High potential of escalating HIV transmission in a low prevalence setting in rural Tanzania

Few countries have presented evidence of declining trends in HIV incidence [1–8]. However, there are signs of escalating rural epidemics even within countries reporting a decline of HIV prevalence [4, 6, 8]. Unfortunately relatively little information is available in rural settings, especially in remote areas with difficult terrain where the dynamics and epidemiology of HIV infection is likely to be different.

To estimate HIV prevalence in remote rural Manyara and Singida regions in Tanzania, we set up a local HIV surveillance system in selected antenatal clinics (ANC). The antenatal surveys conducted at the same clinics at different points in time, identified an imminent but still relatively low HIV epidemic within the catchment area [9, 10]. As a follow up, we conducted a population-based-survey to validate the representativeness of the ANC-based HIV surveillance, and to examine sexual risk behaviours among men and women in the general population in the same catchment area.

The total sample included 1,956 adults aged between 15–49 years. The refusal rate for HIV-1 testing was 13% (253/1,956) for those who completed the questionnaire, among whom 46% (117) were men. Seven participants, who refused the questionnaire survey and accepted HIV testing, had no antibodies for HIV, and were not included in the analysis. Five (0.3%) of the 1,703 tested saliva samples gave indeterminate results.

Figure 1 illustrates the comparisons of HIV prevalence based on results from the ANC-based surveillance and from this population-based survey. Apparently, pregnant and non-pregnant women aged 20–49 years show a similar pattern of infection. However, for the younger age group, women in the general population show a relatively lower HIV prevalence compared to the pregnant women, 0.7% and 3%, respectively. Peak HIV prevalence in women was between the ages of 30 and 34 years, while for men the highest prevalence was between 35 and 39 years. Tables 3 shows the results of the analysis using logistic regression on demographic and selected behavioural indicators adjusted for age. An increase in HIV infection is associated with being a resident in rural areas (OR 4.9; 95% CI 1.2–1.9) and a shorter duration of stay at the present residential address (OR 5.9; 95% CI 1.2–33.3) for men. A four-fold increase of HIV infection was weakly associated with residing in the fishing village of Basotu for both sexes (4.1; 1.0–16.4). In women, the use of contraceptives other than condoms increased the likelihood of HIV infection seven times (OR, 7.7; 95%CI: 2.0–30.2). Men who have received any kind of incentive in exchange for sex, were likely to be infected (OR 6.9; 95% CI 1.2–41.5) compared to those offering the incentive (OR 0.6; 95% CI 0.1–4.0).

