OBJECTIVES: Multi-cancer early detection (MCED) testing along with standard of care (SoC) could improve survival outcomes and lower treatment costs but increase screening costs. These estimates may depend on methods for projecting post-diagnosis survival. Mixture Cure Modeling (MCM) has been proposed for projecting survival impact with MCED tests because it includes a proportion of cured patients which is less sensitive to lead time bias. This study explores the impact of using MCM vs a simple extrapolation of survival on the cost-effectiveness of MCED testing.

METHODS: A Markov model compared annual MCED testing for ages 50-79 plus SoC with SoC alone. Nineteen solid cancer groupings representing 80% of total cancer incidence were considered. Survival by stage at detection was projected from SEER using two methods. The first used Kaplan-Meier survival from SEER for 3 years, followed by MCM projections, including potential long-term excess mortality in survivors. The other similarly used Kaplan-Meier survival for 3 years, but followed by a constant hazard. Both were capped by the age-matched general population survival. Value based prices (VBP) were estimated for a willingness-to-pay threshold of $100,000/QALY.

RESULTS: Stage III/IV cancer diagnoses were reduced by 36% when adding MCED testing with both projections. This yielded 0.15 and 0.13 incremental QALYs per person when using MCM and simple extrapolation, respectively. Cancer-related treatment costs were reduced by $7,792 and $5,282 per person, respectively. VBP for the MCED test ranged from $1,460/test to $1,182/test. VBP was similarly sensitive to dwell time assumptions with both survival methods, suggesting the most important impact of dwell time in this analysis was an increase in interval cancers with more rapid dwell times, rather than lead time bias.

CONCLUSIONS: Different methods for projecting post-diagnosis survival may lead to variation in estimated cost-effectiveness of MCED testing, with modeling the potential for cure supporting larger benefits.