The HIV epidemics in sub-Saharan Africa appear strikingly diverse in maturity and magnitude [1–2], and the building of appropriate national knowledge bases is vital to enable AIDS programmes to guide and target preventive efforts. National surveillance and research systems have evolved over time, but data on patterns and trends of infection are predominantly based on sentinel surveillance among antenatal clinic (ANC) attendees [1–4]. However, the value in terms of validity and reliability of data from these systems, as well as their usefulness in providing the necessary information for timely decision making, continue to be of concern.
Most validations of ANC-based sentinel surveillance estimates of HIV prevalence have used single time-point surveys, and these have revealed that several inherited biases are introduced when extrapolating findings to the population of women in the general population [5–9]. The only validation of the method in monitoring trends showed HIV prevalence among ANC attendees to follow a similar pattern to that found in the general adult population . The study was conducted in a high prevalence population and during a period when the epidemic was declining, particularly in younger age-groups . A similar trend pattern has been observed in several areas in Uganda and has been shown to be associated with increased condom use and a decrease in casual sex [9,12,13].
Data from Zambia indicate a rapid increase of HIV prevalence in the early 1980s, apparently followed by a stabilization during the early 1990s [7,14]. In order to build a stronger national knowledge base, the core Zambian national surveillance and research system was refined in the mid-1990s by (i) extending the geographic coverage of the ANC-based sentinel system, and (ii) establishing a parallel system of population-based HIV surveys in selected sentinel populations in which HIV infection, social and behavioural data are collected concomitantly . We use data generated from this system to assess the continued evolution of the epidemic since early 1990s by examining patterns in spread of infection and linkages to behavioural changes.
National ANC-based sentinel surveillance
The system was established in 1990 (Lusaka and two other sites) and extended to four more sites during 1992–1993 [14,15]. A major extension came in 1994 (a total of 27 sites, all nine provincial capitals and at least one rural site in each province), and in 1998 it was repeated in 22 of these sites (n = 12 001). The reduction from 27 to 22 sites was for sustainability reasons and did not disturb the geographical representation.
PWomen attending the antenatal clinic for the first time during pregnancy were enrolled consecutively. The data collection period was fixed to a maximum of 4 months. A sample size of 500 was the target number for except four sites (Livingstone, Chelstone, Kapiri Mposhi and Ndola), where the target was higher in order to allow for more accurate age-specific analysis.
A structured interview comprising socio-demographic characteristics was established in 1994 (data on age and marital status also recorded in 1993) , and age, marital status, residence, years of school attendance and parity are variables used in the present analyses.
The testing strategy employed before 1994 has been described . The strategy in 1998 was essentially the same as in 1993 and 1994, apart from some variation in the type of test kits used . Serum samples were initially tested at site level using Capillus HIV-1/HIV-2 rapid test (Cambridge Biotechnology, Galway, Ireland). All positive samples and 5% of randomly selected negative samples were re-tested at the two national reference laboratories using Wellcozyme HIV Recombinant HIV–1 (Murex, Johannesburg, South Africa). Samples showing discordant results were confirmed using Bionor HIV-1&2 (BIONOR AS, Skien, Norway). For sites with ‘false-negative’ results, the Wellcozyme enzyme-linked immunosorbent assay was used to test 50% of negative samples in order to judge the need for further quality control.
The samples for the two population-based HIV surveys in 1995–1996 and 1998–1999 were selected from Chelston, Lusaka, and from Kapiri Mposhi district using a stratified random-cluster sampling method. Chelston is a residential area of the capitol City, and Kapiri Mposhi district has low population density with a dispersed, village-based rural population . In the follow-up survey the sample size from rural Kapiri Mposhi district was approximately doubled (five more clusters in addition to the five initially selected). The rationale was to improve power to detect changes in future follow-up surveys. No follow-up was conducted in urban Kapiri Mposhi district due to restricted funds.
