The prevalence of Human Immunodeficiency Virus (HIV) is a measure of disease burden which is important in evaluating HIV prevention programmes and in designing new preventive strategies. Most estimates of HIV prevalence in developing countries have been based on sentinel surveillance of pregnant women 1. HIV prevalence among pregnant women in antenatal clinics (ANC) closely approximates HIV prevalence in the adult population. ANC data is sometimes used to estimates HIV prevalence in general female population, instead of the general adult population (both males and females). This assumption is, however, unfounded 2. Women who adopt abstinence or correct and consistent condom use in response to HIV prevention programmes are less likely to be pregnant and equally less likely to contract HIV infection. As such, sentinel surveillance of pregnant women would over-estimate HIV prevalence in the general female population. On the other hand, there is some evidence that HIV-1 could impair fertility 34. If HIV infected women are less likely to be pregnant because of impairment of their fertility by the HIV virus, the use of sentinel surveillance would under-estimate the HIV prevalence.

Research comparing ANC estimates of HIV prevalence with HIV prevalence from population-based studies has shown that ANC figures closely estimate the actual population prevalence 567. However, the findings have not been consistent 8. Saphonn et al (2002) reported that although the prevalence of HIV in ANC was good for monitoring trends, it over-estimated HIV prevalence in rural Cambodia 9. On the contrary, Changalucha et al (2002) reported that the prevalence of HIV from ANC underestimated community HIV prevalence in Tanzania 10. Grassly, et al (2006) reviewed studies that have been carried out in different countries and compared estimates from ANC sentinel surveillance with population based estimates 11. They found that in 12 out of 15 countries, HIV prevalence from ANC surveillance closely estimated the actual population prevalence. However, prevalence from ANC over-estimated community prevalence in Kenya and Rwanda, but underestimated the community prevalence in Zambia. Assuming that prevalence from ANC surveillance is representative of pregnant women, and that errors resulting from laboratory testing are minimal, one major explanation of inaccuracies in the estimates is the differences in the fertility rates between HIV positive and negative women. The accuracy of HIV estimates from ANC also depends on the contraceptive prevalence and ANC coverage.

Improved methods for estimating HIV prevalence such as population-based surveys can provide useful information on HIV prevalence levels and distribution, and improve HIV estimates at national and regional levels 12. However, surveys may be costly especially when there are frequent and nation-wide.

Research on ANC sentinel surveillance has mainly been focused on the role of behavioural data in ANC surveillance 13, uses of prevalence data 14, coverage 15, assessing the trends of HIV prevalence 161718, comparison between HIV prevalence from ANC and community prevalence 910 and methods of adjustment of estimates from ANC surveillance 1119. In addition some research has addressed the differences in fertility according to HIV status 220. Glynn et al (2000) reported that the birth interval was reduced in HIV positive multiparous women compared to HIV-negative multiparous women in three African cities, namely Yaounde (Cameroon), Kisumu (Kenya) and Ndola (Zambia) 20.

Using sentinel data of pregnant women, the prevalence of HIV in Cameroon rose from 0.5% in 1987 to 10.8% in 2000 and then dropped to 7.3% in 2002 (Ministry of Public Health, 2001 and 2003) 21. A population-based survey conducted in 2004 estimated the overall adult HIV prevalence in Cameroon to be 5.5% 22. It is not clear whether this represents a true decrease in the prevalence of HIV or an artifact due to the differences in the two methods of data collection. Knowledge about the accuracy of sentinel prevalence as a proxy for national HIV prevalence is indispensable especially in a country facing rapid changes in the HIV prevalence. Differential prevalence of HIV in pregnant and non-pregnant women has been reported to be significantly associated with age, marital status, parity, schooling, and contraceptive use 23.

The purpose of our study was to compare the fertility rates of HIV-infected and HIV-uninfected women using population-based data in order to make recommendations on the appropriate adjustments when using sentinel data in designing and evaluating HIV prevention programmes in Cameroon.

