The Association between Cutaneous Squamous Cell Carcinoma and Betapapillomavirus Seropositivity: a Cohort Study

Cutaneous squamous cell carcinoma (SCC) is among the most commonly diagnosed cancers in people with fair skin. Human papilloma viruses (HPV) of the beta-genus (betaPV) are nonenveloped cutanotropic DNA viruses that have been associated with the development of SCC (1). At present, 31 different betaPV types have been fully sequenced (2–4).

Epidemiologic studies have shown that antibodies against the betaPV major capsid antigen, L1, can be found not only in the serum of patients with SCC or the precursor lesion actinic keratosis (AK) but also in people unaffected by these lesions (5–10). BetaPV L1 antibodies have been associated with both SCC and AK in case–control studies (6, 10–20), although the findings are somewhat inconsistent (reviewed in ref. 14). The most recent and largest case–control study found an association between betaPV antibodies and SCC, with an increasing risk for antibodies to multiple betaPV types as well as in people using glucocorticoids (20). It is difficult to assess the independent effect of specific viruses, due to the very high frequency of multiple infection (21), but to date, associations have mostly been identified between HPV8 and HPV38 and prevalent SCC (11–13,16, 17,19) A more recent study also found associations with HPV15 and HPV17 as well as with gammaPV types (18).

Case–control studies assess betaPV exposure at the same time as, or even after, diagnosis of the skin cancer. This prevents assessment of temporality, so the direction of any observed association cannot be determined. Longitudinal studies overcome this issue, but there has been only 1 report of a prospective pilot study published, in which there was no association found between baseline HPV antibodies and subsequent SCC in 39 patients (14). However, as the authors of this article pointed out, their study did not have adequate power to assess the association.

To further explore the issue, we aimed to assess the association between betaPV L1 antibodies and cutaneous SCC using data from a population-based cohort study in Nambour, Australia.

Participants were a subset of the study population of the Nambour Skin Cancer Study described in detail previously (22–24). Briefly, in 1986, 3,000 residents of Nambour, a subtropical township in Australia (latitude 26°S), who were aged 20 to 69 years were randomly selected from the Australian Electoral Roll (enrolment to vote is compulsory) and invited to participate in a skin cancer prevalence survey. From 1992 to 1996, 1,621 of the 2,095 participants of the 1986 survey took part in a trial of sunscreen application and beta-carotene supplementation for the prevention of skin cancer. All participants received full-body skin examinations by a dermatologist in 1992, 1994, and 1996 to ascertain the presence of actinic keratoses, skin malignancies, telangiectasia on the face, and elastosis of the neck. Skin lesions arising between these examinations, and from 1997 to 2007, were ascertained through the local pathology laboratories. Participants completed standardized questionnaires about sun exposure and other possible risk factors for skin cancer including past history of skin cancer. Blood was collected from a randomly selected subsample of participants in 1992 and from all consenting participants in 1996. A total of 1,388 participants had blood collected in either 1992 or 1996. Ethical approval for all aspects of the study was obtained through the Bancroft Centre Human Research Ethics Committee, Queensland Institute of Medical Research.

We tested serum samples for the presence of antibodies to the major capsid antigen L1 of HPV5, 8, 9, 14, 15, 17, 20, 21, 22, 23, 24, 36, 38, 47, 49, 75, 76, 80, 92, 93, and 96 by multiplex serology. This is an antibody detection method based on a glutathione S-transferase capture ELISA, in combination with fluorescent bead technology (25–27). Positive serology cutoff points were standardized at 200 MFI (mean fluorescence intensity; ref. 8).

