HIV Sentinel Surveillance Among Women Seeking Elective Pregnancy Termination, San Francisco

Objective: The objective of this study was to measure HIV prevalence, HIV incidence, and risk factors for infection among women seeking elective pregnancy termination in San Francisco.

Study: The authors conducted a cross-sectional survey comprising a consecutive sample of women seeking elective pregnancy termination in San Francisco's county hospital from August 2002 to July 2003. Demographic and risk behavior information was abstracted from routine clinic records. HIV testing was conducted on blood specimens collected for other purposes after removing identifying information.

Results: Based on 11 HIV-positives among 1,992 tested, HIV prevalence among women seeking pregnancy termination was 0.55% (95% confidence interval [CI], 0.28-0.99). One recent HIV seroconversion was detected for an annual incidence of 0.11% per year (95% CI, 0.23-0.88). In addition, risk factors significantly associated with HIV infection included sex with a known HIV-positive man, history of an abnormal Pap smear, history of genital herpes infection, history of trichomoniasis, and age 25 to 29 years.

Conclusions: Women electing pregnancy termination can serve as a sentinel population to track trends in the HIV epidemic. However, barriers remain to wider implementation of the approach as a surveillance tool.

MONITORING HIV PREVALENCE AMONG women attending prenatal clinics forms the backbone of HIV surveillance in most countries worldwide.¹ The most common method to measure prevalence, referred to as unlinked anonymous testing, entails HIV antibody testing of blood samples drawn for other purposes such as syphilis screening after they have been permanently stripped of identifying information.^2,3 Unlinked anonymous testing reduces participation bias (i.e., women who elect to test for HIV may differ from those who do not). These studies, known in the United States as the Surveys of Childbearing Women, provided a large, unbiased sample of all women delivering infants and an approximation of HIV prevalence in the general heterosexual population. In California, unlinked anonymous surveys were conducted on neonatal dried blood spots collected for metabolic disorder screening from 1989 to 1995. Data were used to track the course of the epidemic, target and evaluate HIV prevention programs, and project HIV-related health care needs for women and children.

However, the surveys were stopped in the mid-1990s. Among key objections was the argument that the priority should be to identify HIV-infected mothers and provide treatment to prevent mother-to-child transmission, particularly after such treatment was proven to be effective.⁴ Emphasis shifted toward universal HIV testing of all pregnant women in prenatal care. Another problem with the Surveys of Childbearing Women was the small number of HIV-positive women detected, making estimates imprecise and interpretation of data difficult. In areas where HIV is concentrated in a few populations such as in San Francisco, childbearing women may not represent those at highest risk for infection. Moreover, many women who know they are HIV-infected may elect pregnancy termination. Thus, prevalence based on estimates for women who deliver may underestimate the true prevalence of HIV in the wider population. For all their shortcomings, the discontinuation of the Surveys of Childbearing Women in the United States left a paucity of recent data on HIV prevalence trends among women.

HIV sentinel surveillance among women seeking pregnancy termination presents some advantages over childbearing women. Several studies show higher HIV prevalence among women terminating pregnancy than those who deliver,^5-8 suggesting inclusion of women at higher risk for infection and therefore a better sentinel population in concentrated epidemics. Moreover, the objection that unlinked anonymous HIV testing misses opportunities to prevent mother-to-child transmission does not directly pertain to cases of women seeking abortions.

Materials and Methods

We conducted an unlinked anonymous HIV sentinel surveillance survey of women electing pregnancy termination at the San Francisco General Hospital, the only public hospital for the county, using leftover sera collected for syphilis screening or blood typing from August 2002 to July 2003. The outpatient abortion clinic in the hospital serves as the referral facility for medically complicated patients and for indigent patients in the late second trimester in Northern California. Data on demographic characteristics and risk factors were abstracted from existing medical records; no additional questions were asked for the purposes of this study. All personally identifying information, including medical record numbers, was unlinked from specimens and data before HIV testing. Voluntary confidential testing for HIV was available on site.

Specimens were tested for HIV antibodies using the Vironostika HIV-1 Microelisa (BioMérieux, Durham, NC) and positive specimens were confirmed by the Fluorognost HIV-1 Immunofluorescence Assay (Waldhiem Pharmazeutika, Vienna, Austria). We additionally conducted the serologic testing algorithm for recent HIV seroconversion (STARHS; Vironostika; BioMérieux) on HIV-positive specimens to detect recent infections (seroconverting within the last 170 days; 95% confidence interval [CI], 145-200) following standard protocols.^9,10

The study was approved by the Committee for Human Research of the University of California, San Francisco. Statistical comparisons are made using the chi-squared test or Fisher exact test when expected counts were less than five.

