Journal of Infection (e-publication ahead of print), 22 September 2006.
Robert S. Van Howe*
* Tel.: +1 906 225 3925; fax: +1 906 225 4838.
E-mail addresses: email@example.com, firstname.lastname@example.org
Accepted 8 August 2006
© 2006 The British Infection Society. Published by Elsevier Ltd. All rights reserved.
KEYWORDS: Circumcision; Human papillomavirus; Penile cancer; Meta-analysis
Background: Determine the relationship of circumcision status to the risk for genital infection with human papillomavirus (HPV).
Methods: A MEDLINE search and a review of references in published articles were used to identify publications from peer-reviewed journals in Index Medicus with data on circumcision status in patients with and without HPV infections. Inclusion criteria included diagnosis by culture, biopsy, or PCR, determination of circumcision status by physical examination, and multiple site sampling including the shaft of the penis. A meta-analysis was performed with sensitivity analyses.
Results: Sixteen articles contained data on circumcision status in patients with and without HPV infections. Eight studies used accurate diagnostic methods. Only three articles satisfied the strict inclusion criteria. There was no significant association between circumcision status and HPV infection (random-effects model summary effect OR Z 1.20, 95%CI Z 0.80–1.79) in these three studies. If the eight studies using accurate diagnostic methods are adjusted for the method of determining circumcision status and failure to sample the penile shaft using meta-regression the summary effects odds ratio is 1.25 (95%CI Z0.95–1.67).
Conclusions: The medical literature does not support the claim that circumcision reduces the risk for genital HPV infection. To correctly assess the risk of HPV infection in circumcised males, the penile shaft needs to be sampled for HPV infection.
This file does not include three tables and one figure.
A decreased risk for sexually transmitted infections in circumcised adult males has long been used as a justification for neonatal circumcision.1-3 Other than general review articles and opinion pieces, the association between circumcision status and sexually transmitted infections has not been subjected to systematic review or analysis until recently.4,5 In 1999, the American Academy of Pediatrics Task Force on Circumcision concluded, ‘‘Existing scientific evidence demonstrates potential medical benefits of newborn male circumcision; however, these data are not sufficient to recommend routine neonatal circumcision.’’ In addressing sexually transmitted infections other than HIV, they merely stated, ‘‘Evidence regarding the relationship of circumcision to STD in general is complex and conflicting.’’6
Human papillomavirus (HPV) is believed to be responsible for cervical cancer in women and penile cancer in men. In an effort to clarify the impact of circumcision status on sexually transmitted HPV infections and indirectly on the female partner’s risk of cervical cancer, the medical literature addressing the impact of circumcision status on sexually transmitted HPV infections was reviewed and analyzed.
The recommendations of Stroup et al. for the meta-analysis of observational studies were followed.7 Inclusion criteria included publication of a cohort, cross-sectional, and case-control study in a peer-reviewed journal, the presence of data on the circumcision status of males both with and without genital HPV infections, diagnosis by culture, biopsy, or HPV DNA detection using PCR or Hybrid capture 2, determination of circumcision status by physical examination, and multiple site sampling including the shaft of the penis. Articles were identified using a MEDLINE search and a review of references in published articles. A MEDLINE search using PubMed was last undertaken on March 13, 2006. ‘‘Circumcision’’ was used as a key word, which identified 3515 articles. The majority were review articles and opinion pieces. Fifteen articles compared the rates of sexually transmitted HPV infections, including genital warts, in men who were and were not circumcised.8-22 One additional study was identified by review of the bibliographies.23
Of these sixteen studies, four made the diagnosis of genital warts clinically,9,11,12,14 one relied on patient report,18 and four lacked documentation of the diagnostic method.8,15,23 These eight studies were excluded from analysis. An additional two studies failed to sample the penile shaft and were excluded.16,19 Three other studies relied on patient report to determine circumcision status and were excluded.17,20,22 This left three studies that fulfilled the inclusion criteria. Analysis was performed with these three studies. Because of the small number studies that met the inclusion criteria, it was elected to analyze the eight studies with sufficient diagnostic methods adjusting for differences in sampling and assignment using metaregression24 and, alternatively, by estimating the number of cases of HPV infection missed by not sampling the penile shaft based on the rates published by Weaver et al.21 In studies in which distinct strata of the subjects, each strata was considered separately in calculating the summary effect.