Participants who had an SCC prior to their first serum measurement were excluded from this analysis (N = 77). We calculated the prevalence of betaPV antibodies for any betaPV type overall and for each of the 21 genotypes tested in 1992 and 1996. We estimated HRs for the association between SCC and the presence of any betaPV antibodies, the number of different antibody types, and selected specific antibody types (HPV5, 8, 9, 15, 20, 23, 24, 36, and 38) using a Cox proportional hazards model, adjusted for age and sex. The specific types were selected on the basis of their seroprevalence and also as they had been shown to be associated with SCC in previous studies. The date of entry into the cohort was the date at which serum antibodies were first measured (1992 or 1996), and the date of censoring was either the date of first SCC diagnosis, the date participant was lost to follow-up or 31 December 2007, whichever occurred first. If participants had both 1992 and 1996 serum measurements available, we used both records to allow for changing antibody status. A robust sandwich covariance matrix was used to account for intraperson correlation using the method described by Lin and Wei (28). We conducted additional analyses, within strata of age, sex, skin color, and tanning ability. Statistical analyses were performed using SAS 9.1.

After excluding people who did not have a blood sample collected in 1992 or 1996 and those who had SCC diagnosed prior to the start of the trial, 1,311 of the 1,621 people initially enrolled in the trial were included in this analysis, 176 with only a 1992 blood sample, 655 only a 1996 blood sample, and 480 with both 1992 and 1996 samples. Their mean age was 49 years (SD: 13, range: 25–75) and 44% were men (Table 1). SCC was diagnosed in 150 participants during follow-up. A single SCC developed in 97 people, and 53 participants developed more than 1 SSC (range: 2–20). Being aged over 50 years, having medium or fair skin color, a propensity to burn when exposed to the sun, and having a high degree of nuchal elastosis were significantly associated with the development of SCC (Table 1).

The overall prevalence of betaPV antibodies in 1992 was 62%, and 46% of participants were positive for multiple types (Table 2). The type-specific seroprevalence was highest for HPV8 (34%), HPV38 (33%), and HPV49 (26%) and lowest for HPV14 (2%) and HPV93 (3%). In 1996, 66% of participants were betaPV seropositive and 48% were positive for more than 1 type (Table 2). Type-specific prevalence varied from 31% for HPV38 to 1% for HPV14. The overall and type-specific prevalences were similar among the 480 people with both 1992 and 1996 sera available (Table 2). The betaPV antibody prevalence of the 480 people with serum samples collected in 1992 and 1996 was generally stable in the intervening 5 years, with only 13% of people changing their antibody status between 1992 and 1996. Seroconversion occurred in 46 people (10%), whereas 17 people (3%) seroreverted between 1992 and 1996.

No association was found between the presence of L1 antibodies to at least 1 betaPV type and the development of SCC [relative risk (RR): 1.0, 95% CI: 0.7–1.4; Table 3]. Also, no association was found with antibodies to multiple types (RR: 1.0, 95% CI: 0.7–1.5 for 1–3 antibody types; RR: 0.9, 95% CI: 0.6–1.4 for 4 or more antibody types). However, among people who were less than 50 years old in 1992, the presence of betaPV antibodies was associated with a 2-fold increased risk of SCC. This association was not evident in those older than 50 years. Stratification by sex, skin color, and burning ability showed no differences in associations between betaPV risk factors and SCC risk. There was no significant association between any of the individual betaPV types examined (HPV5, 8, 9, 15, 20, 23, 24, 36, and 38) and the development of SCC (Table 4). Seroconversion between 1992 and 1996 was not associated with the development of SCC after 1996, but with small sample size, the power to detect an association was limited.

The majority of previous studies finding associations between betaPV and SCC of the skin have been case–control studies (6, 11–13, 16, 17, 19, 20, 29). In this longitudinal study of 1,311 adults followed over 10 to 15 years, we did not observe any association with overall antibody positivity, antibodies to multiple betaPV types or to specific types previously shown to be associated with SCC. There was, however, an association among younger adults. Those under 50 years showed a 2-fold increased risk of SCC in the presence of betaPV antibodies.

The overall prevalence of betaPV antibodies of 62% in 1992 and 66% in 1996 was slightly higher than the 51% previously found in a similar population where the same laboratory technique and cutoff point were used for ascertainment of antibodies (27) and was higher also than the 46% reported in New Hampshire in the United States (20). As betaPV seropositivity increases with age (8, 27), and the mean age was lower in this study than in the previous studies (20, 27), the relatively high seroprevalence here is not due to older age and remains unexplained. The betaPV antibody prevalence of the 480 people with serum samples collected in 1992 and 1996 was quite stable in the short term (5 years), with only around 10% of people changing their antibody status in that time.

