Expected harms and benefits of multi-cancer early detection tests

This calculator provides a transparent computational framework for translating the diagnostic performance of a multi-cancer test to clinically relevant outcomes.

Outcomes of single-occasion testing performed at age 50, 60, or 70 for 100,000 persons with specified sex and race category are generated given test characteristics specified by the user and estimates of disease prevalence and disease-specific mortality in the United States. A user can specify test characteristics for detecting specific cancers, the associated mortality reduction for cancers detected by the test, the time horizon for cancer deaths prevented, and which cancers to include in the test’s target set.

See the Details tab for more information and the rationale for the calculator.

Population characteristics


Time horizon for cancer deaths prevented


Test characteristics

Expected outcomes per 100,000 persons

Download projections in CSV format

Download

Expected harms and benefits of multi-cancer early detection tests

Test characteristics

To configure the multi-cancer test, a user specifies the test’s (1) specificity and (2) sensitivity, localization probability, and mortality reduction for each cancer.

Specificity of a multi-cancer test is the probability that the test returns a “no cancer” signature when none of the target cancers is present. If the test technology can detect a large number of cancers but a only a subset are included in the target set, then the specificity of the test should be lowered accordingly.

The marginal sensitivity of a multi-cancer test for a cancer involves sensitivity and localization probability. Sensitivity is the probability that the test returns a cancer signature given that a targeted cancer is present. Localization is the probability of correct identification of the tissue of origin. For example, the marginal sensitivity for breast cancer is the probability the test correctly returns a cancer signal and correctly identifies the signal as breast cancer.

Note that the sensitivity for a cancer is understood to be the sensitivity of the test to detect latent disease in a prospectively screened population. This may not be the same as the sensitivity to detect known disease in a retrospective study (e.g., as presented in Liu et al. (2020) ). Sensitivity values may be selected to pertain to certain disease stages (e.g., stages I-III).

Mortality reduction reflects the reduction in the risk of disease-specific death among individuals diagnosed early by the test in the stages captured by the test sensitivity. For example, a mortality reduction of 0.1 reflects a 10% reduction in the risk of disease-specific death. Note that this is not the same as the mortality reduction among all persons screened, which is the primary result typically presented in reports of cancer screening trials. If sensitivity is selected to pertain to certain disease stages, the mortality reduction should also be specified to reflect the gain in life expected among persons detected by the test in these stages.

To input sensitivity, localization probabilities, and mortality reduction for different cancers, click on the “Configure multi-cancer test” button on the Calculator tab. Clicking this button opens a table that lists all available cancers. Double-clicking on default values for sensitivity, localization probability, and mortality reduction allows users to edit them. Finally, clicking on the desired target cancers to include in the multi-cancer test highlights the row containing that cancer, and a summary of selected cancers will appear beneath the “Configure multi-cancer test” button.

Disease characteristics

Prevalence of each cancer is approximated using the cumulative incidence for that cancer over the 5 years following the screening age based on age- and race-specific incidence in the Surveillance, Epidemiology, and End Results cancer registry for the calendar period 2000-2002. The cumulative incidence over 5 years (e.g., for ages 50-54) is approximated using the annual incidence for the corresponding 5-year age group multiplied by 5.

Mortality due to each cancer detected at a given screening age (e.g., 50 years) over a given time horizon (e.g., 15 years) is approximated using incidence-based mortality among cases diagnosed over the 5 years following the screening age. For example, the 15-year mortality associated with cancers found at age 50 is given by the 15-year mortality among cancers diagnosed at ages 50-54.

We acknowledge that, particularly for some cancers known or suspected to have longer latencies, the underlying prevalence at the time of the test could be higher than that assumed. Conversely, for cancers with shorter latencies and poorer survival, a 15-year interval for mortality might be too long. Ideally, we would want to project the prevalence of early-stage cancer at the time of the test and the corresponding baseline mortality only for these cancers, but this would require additional data and modeling.

Projected outcomes

Three outcomes are projected per 100,000 persons screened at the specified age:

Exposed to unnecessary confirmation is the number of persons who might undergo one or more unnecessary confirmation tests either due to their test yielding a false positive result (i.e., a cancer signal when no cancer is present) or a true positive for cancer but incorrect localization. We do not consider the possibility that a confirmation test will fail to detect a cancer that is present. We also do not count the actual number of confirmation tests expected, as this will depend on the protocol for post-test workup following a positive test, which may be quite variable in practice.

Cancers detected is the expected number of cancers detected early by the test. This is the total number of cases of cancers detected by the test given the test characteristics and disease prevalence for the specified sex, race category, and screening age.

Cancer deaths prevented is the expected number of cancer deaths prevented over the specified time horizon. This is the total cancer deaths prevented across cancers detected early by the test based on the (e.g., 15-year) mortality for the specified sex, race category, and screening age and the mortality reduction for the target cancers.

Note that the number needed to screen (NNS) to save one life can be calculated as 100,000 divided by the number of cancer deaths prevented.

Additional information

Jiao B, Gulati R, Katki HA, Castle PE, Etzioni R. A Quantitative Framework to Study Potential Benefits and Harms of Multi-cancer Early Detection Testing. medRxiv doi: https://doi.org/10.1101/2021.03.19.21253904
Jiao B, Gulati R, Katki HA, Castle PE, Etzioni R. A Quantitative Framework to Study Potential Benefits and Harms of Multi-Cancer Early Detection Testing. Cancer Epidemiol Biomarkers Prev, Nov 17 2021. DOI: 10.1158/1055-9965.EPI-21-0380

Expected harms and benefits of multi-cancer early detection tests

Please email questions or comments to rgulati@fredhutch.org