Abstract

Objective To investigate the association between alcohol consumption and HIV sero-positivity in a rural Ugandan population.

Methods The adult population residing in a cluster of 15 neighbouring villages has been kept under epidemiological surveillance for HIV infection using annual censuses and sero-surveys since 1989. At the eighth annual survey all respondents were asked about their history of alcohol consumption, the sale of alcohol in their household, and other socio-demographic information. After informed consent, blood was drawn for HIV serology.

Results Of the total adult population 3279 (60%) were interviewed; 48% were males; 905 (27%) had not started sexual activity and were excluded from further analysis. Of the remaining 2374, 8% were HIV infected, 57% had ever drunk alcohol, and 4% lived in households where alcohol was sold. Living in a household where alcohol was sold was associated with a history of having ever drunk alcohol (OR 2.9, 95% CI : 1.7–4.8). HIV prevalence among adults living in households selling alcohol was 15% compared with 8% among those living in households not selling alcohol (OR 2.0, 95% CI : 1.1–3.6). Individuals who had ever drunk alcohol experienced an HIV prevalence twice that of those who had never drunk, 10% versus 5% (OR 2.0, 95% CI : 1.5–2.8). This association remained after adjusting for potential confounders including sale of alcohol in the household and Muslim religion (OR 1.8, 95% CI : 1.2–2.7). Only age, marital status and having ever drunk alcohol independently predicted HIV sero-positivity in a logistic regression model.

Conclusions We have demonstrated an association between a history of alcohol consumption and being HIV sero-positive. This unexplored factor may explain in part the observed lower prevalence of HIV infection among Muslims. Public health campaigns need to stress the relationship between HIV and alcohol.

The predominant mode for HIV transmission in sub-Saharan Africa is heterosexual contact.1 The major risk factors for HIV transmission are now well known and form the basis for preventive public health interventions.2,3 A number of studies from developed countries suggest that alcohol consumption increases risk for HIV infection,4,5 yet only a few studies conducted in sub-Saharan Africa, where nearly half of the global epidemic exists, have reported this relationship.6 Qualitative studies from sub-Saharan Africa demonstrate easy access to and use of alcohol by groups at high risk for HIV infection such as commercial sex workers, truckers7 or migrant workers such as miners.8 Individuals who drink alcohol are less likely to use condoms.9,10 These studies were conducted among urban dwellers and relatively mobile groups, and the results cannot be generalized to rural populations.

Muslims in Uganda, including those living in our area of study, are at a lower risk for HIV infection than non-Muslims; a protective effect that is attributed in part to male circumcision.11 Non-ingestion of alcohol was not controlled for as an alternative explanation or additional protective factor in previous studies.

Alcohol is a central nervous depressant, and in moderate quantities impairs judgement. Alcohol can therefore increase risk for HIV transmission by diminishing personal control, increasing risk-taking, diminishing perception of risk from unprotected sex, or by simply increasing sexual activity. Alternatively, alcohol consumption may be a marker of a personality type given to risk taking, deviant behaviour or unconventionality.12,13

We have studied a rural population to determine the relationship between alcohol consumption and HIV sero-positivity.

Methods

The study population resided in a cluster of 15 neighbouring villages in Masaka district, Southwest Uganda. This cohort has been kept under epidemiological surveillance for HIV infection through annual censuses and sero-surveys since 1989.14 At the eighth annual survey all respondents aged 13 years and above were asked about the sale of alcohol in their household, whether they had ever drunk alcohol, number of days bars were visited and alcohol consumed in the past week. No attempt was made in the questionnaire to distinguish between types of alcohol consumed (e.g. home brewed beer, bottled beer, spirits) or long-term frequency of consumption. In addition, data were collected on religion, ever use of a condom, years of schooling and number of sexual partners in the past 12 months. We used standardized structured questionnaires administered by trained interviewers to collect the data.

After informed consent, blood was drawn for HIV serology. Two independent enzyme immunosorbent assays (EIA) were used to determine HIV status (Wellcozyme HIV-1 recombinant VK 56/57, Murex Biotech Ltd, Dartford, Kent, UK; and Recombigen HIV-1/2, Cambridge Biotech, Galway Ireland Ltd). When EIA were inconclusive, we used Western blot (Cambridge Biotech Corporation, Rockville, USA) to determine final HIV status.15 Test results were issued on request after appropriate counselling.16 The Uganda National Council for Science and Technology gave ethical approval for the study.

