Age-stratified or restricted analyses were available for 35 populations under 30 years of age (Table 1). In Kisumu, Kenya, and Ndola, Zambia, there was no association between educational attainment and HIV among males or females aged 15–24 years. In Yaounde, Cameroon, among females aged 15–24 years, the most educated were at the lowest risk of infection, while there was no association among young males. In Cotonou, Benin, the most educated males aged 15–24 years were at the lowest risk of infection, but there was no association among females. In Manicaland, Zimbabwe, there was a lower risk of infection among the most educated females aged 15–24 years, while among males aged 17–29 years there was no association. In Mwanza, Tanzania, ever having attended school was associated with a lower risk of infection among males, but not females, aged 15–19 years.
Time-series analyses were also available for some populations of young people. In Masaka, Uganda, there was no significant association among males aged 18–29 years in 1989–90 or 1999–2000. However, among females there was no association in 1989–90 but a significantly lower risk of infection among the most educated in 1999–2000. In Fort Portal, Uganda, there was no association among antenatal clinic attendees aged 15–24 years in 1991–4, whereas by 1995–7 there was a lower risk of infection among the most educated. Among Zambian antenatal clinic attendees aged 15–24 years, there was a higher risk of infection among the most educated in rural and urban areas in 1994, 1998 and 2002. In Zambian population-based surveys, educational attainment was not associated with risk of infection among rural males aged 15–24 years in any year. Among urban males, there was no association in 1995 and 1999, but by 2003 there was a lower risk of infection among the most educated. Among Zambian females from urban areas, there was no association among younger women in 1995 and 1999, whereas by 2003 a lower risk was seen among the most educated women aged 15–24 years. In rural Zambia, there was no association among younger women in 1995 and 2003, but a higher risk of infection among the most educated women in 1999.
We present evidence that the epidemiology of HIV infection in sub-Saharan Africa may be changing. Studies conducted before 1996 tended to find either no association with education level or a higher risk of HIV infection among the most educated. A larger proportion of studies conducted from 1996 onwards identified a lower risk of infection among the most educated. Where data over time were available, the trend was generally for strong positive associations to be replaced by weaker or negative associations. Across many settings, HIV prevalence fell more consistently among the higher educated than among the less educated groups, in whom prevalence sometimes rose even while overall population prevalence was falling. Taking these findings together, we suggest that new HIV infections occurring in the latter half of the 1990s and into the 21st century have been occurring disproportionately among the least educated members of society in many sub-Saharan African countries.
Our attempt to synthesize the available evidence might be subject to several limitations. There may have been limitations in the included studies . For instance, population studies may have been biased if eligible individuals excluded from the study differed from those included. However, response rates were high where they were reported. Studies of antenatal clinic attendees are generally unlinked–anonymized, with approximately 100% response rates, but include only pregnant women who attend antenatal clinics , thus excluding the sexually inactive  and those not using clinics, who may be less educated. Since delayed first pregnancy has become more common over time in many settings, young women who do become pregnant may be at particularly high risk of infection, perhaps explaining why antenatal clinic data more often suggested higher levels of infection among the most educated and less evidence of changes over time [40,45].
We only included studies that adjusted for (or restricted by) age and sex, and that did not adjust for sexual-behaviour factors that are likely to be on the causal pathway between education level and HIV. An exception was made for marital status, as many of the studies had controlled for this. In one study that explicitly explored the confounding effect of marriage on the association of education status and HIV risk in four different settings, it had little effect on the results .
We cannot exclude publication bias: perhaps studies finding a lower risk of HIV infection among more educated groups, or a shifting pattern of infection towards this trend, may have been more likely to be published between 2001 and 2006. However, education level was often not the main focus of the studies, so it is unlikely to have biased publication. Finally, data were only available from 11 of the countries in the region, a minority.
