratio of the increased risk of having AUD was 5.5 (95% CI: 1.10, 9.98) (I 2 =90%, P< 0.01). This implied that male HIV/AIDS patients who were on ART were 5.5 times at higher risk of having alcohol use disorder as compared to female patients who were on ART therapy.
The association between cigarette smoking and chat chewing with an alcohol use disorder Among the 22 studies incorporated in the current meta-analysis (7-23, 27, 28, 49-52), four (7,27,49,51) had reported cigarette smoking as an independent factor for AUD in HIV patients. The average adjusted odds ratio of cigarette smoking in these studies was found to be 3.95% (95% CI: 3.00, 4.89) (I 2 =96.2%, P< 0.01).This result suggested that patient's on ART who were smoking a cigarette were on average 4 times at increased risk of developing AUD than patients who were not smoking a cigarette. Similarly, three of the above-indicated studies (7,27,49) had also reported khat chewing as a risk factor for AUD. The average adjusted odds ratio of khat chewing among these studies was found to be 3.34% (95% CI: 1.71, 4.96) (I 2 =98.2%, P< 0.01). This implied that patients who were chewing khat were on average 3.3 times more likely to have AUD than patients who were not chewing khat.

Discussion
In general, AUD in people living with HIV/AIDS is highly prevalent and linked with non-adherence to antiretroviral therapy, decreased help-seeking, and health care utilization as well as poor HIV treatment outcomes (63). Despite this and the presence of multiple single setting studies in Africa, to date as per the knowledge of investigators, this review and meta-analysis on the prevalence of AUD and its associated factors in patients who are on antiretroviral therapy are the rst of its kind in Africa. This study, therefore, aimed and assessed the average prevalence of AUD and its associated factors in the African population, and the data obtained from this meta-analysis study would be signi cant evidence to future researchers, clinical practitioners, public health experts, and policymakers.
Twenty-two studies that assessed AUD in African HIV/AIDS patients were included in the current metaanalysis. The average prevalence of AUD among included studies was found to be 22.03%. This average prevalence of AUD was higher than the average prevalence of alcohol use disorder in the general population (29). This difference might be caused because people with a chronic medical illness like HIV/AIDS, might engage in alcohol use as a coping strategy for a way for psychological distress and anxiety associated with the perceived severity of the illness and medication side-effects (64).
The average prevalence of AUD among PLWHA in the current study was lower as compared to the result of a systematic review and meta-analysis study by Duko et al.2019 (7); which reported the average worldwide prevalence of AUD among HIV/AIDS patients to be 29.8%. The differences in economic, social, and cultural factors in which alcohol use behaviors are favorably higher in developed countries than African countries might cause variance. The average prevalence of AUD in the present study was comparatively lower than the prevalence of AUD in individuals with psychiatric disorders which ranges from 28%-70%(65). The poor judgment and insight in patients with psychiatric disorders could result in the elevated prevalence of AUD in this group of population.
This average prevalence of AUD in our study was considerably higher in South Africa (28.8%) than the average prevalence estimate in Uganda (16.6%). South Africa is a relatively economically advanced nation than Uganda so individuals would have the ability to afford alcohol.
Variation in culture, the ease of access and accessibility of alcoholic drinks, differences in the number of included studies in the analysis might also add to the difference in the prevalence of AUD among HIV/AIDS patients.
As expected, the pooled estimated prevalence of AUD in studies that studied a relatively higher sample (> 450) was signi cantly lower (16.7%) than the average estimated prevalence for studies which assessed smaller sample (< 450) which was 26.5%. This could be due to the decreases in the probability of a standard error when using a larger sample size and so providing a more precise and reliable result with strong power (66, 67).
Besides, the average estimated prevalence of AUD was higher among studies done in 2011 and after (22.9% to 24.9%) than the average estimate of AUD in studies that were done before 2011(13.5%). The increased availability of alcohol and alcohol advertising programs at current times could bring such variation.
The average estimated prevalence of hazardous alcohol use, harmful alcohol use, and dependent drinking in this study were found to be 10.87%, 8.1%, and 3.12 % respectively. The prevalence of hazardous alcohol use is lower than alcohol use disorder in the present study. This could be due to the small number of studies included in the estimated prevalence of hazardous drinking that affected the precision of the estimate. The average estimated prevalence of AUD was 4% in studies that used a larger sample size (>400) and is lower than the average estimated prevalence in studies that assessed a relatively smaller sample size (<400) (13.5%). Furthermore, the estimated prevalence of dependent drinking was also higher in smaller sample studies (6.7%) than larger sample studies (1.8%).
Association between male sex and AUD was reported in this study. The average estimated odds ratio of being male as a risk factor for AUD was 5.5 in this study. This showed that male patients with HIV were 5, Page 12/31 5 times more vulnerable to develop AUD than female patients. Supportive evidence for this existed in a meta-analysis study (7). Factors related to variation in neurochemistry could be responsible for this. This can be illustrated by a US study that revealed that a higher rate of dopamine release was observed in men than women despite the same level of alcohol intake(68) that can further reinforce the alcohol-seeking behavior and heighten the risk of AUD. Some cultures restrict alcohol consumption in women which could further reduce the risk of AUD in women and other factors in the environment may have an additive role for the difference(69).
The average estimated odds ratio of cigarette smoking and chat chewing in this study was 3.9 and 3.3. Patients who were smoking cigarettes and chewing chat were nearly 4 times and 3.3 times more likely to develop AUD than patients who were not smoking cigarettes and chewing khat. These substances have a similar mechanism of action with alcohol (70) and therefore one can potentiate the rewarding effect of the other (71).

