β-HPV indicates β human papillomavirus; OR, odds ratio.
The squares and horizontal lines correspond to the study-specific odds ratios (ORs) and 95% CIs. The diamond represents the pooled OR and 95% CI of the overall population. The vertical dashed line indicates the overall pooled OR of 1.33. cSCC, cutaneous squamous cell carcinoma; HPV, human papillomavirus; NA, not available; OR, odds ratio; RR, relative risk.
The vertical solid line represents the summary effect estimates, and the dotted lines are pseudo 95% CIs
The squares and horizontal lines correspond to the study-specific ORs and 95% CIs. The diamond represents the pooled OR and 95% CI of the overall population. The vertical dashed line indicates the overall pooled OR of 1.35. HPV indicates human papillomavirus; NA, not available; OR, odds ratio
The squares and horizontal lines correspond to the study-specific ORs and 95% CIs. The diamond represents the pooled OR and 95% CI of the overall population. The vertical dashed line indicates the overall pooled OR of 1.35.
HPV indicates human papillomavirus; NA, not available; OR, odds ratio
Chahoud J, Semaan A, Chen Y, Cao M, Rieber AG, Rady P, Tyring SK. Association Between β-Genus Human Papillomavirus and Cutaneous Squamous Cell Carcinoma in Immunocompetent Individuals—A Meta-analysis. JAMA Dermatol. Published online December 30, 2015. doi:10.1001/jamadermatol.2015.4530
Existing epidemiological evidence remains controversial regarding the association between β-genus human papillomavirus (β-HPV) and cutaneous squamous cell carcinoma (cSCC) in immunocompetent individuals.
We aimed to clarify this association and evaluate type-specific β-HPV involvement.
We performed a systematic literature search of MEDLINE and EMBASE for studies in humans through June 18, 2014, with no restriction on publication date or language. The following search terms were used: “human papillomavirus” and “cutaneous squamous cell carcinoma or skin squamous cell carcinoma or cSCC or nonmelanoma skin neoplasms.”
Articles were independently assessed by 2 reviewers. We only included case-control or cohort studies, in immunocompetent individuals, that calculated the odds ratio (OR) for cSCC associated with overall and type-specific β-HPV.
Data Extraction and Synthesis
We first assessed the heterogeneity among study-specific ORs using the Q statistic and I2 statistic. Then, we used the random-effects model to obtain the overall OR and its 95% CI for all studies as well as for each type of HPV. We also tested and corrected for publication bias by 3 funnel plot–based methods. The quality of each study was assessed with The Newcastle Ottowa scale.
Main Outcomes and Measures
Pooled ORs and 95% CIs for overall β-HPV and HPV types 5, 8, 15, 17, 20, 24, 36, and 38 association with skin biopsy proven cSCC.
Seventy-nine articles were assessed for elligibility; 14 studies met inclusion criteria for the meta-analysis and included 3112 adult immunocompetent study participants with cSCC and 6020 controls. For all detection methods, the overall association between β-HPV and cSCC was significant with an adjusted pooled OR (95% CI) of 1.42 (1.18-1.72). As for the type-specific analysis, types 5, 8, 15, 17, 20, 24, 36, and 38 showed a significant association with adjusted pooled ORs (95% CIs) of 1.4 (1.18-1.66), 1.39 (1.16-1.66), 1.25 (1.04-1.50), 1.34 (1.19-1.52), 1.38(1.21-1.59), 1.26 (1.09-1.44), 1.23 (1.01-1.50) and 1.37 (1.13-1.67) respectively. Our subgroup analysis in studies using only serology for HPV detection showed a significant association between overall β-HPV and HPV subtypes 5, 8, 17, 20, 24, and 38 with an increased risk of cSCC development.
Conclusions and Relevance
This study serves as added evidence supporting β-HPV as a risk factor for cSCC in healthy individuals. The subgroup analysis highlights this significant association for HPV 5, 8, 17, 20, and 38, which may help to direct future prevention efforts.