Overall, available data from a number of settings suggested that the trend seen in data collected before 1996, of a higher risk of infection among the most educated groups, was weakening, and in some cases reversing, over time. Since patterns of prevalent HIV infection are relatively slow to respond to changes in HIV incidence, it is possible that these patterns hide much greater relative differences in HIV incidence between the most and least educated in recent years. Where time-series data were available on younger age groups, in whom infection is likely to have been acquired recently, this trend was also seen. Previous studies have attempted to infer time trends by comparing the association between education and HIV between younger and older groups [15,27,32]. However, differential AIDS-related mortality in these groups makes these comparisons difficult  and a strength of our study was the inclusion of data collected in different time periods to address this question. Differential mortality could only explain the changes seen in different time periods if overall rates of AIDS-related mortality were changing rapidly in some age groups over time, which is unlikely as these studies predate the widespread accessibility of antiretroviral drugs.
While an overall pattern emerged, there was also some evidence of differences between population groups, for example with differences between rural and urban Zambia. Zambia was the only country with data from a wide range of regions within the country; other results came from single regions or towns, and for most countries of sub-Saharan Africa there were no data at all. It is possible that patterns in other countries may differ.
Strong supporting evidence of behaviour change among the most educated comes from studies that have consistently found higher levels of reported condom use among more educated individuals in a variety of contexts [28,45,47–53]. However, evidence with regard to other safer-sexual behaviours, such as delayed age at first sex and reduced partner numbers, is less consistent: some studies have suggested lower risk among the most educated but this has not been seen in other settings [28,31,45,48,53–56]. The plausibility of the hypothesis of change occurring earlier in those with more education is also supported by well-established theory. For example, the diffusion of innovations model predicts that more-educated, more-empowered members of a population will seek out information on new practices (such as condom use in sub-Saharan Africa in the latter part of the 20th century) and more readily adopt these .
The evidence presented here should highlight the importance of monitoring future trends within surveillance systems across sub-Saharan Africa. We believe there is already sufficient evidence to support consideration of policy responses to the changing epidemiology of HIV infection we outline in this paper, since such responses will need to be rapidly rolled out if the trends reported here are confirmed in ongoing surveillance. Current approaches, focusing primarily on the provision of information, distribution of condoms and treatment of sexually transmitted infections have worked and should continue. However, these efforts may effectively serve only some sections of the community, and health inequalities in sub-Saharan Africa may be set to increase. Additional efforts are needed to expand the reach of HIV prevention programmes to target socially vulnerable groups more effectively and to address social inequalities. As an example, recent interventions to improve school enrolment, such as the abolition of primary school fees, have met with dramatic success in Kenya, Malawi, Tanzania and Uganda as part of efforts to achieve universal access to primary education . Such interventions may have a role to play in complementing more traditional HIV prevention methods in reducing HIV incidence in all social groups.
We would like to thank Knut Fylkesnes, Simon Gregson, James Lewis and Andrew Nunn for providing further information on some of the studies described here. We would also like to thank Andrew Thomson for his assistance in producing the figures.
1. UNAIDS. Report on the Global HIV/AIDS Epidemic
. Geneva: UNAIDS; 2006.
2. Hargreaves JR, Glynn JR. Educational attainment and HIV-1 infection in developing countries: a systematic review. Trop Med Int Health 2002; 7:489–498.
3. Wojcicki JM. Socioeconomic status as a risk factor for HIV infection in women in East, Central and Southern Africa: a systematic review. J Biosoc Sci 2005; 37:1–36.
4. Rogers EM. Diffusion of Innovations. New York: The Free Press; 1962.
5. Over M, Piot P. HIV infection and sexually transmitted diseases. In: Jamison DT, Mosley WH, Mensham AR, Bobadilla JL, editors. Disease Control Priorities in Developing Countries. Oxford: Oxford University Press; 1993. pp. 455–527.
6. Barongo LR, Borgdorff MW, Mosha FF, Nicoll A, Grosskurth H, Senkoro KP, et al
. The epidemiology of HIV-1 infection in urban areas, roadside settlements and rural villages in Mwanza Region, Tanzania. AIDS 1992; 6:1521–1528.
