Cancer screening simulation models: a state of the art review

Abstract

Abstract Background Nowadays, various simulation approaches for evaluation and decision making in cancer screening can be found in the literature. This paper presents an overview of approaches used to assess screening programs for breast, lung, colorectal, prostate, and cervical cancers. Our main objectives are to describe methodological approaches and trends for different cancer sites and study populations, and to evaluate quality of cancer screening simulation studies. Methods A systematic literature search was performed in Medline, Web of Science, and Scopus databases. The search time frame was limited to 1999–2018 and 7101 studies were found. Of them, 621 studies met inclusion criteria, and 587 full-texts were retrieved, with 300 of the studies chosen for analysis. Finally, 263 full texts were used in the analysis (37 were excluded during the analysis). A descriptive and trend analysis of models was performed using a checklist created for the study. Results Currently, the most common methodological approaches in modeling cancer screening were individual-level Markov models (34% of the publications) and cohort-level Markov models (41%). The most commonly evaluated cancer types were breast (25%) and colorectal (24%) cancer. Studies on cervical cancer evaluated screening and vaccination (18%) or screening only (13%). Most studies have been conducted for North American (42%) and European (39%) populations. The number of studies with high quality scores increased over time. Conclusions Our findings suggest that future directions for cancer screening modelling include individual-level Markov models complemented by screening trial data, and further effort in model validation and data openness.