To stabilize the impact of cells with small numbers, a general variance-based random-effects model25 was run using each study’s exact odds ratios and confidence intervals (Proc LogXact, version 5.0, Cytel Software Corporation, Cambridge, Massachusetts). To test for potential outliers, the dataset from each study group was individually excluded from the analysis to measure the impact on the chi-square measure of betweenstudy heterogeneity. The possibility of publication bias was explored using funnel graphs and linear regression analysis as described by Egger and associates,26 funnel plot regression as described by Macaskill, Walter, and Irvwig,27 and the adjusted rank correlation test described by Begg and Mazumdar28 using statistical analysis software (SAS version 8.02, SAS Institute, Cary, North Carolina). Adjustment for publication bias was performed using the ‘‘trim and fill’’ method described by Duvall and Tweedie.29,30
Characteristics of the studies with data on the circumcision status of males both with and without genital HPV infections are shown in Table 1. Table 2 shows the raw data from the sixteen studies comparing HPV infections by circumcision status. Table 3 shows the results of the meta-analyses. The random-effects summary effect odds ratio of the three studies meeting the inclusion criteria was 1.20 (95%CIZ 0.80–1.79). There was no evidence of significant between-study heterogeneity.
In our attempt to identify datasets that might be outliers, none met the criteria.
A funnel graph, which plots precision (the inverse of variance) on the y-axis and the natural logarithm of the odds ratio on the x-axis, should have a shape like an inverted funnel (Fig. 1). With the largest study representing the apex of the inverted funnel, there is a paucity of studies in the left lower portion of the inverted funnel and cluster of studies in the right lower portion that did not meet the strict inclusion criteria. Methods to determine the presence of publication bias use a p-value threshold of 0.10 for significance. These were applied to the eight studies using reliable diagnostic methods. Using Egger’s regression model the p-values were 0.3642 and 0.1996 for the unweighted and weighted models, respectively. The results of Begg’s adjusted rank correlation test were also not consistent with publication bias (pZ 0.9029 and pZ 0.3290 for standardized treatment effect versus variance of the treatment effect and versus study sample size, respectively). Using Macaskill’s funnel plot regression, neither the unweighted model nor the weighted model was significant for publication bias (p Z0.5537 and p Z 0.3600, respectively).
Using the ‘‘trim and fill’’ technique on the eight studies using reliable diagnostic techniques, three datasets in the right hand portion of the inverted funnel were found to not have a corresponding dataset in the left hand portion of the funnel.
This is the first systematic review of the medical literature looking at the risk of genital HPV infections based on circumcision status. Using strict criteria to include multiple genital sites for sampling, including detection of HPV on the penile shaft the analysis failed to show a significant association between genital HPV infection andcircumcision status.
The limitations of the meta-analysis when applied to observational studies, the role of between-study heterogeneity, the role of publication bias, the use of exact odds ratios, and the limitations of attributing risk of STD to circumcision status have been discussed previously.4 To minimize bias in studies included in the meta-analysis, selection criteria were developed. Relying on patient report or clinical appearance would miss the many HPV infections that are only found through physical examination using acetic acid, biopsy, culture, or HPV DNA testing. Likewise, the correlation between patient report and physical examination can vary widely depending the population studied.9,31-34 Finally, two studies have shown the distribution of HPV infections of the male penis to differ by circumcision status, with circumcised men having the penile shaft as the focus of these infections.11,21 For example, more than half of HPV infections in circumcised men will be missed if the penile shaft, scrotum, and urine are not sampled, while more than 89% of HPV infections will be detected in genitally intact males sampling only the glans and foreskin.21 Studies that failed to sample the penile shaft would fail to diagnose a disproportionate number of circumcised subjects. While some would argue that the relevance of HPV DNA detection from the penile shaft is unclear, genital warts and penile intraepithelial neoplasia have been found on the shaft.10,11 The presence of HPV on the shaft of the penis indicates that the virus is also available forsexual transmission.
With these criteria in pla ce, only three studies satisfied all of them. Rather than limit the analysis to these three studies, it was elected to evaluate the eight studies with reliable diagnostic methods and attempt to minimize the bias introduced by incomplete sampling and determination of circumcision status by patient report. The first method, meta-regression, adjusted for incomplete sampling and method of ascertaining circumcision status using a regression model.24 The second method estimated the number of cases of HPV infection missed by not sampling the penile shaft using the numbers generated by Weaver and colleagues.21 This method depends on the applicability of these ratios to other populations. Knowing that these biases exist and not adjusting for them would generate results know to be biased.
The studies included for analysis only measured prevalence of HPV infection. This study does not attempt to describe the relationship between circumcision status and persistent HPV infection or penile cancer. The impact of HPV and circumcision status on penile cancer is only now enfolding. While HPV DNA is found over 90% of the biopsies of penile intraepithelial neoplasia and carcinoma in situ, it is detected in only about half of the cases with invasive squamous cell carcinomas.35 Two recent studies indicate that phimosis rather than having a foreskin places a noncircumcised male at greater risk for invasive penile cancer.36,37 Non-circumcised men are at no greater risk for less invasive disease.36 There is mounting evidence that balanitis xerotica obliterans, a common cause of pathologic phimosis, may also be associated with penile cancer.38-40 It is likely that HPV associated lesions, such as penile intraepithelial neoplasia, carcinoma in situ, and half of the cases of invasive squamous cell carcinoma occur equally in circumcised and non-circumcised men. The remaining squamous cell carcinomas may develop from the rare case of balanitis xerotica obliterans seen in non-circumcised men. Longitudinal studies documenting the natural history of HPV infection in men are needed to answer this question more directly.