Using Fox-pro,17 data were entered in duplicate and checked for consistency before analysis. Univariate analysis was done using an epidemiological package, EPIINFO 6.0 (CDC/WHO). The strength of association between HIV sero-status and alcohol was calculated as odds ratios (OR) with 95% confidence intervals (CI). Stratified analysis (Mantel-Haenszel) was performed to assess independence of sale of alcohol in the household from reported alcohol consumption. Multivariate analysis was done using the STATA 5 statistical package.18 A baseline socio-demographic model was constructed using sex, age group (10-year bands), marital status and religion (Muslim/Christian). We added number of sexual partners in the past 12 months and ever use of a condom to this model although not statistically significant at univariate analysis because they were considered a priori to be confounders. The model was then used to examine the independent association between HIV and alcohol measures, i.e. sale of alcohol in the household, ever drank alcohol, number of days bars visited or alcohol drunk in the past week which were regarded as both exposures and confounders for each other. Because alcohol variables were collinear, they were not adjusted for each other. Statistical significance was estimated using the log likelihood ratio test, and P values were two-sided.

Results

A total of 5453 adults were censused, of whom 60% were interviewed and bled. Forty eight per cent of the respondents were males. Of those not interviewed, 10% were bled but not interviewed, 17% refused, 13% were persistently absent from their homes and 0.1% died before being interviewed. A total of 905 (27%) respondents had not begun sexual activity and, being at minimal risk of acquiring HIV infection,19 were excluded from further analysis. The crude HIV prevalence for the remaining 2374 sexually active adults was 8%. Only 4% of the respondents lived in households where alcohol was sold. Of these, 79% had ever drunk alcohol compared with 57% among those residing in households where alcohol was not sold. HIV prevalence among adults living in households where alcohol was sold was 15% compared with 8% among those living in households where alcohol was not sold (OR 2.0, 95% CI : 1.1–3.6).

Table 1 shows the association of reported alcohol consumption with selected variables. Males were more likely than women to have ever drunk alcohol (OR 1.7, 95% CI : 1.4–2.0), an observation consistent for all variables with the exception of males still in school and those aged 13–19 years. A history of having ever drunk alcohol increased from 31% in the youngest age group to 68% for those 40 years or older (chi-square [χ2] for trend = 126.8, P < 0.001). With respect to schooling, a history of having ever drunk alcohol was lowest among those still in school (25%), and highest among those who had never attended school (65%). This inverse relationship between alcohol and years of schooling was statistically significant, (χ2 for trend = 68.5, P < 0.001). Compared with those who had never married, a history of alcohol consumption was more often reported by those who had ever married regardless of their current marital status, i.e. widowed or separated (χ2 for trend = 62.7, P < 0.001). Non-Muslims were 14-fold more likely to report having ever drunk alcohol than Muslims (74% versus 5%, χ2 for association = 900, P < 0.001). However, a history of having ever drunk alcohol did not vary with reported condom use (χ2 for association = 0.17, P = 0.7) or by sexual partner history (χ2 for trend = 3.6, P = 0.17, data not shown).

A history of alcohol consumption was higher among HIV positive than HIV negative individuals (72% versus 56%), and adults who reported having ever drunk alcohol experienced twice the prevalence of HIV infection than those who had never drunk, 10% versus 5% (χ2 for association = 19.7, P < 0.001).

