Background: Reporting bias and validity estimation of self-reports in clinical settings is a major problem in sexual behavior research.
Objective: The objective of this study was to determine if a newly described biomarker tool helps clarify sensitivity issues related to condom use self-report.
Methods: A polymerase chain reaction (PCR) assay to detect Y-chromosome (Yc) fragments was used to assess detectability and longevity of YcDNA in vaginal swabs collected from 141 women seeking care for sexually transmitted disease (STD)-related symptoms or as a reported sexual contact to STD. Data analyzed were collected in 1992 to 1994 as part of the Transmission, Acquisition, and Condom Use study. Archived vaginal swab samples were selected from women who in the accompanying survey reported their last intercourse in the previous 14 days. Survey data about partners, sexual behaviors, and self-reported condom use in the previous month was also retrieved.
Results: Overall, 137 (97.2%) vaginal samples had usable PCR assay results; 90 (65.7%) had detectable YcDNA content. Linear regression showed that number of days since last sexual intercourse was a significant predictor of YcDNA concentration. PCR results were compared with self-reported condom use. Of the women, 67 (47.5%) women reported no condom use in the last 14 days, and 36 (25.5%) women reported using condoms consistently. Although YcDNA was detected in the swabs of both condom reporting groups, mean DNA content was significantly lower among the consistent condom users.
Conclusion: The YcPCR assay can detect DNA in a clinical sample of women for up to two weeks. This assay may be useful as a qualitative adjunct to behavioral studies to further understanding of sexual behavior reporting in women.
We used a Y-chromosome polymerase chain reaction assay to validate self-reported sexual behavior in archived vaginal lavage samples from a previous longitudinal study that evaluated incidence of sexually transmitted disease and self-reported condom use.
From the *University of Wisconsin–Eau Claire, College of Nursing and Health Sciences, Eau Claire, Wisconsin; and the † Division of Infectious Diseases, Johns Hopkins University School of Medicine, Baltimore, Maryland
This work was supported by NIH grants K24 AI 001633 and R01 HD 043674.
Correspondence: Jonathan Zenilman, MD, Infectious Diseases Division, Johns Hopkins Bayview Medical Center, 4940 Eastern Avenue, Baltimore, MD 21224. E-mail email@example.com.
Received for publication January 14, 2005, and accepted June 3, 2005.
A MAJOR FOCUS OF SEXUALLY transmitted disease (STD) and HIV prevention interventions is condom promotion.1,2 When consistently and correctly used, condoms are an effective means of preventing HIV/STDs3 and pregnancy.4 Evaluating consistent condom use in persons at high risk for STD is challenging. Because condom use is typically assessed by self-report, the measurement is prone to reporting and social desirability biases, especially in settings where condom use is vigorously promoted.5–7 Because of these problems, biomarkers are an attractive option. Previous biomarker approaches have used incident STDs as a marker of inconsistent condom use. However, using STDs as a condom validation measure requires studying a population at high risk for incident (i.e., mostly bacterial) STDs and having appropriate laboratory support, which is expensive. Furthermore, even in high-incident populations such as our Baltimore City STD clinic attendees, incidence rates approach only 5% per month.5,8 Therefore, only a small minority of persons who have risky exposures become infected, leading to type II error if this is used as a primary condom validation tool.9
A promising approach to assess the validity of self-reported condom behaviors is through the use of nondisease biologic markers. The intrinsic genetic difference between the male and female genome provides opportunity for a sexual exposure biomarker, because Y chromosome sequences are unique to males. Screening for the Y chromosome in vaginal fluid could provide a highly sensitive validity marker for self-reports of condom use.
We developed a polymerase chain reaction (PCR) assay to detect Y-chromosome DNA components in vaginal swabs.10 In controlled experimental settings, we were able to detect Y chromosome in vaginal swabs up to 15 days after unprotected vaginal intercourse.10 In this current study, we extend this work using archival samples to assess the presence of Y chromosome in the vaginal secretions of women in a less controlled setting—women accessing an urban STD clinic—and compare these results with their self-reported condom use and risk behaviors.
We analyzed samples collected in 1990 to 1992 as part of the National Institutes of Health (NIH)-funded Transmission, Acquisition, and Condom use study (TRAC).5 This study was one of the first prospective studies to evaluate the validity of self-reported condom use and the reliability of calendar reports of sexual behavior using data from sexual partners.5,7,11 Our previously reported controversial findings were that self-reported condom use patterns did not predict STD incidence. In part, previous criticisms included the sensitivity and specifications of STD outcomes to measure condom use. With the availability of new biomarker tools, we were interested in ascertaining if the newly described Yc nondisease biomarker could clarify some of these issues. The objectives of this analysis were to 1) assess our ability to detect Y chromosome in vaginal fluids of women accessing an STD clinic, and 2) compare presence of YcDNA and reports of condom use to self-reported risk behaviors.