This is the largest longitudinal study to evaluate the relationship between betaPV antibodies and SCC, and it is consistent with a small prospective pilot study (cases = 39) in the United Kingdom in that both found no association between baseline serology and incident SCC (14). However, we did find an association in people diagnosed with SCC who were less than 50 years at entry into the study. It is possible that SCC diagnosed at younger ages is less strongly related to cumulative ultraviolet radiation exposure with HPV more likely to play a role than in older people. No previous studies have reported results stratified by age, so we do not know whether stronger associations have been apparent in younger age groups in previous studies.

Apart from the association in younger people, our overall results are not in accordance with those from recent large case–control studies which have showed generally positive associations (10–20). Understanding the relationship between HPV and skin cancer is hampered by lack of knowledge about the appropriate measure of infection. BetaPV are almost ubiquitous on the skin, with much higher prevalence of betaPV DNA than of antibodies, so some factor(s) other than simply the presence of betaPV on the skin must influence the development of antibodies. It has been hypothesized that the presence of SCC (or its precursor lesions, actinic keratoses) may result in increased viral load or local inflammation, resulting in presentation of the virus to the immune system and seroconversion (6, 10, 19). If this is the case, it is possible that the results of case–control studies using antibodies as a marker of infection do not indicate a causal association but are due to “reverse causality.” In support of this, in the previous small longitudinal study in the United Kingdom, there was a tendency toward higher seroprevalence in 15 prevalent cases than in 39 incident cases, although this was not statistically significant. However, we did not find that a diagnosis of SCC between 1992 and 1996 increased the likelihood of seroconversion in our study, and there is also no evidence that the diagnosis of AK leads to seroconversion in people without a diagnosis of SCC (30). Furthermore, the lack of association between the presence of antibodies and BCC (basal cell carcinoma) in case–control studies argues against the reverse causality hypothesis (20).

There may be other biases or uncontrolled confounding that have led to the associations observed in case–control studies. However, assuming that neither bias nor reverse causality is responsible and that there is a causal relation, the possible reasons for the overall lack of association in our longitudinal study need to be considered. It is possible that very high levels of sun exposure in Queensland overwhelm any detectable effect of betaPV. However, we have previously shown that the presence of persistent betaPV DNA in eyebrow hairs is associated with the development of AK in this population (31), providing support for an effect of long-term infection. We measured antibody status in mid-adulthood, and although antibodies appear to be relatively stable over a short time period, it is possible that this does not reflect the status at the time of initiation of SCC many years earlier (this limitation would also be true of case–control studies.) Finally, a high proportion of controls in this study had been diagnosed with AK. As these lesions are frequently not histologically diagnosed, we could not ascertain all AKs occurring during the follow-up period, so could not adjust for the presence of AKs. The fact that AKs are on the causal pathway between sun exposure and SCC, and possibly also between betaPV infection and SCC, may have made adjustment inappropriate even had it been possible. The complex interplay among betaPV, sun exposure, AK, and SCC therefore may be responsible for the lack of association between betaPV and SCC in people over 50 years of age at entry into this cohort.

In conclusion, whether betaPV infection of the skin, and indirectly betaPV antibodies, is involved in the oncogenic process in the general population remains unclear, and this study provides only limited support. Research incorporating measures of immune function and genotype, along with multiple measures of betaPV infection and sun exposure, may help to elucidate the role of betaPV in cutaneous carcinogenesis.

No potential conflicts of interest were disclosed.

The authors thank Dr. P. Buttner for critically reviewing the manuscript.

This study was supported by a project grant from the Cancer Council Queensland. E.I. Plasmeijer was supported by a travel grant from the Dutch Cancer Society (KWF). R.E. Neale is supported by an NHMRC (Aust) Career Development Award. M.C.W. Feltkamp was supported by the Netherlands Organization for Health Research and development (ZonMW, Clinical Fellowship grant).

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.