Table 2 shows univariate and multivariate analysis for association of HIV infection with socio-demographic, sexual and alcohol variables. A statistically significant association was demonstrated between having ever drunk alcohol and HIV sero-positivity (OR 2.0, 95% CI : 1.5–2.8). This association remained in the multivariate model after adjusting for potential confounders including ever use of a condom and number of sexual partners in the past 12 months (OR 1.8, 95% CI : 1.2– 2.7). Adults who lived in households where alcohol was sold were more likely to be HIV positive than those living in households where alcohol was not sold in univariate analysis (OR 2.0, 95% CI : 1.1–3.6). To test whether sale of alcohol in the household had any independent effect on HIV sero-positivity from that mediated through alcohol consumption, a Mantel-Haenszel stratified analysis was performed. Sale of alcohol was not significant after stratifying by having ever drunk alcohol (OR 1.8, P = 0.1). However, the association between having ever drunk alcohol and HIV remained after stratifying by sale of alcohol in the household (OR 2.0, P < 0.001). No gender difference in risk for HIV infection was observed (Table 2). A highly significant relationship between HIV sero-positivity and age group was demonstrated, peaking in the 30–39-year age group. While risk of being HIV sero-positive increased modestly with years spent at school in univariate analysis, this effect was largely explained by collinearity between age and schooling. Compared with those who had never married, the odds for HIV sero-positivity were significantly higher among the currently married, widowed and especially in those who were separated. Muslims were half as likely to be HIV sero-positive as non-Muslims (OR 0.5, 95% CI : 0.3–0.7) but the effect of Muslim religion disappeared when alcohol was added to the multivariate model (OR 0.7, 95% CI : 0.4–1.2).

Frequency of recent alcohol consumption as measured by number of days alcohol was consumed or bars were visited in the past week did not show a significant association with HIV sero-positivity, P = 0.8 and P = 0.2, respectively.

Discussion

We have demonstrated a statistically significant association between reported alcohol consumption and HIV sero-positivity in a rural Ugandan population. This association is independent of gender, age, marital status, ever use of a condom, reported number of past sexual partners, religion and sale of alcohol in the household. Individuals who resided in households where alcohol was sold were at a higher risk for HIV infection, and were more likely to have ever drunk alcohol. The effect of sale of alcohol in the household on HIV sero-positivity appears to be mediated through a greater propensity to drink alcohol in such homes.

Muslims were half as likely as non-Muslims to be HIV sero-positive in univariate analysis as reported in other studies.2,11 Male circumcision, almost universally practised by Muslims, is the generally accepted explanation for the lower HIV prevalence observed among Muslims. ‘Closed sexual networks’ where Muslim men may legally have up to four wives, thereby reducing the pressure for casual sexual partners, is another possible explanation. The small proportion of Muslims who gave a history of having ever drunk alcohol suffered an HIV prevalence of about 10%, similar to that of non-Muslims who had ever drunk alcohol. The disappearance of the protective effect of Muslim religion when a history of alcohol consumption was taken into account suggests that abstinence from alcohol may partly explain the protective effect attributable to Muslim religion. We would urge other workers researching the association between Muslim religion, male circumcision and HIV to include alcohol consumption in their studies to see if our findings are replicated elsewhere.20

Because of the effect of alcohol on sexual behaviour in reducing inhibitions, a relationship between alcohol and HIV has long been suspected. This is the first report from sub-Saharan Africa giving population-based estimates of the association after adjusting for confounders. Previous published studies that examined this relationship were either qualitative or conducted among relatively urban and mobile populations and the results therefore cannot be generalized.7,21

Consumption of alcohol may enhance risk of HIV infection by increasing sexual activity22 or number of sexual partners, or by modifying the willingness to have sex without using condoms.23,24 It may also act by inhibiting libido, in turn causing marital disruption. Alternatively inebriated women may be at increased risk of being sexually assaulted and raped. While alcohol induced failure to use condoms may be pertinent in developed countries, condom acceptance and use is still fairly low in our population even without alcohol.25,26

Our study suffers some limitations. Firstly, only 60% of the censused study population was interviewed for this study. Many of those who did not answer the questionnaire either refused or were persistently absent and this may have led to some bias in our findings. Our results should be considered with this in mind. Secondly, we relied on reported history of alcohol intake and we were not able to demonstrate any effects on HIV infection of type, quantity and long-term frequency of alcohol consumption. Further, our measure of alcohol intake ignored the fact that it often fluctuates, occurs over a long time and could have begun before or after individuals became infected with HIV. Since only a few individuals were aware of their HIV status, it is unlikely that knowledge of HIV status could have influenced reported or actual consumption of alcohol. Any bias introduced by imprecisely measuring alcohol consumption would result in underestimation of the association, as individuals who had an occasional drink many years ago would be included. Our failure to demonstrate a significant dose-response relationship between alcohol intake and HIV infection may be due to the short period of alcohol intake we investigated. The association demonstrated between alcohol and HIV suggests that messages for behaviour change should be broadened to include reduction of alcohol intake.