Three of the recent publications comparing the rate of genital HPV in circumcised and non-circumcised men suffer from serious methodological flaws. One study compiled seven studies from five countries from three continents. The study groups were the regular male sexual partners of women diagnosed with invasive cervical cancer or cervical cancer in situ. Control groups were drawn from the same medical facilities.16 Consequently, half of the study’s population was high-risk. A fatal flaw in the study was the small number of circumcised men in all of the countries other than the Philippines and small number of non-circumcised men in the Philippines. Of the twenty data cells from the five countries, seven had five or fewer subjects. The authors used parametric statistical methods, which are notably unreliable in this situation, to report the statistics on the combined data. These studies only sampled the glans, foreskin, and urethra. If the data are adjusted for the number of infections that would be missed from this limited sampling, the odds ratio, adjusted for study site using exact statistics, lowers the reported odds ratio of 3.41 (95%CI Z 1.56–8.03) to an odds ratio of 1.64 (95%CI Z0.87e3.18).
The second problematic study found that Hispanic men were at greater risk for genital HPV infection.19 Since Hispanic men are more likely to be non-circumcised, the authors predictably found a greater risk for genital HPV infection in non-circumcised men. When performing multivariate analysis, the authors inexplicably failed to control for ethnicity when determining the adjusted risk of circumcision status. Since sexual partners are not chosen at random, small, close communities can have accelerated increases in incidence and higher prevalences than the community at large. Controlling for ethnicity was a reasonable tool available to the authors to adjust for transmission rates in close communities, although the colinearity of ethnicity and circumcision status may be problematic. Without controlling for ethnicity it is impossible to determine whether being in the Hispanic community or circumcision status was responsible for the increased risk of genital HPV. This study also did not sample the penile shaft for HPV. Adjusting for this incomplete sampling, the results of the study are reversed (ORZ 0.63, 95%CI Z 0.40–0.98).
The third problematic study was published by Lajous et al.22 In this study fourteen men were identified as circumcised on physical examination, while 95 men identified themselves as being circumcised. Physical examination is considered the gold standard for assigning circumcision status. Instead of using physical examination as the measure of circumcision status, the study published the association between HPV infection and self-report of circumcision. Eighty-eight of the 95 men who reported themselves as circumcised were not circumcised.41
One limitation of this meta-analysis is the inability to find all sources of data using any search strategy. There may be published and unpublished studies not included in the analysis. Adjustment for confounding factors was missing in nearly all of the studies used in the analysis, so raw numbers were used.
The funnel graph and one of the measures for publication bias indicate the presence of publication bias. Since there has been no comparative analyses of the methods of detecting publication bias, it is recommended that all of the tests for publication bias be performed.42 When adjusted for publication bias using the ‘‘trim and fill’’ method, three studies in the right lower portion of the inverted funnel was found to not have a corresponding study in the left lower portion. Adjustment for publication bias needs to viewed with caution as asymmetry of a funnel plot may be due to factors other than publication bias, and, likewise, results generated to correct for the asymmetry may not reflect a correction for publication bias.43 An analysis of strengths and weaknesses of the different measures of publication bias has been performed by Macaskill et al.27
One of the drawbacks seen in some of the observational studies is that specific outcomes can have a small number of patients. When this occurs, the assumptions that allow one to make accurate inferences from the Mantel-Haenszel and GLIMMIX methods may no longer be valid. Likewise, if parametric methods are used to estimate odds ratios and 95% confidence intervals for studies with small cell sizes, these inaccurate estimates can bias summary effects using general-variance based methods. The instability propagated by studies with small cell sizes is reflected in variation of results each method delivers (not shown). In an attempt to minimize the bias of studies with cells of small size, the odds ratios and confidence intervals were calculated using exact methods. These values were then used to calculate a random-effects summary odds ratio using the general-variance based method. This method introduces less bias than models based on parametric assumptions.
In conclusion, the medical literature does not support an association between the prevalence of genital HPV and circumcision status when strict criteria for diagnosis are applied.
The Circumcision Information and Resource Pages are a not-for-profit educational resource and library. IntactiWiki hosts this website but is not responsible for the content of this site. CIRP makes documents available without charge, for informational purposes only. The contents of this site are not intended to replace the professional medical or legal advice of a licensed practitioner.
© CIRP.org 1996-2024 | Please visit our sponsor and host: IntactiWiki.