Human papillomavirus (HPV) is an important human carcinogen, accounting for an estimated 5.2% of the worldwide cancer burden.1 Following initial postulation over 30 years ago,2 epidemiological investigations confirmed the causal role of HPV in all cervical cancers and a majority of other anogenital cancers.3, 4 Mounting epidemiological and molecular evidence also supports a role of HPV in the etiology of a subset of upper aerodigestive tract cancers, particularly tonsillar and oropharyngeal cancers.5 Cutaneous types of HPV are often detected in nonmelanoma skin tumors, particularly in squamous cell carcinomas (SCC) of immunosuppressed individuals.6 The ubiquity and persistence of low levels of HPV on the skin of healthy individuals7 and limited data regarding the temporal sequence of infection and cancer contribute to the uncertain role of HPV in the etiology of skin cancers.
Familial aggregation of cancer reflects etiological contributions of shared environment and inherited genetics. Previous studies report varying degrees of aggregation of anogenital,8–10 upper aerodigestive tract11, 12 and skin13 cancer with familial cancer at the same site. A few studies provide clues into the relative importance of shared environmental and genetic factors in explaining this aggregation. For example, studies including full and half siblings and first through third degree relatives suggest a complex model for cervical cancer aggregation, with shared environmental factors in early childhood and inherited genes accounting for nearly all of the association.14
To date, studies of familial aggregation of HPV-associated cancers have grouped rare malignancies with more common ones which limits our knowledge of aggregation at specific genital (i.e., vaginal, vulvar, penile) and upper aerodigestive tract (i.e. mouth, pharyngeal, tonsillar, laryngeal, etc.) sites. Furthermore, clustering of rare histological subtypes and discordant cancers is largely unknown. Our study was conducted to fill in these gaps in our current knowledge of familial aggregation of cancers with suspected or known links to HPV infection.
AC, adenocarcinoma; CI, confidence interval; EV, epidermodysplasia verruciformis; HPV, human papillomavirus; ICD, International Classification of Diseases; SCC, squamous cell carcinoma; SIR, standardized incidence ratio.
Material and methods
The Swedish Family-Cancer Database was first created in the mid-1990s by linking census information, death notifications and the administrative family-register at Statistics Sweden, to the Swedish Cancer Registry. The 2006 update (version VII, MigMed2) of the Database was used for our study. It includes more than 11 million individuals born in Sweden between 1932 and 2004, aged 0–72 years, who are linked to their biological parents, totaling more than 3.5 million nuclear families. More than 1 million first primary cancers diagnosed between 1958 and 2004 are captured by the Database. Nuclear family linkages are complete, with the exception of a small number of offspring born between 1935 and 1940 who died before 1991. Our study was limited to offspring whose parents were known.
Coverage of cytologically or histologically verified incident cancers is considered to be 100% complete on account of compulsory nationwide registration by clinicians, pathologists and cytologists. Invasive and in situ primary cancer diagnoses were based on the 4-digit code according to the 7th revision of the International Classification of Diseases (ICD-7). Cancer sites and the corresponding first 3 or 4 digits of the ICD-7 codes included in our study are as follows: cervix (171), vulva (1761), vagina (176 except 1761), penis (1790), anus (1541), rectum (154 except 1541), lip (140), tongue (141), mouth, including gums and palate (143 and 144), tonsil (1450), oropharynx (145 excluding 1450), hypopharynx (147), pharynx, parts unspecified (148), larynx (161) and skin (191). Broad histopathological classification (i.e. SCC vs. adenocarcinoma, AC) was based on WHO/HS/CANC/24.1 coding.15
The standardized incidence ratio (SIR) was calculated to measure offspring cancer risk according to cancer in sibling and parental probands at concordant (same site and histology) and discordant (different site and/or histology) sites. Follow-up time began at birth or the study start date of January 1st, 1958, whichever came latest and ended at the first occurrence of a diagnosis of any primary invasive or in situ cancer, death, emigration or the study end date of December 31st, 2004. The SIR was calculated as the ratio of observed to expected number of cancer cases in the offspring, according to observed cancer in the familial proband. The expected numbers were computed from incidence rates standardized for age (up to 6 categories: ≤19, 20–29, 30–39, 40–49, 50–59 and ≥60), time-period (3 categories: 1961–1979, 1980–1989 and 1990–2004), geographical residence (5 categories: Stockholm area, Götebörg-Malmö, Götaland, Svealand and Norrland) and socioeconomic status (6 categories: agriculture, worker, blue collar, professional, private and other), for all offspring lacking a family history.16 Incidence rates for female offspring were also standardized for parity (3 categories: 0, 1–2 and 3 or more). The 95% confidence interval (CI) for the SIR was calculated assuming a Poisson distribution.16 Offspring cancer risks associated with cancer in sibling probands were calculated using the cohort method.17 The variances of the SIRs were inflated due to the fact that families with multiple affected siblings were ascertained multiple times. The 95% CI for the SIRs were corrected for this inflation as described previously.17, 18
Female offspring (3,625,784), followed for 103,685,112 person-years, and 3,808,150 male offspring followed for 111,171,506 person-years contributed to the SIR calculations in our study. In total, 105,418 in situ and 19,577 invasive offspring cancer diagnoses were identified (Table I). SIRs for sex-specific offspring genital cancer risk by histology and site concordant and discordant cancer in sibling and parental probands are presented in Table II. SIRs for in situ and invasive cancers were broadly similar for the majority of sites; accordingly these cases were combined for the calculations of the SIRs presented in Table II, with the exception of cervical SCC.