In conclusion, a history of alcohol consumption is independently and significantly associated with HIV sero-positivity. Other covariates such as living in a household where alcohol is sold, frequency of visiting bars and number of days drinking alcohol in the past week were highly correlated with each other, and with HIV infection. Alcohol most likely operates by promoting risky sex. Abstinence from alcohol may be an additional or alternative explanation for the observed lower HIV prevalence among Muslims. Public health campaigns should stress the relationship between alcohol and HIV. People who drink or sell alcohol need to be advised about the risks of HIV infection.

Table 1

Distribution of reported alcohol consumption and selected variables by sex

 Percentage ever drunk alcohol (No.) 
Variable Males Females ORa χ2 P-value 
a Odds ratios are for the risk of alcohol consumption by the variable for sexes combined (except for the gender variable). 
b χ2 and P-value for trend, otherwise for association. 
Gender 64 (1083) 52 (1286) 1.7 38.2 <0.001 
Age group (years) 
13–19 25 (126) 36 (158) 1.0   
20–29 60 (328) 49 (399) 2.6   
30–39 70 (205) 51 (281) 3.2   
40+ 77 (424) 60 (448) 4.6 126.8 <0.001b 
Level of education 
No school 81 (128) 59 (288) 1.0   
≤7 years of school 67 (721) 52 (804) 0.8   
>7 years of school 60 (156) 45 (150) 0.6   
In school 21 (78) 33 (43) 0.2 68.5 <0.001b 
Marital status 
Never married 46 (323) 37 (167) 1.0   
Currently married 69 (605) 53 (746) 2.0   
Separated 86 (21) 61 (134) 2.5   
Widowed 86 (121) 56 (226) 2.7 62.7 <0.001b 
Religion 
Christian 79 (868) 69 (945) 1.0   
Muslim 7 (215) 4 (341) 0.2 900.0 <0.001 
Sale of alcohol in a household 
Do not sell 64 (1041) 51 (1235) 1.0   
Sell 77 (39) 80 (51) 2.9 18.9 <0.001 
HIV sero status 
HIV– 63 (988) 50 (1157) 1.0   
HIV+ 81 (84) 66 (110) 2.0 19.7 <0.001 
 Percentage ever drunk alcohol (No.) 
Variable Males Females ORa χ2 P-value 
a Odds ratios are for the risk of alcohol consumption by the variable for sexes combined (except for the gender variable). 
b χ2 and P-value for trend, otherwise for association. 
Gender 64 (1083) 52 (1286) 1.7 38.2 <0.001 
Age group (years) 
13–19 25 (126) 36 (158) 1.0   
20–29 60 (328) 49 (399) 2.6   
30–39 70 (205) 51 (281) 3.2   
40+ 77 (424) 60 (448) 4.6 126.8 <0.001b 
Level of education 
No school 81 (128) 59 (288) 1.0   
≤7 years of school 67 (721) 52 (804) 0.8   
>7 years of school 60 (156) 45 (150) 0.6   
In school 21 (78) 33 (43) 0.2 68.5 <0.001b 
Marital status 
Never married 46 (323) 37 (167) 1.0   
Currently married 69 (605) 53 (746) 2.0   
Separated 86 (21) 61 (134) 2.5   
Widowed 86 (121) 56 (226) 2.7 62.7 <0.001b 
Religion 
Christian 79 (868) 69 (945) 1.0   
Muslim 7 (215) 4 (341) 0.2 900.0 <0.001 
Sale of alcohol in a household 
Do not sell 64 (1041) 51 (1235) 1.0   
Sell 77 (39) 80 (51) 2.9 18.9 <0.001 
HIV sero status 
HIV– 63 (988) 50 (1157) 1.0   
HIV+ 81 (84) 66 (110) 2.0 19.7 <0.001 
Table 2

