Tumor markers: value and limitations in the management of cancer patients

Abstract

Sixteen tumor markers are reviewed, and measured to the ideal: produced by the tumor cell alone absent in health and in benign disease present in all patients with a given malignancy level in the blood representative of tumor mass detectable in occult disease. The only marker that approaches the ideal is human chorionic gonadotropin (HCG) in gestational trophoblastic tumors. In this malignancy, the HCG level suggests the diagnosis and stage, confirms response to therapy, and predicts relapse. The three most widely used and intensely studied tumor markers are carcinoembryonic antigen (CEA), alphafetoprotein (AFP), and HCG. CEA cannot be used in screening for cancer, but in carcinoma of the colon its elevation preoperatively increases the likelihood of advanced disease and postoperative recurrence. Postoperatively, elevated titers are often but not invariably associated with recurrent disease. AFP and HCG are useful in the management of nonseminomatous germ cell testicular tumors. Like CEA, they cannot be used for screening. They are more likely to be increased with advancing stage, and after therapy rising levels almost always mean recurrent disease. Some markers are valuable in specific circumstances, such as calcitonin in screening for familial medullary carcinoma of the thyroid. In multiple myeloma, immunoglobulins are useful in determining the tumor mass and response to therapy. In neuroblastoma, catecholamine metabolites are useful primarily in making the diagnosis. In some malignancies, the absence of effective therapy lowers the value of the marker, as for AFP in hepatoma. The remaining markers are too unreliable or too little studied to be useful in the management of an individual patient with cancer. The purpose of this paper is to provide the clinician with an understanding of the limitations of the present tumor markers that will lead to wiser use of the tests, and to provide standards to which future tumor markers should be measured.