HIV prevalence levels by recent use of contraceptives are shown in Table 1 and are contrasted with those for recently pregnant women in Fig. 2b. Women using contraceptives other than condoms are at similar risk of HIV infection to those with a recent pregnancy, whereas those reporting condom use or no method of family planning are at significantly greater risk. The differences are modest, but women who report contraceptive use appear to be at lower risk of HIV infection at young ages (possibly due to later sexual debut) and greater risk at older ages (probably due to HIV-associated infertility and, perhaps, contraceptive use by some women who suspect they may be HIV-positive). Non-contraceptive users and condom users are more likely to be HIV-positive than recently pregnant women at all ages; the degree of excess risk increases monotonically with age from the early twenties onwards. The high HIV prevalence in condom users reflects an association with casual sexual relationships .
HIV prevalence in women tested at local ANCs in the study areas was lower at all ages than in pregnant women from the surrounding areas (Table 1 and Fig. 2c). However, those of the latter who reported attending an ANC elsewhere (24.7%) or who had not been for an antenatal check-up (21.9%) were no more likely to be infected with HIV than those who reported visits to local ANC (27.0%).
Overall, ANC attendance is very high in Manicaland (Table 2). Most of the women interviewed who had not had an ANC check-up were in the first 6 months of pregnancy. Of those whose pregnancy was now complete or within the final trimester, 91% had been for a check-up; 81% of these attended at their local clinic. More educated and unmarried women were less likely to attend their local clinic. Older women and those at advanced stages of pregnancy were more likely to use a local clinic. Local ANC use was more common among women from small towns and commercial farming estates than in those from subsistence farming areas and roadside trading settlements.
The lower HIV prevalence in local ANC attendees than in pregnant women from the surrounding areas was most apparent in women in their twenties, unmarried women, the more educated, and those living in the small towns. There was no difference in HIV infection levels in the commercial farming estates. Only 4% of the women tested at ANC were visitors rather than local residents. Visitors were most common in the towns (10%) where the majority (10/19) were from subsistence farming areas.
Adjustment for bias in pregnant women
Applying the adjustment procedure proposed by Zaba and colleagues  to the HIV prevalence rate in recently pregnant women reporting local ANC attendance (27.0%) yields revised estimates of 22.6% and 23.5%, with and without adjustment for birth interval length, respectively. These estimates are slightly lower than the figure obtained directly for women in the population survey (25.5%). Applying the procedure to the ANC data reduces the HIV prevalence estimate from 20.8% – marginally lower than the figure in Table 1 (21.1%) as 75 (6.2%) women had incomplete data on birth intervals – to 19.2%.
Socio-demographic risk factors
The risk ratios for HIV infection by level of education observed in the population and antenatal surveys are compared in Table 3. The population data highlight how female secondary education has increased in Zimbabwe over the past 20 years. These data also suggest that within the age-range 15–44 years, women with secondary education have a lower risk of HIV infection than those with less education (age-adjusted OR, 0.7; 95% CI, 0.6–0.9;P < 0.001). In contrast, the ANC data point to a slightly higher risk among the more educated women (age-adjusted OR, 1.1; 95% CI, 0.8–1.4;P = 0.659). The data for recently pregnant women in the general population gave a similar result (age-adjusted OR, 1.2; 95% CI, 0.8–1.9;P = 0.301).
The discrepancy is greatest in teenagers: those with secondary education in the ANC sample being 3 times as likely (χ2(1) = 16.58;P < 0.001) to be HIV-positive as those in the population survey. Elsewhere, we have shown that this is because more educated women commence sexual activity later . Thus, those with early pregnancies are a highly selected subset. In contrast, there is almost no difference between the HIV prevalence levels found in young women with less education in the two samples.