Clinical Characterization of the Assay in Sexually Active Women Accessing an Urban Sexually Transmitted Disease Clinic
Samples analyzed in the study were collected in 1990 to 1992 as part of the NIH-funded Transmission, Acquisition, and Condom use study.5,7 Briefly, a systematic sample of men and women were recruited at the Baltimore City Health Department Sexually Transmitted Disease clinics if they were seeking care either for STD-related symptoms or reported sexual contact to a person with an STD. After informed consent, a detailed sexual history and STD symptoms questionnaire was administered. This included a 30-day retrospective calendar on which the subject was asked to indicate dates of sexual activity, types of sexual exposure, initials of the sexual partner, and whether a condom was used.7 Clinical evaluation was performed, including testing for STD. In addition, nonmenstruating women had a cervicolavage (CVL) sample collected. The CVL was obtained by introducing 10 cc of phosphate-buffered saline in the vagina and then rinsing the cervical os and the posterior fornix with an 18-gauge Teflon intravenous catheter attached to a 10-cc syringe. The fluid was aspirated and then transported to the laboratory. In the laboratory, the sample was centrifuged at 5000 rpm × 10 minutes. The cellular pellet was resuspended in 1 cc of supernatant and stored at −70°C.
Identification of Specimens
From the original database and specimen archive, we identified first-visit women from the original study who reported their last intercourse in the previous 14 days, including women who reported either consistent (100%) reported condom use or no (0%) reported condom use.
DNA Extraction from the Cervicolavage Samples
Vaginal fluid components were removed from the CVL by placing in 0.5 mL sterile water, incubating at room temperature for 5 minutes, and rotating vigorously. The specimen was centrifuged for 1 minute at 10,000 g and all but 200 μL of supernatant was removed.
The resulting specimen was extracted for Y-chromosome DNA by a modified two-step differential extraction technique.10 This dual DNA extraction method allows elimination of all male and female epithelial cell sources from the final sperm extraction. Briefly, 2 μL of proteinase-K (10 mg/mL) was added to the sample and incubated at 56°C for 1 hour. The sample was centrifuged for 3 minutes at 10,000 g, the supernatant was removed, and the pellet was resuspended in 0.5 mL wash buffer containing 10 mmol/L Tris-HCL 7.5, 10 mmol/L EDTA, 50 mmol/L NaCl, and 2% SDS. The pellet was vortexed then centrifuged for 3 minutes at 10,000 g. The supernatant was discarded and the washing was repeated two more times. The pellet was washed a final time in 1 mL distilled water, then vortexed, and centrifuged for 3 minutes. The pellet was then resuspended in 150 μL of 5% (w/v) Chelex 100 (100–200 mesh, sodium form, biotechnology grade; Bio Rad), 2 μL of proteinase-K (10 mg/mL), and 7 μL of 1 mol/L DTT. The sample was incubated for 1 hour at 56°C. After incubation, the sample was vortexed for 5 to 10 seconds, centrifuged for 5 to 10 seconds at 10,000 g, and boiled for 8 minutes. The supernatant was then used for DNA amplification.
A PCR assay was developed in a controlled setting for Y-chromosome DNA components from forensic science protocols and is described elsewhere.10,12 PCR reactions were conducted exclusively by a female laboratory worker to prevent possible Y-chromosome contamination in a sterile or clean PCR hood. PCR amplification of the DNA extract was performed using a modification of previously described procedures.10,12 The samples were amplified, and X and Y amplifications were run in separate reaction tubes. Controls for the PCR reaction included DNA used to detect Y-chromosome contamination.
These experiments measured the absolute quantity of DNA present in the CVL sample pellet. YcDNA results were transformed using the natural logarithm (ln). Transformed YcDNA content for swabs was plotted against time since last intercourse, and the mean trend was calculated using both linear regression and a locally weighted smoothing (lowess) curve. The nonparametric lowess curve is way of examining “local” relationships between a response variable and a particular variable over parts of their ranges, which may differ from a “global” relationship determined using the whole dataset. YcDNA content plotted against time since last intercourse was compared by reported condom use groups (never used condoms in the last 14 days, always used condoms in the last 14 days) using linear regression. Similarly, YcDNA was compared with self-reported measures of risk behaviors. The t-test was used to compare independent means, chi square was used to compare independent proportions, and Pearson’s correlation coefficient was used to measure the dependence between variables. All P values are two-sided, with P <0.05 representing statistical significance.
