Assessment of attribution algorithms for resolving CIN3-related HPV genotype prevalence in mixed-genotype biopsy specimens using laser capture microdissection as the reference standard.

To make accurate determinations regarding potential and actual impact of HPV vaccine programs, precise estimates of genotype-specific contributions to disease are required for pre- and post-vaccine populations. Definitive determination of lesion-specific genotypes, particularly where multiple genotypes are detected in a sample, can be technically demanding and resource intensive; therefore, most prevalence studies use mathematical algorithms to adjust for multiple genotype detections. There are currently several algorithms, which can produce genotype estimates within a wide range of variability. The use of these for cervical cytology samples has recently been assessed for accuracy against a definitive reference standard, but none have yet been assessed for multiple-genotype-containing whole biopsy specimens. Using laser capture microdissection (LCM) on biopsy samples, lesion-specific genotype prevalence data were generated for a cohort of 516 young Australian women (aged 18-32 years) with cervical intraepithelial neoplasia grade 3 or adenocarcinoma in situ. Using whole tissue section genotype data from the same cohort, including 71 (13.7%) with multiple genotypes, lesion-associated genotype prevalence was estimated using four different attribution algorithms. The proportion of lesions attributable to HPV16 and HPV18 by LCM were 58.4% and 5%, respectively; hierarchical, proportional, single type/minimum and any type/maximum attribution estimates were comparable across genotypes. For analyses utilising whole tissue biopsy cervical specimens, attribution estimates are appropriate for estimating the proportional contribution of individual genotypes to lesions in a population.