RESEARCH THAT EVALUATES BEHAVIORAL INTERVENTIONS or barrier methods for the prevention of pregnancy or sexually transmitted infection (STI)/HIV often relies on self-reports of sexual activity. A work group sponsored by the U.S. National Institutes of Health to examine the validity of these measures concluded that “most people will provide what they believe are truthful responses” if specific data collection conditions are met.1 However, the group did not provide empirical evidence or references to support their conclusion. In contrast, other research has questioned the validity of self-reported sex and condom use data.2
A biologic marker of unprotected sex, detectable in vaginal fluids, could serve to support or supplant the use of self-reported data. Prostate-specific antigen (PSA), a protein produced by the prostate gland and secreted into the urethra during ejaculation, can be detected in vaginal fluid samples with high sensitivity using solid-phase enzyme-linked immunosorbent assay (ELISA).3,4 PSA clears from vaginal fluid within 48 hours with a mean clearance time of approximately 20 to 27 hours.4–6 We evaluated the validity of self-reports of recent unprotected sex among female sex workers (SWs) participating in a condom promotion study by comparing their reported sexual activity and condom use with PSA detection in vaginal samples.
Materials and Methods
The current work was part of a larger randomized, controlled trial of 1,000 female SWs in Madagascar designed to compare the effect of supplementing community-based (male and female) condom promotion with clinic-based counseling on both biologic and behavioral outcomes at 6-, 12-, and 18-month follow-up visits.7 We recruited the subset of SWs who presented for their last study visit from December 2002 through May 2003. We gave no additional compensation for participation beyond that given for the larger study. Participants gave verbal consent for the current research, which was approved by the ethical review boards of Family Health International and the Madagascar Ministry of Health.
The female interviewers from the larger study asked participants 8 additional dichotomous questions. Participants were asked whether they had sex with any clients, partners, or boyfriends in the last 24 hours and whether any of the acts during this timeframe occurred without male or female condom use. The questions were repeated for the past 48-hour, 7-day, and 14-day periods. We did not specify vaginal sex because formative research showed that SWs did not include nonvaginal sex acts in their answers.
Immediately before collecting samples for the larger study, clinicians collected samples for PSA testing by inserting a cotton-tipped swab into the posterior fornix and rotating 4 times. Although PSA is stable at ambient temperatures for extended periods,8 we sealed the swabs after several hours of air-drying and froze them in individual transport tubes until data collection ended. Specimens were shipped on dry ice to the research laboratory at the University of North Carolina for processing.
We detected PSA in vaginal swab eluates using the Abbot IMx microparticle enzyme immunoassay (Abbott Laboratories, Abbott Park, IL). We inserted each vaginal swab into a 1.5-mL microcentrifuge tube containing 1 mL phosphate-buffered saline and allowed the contents to elute at room temperature for 5 to 15 minutes. Working with individual specimens apart from the others to avoid contamination, we vigorously rotated and pressed the swabs against the side of the tube to maximize sample recovery. We centrifuged capped tubes at 13,000× g for 5 minutes and pipetted 250 μL of the resulting supernatant into an IMx reaction cell. In this assay, PSA concentrations in specimens are automatically calculated from a standard curve of the output from the IMx instrument plotted as a function of the PSA concentration in 6 manufacturers’ standards ranging from 0 to 100 ng/mL. Linear regression analysis of 14 independent standard curves performed over the period of testing for this study was highly reproducible; the assay results were linear up to 100 ng PSA/mL with standard curve R2 values consistently >0.98. We did not dilute and retest specimens to quantify concentrations above 100 ng PSA/mL. Because specimens are diluted as a consequence of the elution process, PSA concentrations measured in vaginal swab eluates are lower than the concentrations of PSA in vaginal fluid. Because the volume of vaginal fluid collected on a swab is uncertain and variable, PSA concentrations may be considered semiquantitative. We assessed the reproducibility of the assay with vaginal swab eluates by retesting approximately 10% of specimens (n = 35) and obtained similar results (correlation coefficient of 0.97).
