Risk Factors for Squamous Cell Carcinoma of the Penis—Population-Based Case-Control Study in Denmark

In the Western world, squamous cell carcinoma (SCC) of the penis is a rare malignancy occurring mainly among elderly men above age 60 years (1, 2) and with a standardized annual incidence of <2 per 100,000 men (1-4). In some developing countries, penile SCC is more common and affects a somewhat wider age range (5-9). The etiology of penile SCC remains incompletely understood (7, 10). Studies have consistently reported neonatal or childhood circumcision to be associated with reduced risk (1, 4, 6, 10-12), which corresponds geographically with reduced rates of penile SCC in populations practicing neonatal circumcision (13). The protective effect of childhood circumcision, but not of circumcision in adulthood (4, 11), seems to be attributable to the elimination of inflammatory conditions related to poor genital hygiene (2, 8-10), such as phimosis (3, 4, 14-16) and balanitis (14). Accordingly, an intact foreskin has been shown not to be associated with increased penile SCC risk in the absence of phimosis (4). Other previously suggested risk factors include infection with high-risk types of sexually transmitted human papillomaviruses (hrHPV; refs. 4, 11, 17) and current tobacco smoking (4, 11, 18). We undertook a nationwide case-control study in Denmark, a country with a largely uncircumcised male population (19) to further address the etiology of penile SCC.

Three study groups were identified, including case patients diagnosed with invasive or in situ penile SCC between 1993 and 1998 and two frequency-matched control groups consisting of male population controls and patients diagnosed with adenocarcinoma of the prostate in the same time period as the penile SCC patients. Information about newly diagnosed cases of penile SCC and prostate cancer was obtained every third month from the files of the Danish Cancer Registry and the Danish Pathology Registry, as well as through manual searches in the files of pathology departments across Denmark. Overall, we identified 259 men with penile SCC (215 invasive and 44 in situ cases) diagnosed between 1993 and 1998. Before inviting these men to participate in the study, we contacted the hospital department or the private practitioner responsible for their treatment to ensure that each eligible participant had been fully informed about his diagnosis, that he had no other health-related reason for nonparticipation known to the doctor (e.g., psychosis, hearing impairment), and that he was a Danish-speaking person. Of the original 259 invasive or in situ penile SCC patients, 88 patients (34%) had died, for 47 patients (18%), we received no reply from the hospital department or private practitioner responsible for the patient's treatment, and 3 patients (1%) were inaccessible due to emigration. Thus, we invited 121 (47%) penile SCC patients who had been informed about their cancer (or carcinoma in situ) and were alive at the time we aimed to approach them between 1997 and 2000. Prostate cancer controls were frequency matched to penile SCC patients on year of diagnosis and birth year (±5 y). From the Civil Registration System, a nationwide demographic database (20), we obtained a list of all Danish men born on two specific dates for each year between 1911 and 1963. As patients with penile cancer were invited to participate in the study, this list was used to select possible population controls, by means of frequency matching to invited penile SCC patients on year of birth. We anticipated somewhat lower participation rates among population controls, so to obtain the intended case-cancer control and case-population control ratios of 1:1, we invited more population controls than patients with penile cancer. Potential participants (121 penile SCC cases, 136 prostate cancer controls, and 172 population controls) received a letter of invitation with information about the study and were asked to sign and return a consent form in a prepaid envelope to participate in the study.

All consenting participants underwent a structured telephone interview between March 1997 and April 2000. The interviews, conducted according to written guidelines by six trained female medical students who were unaware of the study hypotheses and who were kept blinded to the case or control status of the participants, covered a broad range of topics including basic school attendance and postschool education, weight and height, tobacco and alcohol consumption, issues related to general and genital health, fertility, and details about sexual experiences and hygiene as well as venereal history. Patients with prostate cancer were chosen as a supplementary control group because most examined risk factors for penile SCC are not suspected to be risk factors for prostate cancer. Consequently, information provided by these control subjects was anticipated to be comparable with the information obtained from population controls. Moreover, we anticipated the reliability of answers to be high among prostate cancer controls because these men were expected to be similarly motivated as case patients to provide honest answers with respect to sexual and venereal history, genital hygiene, and other sensitive matters. Finally, possible recall problems were expected to apply equally to penile SCC patients and prostate cancer controls because both cancers originated in the male genitalia.