Demographic Characteristics of the Subjects
Samples were available for 141 women who met the criteria outlined. Their mean age was 24.5 (standard deviation [SD], 6.6) years; 112 (78.9%) were single, 12 (8.5%) were separated, 12 (8.5%) were married, and 5 (3.5%) were divorced. The majority was black (87.2%). Demographic data were similar to those from the overall TRAC sample.
The subjects reported one to three sexual partners in the past month and reported an average of eight episodes of vaginal sexual intercourse in the past month. The mean number of days since last sexual exposure was 5.4 days (SD, 4.1 day; range, 0–14 days). The average percent of reported condom use in the past month was 40%; 67 (47.5%) women reported no use in the previous 2 weeks, 36 (25.5%) reported “some” condom use, and 38 (27%) reported using condoms for all sexual episodes in the past 2 weeks.
Longevity of Detectable Y-Chromosome DNA in Vaginal Fluid
Of the 141 vaginal samples, 137 (97.2%) had usable PCR assay results, 90 (65.7%) had detectable YcDNA content, 47 (34.4%) had no detectable DNA content. A linear regression analysis revealed that number of days since last sexual intercourse was a significant predictor of ln(YcDNA) (P <0.001), accounting for 15.7% of the variance in ln(YcDNA). Clinical longevity of YcDNA was also examined by plotting ln(YcDNA) by the number of days since last unprotected episode of vaginal intercourse using linear regression (Fig. 1). The predicted line had a Y intercept (ln[YcDNA] at the time of intercourse) concentration of 2.94 (95% confidence interval [CI], 2.29–3.59); results of the linear regression showed there was a decrease of 0.20 ln(YcDNA) per day since last sexual intercourse (95% CI, −0.294–−0.100). Untransformed data were also fitted with a nonparametric lowess curve (Fig. 2). Figure 2 demonstrates that detectable YcDNA in this clinical sample dropped dramatically at approximately 1 week.
Self-Reported Condom Use and Detection of Y-Chromosome DNA
PCR results were compared with self-reported condom use. Two reported condom use groups were examined: women who reported using condoms for each and every episode of sexual intercourse and women who reported never using condoms (see Table 1). The percentage of participants with positive PCR results did not differ by self-reported condom use: 70.1% of those who did not use condoms had detectable YcDNA compared with 55.6% of self-reported consistent condom users (P = 0.1).
However, when examining the presence of YcDNA in the positive PCR samples between reported condom use groups, more predictable results emerged. Although YcDNA was detected in the swabs of both groups, the mean DNA content was significantly less among the consistent reported condom users (P <0.001). The predicted lines based on linear regression are shown in Figure 3. Consistent reported condom users had a mean ln(YcDNA) of 1.26 (95% CI, −0.36–2.88); there was a decrease of 0.10 ln(YcDNA) per day since last intercourse (95% CI, 0.03–0.12). Respondents who reported no condom use had a mean ln(YcDNA) of 4.04 (95% CI, 3.25–4.82); there was a decrease of 0.27 ln(YcDNA) (95% CI, −.38–−.15) per day since last intercourse.
Linear regression analysis revealed that number of days since last sexual intercourse was not a significant predictor of ln(YcDNA) among the consistent reported condom users (P = 0.384). Conversely, days since last sexual intercourse was a strong predictor of ln(YcDNA) among respondents who reported no condom use (P <0.001).
Polymerase Chain Reaction and Reported Condom Use Groups Compared With Selected Risk Behaviors
Presence of YcDNA and reported condom use groups were compared with a number of risk behaviors, including previous sexually transmitted infection (STI), number of partners, drinking, and drug-use behaviors. Overall, there were no significant differences in the groups by previous STI, current STD, drinking, and drug-use behaviors. However, significant comparisons were found for number of sexual partners. Significantly, more persons with 2 or more partners in the past month had positive PCR results even after adjusting for self-reported condom use (P <0.02).
Our goal was to revisit a prospective cohort study published in 1995 of self-reported condom use among STD clinic attendees.10 In that study, participants who reported “always” using condoms in the past 30 days had the same number of incident STIs as those who reported “never” using condoms. After its publication, several hypotheses were advanced to explain the anomalous findings. These included methodological deficiencies, selection/exposure bias, and outcome identification biases.14 This follow-up study allowed us to apply a biomarker for reported condom use10 to better elucidate those controversial findings.
We found that presence of DNA did not differ between reported condom use groups and represents a limitation to the use of this test to determine whether unprotected intercourse took place. Similar to our previous studies, these discordant suggest that relying on self-reports of consistent reported condom use may underestimate prevalence of risk behaviors.5,11 However, when examining the amount of YcDNA between condom-use groups, more predictable patterns emerge. Regression analyses showed significantly more DNA present among those reporting no condom use. The mean baseline YcDNA was higher by a factor of three for those who reported no condom use as compared with those who reported consistent condom use.
