Personalised cervical cancer screening
Background
Screening for cervical cancer is currently recommended every third year for women aged 25 to 69 years. If all tests are normal, this corresponds to around 16 screening tests over a lifetime. Current knowledge indicates that after a series of normal samples, the risk of pre-stages of cervical cancer is low. Nevertheless, the program treats all women equally, regardless of personal test history.
Aim
This project promotes the idea of personalized screening in cancer prevention. The aim of the project is to create more flexible cancer prevention by moving from a "one size fits all" model with standardized recommendations to recommendations based on a personal risk assessment. By combining expertise from both the medical and computer technology worlds, we want to develop an algorithm, a guideline, which, tailors recommendations for disease prevention based on the individual's risk profile. This will be done using health data.
The project will develop statistical methods for risk assessment, with a long-term goal of improving health services offered to the population.
Data material
The cervical cancer screening programme has been shown to have a good effect when it comes to cancer prevention in the population, and the Norwegian Cervical Cancer Screening Programme has generated large amounts of data over the years. These data are available via the Cancer Registry's central health registers and can thus be used to research even better cancer prevention. This makes screening for cervical cancer a good area to develop such an algorithm.
In addition to statistical data from the Norwegian Cervical Cancer Screening Programme, we will also use information about human papillomavirus infection obtained from a previous study, as well as data from questionnaire surveys that deal with risk factors which can be linked to the development of cervical cancer. Having access to such a data set, combined with new advances in understanding how cervical cancer develops and not least the development in the world of technology, gives us access to completely new opportunities to process data and assess the risk of cancer in each individual woman.
Status
In collaboration with Lawrence Livermore National Laboratory, CA, USA, SimulaMet, Oslo, Norway, Karolinska Institutet, Stockholm, Sweden, and University of Tartu, Tartu, Estonia, we develop statistical models that calculate the risk level for an abnormal sample and cell changes by taking into account the individual screening history.
Publications
Improving five-year survival prediction via multitask learning across HPV-related cancers
PLoS One, 15 (11), e0241225
DOI 10.1371/journal.pone.0241225, PubMed 33196642
A hidden Markov model for population-level cervical cancer screening data
Stat Med, 39 (25), 3569-3590
DOI 10.1002/sim.8681, PubMed 32854166
Goncalves A, Ray A, Soper B, Widemann D, Nygård D, Nygård JF, Sales AP (2019). Bayesian Multitask Learning Regression for Heterogeneous Patient Cohorts.
Journal of Biomedical Informatics Vol 4 2019 December
In media:
Laboratory and Norwegian researchers collaborate to imporve cancer screening