In the sampled clusters members of all households aged 15 years and over were listed and contacted at home. The follow-up survey took place between mid- October 1998 and the end of May 1999. An interview with heads of households established a record of residents in the selected strata, and a listing system was designed to make individual linkages of information collected in the two surveys. Records on participants in the initial survey who were not found included date of leaving for those who had migrated from the study area and the date of death for those known to have died. The interviews and collection of saliva samples followed the same procedures as in the initial survey . In the initial survey all saliva samples were tested using Gacelisa HIV 1&2 (Welcome Diagnostics, Dartford, Kent, UK) and 450 randomly selected samples were tested using Bionor HIV-1&2 (BIONOR AS) magnetic particle assay following modifications for saliva. The two test kits showed a 99.8% agreement in test results . The accuracy of Gacelisa was validated based on paired saliva and serum samples collected from 494 antenatal care attenders, and both sensitivity and specificity was 100% . In the follow-up survey samples were tested using Bionor HIV-1&2.
The analyses used EpiInfo (Centre for Disease Control and Prevention, Atlanta, Georgia, USA) and Statistical Package for Social Sciences (SPSS for Windows; SPSS, Chicago, Illinois, USA).
The age-specific prevalence rates from the ANC-based data are presented here for groups of sites within urban and rural site categories, and sites were regrouped, based on similarity in prevalence levels and trends. We weighed proportions equally (for age-specific prevalence only), but statistical tests were based on pooled data from grouped sites (as for age-specific trends by education). Prevalence change from 1994 to 1998 was adjusted for changes in educational status over time, and for over-representation of women with low education ; that is, for adjustment of age-specific rates the standard population used was the educational structure of women in the general population, as revealed in population-based surveys conducted in 1995–1996 and 1998–1999. In some sentinel sites the structure of antenatal care services changed from 1994 to 1998, and the possibility of significant change in coverage was examined. The S2 tests were used to test for difference between rates (Mantel extension). Odds ratio (OR) and Cornfield 95% confidence interval (CI) for OR was used to test differences in proportions between groups. Rate ratio (RR) and Taylor 95% CI for RR was used to test differences in proportions between different years.
For the population-based HIV prevalence data, the 95% CI were calculated by taking into account the design effect of cluster sampling (CSAMPLE; EpiInfo). HIV prevalence rates were standardized for age using the Zambian population (1990 census projected for 1995 for both time points. Logistical regression models were used to analyse the relation between HIV infection and (i) educational achievement (years in school), (ii) school attendance (age-group 15–19 years only), and (iii) behavioural indicators. These analyses were stratified by urban/rural residence and sex. Age-group (5-year) and marital status (married, single, other) were entered as dummy variables whereas education and behavioural indicators were categorical. Migration was analysed using information on length of present residence, place migrated from (village or town), and out-migration between surveys.
Protocols were approved by the National AIDS Research Committee. Procedures were instituted to fulfil the requirements for unlinked anonymous testing of blood samples collected as part of the routine standards of antenatal care. Participation in the population survey was based on informed consent. Participants were informed that saliva-based testing was strictly for research purposes, and voluntary HIV counselling and testing was offered . A total of 10% of the urban and 6% of the rural participants reported to have been tested for HIV. Based on the experience from the initial survey, the voluntary HIV counselling and testing design was changed to improve acceptability . This resulted in a substantial improvement in terms of acceptability, particularly when counselling was offered outside a medical facility .
Overall HIV prevalence declined from 1994 to 1998 (pooled data, 22 sites) from 28.5 to 26.2% (RR, 0.92; 95% CI, 0.87–0.98) among urban residents and from 12.1 to 11.7% among rural residents. The urban/rural HIV prevalence ratio did not change significantly (RR 2.35 in 1994, 2.24 in 1998). The dominant age-specific prevalence pattern of individual sites was a declining trend among those aged 15–19 years, a diverse trend pattern across sites in other age-groups, and no significant change in the overall prevalence (Table 1). Among the sites with age-specific measurements since 1993 the decline among 15–19 year old women was strongest in the four Lusaka sites (down from 28.4% in 1993 to 14.8% in 1998; RR, 0.53; 95% CI, 0.41–0.71) and in the rural site Macha (down from 10.1 to 5.2%; RR, 0.51; 95% CI, 0.21–1.25). One rural site (Kalabo) showed a distinct pattern of increasing prevalence in most age-groups (Table 1). Considering sites from which only two age-specific measurements were available, most showed a decline in prevalence among 15–19 year olds similar to those for either Lusaka or Macha (Table 1). There was no difference in change in prevalence by parity.