For 37 (52%) of the 71 penile SCC cases who participated in the study, we identified paraffin-embedded archival tumor specimens in the relevant pathology department. These tumor tissues were examined for HPV DNA by the PCR technique using general GP5+/6+ primers (GP5+/6+ PCR-EIA assay) as described in detail elsewhere (21, 22). In brief, paraffin-embedded tissues were cut for PCR analyses; outer sandwich sections were H&E stained and analyzed for the presence of tumor cells, whereas inner sections were used for DNA extraction and subsequent PCR analyses. All DNAs were tested for B-globin PCR positivity after agarose gel analysis to check for the quality of DNA for PCR purposes (23). HPV was detected using the GP5+/6+ PCR-EIA using cocktail probes for 14 hrHPV types and 23 low-risk HPV types as previously described (22). HPV-positive samples were subsequently analyzed for their HPV types using the reverse line blotting method. The following hrHPV types were detected and identified as follows: HPV 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66, and 68; low-risk HPVs detected were HPV 6, 11, 26, 34, 40, 42, 43, 44, 53, 54, 55, 57, 61, 70, 71 (CP8061), 72, 73, 81 (CP8304), 82/MM4, 82/IS39, 83 (MM7), 84 (MM8), and CP6108.

To ensure comparable exposure ascertainment periods in all 3 groups of study participants, we assigned a pseudo-year of diagnosis between 1993 and 1998 to population controls according to the distribution of year of diagnosis among penile SCC patients. Specifically, the distribution of pseudo-year of diagnosis among population controls was similar to the distribution of year of diagnosis among penile SCC cases (i.e., similar proportions of population controls having pseudo-year of diagnosis in 1993, 1994, 1995, 1996, 1997, and 1998 as penile SCC patients with year of diagnosis in 1993, 1994, 1995, 1996, 1997, and 1998, respectively). We considered only those exposures that preceded the pseudo-year of diagnosis to be relevant in population controls, similar to the situation in penile SCC cases and prostate cancer controls, for whom only exposures that preceded the year of cancer diagnosis counted as exposures of potential etiologic relevance.

Univariate Analyses. Because most risk factors are similar for both diagnostic entities (4), we considered as our case group all patients with invasive or in situ penile SCC. Initially, we calculated univariate odds ratios (OR) separately for penile SCC patients versus each of the two control groups, adjusting only for age at diagnosis in 10-y age groups (< 40, 40-49,…, ≥70 y). ORs obtained with either of the two control groups were generally similar, so we subsequently tested the appropriateness of combining the two control groups by doing a logistic regression analysis restricted to prostate cancer controls and population controls, with adjustment for age in 10-y age groups. This analysis revealed that none of the examined explanatory variables were associated with statistically significantly increased risk of prostate cancer, so to reduce complexity and gain statistical power, we combined the two control groups in all subsequent analyses. Logistic regression analyses were done to calculate exposure ORs for penile SCC versus the combined group of control subjects, again with adjustment only for age at diagnosis in 10-y age groups. The results of these age-adjusted analyses are called univariate ORs. Likelihood ratio tests for homogeneity determined possible differences in risk between exposure categories. For categorized continuous variables, we also did likelihood ratio tests to determine the appropriateness of reducing the observed associations to a linear specification. Trend tests were done only when the preceding likelihood ratio test for linearity yielded a reasonable fit (P > 0.05). In tests for trend, categorized continuous variables were treated as continuous variables, using the median for each category as the category value.