Cutaneous squamous cell carcinoma (cSCC) is one of the most common cancers in men and women worldwide, with more than 700 000 newly diagnosed cases yearly in the United States alone, compared with less than 15 000 newly diagnosed cases yearly of cervical cancer.1,2 The annual incidence of cSCC has increased at an alarming rate in the last 3 decades, with around 8000 attributable deaths yearly, twice the death rate from invasive cervical cancer. Furthermore, the estimated annual cost of treating cSCC cases in the United States is about $3.8 billion. This highlights the important public health burden that cSCC places on our health care system.3,4 Therefore, a deeper understanding of cSCC’s risk factors should help us to develop more effective preventive measures, in turn decreasing the number of newly diagnosed cases and the financial burden.
The known risk factors implicated in the development of cSCC are UV radiation exposure, immunosuppression, and fair skin.5 In the context of the recent increase in the rates of newly diagnosed cSCC, a viral etiology has been hypothesized, with human papillomavirus (HPV) being the major incriminated virus.6- 8 Human papillomaviruses are a large and diverse group of more than 170 subtypes with 5 major HPV genera: α papillomavirus, β papillomavirus, γ papillomavirus, μ papillomavirus and ν papillomavirus.9,10 β-Genus HPV (β-HPV) is the most detected genus in cancerous, precancerous, and normal keratinocytes. The analysis of human papillomatous skin lesions and their relationship to virus infections and carcinogenesis had a slow start because they were considered a cosmetic problem with no significant medical implications. This view gradually changed after 1922, when Lewandowsky and Lutz described a hereditary condition characterized by an extensive verrucosis, called epidermodysplasia verruciformis (EV). The first description of β-HPV infection and cutaneous carcinoma was mainly the work of Stefania Jablonska, who pointed out the potential role of HPV 5 and 8 in these warts as causal factors for the subsequent development of cSCC.11,12 This later led the International Agency of Research on Cancer (IARC) to consider β-HPV types 5 and 8, found in 90% of cSCC lesions of EV cases, as possibly carcinogenic. Also, the increased occurrence of cSCC in solid organ transplant recipients has been associated with significantly higher rates of β-HPV.13,14 More recently, the association of β-HPV and cSCC in immunocompetent individuals was evaluated in multiple epidemiological case-control studies6,8 with controversial results.
The molecular pathways explaining β-HPV’s implication in the carcinogenesis of cSCC are not yet fully clarified and could be explained by a number of mechanisms of action, including the following 3 pathways: (1) Increased susceptibility to UV-induced oncogenesis in transgenic mice expressing HPV type 38 E6 and E7 oncoproteins15- 17; (2) HPV type 8 E6 oncoprotein’s capacities to inhibit the PDZ (Psd95-DlgA-ZO1) domain protein syntenin-2, a critical element in the control of viral oncogenic potential (the downstream pathway for syntenin-2 remains to be fully understood)18,19; and (3) β-HPV types 5, 8, 20, and 38 through E2, 6, and 7 oncoprotein increase the quantity of stem cell–like cells available during early carcinogenesis, thus enabling the persistence and accumulation of DNA damage necessary to generate malignant stem cells. These pathways are only a few of the many plausible molecular mechanisms for type-specific β-HPV involvement in the initiation and progression of the onocogenic process in cSCC. However, type-specific β-HPV epidemiological evidence remains controversial. Thus, the primary endpoint of this meta-analysis is to evaluate the existing epidemiological data on type-specific β-HPV association with cSCC in immunocompetent individuals.
We systematically searched for studies in humans through June 18, 2014, with no language restrictions or specified start date. We searched the MEDLINE and EMBASE databases using the following search terms: “human papillomavirus or HPV or β-HPV” and “cutaneous squamous cell carcinoma or skin squamous cell carcinoma or cSCC or nonmelanoma skin neoplasms.” The initial inquiry was conducted by one of us (J.C.) and independently verified by the university medical reference librarian. In addition, to ensure comprehensiveness, we examined the reference lists from retrieved articles for supplementary relevant studies. The results were uploaded to Mendeley, a citation database program for review and selection.