Change in HIV prevalence differed by level of education (Table 2). The dominant pattern was decline in the higher educational groups and no change or rising rates among groups with lower education. Differences varied across age-groups and were most prominent in urban sites. Among urban women aged 20–24 years the prevalence declined significantly in women with more than 7 years’ schooling (RR, 0.76; 95% CI, 0.66–0.88) and tended to increase among those with less than 7 years of schooling (RR, 1.25; 95% CI, 0.98–1.59). Furthermore, the prevalence OR between those with more than 7 years versus less than 7 years schooling changed from 2.06 (95% CI, 1.50–2.83) in 1994 to 1.01 (95% CI, 0.79–1.30) in 1998. There was a divergence from this main pattern in two sites (Mongu and Chipata) where prevalence was stable across educational groups.
An over-representation of pregnant women with low education relative to women in the general population was found in urban sites. The proportion with 7 or more years of school attendance was 59% in the ANC-based sample from 1998 (65% in 1994), but 27% in the population-based sample of women (37% in 1996). Adjusted prevalence rates (using educational structure in the general population) also showed a significant urban decline in the age-group 20–24 years and for all ages. In the Lusaka sites the adjusted rates declined in the 20–24 years age-group from 34.1% in 1994 to 27.7% in 1998 (adjusted RR, 0.81; 95% CI, 0.69–0.94), and for all ages from 32.1 to 27.8% (adjusted RR, 0.87; 95% CI, 0.79–0.95).
Follow-up survey: participation and migration
A total of 6235 individuals (3515 urban and 2720 rural) were listed as residents (51% women). The interview team found 4472 individuals (2509 urban and 1963 rural) at home (de facto eligible population), and 4419 of these were interviewed (59% women). Among those asked to provide a saliva sample, 91% consented (rural consent rate 94%, urban 89%, no difference by sex). A full dataset linking information from interviews with saliva-based test results was obtained from 3757 individuals (59% women).
The proportion of participants in the 1996 survey who had migrated out from the area was 56% for the urban and 31% for the rural survey areas. A total of 89% of the urban in-migration was from towns, whereas 69% of the rural in-migration was from villages. Among urban men aged 15–19 years there was a significant excess prevalence rate among in-migrants compared with out-migrants (14.8 versus 3.2%, P < 0.01).
Change in HIV prevalence over time
Prevalence among urban men and women tended to decline in most age-groups and overall (Table 3), but the prevalence reduction was only significant among women aged 15–29 years (28.3 versus 24.1%; RR, 0.85; 95% CI, 0.74–0.99). Prevalence in the rural sample tended to decline among women aged 15–24 years (down from 16.1 to 12.2%; RR, 0.76; 95% CI, 0.48–1.23) whereas among men in this age-group there was a non-significant increase (5.7 versus 9.6%; RR, 1.69; 95% CI, 0.77–3.69).
Changes in prevalence by educational status followed a similar diverse pattern as revealed in the ANC-based data. For example, in age-groups under 30 years, regardless of gender and residence, prevalence reduction was associated with higher status and stable or increased risk with low status. There was a marked increase in prevalence particularly among men aged 15–19 years with low education. All in all, the positive association between risk of infection and educational status was reduced during the period. In 1999 there was a significant negative association between HIV prevalence and years of school in the 15–24 years age-group for both urban men (age-adjusted OR, 0.85; 95% CI, 0.74–0.98) and women (age-adjusted OR, 0.93; 95% CI, 0.85–0.99) but no difference in the rural sample (adjusted OR, 1.04 in men and 1.01 for women).