Multivariate Analyses. All explanatory variables with a P value of <0.10 in the test for homogeneity in the univariate analysis were entered simultaneously in a logistic regression model along with age in 10-y age groups. Using backward elimination to remove the least significant variables one by one until all remaining variables in the model were statistically significant (P value in test for homogeneity < 0.05), we arrived at a multivariate model. Subsequently, excluded explanatory variables were given an extra chance to enter the final model by means of forward inclusion, but none of the previously excluded variables reached statistical significance. Tests for homogeneity and tests for trend were done as for the univariate analyses. Comparisons of proportions were made using the χ² test or, when numbers were small (<5 in any one cell), by means of Fisher's exact test. Throughout, two-sided P values of <0.05 and 95% confidence intervals (CI) excluding unity were considered indicators of statistical significance. All logistic regression analyses were carried out in Statistical Analysis System software Version 9.1 (SAS Institute).

The study was approved by The Danish National Committee on Biomedical Research Ethics, approval no. C-1995-20, and The Danish Data Protection Agency, approval no. 1995-1200-187. Delayed reporting was due to shortage of funding.

Of the 121 patients with invasive or in situ penile SCC, 136 prostate cancer controls and 172 male population controls who were invited to participate in the study, 71 penile SCC cases (58.7%), 86 prostate cancer controls (63.2%), and 103 population controls (59.9%) agreed to participate in the telephone interview. The most common reasons for nonparticipation were unspecified unwillingness (22.3% of invited penile SCC cases, 15.4% of invited prostate cancer controls, and 26.7% of invited population controls), illness or death (8.3%, 13.2%, and 7.0%, respectively), or lack of telephone (10.7%, 8.1%, and 6.4%, respectively).

Men who agreed to participate in the study were slightly younger than those who declined the invitation to participate. The median age at cancer diagnosis was 62 years among participating penile SCC patients (64 years among 53 patients with invasive SCC, 57 years among 18 patients with SCC in situ), and 68 years among prostate cancer controls, whereas the corresponding median age at the pseudo-year of diagnosis was 62 years among population controls. Corresponding median ages among 50 penile SCC cases, 50 prostate cancer controls, and 69 population controls who declined the invitation to participate in the study were, respectively, 4, 2, and 5 years older. Of the 37 penile SCC patients whose tumor tissues were PCR examined for the presence of HPV DNA, 24 (65%) were hrHPV positive, and 1 (3%) was positive to a low-risk HPV type (HPV6; Table 1). By far, the predominant HPV type was HPV16, which was detected in 22 (59.5%) of the 37 examined tumors, corresponding to 92% of the 24 hrHPV-positive tumors. All 9 examined lesions from patients with penile SCC in situ were hrHPV positive versus 15 (53.6%) of 28 lesions from patients with invasive penile SCC (Fisher's exact test; P = 0.02, two sided). Penile SCC lesions involving the preputium were slightly more likely to be hrHPV positive (71% of 14) than lesions of the penile glans (61% of 23) and lesions of unspecified penile location (60% of 5), but these differences were not statistically significant (X² = 0.47; degrees of freedom = 2; P = 0.79). A series of 12 tumor specimens from controls with prostate cancer subjected to the same PCR analysis as the penile SCC cases were consistently negative for all examined HPV types (Table 1).

Sexual and Venereal Risk Factors. Measures of early and high sexual activity were positively associated with penile SCC risk (Table 2). Statistically significant trends were seen for lifetime number of female sexual partners (P_trend = 0.002), notably the number of female sexual partners before age 20 years (P_trend < 0.001), age at first sexual intercourse (P_trend = 0.009), number of female partners performing oral sex (P_trend < 0.001), and female partner's lifetime number of other sexual partners (P_trend = 0.05). A history of anogenital warts (OR, 6.34; 95% CI, 2.87-14.0) and ever having been HIV tested (OR, 2.76; 95% CI, 1.06-7.23) were also significantly linked with increased penile SCC risk (Table 2).

