A Quantitative Framework to Study Potential Benefits and Harms of Multi-Cancer Early Detection Testing

Abstract

Background:

Multi-cancer tests offer screening for multiple cancers with one blood draw, but the potential population impact is poorly understood.

Methods:

We formulate mathematical expressions for expected numbers of: (i) individuals exposed to unnecessary confirmation tests ( ${\rm{EUC}}\(), (ii) cancers detected ( \){\rm{CD}}$ ), and (iii) lives saved ( ${\rm{LS}}\() given test performance, disease incidence and mortality, and mortality reduction. We add colorectal, liver, lung, ovary, and pancreatic cancer to a test for breast cancer, approximating prevalence at ages 50, 60, or 70 using incidence over the next 5 years and mortality using corresponding probabilities of cancer death over 15 years in the Surveillance, Epidemiology, and End Results registry.

Results:

\){\rm{EUC}}\(is overwhelmingly determined by specificity. For a given specificity, \){\rm{EUC}}/{\rm{CD}}\(is most favorable for higher prevalence cancers. Under 99% specificity and sensitivities as published for a 50-cancer test, \){\rm{EUC}}/{\rm{CD}}$ is 1.1 for breast + lung versus 1.3 for breast + liver at age 50. Under a common mortality reduction associated with screening, ${\rm{EUC}}/{\rm{LS}}\(is most favorable when the test includes higher mortality cancers (e.g., 19.9 for breast + lung vs. 30.4 for breast + liver at age 50 assuming a common 10% mortality reduction).

Conclusions:

Published multi-cancer test performance suggests a favorable tradeoff of \){\rm{EUC}}$ to ${\rm{CD}}$, yet the full burden of unnecessary confirmations will depend on the posttest work-up protocol. Harm–benefit tradeoffs will be improved if tests prioritize more prevalent and/or lethal cancers for which curative treatments exist.

Impact:

The population impact of multi-cancer testing will depend not only on test performance but also on disease characteristics and efficacy of early treatment.

See related commentary by Duffy and Sasieni, p. 3