In the initial survey, urban out-of school women aged 15–19 years of age were more likely to be infected than women still attending school (OR, 3.3; 95% CI, 1.46–7.46). This association was less pronounced in the 1999 data (OR, 2.1; 95% CI, 0.95–4.50), but was now also appearing among young rural women (results not given due to small numbers).
The proportion of discordant couples (one partner HIV-positive and the other HIV-negative) was 31.8% in the 1996 urban sample (men accounting for 58% of HIV-positive partners in discordant partnerships; number of couples (n) = 201) and 20.1% in the rural sample (men accounting for 55%, n = 154). In 1999 the respective proportions of discordant couples were 25.5% (62% men, n = 149) and 19.7% (60% men, n = 152). Concordant HIV-negative couples changed from 49.8 to 51% and from 73.4 to 70.4% in the urban and rural populations, respectively.
Figure 1 illustrates the change in the proportion of HIV-negative women who have ever given birth. Delayed age at first birth was found among urban women aged 17–22 years (P < 0.001), a delay that had resulted in a 40% decline in fertility in the age-group 15–24 years (measured as number of children ever given birth to, P < 0.001). No similar change was found among rural women.
In the aggregate analyses (15–49 years), urban men and women reported less sexual activity, fewer multiple sexual partners, and more consistent use of condoms in 1999 compared with 1996 (Table 4). Changes in sexual activity were most pronounced in the younger age-groups. The proportion of urban men aged 15–19 years reporting any sexual activity in the last 12 months declined from 47 to 23%, whereas the proportion with two or more sexual partners fell from 52 to 38% (Table 4). In contrast, no evident change was revealed in most behavioural indicators in the rural population (except that more than 90% reported to have less sexual partners compared with some years ago). Urban/rural differences in condom use were marked and increased during the period, regardless of age and gender (Table 4). The likelihood that urban men (15–49 years) used condoms when last having non-regular sex was 68% compared with 15% among rural men. There was a higher likelihood of rural residents reporting barriers to condom use related to availability constraints and religious beliefs, whereas ‘trust’ in the safety of condoms did not differ by residence (Table 4).
Condom use was strongly associated with educational status at both time points (and tended to have increased during the period). Taking men aged 20–29 years as an illustration, the likelihood of ‘use last time you had sex’ was 29% in the group with less than 8 years of schooling versus 69% among those with 10 or more years of school attendance. The rural contrast was 8 versus 54%, respectively. The association between number of different sexual partners (in the last 12 months) and education tended to have changed over time. Among urban men aged 20–29 years the mean number of sexual partners was 2.5 in the group with less than 8 years’ schooling versus 1.1 in that with 12 or more years. The findings in 1996 was 2.0 versus 2.1, respectively.
The logistic regression analyses to estimate the relation between HIV infection and behavioural indicators showed weak associations. However, the association between infection and number of sexual partners was consistently more prominent in the 1996 data (men aged 20–29 years: adjusted OR, 1.14; 95% CI, 0.99–1.31) compared with data from 1999 (adjusted OR, 0.99; 95% CI, 0.81–1.22). In the 1999 survey we introduced the question, ‘have you ever had sexual intercourse?', and the proportion with no sexual experience was particularly high in the 15–19 years age-group. However, among men the likelihood of HIV infection in the sexually active was not different compared with the sexually inactive (OR, 1.05), but stronger in women (OR, 1.89; 95% CI, 0.90–4.00). The predictive value was stronger in the 20–24 years age-group (20% reported no sexual experience), showing an OR of 5.6 (95% CI, 2.01–16.30) in women and 3.9 (95% CI, 0.51–30.0) in men.
The data from the ANC-derived national sentinel surveillance revealed a dominant trend of declining HIV prevalence among younger women since the early 1990s. Furthermore, the population-based observations over 3 years showed a reduction in prevalence among younger women. A similar age-specific pattern of decline has been reported from various areas in Uganda and one area in Tanzania [9–13]. It should be noted that our ANC-based decline was primarily seen in the 15–19 years age-group, but that age-specific rates adjusted for level of education revealed significant reduction in the urban prevalence also in the 20–24 years age-group and for all ages. Two phenomena contributed in this regard. First, a striking diversity in prevalence trends by level of education, i.e. declines in people with higher levels of education and stable or rising prevalence in less-educated groups. Second, women with low education were over-represented in the ANC-based data relative to the general population of women because of their higher fertility . A similar pattern of differences by educational status has been revealed in one area in Uganda , but in the Zambian data the contrasts were more pronounced and there were signs of continued increase in prevalence among the least educated. Men showed the same diverse pattern of change in prevalence by education as for women, but the prevalence change did not reach statistical significance.