Penile Conditions. There was no significant difference in the proportion of participants who had undergone childhood circumcision between patients with penile SCC (0 of 71) and controls (5 of 189; Fisher's exact test, P = 0.33, two sided). However, phimosis (OR, 3.39; 95% CI, 1.62-7.11), balanitis (OR, 3.07; 95% CI, 1.36-6.93) and priapism (OR, 4.90; 95% CI, 1.36-17.7) were each significantly associated with risk of penile SCC (Table 3). For each of the 7 penile SCC patients who reported a history of priapism, >10 years had passed from the priapic episode to the time of penile SCC diagnosis (mean, interval 27 years; range, 10-62 years). All 4 controls reporting priapism were younger than 26 years (on average, 19 years; range, 16-25 years) at the time they experienced priapism, whereas the case patients were on average ages 32 years (range, 15-60 years) at their reported priapic episode.

Social and Nonsexual Risk Factors. Although smoker status (lifelong nonsmoker, former smoker, or current smoker) was not associated with risk, cumulative tobacco consumption was dose-dependently associated with risk of penile SCC (P_trend = 0.009). There was no dose-response relationship between alcohol consumption and risk of penile SCC, but individuals reporting no alcohol consumption were at significantly reduced risk (OR, 0.27; 95% CI, 0.07-0.98 for 0 versus >20 consumption-years). School attendance and postschool education were not associated with risk (Table 4).

Sexual and Venereal Risk Factors. Associations with the lifetime number of female partners (P_trend = 0.01), the number of female partners before age 20 (OR, 5.11; 95% CI, 1.89-13.8; for ≥4 versus 0 female partners, P_trend < 0.001), age at first sexual intercourse (P_trend = 0.01), and the number of female partners performing oral sex (OR, 3.65; 95% CI, 1.14-11.7; for ≥3 versus 0 oral sex partners, P_trend = 0.04) all remained statistically significant in multivariate analysis, as did a history of anogenital warts (OR, 7.01; 95% CI, 2.48-19.8; Table 2). Never having used condoms (OR, 2.45; 95% CI, 1.04-5.80) became statistically significant after adjustment, whereas ever having been HIV tested and female partner's history of other sexual partners lost significance upon adjustment for confounders (Table 2).

Penile Conditions. Associations with histories of phimosis (OR, 4.91; 95% CI, 1.85-13.0) and priapism (OR, 9.23; 95% CI, 1.09-78.3) became stronger in multivariate analyses, whereas the association with balanitis lost statistical significance (OR, 2.37; 95% CI, 0.81-6.92). Most participants had noticed the presence of smegma under their foreskin at some point in their life. However, the presence of smegma was not associated with penile SCC risk (Table 3).

Social and Nonsexual Risk Factors. The significant dose-response relationship with cumulative tobacco consumption disappeared after adjustment for confounding variables (P_trend = 0.22; Table 4). Although alcohol abstinence remained inversely associated with the risk of penile SCC (OR, 0.18; 95% CI, 0.04-0.84; for 0 versus >20 consumptions-years), there was no dose-response relationship between alcohol consumption and penile SCC risk (Table 4).

Our study adds to the accumulating international evidence that sexually transmitted hrHPVs, notably HPV16, are important etiologic factors in penile SCC (4). Of tumors tested in our study, more than half (59.5%) were DNA positive for HPV16, constituting 92% of all hrHPV-positive tumors. These findings are in agreement with those of Daling et al. (4) who observed 69% of 94 examined penile SCCs to be DNA positive for HPV 16 (87% of all HPV-positive tumors), and findings in a recent study by Pascual et al. (17) who reported that 65% of 49 penile SCCs were HPV 16 positive (84% of all HPV-positive tumors). The central involvement of oncogenic HPV infection in the etiology of penile SCC was further emphasized by our finding that, among a number of sexually transmitted diseases, only a history of anogenital warts obtained status as an independent risk factor in our study. Specifically, a history of anogenital warts was reported by 33% of penile SCC patients as contrasted by only 6% to 7% of controls, corresponding to an OR of 7.0, which is close to the OR of 7.6 reported for genital warts in a Seattle-based case-control study (4).

