Cost-effectiveness of Colorectal Cancer Screening Tests: Letter

The recently published article from Rui and colleagues (1) describes a Markov modeling analysis of clinical and economic outcomes for three novel, noninvasive colorectal cancer screening tests multitarget stool RNA (mt-sRNA), multitarget stool DNA 2.0 (mt-sDNA 2.0), and cell-free DNA (cf-DNA), with comparisons to other guideline-recommended screening options or no screening. Per the authors, health economic evidence is critical for informing key decision-makers, including providers, payers, and policymakers (1). Although this is a worthy perspective, several issues related to the reported model inputs and resulting outcomes must be carefully considered to appropriately determine the relative contributions of this study.Most notably, many of the screening test–specific performance parameters listed in Table 1 (and elsewhere) of the Rui, et al. article appear to differ from the referenced data sources (1). For the mt-sDNA 2.0, mt-sRNA, and cf-DNA tests, many of the listed sensitivity values for advanced adenoma or nonadvanced adenoma are not readily identifiable in the cited studies (listed as references 9–11 in Rui, et al.; refs. 2–4). Although it is possible to derive some of these performance metrics from previously reported data, accurate calculations still yield different values from those shown in the table. Also, the modeled performance parameters are inconsistently defined across tests (i.e., sensitivity for all colorectal cancers vs. stages I–III only, advanced precancerous lesions vs. advanced adenomas), which creates non-parity and biases the resulting outcomes. For example, when specificity is defined as nonneoplastic or negative findings at colonoscopy, the value for mt-sDNA 2.0 does not change (92.7%), whereas the value for mt-sRNA decreases from 88.0% to 86.9% (2, 3).It is worth recognizing that for both the mt-sDNA 2.0 and mt-sRNA tests, modifications to the clinical validation study dataset were applied during the U.S. Food and Drug Administration (FDA) review process to better align the analysis cohort with the intended use population and to standardize the reported endpoint categories with other approved, noninvasive CRC screening tests. Based on the FDA approval datasets, colorectal sensitivity, advanced precancerous lesion sensitivity, and no colorectal neoplasia or negative colonoscopy specificity are 95.3%, 43.3%, and 92.7% for the mt-sDNA test and 92.6%, 41.3%, and 86.7% for the mt-sRNA test, respectively (5, 6). Applying these FDA-approved performance estimates, which best represent clinically expected performance characteristics for on-label screening, in the Markov modeling analyses would further challenge the results and conclusions reported by Rui and colleagues.With respect to the modeled screening strategy adherence assumptions, no peer-reviewed publications to date have reported completion rates for the mt-sRNA or cf-DNA tests in routine clinical practice (as the mt-sDNA and mt-sDNA 2.0 tests are provided and navigation-supported by the same company, it is reasonable to generalize adherence data for these tests). Therefore, any direct or implied reference to the adherence inputs for the mt-sRNA and cf-DNA tests as “real world” (i.e., see Abstract, Conclusions) is incorrect and misleading.Given that sensitivity and specificity inputs were identified as the most influential factors in the present study (1), colorectal cancer screening decision-makers would benefit from the authors’ correction and/or clarification of the modeling assumptions and associated outcomes so that the reported findings can be accurately interpreted and adjudicated with respect to potential practice, policy, and payer implications.P.J. Limburg reports other support from Exact Sciences outside the submitted work.