Using the same assay, Macaluso et al described the clearance kinetics of PSA in vaginal swabs from women after insemination.4 The authors eluted vaginal swabs into 3 mL of buffer and identified a concentration of 1 ng PSA/mL eluate as the threshold for a positive PSA result. As a marker for exposure to semen within the last 48 hours, the specificity of the test using this cut point for PSA concentration in vaginal swab eluates is estimated at 97%.4 Because our specimens were eluted into one third of the volume used by Macaluso et al, we defined a positive result as >3 ng PSA/mL vaginal swab eluate.
Using SAS 8.01 for data analysis, we calculated McNemar’s tests for concordance between self-reports and biomarker detection. Discordance could occur if women reported 1) no unprotected sex but tested positive for PSA or 2) unprotected sex but tested negative for PSA. We focused on the former because the latter could be the result of reasons other than misreporting. Macaluso and colleagues conducted a controlled clinical trial in which they inseminated 40 women with 1 mL of their partner’s semen and then collected vaginal fluid specimens at periodic intervals to test for PSA.4 Using a cut point of 1 ng/mL for identifying PSA, almost all women (98%) tested positive immediately after exposure. The proportion testing positive declined to 29% by 24 hours and to 3% at 48 hours after insemination. Thus, high proportions of women who report semen exposure within the past 48 hours could be expected to test negative for PSA because of the marker’s clearance time.
We used Kruskal-Wallis test to compare PSA concentrations between subgroups defined by their self-reports (i.e., no sex, protected sex only, and at least one unprotected act) because the PSA concentrations were not normally distributed and were bound by 100 for the upper range. We used logistic regression to identify predictors of discordance between self-reports and biomarker detection based on data collected from the larger study. We fit the model for logistic regression with all of the available variables that were hypothesized to be potential predictors of discordance and then performed backward stepwise elimination of variables that were not predictive. We performed sensitivity analyses using higher cut points (up to 9 ng/mL) for a positive PSA result to test the robustness of all findings that involved PSA. Finally, we assessed the internal consistency of participant responses to study questions.
Although all 347 women who were recruited to participate in the study consented, samples or questionnaires were missing for 15 women. Thus, we measured PSA in vaginal swabs from 332 women (96% of the sample). The sample’s baseline characteristics were similar to those previously reported in the larger study.7 Figure 1 shows the distribution of the PSA concentrations for the 332 specimens.
Twenty-one percent of women who reported no sex and 39% of women who reported protected sex only for the prior 48 hours had PSA concentrations >3 ng/mL (Table 1). Even when reporting events for the prior 7- or 14-day periods, high proportions of women who reported no unprotected sex had PSA detected. Among women testing positive for PSA, concentrations for those who reported no sex in the past 48 hours (median, 100 ng PSA/mL; range, 3.4–100 PSA/mL); protected sex only (83.3; range, 4.0–100 PSA/mL); or at least one unprotected act (100; range, 3.1–100 PSA/mL) did not differ (Kruskal-Wallis test, P = 0.12). Sensitivity analysis using higher cut points did not reveal differences.
Predictors of Discordance
We hypothesized that the randomization group might be a predictor of discordant results (i.e., PSA identified but reporting no recent unprotected sex); women receiving more intensive condom counseling might be more likely to underreport unprotected sex. We assessed study site as a predictor because it could be a marker of interview-related effects. However, none of the characteristics assessed in Table 2 were statistically significant predictors of having discordant PSA and self-reported behavioral results except for testing positive for chlamydia at the study visit (odds ratio = 2.7; 95% confidence interval = 1.2–5.8).
Internal Consistency of Responses
Participant responses to the questionnaires for the larger and the current studies had high internal consistency. Of the 26 women who reported having no clients or partners in the last 7 days (larger study), one woman reported sex in the last 7 days (current study). Conversely, every woman (N = 286) who reported sex with at least one client or partner in the last 7 days (larger) also reported having sex in the last 7 days (current). All women (N = 247) who reported that last sex with a client had occurred within the last 7 days (larger) also reported having sex in the last 7 days (current). Of those (N = 67) reporting that the last time they had unprotected sex with a client occurred in the past 7 days (larger), 94% reported having unprotected sex in the last 7 days (current). Every woman (N = 101) who reported that last sex with a client had occurred today or yesterday (larger) also reported having sex in the last 48 hours (current).