Our findings support the view that transmission of the responsible etiologic agents, presumably HPV16 and other hrHPVs, usually takes place through intimate sexual contact. We observed a statistically significant trend of increasing risk with increasing numbers of female sexual partners. Notably the partner number before age 20 years seemed to differ between penile SCC patients and controls in our study. To the best of our knowledge, our study is the first to show a strong and statistically significant association between the practice of heterosexual penile-oral sex and risk of penile SCC, suggesting that oral sex, a common sexual practice reported by almost half of our population controls, may be a major and previously underestimated means of viral transmission from the female partner's oral cavity/pharynx to the penis. Men who had practiced anal sex with two or more female partners were also at increased risk, but numbers were unstable. Although some study participants may have been reluctant to report certain sexual details, the finding that men with same-sex experience are not at increased risk of developing penile SCC accords well with prior reports of no conspicuous penile SCC risk in gay men (4, 24).

Circumcision in childhood is uncommon in Denmark with an estimated overall proportion of <2% of boys having undergone the procedure before age 15 years in recent decades (19). No patient with penile SCC had been circumcised in childhood or adolescence versus 5 (2.6%) of 189 controls, a difference that was not statistically significant. Self-reported phimosis, however, was reported by 27% of penile SCC patients versus 10% to 11% of controls, thus corroborating the current view that having an unretractable foreskin, notably in combination with penile hrHPV infection, may constitute the single most important risk factor for penile SCC (1). Balanoposthitis, inflammation of the penile glans and foreskin, has also been linked to penile SCC risk in previous studies with ORs of 3.5 (95% CI, 1.5-8.5) and 9.5 (95% CI, 5.2-17.2), respectively (4, 14). With 22% of penile SCC patients versus only 7% to 8% of controls reporting a history of balanitis, our study confirms this association, although it was not statistically significant in the multivariate analysis. However, unlike in some other studies (4, 25) the occasional presence of smegma under the foreskin, which was noticed by a majority of all participants, was not associated with penile SCC risk in the present study.

A history of penile trauma was not reported more frequently by penile SCC patients than controls in our study. This is in contrast to findings in a case-control study in Western Washington/British Columbia, where researchers reported statistically significant ORs of 3.2 (95% CI, 1.5-6.8) and 5.2 (95% CI, 3.1-8.7) for penile injury and penile tear, respectively (4). The reason for this difference between studies is unclear.

Tobacco smoking, particularly current smoking, has been reported in a number of studies to be linked to increased risk of penile SCC (4, 10, 11, 14, 18, 26). For instance, current smokers were at double risk of penile SCC compared with lifelong nonsmokers in the Seattle-based case-control study (4). However, not all studies are in support of tobacco smoking as an important etiologic factor. A case-control study in China failed to find support for a causal role of tobacco smoking (25), and researchers in Sweden reported only a weak, positive association with current consumption of >10 cigarettes per day (14). In the present study, current smoking was not associated with penile SCC risk and cumulative tobacco consumption was associated with risk only in univariate analysis, but upon adjustment for confounders in the multivariate model, there remained no significant role for tobacco consumption. Differences in reported ORs linking tobacco smoking to penile cancer risk is likely to be partly due to differences in smoker prevalence across countries. In the present Danish study, 40% of population controls were current smokers, which is almost twice the proportion (22%) of current smokers among population controls in the Seattle-based study (4). Considering the much higher smoker prevalence in Denmark, smoking may be less strongly associated with other social and behavioral penile SCC risk factors among Danes than among citizens in areas with a lower smoker prevalence. Theoretically, some of the previously reported effect of tobacco smoking might reflect associations with behavioral risk factors that study participants were reluctant to provide honest answers about.

Our finding that priapism may be a risk factor for penile SCC is potentially interesting but requires cautious interpretation. During interviews, a nontrivial 10% of penile SCC patients affirmed a question of whether they had ever experienced an unwanted penile erection of long duration that would not wear off. In contrast, only 4 (2.1%) of 189 controls reported ever having had such an experience, thus resulting in a statistically significant multivariate OR of 9.2. We deliberately excluded those priapic episodes that occurred <5 years before the time of diagnosis to avoid inclusion of cancer-related priapism (27-29). All seven penile SCC patients with a positive history reported that the priapic episode antedated the cancer diagnosis by 10 or more years, thus reducing the possible distorting influence of priapism secondary to malignancy. The lack of prior supporting evidence of a causal connection constitutes no strong argument against an association because to the best of our knowledge, no prior penile cancer case-control study has examined the possible role of priapism. Obviously, further data from other settings are needed before a causal association of priapism with later penile SCC development can be established.

