Secondary Logo

Journal Logo

ARTICLE

The relationship of HIV prevalence in pregnant women to that in women of reproductive age

a validated method for adjustment

Nicoll, Angus1,5; Stephenson, Judith2; Griffioen, Anja2; Cliffe, Susan1; Rogers, Pauline3; Boisson, Eldonna4

Author Information
  • Free

Abstract

Introduction

One of the aims of unlinked HIV seroprevalence studies using specimens from sentinel patient groups is to estimate HIV prevalences in the source populations from which the patients are drawn [1]. These prevalences can then be applied to population denominators to give total numbers of prevalent infections [2–4]. Equally, prevalence may be monitored to detect trends in the source population [5,6]. Pregnant women are considered especially important for this surveillance because they are often taken to be representative of all women, or both men and women, in the child-bearing years [1,2,7]. It is common for pregnant women receiving antenatal care, their newborns and, in some countries, women having terminations of pregnancy to be screened for other conditions, thus providing readily available residual blood specimens for unlinked sero-prevalence studies [8,9].

The application of sentinel seroprevalence to population denominators relies on an assumption that HIV-infected women have an equal probability of becoming, and staying, pregnant as uninfected women [10–12]. That is fertility is independent of HIV status. In this paper, fertility is used in its epidemiological and demographic meaning, that is the number of births a woman experiences [13]. There are a number of reasons why this is unlikely to be the case (Fig. 1) [11]. The fact that HIV prevalences among women choosing termination of pregnancy are almost always higher than those seen in women proceeding to a live birth demonstrates this point [12,14–16]. Some population studies have found HIV-infected women to have lowered fertility rates [17–19], but others have not [12]. To allow for these effects, it is important to develop methods of measuring or estimating the numeric relationship between fertility in HIV-infected and uninfected women. This is equivalent to the ratio of live birth (or termination of pregnancy) rates in HIV-infected women to uninfected women. In countries with low and medium HIV prevalences, the latter can be replaced by whole population rates from vital statistics. This approximates to the ratio between the probabilities that specimens from an HIV-infected and from an uninfected woman will be included in a seroprevalence study, the summary relative inclusion ratio or RIR [10]. This paper describes and validates a simple method of estimating summary RIR data for pregnant women.

Fig. 1
Fig. 1:
. Factors Influencing observed fertility in HIV-infected women relative to fertility in uninfected women.

Methods

The methods used are summarized in Figure 2. Seroprevalence data were derived from unlinked anonymous prevalence monitoring of pregnant women (A in Fig. 2). Fertility data from women known to be HIV infected (B in Fig. 2) were here derived from the United Kingdom Medical Research Council (MRC) Collaborative Study of Women with HIV. This is a voluntary study of women with diagnosed HIV infections attending 14 collaborating centres in the UK. Detailed methods are described elsewhere [20]. The study comprises of 507 women from three major exposure categories (see below) each of which has been found to be representative of these groups in the general population when compared with HIV and AIDS reports to the national surveillance centre database for the UK, held at the Public Health Laboratory Service (PHLS) AIDS and STD Centre at the Communicable Disease Surveillance Centre (CDSC). From study data it was possible to obtain details of each woman's lifetime births and terminations to date and demographic information as to age, date of HIV diagnosis and probable mode of infection. Data were collected at recruitment and at regular study follow-up visits.

Fig. 2
Fig. 2:
. Adjusting HIV prevalence in pregnant women to allow for fertility – summary of methodology. Variables in square brackets and column D will only be necessary in complex HIV epidemics, i.e. if there are important sub-categories (exposure categories) of HIV-infected women and numbers of women who are diagnosed.

These birth and termination rates were than compared with those for HIV-uninfected women using vital statistics as a proxy (C in Fig. 2). The MRC study showed that birth and termination rates differed between women according to their demographic characteristics, HIV risk behaviours and likely route of HIV infection [21]. Different live birth and termination rates were, therefore, derived for three major groups of HIV-infected women who, according to routine surveillance data, are the most numerous in the United Kingdom [22]: black African women (black African), women infected through injecting drugs [injecting drug users (IDU)] and women without either of these risk factors for HIV infection (other categories). The last were almost exclusively white women probably infected through heterosexual sex. The MRC study also demonstrated that HIV-infected women experience significantly different birth and termination rates before and after their infections are diagnosed [21]. Therefore, rates were expressed separately for before and after HIV diagnosis, giving a total of six pairs of live birth and termination rates. These were the fertility rates used in this analysis (Table 1). Analyses were restricted to the age group 20 to 34 years (which was subdivided into the groups 20–24, 25–29 and 30–34) because of the scarcity of data outside those years [21].

