Which disseminated neoplasms can we cure? In adults these are lymphomas, some acute leukemias, and germ cell neoplasms, which arise from those normal tissues that are most severely damaged by our drugs (leukopenia and male sterilization are ubiquitous side effects of chemotherapy). Beyond these clinical facts, and the unique activity of glucocorticoid agents and antimitotic vinca alkaloid agents in some of these tumors, we do not really know why we can cure them. The high rate of neoplastic cell mutation leading to cell resistance, to which we ascribe our failure to cure other tumors, does not prevent us from curing these. The reasons for our success in treating certain pediatric tumors (such as Wilm and embryonal rhabdomyosarcoma) with an entirely different histogenesis are even more obscure. Perhaps, as in retinoblastoma, we are dealing with tumors that arise from cells with unique proliferative programs completed early in postnatal life.

Every new cytotoxic agent, of any class, is reported to show some activity against lymphomas, but effective new drugs for epithelial neoplasms are rare. Is there not, perhaps, a qualitative difference between the neoplasms we can cure with cytotoxic drugs and those we cannot? May not the relative efficacy of cytotoxic agents in the treatment of germ cell and immunocyte neoplasms have misled us by implying a broader role for such drugs in neoplastic disease than really exists?

Consider the results of the cytotoxic therapy for adult epithelial tumors [1] and sarcomas. We can cure a few cancers, at an early stage of dissemination, using adjuvant treatment and induce temporary regression of metastatic disease in certain types (such as breast and small cell cancer), but we cannot cure most cancers. Autologous bone marrow transplantation permits us to use higher doses of these drugs, with little improvement in results. It is hard to believe that the outgrowth of subsets of drug-resistant tumor cells is the problem; most disseminated epithelial tumors are unresponsive to primary therapy, regardless of their proliferation rates. Cells in cycle should be vulnerable to our drugs; however, rapidly proliferating, anaplastic tumors are often insensitive [2].

We often forget how little we know. Our understanding of the biological effects of most cytotoxic drugs is incomplete. We have some understanding of how antimetabolite agents work, but they have a limited therapeutic spectrum. It is clear that antiproliferative agents are not lethal to most types of neoplastic cells, although we do not know why they are not lethal. We may be approaching the limit of conventional cytotoxic therapy for neoplastic diseases; it has been, and remains, a source of real progress. Promising agents (such as taxol) still deserve testing, but the large-scale search for, and protocol study of, such drugs may no longer be appropriate. Medical oncologists will hardly want conventional cytotoxic therapy to remain their major weapon against cancer.

The effectiveness of immunologic therapies (such as the use of monoclonal antibodies or primed T and natural killer cells [3, 4]) remains limited and may be no more specific than chemotherapy. The limited but definite efficacy of interferons in the treatment of hairy cell leukemia, myeloproliferative syndromes, and hypernephromas is of great interest but does not imply tumor antigenicity. Induction of differentiation may, in fact, be their true mechanism of action (vide infra). How important can we assume defective immune surveillance to be in oncogenesis? For example, the only neoplasms whose incidence is increased (in congenitally immune deficient children, chronically immunosuppressed transplant recipients, and the tens of thousands of patients with the acquired immunodeficiency syndrome we have observed for more than a decade) are certain cutaneous tumors, B lymphomas, and Kaposi sarcoma (if the latter is a neoplasm). The immune system, like cytotoxic drugs, works by killing neoplastic cells and may be no more effective, specific, or less toxic [5]. We know as little about why immunologic approaches are successful in some patients with hypernephroma or melanoma as we know about the efficacy of chemotherapy, but immunologic approaches have been used for only a decade and deserve further study (> 20 years were needed to learn to use cytotoxic drugs effectively). Immunologic approaches have, however, been overpublicized and are hardly enough of a cutting edge to justify either a large-scale national effort to test them or designating them as standard therapy.

The delivery of toxins to tumor cells by their conjugation to monoclonal antibodies [6] is an interesting but still experimental approach. It certainly deserves further study, but neoplastic cells are a proliferating population not governed by the usual controls, have great genetic diversity, and remissions are likely to be short. This method, like most others, is designed to destroy noxious cells. But in many other human diseases, from autoimmunity to diabetes and hypertension, our treatments are based on subtle perturbations of cell and organ physiology with the use of precisely defined molecules. Has our understanding of neoplastic cell function reached the point that we can use such subtle approaches?

The advances in our understanding of oncogenesis in the past 20 years should claim more of our attention, because they may point the way to more specific and effective therapies. The study of oncogenes has already showed mutations that result in proteins specific for certain neoplasms. Examples are the bcr-abl chimeric gene product in chronic myelogenous leukemia (a proven hematopoietic stimulator [7]) and the abnormal retinoid receptor that arises from the 15:17 translocation in M-3 acute leukemia (which may help to account for incomplete, retinoid-dependent myeloid differentiation [8]). In other cases, such mutations lead to quantitative but well-defined abnormalities. The translocations of the myc gene in Burkitt and certain other high-grade lymphomas are associated with excess production of a normal nuclear protein that has been shown to promote cell proliferation. A mutation in the Burkitt-associated myc locus eliminates binding of another nuclear protein that may control myc gene expression [9]. Conversely, the translocation of the bcl-2 gene in follicular lymphoma appears to prevent normal apoptosis (programmed cell death) and may lead to the accumulation of these slowly proliferating cells [10].

The study of familial human neoplasms has shown the presence of anti-oncogenes (deletion or mutations of anti-oncogenes are oncogenic); they occur as acquired abnormalities in sporadic tumors. The pathogenesis of colorectal carcinoma, one of the most common human epithelial neoplasms, is being rapidly elucidated by anti-oncogene studies, which have revealed the presence of a new gene deletion on the long arm of the fifth chromosome [11]. Deletions of the p53 tumor suppressor gene (or mutations, yielding a product that inhibits the function of the normal allele) occur as primary events in the heterozygous state in families and persons with multiple neoplasms and occur as secondary events in many sporadic cases [12, 13]. Not one of these mutations is sufficient for oncogenesis, and altering these mutations is likely to be difficult. However, inhibition of the synthesis of bcr-abl [14] and myc [15] proteins in the cytoplasm with specific anti-sense oligonucleotides has already led to control of leukemic cell proliferation and induction of differentiation in vitro. In-vivo, high retinoid doses induce remissions in M-3 leukemia and certain squamous cell carcinomas [16-18]. Induction of differentiation is perhaps the most potentially important advance in neoplastic treatment in decades, because it points the way to an entirely new approach, based not on cell destruction but on cell control. A similar mechanism may be the basis of effective hormonal treatment of breast and prostate carcinoma. The fact that neoplasms arise from changes in the genome does not imply that DNA itself must always be modified to treat them. Specific molecules that might antagonize transforming signal transduction, or compete for abnormal growth-factor receptors, can also be envisioned [19, 20], and one (suramin) is already in use [21]. Additionally, the alteration of neoplastic cells by gene transfer has been proposed [22].

These novel approaches [23] to the treatment of human neoplasms are as likely to be useful as the development of yet another nonspecific cytotoxic drug, and pursuing them should certainly teach us more. The real successes we have achieved with such drugs justify their continued investigation but on a smaller scale. Our limited knowledge of neoplastic cell physiology makes attempts to base treatment on it a long shot but chemotherapy has been and will remain a long shot. It is time to devise rational methods to treat the human neoplasms we have begun to understand.