HIV serosurveillance traditionally relies on HIV testing specifically performed for the purpose of surveillance. Specimens for testing may be collected through population-based sampling, such as demographic health surveys, AIDS indicator surveys, or behavioral surveillance surveys. In generalized epidemics, in which HIV prevalence in the adult population exceeds 1%, HIV surveillance is most often conducted in antenatal clinics (ANC), where blood is collected as part of antenatal care and screened for syphilis, anemia, or blood group typing. Unlinked anonymous testing (UAT) for HIV without informed consent for the purpose of surveillance is performed on the blood left over after these routine screening tests . This type of surveillance is referred to as (UAT-based) ANC surveillance in this paper. ANC surveillance has been widely used since the late 1980s or early 1990s in countries with generalized epidemics, and has provided a wealth of data for monitoring HIV trends and estimating the burden of HIV disease. UNAIDS updates on the global state of the HIV/AIDS epidemic rely to a large extent on data from UAT-based ANC surveillance.
In addition to HIV testing data generated specifically for surveillance, testing data from HIV prevention programmes are increasingly available. The expansion of HIV prevention and treatment programmes in resource-constrained countries, especially in sub-Saharan Africa, has resulted in a massive increase in HIV testing. In many high-prevalence countries (e.g. Rwanda, Kenya, Botswana) the number of ANC providing prevention of mother-to-child transmission (PMTCT) services already exceeds the number of clinics included in ANC surveillance programmes. Voluntary counseling and testing (VCT) services have also expanded rapidly in recent years.
The increased availability of HIV prevalence data from VCT and PMTCT programmes raises the question of their utility for HIV surveillance. PMTCT data in particular share the same target population, and may thus have the potential to complement or replace UAT-based ANC surveillance. In contrast to UAT, individuals may refuse HIV testing for the PMTCT or self-select themselves for testing at VCT centers, and HIV prevalence in those tested may differ from those not tested. Furthermore, little is known about the comparability of HIV prevalence estimates obtained from PMTCT and UAT-based ANC surveillance programmes. It is thus crucial to understand the potential respective disadvantages and biases of these programmes, which may have significant consequences for data quality and trend observations in surveillance.
At the HIV Surveillance in Resource-constrained Settings Conference held in Addis Ababa, Ethiopia, in January 2004, several presentations focused on the utility of VCT and PMTCT programme data for HIV surveillance. A brief summary of four presentations on this topic, one from Uganda on the use of VCT data, and more from Thailand, Botswana, and Kenya on the utility of PMTCT data and their comparability with ANC surveillance data are presented here. A discussion of the theoretical and practical considerations of the use of such programme data for surveillance follows.
Examples of HIV programme data for surveillance from four countries
Use of voluntary counseling and testing data for surveillance in Uganda
In Uganda, as in most sub-Saharan African countries, HIV surveillance is mainly ANC based, comprising 25 sentinel sites. There are approximately 160 VCT centers countrywide. Data are available on more than 700 000 VCT sessions dating back as far as 1992, and include HIV testing, demographic, and behavioral data.
To assess trends in VCT-based prevalence and their comparability with ANC surveillance data, more than 200 000 records of first-time and non-symptomatic clients from 1992 to 2000 were analysed . The overall HIV prevalence in first-time VCT clients declined from 23% in 1992 to 13% in 2000 (P < 0.001). Prevalence declined both in men (from 17 to 9%, P < 0.001) and in women (31–18%, P < 0.001). The decline in VCT-based HIV prevalence paralleled a decline in ANC-based prevalence, although prevalence among individuals aged 25–29 years in the VCT group was consistently higher than that in the ANC group. VCT data also demonstrated a shift in the age-specific peak prevalence over time. From 1992 to 2000, the peak HIV prevalence in men shifted from the age range 25–29 years to 35–39 years; peak prevalence among women shifted from the age range 35–39 years to 40–44 years. ANC surveillance data often tend to overestimate HIV prevalence in women aged less than 20 years and underestimate prevalence in women 25 years and older. Although VCT clients constitute a self-selected group, VCT programmes are an important additional data source for surveillance. VCT provides data on non-pregnant women who are not subject to this bias, and offers data on men not included in ANC surveillance.