7. Fontanet AL, Woldemichael T, Sahlu T, van Dam GJ, Messele T, Rinke de Wit T, et al
. Epidemiology of HIV and Schistosoma mansoni
infections among sugar-estate residents in Ethiopia. Ann Trop Med Parasitol 2000; 94:145–155.
8. Grosskurth H, Mosha F, Todd J, Senkoro K, Newell J, Klokke A, et al
. A community trial of the impact of improved sexually transmitted disease treatment on the HIV epidemic in rural Tanzania: 2. Baseline survey results. AIDS 1995; 9:927–934.
9. Kwesigabo G, Killewo J, Godoy C, Urassa W, Mbena E, Mhalu F, 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.
10. 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 1996; 11:484–491.
11. Nunn AJ, Wagner HU, Okongo JM, Malamba SS, Kengeya-Kayondo JF, Mulder DW. HIV-1 infection in a Ugandan town on the trans-African highway: prevalence and risk factors. Int J STD AIDS 1996; 7:123–130.
12. Quigley MA, Morgan D, Malamba SS, Mayanja B, Okongo MJ, Carpenter LM, Whitworth JA. Case–control study of risk factors for incident HIV infection in rural Uganda. J Acquir Immune Defic Syndr 2000; 23:418–425.
13. Senkoro K, Boerma JT, Klokke A, Ng'weshemi JZL, Muro AS, Gabone R, Borgdorff MW. HIV incidence and HIV-associated mortality in a cohort of factory workers and their spouses in Tanzania, 1991 through 1996. J Acquir Immune Defic Syndr 2001; 23:194–202.
14. Fylkesnes K, Masonda RM, Kasumba K, Ndhlovu Z, Mluanda F, Kaetano L, Chipaila CC. The HIV epidemic in Zambia: sociodemographic prevalence patterns and indications of trends among childbearing women. AIDS 1997; 11:339–345.
15. Fylkesnes K, Musonda RM, Sichone M, Ndhlovu Z, Tembo F, Monze M. Declining HIV prevalence and risk behaviours in Zambia: evidence from surveillance and population based surveys. AIDS 2001; 15:907–916.
16. Gregson S, Anderson RM, Ndlovu J, Zhuwau T, Chandiwana SK. Recent upturn in mortality in rural Zimbabwe: evidence for an early demographic impact of HIV-1 infection? AIDS 1997; 11:1269–1280.
17. Gregson S, Mason PR, Garnett GP, Zhuwau T, Nyamukapa CA, Anderson RM, Chandiwana SK. A rural HIV epidemic in Zimbabwe? Findings from a population-based survey. Int J STD AIDS 2001; 12:189–196.
18. Gregson S, Zhuwau T, Anderson RM, Chimbadzwa T, Chiwandiwa SK. Age and religion selection biases in HIV-1 prevalence data from antenatal clinics in Manicaland, Zimbabwe. Cent Afr J Med 1995; 41:339–346.
19. Kilian AH, Gregson S, Ndyanabangi B, Walusaga K, Kipp W, Sahlmuller G, et al
. Reductions in risk behaviour provide the most consistent explanation for declining HIV-1 prevalence in Uganda. AIDS 1999; 13:391–398.
20. Kirunga CT, Ntozi JP. Socio-economic determinants of HIV serostatus: a study of Rakai District, Uganda. Health Transit Rev 1997; 7(Suppl):175–188.
21. Konde-Lule JK, Wawer MJ, Sewankambo NK, Serwadda D, Kelly R, Li C, et al
. Adolescents, sexual behaviour and HIV-1 in rural Rakai district, Uganda. AIDS 1997; 11:791–799.
22. Serwadda D, Wawer MJ, Musgrave SD, Sewankambo NK, Kaplan JE, Gray RH. HIV risk factors in three geographic strata of rural Rakai District, Uganda. AIDS 1992; 6:983–989.
