Barry B. Lowitz and Dennis A. Casciato
“If two treatments are equally ineffective, and one has fewer side effects, use neither.”
B.B.L.
I. SOME PRINCIPLES OF CANCER BIOLOGY AND CANCER TREATMENT
A. Normal cell reproduction
1. The cell cycle is depicted in Figure 1.1. Cell replication proceeds through a number of phases that are biochemically initiated by external stimuli and modulated by both external and internal growth controls. Certain oncogenes and cell cycle–specific proteins are activated and deactivated synchronously as the cell progresses through the phases of the cell cycle. Most cells must enter the cell cycle to be killed by chemotherapy or radiation therapy. Many cytotoxic agents act at more than one phase of the cell cycle, including those classified as phase specific.
a. In the G0 phase (gap 0 or resting phase), cells are generally programmed to perform specialized functions. An example of drugs that are active in this phase is glucocorticoids for mature lymphocytes.
b. In the G1 phase (gap 1 or interphase), proteins and RNA are synthesized for specialized cell functions. In late G1, a burst of RNA synthesis occurs, and many of the enzymes necessary for DNA synthesis are manufactured. An example of drugs that are active in this phase is L-asparaginase.
c. In the S phase (DNA synthesis), the cellular content of DNA doubles. Examples of drugs that are active in this phase are procarbazine and antimetabolites.
d. In the G2 phase (gap 2), DNA synthesis ceases, protein and RNA syntheses continue, and the microtubular precursors of the mitotic spindle are produced. Examples of drugs that are active in this phase are bleomycin and plant alkaloids.
e. In the M phase (mitosis), the rates of protein and RNA synthesis diminish abruptly, while the genetic material is segregated into daughter cells. After completion of mitosis, the new cells enter either the G0 or the G1 phase. Examples of drugs that are active in this phase are plant alkaloids.
Figure 1.1. Phases of cell growth.
2. Cyclins activate the various phases of the cell cycle. Most normal cells capable of reproduction proliferate in response to external stimuli, such as growth factors, certain hormones, and antigen–histocompatibility complexes, which affect cell-surface receptors. These receptors then transduce the signal that results in cell division. Tyrosine kinases (TKs) are an essential part of the cascade of proliferative signals, from extracellular growth factors to the nucleus. Cyclins combine with, activate, and direct the action of special TKs, called cyclin-dependent kinases.
3. Cell cycle checkpoints. Cells that are capable of reproducing are normally stopped at specific phases of the cell cycle called checkpoints. The most important of these are immediately preceding the initiation of DNA synthesis and immediately preceding the act of mitosis. These histologically quiescent periods are probably mediated by decreased activity of cyclin-associated kinases and tumor-suppressor proteins. In fact, the cells in these phases are biochemically active as they prepare proteins to enter the next phase of the cell cycle and correct any genetic defects before going on to reproduce.
a. Normal cells have mechanisms that detect abnormalities in DNA sequences. When DNA is damaged, a number of repair mechanisms replace damaged nucleotides with normal molecules. These mechanisms are most important during cellular reproduction to ensure that new genetic material in daughter cells is an exact copy of the parent cell.
b. The first checkpoint occurs in the late G1 phase, just before cells enter the S phase. Even if the proper extracellular signals are received and all of the machinery is in place for DNA synthesis, the DNA must be in an acceptable state, with no lesions, before the cell can leave G1. If lesions are detected, either they are repaired or the cell is made to undergo apoptosis. This stopping point is one of the actions of the p53 protein (see Section I.C.3.b).
c. The second checkpoint occurs just before the cell enters the M phase; the cell cycle inhibitors stop the cell until it is determined whether the new progenies are worthy successors with accurate genetic copies of the parent. A cell that has not completely and accurately replicated all of its DNA or that does not have the full complement of proteins, spindle materials, and other substances to complete mitosis is arrested at this checkpoint until everything is in order and before the M phase can begin.
4. Normal populations of cells have a small component of “immortal cells” that, when called on by signals from other parts of the organism, can replenish themselves and also supply daughter cells that mature and differentiate into specialized tissue cells necessary for the function of the whole organism. Although a few types of tissue can dedifferentiate, most cell types lose their vitality as they differentiate, enter senescence, and eventually die. The following four populations of normal cells can be identified in eukaryotes:
a. Germ cells, which are capable of reproducing themselves indefinitely, possibly as a result of going through meiosis. Unlike cancer cells, these cells must undergo a meiotic “event” to produce an immortal cell line.
b. Stem cells, whose only two functions are to reproduce and to produce cells destined to differentiate and perform specialized functions for the host. Unlike cancer cells, these cells have a limited biologic number of reproductive cycles.
c. Partially differentiated cells, which have limited capability to reproduce and whose progeny eventually become fully differentiated nonreproducing cells
d. Fully mature specialized cells, which cannot reproduce further generations
5. Differentiation is inversely related to immortality. Unlike cancer cell lines that are immortal by definition, differentiated normal cells have a biologic “clock” that counts the number of times the cell can divide, after which no further division is possible. For example, a human fibroblast in culture can divide about 50 times, no matter what it is fed or the conditions of its culture, and then it and its progeny can divide no further.
