WITH MORE THAN 25 MILLION people living with HIV/AIDS in sub-Saharan Africa,1 the HIV epidemic is one of the greatest public health challenges facing the region. In the absence of an effective vaccine, modification of social and culturally rooted behaviors that are associated with HIV transmission remains the most promising strategy to curtail and reverse the spread of the disease. Alcohol use is one such social behavior that has been associated with HIV prevalence and implicated in its transmission.
A recent review of the literature,2 as an example, documented an association between alcohol consumption and presence of sexually transmitted diseases (STDs) that are often precursors to HIV infection. While another review of empirical studies conducted in sub-Saharan Africa3 found alcohol consumption was consistently related to risky sexual behaviors among men, greater consumption by men was associated with increased sexual risk-taking and risks among women were associated with partner drinking.
We, in turn, conducted a systematic review and meta-analysis of studies conducted in Africa that investigated a direct relationship between alcohol use and the risk of HIV infection.4 Pooled results from the 20 studies identified indicated that drinkers have 57% to 70% greater risk of being HIV+ than nondrinkers. Men and women had similar risk profiles whereas high-risk groups such as bar and hotel workers, miners, and soldiers, were at greater risk than were members of the general population. Further, problem drinkers, defined by quantity and frequency indicators or screening instruments, e.g., CAGE or AUDIT, were also at significantly greater risk to be HIV+ when compared with nonproblem drinkers.
The meta-analysis was limited by several factors including the lack of commonality in the way alcohol consumption was measured across studies. No consistent measures of quantity consumed or recency and frequency of drinking were employed. As a consequence findings could only be pooled on a rudimentary basis into broadly defined groups, nondrinker, drinker, and problem drinker.
This study attempts to fill in some of the information gaps we encountered when conducting the review and meta-analysis. Specifically, the investigation has two objectives: (a) to describe the relationship between alcohol consumption and HIV infection in greater depth by using detailed measures of drinking behavior, and (b) to extend the investigation of the association between problem drinking and HIV infection and known HIV risk factors such as prevalence of other STDs and risky sexual behavior.
To accomplish these objectives we examine the drinking behavior of female bar and hotel workers in Moshi, Tanzania. In most African countries, alcohol is sold in small bars and hotels patronized by male clients who offer drinks and may seek sexual encounters with women employed in these settings. Thus, women working in these places are more likely to drink, have multiple sex partners and are at increased risk of acquiring HIV and other STDs.
Study Population and Data Collection
This study is based on analysis of data collected at baseline as part of a prospective study to examine predictors of HIV seroincidence. A more complete description of the study protocol, data gathering instrument and procedures and laboratory methods is described elsewhere5 and will be reviewed briefly below.
Between December 2002 and November 2003, women employed in the bars and hotels located in 7 of 15 local administrative wards of Moshi, Tanzania were recruited to take part in a prospective cohort study designed to determine HIV incidence and to identify factors affecting transmission. These wards were selected because they have the largest number of bars and hotels in the study area. Women aged ≥14 years, working in the registered bars/hotels, and willing to provide written consent and participate in the study procedures were eligible to be enrolled in the study. The study was part of the Kilimanjaro Reproductive Health Program, a collaborative research program of the Kilimanjaro Christian Medical Centre (KCMC), Moshi Municipal Council, and Harvard School of Public Health.
At baseline, women were interviewed to obtain information about sociodemographic characteristics, sexual behaviors, and other potential risk-factors for HIV/STDs. Detailed information was also collected about a variety of alcohol drinking behaviors, including whether the respondent drank currently or in the past, the type of beverage typically consumed, quantity and frequency of drinking, and a self-assessment of whether the woman felt she was a problem drinker. Alcohol consumption was standardized across product forms and expressed in bottles of beer units.
In addition, the CAGE questionnaire6 was used to categorize women about problem drinking. The mnemonic CAGE refers to the four yes or no, items that make up the instrument, including: 1) Have you ever felt you ought to Cut down on your drinking? 2) Have people ever Annoyed you by criticizing your drinking? 3) Have you ever felt bad or Guilty about your drinking? 4) Have you ever had a drink first thing in the morning to steady your nerves or get rid of a hangover (Eyeopener)? Problem drinkers are defined as those who answer yes to 2 or more of the questions.