Approximately one third of tested penile SCCs were HPV negative, suggesting that a proportion of penile cancers may arise through causal pathways that do not involve infection with anogenital types of HPV. Theoretically, rare HPV types that were not included in the broad panel of HPV types examined might have been present in some of our negative cases. However, some subtypes of penile SCC are probably genuinely HPV negative, notably those with keratinizing SCC histology (30). Some cancers arising in the penile skin may resemble cutaneous SCCs elsewhere on the body, which are not etiologically related to HPV16 or other mucosotropic types of HPV. Unfortunately, the limited number of HPV-tested penile SCC patients in our study precluded meaningful risk factor analyses stratified by HPV status.

Our study has some limitations that need to be considered. Participation rates were modest (around 60%) among penile SCC patients and among prostate cancer and population controls. We share these difficulties in recruiting penile SCC patients with other case-control studies whose participation rates among invited penile SCC patients ranged between 50% and 70% (4, 10, 11), although very high levels of participation (respectively, 90% and 91%) have been seen in case-control studies carried out in China (25) and Sweden (14). Theoretically, the ∼40% of invited penile SCC patients and controls who declined the invitation to participate in this and other studies could represent socioeconomic or behavioral segments of the population that would make generalization to the entire population problematic. Participants in our study were slightly younger than those who declined the invitation to participate in the study, a general pattern that applied to all three groups of study participants. Also, by study design, participants were restricted to those invited subjects who were in possession of a telephone, a criterion that precluded the participation of only slightly different proportions of invited penile SCC patients (10.7%) and controls (8.1% of prostate cancer controls, 6.4% of population controls), thus making telephone ownership an unlikely major biasing factor. However, penile SCC patients in our study were diagnosed during the study period 1993 to 1998 so, because interviews only started in 1997, participating penile SCC patients might be somewhat skewed toward patients with localized cancers. We cannot exclude that this might have introduced bias in some analyses. Specifically, to the extent that some of the studied risk factors have a negative effect on survival after penile cancer, some of the ORs in our study may be somewhat conservative. Another limitation is the limited number of penile SCC patients, which made findings in some of our statistical analyses unstable.

Our study also had some methodologic assets. It was population-based, which adds to the generalizability of our findings and, unlike other case-control studies, our study benefited from having two independent control groups. Only after (a) having seen that risk factor associations were fairly similar in initial logistic regression analyses obtained with the two control groups separately, and (b) after having formally examined the appropriateness of combining the two control groups in a logistic regression analysis comparing the two control groups, did we combine them to gain statistical power, thus adding robustness and credibility to the reported risk factor associations. Another advantage of our double control group study design, which has been used previously in case-control studies of risk factors for anal and vulvo-vaginal cancers (21, 31), is that blinding of interviewers to the case or control status of participants is more easily maintained. Expectations among interviewers may color the way answers are perceived and noted on the questionnaires, but interviewers in our study could not plausibly distinguish between interviews conducted with penile SCC cases and those conducted with prostate cancer controls. Any influence of interviewer bias is therefore likely to have been small and nondifferential.

In conclusion, our study provides population-based evidence that sexually transmitted hrHPV infections, notably infections with HPV16, constitute a central risk factor for penile SCC and that heterosexual penile-oral sex may be a hitherto unrecognized means of viral transmission. Phimosis was confirmed as a strong risk factor, but the novel association with priapism needs replication in other studies.

Dr. Frisch has served as a member of a national advisory board for Glaxo Smith Kline, the manufacturer of a vaccine against HPV 16 and HPV 18. The other authors disclosed no potential conflicts of interest.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.