ASO Author Reflections: Does Genomic Testing of DCIS Provide Added Value? And Is It Worth the Cost?

Past:

In tailoring treatment for ductal carcinoma in situ (DCIS), the aim is to avoid both under- and over-treatment. This requires risk assessment for the individual patient and estimation of the risk reduction for each intervention.

For DCIS, many clinicopathologic factors affect the risk of local recurrence after breast-conserving surgery, for example: margin width,¹ age,² grade, extent of disease, and clinical vs. screen-detection. In four mature randomized trials (RCTs), radiation reduced the risk of recurrence by approximately 50%.³ Other RCTs (NSABP-B24, UKANZ, NSABP-B35, IBIS-II) proved that endocrine therapy (ET) also reduces recurrence. Furthermore, recurrence rates have decreased over time, and currently are substantially lower than those of patients treated in the mature RCTs, i.e., before 1999.⁴

Present:

The Memorial Sloan Kettering Cancer Center (MSKCC) DCIS Nomogram integrates known clinicopathologic and treatment factors to estimate risk of recurrence, is available online free-of-charge, and has been validated in at least 5 independent external populations, with good discrimination (area under the ROC curve [AUC]) and calibration (matching of actual and estimated risk). Risk estimation by the MSKCC nomogram allows both patient and clinician to weigh the benefit of various treatment options.

The Oncotype DX DCIS Score™ (Genomic Health, Redwood City, CA) is a genomic test developed to achieve the same goal. Because genomic tests have been helpful in determining the benefit of chemotherapy in invasive cancer, a genomic test for DCIS was very attractive to many. However, the DCIS Score is not predictive of benefit of radiation; it simply estimates recurrence risk. Furthermore, application of the DCIS Score in the Ontario DCIS Cohort showed us that it estimates risk no better than some standard clinicopathologic factors, such as age and size. Because of this, the test was recently “refined”, so that currently reported results incorporate age, size, and year of surgery (although not readily apparent to many users).

Here we compare the risk estimates from the MSKCC Nomogram and the “Refined DCIS Score” (RDS).⁵ We find remarkable agreement for patients with margin width > 2 mm, but for those with close margins (≤ 2 mm)—a known risk factor—the RDS appears to underestimate risk. Further, since some of the RDS development population received ET and the RDS did not incorporate ET use, the risk estimates are inherently too high for a woman taking ET and too low for one who does not.

Future:

Given the agreement between the MSKCC Nomogram (free) and the RDS ($4,620) for those with margins > 2 mm, and the apparent underestimation of risk by the RDS for those with close margins, we conclude that the cost of the RDS is not justified. Amid growing concern over health care costs, both metrics of predictive accuracy (i.e., AUC and calibration curves) and rigorous proof of significant added clinical benefit should be available before adoption of any expensive new genomic test that claims to predict outcomes.