The presence of DNA among reported condom users raises interesting questions. Persons who report consistent condom use, yet test positive for YcDNA, represent a group who may be using condoms incorrectly or represent individuals who may be providing inaccurate self-reports. Similar concerns have been reported about other self-reported risk behaviors. In clinical samples, recollection for events in the past month may come in to play, leading to concern about the extent to which respondents can accurately recall behaviors over time.5,11 Researchers have suggested that the window for accurate self-report of sexual variables might be at best 1 week and at most 2 weeks.10 Other researchers have described how accuracy of recall may also be influenced by other risk variables such as number of partners.13 We describe similar findings in that women with more than two partners were more likely to have positive PCR results. YcDNA used as a biomarker could provide clinicians and researchers with additional tools to accurately assess sexual risk behaviors.
Use of the YcDNA assay with clinical samples may be limited by the number of days since last intercourse. In this sample, 30% of those who had sexual intercourse in the past 2 weeks and reported “never” using condoms had negative PCR results. When examining only those who reported no condom use, the mean number of days since last intercourse for those with negative PCR results was 7.4 days as compared with 5.4 days for those with positive PCR results. It may be that in clinical settings, the ability of the assay to detect YcDNA in vaginal samples decreases more quickly over time than in controlled settings because of increased variability in measurement.
This study is also limited by the fact that all changes in mean YcDNA are described based on cross-sectional data. Longitudinal data may provide more accurate estimates by accounting for individual variability. However, the overall hypotheses that YcDNA can be detected and that the amount detected decreases as a function of time are supported by this study. In addition, these data are similar to findings from the pilot study in which data were longitudinally collected10 and provide further support for the use of YcDNA as a biomarker of consistent reported condom use.
We have shown that the YcDNA assay can detect YcDNA components for up to 2 weeks postcoitus in this clinical sample. The YcDNA assay in this setting should be thought of as qualitative and as an adjunct to behavioral studies. Quantitative determination in actual field conditions is difficult because of the tremendous variability potential in samples, including concentration and the amount of semen deposited in the vagina, physical characteristics of the swab, and the vaginal environment.
This assay has the potential for being a nondisease biomarker of unprotected sexual behavior. We found that the amount of YcDNA detected is greater for women who reported “never” condom use as compared with those who reported “always” use but that YcDNA was still detectable in the latter group. These findings echo the original paper’s conclusions that, even in a research setting, self-reported condom-use behaviors are subject to substantial reporting bias.
1. The Hidden Epidemic. Washington, DC: National Academy Press, 1997.
2. Robin L, Dittus P, Whitaker D, et al. Behavioral incidence of HIV, STD, and pregnancy among adolescents: A decade in review. J Adolesc Health 2004; 34:3–26.
3. Jones EF, Forrest JD. Contraceptive failure rates based on the 1988 National Survey of Family Growth. Fam Plan Perspect 1992; 24:12–19.
4. Ku L, Sonenstein FL, Pleck JH. Patterns of HIV risk and preventive behaviors among teenage men. Public Health Rep 1992; 107:131–138.
5. 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.
6. Catania JA, Gibson DR, Chitwood DD, Coates TJ. Methodological problems in AIDS behavioral research: Influences on measurement error and participation bias in studies of sexual behavior. Psychol Bull 1990; 108:339–362.
7. Upchurch DM, Weisman CS, Shepherd M, et al. Interpartner reliability of reporting of recent sexual behaviors. Am J Epidemiol 1991; 134:1159–1166.
8. Zenilman JM, Erickson B, Fox R, Reichart C, Hook EW. Effect of HIV posttest counseling on STD incidence. JAMA 1992; 267:843–845.
9. 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.
10. 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; 32:90–94.
11. Ellish NJ, Weisman CS, Celentano D, Zenilman JM. Reliability of partner reports of sexual history in a heterosexual population at a sexually transmitted disease clinic. Sex Transm Dis 1996; 23:446–452.
12. Neeser D, Liechti-Gallati S. Sex determination of forensic samples by simultaneous PCR amplification of alpha-satellite DNA from both the X and Y chromosomes. J Forensic Sci 1995; 40:239–241.
13. Jaccard J, McDonald R, Wan CK, Guilamo-Ramos V, Dittus P, Quinlan S. Recalling sexual partners: The accuracy of self reports. J Health Psychol 2004; 9:699–712.
© Copyright 2006 American Sexually Transmitted Diseases Association
14. Turner CF, Miller HG. Zenilman’s anomaly reconsidered: Fallible reports, ceteris paribus, and other hypotheses. Sex Transm Dis 1997; 24:522–527.