Using prevention of mother-to-child transmission data for HIV surveillance in Thailand
Thailand began UAT-based ANC surveillance in 1989. In 2000, a national programme to provide PMTCT services was introduced in Thailand , and by 2003 PMTCT replaced UAT as a surveillance data source. This situation no longer allows a comparison of UAT with PMTCT data; however, it provides a practical example of PMTCT use for surveillance. Thailand's PMTCT programme is very successful, with 97% of ANC attendees receiving their HIV test result and 95% delivering in hospitals. PMTCT data are reported to the perinatal HIV intervention monitoring system (PHIMS). Monthly routine, aggregated reporting to PHIMS includes data from the antenatal period (P-ANC) and the delivery room (P-DEL). For the purpose of surveillance, data from individual PHIMS records (termed SS for sentinel surveillance) are reported for 1–3 months each year. P-ANC data cover approximately 60% of all pregnancies in Thailand, whereas the P-DEL data system covers approximately 75% of all deliveries. SS data include HIV and syphilis status, age, parity, and race. Table 1 compares the HIV prevalence based on the individual SS data, the aggregated P-ANC data and the aggregated P-DEL data for the years 2001 and 2002. Prevalence is very similar across the three data sources. P-DEL-based prevalence may be an underestimate because it does not include most pregnancies resulting in spontaneous abortions, a risk higher in HIV-positive women. Thailand's use of PMTCT data for surveillance may provide a ‘best-case scenario’, taking advantage of near universal ANC attendance, very good acceptance of HIV testing, and a well-functioning data collection system.
Comparing prevention of mother-to-child transmission and antenatal clinic surveillance data in Botswana
More than 95% of pregnant women in Botswana receive antenatal care in public health facilities. The core element of HIV surveillance is ANC-based UAT. Botswana's PMTCT programme, the first in Africa, has expanded rapidly in recent years covering all ANC. The abundance of PMTCT-related HIV testing data prompted a comparison of this data source with the traditional ANC surveillance system. HIV testing for PMTCT is performed in district laboratories. The form with the HIV result is returned to the clinic, and a copy is routinely forwarded to the Ministry of Health. In 2002, these HIV request forms were available from only 13 of the 22 districts at the Ministry of Health. A further three districts submitted fewer than 50 forms and were excluded from analysis. For the remaining 10 districts, representing 47% of Botswana's female adult population, 4136 forms were available for analysis. These 4136 forms represent only 18% of the nearly 23 000 new ANC clients seen in these 10 districts, as a result of refusal of pre-test counseling by 19% of ANC clients, testing refusal after pre-test counseling by 58% of ANC clients, and gaps in electronic data entry. Figure 1 compares the UAT and PMTCT-based HIV prevalence. The UAT and PMTCT groups were similar in mean age and overall HIV prevalence, and prevalence did not differ significantly by the age group or geographical district.
These results indicate that in Botswana, PMTCT-based prevalence data may be used as a proxy for sentinel surveillance despite low acceptance of HIV testing and incompleteness of data (data missing from 12 districts). HIV testing for PMTCT was recently implemented as part of routine ANC care, a measure that is likely to improve the acceptance of HIV testing. The eventual introduction of HIV rapid testing for PMTCT at ANC may result in the discontinuation of HIV request forms for PMTCT-related testing at district laboratories, and may affect the availability of individual PMTCT testing data for surveillance.
Evaluating the usefulness of prevention of mother-to-child transmission data for HIV surveillance in Kenya
The ANC surveillance system in Kenya is UAT based and includes 42 sentinel sites. The PMTCT programme is rapidly expanding, covering approximately 140 ANC, a growing number that are also used for UAT. Individual PMTCT records were collected at a Coast Province clinic and compared with the clinic's UAT surveillance data. Analysis was restricted to the ANC surveillance sampling period. The sample size in the UAT group was 283 ANC clients; in the PMTCT group, 135 out of 266 (51%) accepted HIV testing and were included in the analysis. Acceptance of PMTCT-related HIV testing was higher in ANC clients aged 19 years or less (68%) than in ANC clients aged 20 years or over (46%). Figure 2 compares HIV prevalence in the UAT and PMTCT groups. The overall HIV prevalence in the PMTCT group (9.5%) was not significantly different from that in the UAT group (7.4%). Both the UAT and PMTCT samples include the same core group of ANC clients; therefore, the relative difference of 22% in HIV prevalence in the PMTCT group and the UAT group probably results from test refusals by women in the PMTCT group. Investigators also reported difficulty accessing the PMTCT logbooks, which are in constant use during clinic hours, and logbook formats lacked standardization across clinics. Furthermore, the legibility and accuracy of routine PMTCT data did not always meet expected standards. These factors must be considered if individual PMTCT records are to be used for surveillance.