B. Characteristics of cancer cells. Cancer can be defined as a cellular disorder characterized by progressive accumulation of a mass of cells, as a result of excessive reproduction of cells not compensated by appropriate cell loss; these cells progressively invade and damage the tissues and organs of the host. Although cancer cells are abnormal and die at a faster rate than their normal counterparts, the death rate is not able to keep up with the formation of new cells. This imbalance is the result both of genetic abnormalities in cancer cells and of the inability of the host to detect and destroy such cells. Some of the unique characteristics of cancer cells follow.
1. Clonal origin. Most cancer cells appear to originate from a single abnormal cell. Some cancers arise from multiple malignant clones either as a result of a field defect, in which multiple cells of a tissue are exposed to a carcinogen (such as the upper airways in smokers), or as a result of inherited defects in certain genes.
2. Immortality and telomeres. Most normal cells have limits on the number of reproductive cycles that a cell can have as it matures. Cancer cells, on the other hand, can proliferate indefinitely, providing an inexhaustible pool of precursor cells. One mechanism for immortality involves telomeres, the ends of chromosomes. Telomeres of most types of normal cells progressively shorten as the cells differentiate. In contrast, the telomeres of cancer cells and stem cells are replenished by the enzyme telomerase. This enzyme normally progressively decreases in a programmed manner as cells differentiate; the fully differentiated cell becomes senescent and eventually dies as it loses its ability to reproduce. In contrast, telomerase production is preserved or activated in many types of cancer cells; consequently, the length of the telomeres remains intact, and the cell remains “immortal.”
3. Genetic instability due to defects in DNA repair and in detection of DNA mismatches leads to heterogeneity of cancer cells. The cancer cells produce clones that become progressively less responsive to control mechanisms of proliferation and have an increased capacity to survive in “foreign environments” as metastases.
4. Loss of contact inhibition and anchorage-dependent growth. Normal cells grown in tissue culture do not divide unless they become anchored to a solid substratum to which they can adhere. Normal cells also stop dividing when they attain a confluent monolayer, even if the culture medium contains all the growth factors and nutrients necessary for further division. Cancer cells can grow independently in a semisolid medium without the requirement for substrate adherence; they continue to proliferate beyond a confluent monolayer in cell culture.
5. Progressive independence of proliferation from growth factors and nutrients is noted in cancer cell cultures. Cancer cells can actually self-destruct by continuing to divide even after they have consumed the nutritional factors in the culture media necessary for their survival.
6. Metastasis is a feature of cancer that is not found in normal tissues or benign tumors. The ability to metastasize results from the loss of or abnormalities of cellular proteins responsible for adhesion to the extracellular matrix, abnormalities in the interaction between cells, abnormal attachment to basement membrane, abnormal production of basement membrane, destruction of basement membrane by enzymes such as the metalloproteases (collagenases), and many other factors.
C. Causes of overproduction of cancer cells
1. Failure of abnormal cells to undergo apoptosis. Apoptosis is programmed cell death. An initial stimulus sets off an extremely complex cascade of events eventually resulting in apoptosis.
a. Apoptosis occurs in normal tissue reabsorption; the classic example is the disappearance of tadpoles’ tails. Apoptosis also results in the disappearance during embryogenesis of webs between fingers of primates, allowing the formation of individual digits. Apoptosis results in the elimination of normal senescent cells when they become old and useless and of thymic T cells that recognize “self” and thereby prevent immune attack by these cells on the host.
b. Apoptosis eliminates cells with abnormal DNA caused by either irreparable DNA damage or by inaccurate, incomplete, or redundant transcription of DNA. This is a major mechanism for maintaining chromosome number in cells of a particular species and in preventing aneuploidy. The process ensures that only cells that have fully and accurately replicated their entire DNA can enter mitosis.