Previous studies have established that data on pregnant women frequently provide under-estimates of female HIV prevalence in sub-Saharan African populations  but reasonable estimates of both-sex prevalence  as HIV levels are generally considerably lower in men than in women in established epidemics . Uncorrected ANC estimates are therefore used as the basis for official estimates of HIV prevalence in all adults [2,19] with figures for men and women being adjusted downwards and upwards, respectively.
However, existing studies have been conducted almost exclusively in low income, high fertility populations. Here, we have presented data on bias in HIV prevalence estimates from pregnant women and ANC attendees in a high contraceptive use population. In this setting, the ANC data again understate female population HIV prevalence. But the reasons for this are different: there is no difference in HIV prevalence between pregnant women and women as a whole over the reproductive age range but the level of HIV infection is lower in women attending local ANC than in pregnant women living in the immediately surrounding areas.
The first of these results confirms earlier predictions that the bias in HIV estimates based on data from pregnant women would be smaller in populations at advanced stages of fertility transition. There are two reasons. The first follows from the fact that young women in these populations are typically less sexually active and more likely to use condoms or other effective methods of contraception. In rural Zimbabwe, female median age at first sex is 18.5 years  and more than one-third of the young women who have started sex have recently used contraception. Thus, over-representation of more sexually active, higher risk for HIV-infection women at young ages is greater and persists to older ages in this population.
The second reason is that bias due to subfertility in HIV-infected women [6,7] has a weaker effect in high contraceptive prevalence populations because subfertility also occurs in uninfected women who use contraception. In rural Zimbabwe, 60% of women reported recent use of a modern method of family planning. Those who did so were at lower risk of HIV infection than non-users, possibly because they are more open to adopting new patterns of behaviour. These women also become pregnant less often so are under-represented in samples based on pregnancy. At older ages, fewer women use contraceptives but those that do are seeking to stop rather than to space future births so very rarely become pregnant.
The reasons for the lower HIV prevalence in pregnant women attending local ANC compared to those living in the immediately surrounding areas are less clear. Most of the pregnant women interviewed in the population survey reported attending local ANCs. Younger, more educated and unmarried women – groups that are typically more geographically mobile – were more likely to report visiting a clinic outside their home area. Women from subsistence farming areas and adjoining roadside trading settlements were also more likely to have done so, perhaps because limitations in the rural transport system can make it easier to reach a clinic in a nearby town than to get to the nearest rural clinic. However, these women had a similar likelihood of HIV infection to those who attended their local clinics.
The discrepancy between the HIV prevalence results for local ANC attendees and local pregnant women could however reflect use of local ANCs by women from beyond the population survey catchment areas which tended to be concentrated around more accessible locations such as small towns, estates and roads. The largest difference was observed in the clinics in the small towns. Women from nearby rural areas may have received ANC check-ups at these clinics and, given the steep urban–rural gradient in female HIV prevalence recorded in the population survey (46% in towns; 22% in subsistence farming areas), would tend to suppress the HIV prevalence results obtained at these clinics. However, we were not able to confirm this directly as full details of residence were not recorded in the ANC survey. The smallest difference was seen in the estate clinics whose services are restricted to estate workers and their families who live in densely populated compounds.
There have been few previous studies of bias in estimates of HIV prevalence derived from samples of pregnant women in low fertility settings. One study in Addis Ababa, Ethiopia, conducted in the mid-1990s found much higher HIV prevalence in ANC attendees than in women in the general population  but the results are difficult to interpret as the ANC samples were collected 2 years after the population survey during a period when the epidemic may have been growing. A recent study in Ndola, Zambia, where contraceptive use was lower than in Manicaland but still relatively high by current sub-Saharan African standards (27% of recent ANC attendees) [21,22], found lower HIV prevalence in local ANC attendees than in the general female population. In part, this was because age at first sexual intercourse was still early (mean, 16.85 years).
Clearly, further studies in low-fertility populations are needed to test the broader relevance of our findings. As they stand, they suggest that figures for recently pregnant women would represent over-estimates of HIV prevalence in men and women combined in such settings. It is currently less certain that the compensating difference found here between HIV prevalence levels in local ANC attendees and pregnant women would apply elsewhere.
