This study on the distribution of 14 hrHPV types revealed marked differences between cervical adenocarcinoma and its precursors, and SCC. The prevalence of HPV16 is increased in SCC compared to normal cytology, whereas HPV18 is more prevalent in adenocarcinoma and its precursor. However, when accounting for the distorting effect of extremely high prevalence types, HPV16 and HPV45 are also associated with adenocarcinoma and its precursors, and HPV18 is associated with SCC as well. These data suggest that within the group of high-risk types of which the association with cervical cancer has already been established beyond any doubt (Munoz et al, 2003
), infections with either HPV16, HPV18, or HPV45 confer a preferential risk to develop a malignancy of the uterine cervix.
Some methodological aspects of this study need to be discussed. Firstly, we compared cross-sectional data of women with normal cytology obtained from the POBASCAM trial with retrospectively collected cases of cervical cancer. This approach may have biased our estimates of risk associated with specific hrHPV types. Women with cervical cancer were on average older than the women with normal cytology, and age has been shown to be associated with hrHPV type-specific prevalence (Castle et al, 2005a
). We used two methods for to correct for the age difference. Firstly, we analysed our data stratified in age categories. Secondly, a potential source of bias may be that it takes at least 8–10 years to develop invasive carcinoma of the cervix (van Oortmarssen and Habbema, 1995
; Zielinski et al, 2001
). Women with cancer may have contracted an hrHPV infection 10 years before they were diagnosed with cancer. We repeated the analyses matching cases of cancer with normal controls 10 years younger, and our estimates were not affected, indicating the robustness of our data. Also, we defined women with normal cytology as women having a screening smear diagnosed as normal without either histologically or cytologically diagnosed lesions in the 2 years preceding the screening smear. In the Dutch screening programme, these women are considered to be free of cervical disease, and they will not be called for cervical screening for the next 5 years (van Ballegooijen and Hermens, 2000
). However, our population with normal cytology may have contained a small number of women with either an underlying high-grade lesion or who may develop a high-grade lesion during follow-up. This diagnostic bias would have a diluting effect on the risk estimates obtained by our study, as hrHPV prevalence in the normal population would have resembled the cervical lesion cases more closely. A strong point of our cross-sectional approach is that we were able to use a reference group of women with normal cytology taken from the same geographic region as cases with cervical cancer. In contrast, other studies relating hrHPV prevalence to histological type have relied on pooling of data obtained from worldwide studies to provide estimates of hrHPV prevalence in cancer and its precursor stages and regional variations in type-specific prevalence may have affected comparisons in these studies (Clifford et al, 2003a
Thirdly, we have performed a type-specific E7 PCR in women with carcinoma in order to diagnose integrated hrHPV infections, whereas women with normal cytology were not evaluated using the type-specific PCR. This may have resulted in a confirmation bias in the diagnosis of hrHPV infections, as integrated hrHPV infections in women with normal cytology may not have been diagnosed. However, integration of the hrHPV occurs late in the progression from normal epithelium to carcinoma, and viral integration is extremely rare in a population of women with normal cytology (Munoz et al, 2003
). Therefore, we do not consider our approach biased.
In addition to other studies demonstrating that adenocarcinomas are more often HPV18 positive than HPV16 positive (Pilch et al, 2001
; Schwartz et al, 2001
; Altekruse et al, 2003
; Burk et al, 2003
; Clifford et al, 2003b
; Huang et al, 2004
), we have now shown that when comparing invasive adenocarcinoma cases to cytologically normal controls, the ORMH
is only 1.3 for HPV16 and 15.0 for HPV18. Conversely, when comparing invasive SCC cases to normal controls, ORMH
is 7.0 for HPV16 and only 1.3 for HPV18. This suggests that HPV16 and HPV18 are associated with a preferential risk compared to the other high-risk types of hrHPV for the development of either SCC or adenocarcinoma. Combining our analyses both including HPV18 and excluding HPV18, our data suggest that HPV18 is mainly a risk factor for the development of adenocarcinoma whereas the highly aggressive HPV16 is associated with both SCC and adenocarcinoma. Alternatively, HPV16 and HPV18 might preferentially induce differentiation in either squamous or columnar direction respectively after infection of epithelial stem cells localised in the basal layer of the epithelium. The hrHPV types tested for in this study other than HPV16, HPV18, and HPV45, did not reveal an increased prevalence in either histological subtype of cervical cancer, suggesting that the other types all pose a similar relatively low risk of progression to cancer. This also includes HPV33, which is prevalent in lesions
CIN2 (Bulkmans et al, 2005
). However, a plausible explanation might be that HPV33 has the potential to induce high-grade CIN lesions relatively easy, whereas its capacity to induce progression from high-grade CIN to invasive carcinoma might be relatively low (Bulkmans et al, 2005
What are the consequences of our findings for cervical screening? Recently, it was shown that both HPV16 and HPV18 infections in women with normal cytology are associated with an increased 10-year absolute risk for high-grade lesions and cervical cancer (Khan et al, 2005
). However, two other recently published studies did not demonstrate an association of HPV18 with cytological abnormalities and high-grade histological lesions in either short-term follow-up or in a cross-sectional design (Bulkmans et al, 2005
; Castle et al, 2005b
). These data suggest that HPV18 infections, which we have shown to be preferentially increased in prevalence in cervical adenocarcinomas, either do not induce cervical lesions diagnosed as abnormal cytologically or that cervical lesions associated with HPV18 are not diagnosed as a result of sampling error, since these lesions are more often localised high in the endocervical canal (Woodman et al, 2003
). Whatever the cause, our results show that HPV18 has a preferential risk for AdCx and ACIS. Being aware of this association should warrant a less expectant attitude for women with persistent HPV18 infections and normal cytology to refer women to colposcopically mediated biopsies and endocervical sampling in case of a normal transformation zone.
In conclusion, we have shown that HPV16, HPV18, and HPV45 display an increased prevalence in cervical cancer compared to cytologically normal smears. HPV16 confers the greatest risk for SCC and HPV18 for adenocarcinoma of the cervix. These data strongly argue for hrHPV type-specific risk stratification of women with normal cytology and a positive hrHPV test participating in cervical screening programmes.
The study was approved by the Medical Ethics Committee of the VU University Medical Center, Amsterdam, The Netherlands (nr 96/103), and by the Ministry of Public Health, The Hague, The Netherlands (VWS nr 328 650).
POBASCAM study collaborators other than authors
Dr K v Groningen (Spaarne Ziekenhuis, Heemstede, The Netherlands), Dr W Ruitinga (Stichting PA Laboratorium Kennemerland, Haarlem, The Netherlands), Dr ME Boon (Leiden Cytology and Pathology Laboratory, Leiden, The Netherlands), Dr M van Ballegooijen (Department of Public Health and Social Medicine, Erasmus University Rotterdam, The Netherlands), Dr AJP Boeke (Institute for Research in Extramural Medicine, VU University Medical Center, Amsterdam, The Netherlands), Prof Dr RHM Verheijen (Department of Obstetrics and Gynaecology, VU University Medical Center, Amsterdam, The Netherlands).