Table 1
Table 1:
. Live birth rates and termination rate observed in HIV-infected women prior to and after HIV diagnosis: by main demographic/exposure group.

Unlinked seroprevalences derived from pregnant women in the UK could not be divided into behavioural subgroups as exposure category data were not collected with these unlinked specimens [14]. Hence to derive unified adjustment factors that could be applied to unlinked HIV prevalences, the six fertility rates were weighted and combined according to the proportions of each of the six categories among the HIV-infected population (D in Fig. 2). A distinction was made between Greater London and elsewhere in England and Wales because of the important differences in exposure category composition of the population of HIV-infected pregnant women in these two areas, and the considerably higher prevalence of HIV observed in London [23].

The proportion of HIV-infected population in each of the six categories was derived from the results of national voluntary confidential AIDS case reporting to CDSC; AIDS cases diagnosed in 1992 to 1996 and reported by January 1997 for women aged 15 to under 40 years were used (Table 2) [24]. AIDS case reporting was used as it is considered to be 90% complete [24] and cases can be subdivided according to the woman's likely exposure category, her geographical location of residence (Greater London or elsewhere in England and Wales) and whether or not her HIV infection had been previously diagnosed. If the first positive HIV test was reported less than 9 months before the date of AIDS diagnosis then the woman's infection was considered to have been undiagnosed until she developed significant HIV disease [25]. It was not possible to derive the age-specific proportions that corresponded to the age subgroups within the MRC study because of the small number of AIDS reports in some categories. In the MRC study, there was little variation between age groups in age-specific rates apart from among undiagnosed IDU. However, IDU are a small proportion of the total population and using the overall 20–34 year rate instead of age-specific rates made little difference to the weighted fertility rates. Therefore, the numbers of AIDS cases was applied to the overall rates for 20–34 years.

Table 2
Table 2:
. Distribution of AIDS cases in women aged 15 to under 40 years reported in years 1992–1996 in England and Wales by main group and geographic strata.

The weighted fertility rates (Table 3) were then compared with the observed live birth rates for all women (HIV and non-HIV infected) for 1994 (Fig. 2) [26]. Strictly the latter should have been the live birth rate for HIV-uninfected rather than all women, but in populations of low and medium HIV prevalence the two are almost identical. The ratio of the rate for HIV-infected women and that for all women represents the relative probability that an HIV-infected woman would be included in a representative survey of women having live births compared with an uninfected woman, the summary RIR. The same process was repeated for termination rates (Table 3) [27]. The 95% confidence limits (CI) for the weighted fertility rates were calculated assuming independence between the rates in each exposure category, and fixed weights and 95% CI for the RIR were then calculated (Table 3). These summary RIR can then be applied to the unlinked HIV seroprevalence data from either women proceeding to live births [3] or women experiencing terminations. This produces as final outputs representative population prevalences (Fig. 2).

Table 3
Table 3:
. Calculation of weighted rates for HIV-infected women and relative inclusion ratios (RIR) for live births and terminations of pregnancies.

Validation was undertaken by establishing how well the live birth and termination RIR values calculated using the above method predicted the observed differences in HIV prevalence among samples from women having live births and termination in the unlinked anonymous seroprevalence surveys in London [14].

A sensitivity analysis was undertaken to discover the most important variables influencing the final RIR values (Table 4). For example, it was calculated what the effect on the RIR would be if all HIV-infected women were black African and their infections were undiagnosed. This analysis was confined to live birth rates in London.

Table 4
Table 4:
. Sensitivity analysis: effects on relative inclusion ratios (RIR) for live births of changes in the composition of the population of HIV-infected women and proportion of infections diagnosed in London.

Results

The 505 HIV-infected women in the MRC study for whom data were available had experienced 656 reported pregnancies as of 1 December 1996: 429 live births and 227 terminations of pregnancy. The number of live births and the fertility rates before and after the reported date of diagnosis are shown in Table 1 for each of the three groups of women.

