Post cancer treatment surveillance
Is the benefit of the benefit of detecting tumor markers overestimated?
For decades, the medical community has known that a variety of cancers produce proteins that circulate in the blood and can be quantified to determine the response of a cancer to a specific treatment. These proteins are commonly referred to as tumor markers. Identifying these tumor markers can be especially helpful in cases where the cancer may be difficult to measure in regard to treatment response, such as breast cancer in a patient who presents with isolated bone metastasis.
Unfortunately, many patients—and physicians—overestimate the benefit of detecting tumor markers. There is no question that the advent of biomarker detection in malignancies has helped physicians monitor various treatment modalities. As a result, patients often insist on having these markers checked routinely, even when not on therapy. Physicians, too, encourage the practice. In fact, a recent study found that almost half of US oncologists order tumor markers indiscriminately, despite evidence that it does not save or lengthen the life of a patient, in most cases.
Frequent measuring of tumor markers seems to promote a sense of security among patients in that they feel if the markers have not increased, their cancer is under control. Some patients keep log books of their tumor markers that have been obtained over the years, and while they feel comfortable when the tumor marker is stable or reducing, even if it increases within normal ranges, they may panic unnecessarily. Physicians often refer patients to me for consultation due to an elevated screening CEA or CA-19-9 in an asymptomatic patient. Many times imaging studies, which are very difficult to get insurance carrier approval for based on this information, turn out to be normal. Nevertheless, much anxiety is produced in both the patient and the referring physician. The bottom line is that we have yet to demonstrate that measuring these tumor markers in a patient with no clinical evidence of active cancer is necessary, and the knowledge obtained from doing so has yet to result in a better outcome for most patients.
Recently a high level, randomized, phase III study determined that in ovarian cancer, regardless of whether a patient suspected the cancer had returned based on symptoms versus an elevated level of the tumor marker CA-125 in an otherwise asymptomatic situation, survival was not affected by what triggered reinstituting therapy. Despite these findings in patients with ovarian cancer, some NCCN recommendations suggest obtaining specific tumor markers ahead of any evidence or symptoms of disease progression. CEA levels for colorectal cancers and alpha fetal protein and beta HCG levels for testicular cancer are examples of this practice, as the belief is that these are useful due to issues related to potential salvage therapeutics.1
Problems with any test used to detect an illness start with issues such as sensitivity, specificity, and both positive and negative predictive values. Sensitivity means that when a condition exists the test has the ability to detect it, and therefore lacks false negative results. Specificity is when a disease is not present and the test has the ability to not show positivity, in other words avoid false positives. The statistical measurements known as positive and negative predictive values can be a little more complicated. Whereas sensitivity and specificity are not determined based on having a particular population in mind, predictive values are population-based.
A positive predictive value is when a test result demonstrates a certain finding, such as presence of a disease among a defined population and it really exists. Negative predictive value is the opposite. For example, a test used to diagnose coronary artery disease, such as a stress test, would have a higher positive predictive value in the elderly than in an elementary school population, simply because the likelihood of cardiac disease is so much higher among elderly people.
In regard to using tumor markers, these issues play a major role in demonstrating their lack of effectiveness. One of the most feared cancers is pancreatic cancer due to its extremely low cure rate. The best commercially available assay, the CA-19-9, has sensitivity and specificity rates as low as 41% and 33%, respectively, meaning the test can erroneously respectively miss and indicate pancreatic cancer more than it can lead to a correct diagnosis.2,3