23. Smith J, Nalagoda F, Wawer MJ, Serwadda D, Sewankambo N, Konde-Lule J, et al
. Education attainment as a predictor of HIV risk in rural Uganda: results from a population-based study. Int J STD AIDS 1999; 10:452–459.
24. Abebe Y, Schaap A, Mamo G, Negussie A, Darimo B, Wolday D, Sanders EJ. HIV prevalence in 72 000 urban and rural male army recruits, Ethiopia. AIDS 2003; 17:1835–1840.
25. Boisier P, Ouwe Missi Oukem-Boyer ON, Amadou Hamidou A, Sidikou F, Ibrahim ML, Elhaj Mahamane A, et al
. Nationwide HIV prevalence survey in general population in Niger. Trop Med Int Health 2004; 9:1161–1166.
26. Crampin AC, Glynn JR, Ngwira BM, Mwaungulu FD, Ponnighaus JM, Warndorff DK, Fine PE. Trends and measurement of HIV prevalence in northern Malawi. AIDS 2003; 17:1817–1825.
27. de Walque D, Nakiyingi-Miiro JS, Busingye J, Whitworth JA. Changing association between schooling levels and HIV-1 infection over 11 years in a rural population cohort in south-west Uganda. Trop Med Int Health 2005; 10:993–1001.
28. Glynn JR, Carael M, Buve A, Anagonou S, Zekeng L, Kahindo M, Musonda R. Does increased general schooling protect against HIV infection? A study in four African cities. Trop Med Int Health 2004; 9:4–14.
29. Glynn JR, Ponnighaus J, Crampin AC, Sibande F, Sichali L, Nkhosa P, et al
. The development of the HIV epidemic in Karonga District, Malawi. AIDS 2001; 15:2025–2029.
30. Gregson S, Terceira N, Kakowa M, Mason PR, Anderson RM, Chandiwana SK, Carael M. Study of bias in antenatal clinic HIV-1 surveillance data in a high contraceptive prevalence population in sub-Saharan Africa. AIDS 2002; 16:643–652.
31. Gregson S, Waddell H, Chandiwana SK. School education and HIV control in sub-Saharan Africa: from discord to harmony? J Int Dev 2001; 13:467–485.
32. Michelo C, Sandoy IF, Fylkesnes K. Marked HIV prevalence declines in higher educated young people: evidence from population-based surveys (1995–2003) in Zambia. AIDS 2006; 20:1031–1038.
33. Nyambi P, Zekeng L, Kenfack H, Tongo M, Nanfack A, Nkombe I, et al
. HIV infection in rural villages of Cameroon. JAIDS 2002; 31:506–513.
34. Obasi AI, Balira R, Todd J, Ross DA, Changalucha J, Mosha F, et al
. Prevalence of HIV and Chlamydia trachomatis
infection in 15–19-year olds in rural Tanzania. Trop Med Int Health 2001; 6:517–525.
35. Todd J, Grosskurth H, Changalucha J, Obasi A, Mosha F, Balira R, et al
. Risk factors influencing HIV infection incidence in a rural African population: a nested case–control study. J Infect Dis 2006; 193:458–466.
36. Fabiani M, Nattabi B, Opio AA, Musinguzi J, Biryahwaho B, Ayella EO, et al
. A high prevalence of HIV-1 infection among pregnant women living in a rural district of north Uganda severely affected by civil strife. Trans R Soc Trop Med Hyg 2006; 100:586–593.
37. Auvert B, Buve A, Ferry B, Carael M, Morison L, Lagarde E, et al
. Ecological and individual level analysis of risk factors for HIV infection in four urban populations in sub-Saharan Africa with different levels of HIV infection. AIDS 2001; 15(suppl 4):S15–S30.
38. Lagarde E, Congo Z, Meda N, Baya B, Yaro S, Sangli G, et al
. Epidemiology of HIV infection in urban Burkina Faso. Int J STD AIDS 2004; 15:395–402.