Unless this is the case, our results could explain some of the difference between contemporary estimates of HIV prevalence for central/eastern African countries and the higher income countries of southern Africa. The results also have implications for interpretation of future trends in ANC-based HIV prevalence estimates. Given the nature of the bias in estimates based on pregnant women, its impact in less developed countries would be expected to reduce as family planning becomes more widespread. Under these conditions, ANC-based HIV prevalence estimates could show artificial declines. Thus, parallel behavioural surveillance [23,24] will be especially important if correct inferences regarding the impact of behaviour change and HIV control strategies are to be drawn.
The possibility that urban and peri-urban ANCs might attract significant numbers of women from the surrounding rural areas where HIV prevalence is lower warrants further investigation as it could mitigate current concerns that national ANC-based surveillance estimates tend to exaggerate HIV prevalence levels due to over-representation of urban-based clinics . It may also be a bigger factor in more developed countries where better transport infrastructure and higher income levels facilitate rural–urban mobility and could be particularly significant in trading centres where the stable resident population is often relatively small.
Within countries, ANC-based sentinel surveillance systems generally show that infection levels are highest in urban areas. Use of urban ANCs by women from rural areas where HIV prevalence is lower would introduce an element of understatement into this differential. However, this is probably counter-balanced by the greater bias due to later onset of sexual relations and more widespread family planning in urban populations. The Zimbabwe results also show how ANC-based estimates can give a distorted picture of other socio-economic risk factors for HIV infection. In the case of education, the ANC data indicated that greater education was a risk factor for HIV infection particularly at young ages, when, in fact, the reverse was actually the case. A similar finding was obtained recently in Zambia .
Finally, a new procedure for adjusting HIV prevalence estimates obtained from samples of pregnant women has been proposed that takes account of underlying differences in fertility-related behaviour . This procedure works well in high fertility settings and is reasonably robust to changes in population structure . Our results suggest that it also differentiates well between high and low family planning use settings. However, further evaluations using data from independent sources are needed to confirm this. In the meantime, the minor modifications needed to ANC-based surveillance systems to incorporate collection (where this is not done already) and processing of data on parity and previous birth interval so that the procedure can be applied would seem to be worthwhile. At the same time, periodic population-based surveys that record details of fertility and its proximate determinants (such as the Demographic and Health Surveys conducted in many developing countries) need to be maintained so that reliable results can be obtained in circumstances where fertility-related behaviour is changing rapidly.
The authors thank T. Zhuwau, T. Mutevedzi, C. Nyamukapa, E. Dauka, L. Chisvo, M. Mlilo and the Manicaland Study fieldwork team for data collection assistance, J. Mutsvangwa and J. Magwenzi for conducting the laboratory tests, and the people of Manicaland for their kind co-operation and participation in the study.
1. Chin J, Mann J. Global surveillance and forecasting of AIDS. Bull WHO 1989, 67: 1–7.
2. UNAIDS. Report on the Global HIV/AIDS Epidemic.
Geneva: UNAIDS; 2000: 135135.
3. US Bureau of the Census. Recent HIV Seroprevalence Levels by Country: June 1996.
Washington, DC: US Bureau of the Census; 1996.
4. Zaba B, Gregson S. Measuring the impact of HIV on fertility in Africa. AIDS 1998, 12: S41–S50.
5. Widy-Wirski R, Berkley S, Downing R. et al
. Evaluation of the WHO clinical case definition for AIDS in Uganda. J Am Med Assoc 1988, 260: 3286–3289.
6. Carpenter LM, Nakiyingi JS, Ruberantwari A, Malamba SS, Kamali A, Witworth JAG. Estimates of the impact of HIV-1 infection on fertility in a rural Ugandan population cohort. Health Trans Rev 1997, 7: 113–126.