After the interviews, blood and genital samples were collected for detection of STDs and other genital infections. Samples were transported in cold boxes within 3 to 4 hours of collection to the Department of Clinical Laboratories at KCMC for further processing. Study participants returned to the clinic within 10 to 14 days for results and post-test counseling and women with STD-related symptoms or laboratory confirmed infections received free treatment based on guidelines of the Tanzania government. The Ethics Committee of the Tanzania National Institute for Medical Research and KCMC and the Institutional Review Board of Harvard School of Public Health approved the study protocol.
HIV-1 infection was determined by using 2 enzyme-linked immunosorbent assays. Indeterminate results were resolved by Western blot (Genetic Systems, Bio-Rad Laboratories, Redmond, WA). HSV-2 infection was detected by using the type-specific HSV-2 enzyme-linked immunosorbent assay according to manufacturer’s instructions (Focus Technologies, Cypress, CA). Active/recent syphilis was diagnosed if serum was reactive on the rapid plasma reagin card test (Becton-Dickinson, Cockysville, MD) and the Treponema pallidum hemagglutination assay (TPHA) (Murex Biotech Ltd., England, UK). A wet mount of a vaginal swab was examined microscopically for the presence of clue cells, motile Trichomonas vaginalis and yeast cells. Candidiasis was diagnosed by isolation of the Gram-negative yeastlike cells on Sabouraud’s dextrose agar and confirmed by using the germ tube test (Remel, Lenexa, KS). We attempted to isolate Neisseria gonorrhoeae by inoculation of the endocervical swab on modified Thayer Martin media but none of the study subjects had this infection. Chlamydia trachomatis antigen was detected by using an antigen detection enzyme immunoassay (Murex Biotech Ltd., Dartford, UK). Positive samples were confirmed by a blocking assay from the same manufacturer.
Initially, we examined the direct relationship between current HIV status and drinking behavior. Individual measures of alcohol use including recency and frequency of drinking, quantity typically consumed when drinking, total drinks per week and self-reported problem drinking were related to whether a woman was HIV+. Responses to each CAGE question and overall CAGE scores were also associated with HIV status in this phase of analysis.
Each indicator of drinking behavior, 12 in all, was treated as a separate exposure and related individually to the single outcome, presence of HIV infection. For every comparison the bivariate relationship between the drinking measure and HIV status was expressed using an odds ratio (OR) and 95% confidence interval (CI). Additionally, a logistic model was created for each relationship investigated that included the drinking measure and other sociodemographic and employment characteristics (Table 1) that have been shown to be associated with HIV infection in this sample.7 Adjusted odds ratios (AOR) and 95% CIs from the models were used to assess the strength of the relationship between the drinking measures and HIV infection prevalence while holding constant the influence of the covariates.
HIV risk factors and problem drinking were investigated in the second stage of the analysis. Specifically, we examined the association between the presence of other STDs and engagement in high-risk sexual behaviors and degree of involvement with alcohol. Alcohol involvement was categorized as 1 of 3 states: nondrinker (never drank), nonproblem drinker (1 or fewer “yes” answers to CAGE questions), and problem drinker (2 or more “yes” answers to CAGE questions).
In this case, there was a single exposure, problem drinking, and 13 different outcomes, i.e., STDs and HIV risk factors. As in the first analysis, the univariate relationship between the exposure and each individual risk factor outcome was expressed as an OR and associated 95% CI. A logistic model was then created for each risk factor that expressed it as a function of problem drinking and the covariates. AORs and 95% CIs were used to quantify the relationship between the risk factor and alcohol use holding constant the influence of the covariates.
Out of 1050 women enrolled in the study, 1044 women provided a blood sample for detection of HIV and other STDs. One hundred and ninety-nine women (19.0%) were HIV seropositive (95% CI, 16.6–21.4). Overall, 771 (73.9%) women in the cohort had consumed alcoholic beverages. Among women who used alcohol, 361 answered 2 or more of the CAGE questions positively. Thus, the prevalence of problem drinking at baseline was 34.6% (95% CI, 31.7%–37.5%).