Men (n = 741)				Women (n = 904)
Characteristic	N (%)	OR (95%CI)	P-value	N (%)	OR (95%CI)	P-value
Residential area
Remote	345 (46.6)	1	-	369 (39.5)	1	-
Rural	396 (53.4)	4.9 (1.2–19.9)	0.03	565 (60.5)	4.6 (0.9–22.9)	0.06
Village
Haydom	153 (20.7)	1	-	220 (24.3)	1	-
Muslur	118 (15.9)	0.6 (0.1 – 6.0)	0.61	133 (14.7)	0.6 (0.1–5.0)	0.61
Basotu	117 (15.8)	4.1 (1.0 – 16.4)	0.05	116 (12.8)	4.1 (1.1–14.7)	0.03
Dangayda	79 (10.7)	0.9 (0.1 – 6.1)	0.93	81 (9.0)	-	-
Mwanga	148 (20.0)	0.5 (0.1–5.6)	0.58	142 (15.7)	0.5 (0.1–5.4)	0.59
Nduguti	126 (17.0)	2.5 (0.4–15.3)	0.30	212 (23.5)	1.5 (0.4–6.0)	0.58
Age group (years)
15–19	90 (12.2)	-	-	119 (13.2)	-	-
20–24	155 (20.9)			208 (23.0)
25–29	149 (20.1)			210 (23.2)
30–34	127 (17.1)			143 (15.8)
35–39	93 (12.6)			105 (11.6)
40–49	127 (17.1)			119 (13.1)
Marital status
Single	241 (32.5)	1	-	183 (20.2)	1	-
Married/engaged/cohabit	479 (64.6)	0.62 (0.1–3.8)	0.61	676 (74.8)	0.2 (0.0–1.3)	0.10
Divorced/separated/widow	21 (2.8)	-	-	45 (5.0)	1.9 (0.2–16.8)	0.55
Duration of stay at current residence
> 1 Year	655 (88.4)	1	-	777 (86.0)	1	-
≤1 Year	86 (11.6)	5.9 (1.2–33.3)	0.03	127 (14.0)	1.8 (0.4–7.7)	0.46
Education level
None	58 (7.8)	1	-	107 (11.8)	1	-
< 7 years	110 (14.8)	0.6 (0.1–7.5)	0.50	158 (16.9)	3.8 (0.4–4)	0.24
≥7 years	573 (77.3)	0.7 (0.1–5.4)	0.60	639 (70.7)	2.3 (0.3–16.5)	0.40
*Income
None	38 (5.1)	1	-	20 (2.21)	-	-
Employed	94 (12.7)	1.6 (0.1–12.6)	0.63	123 (13.61)	0.3 (0.0–3.7)	0.36
Farmer/pastoralist	609 (86.8)	0.5 (0.1–2.6)	0.38	761 (84.18)	0.8 (0.2–3.5)	0.78
Age at first sex
≥18 years	457 (61.7)	1	-	590 (65.27)	1	-
< 18 years	284 (38.3)	2.0 (0.7–6.0)	0.21	314 (34.73)	2.2 (0.8–6.0)	0.12
Ever condom
No	541 (73.0)	1	-	737 (81.5)	1	-
Yes	200 (27.0)	0.5 (0.09–2.5)	0.37	167 (18.5)	1.2 (0.3–5.0)	0.83
Multiple sex partner(s)
No	393 (53.0)	1	-	605 (66.9)	1	-
Yes	348 (47.0)	5.6 (1.5–20.8)	0.01	299 (33.1)	4.6 (1.3–16.9)	0.04
History of STIs
No	628 (84.8)	1	-	674 (74.8)	1	-
Yes	113 (15.3)	4.6 (1.4–15.6)	0.01	230 (25.4)	1.6 (0.7–3.9)	0.29
Ever VCT
No	512 (69.1)	1	1	567 (62.7)	1	-
Yes	229 (30.9)	1.3 (0.5–3.1)	0.57	337 (37.3)	0.9 (0.4–2.1)	0.77
Uses of contraceptives other than condoms
No	-	-	-	576 (63.7)	1	-
Yes				386 (36.3)	7.7 (2.0–30.2)	0.008
"Incentive sex "
Offer incentive for sex
No	684 (92.4)	1	-	899 (99.5)	-
Yes	56 (7.6)	0.6 (0.1–4.0)	0.55	5 (0.6)
Receive incentive for sex
No	731(98.9)	1	-	859 (95.0)	1	-
Yes	8 (1.1)	6.9 (1.2–41.5)	0.03	45 (5.0)	2.2 (0.5–10.8)	0.33

Adjusted for age.

* None of the women without income had HIV infection; the reference category was thus dropped.

Among 1,698 participants, 97% (1645) had been sexually active within the past year, and of these, 55% (904) were women. The mean age at first sex was 17 years ± 3 SD. Sexually inactive individuals belonged to the age group 15–19 years. The majority of currently sexually active individuals, 78% (n = 1278), have never used a condom. Among the youngest sexually active age group, 88% had never used condoms. The most important reasons for not using a condom were values based on traditions and religion (91–92%) for both men and women, though a smaller proportion (31% men and women) reported access difficulties associated with cost and availability.