The four examples of evaluations or use of programme data illustrate the potential use of VCT and PMTCT data for surveillance purposes. Experience is still very limited, and there is a lack of international recommendations about their use in surveillance. These roles may well be different for VCT and PMTCT data. VCT programmes are a complementary data source providing prevalence data for both sexes, information on the prevalence sex ratio, and can be used to confirm trends observed in ANC surveillance.
PMTCT has the potential to replace traditional ANC surveillance systems, as this programme targets the same population and sometimes even the same sampling points as UAT-based ANC surveillance. A select few countries (e.g. Thailand) have already taken this step, and have discontinued UAT. As prevention and treatment programmes are expanding in high-prevalence countries, the absence of any direct benefit for the tested individual in UAT is becoming more apparent. However, UAT is deemed ethically acceptable , and it is worth remembering its merits, which led to the methodology's broad use for surveillance in generalized epidemics. The World Health Organization's Global Programme on AIDS [1,5] recommended UAT in 1989 because of its absence of selection bias for HIV testing [6,7]. In most ANC surveillance systems, the ANC client is unaware that her left-over blood, taken for routine care, is tested for HIV. When HIV testing is based on informed consent, such as in PMTCT programmes, individuals refusing testing more often have a higher risk of HIV infection than consenters, a phenomenon well documented in many testing populations [8–13]. UAT-based ANC surveillance is relatively low cost as blood is taken for routine care, sampling is facility based, and data collection is minimal. Furthermore, staff need not be trained on counseling, and quality assurance for on-site testing is unnecessary as testing is often performed at a central laboratory. UAT-based ANC surveillance methods are standardized to a degree that allows easy comparisons across countries and time.
An important advantage of UAT data is the availability of individual patient data. These typically include data at a minimum age and parity/gravidity. HIV prevalence in 15–24 year-olds is a key indicator for monitoring the HIV epidemic . However, most PMTCT reporting systems currently rely on aggregated, not individual data. PMTCT programmes typically report aggregate data for monitoring and evaluation purposes, in which a series of summary indicators (number of clients attending, counseled, tested, treated) is deemed sufficient, and costs and staff capacity are seen as factors impeding individual data collection.
The biggest advantage PMTCT data pose for surveillance is their routine availability, anticipated low cost of use in surveillance (costs may be limited to additional steps beyond the routine activities already in place, e.g. data transcription from PMTCT logbooks), and the actual or anticipated wide coverage of this programme. As these programmes expand geographically their representativeness will increase. In contrast to surveillance, PMTCT is offered year-round, and the number of pregnant women tested for PMTCT already exceeds the sample size in UAT-based ANC surveillance. Existing routine reporting mechanisms for monitoring and evaluation purposes could be used for surveillance. The benefits and ethics of PMTCT-based HIV testing, informed consent, counseling, and the provision of antiretroviral medication contrast with their absence in UAT. It is important to note, however, that the presence of UAT does not preclude linked testing, i.e. prevent anyone from getting tested, counseled, and treated. This is well demonstrated at ANC that serve both as UAT SS sites and offer PMTCT programmes. Here ANC clients tested in an unlinked and anonymous fashion are also offered linked testing for the PMTCT.
Should prevention of mother-to-child transmission programme data replace unlinked anonymous testing-based antenatal clinic surveillance?
