In some ways, the Pathway Genomics test reflects the serious problem of lead time bias in the oncology arena.
Maurie Markman, MD
A company’s rather stunning recent announcement that it had developed for sale to physicians a molecular-based screening test able to detect cancer in a patient without a prior diagnosis of the malignancy prompted a very stern— and highly appropriate—response from the FDA.1 What was clearly missing in the pronouncements and advertisements from Pathway Genomics, Inc was any objective evidence that the marketed test was actually able to detect cancer, whether the assay could identify up to 10 different cancers as claimed, or whether it could find a particular type of cancer, according to the FDA.1
Further, there is no peer-reviewed evidence that any particular result on this test would impact the ability for early detection and subsequent treatment to favorably influence an overall survival outcome. Finally, and perhaps most importantly, there have apparently been no peer-reviewed reports regarding the specificity of a positive finding with this particular assay, meaning the false positive rate of the test appears to be unknown.
Of course, the substantial concern here is for the morbidity, anxiety, cost, and possibly even mortality associated with follow-up evaluation if the test result was in error and there was actually no cancer present.
Although these questions pertain to this particular test, there are broader issues at work here that bear discussion.
In some ways, this test reflects the serious problem of lead time bias in the oncology arena. This bias involves the perception that if a cancer is found at an “earlier point in time,” the patient may experience a more favorable outcome when in fact there is no change in the ultimate natural history of the illness compared with discovery of the cancer at a later date. This particular concern is highly relevant in all proposed approaches to cancer screening and early detection.A unique issue with any blood test used to detect a solid tumor is the fundamental question of whether a positive test influences the ability to identify the cancer “early” even if the result is a true positive. This is in striking contrast to any imaging-based cancer screening approach where the anatomic location of the abnormality is evident (although the observed abnormality may not actually represent a malignancy).
For blood tests, a true positive result for the presence of cancer may be of quite limited or no clinical utility if subsequent evaluation by radiographic studies (eg, lung) or direct visualization (eg, endoscopy) is unable to identify the malignancy.
Further, the laboratory marker may result in frequent noninvasive or even invasive testing in a search for what may remain for extended periods of time solely microscopic cancer.
Beyond these technical questions, it is essential to acknowledge that there is unfortunately no inherent reason to believe that “discovering the cancer” in this manner will actually improve the patient’s survival outcome.
In fact, the particular cancers that are able to be detected by their molecular signatures in the blood may be the very cancers that metastasize early in their natural history and such screening may not be able to impact the ultimate outcome.
Of course, that statement remains speculative at this point. Appropriately conducted trials would be needed to determine whether cancer screening undertaken by this approach based on a particular blood-based molecular test is an effective strategy for the early detection of malignant disease.Although it is not a molecular marker, the previously published experience with employing the CA-125 serum tumor antigen in the early detection of epithelial ovarian cancer is an excellent case in point regarding the potential danger of using a bloodbased test in an attempt to increase the opportunity for early detection of a nonhematologic cancer.2
Despite multiple attempts to develop an approach to employ this well-established blood test, routinely used to monitor the course of established ovarian cancer, a large-scale phase III randomized trial not only failed to demonstrated the utility of this test in the detection of early-stage disease, but also revealed that the testing strategy resulted in increased morbidity, likely due to subsequent unnecessary invasive testing.2A critically relevant potential exception to these general statements regarding current concerns with blood-based molecular testing in the diagnosis of new cancers would be secondary molecular data that may be obtained from noninvasive prenatal testing.3
In this scenario, it is important to remember that the molecular test is being undertaken for, and is justified by, an entirely different purpose— the detection of fetal aneuploidy.
If the laboratory, while examining maternal plasma cell-free DNA in an individual without a known history of cancer, identifies a molecular abnormality that is not related to either the germline of the mother or of the developing fetus, it is appropriate to inquire if the result may reflect a previously unidentified malignant condition in the mother. In a recent report that examined the plasma cellfree DNA from more than 4000 pregnant women, three individuals were found to have such abnormalities and were subsequently shown to have a previously undiagnosed malignancy (follicular lymphoma, Hodgkin lymphoma, ovarian cancer).3
It is relevant to acknowledge that, while highly provocative, these results indicate just how uncommon it was to discover a malignant condition in a population of individuals with no other signs or symptoms of harboring a cancer.
Of course, the overall incidence of cancer would be greater in an older patient population but the likelihood of discovering and subsequently improving the survival outcome in this clinical scenario, in absence of a justifiable suspicion for the presence of a malignancy, remains to be established.
Maurie Markman, MD, editor-in-chief, is president of Medicine & Science at Cancer Treatment Centers of America, and clinical professor of Medicine, Drexel University College of Medicine. firstname.lastname@example.org.