Women screened for cervical cancer in Sweden and Norway are currently treated under a one-size-fits-all programme, which has successfully reduced the incidence of cervical cancer but relies on guidelines for scheduling screening. Utilising the information in the screening registries, guidelines could be replaced with individual risk estimates based on screening histories and ancillary information about the attendants. Adjusting the screening density at an individual level this way allows for increased observation of high-risk individuals while freeing up resources that would otherwise be spent on low-risk individuals that do not benefit from a high screening frequency. In this study, we applied a method for stratifying women into risk groups using their screening histories, based on the Swedish screening programme1, and validated these results with data from the Norwegian programme, adapting the method to make use of the detailed HPV tests and biopsy samples available there.
We previously developed a method for stratifying women into risk groups using their screening histories using the Swedish Quality Register for cervical cancer (NKCx). Each screening diagnosis was assigned a ‘risk score’ by an expert, and the history ‘risk score’ totals for each participant were computed by an algorithm that accounted for time and delay. The resulting method could identify some very high-risk individuals in the data (up to 15% of all Swedish cancer cases since 2001 were found in this group).
Preliminary results show the same risk-identifying patterns in Norway as in Sweden, even though the screening programmes follow different clinical practices and guidelines. The ‘risk scores’ show exceptional probability of cancer in the upper ranges, and below normal risk in the lowest ranges. HPV status is often combined with low-risk diagnoses for prediction, but not high-risk, i.e. a result of normal is treated differently if it’s HPV positive or negative, but the HPV status of a HSIL diagnosis doesn’t matter for predictive purposes.
Preliminary results show strong similarities between Norway and Sweden in terms of which screening patterns can be used to predict likelihood of cancer. This suggests that the methodology is robust, and that somewhat different screening guidelines do not affect the outcome of the algorithm as long as the data is accurate and complete. Furthermore, the addition of HPV status divides diagnoses into HPV groups, enabling more refined comparisons and predictions. As HPV tests become more widespread, the algorithm is likely to improve in prediction accuracy and detail.
1. Baltzer N, Sundström K, Nygård JF, Dillner J, Komorowski J. Risk stratification in cervical cancer screening by complete screening history: Applying bioinformatics to a general screening population. Int J Cancer 2017;141:200–9.