Difference between studies included in the current review and meta-analysis study
This meta-analysis study had a signi cant heterogeneity from the variance between the included studies. Therefore, we further perform a subgroup analysis to explore the source of heterogeneity. The subgroup analysis revealed that the country of origin for the study and the sample size used among included studies and study period was responsible for the variance in the prevalence of alcohol use disorder between included studies. On top of this, we conducted one study leave out sensitivity analysis but the result showed that the overall average estimated prevalence of AUD was not under the in uence of a single particular study. This meta-analysis study on AUD has limitations to be considered in using its result. The primary limitation is that it was under the in uence of signi cant heterogeneity. Furthermore, at some sub-groups, we integrate a few numbers of studies so that the precision of estimate might be affected.

Implications of the ndings to future research, clinical practice, and policy
This meta-analysis study has placed some important implications for future researchers, clinical practitioners, and policymakers. The high average prevalence of AUD among people living with HIV/AIDS in Africa obtained in the current study as compared to the average estimated prevalence of AUD in the general population urges future researchers to have an area of research for why and what determining factors were responsible for this. Secondly, the result of this study informed that clinical practitioners working in the health care facility especially those working on anti-retroviral treatment centers have to be aware of the problem and intervening regularly while managing such patients. Finally, policymakers and program planners need to give attention to this public health problem so that better ways of treatment and prevention approaches for the holistic implementation of alcohol use service of people living with HIV/AIDS with the existing health care services.

Conclusion
This review and meta-analysis study found a high pooled magnitude of AUD in the African population living with HIV; nearly one in ve (22%). This average estimated prevalence of AUD was subjected to substantial heterogeneity. Therefore, we did a sub-group analysis and it was found that country of origin for the study; sample size and year of the study were among the responsible factors for between studies variance. The average prevalence of AUD was considerably higher in South Africa (28.8%) than in Uganda (16.6%). The average estimated prevalence of AUD in studies with a higher sample (> 450) was lower (16.7%) than studies with a smaller sample (< 450) (26.5%). The average prevalence of AUD was also 13.47%, 24.93%, and 22.88% in studies published before 2011, 2011-2015, and after 2015 respectively.
Our narrative synthesis revealed that actors such as being male, Cigarette smoking, chat chewing, family history of alcohol use, missing ART medication, mental distress, low CD4 count, and low income were some of the associated factors with AUD in people with HIV AIDS in Africa. Our quantitative metaanalysis for the associated factors of AUD in HIV/AIDS patients also showed that the AOR for the male sex, cigarette smoking, and khat chewing were 5.5, 3.95, and 3.34 respectively. Therefore, clinical service