Change in behaviour showed a similar pattern of diversity by educational status as that seen for prevalence. There was evidence in the urban data of increased condom use, decreased sexual activity and number of sexual partners, and among younger women a significant increase in age at first birth. It should be noted that these behavioural observations were limited to a 3 year period. National preventive efforts were in place in early 1990s, and it is likely that favourable changes occurred prior to 1996. Using the 1990 Lusaka population survey as a reference, this is demonstrated with regard to use of condoms and non-regular sex . The marked increase in the proportion of sexually active who ever used condoms (from 28% in 1990 to 60% in 1995) agrees with the recorded steep increase in condom distribution from 1992, and there can be no doubt that the most substantial increase in condom protection occurred during the first half of the 1990s. However, condom protection in the rural population continued to be far below the urban level; a finding that is in agreement with data from two nationally representative Zambian studies [20,21]. This particular urban–rural contrast appeared to be partly associated with religious and availability barriers.
HIV prevalence reflects mainly the balance between incidence, mortality and migration, and the prevalence decline could be attributed to changes in any of these factors. Declines in prevalence among younger women without similar reduction in incidence have been reported [13,22]. Although these studies have either been based on a limited period of observation or low power to detect incidence trends, they might point to HIV-associated mortality as an important contributing factor of such declines. In our data it might particularly have contributed to the decline in prevalence among women in the urban general population aged 20–29 years. Prevalence among those aged 15–19 years reflects incidence most closely since infections have occurred recently and mortality is assumed to be relatively low or stable. Kilian et al. compared age-specific, ANC-based prevalence trends with output from a mathematical model simulating behaviour change versus an uninfluenced epidemic . Their findings strongly supported the assumption that the observed decline in HIV prevalence in the age-group 15–19 years, and later among those aged 20–24 years, to a large degree reflects changes in incidence resulting from behavioural change.
Several studies have indicated migration to be associated with HIV infection [23–25], but our analyses did not reveal any such effect that could explain the prevalence decline among younger women. The only effect was among urban men aged 15–19 years, showing significant excess prevalence rate among in-migrants compared with out-migrants. The use of saliva-based testing appeared to be a highly acceptable method [7,16], and non-response due to refusal to provide saliva (6% in 1996 and 9% in 1999) does not introduce any significant bias. In the ANC-based data three sites were identified where coverage had changed over time. This introduced bias in the measured change in prevalence, and results from these sites were excluded from the trend assessments. Such bias also involved the particular findings from rural Kapiri Mposhi, and the direct comparison with population-based data should be interpreted with great caution. From the urban site this type of comparison using the 1999 data sets showed antenatal-clinic attendees to be fairly representative of women in the general population for all ages, and in the age-range 15–24 years. The respective comparison based on the 1994–1996 data revealed ANC-based estimates to underestimate the true overall prevalence but significantly overestimate the true prevalence in the 15–19 year age-group . However, at both time points pregnant women aged 25 years and over underestimated the true prevalence levels, and this is in agreement with data from a number of studies [5,6,8–10]. The documented lower fertility of HIV-infected as opposed to uninfected women appears to be an important contributing factor to this phenomenon [7,26,27].