Table 1 summarizes the prevalence of HIV infection (column 2), drinking (column 3), and problem drinking (column 4) for the total sample and within levels of the sociodemographic and employment characteristics that were used as covariates in the logistic model. To facilitate comparisons across variables we have expressed the prevalence of each outcome as an index of the total sample prevalence, i.e., indexed prevalence = (% within the level/% in the total sample) × 100 or in Table 1 (row %/column %) × 100. Thus, a value of 100 indicates that the prevalence of HIV infection, drinking or problem drinking within a particular sociodemographic or employment group is the same as the overall prevalence in the full sample, values below 100 indicate a lower than average prevalence and above 100 higher than average prevalence.
To illustrate, the indexed prevalence of HIV infection among 31 to 35 year olds is 159 indicating that the prevalence of HIV among women in this age group is 1.59 times or 59% greater than in the sample as a whole. Since the overall prevalence is 19.0%, in the 31 to 35 year old group it is 30.2%, i.e., 1.59 × 19.0% = 30.2%. Conversely, for those 20 or younger, the index is 35, so the prevalence is 65% below the sample average or 6.7%, i.e., 0.35 × 19.0% = 6.7%. Indices for the prevalence of drinking and problem drinking in each age group can be interpreted in the same manner.
HIV infection increases with age therefore and is highest among 31 to 40 year olds. It is also more prevalent among those with less education, Muslims, ethnic Pare, women who have been but are not now married or who live with a man but are not married, and those who have given birth. Concerning employment circumstances, women who do not receive a salary and those working in the Kiusa and Korongoni wards have the highest rates of HIV infection.
With respect to drinking behavior, rates of alcohol consumption increase with age and number of children but decrease with greater levels of education. Drinking prevalence is also higher among Catholics, those who are or have been married and those who work in the Kilimanjaro ward and are not salaried. Among drinkers, women who are less well educated, were formerly married or are cohabiting, work in the Kiusa and Korongoni wards and do not have receive a salary are more apt to be problem drinkers.
The association between various measures of alcohol use and HIV status are presented in Table 2. The prevalence of HIV among nondrinkers was 9.5% a figure that is almost exactly that observed in a population based survey of 20- to 44-year-old women in the surrounding area (10.3%; 95% CI, 0.86–12.0).8 By comparison, the prevalence of HIV infection was 22.4% among alcohol drinkers (OR, 2.75; 95% CI, 1.77–4.26). After adjusting for the sociodemographic variables, drinkers have significantly increased risk of HIV infection when compared with nondrinkers (AOR, 2.10; 95% CI, 1.29–3.42).
Among 771 alcohol users, three-quarters drank alcohol in the past month (recent drinkers). The risk of being HIV+ was significantly increased among recent drinkers when compared with nondrinkers (AOR, 2.50; 95% CI, 1.52–4.11) and nonrecent drinkers (AOR, 2.20; 95% CI, 1.33–3.62). However, the risk of being HIV+ among nonrecent drinkers was not significantly different from nondrinkers (AOR, 1.22; 95% CI, 0.65–2.29).
HIV+ risk was positively correlated with the quantity of alcohol typically consumed per occasion (P-value, test for linear trend <0.0001), with the highest risk among women consuming more than 3 drinks per occasion (AOR, 3.49; 95% CI, 1.92–6.35). Similarly, the risk of HIV infection increased with increasing frequency of drinking based on the number of days women drink alcohol per week (P-value, test for linear trend = 0.0001). Daily drinkers were nearly four times more likely to be HIV+ (AOR, 3.76; 95% CI, 1.84–7.65) when compared with nondrinkers.
A composite measure of the total alcohol consumed per week was derived from the number of days a woman reported drinking per week and the amount of alcohol usually consumed when drinking. Based on this measure, the observed the risk of being HIV+ increased with increasing total amount of alcohol consumed per week (P-value, test for linear trend <0.0001). Those who drank the equivalent of 11 or more bottles of beer a week were over 3 times more likely to be HIV+ than nondrinkers (AOR, 3.28; 95% CI, 1.75–6.16).