There is limited experience in using PMTCT data for surveillance. In particular, their accessibility, quality, and comparability with UAT data need further evaluation. The operational aspects of PMTCT programmes are not uniform across countries, and may even differ within the same country. Factors that influence testing behavior, such as fee-based or free testing, individual or group-based pretest counseling ; ‘opt-in’ versus ‘opt-out’ of PMTCT [16,17]; the timing of testing (same or different visit), or the locality of testing (in the presence of the tested person, at site, or off-site) may affect the testing behavior and the resulting prevalence. In ‘opt-in’ testing, the pregnant woman is given information about HIV and is offered an HIV test. She must specifically ‘opt-in’ (give consent) for the test. The healthcare facility routinely offers HIV tests as part of a standard package of care. Women have the right to ‘opt-out’ or refuse. The magnitude of stigma also varies across countries and influences testing behavior. If PMTCT data were used for surveillance, several factors that may make comparisons over time and place more difficult need to be considered. The magnitude and direction of bias in the observed PMTCT-based HIV prevalence is subject to the proportion of ANC clients refusing PMTCT-related testing and the HIV prevalence ratio of refusers to consenters. As these ratios change, so may the observed prevalence, masking true changes in prevalence or suggesting false ones. Surveillance systems that would rely on ANC clients participating in PMTCT may expose themselves to such bias. However, as the number of child-bearing women knowing their status increases with the expansion of PMTCT programmes, differentials in family planning behavior by HIV status may even affect UAT-based surveillance if HIV-positive women tend to become pregnant more or less often than their HIV-negative peers. Cost estimates were not included in the three presentations outlined here, but are needed to evaluate more fully the utility of PMTCT data.
Even if PMTCT programme data may be used for surveillance, UAT data could be useful for calibrating PMTCT data, i.e. to estimate the size of error introduced through testing by consent. This may result in more accurate PMTCT-based prevalence estimates. UAT can also provide data for PMTCT monitoring and evaluation, as its comparison with PMTCT data allows an estimation of the HIV prevalence among test-refusers and thus the number of HIV-positive pregnant women missed for PMTCT. Furthermore, UAT can assist in monitoring the quality of HIV testing of PMTCT programmes. If the PMTCT-related HIV result is recorded on the UAT data collection sheet it can be compared with the HIV result based on UAT at the central laboratory, which also often functions as the reference laboratory. In Guyana, ANC-based HIV surveillance currently relies on both (individual) PMTCT data and UAT (on refusers for PMTCT) at selected ANC (personal communication).
What would be lost by discontinuing unlinked anonymous testing-based antenatal clinic surveillance?
UAT provides individual record data, and is largely unaffected by selection bias, two very important aspects. Its methods are sufficiently standardized across time and place, and may meet higher data quality standards than PMTCT logbook data. In contrast to PMTCT, many UAT-based sentinel sites provide important trend data going back to the early 1990s. Whether the benefits UAT can provide for surveillance and PMTCT are important enough to maintain UAT-based surveillance must be weighed against its cost and lack of direct benefit for the tested individual. PMTCT data may eventually be useful for measuring HIV prevalence and trends over time, once acceptance rates for testing are high enough to ensure minimal bias. However, caution seems prudent before discontinuing UAT-based surveillance, which has provided reliable and valid data for many years, still remains the most important data source for HIV surveillance for generalized epidemics, and does not preclude (additional) testing by informed consent. More data and experiences are needed to obtain a better understanding of the potential implications of discontinuing UAT. It is possible that answers and solutions may differ by country, depending on PMTCT data quality and quantity, test refusal rates, the prevalence ratio of test consenters to refusers, and available resources. Additional research in this area would be facilitated by the addition of PMTCT-related data variables in UAT data collection sheets. Variables on counseling (pre- and post-test), consenting, and testing for PMTCT as part of UAT data collection would aid in describing acceptors and refusers for PMTCT, and the bias caused by test refusers. Such work would allow the direct estimation of HIV prevalence among PMTCT refusers, and would also be highly beneficial for PMTCT monitoring and evaluation. Mechanisms on individual PMTCT data reporting would also benefit from further work.
The thorough analysis of Ugandan VCT data demonstrates the potential this data source may have for surveillance. Although VCT cannot replace ANC-based surveillance systems, it provides additional data for the interpretation of trend data. Furthermore, VCT data provide data on men, non-pregnant women, and may facilitate the estimation of the male: female prevalence ratio. Client behavioral data may provide additional insights for the interpretation of temporal prevalence trends. However, when using VCT data for surveillance, it should be remembered that VCT clients may not adequately represent the general population, that bias may be introduced by self-selection or refusal for testing, and data quality may vary. Data on client age, testing history (to identify first-time testers), and reasons for testing (to exclude symptomatic clients) are important for the interpretation of VCT data. This work from Uganda needs to be repeated elsewhere to assess the generalizability of the findings. Many VCT centers routinely enter their client data in electronic databases; this would facilitate a comparison with ANC surveillance or population-based survey data.
