Our predominantly null results do not support a link between sexual behavior or STIs and PSA concentrations in this nationally representative sample of US men 40–59 years of age, with the exception of recent sex and condom use. Therefore, sexual factors are unlikely to lead to falsely elevated PSA tests in this population.
We found a statistically significant lower PSA concentration and a higher %fPSA in men who had had sex without a condom in the last month. We are unsure as to why men who have had recent unprotected sex would have a lower PSA than those that have not. It should be noted that these analyses were limited by small numbers of men, and the possibility remains that this statistically significant result is a chance finding.
There are several reasons for our predominantly null results in this analysis. The lack of association between PSA and the number of lifetime sexual partners or the number of sexual partners in the last year may be because past STIs never infected the prostate, or infected the prostate but were treated and thus no longer induce inflammation. Additionally, men may have been infected asymptomatically, and now report never having had an STI. The antibody data collected in the NHANES do not allow us to consider the precise time of infection or current infection status of those testing positive for exposure to HSV-2 or syphilis. These 2 organisms are not believed to infect the prostate, and hence are unlikely to cause prostatic inflammation themselves; we considered these agents as surrogates for sexual behaviors and the opportunity for the acquisition of other STIs that do infect the prostate. Of the men self-reporting a past diagnosis of an STI (n = 77), most reported either a genital warts or genital herpes diagnosis (n = 71), and only 6 men reported a diagnosis of gonorrhea or chlamydia. Given that the HPV subtypes responsible for genital warts are rarely oncogenic, and that HSV-2 is not known to infect the prostate, we likely misclassified some of the men regarding likelihood of prostatic inflammation. Furthermore, men were asked to report gonorrhea and chlamydia diagnoses only within the past 12 months. Therefore, men who had had these infections in their young adulthood were combined with men who had never been infected with N. gonorrhea or C. trachomatis. All of these possible sources of misclassification, which would have biased our results toward the null, may have limited our ability to detect associations of a small magnitude between sexual behavior and PSA concentrations.
The PSA component of the NHANES excluded any man self-reporting current prostatic inflammation or infection, a well-established elevator of serum PSA. [15] This criterion likely excluded men with symptomatic prostatitis from our study, but we likely included men with asymptomatic chronic prostatitis because they are usually unaware of their condition. Men with mild prostatitis that has not been diagnosed by a doctor are also possibly included in this analysis. Because men with symptomatic prostatitis were excluded we may have underestimated the associations of STIs and sexual behavior with PSA concentration. We do not believe this is a substantial underestimation, because in the antibiotic era it is likely that men undergo curative treatment for a symptomatic STI before it infects the prostate. Men who do not seek treatment for a symptomatic STI likely make up a small portion of the general US population.
It is likely that some proportion of the men in this analysis had occult prostate cancer, which might have influenced PSA concentration. However, the prevalence of cancer in our study population is likely to be low, because all of our participants were younger than 60. [16] The majority of our analysis was restricted to men between the ages of 40 and 59. We do not feel this negatively impacts our analysis, and it may be a positive aspect of it. The inflammatory events that predispose men to later developprostate cancer are likely to have occurred many years prior to the diagnosis. The chronic inflammatory state induced by these events may be detectable as elevated PSA levels during these pre-diagnosis years, the time period where our cross-sectional analysis was conducted. One of the reasons our analysisdid not find a positive association with PSA could be that the infections precipitating the chronic inflammatory state thatleads toprostate cancer occurmostly while men are older than the men in our study. Prostate volume is also associated with serum PSA, [17] and we were unable to control for it in our analysis. The extent to which these 2 factors are associated with sexual behavior or STIs is unknown, and it is possible that they are confounders in this analysis.
The NHANES is not an ideal population in which to study the association between PSA values and a specific STI for a number of reasons. For one, the population is selected and weighted to represent the US population as a whole, not the high-risk sub-population you would expect to find at an STI clinic. This limited the number of STIs captured in this analysis, and forced us to collapse information into binary variables, in the process losing information on the association between PSA values and specific STIs. Another methodological limitation is that the NHANES is a cross-sectional study, and as such we have no information on the participants' numbers of previous STIs, in the cases where there was more than one episode of infection. This would have been valuable information, as men with multiple previous infections are more likely to have had prostate involvement. For these reasons we cannot rule out the role of sexual factors in causing false positive PSA tests in subgroups of the population that have a higher prevalence of high-risk sexual behavior, and more protracted or recent exposures to these agents.
However, the NHANES population-based sampling is also a strength of this analysis, allowing us to extrapolate these results to the male US population. Nationally representative analyses are ideal for studies of population-wide screening tools such as the PSA test, because they reveal the magnitude of potential associations over the entire population that uses the test. This study suggests that the prevalence of sexually transmitted infections of the prostate in the overall US population is relatively low, leading one to conclude that they do not greatly affect the accuracy of the PSA test at this population level.
Self-reported data on sexual behavior are susceptible to a number of response biases. [18] The sexual behavior data used in this analysis were collected using a computer-based system that allowed respondents complete privacy. Because of this enhanced privacy, computer-assisted self-interviewing methods are believed to be least susceptible to the motivational biases encountered commonly in these assessments. However, over-reporting of low frequency estimates and under-reporting of high frequency estimates is a common source of misclassification and may have influenced these results toward the null. [19]