This study compared fertility rates in HIV-infected and uninfected women of reproductive age using population-based data. Our findings indicate that fertility rates are lower in HIV infected women compared to uninfected women in Cameroon.

Our finding of low fertility in HIV positive women is consistent with reports from other authors 34. We found low fertility rates in HIV positive women of all age groups but for urban HIV-infected teenagers (that is, 15–19 year olds) who had a higher fertility rate than their corresponding HIV negative counterparts in cities. We could not find any suitable explanation to this finding. However, since the teenagers in rural areas were similar to other adult women in terms of their RIR, the cause is unlikely to be biological. We therefore attributed the high fertility in urban HIV-infected teenagers to differences in sexual and reproductive attitudes and practices. Monitoring of HIV epidemic via antenatal sentinel surveillance requires regular adjustments for many factors such as the differences in distribution of HIV across different age groups and different sub-populations, and differences in fertility rates between the infected and the uninfected women 26. Adjustment for fertility rates is an important factor because modest changes in fertility can have profound effects on the validity of estimates from pregnant women 27. Nicoll et al suggested that adjusting for the differences in fertility rates was sufficient in countries with overwhelming epidemics where the infection was transmitted through one source (heterosexual route) and many people are unaware of their HIV status 2. The authors argued that it might be possible to use summary RIR in wider geographical area such as Sub-Saharan Africa. Our study reports the summary RIR for Cameroon. Since the study population was a representative sample of the all Cameroon women of reproductive age, our RIR is self-weighted for regional and age differences. Eighty three percent (83%) of Cameroonian women attend at least one antenatal visit during pregnancy 22. However, the differences in utilization of antenatal services by HIV-positive and HIV-negative women could introduce bias in the ANC-based prevalence estimates.

Our study used fertility rates (number of births per 1000 woman years of exposure over 36 months) instead of live birth rates (number of live births per 100 woman years of exposure) as reported by Nicoll et al 2, because we consider that all pregnant women are equally likely to be tested for HIV infection irrespective of the birth outcome (dead or live births). Desgrées du Loû et al used parity as a proxy for fertility rate 28. Parity is a cumulative measure of fertility and is insensitive to recent changes in fertility. In Zambia, ANC-based HIV estimates were found to substantially underestimate declines in HIV prevalence in the general population 29. This was partly explained by changes in fertility-related behaviours among young women.

We found a summary RIR of 0.69 which is lower than that reported by other authors 42830. Nicoll et al reported a summary RIR of 1.03 for London and 0.80 for women else where in England and Wales 2. The differences may be explained by socio-cultural differences of the two populations. Desgrées du Loû et al in Ivory Coast reported a summary RIR of 0.84 which is higher than our summary RIR 28. This may be explained by the fact they used parity instead of fertility rates and collected data from antenatal clinics which may not accurately represent the whole population. Gray et al 4 is a population-based study in Uganda reported an unadjusted RIR of 0.78. Similarly, in another population-based study Terceira et al reported an unadjusted RIR of 0.78 in rural Zimbabwe 30. The RIRs of these two studies are closer to that reported in our study and this could be explained by the fact that they all used population-based data for analysis.

The application of HIV seroprevalence from pregnant women to the whole female population needs adjustment for the differences in fertility rates among HIV-infected and HIV-uninfected women. Following earlier publications, fertility rate is now taken into consideration by the Joint United Nations Programme on HIV/AIDS (UNAIDS) when estimating and projecting HIV prevalences 31.

In conclusion, fertility rates are lower in HIV infected women compared to uninfected sexually active women in Cameroon. The findings of this study support the use of summary RIR for the adjustment of HIV prevalence (among adult female population) obtained from sentinel surveillance in antenatal clinics. Continuous monitoring of the fertility rates in HIV positive and HIV negative women should be an adjunct to HIV serosurveillance because fertility rates are not static, but change over time.

The authors declare that they have no competing interests.

EJ Kongnyuy conceived the study, and analyzed the data, and had primary responsibility for writing this paper. CS Wiysonge assisted in the critical revision of the manuscript for important intellectual content.