The increasing availability of HIV programme data alone should not be reason enough to decrease the role of UAT data for surveillance. To be used for surveillance, programme data need to satisfy qualitative criteria such as the availability of individual data, data quality, representativeness, and sufficient knowledge about the magnitude and direction of bias as a result of self-selection (VCT) or refusal (PMTCT) for testing. Upon fulfilling such criteria, VCT, PMTCT, and other programmes may play important supplemental roles for surveillance and for triangulating data with ANC surveillance.
1. Anonymous. Guidelines for conducting HIV sentinel serosurveys among pregnant women and other groups
. Geneva: UNAIDS/WHO Working Group on Global HIV/AIDS/STI surveillance; 2003.
2. Baryarama F, Bunnell RE, Ransom RL, Ekwaru JP, Kalule J, Tumuhairwe EB, et al
. Using HIV voluntary counseling and testing
data for monitoring the Ugandan HIV epidemic, 1990–2000. J Acquir Immune Defic Syndr 2004; 37:1180–1186.
3. Anonymous. Evaluation of a regional pilot program to prevent mother–infant HIV transmission – Thailand, 1998–2000
4. de Zulueta P. The ethics of anonymized HIV testing of pregnant women: a reappraisal. J Med Ethics 2000; 26:16–21, discussion 22–16.
5. Chin J. Public health surveillance of AIDS and HIV infections. Bull WHO 1990; 68:529–536.
6. Nicoll A, Gill ON, Peckham CS, Ades AE, Parry J, Mortimer P, et al
. The public health applications of unlinked anonymous seroprevalence monitoring for HIV in the United Kingdom. Int J Epidemiol 2000; 29:1–10.
7. Dondero TJ Jr, Pappaioanou M, Curran JW. Monitoring the levels and trends of HIV infection: the Public Health Service's HIV surveillance program. Public Health Rep 1988; 103:213–220.
8. Behrendt C, Kendig N, Dambita C, Horman J, Lawlor J, Vlahov D. Voluntary testing for human immunodeficiency virus (HIV) in a prison population with a high prevalence of HIV. Am J Epidemiol 1994; 139:918–926.
9. Lee JH II, Branan L, Watkins L, Hoff GL, Bayer WL, Tegtmeier G. Voluntary human immunodeficiency virus testing: acceptance levels and identification of seropositive individuals. South Med J 1991; 84:198–199.
10. Hull HF, Bettinger CJ, Gallaher MM, Keller NM, Wilson J, Mertz GJ. Comparison of HIV-antibody prevalence in patients consenting to and declining HIV-antibody testing in an STD clinic. JAMA 1988; 260:935–938.
11. Krasinski K, Borkowsky W, Bebenroth D, Moore T. Failure of voluntary testing for human immunodeficiency virus to identify infected parturient women in a high-risk population. N Engl J Med 1988; 318:185.
12. Schwarcz SK, Bolan GA, Kellogg TA, Kohn R, Lemp GF. Comparison of voluntary and blinded human immunodeficiency virus type 1 (HIV-1) seroprevalence surveys in a high prevalence sexually transmitted disease clinic population. Am J Epidemiol 1993; 137:600–608.
13. Meda N, Zoundi-Guigui MT, van de Perre P, Alary M, Ouangre A, Cartoux M, et al
. HIV infection among pregnant women in Bobo-Dioulasso, Burkina Faso: comparison of voluntary and blinded seroprevalence estimates. Int J STD AIDS 1999; 10:738–740.
14. United Nations General Assembly, Special Session on HIV/AIDS. Guidelines on construction of core indicators
. Geneva: UNAIDS; 2002.
15. Cartoux M, Sombie I, Van de Perre P, Meda N, Tiendrebeogo S, Dabis F. Evaluation of 2 techniques of HIV pre-test counselling for pregnant women in west Africa. DITRAME Study Group. International Journal of STD & AIDS 1999; 10:199–201.
16. Anonymous. HIV testing among pregnant women – United States and Canada, 1998–2001
17. Jayaraman GC, Preiksaitis JK, Larke B. Mandatory reporting of HIV infection and opt-out prenatal screening for HIV infection: effect on testing rates [Comment]. Can Med Assoc J 2003; 168:679–682.