Eligibility was restricted to studies, with human immunocompetent participants only, that examined the association between HPV and cSCC development as primary episode. Studies were excluded if their outcome of interest was the development of melanoma, cutaneous basal cell carcinoma, anogenital cSCC, or recurrent cSCC. As for study design, case-controls and cohorts of at least 40 patients were eligible for inclusion, while we excluded abstracts, letters to the editor, case reports, review articles, case-controls (<20 cases and <20 controls), and cohort studies (<40 patients). Studies including less than 40 patients were excluded, because they lacked statistically significant power. Furthermore, studies not reporting the associated risk for at least 2 of the HPV subtypes 5, 8, 15, 17, 20, 24, 36, and 38 were excluded from the analysis. To be included in our analysis, studies had to report type-specific β-HPV odds ratio (OR) or relative risk (RR) with 95% CIs. Restrictions were not placed on the method of detection of HPV.
Two of us (J.C.) and (A.S.) independently reviewed the titles and abstracts of the previously searched databases. Based on the prespecified selection criteria, both authors independently identified studies. Disagreements were resolved by discussion with a prior agreement that any unsettled conflict would be determined by a third author (A.G.R.).
Data collection forms were used by both authors to extract the required data from eligible studies. Then both authors assessed extracted studies for duplication by comparing authors’ names, dates of publication, and population sizes. Both authors were unblinded to the studies’ authors’ names, population sizes, journals of publication, and locations.
Primary outcome measures were ORs with the corresponding 95% CIs for the association of β-HPV and primary episodes of cSCC. We identified the ORs reflecting the greatest degree of adjustment for possible confounding factors. The adjusted ORs with 95% CIs for overall β-HPV, HPV types 5, 8, 15, 17, 20, 24, 36, and 38 were extracted, when applicable. Other data of interest included study general information: first author name, year of publication, location, and estimated latitude where the study population was enrolled. Information on study design, population size and characteristics, the method of HPV detection, and the number of β-HPV subtypes analyzed was also collected. In certain circumstances when required data were not reported, authors corresponded with study authors via email.
Two of us (A.S. and J.C.) independently used the Newcastle-Ottawa Scale (NOS) for assessing the individual quality of each study. We used ORs to approximate RR given the rare disease assumption. We first assessed the heterogeneity among study-specific ORs using the Q statistic and I2 statistic. We also tested and corrected for publication bias by 3 funnel plot–based methods: the Egger test, the Begg test, and the trim & fill method.
The meta-analysis was performed on qualifying studies that had reported adjusted ORs of the association between global and type-specific β-HPV with cSCC regardless of the detection method. As for the subgroup analyses, they were restricted to studies using seroprevalence detected by multiplex serology or enzyme-linked immunosorbent assay (ELISA). The adjusted ORs with 95% CIs were identified based on the meta-analysis results. Random-effects method was used to pool the ORs and 95% CIs. All study analyses were performed using the R (R Development Core Team, 2008) and metafor package (Wolfgang Viechtbauer, 2010).
The literature search of the MEDLINE and EMBASE databases yielded 916 articles. After review of the titles and abstracts, 837 articles were excluded for lack of adherence to our inclusion criteria. We reviewed the full text of the selected 79 articles and assessed their reference lists for relevant publications, retrieving 8 additional publications. Based on this review, we excluded 67 publications for nonadherence with the inclusion criteria. Furthermore, 1 publication was excluded for data overlap with another study, and we included the one with the highest adjustment for OR in the meta-analysis. We also excluded 5 publications for not providing the associated risk of β-HPV subtypes, based on our correspondence with the authors. The flow diagram of the systematic literature review is illustrated in Figure 1.