We next examined the association between HIV+ status and problem drinking indicators (Table 3). Using a subject self-assessment measure, women who thought they had a drinking problem were more likely to be HIV+ after controlling for other factors (AOR, 2.22; 95% CI, 1.14–4.33). A more objective appraisal, albeit still based on self-reports, was provided the four questions that make up the CAGE problem drinking scale. Affirmative response to each of the questions was associated with more than twofold increased risk of HIV when compared with nondrinkers. Among drinkers affirmative response was related to elevated risk in all cases except for the “Cut down” item.
HIV+ risk was also positively correlated with the number of affirmative responses (P-value, test for linear trend <0.0002), with the highest risk among women who answered affirmatively to all four CAGE questions (AOR, 6.80; 95% CI, 2.59–17.88). Similarly, the risk of being HIV+ was significantly increased among women classified to have problem drinking, as defined by the customary cut point of two or more yes answers to the CAGE questions (AOR, 2.43; 95% CI, 1.45–4.06).
Table 4 presents the association between alcohol use and problem drinking and the presence of STDs and other genital infections, and sexual behaviors related to HIV/STDs. Compared with nondrinkers, problem drinkers were more likely to report abnormal genital discharge in the past year (AOR, 1.96; 95% CI, 1.28–2.99), and to have trichomoniasis (AOR, 2.08; 95% CI, 1.04–4.15). The prevalence of HSV-2 among nondrinkers was 37.4% as compared with 59.3% for nonproblem drinkers and 67.6% among problem drinkers. After adjusting for demographic variables, both groups of drinkers have significantly elevated risk of HSV-2 when compared with nondrinkers.
Problem drinkers were more likely to report high-risk sexual behaviors, including multiple sexual partners during the past 5 years or the past year, concurrent partners, exchange of gifts or money for sex and having male partners who had other partners. Problem drinkers and nonproblem drinkers were also more likely to have met their last partners in a bar/hotel and to have initiated sexual activity by 17 years of age.
Unprotected sex is a risk factor for HIV infection and interestingly, nondrinkers were more likely to have not used a condom during their last sex. About this single risk factor then, problem drinkers were at significantly lower risk (AOR, 0.56; 95% CI, 0.37–0.84) than were nondrinkers. Drinkers were at lower risk as well but not significantly so (AOR, 0.76; 95% CI, 0.50–1.15).
Our findings indicate a strong direct relationship between alcohol use and HIV infection among the bar and hotel workers studied. After adjustment for the sociodemographic and employment characteristics, drinkers have more than twofold increased risk of being HIV+ when compared with nondrinkers. Moreover, there was consistent evidence across a series of measures that greater involvement with alcohol was associated with increased risk of HIV infection. Recent drinkers were more likely to be HIV+ than less recent drinkers or nondrinkers while more frequent drinking and increased consumption per occasion or during the week, were associated in a dose response manner with increased presence of HIV infection.
We also observed an increased risk of HIV infection among women who were problem drinkers. Those who thought they had a drinking problem, who answered affirmatively to each of the four CAGE questions, or were classified as problem drinkers, based on the customary cutoff of two or more yes responses to the four CAGE questions were more likely to be HIV+ when compared with nondrinkers. Furthermore, the number of problem drinking symptoms was related in a dose response fashion to HIV infections. These results are consistent with the findings of the previous studies conducted in Tanzania,7–11 and in other African countries.4,12–14
Problem drinking was also associated with other STD infections and high-risk sexual behaviors. Both drinkers and problem drinkers had elevated risk of HSV-2 infection, although problem drinkers also had increased risk of STD-related symptoms and trichomoniasis. The increased risk of STDs associated with alcohol use and problem drinking is consistent with the findings of other studies.2 In addition, because other STDs are known to facilitate HIV transmission15 and were shown to be related to HIV infection in this sample7 our findings provide evidence that alcohol use increases the risk of HIV infection indirectly by increasing the occurrence of other STDs.