Survey data on sexual behaviour are rarely validated [28,29], and population studies where trends in infection and behaviour are studied concomitantly might contribute in this regard. It is noteworthy that the pattern of educational status diversity in HIV prevalence trends corresponded well with that appearing in behavioural changes. Particularly illustrative was the inverse relation between number of sexual partners and educational status found in the 1999 data. This is in sharp contrast to survey findings from 1990 showing that a common pattern in African populations was the number of sexual partners to rise steeply by level of education . Nevertheless, the measured change in behaviour can to some degree be attributed to change over time in perceptions of acceptable and desirable responses to indicator questions. Our data showed signs of this phenomenon in the reporting of sexual activity among adolescents (15–19 years), particularly men. Accordingly, the measured sharp decline in the proportion of sexually active urban men should be interpreted with caution. With regard to the observed increase in age at first birth among uninfected younger women, information on whether women have given birth or not is assumed to be less prone to this type of bias. The observation indicates behavioural changes among uninfected women that have strong impact on both transmission and fertility, and there is a need for particular studies of the mechanisms involved.
Behaviour change has often been observed as a slow process that takes place through diffusion of innovations from higher social status groups and downwards , and the resemblance with the marked educational status differential in change over time described here is noteworthy. There can be no doubt that the well-educated segments of the population changed behaviour substantially over a relatively short period of time. Preventive efforts to sustain the ongoing process of change in these sub-populations will be of great importance. A conglomerate of preventive programmes were operative in Zambia in the 1990s, and the extent to which they directly affected behavioural changes can not be answered here. However, the modest change identified among the most deprived groups represents a major preventive challenge. The contrasts in behavioural responses appeared to reflect underlying social and environmental determinants of risk , and the necessary adjustment of preventive efforts needs to be focused on such barriers to change.
The authors appreciate the assistance of participants, the staff of Central Statistical Office, Lusaka, and the Zambian National AIDS/STD/TB & Leprosy Programme, the research assistants, counsellors and drivers, and the staff of the national reference laboratories at University Teaching Hospital, Lusaka and at the Tropical Diseases Research Centre, Ndola.
1. Schwartländer B, Stanecki KA, Brown T. et al
. Country-specific estimates and models of HIV and AIDS: methods and limitations. AIDS 1999, 13: 2445 –2458.
2. Tarantola D, Schwartländer B. HIV/AIDS epidemics in sub-Saharan Africa: dynamism, diversity and discrete declines? AIDS 1997, 11 (Suppl B): S5 –21.
3. Chin J. Public health surveillance of AIDS and HIV infection. Bull WHO 1990, 68: 529 –536.
4. United Nations Programme on HIV/AIDS and World Health O rganization. Guidelines on Second Generation Surveillance of HIV Infection.
WHO/CDS/CSR/EDC/2000.5; UNAIDS/00.03E. Geneva: World Health Organization.
5. Borgdorff M, Borongo L, van Jaarsveld E. et al
. Sentinel surveillance for HIV-1 infection: how representative are blood donors, outpatient with fever, anemia, or sexually transmitted diseases, and antenatal clinic attenders in Mwanza Region, Tanzania? AIDS 1993, 7: 567 –572.
6. Kigadye R-M, Klokke A, Nicoll A. et al
. Sentinel surveillance for HIV-1 among pregnant women in a developing country: 3 years’ experience and comparison with a population sero survey. AIDS 1993, 7: 849 –855.
7. Fylkesnes K, Nhlovu Z, Kasumba K, Musonda RM, Sichone M. Studying dynamics of the HIV epidemic: population-based data compared with sentinel surveillance in Zambia. AIDS 1998, 12: 1227 –1234.
8. Kwesigabo G, Killewo JZJ, Sandstom A. Sentinel surveillance and cross sectional survey on HIV infection prevalence: a comparative study. East Afr Med J 1996, 73: 298 –302.
9. Kilian AHD, Gregson S, Ndyanabangi B. et al
. Reductions in risk behaviour provide the most consistent explanation for declining HIV-1 prevalence in Uganda. AIDS 1999, 13: 391 –398.
10. Kwesigabo G, Killewo JZJ, Urassa W. et al
. Monitoring of HIV-1 infection prevalence and trends in the general population using pregnant women as a sentinel population: 9 years experience from Kagera region of Tanzania. J Acquir Immune Defic Syndr 2000, 23: 410 –417.
