Since 2000, UNAIDS and WHO have been promoting second-generation surveillance (SGS) for HIV infection.1,2 Whereas earlier efforts merely measured overall prevalence in various populations, second-generation surveillance aims to monitor HIV and high-risk behavior trends over time, providing data essential for the planning of interventions and the evaluation of their impacts.1,2 In countries where HIV prevalence is greater than 1%, UNAIDS recommends conducting behavioral surveillance surveys in representative samples of the general adult population and high-risk groups [sex workers (SW) and their clients], whereas HIV prevalence is measured in sentinel sites. Data collected on age, sex, socioeconomic, educational, and marital status enable the comparison of behaviors and HIV prevalence in subgroups. However, behavioral measures relying on questionnaires are bias prone, as many individuals tend to downplay undesirable behaviors.1 This social desirability bias is especially problematic among SW who exaggerate the frequency of condom use with their clients, compared with reports made by clients.3 Prevalence of sexually transmitted infections (STI) is used as a biological proxy for risky behavior in such core groups.
A biological marker of unprotected sex that would be quantitative, reproducible, cheap, easy to perform on a large number of samples, and adaptable for use on specimens collected for STI screening would be invaluable in avoiding these biases. It occurred to us that such an assay already exists but has been overlooked by HIV epidemiologists. Forensic scientists are using biological markers of the presence of semen in vaginal samples in circumstances in which it is crucial to determine for how long semen has been present.4 The same technology was used by family planning scientists to measure the frequency of condom leakage.5-7 The prostate-specific antigen (PSA), widely available as a screening assay for prostate cancer, emerged as the best marker of semen presence because of its kinetics: high levels are measured immediately after intercourse, but return to background levels within 24 to 48 hours, as validated in studies during which samples were obtained at intervals after unprotected intercourse or after intravaginal inoculation of varying quantities of sperm.4-7 We sought to determine the performance of PSA measures in vaginal secretions with the use of samples obtained for the detection of STI during a clinical trial on the syndromic management of vaginal discharge (to be reported later).
This study was approved by the ethics committees of each country and the institutional review board of the Centre Hospitalier Universitaire de Sherbrooke, Canada. PSA levels were measured in the vaginal secretions of 508 SW and 658 non-SW women who presented with vaginal discharge in health facilities of Ghana (Accra, Kumasi), Guinea (Conakry), Mali (Bamako), and Togo (Lomé) between January and December 2004. Upon enrollment, a cervical swab was obtained for the detection of Neisseria gonorrhoeae (NG), Chlamydia trachomatis (CT), and Mycoplasma genitalium (MG) using polymerase chain reaction (PCR); vaginal swabs were obtained for detecting Trichomonas vaginalis (TV) by PCR and for Gram staining to look for bacterial vaginosis (Nugent score), yeasts, and spermatozoa.8,9 Swabs to be used for PCR were agitated individually in the Roche Amplicor transport vial (Roche Diagnostics, Branchburg, NJ); the swabs were discarded and the transport medium stored at −20°C within a few hours of sampling. This transport medium contains 1 mL of 0.4% sodium dodecyl sulphate (an ionic detergent) with 0.13% Tris buffer at pH 7.0. For PCR, an equal volume of CT/NG specimen diluent (6 mmol/L MgCl2, <25% nonionic detergent, 0.005% sodium azide) was added to neutralize the detergent. We measured, in March to May 2005, PSA levels in leftover samples of this transport medium neutralized with specimen diluent and kept at 4°C after PCR. HIV serology was performed on capillary blood using 2 enzyme immunoassays (Detect-HIV, Adaltis, Montreal, Quebec, Canada and Genie II HIV-1/HIV-2, Bio-Rad Laboratories, Marnes-la-Coquette, France) with INNO-LIA HIV I/II Score (Innogenetics, Ghent, Belgium) confirmation of discordant results between the first and second enzyme immunoassay.
