Available information for the statistical analysis was: country, age at and year of diagnosis, gender, histopathological diagnosis, HPV DNA presence, HPV type, and p16INK4a expression. Histological subtypes in anal cancers were grouped into the following categories: squamous cell carcinomas (SCC) 100% warty-basaloid (included exclusively or combinations of warty, basaloid or papillary basaloid histologies), SCC 100% non warty-basaloid (SCC without warty-basaloid morphological features), SCC mixed histologies (mix of previous histological subtypes), other (undifferentiated and poorly differentiated carcinomas, neuroendocrine and adenosquamous carcinomas).
HPV prevalence and HPV type-specific detection percentages were determined according to geographical regions, histopathological categories, gender, patient´s age at diagnosis and year of diagnosis. Prevalence ratios (PRs) were estimated using bivariate and multivariate Poisson regression models with robust variance21 (link). In the final model we included region, year of and age at diagnosis, and gender. Histological diagnosis was not included in the regression analysis since it was considered as an intermediate variable in the carcinogenic process. The best fitting model was selected based on the log-likelihood ratio test. PR were only estimated for anal cancers because AIN 2/3 subset of cases was small and showed a high HPV DNA detection rate (only two cases out of 43 AIN 2/3 were HPV DNA negative).
HPV DNA prevalence was estimated among finally included cases and HPV type specific relative contribution was calculated among HPV DNA positive cases. Multiple infections were added to single types under a weighting attribution proportional to the detection found in cases with single types as previously described14 (link). In order to evaluate the increase or decrease on HPV type specific relative contributions between type of lesions, relative contribution ratios and their 95% confidence intervals (CI) were estimated (ratio of type specific relative contribution: percentage of a specific type in anal cancer/percentage of the same type in high-grade pre-neoplastic lesions).
Agreement between HPV DNA detection and p16INK4a was assessed by Kappa score. The McNemar chi-squared test for matched pair data was used for assessing unequal distribution of discordant results.
Statistical significance for all analyses was set at the two-sided 0.05 level. Data analyses were performed with the Statistical Package for the Social Sciences (SPSS) version 13.0 (SPSS Inc, Chicago, IL, USA) and with STATA version 10.0 (Stata Corporation, Computing Resource Center, College Station, Texas).
HPV prevalence and HPV type-specific detection percentages were determined according to geographical regions, histopathological categories, gender, patient´s age at diagnosis and year of diagnosis. Prevalence ratios (PRs) were estimated using bivariate and multivariate Poisson regression models with robust variance21 (link). In the final model we included region, year of and age at diagnosis, and gender. Histological diagnosis was not included in the regression analysis since it was considered as an intermediate variable in the carcinogenic process. The best fitting model was selected based on the log-likelihood ratio test. PR were only estimated for anal cancers because AIN 2/3 subset of cases was small and showed a high HPV DNA detection rate (only two cases out of 43 AIN 2/3 were HPV DNA negative).
HPV DNA prevalence was estimated among finally included cases and HPV type specific relative contribution was calculated among HPV DNA positive cases. Multiple infections were added to single types under a weighting attribution proportional to the detection found in cases with single types as previously described14 (link). In order to evaluate the increase or decrease on HPV type specific relative contributions between type of lesions, relative contribution ratios and their 95% confidence intervals (CI) were estimated (ratio of type specific relative contribution: percentage of a specific type in anal cancer/percentage of the same type in high-grade pre-neoplastic lesions).
Agreement between HPV DNA detection and p16INK4a was assessed by Kappa score. The McNemar chi-squared test for matched pair data was used for assessing unequal distribution of discordant results.
Statistical significance for all analyses was set at the two-sided 0.05 level. Data analyses were performed with the Statistical Package for the Social Sciences (SPSS) version 13.0 (SPSS Inc, Chicago, IL, USA) and with STATA version 10.0 (Stata Corporation, Computing Resource Center, College Station, Texas).