11. Kwesigabo G, Killewo J, Godoy C. et al
. Decline in the prevalence of HIV-1 infection in young women in the Kagera region of Tanzania. J Acquir Immune Defic Syndr 1998, 17: 262 –268.
12. Asiimwe-Okiror G, Opio AA, Musinguzi J, Madraa E, Tembo G, Caraël M. Change in sexual behaviour and decline in HIV infection among young pregnant women in urban Uganda. AIDS 1997, 11: 1157 –1163.
13. Kamali A, Carpenter LM, Whitworth GAG, Pool R, Ruberanter A, Ojwiya A. Seven-year trends in HIV-1 infection rates, and changes in sexual behaviour, among adults in rural Uganda. AIDS 2000, 14: 427 –434.
14. Fylkesnes K, Musonda RM, Kasumba K. et al
. The HIV epidemic in Zambia: socio-demographic prevalence patterns and indications of trends among childbearing women. AIDS 1997, 11: 339 –345.
15. Fylkesnes K, Musonda RM, Luo NP, Msiska R. HIV infection in pregnant women in Zambia 1990–1993 [letter]. AIDS 1996, 10: 555 –556.
16. Fylkesnes K, Kasumba K. The first Zambian population-based HIV survey: saliva-based testing is accurate and acceptable [letter]. AIDS 1998, 12: 540 –541.
17. Fylkesnes K, Haworth A, Rosensvärd C, Kwapa PM. HIV counselling and testing: overemphasising high acceptance rates a threat to confidentiality and the right not to know. AIDS 1999, 13: 2469 –2474.
18. Fylkesnes K. Consent and HIV counselling and testing. Lancet 2000, 356 (Suppl. 1): 43. 43.
19. Cleland J, Ferry B (eds). Sexual Behaviour and AIDS in the Developing World
. London: World Health Organization, Taylor & Francis Publishers; 1995.
20. Central Statistical Office [Zambia], Ministry of Health, Macro International Inc. Zambia Demographic and Health Survey 1996.
Lusaka and Maryland: Central Statistical Office [Zambia], Ministry of Health, Macro International Inc.; 1997.
21. Central Statistical Office (Zambia), Project Concern International, Zambia, and MEASURE Evaluation, University of North Carolina at Chapel Hill. Zambia Sexual Behaviour Survey 1998
. Lusaka: Central Statistical Office (Zambia); 1999.
22. Wawer MJ, Serwadda D, Gray RH. et al
. Trends in HIV-1 prevalence may not reflect trends in incidence in mature epidemics: data from the Rakai population-based cohort, Uganda. AIDS 1997, 11: 1023 –1030.
23. Decosas J, Kane F, Anarfi JK, Sodji KDR, Wagner HU. Migration and AIDS. Lancet 1995, 346: 826 –828.
24. Kane F, Alary M, N'Doye I. Temporary expatriation is related to HIV-1 infection in rural Senegal. AIDS 1993, 7: 1261 –1265.
25. Nunn AJ, Wagner HU, Kamali A, Kengeya-Kayondo JF, Mulder DW. Migration and HIV-1 seroprevalence in a rural Ugandan population. AIDS 1995, 9: 503 –506.
26. Gray RH, Wawer MJ, Serwadda D. et al
. Population-based study of fertility in women with HIV infection in Uganda. Lancet 1998, 351: 98 –103.
27. Zaba B, Gregson S. Measuring the impact of HIV on fertility in Africa. AIDS 1998, 12 (Suppl 1): S41 –50.
28. Konings E, Banteby G, Carael M, Bagenda D, Mertens T. Validating population surveys for the measurement of HIV/STD prevention indicators. AIDS 1995, 9: 375 –382.
29. Japheth ZL, NG'weshemi J, Ties Boerma J. et al
. Changes in male sexual behaviour in response to the AIDS epidemic: evidence from a cohort study in urban Tanzania. AIDS 1996, 10: 1415 –1420.
30. Rogers EM. Diffusion of Innovations
. 3rd edn. New York: The Free Press; 1983.
31. d'Cruz-Grote D. Prevention of HIV infection in developing countries. Lancet 1996, 348: 1071 –1074.