Detection of PSA was performed using the ADVIA Centaur immunoassay (Bayer HealthCare, Tarrytown, NY). The standard PSA testing procedure allows detection of 0.01 to 100 μg/L; serial dilutions were made for results ≥100 μg/L. Briefly, for each sample, 150 μL of the transport medium had been mixed with 150 μL of Amplicor diluent. Samples were centrifuged for 5 minutes at 13,000 rpm, and analysis was performed on the supernatant. Reproducibility of the assay was monitored by retesting quality controls within each test batch. We used a pool of female sera (PSA = 0 μg/L) and 3 levels of Lyphochek Immunoassay Plus Controls (Bio-Rad, Hercules, CA). Quality control samples were diluted 1:20 with the Amplicor transport medium solution and prepared in the same way as patient samples. This level of dilution avoids matrix effects. Results were expressed as the PSA concentration in the diluted eluted sample (eg, PSA value of 1 μg/L represents 2 μg/L in the initial eluted sample). We evaluated the stability of the initial transport medium and the diluted preparation samples kept at 4°C by retesting 7 months after initial analysis a random sample of 20 specimens: results were similar from both preparations, and decay of PSA was estimated at ≈5% per month.
PSA values ranged between 0.02 and 1953 μg/L. To define background levels considered as indicating a negative assay, we examined PSA levels of 95 non-SW women who claimed having had no sexual intercourse during the previous 3 months. Only 2 of them had spermatozoa seen on the Gram stain, and only 4 (1 each) were infected with NG, CT, MG, or TV. The PSA levels of these 95 women varied between 0.02 and 0.35 μg/L (median, 0.06 μg/L). We thus defined as a positive assay any PSA level ≥0.4 μg/L. Among 58 non-SW women considered to have a positive PSA assay, 13 (22%) had levels between 0.4 and 1.0 μg/L, 15 (26%) between 1.1 and 10.0 μg/L, and 30 (52%) >10.0 μg/L. Among 94 SW with a positive PSA assay, 25 (27%) had levels between 0.4 and 1.0 μg/L, 36 (38%) between 1.1 and 10.0 μg/L, and 33 (35%) >10.0 μg/L. Among 25 SW claiming not to have had any intercourse during the previous 3 months, 7 (28%) had a positive PSA assay (5 with values >1.0 μg/L).
A positive PSA assay was strongly associated with spermatozoa being seen on the Gram stain and with the presence of NG, CT, and MG (Table 1). Among SW, there was no association between PSA and HIV status; however, among non-SW, a positive PSA was more common in the HIV infected. Among non-SW, the prevalence of a positive PSA was significantly lower among women better educated. Ghana was the country with the lowest proportion of PSA-positive women, both among SW and non-SW, and also the country with the lowest prevalence of STI, among SW [71 of 254 (28.0%) of SW had either NG, CT, MG, TV, or any combination thereof, compared with 96 of 253 (37.9%) in other countries, P = 0.02] and non-SW women [18 of 260 (6.9%) of non-SW in Ghana had either NG, CT, MG, TV, or any combination thereof, compared with 62 of 397 (15.6%) elsewhere, P = 0.001].
PSA, present at high concentrations in semen, emerged as the best marker for unprotected intercourse in forensic medicine.4,10 It is relatively stable, follows a regular pattern of postcoital decline, and is not normally present in vaginal secretions, in contrast to acid phosphatase.4,10 The current study showed that PSA could be measured as a marker of unprotected intercourse during surveys aiming to measure the prevalence of HIV and other STI or those primarily interested in sexual behaviors, as long as participants agree to provide a self-inserted vaginal swab (usually acceptable for sex workers). Measuring PSA levels is simple, rapid to perform, relatively inexpensive (US $6 per test), and requires equipment normally present in a clinical biochemistry laboratory. PSA is somewhat unstable to freeze-thaw cycles, which should be avoided.11 False negatives can occur in samples contaminated with detergents,11 but detergents used in our assay did not seem to interfere with detection of positive controls. The sodium dodecyl sulphate in Amplicor transport medium is neutralized by the specimen diluent, which prevents inactivation of the DNA polymerase enzyme used in PCR. It might thus be expected that it would not interfere with immunologic detection of PSA. Although the Amplicor transport medium was relatively stable for PSA, using a saline solution immediately frozen and stored at -20°C until testing would be preferable in future surveys. Given that our samples were diluted 1:1 and that some decay occurred during storage, our 0.4 μg/L cutoff for positivity is in line with the 1.0 μg/L cutoff determined in studies that used saline eluents from swabs obtained after intercourse or inoculation of sperm and kept frozen until testing.5-7
For the purpose of measuring sexual behaviors, the kinetics of PSA in vaginal secretions (which reflect unprotected intercourse during the last 24-48 hours) are more interesting than those of spermatozoa (detectable up to 6-12 days after intercourse) or Y chromosome markers (present for up to 2 weeks).4-7,12-14 This may explain why only half of samples with spermatozoa seen on Gram stain were PSA positive. PSA is especially attractive for measures among SW, who might have on average ≈20 sexual partners per week:3 in such circumstances, an assay remaining positive for several days after a single unprotected intercourse could not reliably estimate the frequency of unprotected intercourse.