The studies included in our meta-analysis comprised 12 case-control studies and 2 cohort studies (Table).11,20- 32 Publication years ranged from 2003 to 2014, with data collected from as late as 1992. All studies included adult immunocompetent participants, males and females of all ages, a total of 3112 cases and 6020 controls. Our subgroup analysis comprised 12 studies that exclusively used serology as the HPV detection method and included 2789 cases and 5359 controls. Because UV light exposure is a major risk factor for cSCC and has been implicated as a cofactor with β-HPV in the initiation of carcinogenesis in multiple molecular models, we evaluated the latitude of each study region.33 We determined that 11 studies were conducted in the northern hemisphere between estimated latitudes 40° N to 60° N, while only 2 studies were conducted in the southern hemisphere at an estimated latitude of 26° S. All included studies assessed cSCC as primary outcome in both controls and cases by skin biopsy pathological evaluation. All 14 studies adjusted for age and sex when calculating the β-HPV and cSCC associated ORs. Only 8 studies also adjusted for 1 or more of the following: skin sensitivity, region of residence, eye color, number of lifetime painful sun burns, smoking history, or lifetime exposure to sunlight. The most used detection technique for different types of HPV was serology by multiplex polymerase chain reaction (PCR) (8 studies); the ELISA assay was used in 4 studies and eyebrow hair DNA in 5 studies. ELISA serology is a less sensitive and specific method, compared with multiplex serology, and was used only in studies published between 2003 and 2006 when multiplex serology was not readily available. Serology is a highly cost-effective HPV detection method, because it allows for large-scale studies as well as tracking a population over time for analyses in HPV infections. Previous studies32,34 have shown acceptable comparability in sensitivity between β-HPV collected by eye browhair, serology, and surgical biopsy. The selection of the β-HPV subtypes included in both our meta-analysis and subgroup analysis was based on the number of studies evaluating each subtype: β-HPV 8 and 15 (13 studies); β-HPV 5, 20, 24, and 38 (12 studies); β-HPV 36 (10 studies); and β-HPV 17 (7 studies). HPV type 17 is one subtype of interest that was not evaluated in most studies. Additional information and extracted data from the included studies are presented in eTable 1 in the Supplement.
Quality assessment was performed using the Newcastle Ottowa scale, which is specifically used for nonrandomized studies and has been endorsed by the Cochrane collaboration. We adequately used the version for case-control studies or cohort studies as applicable, addressing subject selection, study comparability, and the assessment of outcome or exposure. NOS scores from 6 to 9 (9 being the highest possible score), with a mean of 7.8, median and mode of 8. All studies earned a star for comparability with regard to age and sex adjustment. Eight studies received an additional star for comparability, because they also adjusted for skin sensitivity, region of residence, or lifetime exposure to sunlight. A summary of the included study evaluation as assessed using the NOS is shown eTable 2 in the Supplement.
Publication bias was not detected for the pooled studies, or the subgroup analysis studies using the Egger test35 and the Begg test.36 We also conducted sensitivity analyses on both the pooled studies and the subgroup data sets to evaluate the potential impact of publication bias on the conclusions. We specifically applied the trim and fill method to evaluate the subgroup analysis data set and imputed 5 studies.37 The corrected ORs remained similar to the uncorrected ones, suggesting that the impact of the publication bias for this subgroup is small.
Our meta-analysis comprised a total of 14 studies (3112 cases; 6020 controls). In the pooled analysis, overall β-HPV–cSCC association was significant with an adjusted pooled OR of 1.4 (95% CI, 1.2-1.7). For the type-specific analysis, types 5, 8, 15, 17, 20, 24, 36, and 38 showed a significant association with adjusted pooled ORs of 1.4 (95% CI, 1.2-1.7), 1.4 (95% CI, 1.2-1.7), 1.2 (95% CI, 1.0-1.5), 1.3 (95% CI, 1.2-1.5), 1.4 (95% CI, 1.2-1.6), 1.3 (95% CI, 1.1-1,4), 1.2 (95% CI, 1.0-1.5), and 1.4 (95% CI, 1.1-1.7) respectively (Figure 2 and Figure 3). A random-effects model was used because heterogeneity was identified among the 13 studies. Visual inspection of the funnel plot revealed no publication bias, later confirmed by the Begg adjusted rank correlation test (Figure 4). The funnel plots and their respective Begg adjusted rank correlation tests are represented in supplement eFigures 1-9 in the Supplement.