An HIV prevalence of 1.8% (95% CI; 1.2–2.6) in the general population aged 15–49 years is relatively low compared with other rural areas in the country [18–20] and elsewhere in Eastern Africa [21]. However, HIV entered these remote communities rather recently, as indicated by the very few cases found as late as the mid-1990s, and moderately increased a decade later [9–11, 13]. Furthermore, the revealed patterns of sexual behaviour, particularly the combination of a high proportion with multiple sexual partners, history of sexually transmitted infections and insignificant condom use, are indicative of a great potential of this epidemic to increase in the near future. The higher prevalence of HIV observed in the fishing community underscores the need for focused preventive initiatives that will consider the local social-economic factors that potentially increase HIV transmission.

Population-based surveys are always prone to different types of bias. Firstly, non-response bias, particularly that due to refusal, might be of particular concern in HIV surveys. In this survey, the 13% who refused to participate did not differ significantly from responders by socio-demographic characteristics. Although one cannot exclude the possibility that non-responders differed in prevalence from responders, the low refusal rate and the similarities in socio-demographic characteristics indicate non-response bias may not have significantly biased the results. The close match between the local ANC-based surveillance and the population-based survey from the same areas is suggestive of the representativeness of pregnant women in estimating the local general adult HIV prevalence albeit the inherent biases [2, 22, 23]. Secondly, at the time of the survey the two regions were experiencing severe drought and food shortage, as well as a lack of natural water sources for cattle. We may have missed some community members (empty or abandoned houses), especially in Dangayda village. This is due to the temporary migration that was necessary in order to search for better pasture for cattle. Though it is likely that non-responders have higher risks of HIV infection compared to responders, we found no difference in socio-demographic factors between them. Thirdly, we lacked a reliable listing of household members and estimated the number of eligible participants within households using household clusters from a list of balozi leaders. This may have led to uncertainty in the estimates of the actual eligible population. Thus, for the reasons above, the uncertainty of the estimated prevalence will always remain.

The use of saliva for HIV antibody testing is particularly well suited for the screening of large populations as well as hard-to-reach populations in non-clinical settings, such as this rural area. A large number of specimens can be collected within limited time and the collection technique does not require specialised training skills. Moreover, studies have shown that testing oral fluids for HIV antibody testing has excellent sensitivity and a specificity for both low and high risk populations [24–26].

Our findings indicate a beginning HIV epidemic in this rural community, as opposed to the general stabilisation of HIV infection in the country as a whole [27]. In early stages of the HIV epidemic, most infections occur in individuals with higher sexual risk behaviours than the general population, and in our study men presented a higher risk of infection [28]. Infection in men occurs at older ages than in women, as has been demonstrated by other studies as well as ours. The male to female ratio of 0.8 puts men in the general population at higher risk, and this suggests that men play a significant role in spreading HIV infection in this area.

The positive association of mobility and increased HIV transmission risks has been shown in other studies [29, 30], and in our study it is likely that those who have recently moved into the area are relatively more mobile compared with the permanent residents. The shift of infection from more urban sites is evident, as indicated by the "rural/remote" prevalence ratio of 3.75. This seems to suggest that the "rural HIV epidemic" might similarly increase the remote HIV prevalence, as reported by other studies [31]. The "rural" site in this setting resembles an urban site where different economic activities prevail with the related social mixing patterns that may be scenarios where sexually transmitted infections are common.

We found relatively higher odds of HIV infection in one village where fishing is among the most important sources of income. This village is among the "rural" strata and considered relatively affluent. This observation is in keeping with findings reported in recent studies that have demonstrated increased HIV risk among those involved in fishing [32]. It is likely that fishing provides a steady income compared to farming and cattle keeping, and thus provides a power of influence to engage in commercial sex or to negotiate sex, or alcohol consumption leading to high-risk sexual practices.

