OBJECTIVES: Using discrete choice experiments (DCEs) to elicit treatment preferences in oncology can be challenging, particularly if trade-offs between survival and severe treatment risks are required, or sample sizes are small. To understand the state of practice in DCEs in oncology treatments, a systematic literature review was conducted.

METHODS: The search updated previous reviews of health-related DCEs to include studies published 1990-2020. Screening was piloted and subsequently conducted by seven analysts, with 10% of studies double screened. Data was extracted using a pre-specified template.

RESULTS: 1,631 studies were screened; 76 were eligible and included in the review. Common indications were breast (n=10; 13%), prostate (n=9; 12%), skin (n=9; 12%), and blood cancer (n=8; 11%). Studies were conducted in patients (n=57; 75%), healthcare providers (n=25; 33%), general population (n=11; 14%) and/or caregivers (n=5; 7%). While sample sizes varied widely (range 18-654), 53% (n=18) of patient studies included a sample size of 150 or fewer. 71% of studies (n=54) used qualitative insights to inform attribute development. Studies included a median of 6 attributes (range 3-11) and an average of 12 experimental choice tasks (range 6-25). Attributes included benefits (n=58; 76%), such as progression-free survival (PFS; n=35; 46%) or overall survival (OS; n=26; 34%), and risks (n=65; 86%). While qualitative pre-testing (n=50; 66%) was common, quantitative pilots (n=9; 12%) were rare. Data was most commonly analysed with multinomial (n=31; 41%) or mixed (n=29; 39%) logit models. Outputs included coefficients (n=64; 89%), marginal rates of substitution (n=34; 45%), and relative attribute importance (n=33; 43%). When included, survival often ranked as the first (OS: 60%; PFS: 33%) or second (OS: 19%; PFS: 24%) most important attribute.

CONCLUSIONS: DCEs are increasingly being used to evaluate oncology treatments. Future studies should carefully consider methodological challenges such as small sample sizes and potentially dominant survival outcomes.