An analysis of AIDS case reports received at CDSC by the end of January 1997 showed that there were 694 cases of AIDS in women aged over 15 and under 40 years reported to the PHLS AIDS and STD centre over the period 1992–1996 (date of report): 470 for women resident in London and 224 for women resident elsewhere in England and Wales (Table 2). Women in the black African group predominated in London, where they were equally divided between those diagnosed and undiagnosed. Outside London, the most numerous group was women in the ‘other categories’ group (Table 2).

The live birth rates and the estimated proportions of HIV-infected women in the population were multiplied together for each of the six groups (Table 3). When these were combined they produced estimated weighted live birth rates for HIV-infected women of 9.27 (95% CI, 8.09–10.47) births per 100 woman-years (age group 20–34 years) for women resident in London and 7.55 (95% CI, 6.71–8.39) for women resident elsewhere in England and Wales. When compared with the fertility rates for all women aged 20–34 years, 8.98 per 100 woman-years in London and 9.48 per 100 woman-years in the rest of England and Wales, this gave summary RIR values of 1.03 and 0.80, respectively (Table 3). The estimated weighted termination rates were 6.11 (95% CI, 4.99–7.21) for women resident in London and 5.06 (95% CI, 4.33–5.87) for women resident elsewhere in England and Wales, giving summary RIR values of 1.79 and 3.24, respectively (Table 3). The RIR figures for termination and live births for women resident in London of 1.79 and 1.03, respectively, suggests that prevalence of HIV in women having terminations of pregnancy would be expected to be 1.74 (1.79/1.03) times higher than in women having live births. Over the five-year period 1992–1996 in eight hospitals undertaking surveillance among both women choosing termination of pregnancy and receiving antenatal care, HIV prevalence was 0.62% (214 of 34 351) and 0.30% (463 of 154 740), respectively, giving a ratio of 2.07, which is close to that predicted (1.74).

The summary RIR for live births and terminations in London would be affected by any systematic overestimation or underestimation of birth or termination rates in HIV-infected women. Any effect in one of the six groups would on its own produce a smaller effect, which would suggest that RIR values would be relatively insensitive to such changes in fertility within groups. However, because of the substantial intergroup differences in fertility, the RIR values would be more sensitive to any changes in the relative proportions of the groups. Similarly, because of the effects of diagnosis on fertility and birth and termination rates [21,28], there would also be significant changes in the RIR if the proportions of pregnant women who had their infections diagnosed changed substantially. Such effects are shown in Table 4, where RIR has been calculated for each of the six exposure categories. These analyses show that in plausible circumstances the RIR could vary three-fold, from 0.47 to 1.56. For example, if all HIV-infected women were within the black African category and their infections were all undiag-nosed then the overall RIR would rise to 1.41 and with it the observed HIV prevalence among women experiencing live births. Such a change would take place without any alteration in the true population prevalence. Alternatively, if most infections were diagnosed, the RIR would decline and with it the observed prevalence.

Discussion

This analysis has demonstrated that fertility effects are important when applying seroprevalence results from pregnant women. A method has been devised for adjusting for these effects and it is now used routinely in England and Wales when numbers of prevalent infections are being estimated [3]. The population denominators to which pregnant women prevalence are applied can be very large, so even a modest difference from unity can be significant [11]. For example, for England and Wales with a population denominator for ‘low-risk heterosexuals’ of around 28 million and an HIV prevalence of 0.012% [2], there would be an estimated 3360 prevalent infections. However, if we applied the observed RIR of 1.03 for London and 0.80 for elsewhere in England and Wales (Table 3), this estimate would change to 3262 or 4200, respectively, a difference of 940, or 28% of the total. Application of higher prevalences to larger population denominators, as is undertaken for international calculations [4], would result in much greater absolute changes.