Association between HIV and socio-demographic, sexual and alcohol variables

 OR 
Variable Univariate analysis Multivariate analysis 
NB χ2 and P-value for association given. 
a Main model includes socio-demographic, sexual and alcohol variables; i.e. sex, age group, current marital status, religion, ever use of a condom, number of sexual partners over the past 12 months, and ever drunk alcohol. 
b Model includes: sex, age, marital status, religion, ever use of a condom, number of sexual partners in the past 12 months and either number of days alcohol consumed in the past week or bars visited in the past week. 
Alcohol ingestion 
Never drunk 1.0 1.0 
Ever drunk 2.0 (1.5–2.8) 1.8 (1.2–2.7) 
 χ2 = 19.7 χ2 = 7.4 
 P < 0.001 P = 0.007a 
Alcohol sale 
Do not sell 1.0 1.0 
Sell 2.0 (1.1–3.6) 1.9 (1.0–3.6) 
 χ2 = 4.2 χ2 = 3.39 
 P = 0.04 P = 0.07a 
Gender 
Female 1.0 1.0 
Male 0.9 (0.6–1.2) 1.1 (0.8–1.5) 
 χ2 = 0.6 χ2 = 0.18 
 P = 0.4 P = 0.7a 
Age group (years) 
13–19 1.0 1.0 
20–29 5.1 (2.3–11.2) 3.5 (1.5–8.4) 
30–39 6.1 (2.7–13.5) 3.4 (1.4–8.4) 
40+ 2.0 (0.9–4.5) 0.8 (0.3–2.0) 
 χ2 = 54.9 χ2 = 62.3 
 P < 0.001 P < 0.001a 
Marital status 
Never married 1.0 1.0 
Currently married 1.8 (1.1–2.8) 2.1 (1.2–3.5) 
Separated/divorced 4.5 (2.5–8.2) 10 (4.7–21.4) 
Widowed 2.8 (1.6–4.7) 4 (2.1–7.4) 
 χ2 = 29.3 χ2 = 39.8 
 P < 0.001 P < 0.001a 
Religion 
Christian 1.0 1.0 
Muslim 0.47 (0.3–0.7) 0.7 (0.4–1.2) 
 χ2 = 14.0 χ2 = 1.8 
 P < 0.001 P = 0.18a 
Condom use 
Never used 1.0 1.0 
Ever used 1.5 (1.1–2.1) 1.2 (0.8–1.8) 
 χ2 = 5.4 χ2 = 0.92 
 P = 0.02 P = 0.33a 
Sexual partners past 12 months 
1.0 1.0 
0.9 (0.7–1.4) 0.9 (0.6–1.5) 
2+ 1.3 (0.8–2.1) 1.3 (0.7–2.6) 
 χ2 = 2.0 χ2 = 1.64 
 P = 0.4 P = 0.4 
No. of days alcohol was drunk in the last week 
1.0 1.0 
1–3 1.1 (0.8–1.6) 1.0 (0.6–1.4) 
4–6 1.3 (0.6–2.8) 1.4 (0.6–3.1) 
1.5 (0.8–2.8) 1.3 (0.7–2.6) 
 χ2 = 2.2 χ2 = 1.2 
 P = 0.59 P = 0.80b 
No. of days bars were visited in the last week 
1.0 1.0 
1–3 1.0 (0.6–1.5) 0.8 (0.5–1.4) 
4–6 1.3 (0.6–2.7) 1.4 (0.6–3.1) 
2.0 (1.2–3.2) 1.6 (0.9–2.8) 
 χ2 = 7.2 χ2 = 4.4 
 P = 0.07 P = 0.20b 
 OR 
Variable Univariate analysis Multivariate analysis 
NB χ2 and P-value for association given. 
a Main model includes socio-demographic, sexual and alcohol variables; i.e. sex, age group, current marital status, religion, ever use of a condom, number of sexual partners over the past 12 months, and ever drunk alcohol. 
b Model includes: sex, age, marital status, religion, ever use of a condom, number of sexual partners in the past 12 months and either number of days alcohol consumed in the past week or bars visited in the past week. 
Alcohol ingestion 
Never drunk 1.0 1.0 
Ever drunk 2.0 (1.5–2.8) 1.8 (1.2–2.7) 
 χ2 = 19.7 χ2 = 7.4 
 P < 0.001 P = 0.007a 
Alcohol sale 
Do not sell 1.0 1.0 
Sell 2.0 (1.1–3.6) 1.9 (1.0–3.6) 
 χ2 = 4.2 χ2 = 3.39 
 P = 0.04 P = 0.07a 
Gender 
Female 1.0 1.0 
Male 0.9 (0.6–1.2) 1.1 (0.8–1.5) 
 χ2 = 0.6 χ2 = 0.18 
 P = 0.4 P = 0.7a 
Age group (years) 
13–19 1.0 1.0 
20–29 5.1 (2.3–11.2) 3.5 (1.5–8.4) 
30–39 6.1 (2.7–13.5) 3.4 (1.4–8.4) 
40+ 2.0 (0.9–4.5) 0.8 (0.3–2.0) 
 χ2 = 54.9 χ2 = 62.3 
 P < 0.001 P < 0.001a 
Marital status 
Never married 1.0 1.0 
Currently married 1.8 (1.1–2.8) 2.1 (1.2–3.5) 
Separated/divorced 4.5 (2.5–8.2) 10 (4.7–21.4) 
Widowed 2.8 (1.6–4.7) 4 (2.1–7.4) 
 χ2 = 29.3 χ2 = 39.8 
 P < 0.001 P < 0.001a 
Religion 
Christian 1.0 1.0 
Muslim 0.47 (0.3–0.7) 0.7 (0.4–1.2) 
 χ2 = 14.0 χ2 = 1.8 
 P < 0.001 P = 0.18a 
Condom use 
Never used 1.0 1.0 
Ever used 1.5 (1.1–2.1) 1.2 (0.8–1.8) 
 χ2 = 5.4 χ2 = 0.92 
 P = 0.02 P = 0.33a 
Sexual partners past 12 months 
1.0 1.0 
0.9 (0.7–1.4) 0.9 (0.6–1.5) 
2+ 1.3 (0.8–2.1) 1.3 (0.7–2.6) 
 χ2 = 2.0 χ2 = 1.64 
 P = 0.4 P = 0.4 
No. of days alcohol was drunk in the last week 
1.0 1.0 
1–3 1.1 (0.8–1.6) 1.0 (0.6–1.4) 
4–6 1.3 (0.6–2.8) 1.4 (0.6–3.1) 
1.5 (0.8–2.8) 1.3 (0.7–2.6) 
 χ2 = 2.2 χ2 = 1.2 
 P = 0.59 P = 0.80b 
No. of days bars were visited in the last week 
1.0 1.0 
1–3 1.0 (0.6–1.5) 0.8 (0.5–1.4) 
4–6 1.3 (0.6–2.7) 1.4 (0.6–3.1) 
2.0 (1.2–3.2) 1.6 (0.9–2.8) 
 χ2 = 7.2 χ2 = 4.4 
 P = 0.07 P = 0.20b 

