The overall objective of this project is to improve the biologic understanding of adenocarcinoma and adenosquamous cancer of the cervix uteri (in the following referred to as adenocarcinoma) through investigation of its viral etiology, and genes encoding certain human leukocyte antigen (HLA) haplotypes. The long-term objective is to propose rational prevention strategies for this disease. Our main specific aims are to: 1) estimate the relative risks for adenocarcinoma both as a function of HPV presence overall and as a function of time since first detected infection with human papillomavirus (HPV); 2) to assess whether persistence and/or a high initial HPV 16/18/45 viral load is a determinant of adenocarcinoma; 3) to assess whether certain HLA haplotypes are associated with risk for adenocarcinoma, and if the association is mediated via a higher viral load and/or persistence of HPV 16/18/45. This project will build on experience from two previous studies examining squamous cell cervical cancer and HPV (both funded by NIH). Our capacity to conduct these studies is derived from the utilization of unique prerequisites in Sweden, namely the extensive documentation present in computerized registers for population-based PAP-smear screening, potential ascertainment of all incident cases of adenocarcinoma, and access to archival smears and tissue specimens. Using a nested case-control design in this large study base, with up to 32 years of complete follow-up, 511 women with adenocarcinoma and 511 individually matched control women will be identified. Using validated and sensitive PCR-assays, the presence of HPV DNA will be analyzed in all available smears from each participant (on an average four smears per individual, giving totally about 4,088 smears).HPV persistence, HPV16/18/45 viral load and HLA (types DQ6 and DR15) will be analyzed for all included women. Relative risks of cervical adenocarcinoma for oncogenic HPV infections, HPV persistence,HPV16/18/45 high viral load, HLA haplotype and interactions between these factors will be estimated with conditional logistic regression. The likelihood ratio test will be used to discriminate between nested models. Statistical power calculations produced for the specific aims show that our study is adequately sized.