The approach that has been taken here has been simplistic and no doubt could be improved. The use of AIDS reports for the weighting is not optimal as it represents a cumulative and historical view of what data should be used, namely the distribution of prevalent HIV infections of women in their child-bearing years. Other data were explored, such as laboratory reports of diagnosed HIV infections, but the results were little different from using AIDS reports. These have the advantage of being available in most countries. The RIR statistic used was only a summary, in reality it is made up of many individual components that will vary with age group and time since infection [10]. Other more complex analyses could be attempted, but there were insufficient data to allow this. We believe the approach shown has the value of being simple and applicable elsewhere, notably in developing countries where such corrections will be particularly important because of high prevalences, which mean that even small alterations in RIR could have gross effects on estimated numbers of prevalent infection [4]. Aside from seroprevalence data, the analysis relies principally on measures of fertility rates in HIV-infected and uninfected women (Fig. 2). These can be derived from observational or cohort studies, from population studies or from surveys of women attending hospital and receiving diagnostic testing [17,18,29]. It is important to ensure that the women are likely to be representative of the population from whom seroprevalence data are drawn. For example, pregnant women themselves would not be suitable as a source of fertility data as they would exclude infertile women. In the analysis for the UK, adjustments and weightings were needed (D in Fig. 2) because of the complexity of the distribution of infection across three different exposure categories, and between women aware and unaware that they were infected. In many developing countries, such complexities would not be necessary because the overwhelming majority of infection is acquired through one exposure (sex between men and women) and most women are unaware of their HIV status. In these settings, it may be possible to generalize about summary RIR across a large region, such as sub-Saharan Africa [30].

The relationship between HIV infection and fertility is complex (Fig. 1) and it should not be assumed that seroprevalence among pregnant women will universally underestimate, or overestimate, population seroprevalence [17]. The data here indicate that in the UK either could occur under plausible conditions (Table 4). The primary aim of collecting sentinel unlinked prevalence data has been to monitor trends in HIV prevalence in the source population [5,6,31]. It has been suggested that differential fertility among HIV-infected women will make HIV prevalence data difficult to interpret for trend analysis [32]. However, the method described here shows how to allow for fertility effects. The results indicate that with a mature epidemic and if there are no alterations in the proportions of infections that are diagnosed there is no prima facie reason to expect unlinked monitoring not to reflect trends in underlying population prevalence among women. Equally, however, if there are changes in the proportion diagnosed or in the mix of exposure categories of prevalent HIV-infected women then RIR values are likely to change and with them observed seroprevalence. For example, intended campaigns to enhance levels of diagnosis among pregnant women in London would produce an apparent fall in RIR and thus the observed antenatal prevalence would decline even if there was no actual change in prevalence (Table 4). This indicates the importance of routinely monitoring fertility in HIV-infected women as an adjunct to serosurveillance in pregnant women wherever that is taking place on an on-going basis.

Acknowledgements

We are grateful to the investigators for the MRC study and their staff for provision of data. Analysis on AIDS cases were provided by Mr Neil McDonald and Mrs Janet Mortimer of the HIV and AIDS Reporting Sections at CDSC. Useful comments on earlier drafts of this paper were provided by Drs Tony Ades, John Karon and Paddy Farrington and the manuscript was prepared by Mrs Virginia Walker.