c. Apoptotic cells can be recognized microscopically. Apoptotic cells show clumps of intracellular organelles in the absence of necrosis. The nuclei are condensed and fragmented; intracellular structures are degenerated and compartmentalized. As the cell falls apart, phagocytes take up the fragments. Unlike the process of cell necrosis, apoptosis does not cause an inflammatory response. Apoptosis requires synthesis of specific proteins that have been highly conserved throughout evolution.
d. Apoptosis is genetically regulated and may be perturbed in malignant cells. Cancer cells and some immunologic cells produce substances that promote inappropriate apoptosis in normal tissues (and may contribute to the cachexia of malignancy). For example, the p53 tumor-suppressor oncogene stimulates apoptosis. The Bcl-2 oncogene inhibits apoptosis, decreases normal cell death, and increases cell populations. Apoptosis may be the major mechanism by which tumor cell populations are decreased by hormones, cytotoxic chemotherapy, and radiation therapy.
e. Caspases. The final stage of the various death pathways is mediated through activation of the caspases, which represent a family of cysteine proteases. The activation of caspases is determined by the intrinsic and extrinsic pathways of apoptosis.
The intrinsic pathway is a mitochondrial-dependent pathway mediated by the Bcl-2 family of proteins. Exposure to cytotoxic stress results in disruption of the mitochondrial membrane, which then leads to release of protease activators. Caspase-9 is subsequently activated, setting off a cascade of events that commits the cell to undergo apoptosis.
The extrinsic pathway is mediated by ligand binding to the tumor necrosis factor (TNF) family of receptors, which includes TNF-related apoptosis-inducing ligand and others, and certain essential adaptor proteins. These adaptor proteins recruit various proteases that cleave the N-terminal domain of caspase-8, which leads to activation of the caspase cascade.
2. Genetic abnormalities that inappropriately stimulate cell proliferation, independent of normal proliferation signals, occur through a variety of mechanisms. Mutations or overproduction of receptors or transducing proteins can cause the cell to become independent of growth factor or other triggers and to initiate cell division independently. These gene abnormalities are usually dominant (i.e., normal cells hybridized with abnormal cells become phenotypically malignant).
3. Abnormalities of tumor-suppressor genes (genes that are responsible for suppressing cell division) probably result in cancer through failure of the host to destroy genetically abnormal cells. These genes are recessive; malignant cells hybridized with normal cells become normal.
a. Hereditary tumors. Retinoblastoma gene (RB1) was the first of these abnormal genes to be discovered. Subsequently, a number of other suppressor gene abnormalities have been found, particularly in uncommon or rare hereditary diseases. Examples include Wilms’ tumor (WT1), familial polyposis (APC), familial melanoma (CDKN20), and familial breast and ovarian cancers (BRCA-1 and BRCA-2).
b. p53 Suppressor gene. The most important example of these genes is the p53 suppressor gene. The p53 protein is a gene product that suppresses the cell cycle with multiple complex activities. It can detect DNA lesions, such as nucleotide mismatches and DNA strand breaks, including those caused by radiation and chemotherapy. This function of p53 is thought to be critical in preserving the integrity of the cellular genome.
(1) When DNA lesions are detected, the p53 protein arrests cells in the quiescent G1 and G2 phases of the cell cycle, preventing cells from entering the DNA synthetic (S) phase of the cell cycle. The p53 protein can then induce repair mechanism proteins or trigger proteins, which cause apoptosis.
(2) In the absence of intact apoptosis, cancer cells can continue through sequential cell divisions and accumulate nucleotide mismatches and progressive DNA mutations.
(3) In vitro studies have shown that chemotherapy and radiation kill cancer cells through DNA damage, which triggers p53 protein–induced apoptosis. In contrast, p53 protein–deficient mouse thymocytes and resting lymphocytes remain viable after irradiation.
(4) Many human cancers are found to have mutant p53 suppressor genes. Mutant p53 is characteristic of Li-Fraumeni syndrome, a hereditary autosomal dominant syndrome of both soft tissue and epithelial cancers at multiple sites starting at an early age.
4. Tumor angiogenesis. Cancer colonies are not larger than about 1 mm in diameter unless they have a blood supply. Colonies without adequate blood supply are not resting (not in the G0 cell cycle phase); they typically have a high rate of proliferation but a fully compensating cellular death rate. After the blood supply is established, the cellular death rate decreases, and the tumor grows rapidly.
a. Several substances are required to promote formation of new blood vessels (angiogenesis) in normal tissues. Almost all measurable cancers, however, are limited to the production of only one of these factors, called vascular endothelial growth factor (VEGF), which induces blood vessel formation. VEGF has a number of interesting properties that may be useful for cancer treatment including the following:
(1) VEGF induces receptors for itself on mature and nonproliferating blood vessel endothelial cells. These normal, resting endothelial cells do not have the receptor until they are exposed to VEGF.