In Ghana, the frequency of condom use during transactional sex was measured through surveys of clients, less prone to social desirability biases than SW.3 According to clients, a condom had been used during the last intercourse in 89% of occasions in 2002 and 94% of occasions in 2003.3,15 This is remarkably similar to the current measure that 87% of sex workers had no evidence of semen in their vaginal secretions, specially considering that some SW have nonpaying partners, with whom condoms are used less often.3 Current measures were made a posteriori during a study of women with vaginal discharge. In West Africa, most such women, even SW, have vaginitis rather than cervicitis.9 However, given that sexually transmitted T. vaginalis causes about one fourth of episodes of vaginal discharge,9 our measures probably somewhat underestimated the frequency of condom use. Measuring PSA on a representative sample of all SW from a given community, regardless of the presence of genital complaints, would provide a less biased estimate of condom use.
The associations that we have found between a positive PSA assay and various STI cannot reflect transmission of their etiological agents during the unprotected intercourse that just preceded our measure. They rather point to the fact that women who had an unprotected intercourse in the previous 24 to 48 hours were presumably more likely than others to have had other unprotected intercourses in the preceding weeks or months. A positive PSA may reflect risky behavior in general and not only risky behavior at a very specific moment.
Measures of the frequency of unprotected intercourse are of paramount importance for the evaluation of interventions aiming to modify high-risk behaviors and the identification of subpopulations at high risk for HIV, toward whom more intensive efforts should be targeted. Measuring PSA should be considered in future surveys of sex workers that aim to verify compliance with condom promotion activities to avoid the biases inherent to questionnaires about sexual behavior.
1. UNAIDS/WHO. Guidelines for Second Generation Surveillance
. WHO/CDS/CSR/EDC/2000.5 UNAIDS/00.003E.
2. UNAIDS. Initiating Second Generation HIV Surveillance Systems: Practical Guidelines
3. Coté AM, Sobela F, Dzokoto A, et al. Transactional sex is the driving force in the dynamics of HIV in Accra, Ghana. AIDS
4. Graves HCB, Sensabaugh GF, Blake ET. Postcoital detection of a male specific protein. Application to the investigation of rape. N Engl J Med
5. Macaluso M, Lawson L, Akers R, et al. Prostate-specific antigen in vaginal fluid as a biologic marker of condom failure. Contraception
6. Maculuso M, Lawson ML, Hortin G, et al. Efficacy of the female condom as a barrier to semen during intercourse. Am J Epidemiol
7. Lawson ML, Maculuso M, Bloom A, et al. Objective markers of condom failure. Sex Transm Dis
8. Pépin J, Deslandes S, Khonde N, et al. Low prevalence of cervical infections in women with vaginal discharge in West Africa: implications for syndromic management. Sex Transm Infect
9. Pépin J, Labbé AC, Khonde N, et al. Mycoplasma genitalium
: an organism commonly associated with cervicitis among west African sex workers. Sex Transm Infect
10. Young WW, Bracken AC, Goddard MA, et al. Sexual assault: review of a national model protocol for forensic and medical evaluation. Obstet Gynecol
11. Johnson ED, Kotowski TM. Detection of prostate specific antigen by ELISA. Forensic Chem
12. Willott GM, Allard JE. Spermatozoa-their persistence after sexual intercourse. Forensic Sci Int
13. Zenilman JM, Yuenger J, Galai N, et al. Polymerase chain reaction detection of Y chromosome sequences in vaginal fluid: preliminary studies of a potential biomarker for sexual behaviour. Sex Transm Dis
14. Chomont N, Grésenguet G, Hocini H, et al. Polymerase chain reaction for Y chromosome to detect semen in cervicovaginal fluid: a prerequisite to assess HIV-specific vaginal immunity and HIV genital shedding. AIDS
15. Labbé AC, Dzokoto A, Khonde N et al. A randomised placebo-controlled trial of routine monthly antibiotics against gonococcal and chlamydial infections among female sex workers in Ghana and Bénin: intention-to-treat analysis.15th Biennial Congress of the International Society for Sexually Transmitted Diseases Research; July 2003; Ottawa. Abstract 0524.