Compared with nondrinkers, problem drinkers were more likely to report high-risk sexual behaviors, including earlier age at sexual debut, multiple and concurrent sexual partners, and exchange of gifts or money for sex. Alcohol is known to impair judgment and diminish personal control and perception of risk from multiple sexual encounters.16 It is also conceivable that the expense of acquiring alcohol may contribute to poverty and the need to engage in commercial sex work. These findings indicate that alcohol use and problem drinking could increase HIV/STD vulnerability by influencing sexual behaviors associated with these infections.
The increased condom use among alcohol users appears at first glance to be counterintuitive. However, increased condom use has been observed among subjects reporting multiple sexual partners,17 and other high-risk behaviors such as alcohol consumption.18 This might be indicative of some success in increasing condom use among groups at highest risk of HIV infection, although overall condom use remains low and in most cases condoms are not used consistently with all partners. It is also possible that condom use is a marker of high-risk behaviors, as availability of condoms and willingness to use them may increase the tendency to engage multiple sex partners because of perceived HIV protection. This is likely to be the case in this population, in view of increased reporting of high-risk sexual behaviors and relatively high prevalence of HIV/STDs among alcohol users.
Another important finding is limited amounts of alcohol consumption, as indicated by infrequent drinking or limited number of drinks typically consumed per occasion, was not associated with increased risk of being HIV+ when compared with nondrinkers. Similarly, nonproblem drinkers were less likely to be HIV+, have other STDs and be engaged in high-risk sexual behaviors than were problem drinkers.
This is particularly important given the context within which the women work and may provide a point of departure for effective intervention design or implementation of existing prevention measures such as brief interventions19 or other targeted harm reduction strategies.20 Specifically, it may be a more realistic goal for bar and hotel workers to drink in a controlled fashion than to expect abstinence particularly in settings intended to promote drinking and in which the women perform hospitality duties that may impose significant pressures to drink.
The current study is not without limitations. Our sample was drawn from women employed in the bars and hotels and hence the participants are at greater risk for alcohol use, problem drinking, and high-risk sexual encounters because of their occupation and place of employment. For instance, the prevalence of alcohol use in our sample, 73.9%, was substantially higher than that reported in studies conducted in the general populations from Africa21–24 while the HIV infection rate was nearly double that of women in the surrounding area.
Many of the measures used in the study are self-reported. Given that female drinking is often disapproved and stigmatized in many African communities and is therefore underreported,22,23 we may have underestimated the true prevalence of alcohol use among women surveyed. Similarly, information about sexual behaviors is based on self-reports and might have been affected by the social desirability bias. This is also likely to have resulted in underreporting of these behaviors.
Although we observed strong associations between alcohol use and HIV, with evidence of a dose-response relationship, the results are based on a cross-sectional relationship and causality cannot be established. Results from the ongoing and other cohort studies will be required to confirm these associations. It is significant in this regard that two prospective cohort studies conducted in Africa, one among women receiving family planning services9 and the other population based,24 reported incidence rate ratios between HIV infection and alcohol use that are consistent with the elevated risks observed here.
Lastly, we recruited women working in bar and hotels located in 7 of 15 administrative wards of Moshi. The sampling approach identified all the hotels and bars in the chosen wards and recruited an exceptionally high proportion (91%) of eligible women, and thus, our findings are generalizable to other women working in these settings. However, the women represent a high-risk sample as a function of their hospitality duties and the availability of alcohol and the results may not apply to women in the surrounding area8 and the risk relationships may be lower than those based on population sampling schemes.4
In summary, our results join a growing body of evidence, for the most part observational, that implicates alcohol use as a risk factor for HIV infection. The relationship exists at the intersection of 2 complex social behaviors, alcohol use, and sexual encounters and further investigation and specification with prospective studies, more highly refined and comprehensive measures of alcohol consumption, drinking patterns linked to sexual incidents and potential indirect effects of alcohol use through STDs and high-risk sexual behaviors is needed. Problem drinking, finally, proved to be a useful construct in differentiating direct and indirect risk for HIV infection and as a consequence may represent a starting point for prevention efforts.