- Dr Robert S. Van Howe, 413 E Ohio St, Marquette, MI 49855 (vanhowe@miuplink.net or vanhowe@msu.edu)
To the Editor—I read with interest the recent study of human papillomavirus (HPV) infection and circumcision by Auvert et al [1]. Although the authors believe they have demonstrated that circumcision prevents HPV infection, their findings can be completely attributed to sampling bias, which the authors fail to consider
Sampling Bias Explains Association between Human Papillomavirus and Circumcision
The authors chose to sample only the urethra for HPV, stating that “[t]he urethra was chosen because the detection of HPV in this anatomical site is probably not affected by circumcision status” [1, p 15]. Unfortunately, several studies have demonstrated that the yield of HPV cultures differs at various locations on the penis depending on circumcision status [2 –6]. Most notably, HPV is more likely to be cultured from the shaft of the circumcised penis and from the glans of the normal penis. Weaver et al [2] found that, among men who had HPV recovered from any of numerous locations, 48% of circumcised men and 65% of uncircumcised men had a positive glans culture result. This indicates that an infected man with an uncircumcised penis for whom only the glans was cultured would be 1.35 times more likely to have a positive result than would an infected man with a circumcised penis. If the numbers of Auvert and colleagues are adjusted for this difference in yield, their findings are less dramatic and are no longer statistically significant (intention-to-treat odds ratio [OR], 0.87 [95% confidence interval {CI}, 0.66–1.14]; as-treated OR, 0.77 [95% CI, 0.59–1.01])
Auvert and colleagues state that their findings are consistent with those of other studies, yet these studies likewise fail to address sampling bias. For example, Hernandez et al [5] found that oncogenic HPV was more common on the glans of men with an uncircumcised penis; however, when oncogenic HPV obtained from any site is considered, no difference is seen (OR, 0.88 [95% CI, 0.45–1.74]). A meta-analysis of circumcision status and HPV infection that used metaregression also demonstrated that finding an association between HPV and circumcision status was less likely in studies in which sampling the shaft of penis for HPV was performed [7]. A recently published birth cohort study from New Zealand found that the combined seroprevalence of HPV-16 and HPV-18 was not affected by circumcision status (OR, 1.2 [95% CI, 0.75–1.9]) [8]. When studies published since the meta-analysis are included [1, 5, 6, 8] and the analysis is repeated using metaregression, not ascertaining circumcision status by physical examination (OR, 1.77 [95% CI, 1.15–2.73]) and not sampling the penile shaft (OR, 1.57 [95% CI, 1.16–2.12]) are both significantly associated with finding an association between circumcision status and HPV infection. Adjusting for these 2 factors, the summary OR for circumcision is 0.91 (95% CI, 0.74–1.11), and the difference is not statistically significant
When Auvert and colleagues chose to sample only the urethra, sampling bias guaranteed that their study would find a negative association between circumcision status and HPV infection. Consequently, their study fails to confirm an association between HPV infection and circumcision status, but it does provide further evidence that the sampling bias seen in this and other studies is unidirectional
- © 2009 by the Infectious Diseases Society of America