Higher risks in men is associated with multiple sexual partners, either by polygamous marriages, infidelity or multiple sexual relationships that are culturally acknowledged [33]. Higher prevalence in men poses a potential threat to women in this area, due to the higher vulnerability of women created by socio-cultural values that deny them the power of negotiation in sexual relationships with their partners [34]. Among selected behavioural indicators, the "mean age at first sex" (from 17 years or more) is no different from the national estimate [13]. Since the majority of the population (62%) begin their first sexual experience at ages above 18 years, 38% of young people have an increased risk of HIV transmission by their early age of sexual experience. More importantly, the low condom utilisation by the younger age group further magnifies their exposure to HIV transmission risk since a very low proportion of this group (6.2%) used condoms at last sexual activity.

The predominant non-utilisation of condoms possibly based on religious and traditional values, irrespective of associated sexual risks, places this community at a higher risk of HIV and other sexually transmitted diseases. The results of the proxy measure of commercial sex "incentive sex" suggests that men who receive gifts of any kind for sex, are more likely to be infected compared to those who offer the incentive. We are uncertain of the existence of commercial sex by men, and who would be the recipients of such transactions since the main mode of HIV transmission has predominantly been heterosexual in most of sub-Saharan African countries. Studies on HIV transmission between men are scarce, and almost non-existent in Africa [35]. An anthropological study from the same area identified fertility as a central means of achieving status in this community [33]. Thus, measures such as condoms may be in conflict with this dominant socio-cultural norm. Under such circumstances, an increase in condom utilisation is unlikely within a short time. Similarly, a recent critique on international indicators of sexual risk behaviour elaborated the variability of interpretation of these indicators to assess sexual risk for HIV infection mainly due to the different factors that occur in various communities [36]. We are of the opinion that local understanding of sexual related behaviours based on socio-cultural contexts in this setting and community participation would have important consequences for designing and evaluating efforts to encourage self-protective behaviours. Based on these factors and our findings, we are likely to observe an escalation in the HIV prevalence in this setting. A similar situation was reflected by an established cohort within rural Mwanza in Northern Tanzania where an increasing HIV incidence/prevalence was observed despite various interventions in the community [37]. Changes in sexual behaviour had not been observed in the latter setting, reflected in findings from the rest of the country, highlighting the difficulty in achieving behavioural changes [13, 38].

Sexual risk behaviours revealed in this survey, such as multiple sexual partners and mobility, are well known risk factors for HIV transmission that are not different form elsewhere in Africa, [39, 40]. Despite these factors, the HIV prevalence has remained relatively low in this area when compared to other areas within Tanzania and elsewhere in Eastern Africa. Thus, identifying factors involved in maintaining the relatively low HIV prevalence in this setting to-date will be a key point in controlling the spread of HIV. There is convincing evidence that male circumcision has a protective effect against HIV transmission [41–44]. This factor is likely to have significantly contributed since male circumcision is a norm in this area. We can also hypothesize that the initial remoteness, lack of infrastructure, and hence, limited mobility, may have been protective. However, there are ongoing improvements in roads and trade leading to increasing mobility. These developments and the fact that HIV has infected close to 2% of the adult population aged 15–49 years need urgent and effective responses in terms of efforts to reduce transmission risk in this area. In this regard, there are particular challenges in reducing cultural practices that are likely to magnify the recently introduced virus. For instance in some tribes, the "rata" practice that essentially allows women to have "sexual and procreative relationships with a husband's clan brothers (rata)" [34]. Furthermore, there are obvious difficulties in identifying high-risk core group such as sex workers in a setting where sex work is informal.

In summary, the HIV prevalence from the general population was comparable to that of pregnant women attending antenatal clinics. We have identified a young and apparently potentially increasing HIV epidemic. The revealed patterns of sexual risk behaviours for example close to 50% of men having multiple partners and 78% of the population have never used a condom; it is likely that HIV infection will rapidly escalate. Immediate and effective preventive efforts that take into account the socio-cultural contexts are necessary to reduce the spread of the infection.