Results

A total of 2,025 women were seen at the clinic during the study period for elective pregnancy termination. Specimens from 1,992 women (98.4%) were tested for HIV. Of the 33 not tested, 19 specimen tubes cracked in storage, one had insufficient serum, and for 13, the laboratory did not receive a specimen for unknown reasons. The majority of women in the study was under 25 years (53.8%) and had public medical insurance (85.0%); the largest proportion (33.2%) was black; 2.7% were homeless. Excluding five indeterminate test results, 11 women were confirmed HIV-positive for an overall prevalence of 0.55% (95% CI, 0.28-0.99). Among the 11 HIV-positive women, one recent infection was detected by STARHS, yielding an annual HIV incidence estimate of 0.11% per year (95% CI, <0.001-0.81). Of note, six of the 11 HIV-positive women gave a history of a positive HIV test. Additionally, two of the 13 women whose specimens were missing also gave a history of a prior HIV-positive test result. If these two are added to the overall calculation, HIV prevalence would be 0.65% (95% CI, 0.34-1.1).

Table 1 shows HIV prevalence by demographic characteristics and potential risk factors for HIV infection. Variables significantly associated with HIV infection were age 25 to 29 years, reporting sex with a known HIV-positive man, history of genital herpes infection, history of trichomoniasis, and history of an abnormal Pap smear. Of note, no HIV infections were detected among women with a history of injection drug use, sex with an injection drug user, sex with a man who has sex with men, or among teenage women seeking abortion.

Table 1

Discussion

Our study demonstrates the feasibility of monitoring HIV prevalence and, with STARHS, HIV incidence among women seeking pregnancy termination in a concentrated epidemic setting. We were also able to identify several significant risk factors for HIV infection in this population of women. Additional strengths of using the clinic as a sentinel site is its position as one of the largest providers of abortion services in Northern California and the only public hospital providing such services in the late second trimester to indigent women in the area.

HIV prevalence of 0.55% among women in our study in 2002 to 2003 was higher than among childbearing women from 1989 to 1995. The median prevalence in those surveys was 0.21%,¹¹ which was outside our confidence limits of 0.28% to 0.99%. We believe the higher HIV prevalence among women seeking pregnancy termination compared with women in the surveys of childbearing several years ago primarily reflects elevated risk in the clinic population rather than a temporal increase in HIV transmission among women since the mid-1990s.^5-8 For additional comparison, a population-based household survey of young women in low-income neighborhoods of the San Francisco Bay Area found HIV prevalence at 0.3% (95% CI, 0.1%-0.4%) in 1996 to 1998 to be more consistent with the Surveys of Childbearing Women.¹²

Although feasible, there remain obstacles to widely implementing the use of women seeking elective pregnancy termination as sentinel populations for tracking the HIV epidemic. Some may find unlinked anonymous HIV testing of women seeking abortions objectionable for reasons similar to those raised for the Surveys of Childbearing Women, namely that the priority should be placed on counseling and testing women at risk for infection. Although there is no immediate missed opportunity for preventing mother-to-child transmission, some HIV-positive women seeking abortions will choose to deliver children in subsequent pregnancies. Although the uptake of universal testing of women seeking prenatal care would identify the majority of them, seeking abortion is another opportunity to identify women with HIV infection and refer them to care. We agree that the promotion of voluntary HIV testing should remain a priority for all women at risk for infection; however, unlinked anonymous testing for surveillance purposes is not at odds with this goal. We recognize, nonetheless, that the issue of abortion is politically charged in the United States. Any attempt to integrate data collected from abortion clinics into a nationally supported surveillance system is likely to be met with opposition.

We also found some logistic barriers to the approach. Like with the Survey of Childbearing Women, we were faced with difficulties interpreting small numbers of HIV infections. Errors for a few HIV-positive cases can substantially alter findings. For example, two of 13 women without unlinked specimens had a history of a prior HIV-positive test. We also had a relatively large number of indeterminate results, a finding seen in other studies of pregnant women.^13,14 Moreover, the accuracy and completeness of behavioral information was uncertain. Finding no infections among women with a history of injection drug use is a case in point.

Although no epidemiologic data are without potential biases and limitations, there may be few efficient alternatives to widely tracking HIV infection among women in countries with concentrated epidemics. Along with data on sexually transmitted diseases and risk behavior, HIV prevalence and incidence among women electing pregnancy termination can serve as one component of a comprehensive HIV/AIDS surveillance system and provide relevant data to help fill current gaps in our knowledge of the epidemic among women in the United States.