Genotype attribution in high-grade cervical lesions (CIN3+) can be calculated by the hierarchical or proportional method, but these do not account for the genotype distribution in the general population and cannot assess the number of genotype-specific high-grade cervical lesions (CIN3+).
We present a statistical method for estimating genotype-specific CIN3+ risks and genotype attribution in CIN3+ from cervical screening samples. A key assumption is that genotype-specific infections in women with multiple infections have independent progression risks. We applied the method to 512 human papilloma virus (HPV)-positive women referred for colposcopy and validated it by laser-capture microscopy-polymerase chain reaction (PCR). We also compared performance by simulation.
For endpoint CIN3+, the summed deviation of attributable fractions (SDAF) between the estimated genotype-specific attributable fractions (AFs) and laser-capture microscopy PCR-based AFs was similar for the three methods: 0.17 for the new method (95% confidence interval (CI): 0.091 – 0.28), 0.19 (95% CI: 0.11 – 0.33) for the hierarchical method and 0.15 (95% CI: 0.085 – 0.26) for the proportional method. Simulations indicated that the new method outperformed the other methods for endpoint CIN3+ when the number of HPV-positive women was large. Exclusion of HPV16-positive women had only a small effect on the estimated genotype-specific risks, supporting the independence assumption.
Genotype-specific attribution in CIN3+ can be accurately predicted by a model that assumes independence between genotypes with respect to disease progression. The method can be used to monitor HPV vaccine effectiveness for prevention of genotype-specific CIN3+ and to assess disease risk after vaccination.
1. Amsterdam UMC, Vrije Universiteit Amsterdam, Epidemiology and Biostatistics, Amsterdam Public Health Research Institute, Amsterdam, the Netherlands.
2. Center for Infectious Disease Control, National Institute for Public Health and the Environment, Bilthoven, the Netherlands
3. DDL Diagnostic Laboratory, Rijswijk, the Netherlands
Conflicts of Interest: Birgit I. Lissenberg-Witte: none declared
Johannes A. Bogaards: none declared
Wim Quint: DDL Diagnostic laboratory offers LCM and SPF10 technology
Johannes Berkhof: received consultancy fees from GlaxoSmithKline and Merck/SPMDS and travel support from DDL. All fees were collected by his employer.
Source of Funding: This work was supported by the COHEAHR Network, coordinated by Amsterdam UMC, Vrije Universiteit Amsterdam (Amsterdam, The Netherlands) and funded by the 7th Framework Programme of the European Commission (Brussels, Belgium), grant Health-F3-2013-603019. This work was also supported by the project EVAH studie from Stichting Pathologie Ontwikkeling en Onderzoek (SPOO).
Data and Code Availability: Data are available upon request by the corresponding author, the R-code of the method with an example data-file is available in eAppendix 2.
Corresponding Author: Birgit I. Lissenberg-Witte, De Boelelaan 1117, 1081 HV Amsterdam, The Netherlands, telephone number: +31 20 – 444 46 39, e-mail address: B.Lissenberg@vumc.nl