Table I. Frequency of Incident Primary Cancers in Offspring by Histological-Specific Site in Sweden between 1958 and 20041
Offspring in situ cervical SCC risk was significantly increased by ∼2-fold by cervical SCC, cervical AC, vulvar SCC, vaginal SCC, and anal SCC in sibling probands, and less so by maternal probands (1.78-, 1.31-, 1.09-, 1.39- and 1.36-fold, respectively, Table II). Offspring invasive cervical SCC risks were similar to those for in situ cancer when cervical SCC or AC was diagnosed in a sibling or mother. Invasive cervical SCC risk associated with a sibling with anal SCC was 5.59. Upper aerodigestive tract SCC in siblings, mothers and fathers similarly increased offspring in situ cervical SCC risk (SIR = 1.35, 1.26 and 1.26, respectively), however some site-specific differences were observed. For example, offspring in situ cervical SCC risk was significantly increased by oropharyngeal cancer in sibling (SIR = 3.17) and maternal (SIR = 3.08) probands, but not paternal probands, and tonsillar cancer in sibling (SIR = 1.84) and paternal (SIR = 1.55) probands, but not maternal probands. Familial vulvar, vaginal, and upper aerodigestive tract AC did not significantly influence offspring in situ cervical SCC risk, although these SIR calculations were based on a low number of familial cases. Familial skin SCC was also unassociated with in situ cervical SCC risk in offspring.
Offspring cervical AC risk was significantly increased by cervical SCC in siblings (SIR = 1.69) and mothers (SIR = 1.39). Significant increased offspring vulvar SCC risk was associated with cervical SCC in siblings (SIR = 1.80) and mothers (SIR = 1.76), and upper aerodigestive tract SCC in fathers (SIR = 1.82). A low numbers of vulvar AC cases in our population prohibited a separate analysis of this histological subgroup. A sensitivity analysis including all vulvar cancers, which added 32 additional familial cases, revealed very similar results to those presented in Table II, with 1 exception. Offspring vulvar cancer risk (of any histology) was significantly increased by vulvar cancer in siblings (SIR = 3.72). Offspring vaginal SCC risk was significantly increased by cervical SCC in siblings (SIR = 2.01). The addition of the vaginal non-SCC cases did not change the risks estimates or statistical significance of the results presented in Table II. Significant increased offspring penile SCC risk was associated with penile SCC (SIR = 6.86) and rectal SCC (SIR = 1.98) in fathers.
Sex-specific SIRs were also calculated for offspring anal SCC, rectal AC, upper aerodigestive tract SCC and skin SCC risk according to concordant and discordant cancer in familial probands (Table III). Anal cancer risk in male, but not female, offspring was significantly increased by cervical SCC in siblings (SIR = 1.90). Upper aerodigestive tract SCC risk in female offspring was significantly increased by siblings with cervical SCC (SIR = 1.37), mothers with skin SCC (SIR = 1.50) or fathers with upper aerodigestive tract SCC (SIR = 1.92). Male offspring were at a significantly increased risk of upper aerodigestive tract SCC by siblings with upper aerodigestive tract SCC (SIR = 2.36), mothers with vulvar SCC (SIR = 2.47) or fathers with upper aerodigestive tract SCC (SIR = 1.66). Skin SCC and rectal AC risk in female and male offspring were increased ∼2-fold with concordant cancer in siblings, mothers or fathers. Figure 1 summarizes cancer risks for offspring according to concordant cancer in any familial proband (sibling, mother or father). The SIRs range from 1.57 for upper aerodigestive tract SCC to 7.54 for penile SCC.
Table III. Risk of Rectal, Anal, Upper Aerodigestive Tract and Skin Cancer According to Concordant and Discordant Cancer in Siblings and Parents
We analyzed a group of primary cancers, selected a priori on account of known or suspected etiological links to HPV infection, for aggregation with cancers at the same site and different sites in sibling and parental probands. We observed several notable findings with respect to familial cancers at concordant sites, including statistically significant aggregation of cervical AC (SIR = 2.31), vulvar SCC (SIR = 2.27) and penile SCC (SIR = 7.54). Statistically significant clustering of discordant anogenital cancers and upper aerodigestive tract SCCs was also observed, which was unchanged when the proband and offspring cancer sites were switched. In situ cervical SCC risk was particularly strongly influenced by familial oropharyngeal SCC (SIR = 3.17) and tonsillar SCC (SIR = 1.84). Our data do not allow us to conclude whether the observed aggregation is a result of shared environmental or genetic factors. In part, the aggregation likely reflects factors relating to acquiring an HPV infection. However, most genital HPV infections are transient,19 which stresses the importance of environmental or genetic cofactors for progression to cancer. These cofactors may include cigarette smoking or inherited variation in genes regulating the immune response to HPV and/or tumor development.