39. Bloom SS, Urassa M, Isingo R, Ng'weshemi J, Boerma JT. Community effects on the risk of HIV infection in rural Tanzania. Sex Transm Infect 2002; 78:261–266.
40. Sandoy IF, Kvale G, Michelo C, Fylkesnes K. Antenatal clinic-based HIV prevalence in Zambia: declining trends but sharp local contrasts in young women. Trop Med Int Health 2006; 11:917–928.
41. Fabiani M, Nattabi B, Ayella EO, Ogwang M, Declich S. Using prevalence data from the programme for the prevention of mother-to-child-transmission for HIV-1 surveillance in North Uganda. AIDS 2005; 19:823–827.
42. Mnyika KS, Klepp KI, Kvale G, Nilssen S, Kissila PE, Ole-King'ori N. Prevalence of HIV-1 infection in urban, semi-urban and rural areas in Arusha region, Tanzania. AIDS 1994; 8:1477–1481.
43. Michelo C, Sandoy IF, Dzekedzeke K, Siziya S, Fylkesnes K. Steep HIV prevalence declines among young people in selected Zambian communities: population-based observations (1995–2003). BMC Public Health 2006; 6:279.
44. Fylkesnes K, Ndhlovu Z, Kasumba K, Mubanga MR, Sichone M. Studying dynamics of the HIV epidemic: population-based data compared with sentinel surveillance in Zambia. AIDS 1998; 12:1227–1234.
45. Sandoy IF, Michelo C, Siziya S, Fylkesnes K. Associations between sexual behaviour change in young people and decline in HIV prevalence in Zambia. BMC Public Health 2007; 7:60.
46. Cochrane Collaboration. Cochrane Handbook for Systematic Reviews of Interventions
, version 4.2.5. Chichester, UK: John Wiley; 2005 www.cochrane.org/resources/handbook/
[updated May 2005].
47. Agha S, Hutchinson P, Kusanthan T. The effects of religious affiliation on sexual initiation and condom use in Zambia. J Adolesc Health 2006; 38:550–555.
48. Blanc AK. The Relationship Between Sexual Behavior and Level of Education in Developing Countries
. Columbia, MD: Blancroft Research International, for UNAIDS; 2001.
49. Camlin CS, Chimbwete CE. Does knowing someone with AIDS affect condom use? An analysis from South Africa. AIDS Educ Prev 2003; 15:231–244.
50. Kapiga SH, Lugalla JL. Sexual behaviour patterns and condom use in Tanzania: results from the 1996 Demographic and Health Survey. AIDS Care 2002; 14:455–469.
51. Kongnyuy EJ, Wiysonge CS, Mbu RE, Nana P, Kouam L. Wealth and sexual behaviour among men in Cameroon. BMC Int Health Hum Rights 2006; 6:11.
52. Lagarde E, Carael M, Glynn JR, Kanhonou L, Abega SC, Kahindo M, et al
. Educational level is associated with condom use within nonspousal partnerships in four cities of sub-Saharan Africa. AIDS 2001; 15:1399–1408.
53. Magnani RJ, Karim AM, Weiss LA, Bond KC, Lemba M, Morgan GT. Reproductive health risk and protective factors among youth in Lusaka, Zambia. J Adolesc Health 2002; 30:76–86.
54. Filmer D. The socioeconomic correlates of sexual behavior: a summary of results from an analysis of DHS data
. In: Ainsworth M, Fransen L, Over M, editors. Confronting AIDS: Public Priorities in a Global Epidemic
. New York: Oxford University Press; 1998.
55. Lugoe WL, Klepp KI, Skutle A. Sexual debut and predictors of condom use among secondary school students in Arusha, Tanzania. AIDS Care 1996; 8:443–452.
56. Meekers D, Ahmed G. Contemporary patterns of adolescent sexuality in urban Botswana. J Biosoc Sci 2000; 32:467–485.
57. UNICEF. The State of the World's Children
. New York: UNICEF; 2004.