Biologic and behavioral studies related to sex depend extensively on self-reports. However, the validity of these data is difficult to establish. For example, Zenilman et al failed to find a correlation between self-reported condom use and STI risk reduction.2 Some have interpreted this as evidence for the fallibility of self-reports, whereas others argued that STI is a poor surrogate for unprotected sex.9 Choosing an appropriate surrogate is vital to avoid incorrect research conclusions.10 Both PSA and Y chromosome appear to be good biomarkers of recent exposure to semen through unprotected coitus.11,12 We tested for PSA, which is found in high concentration in semen and can be reliably detected in the vaginal vault after unprotected intercourse.
We detected PSA in the vaginal vault of 21% of women who reported no sex and 39% of those who reported protected sex only within the past 48 hours. These results likely underestimate misreporting because PSA concentrations begin to decline immediately after exposure. Macaluso et al found that 71% of specimens tested negative for PSA 24 hours after clinical insemination of 1 mL of their partner’s semen.4 This rapid clearance of PSA from vaginal fluid after exposure to semen prevents us from measuring discordance in the other direction (i.e., testing negative for PSA despite reporting recent unprotected sex). Furthermore, menstruation, vaginal washing (an apparently common practice in the study population), or other behaviors hypothetically could have interfered with PSA detection or hastened its clearance.
In contrast, a positive test for PSA is unlikely to occur in the absence of recent semen exposure given the high test specificity.4,13,14 Incorrect condom use or condom malfunctions could have resulted in female exposure to semen.15,16 If this were the case, however, among those testing positive for PSA, PSA concentrations should be lower for those who reported protected sex only compared with those who admitted to at least one unprotected act. Despite the careful phrasing of questions, women could have misunderstood the questionnaire. Also, participants might have had difficulty considering the required time periods. However, when we compared participant responses for our and the larger study’s questionnaires (both of which were completed at the same visit), we found high internal consistency. Because achieving consistency between the 2 questionnaires often required both an affirmative and negative answer, women could not have achieved consistency simply by giving the same responses. Given the high internal consistency, poor comprehension does not appear to explain the apparently substantial misreporting.
We studied female SWs participating in an 18-month condom promotion study in Madagascar; the findings might not be generalizable to other populations. For example, social desirability bias could have caused the participants to overreport condom use during the face-to-face interviews. Also, the choice of study population could have resulted in selection bias; the subset of participants in the larger study who adhered to the study protocol by returning for their last scheduled visit might have been more likely to underreport unprotected sex. Alternatively, participants might have been less likely to misreport condom use than non-SW populations because the interviewers were aware of their occupation and the resistance to condom use that they often encounter from their clients.
Self-reported data are used for informing policy, research, and funding decisions regarding STI/HIV and pregnancy prevention efforts. Participants might give inaccurate responses as a result of self-presentation or courtesy bias,7,17–19 imperfect recall,20–23 poor question comprehension,24 limited topical vocabulary, exaggeration resulting from social norms or to comply with study eligibility criteria, personal salience of the sexual event, or emotional responses to sensitive questions.25–27 The high level of misreported recent exposure to semen that we demonstrated substantiates that self-reports of unprotected sex cannot be assumed to be valid measures. Future STI/HIV and pregnancy prevention studies should establish the veracity of self-reported measures of sex and condom use or should use end points that do not rely on self-reported data.
1. Pequegnat W, Fishbein M, Celentano D, et al. NIMH/APPC workgroup on behavioral and biological outcomes in HIV/STD prevention studies: A position statement. Sex Transm Dis 2000; 27:127–132.
2. Zenilman JM, Weisman CS, Rompalo AM, et al. Condom use to prevent incident STDs: The validity of self-reported condom use. Sex Transm Dis 1995; 22:15–21.
3. Lawson ML, Macaluso M, Bloom A, Hortin G, Hammond KR, Blackwell R. Objective markers of condom failure. Sex Transm Dis 1998; 25:427–432.
4. Macaluso M, Lawson L, Akers R, et al. Prostate-specific antigen in vaginal fluid as a biologic marker of condom failure. Contraception 1999; 59:195–201.
5. Graves HC, Sensabaugh GF, Blake ET. Postcoital detection of a male-specific semen protein. Application to the investigation of rape. N Engl J Med 1985; 312:338–343.