7. Gray R, Wawer M, Serwadda D. et al
. Population-based study of fertility in women with HIV-1 infection in Uganda. Lancet 1998, 351: 98–103.
8. Boerma JT, Urassa M, Isingo R. Female infertility and its association with sexual behaviour, STD and HIV infection in Tanzania.TANESA
Working Paper no 14; 1996.
9. Zaba B, Carpenter LM, Boerma JT, Gregson S, Nakiyingi J, Urassa M. Adjusting ante-natal clinic data for improved estimates of HIV prevalence among women in sub-Saharan Africa. AIDS 2000, 14: 2741–2750.
10. Gregson S, Zhuwau T, Anderson RM, Chimbadzwa T, Chandiwana SK. Age and religion selection biases in HIV-1 prevalence data from antenatal clinics in Manicaland, Zimbabwe. Central African J Med 1995, 41: 339–345.
11. Fylkesnes K, Ndhlovu Z, Kasumba K, Mubanga Musonda R, Sichone M. Studying dynamics of the HIV epidemic: population-based data compared with sentinel surveillance in Zambia. AIDS 1998, 12: 1227–1234.
12. Blanc A. The Relationship Between Sexual Behaviour and Level of Education in Developing Countries.
Geneva: UNAIDS; 2000.
13. Ray CS, Mason PR, Smith H, Rogers L, Tobaiwa O, Katzenstein DA. An evaluation of dipstick-dot immunoassay in the detection of antibodies to HIV-1 and HIV-2 in Zimbabwe. Trop Med Int Health 1997, 2: 83–88.
14. Central Statistical Office [Zimbabwe] and Macro International Inc. Zimbabwe Demographic and Health Survey 1999.
Calverton, Maryland: Central Statistical Office and Macro International Inc.; 2000.
15. Gregson S, Nyamukapa C, Garnett GP, et al
. Sexual mixing patterns and sex-differentials in teenage exposure to HIV infection in rural Zimbabwe.Lancet
16. 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.
17. UNAIDS. Report on the UNAIDS Epidemiology Reference Group Meeting, Rome, October 8–10, 2000.
Oxford: Oxford University; 2000.
18. Gregson S, Garnett GP. Contrasting gender differentials in HIV-1 prevalence and associated mortality increase in eastern and southern Africa: artefact of data or natural course of epidemics? AIDS 2000, 14: S85–S99.
19. 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.
20. Fontanet A, Messele T, Dejene A. et al
. Age- and sex-specific HIV-1 prevalence in the urban community setting of Addis Ababa, Ethiopia. AIDS 1998, 12: 315–322.
21. Glynn JR, Buvé A, Caraël M. et al
. Factors influencing the difference in HIV prevalence between antenatal clinic and general population in sub-Saharan Africa. AIDS 2001, 15: 1717–1725.
22. Glynn JR, Buvé A, Caraël M. et al
. Decreased fertility among HIV-1-infected women attending antenatal clinics in three African cities. J Acquir Immune Defic Syndr 2000, 24: 345–352.
23. Mills S, Saidel T, Bennett A. et al
. HIV risk behavioral surveillance: a methodology for monitoring behavioral trends. AIDS 1998, 12: S37–S46.
24. Amon J, Brown T, Hogle J. et al
. Behavioral Surveillance Surveys (BSS): Guidelines for Repeated Behavioral Surveys in Populations at Risk of HIV.
Arlington, Virginia: USA Family Health International; 2000: 350350.
25. Fylkesnes K, Musonda R, Sichone M, Ndhlovu Z, Tembo F, Monze M. Declining HIV prevalence and risk behaviours in Zambia during the 1990s: evidence from national surveillance and population-based surveys. AIDS 2001, 15: 907–916.
Keywords:© 2002 Lippincott Williams & Wilkins, Inc.
HIV; surveillance; family planning; bias; fertility; Zimbabwe