We thank Professor John Ziegler for his comments on early drafts, the population for their participation and the Ministry of Health of Uganda for allowing this work to be published.

References

1
Piot P, Kapita BM, Were JBO, Laga M, Colebunders RL. AIDS in Africa: the first decade and challenges for the late 1990s.
AIDS
 
1991
;
5:
S1
–S5.
2
Nunn AJ, Kengeya-Kayondo JF, Malamba S, Seeley JA, Mulder DW. Risk factors for HIV-1 infection in adults in a rural Ugandan community: a population study.
AIDS
 
1994
;
8:
81
–86.
3
Serwadda D, Wawer MJ, Musgrave SD et al. HIV risk factors in three geographic strata of rural Rakai District, Uganda.
AIDS
 
1992
;
6:
983
–89.
4
Bastani R, Erickson PA, Marcus AC et al. AIDS-related attitudes and risk behaviours: a survey of a random sample of California heterosexuals.
Prev Med
 
1996
;
25:
105
–17.
5
Kusseling FS, Shapiro MF, Greenberg JM, Wenger NS. Understanding why heterosexual adults do not practice safer sex: a comparison of two samples.
AIDS Educ Prev
 
1996
;
8:
247
–57.
6
Mnyika KS, Klepp KI, Kvale G, Ole-King’ori N. Risk factors for HIV-1 infection among women in the Arusha region of Tanzania.
J Acquir Immune Defic Syndr Hum Retrovirol
 