6. Kamenev L, Leclercq M, Francois-Gerard C. An enzyme immunoassay for prostate-specific p30 antigen detection in the postcoital vaginal tract. J Forensic Sci Soc 1989; 29:233–241.
7. Feldblum PJ, Hatzell T, Van Damme K, Nasution M, Rasamindrakotroka A, Grey TW. Results of a randomised trial of male condom promotion among Madagascar sex workers. Sex Transm Infect 2005; 81:166–173.
8. Armbruster DA. Prostate-specific antigen: Biochemistry, analytical methods, and clinical application. Clin Chem 1993; 39:181–195.
9. Turner CF, Miller HG. Zenilman’s anomaly reconsidered: Fallible reports, ceteris paribus, and other hypotheses. Sex Transm Dis 1997; 24:522–527.
10. Grimes DA, Schulz KF. Surrogate endpoints in clinical research: Hazardous to your health. Obstet Gynecol 2005; 105:1114–1118.
11. Steiner MJ, Feldblum PJ, Padian N. Invited commentary: Condom effectiveness—Will prostate-specific antigen shed new light on this perplexing problem? Am J Epidemiol 2003; 157:298–300.
12. Zenilman JM, Yuenger J, Galai N, Turner CF, Rogers SM. Polymerase chain reaction detection of Y chromosome sequences in vaginal fluid: Preliminary studies of a potential biomarker for sexual behavior. Sex Transm Dis 2005; 2:90–94.
13. Hobbs MM, Steiner MJ, Feldblum PJ, et al. Reply. Am J Epidemiol 2005; 162:706.
14. Macaluso M, Lawson ML, Warner DL, et al. Reply. Am J Epidemiol 2005; 162:705.
15. de Visser RO, Smith AM. When always isn’t enough: Implications of the late application of condoms for the validity and reliability of self-reported condom use. AIDS Care 2000; 12:221–224.
16. Crosby R, Sanders S, Yarber WL, Graham CA. Condom-use errors and problems: A neglected aspect of studies assessing condom effectiveness. Am J Prev Med 2003; 24:367–370.
17. Catania JA. The reliability of partner reports of sexual histories in a heterosexual STD clinic population. Sex Transm Dis 1996; 23:522.
18. Weinhardt LS, Forsyth AD, Carey MP, Jaworski BC, Durant LE. Reliability and validity of self-report measures of HIV-related sexual behavior: Progress since 1990 and recommendations for research and practice. Arch Sex Behav 1998; 27:155–180.
19. Fenton KA, Johnson AM, McManus S, Erens B. Measuring sexual behaviour: Methodological challenges in survey research. Sex Transm Infect 2001; 77:84–92.
20. Hornsby PP, Wilcox AJ. Validity of questionnaire information on frequency of coitus. Am J Epidemiol 1989; 130:94–99.
21. Lagarde E, Enel C, Pison G. Reliability of reports of sexual behavior: A study of married couples in rural west Africa. Am J Epidemiol 1995; 141:1194–2000.
22. Ellish NJ, Weisman CS, Celentano D, Zenilman JM. Reliability of partner reports of sexual history in a heterosexual population at a sexually transmitted diseases clinic. Sex Transm Dis 1996; 23:446–452.
23. Weir SS, Roddy RE, Zekeng L, Ryan KA, Wong EL. Measuring condom use: Asking ‘do you or don’t you’ isn’t enough. AIDS Educ Prev 1998; 1:293–302.
24. Geary CW, Tchupo JP, Johnson L, Cheta C, Nyama T. Respondent perspectives on self-report measures of condom use. AIDS Educ Prev 2003; 15:499–515.
25. Berk R, Abramson PR, Okami P. Sexual activities as told in surveys. In: Abramson PR, Pinkerton SO, eds. Sexual Nature, Sexual Culture. Chicago: University of Chicago Press, 1995:371–386.
26. Orr DP, Fortenberry JD, Blythe MJ. Validity of self-reported sexual behaviors in adolescent women using biomarker outcomes. Sex Transm Dis 1997; 24:261–266.
27. Cecil H, Zimet GD. Meanings assigned by undergraduates to frequency statements of condom use. Arch Sex Behav 1998; 27:493–505.