A comparison of international modelling methods to evaluate health economics of colorectal cancer screening: a systematic review

Purpose: To identify cost-effectiveness analysis (CEA) models and their corresponding methods and assumptions used internationally to evaluate colorectal cancer screening.

Methods: A systematic search of the literature was performed using MEDLINE, EMBASE, Web of Science and Scopus. The Consolidated Health Economic Evaluation Reporting Standards (CHEERS) statement was used as a template to extract relevant data from studies retrieved. There was an additional focus on extracting specific modelling information such as health states and structures, model type, parameters considered, modelling assumptions, data sources, and justification of all these choices.

Results: A total of 80 papers were identified for inclusion in the review, corresponding to 11 distinct models. Of those, 51 papers cited the use of Markov modelling, 26 used microsimulation modelling and 3 used a decision tree. In 45 cases, the modelling methods consisted of imprecise descriptions of the model type, for example, Markov model use was stated, with no other details such as being time-homogeneous or inhomogeneous, whether it was cohort-based or used subgroups. Most papers included extensive lists of parameters and costs used in the model, with 60 and 69 papers respectively reporting them either in the main body of the paper or in the supplementary material. Our review highlights that justification of model choices was limited; 16 papers included a brief paragraph stating the modelling choices. There was a broad consensus of health states used in the models, with approximately 50 models following the same format of healthy leading to the discovery of an adenoma or polyp discovery, followed by un/-diagnosed Stage I/II/III/IV CRC, ending with CRC or non-CRC related death. Although papers offered similar attribute content yet were country-specific adjusted, a robust model comparison was obscured by documented reporting.

Conclusions: This systematic review of CRC screening models is the first to address the methods and assumptions used to construct CEA models as recommended by the CHEERS checklist. The most standard modelling approach identified was microsimulations with a Markov process, depicting a realistic country-specific population, varying by age, sex and age-/sex-specific incidence and mortality, spanning their lifetime. The insights gained from this review are crucial for understanding variation in model development and that without clarity in model analysis, any decisions made can be highly uncertain due to a lack of understanding of the model inputs and assumptions.