Cervical AC, which accounted for less than 1% of the in situ and 25% of the invasive cervical cancers in our study, differs from the more common cervical SCC histological subtype in strength of association with specific HPV types and cofactors. For example, HPV 16 is the predominant type in cervical SCC while type 18 is more often implicated in cervical AC,20 and cigarette smoking is associated with up to 2-fold increased risk of cervical SCC and slightly reduced risk of cervical AC.21 No known studies have previously examined familial aggregation of concordant cervical ACs. However, 1 study found a slight, nonstatistically significant, increased risk of cervical AC associated with any familial cervical cancer (OR = 1.3, 95% CI 0.43, 3.9),22 while another study reported a significant increased risk of cervical SCC (SIR = 1.47, 95% CI 1.17, 1.80), associated with familial cervical AC,23 which was similar to the association observed in our study.
Vulvar SCC is suspected of having at least 2 distinct etiologies. A subset of vulvar SCCs are clearly related to oncogenic HPV and cigarette smoking, similar to cervical SCC.24, 25 While no prior studies have assessed the role of family history in vulvar SCC risk, a few studies suggest a modest influence of inherited genetic variation in cancer risk.26, 27 HPV is associated with the majority (∼80%) of penile cancers, and cigarette smoking, lack of circumcision and phimosis have also been implicated as risk factors.28, 29 The concordant familial aggregation of penile SCC was the greatest among the HPV-associated cancers assessed, although it was based on a small number of concordant familial cases (n = 6).
The clustering of cervical and vulvar SCC with upper aerodigestive tract SCC is consistent with the long-established etiologic role of tobacco use in all of these cancers, and the more recent extensive evidence for a role of HPV in oropharyngeal and tonsillar malignancies.5 The most common mucosal high-risk HPV types (16 and 18) associated with anogenital cancers are also the most common types found in upper aerodigestive tract SCCs. Our results are supported by a prior report of excess tonsillar cancer inhusbands of women with cervical cancer,30 and molecular evidence linking HPV to these particular cancer sites.
The link between HPV and nonmelanoma skin cancer was first demonstrated in patients with epidermodysplasia verruciformis (EV), a rare genetic disease characterized by disseminated and persistent HPV-associated warts and lesions, which predisposes to skin SCC.31 Subsequent studies have found an association between certain cutaneous types of HPV and skin SCC in non-EV immunosuppressed and immunocompetent populations.32 These cutaneaous HPVs are phylogentically and also probably pathologically and biologically different from the causal mucosal types in anogenital and upper aerodigestive tract cancers.33 The lack of aggregation of skin SCCs with cancers at other sites in our study may reflect site-specific cofactors of HPV in carcinogenesis, such as cigarette smoking in anogenital and upper aerodigestive tract cancers and ultraviolet radiation in skin SCC, or differences in the biology or pathology of, or immune response to, the relevant HPV types.
Cigarette smoking is an important HPV-cofactor in anogenital SCCs, and an independent risk factor in the majority of upper aerodigestive tract SCCs, and may have influenced the familial aggregation of theses cancers. However, smoking cannot explain the concordant aggregation of cervical AC or discordant aggregation of cervical SCC/AC since cervical AC is slightly inversely associated with cigarette smoking,21 which implicates other environmental or inherited HPV-cofactors. Offspring female genital cancer risk was increased by a positive maternal or sibling anogenital cancer history, however often the risks were higher with sibling probands. This may reflect the influence of shared childhood environmental factors or recessive genetics effects.
Several limitations should be considered when interpreting these results. Only offspring linked to both biological parents were included in our study, and therefore the study results may not be generalizable to the entire Swedish population, particularly first generation immigrants. Furthermore, there may be differences in familial risks that are masked by combining in situ and invasive cancer into 1 analysis. However, for the majority of the sites included in our study, in situ cancer is a recognized precursor to invasive cancer.24, 29, 34, 35 Furthermore, prior investigations indicate that in situ and invasive familial risks are similar.13, 23
The main strength of our study is the complete ascertainment of incident cancers and reconstruction of nuclear families through data linkage, removing any possibility of recall bias which may exist in case-control studies. A majority of women are exposed to genital HPV infection at some point in their lives, yet only a small proportion develop associated cancers. The aggregation observed in our study is likely to reflect HPV infection as well as HPV cofactors, both genetic and environmental, which influence the progression to cancer. Family history may be a useful factor in identifying individuals at high-risk of particular subsets of HPV-associated cancers.
The Family-Cancer Database was created by linking registers maintained at Statistics Sweden and the Swedish Cancer Registry.