Our subgroup meta-analysis comprised a total of 12 studies (2789 cases; 5359 controls). The pooled analysis of the overall β-HPV association with cSCC was significant, with adjusted pooled ORs of 1.4 (95% CI, 1.2-1.8), an increased associated risk of 45% for cSCC development. In respect to the type-specific pooled analysis, types HPV 5, 8, 17, 20, 24, and 38 were significantly associated with increased risk of cSCC; the adjusted pooled ORs were 1.4 (95% CI, 1.1-1.8), 1.5 (95% CI, 1.2-1.8), 1.4 (95% CI, 1.2-1.6), 1.3 (95% CI, 1.1-1.6), 1.3 (95% CI, 1.0-1.6), and 1.4 (95% CI, 1.1-1.8), respectively. Human papillomavirus 15 and 36 did not show any significant increased risk for cSCC with respective adjusted pooled OR of 1.1 (95% CI, 0.9-1.5) and 1.11 (95% CI, 0.9-1.4) (Figure 5 and Figure 6). Heterogeneity was identified among the 12 studies. The funnel plots and their respective Begg adjusted rank correlation tests are represented in eFigures 10 through 21 in Supplement 2.
This pooled analysis included a large data sample of case-control and cohort studies evaluating the association between type-specific β-HPV and cSCC. The findings suggest that β-HPV is associated with a 42% increase in the risk of cSCC among immunocompetent individuals. This analysis highlights specifically types 5, 8, 15, 17, 20, 24, 36, and 38, with an associated increased risk of 40%, 39%, 25%, 34%, 38%, 26%, 23%, and 36%, respectively. Notably, to our knowledge, this study is the largest to evaluate the type-specific β-HPV associated risk of cSCC. Our subgroup analysis showed similar results with the exception of HPV type 15. In concordance with our pooled analysis, Aldabagh et al7 performed an extensive meta-analysis evaluating the association of cSCC and β-HPV in both immunocompromised and healthy individuals, restricted to biopsy PCR HPV detection, and that suggested an increased risk of cSCC associated with HPV. Their analysis was not restricted to immunocompetent individuals and included only studies using biopsy specimens with PCR-based detection of HPV, with significant evidence of statistical heterogeneity in the included studies. A major source of heterogeneity in their analysis was the inclusion of 10 studies that did not strictly account for cutaneous HPV subtypes. Furthermore, the results did not account for the HPV type-specific associations, which represents another limitation to their findings.
Moreover, a concordance exists between our epidemiological HPV type-specific findings and the established type-specific molecular explanations of HPV mechanism in cSCC carcinogenesis. Specifically, the E6 proteins of HPV 5 and 8 are known to inhibit the transforming growth factor β (TGFβ) signaling pathway through degradation of the SMAD3 transcription factor.38,39 A pathway that normally plays an essential role in the cell cycle, which could negatively affect viral DNA replication and cell transformation. Also, HPV 5 and 8 E6 proteins induce the recruitment of MAML1, which represses the cutaneous tumor-suppressive Notch signaling pathway, favoring RAS oncogene.40,41 This in turn can upregulate AP1-activating Wnt5a signaling in keratinocytes, a classical skin carcinogenesis pathway.15,42 Finally, HPV 5 and 8 E6 expression increases the carcinogenic potential of UV-B exposure by promoting p300 degradation, thus increasing thymine dimer persistence and UV-induced double-strand breaks.18 However, HPV 20 has been shown to upregulate the p16INK4a and Akt–PI3K pathway, interfering with the cell cycle involved in progression to basal cell carcinoma. It is also known that the E6 proteins of HPV 20 can prevent UV-treated keratinocytes from undergoing apoptosis,15,16,38,43 while HPV38 E7 can degrade pRb increasing the lifespan of human keratinocytes by deregulating the cell cycle.15,44,45 In addition, the activation of nuclear factor kappa beta is believed to protect β-HPV immortalized human keratinocytes against tumor necrosis factor (TNF) and UV-mediated apoptosis.46 Among all of the identified HPV types in our pooled analysis, only HPV 17 E7 binds to UBR4, the effect of this interaction contributes to cellular transformation and anchorage independent growth.41 Interestingly the recent effort by Cohen et al,47 evaluating BRAF induced-cSCC, highlighted HPV 17 as the most frequently isolated genotype. Molecular evidence also supporting the involvement of β-HPV in cSCC carcinogenesis is the recent discovery of the rs7208422 polymorphism in the EVER2 gene, which was associated with an increased risk of cSCC in healthy individuals who did not have EV.48 Previously, it was known that the homozygous mutations in EVER2 genes cause EV, which in turn leads to the development of cSCC associated with β-HPV infections. EVER2 loss facilitates activation of the HPV 5 long control region through a JNK-dependent pathway facilitating HPV replication and cSCC development.49
This constitutes the most extensive meta-analysis on the topic. Although most of the studies included in our meta-analysis were case-control studies, they were of higher quality, with NOS scores ranging between 6 and 9. We addressed the interstudy heterogeneity, a major limitation of previous research on this topic. First, the statistical tests showed no significant heterogeneity between the included studies in both our meta-analysis and subgroup analysis. Second, to account for the different HPV collection and detection methods, we performed a subgroup analysis restricted to studies that employed HPV serology. Third, we individually evaluated the association of each β-HPV type with cSCC. This type-specific meta-analysis constitutes another major quality of the evidence presented in this study.