References

1. Colebunders RL, Heyward WL: Surveillance of AIDS and HIV infection: opportunities and challenges. Health Policy 1990, 15:1–11.
2. Giesecke J, Johnson A, Hawkins A, et al.: An estimate of the prevalence of human immunodeficiency virus infection in England and Wales by using a direct method. J R Statist Soc 1994, 157:89–103.
3. Petruckevitch A, Nicoll A, Johnson AM, Bennett D: Direct estimates of prevalent HIV infection in adults in England and Wales for 1991 and 1993. Genitourin Med 1997, 73:348–354.
4. Burton A, Mertens TE: Provisional country estimates of prevalent adult human immunodeficiency virus infections as of end 1994: a description of the methods. Int J Epidemiol 1998, 27:101–107.
5. Gill ON, Adler MW, Day NE: Monitoring the prevalence of HIV: foundations for a programme of unlinked anonymous testing in England and Wales. BMJ 1989, 299:1295–1298.
6. Chin J: Public health surveillance of AIDS and HIV infection. Bull World Health Org 1990, 68:592–536.
7. Karon JM, Rosenberg PS, McQuillan G, Khare M, Gwinn M, Petersen LR: Prevalence of HIV infection in the US, 1984 to 1992. JAMA 1996, 276:126–131.
8. US Bureau of the Census: Recent HIV Seroprevalence Levels By Country. 21. 1996. Washington: US, Health Studies Branch, International Programs Center, Population Division.
9. European Centre for the Epidemiological Monitoring of AIDS: Unlinked anonymous testing among pregnant women and neonates in Europe. AIDS Surveillance in the European Community Quarterly Report 1991, December:38a-45a.
10. Ades AE: Serial HIV seroprevalence surveys: interpretation, design, and role in HIV/AIDS prediction. J Acquir Immune Defic Syndr 1995, 9:490–499.
11. Boisson E, Nicoll A, Zaba B, Rodrigues LC: Interpreting HIV seroprevalence data from pregnant women. J Acquir Immune Defic Syndr 1996, 13:434–439.
12. Karon JM: Methods for estimating HIV prevalence in the United States. Atlanta, Georgia: US Department of Health and Human Services; 1996: D821, 1–49.
13. Last JM: A Dictionary of Epidemiology 2nd edn. Oxford: Oxford University Press; 1988.
14. Unlinked Anonymous Surveys Steering Group: Unlinked Anonymous HIV Prevalence Monitoring Programme: England and Wales, data to the end of 1996. London: Department of Health, Public Health Laboratory Service, Institute of Child Health (London); 1997.
15. Goldberg DJ, MacKinnon H, Smith R, et al.: Prevalence of HIV among childbearing women and women having termination of pregnancy: multidisciplinary steering group study. BMJ 1992, 304:1082–1085.
16. Couturier E, Brossard Y, Six C, et al.: Prevalence de l'infection V. I. H. chez les femmes enceintes de la region parisienne. Une enquete anonyme non correlee: Prevagest. Bull Epidemiol Heb 1993, 38:175–177.
17. Gray RH, Wawer MJ, Serwadda D, et al.: Population-based study of fertility in women with HIV-1 infection in Uganda. Lancet 1998, 351:98–103.
18. Carpenter LM, Nakiyingi J, Ruberantwari A, et al.: Estimates of the impact of HIV-1 infection on fertility in rural Ugandan population cohort. Health Transit Rev 1997, 7:113–126.
19. 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.
20. The Study Group for the MRC Collaborative Study of HIV Infection in Women: Ethnic differences in women with HIV infection in Britain and Ireland. AIDS 1996, 10:89–93.
21. Stephenson JM, Griffioen A and the Study Group for the Medical Research Council Collaborative Study of Women with HIV: The effect of HIV diagnosis on reproductive experience. AIDS 1996, 10:1683–1687.
22. Hughes G: An overview of the HIV and AIDS epidemic in the United Kingdom. CDR Rev 1997, 7:121–122.
23. Nicoll, A, McGarrigle, C, Brady, et al.: Epidemiology and detection of HIV-1 among pregnant women in the United Kingdom: results from national surveillance systems 1988–1995. BMJ 1998, 316:253–258.
24. Mortimer JY, Evans BG, Goldberg DJ: The surveillance of HIV infection and AIDS in the United Kingdom. CDR Rev 1997, 7:118–120.
25. Report of a Working Group (Chairman Day NE) convened by the Director of the Public Health Laboratory Service on behalf of the Chief Medical Officers: The incidence and prevalence of AIDS and prevalence of other severe HIV disease in England and Wales for 1995 to 1999: projections using data to the end of 1994. Communicable Dis Rep Rev 1996, 6:R1-R24.
26. Office of National Statistics: Birth Statistics, 1994. Series FM1 No. 23. London: HMSO; 1997.
27. Office of National Statistics: Abortion statistics. Legal Abortions carried out under the 1967 Abortion Act in England and Wales, 1994. Series AB No. 21. London: HMSO; 1997.
28. Thackway SV, Furner V, Mijch A, et al.: Fertility and reproductive choice in women with HIV-1 infection. AIDS 1997, 11:663–667.
29. Ryder RW, Batter VL, Nsuami M, et al.: Fertility rates in 238 HIV-1-seropositive women in Zaire followed for 3 years post-partum. AIDS 1991, 5:1521–1527.
30. Zaba B, Gregson S: Measuring the impact of HIV on fertility in Africa. AIDS 1998, 12 (suppl.):S41–S50.
31. World Health Organization: Unlinked anonymous screening for the public health surveillance of HIV infections. Proposed international guidelines. In Global Programme on AIDS 1989, GPA/SFI/89, Vol 3. Geneva, Switzerland: World Health Organization; 1989:1–7.
32. 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.
Keywords:

Fertility; HIV prevalence; pregnant women; surveillance

© 1998 Lippincott Williams & Wilkins, Inc.