Are Short-Term or Long-Term Recurrence Rates More Important in Breast Cancer Screening?

IN RESPONSE:

We appreciate the points made by Dr. Rutgers and colleagues, which underscore messages of our article and the editorial by Simon (1). The authors are correct that the populations on which the 2 tests were validated differed in terms of pretreatment. They also differed by estrogen receptor status: The 70-gene test was validated on a mixed population of estrogen receptor–positive and –negative patients, whereas all of the OncotypeDX patients were estrogen receptor–positive. Although there may be an explanation for the reported differences, these are the numbers reported from the validation studies on which women have to base decisions, and they point out the difficulty of interpreting numbers for populations that may have mixed prognoses (estrogen-receptor–positive and –negative) and a mix of therapeutic choices (estrogen therapy or chemotherapy).

With respect to the choice of end point, the claim that the tests would predict similar risks in the short term in the same patients is based on a series of assumptions, required because of the mixed validation population and the broader indications of the MammaPrint test.

Although chemotherapy does have its greatest effect on early relapse, we believe that most women are interested in their long-term prospects, and if the probability of long-term relapse is very low, this puts a bound on how much benefit could be gained from therapy. If the indices are approximately equally predictive in the short term and one predicts better in the long term, it is hard to see how they could be of equal clinical value.

Finally, we must respectfully disagree that the measure of clinical utility is whether a test is superior to existing predictors. Whether superior predictions translate into clinical benefit depends on whether the prediction would justify a change in a therapeutic decision and whether the change in the therapeutic decision would result in net therapeutic benefit (from reduced morbidity and mortality from treatment or disease). Our conclusion remains that the evidence for the decision-making element for MammaPrint needs strengthening (by evaluation in prognostically and therapeutically homogeneous populations to avoid the kinds of calculations that Dr. Rutgers and colleagues are forced to make), and that the direct evidence for therapeutic benefit is absent.