One limitation of this study is the effect of UV light that might represent a confounder.50 We attempted to minimize this effect by addressing it in our study quality assessment. Furthermore, most included studies were case-control studies that recruited from the same latitude level and adjusted for UV sunlight exposure.
Even though we have found a strong association between β-HPV and cSCC it is difficult to prove causality as our analysis is constituted mostly of case-control studies. However, the findings of 2 large prospective cohort studies26,27 were in concordance with current evidence suggesting HPV involvement as a causal agent of the initiation process of cSCC. Moreover, our pooled analysis did not include every β-HPV type, a limitation that pertains to the majority of studies assessing the topic. In addition, the method of HPV collection in our analysis was not PCR for HPV DNA from skin biopsy, the gold standard for detection. However, previous studies7,34 demonstrated that eyebrow hair DNA and multiplex serology are reliable substitutes.
The concordance between our epidemiological data in immunocompetent individuals and the molecular plausibility of type-specific involvement of β-HPV type 5, 8, 15, 17, 20, 24, and 38 in cSCC represents the evidence to strongly suggest a role for the HPV types in viral oncogenesis described herein. The associated increased risk of cSCC with these HPV types (30%-45%), the reported morbidity and mortality associated with cSCC and the increasing cost of care, constitute a burden on our health care system. This highlights the need to develop and test the efficacy of incorporating the above mentioned β-HPV subtypes into available HPV vaccines. Moreover, these efforts could render HPV vaccinations more widely accepted, possibly increasing the compliance rates for all HPV vaccines. A vaccine that includes these types of β-HPV will also be a major step toward precision medicine in the prevention of cSCC in solid organ transplant recipients, stem cell transplant recipients, and patients with melanoma undergoing BRAF inhibitor therapy.
This article represents, to our knowledge, the most extensive meta-analysis appraising the epidemiological association of β-HPV subtypes implicated in the pathogenesis of cSCC. This meta-analysis provides additional evidence of the involvement of β-HPV in the development of cSCC in immunocompetent individuals. Furthermore, this adds precision to the existing epidemiological findings by highlighting a significant, type-specific HPV association, which in turn is in concordance with the new arising molecular type-specific evidence.
Corresponding Author: Stephen K. Tyring, MD, PhD, MBA, Center for Clinical Studies, Department of Dermatology, University of Texas Health Science Center at Houston, 1401 Binz Str, Suite 200, Houston, TX 77004 (firstname.lastname@example.org).
Published Online: December 30, 2015. doi:10.1001/jamadermatol.2015.4530.
Author Contributions: Drs Chahoud and Chen had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Chahoud.
Acquisition, analysis, and interpretation of data: All authors.
Drafting of the manuscript: Chahoud, Semaan, Chen, Cao.
Critical revision of the manuscript for important intellectual content: Chahoud, Semaan, Rieber, Rady, Tyring.
Statistical analysis: Chahoud, Chen, Cao.
Administrative, technical, or material support: Chahoud, Semaan, Rady, Tyring.
Study supervision: Chahoud, Rieber, Tyring.
Conflict of Interest Disclosures: Dr Chahoud is a recipient of the 2015 American Society of Hematology–Hematology Opportunities for the Next Generation of Research Scientists. Dr Rieber received salary support as part of a grant from Gilead for an unrelated project. Dr Rieber’s spouse is employed at a medical company, Steris Corp. No other conflicts were reported.