For almost a century, clinicians reporting the treatment of patients with cancer have been analyzing the cancers, but not the patients. The formal systems we have developed to classify patients, estimate prognosis, choose treatment, and evaluate outcomes depend almost exclusively on the morphology of the cancer. The morphology is described with "staging systems" or other taxonomies that identify a cancer's primary site, its gross anatomic spread beyond the primary location, and such microscopic attributes as histologic type, mitotic grade, and vascular relationships.
Statistical appraisals of prognosis and therapy have been greatly improved by these morphologic classifications. They have helped put an end to many past acts of statistical malpractice, such as the comparison of results for treatment A, given to patients with localized cancer, with the results of treatment B, given to patients with disseminated cancer. By insisting that treatments be contrasted only for patients having similar morphologic stages of