1996
;
11:
484
–91.
7
Mbuguzi G, Hearst B, Linan C et al. Risk factors associated with HIV infection among long distance truck drivers in Kenya.
Int Conf AIDS
 
1992
;
8:
D448
(abstract no. PoD 5367).
8
Boets L, Lurie M, Mini C, Field ML. Health seeking behaviours for sexually transmitted disease and the social context of commercial sex in a gold mining community; a case study of Welkom, South Africa.
Int Conf AIDS
 
1996
;
11:
368
(abstract no. Th.C.4745).
9
Peters M, Alary M, Laga M, Piot P. Determinants of condom use in European female sex workers. The European Working Group on HIV infection in female prostitutes.
Int Conf AIDS
 
1992
;
8:
D497
(abstract no. PoD 5649).
10
Wilson D, Lavelle B, Mwoboto B, Armstrong M. Use of a retrospective timeline calendar to examine alcohol use, sex behaviour and condom use among Zimbabwean men.
Int Conf AIDS
 
1992
;
8:
C332
(abstract no. PoC 4522).
11
Malamba SS, Wagner HU, Maude G et al. Risk factors for HIV infection in adults in a rural Ugandan community: a case-control study.
AIDS
 
1994
;
8:
253
–57.
12
Adalaf EM, Smart RG. Risk taking and drug use behaviour: an examination.
Drug Alcohol Depend
 
1983
;
11:
287
.
13
Zukerman M. Sensation-seeking: Beyond the Optimal Level of Arousal. Hillsdale, NJ: Erlbaum, 1979.
14
Mulder DW, Nunn A, Kamali A, Nakiyingi J, Wagner HU, Kengeya-Kayondo JF. Two-year HIV-1 associated mortality in a rural Ugandan community.
Lancet
 
1994
;
343:
1021
–23.
15
Nunn A, Biryahwaho B, Downing RG, Van der Groen G, Ojwiya A, Mulder DW. Algorithms for detecting antibodies to HIV-1: results from a rural Ugandan cohort.
AIDS
 
1993
;
7:
1057
–61.
16
Seeley JA, Wagner HU, Mulemwa J, Kengeya-Kayondo JF, Mulder DW. The development of a community-based HIV/AIDS counselling service in a rural area in Uganda.
AIDS Care
 
1991
;
3:
207
–17.
17
Fox-Pro for Windows. Version 2.6 ©1989–1994 Microsoft Corporation.
18
Stata Corp. 1995. Stata Statistical software: Release 5.00. College Station, TX: Stata Corporation.
19
Mulder DW, Nunn A, Kamali A, Kengeya-Kayondo JF. Post-natal incidence of HIV-1 infection among children in a rural Ugandan population: no evidence for transmission other than mother-to-child transmission.
Trop Med Int Health
 
1996
;
1:
81
–85.
20
Allen P, Linden C, Semfilina A et al. Human Immune Deficiency Virus infection in urban Rwanda. Demographic and behavioural correlates in a representative sample of childbearing women.
JAMA
 
1991
;
266:
1657
–63.
21
O’Farrel N, Egger M. Circumcision and the prevention of HIV infection: a ‘meta-analysis’ revisited.
Int J STD AIDS
 
2000
;
11:
137
–42.
22
Leigh BC, Temple MT, Trocki FK. The relationship of alcohol use to sexual activity in a US national sample.
Soc Sci Med
 
1994
;
39:
1527
–35.
23
Gossop M, Pwis B, Griffiths P, Strang J. Female prostitutes in south London: use of heroin, cocaine and alcohol and their relationship to health risk behaviours.
AIDS Care
 
1995
;
7:
253
–260.
24
Seeley JA, Malamba SS, Mulder DW et al. Socio-economic status, gender and risk of HIV-1 infection in a rural community in south west Uganda.
Med Anthropol Q
 
1994
;
8:
78
–79.
25
Mulder D, Nunn AJ, Kamali A, Kengeya-Kayondo J. Decreasing HIV-1 prevalence in young adults in a rural Ugandan cohort.
BMJ
 
1995
;
311:
833
–36.
26
Pickering H, Okongo M, Nalusiba B, Bwanika K, Whitworth J. Sexual networks in Uganda: casual and commercial sex in a trading town.
AIDS Care
 
1997
;
9:
199
–207.