(2) VEGF induces the production and activity of multiple other growth factors that contribute to blood vessel formation.
(3) VEGF can be induced by c-ras and by other oncogenes and growth factors, which then induce further production of VEGF.
(4) Unlike normal blood vessels that require other factors for normal development, blood vessels induced by VEGF are “leaky.” VEGF-induced plasma proteins, such as fibrinogen, can leak out of the new vessels, forming a spongy gel around the tumor. This gel contains VEGF that induces further angiogenesis.
(5) VEGF appears to prevent apoptosis in induced endothelial cells.
b. Tumors also elaborate angiogenesis inhibitors, which can decrease growth of tumor at distant sites. One form of murine lung cancer gives rise to metastases that elaborate such inhibitors and suppress the growth of the primary site. This mechanism may account for the difficulty in locating primary tumors that present with metastases and for the absence of a detectable primary tumor (metastases of unknown origin).
5. Population kinetics. Tumor growth depends on the size of the proliferating pool of cells and the number of cells dying spontaneously. The larger the tumor mass, the greater the percentage of nondividing and dying cells and the longer it takes for the average cell to divide. Figure 1.2 demonstrates the theoretic tumor growth curve based on the Gompertzian model of tumor growth and regression. Some features of this sigmoid-shaped curve on logarithmic coordinates are as follows:
Figure 1.2. Tumor growth in diameter found on chest x-ray film or by breast examination. Tumors with 1012 and 1013 cells (about 2 to 20 lb of cancer) usually result in damage to vital organs and death of the patient.
a. The lag phase. During the earliest phase of tumor development, a small mass of a tumor does not enlarge very much. The working hypotheses about this lag phase are that the “pre-cancer” cells are dividing, but the rate of birth of new cells is offset by cell death. During this phase, the dividing cells are accumulating various mutations. These mutations help the surviving cells improve their adaptivity to the supply of nutrients, increase the rate at which the mutated cells divide, decrease the rate of apoptosis sensitivity (e.g., c-kit factor), provide them with invasive properties, make the mutated cells more responsiveness to host factors, and produce angiogenesis factors. Before the angiogenesis factors are expressed, the small tumor does not have its own blood supply and is dependent on local factors to get all the necessary nutrients. Although not shown on the graph, animal models suggested that these tiny cancers may remain unchanged in size and undetectable for many years before they enter the logarithmic phase and are large enough to be detectable.
b. The log phase. The tumor now shows rapid exponential growth of the tumor mass. Hypothetically, the reasons for this phase are a relatively high proportion of cells are undergoing division, with rapidly declining rates of cell death, and the growth fraction (ratio of dividing to total cells) is high. This rapid growth also reflects the adaptivity of the cells and the production by the tumor cells of angiogenesis factors that induce the surrounding tissues to form new blood vessels that “feed” the tumor mass. When the tumor’s growth fraction is at its highest level, it is still clinically undetectable. Although the reduction in cell number is small, the fractional cell kill from a dose of effective chemotherapy would be significantly higher than later in the course of the tumor.
c. The plateau phase. Tumor growth slows down as the percentage of dividing cells decreases, and a larger percentage of cells are dying. The hypotheses are that growth rates eventually plateau because of restrictions of space and nutrient availability, of blood supply, and of genetic mutations, which cause a higher death rate of cells. The curve becomes asymptotic with some maximum.
d. 1 × 109 cells represents 1 g or 1 cc of tissue (equivalent to a tumor that is 1 cm in diameter). A 50% reduction in tumor mass represents only a one-third log decrease in tumor volume. For example, a tumor mass on x-ray film containing 8 × 1010 cells that is reduced to half its volume by chemotherapy still contains 4 × 1010 cells.
II. PRINCIPLES OF CANCER CHEMOTHERAPY
A. Categories of chemotherapeutic drugs. Cytotoxic agents can be roughly categorized by their activities relative to the cell generation cycle.
1. Phase nonspecific
a. Cycle-nonspecific drugs kill nondividing cells (e.g., steroid hormones, antitumor antibiotics except bleomycin).
b. Cycle-specific, phase-nonspecific drugs are effective only if the cells proceed through the generation cycle, but they can inflict injury at any point in the cycle (e.g., alkylating agents).
c. Pharmacokinetics. Cycle-nonspecific and cycle-specific, phase-nonspecific drugs generally have a linear dose–response curve: The greater the amount of drug administered, the greater the fraction of cells killed.
2. Phase specific
a. Cycle-specific, phase-specific drugs are effective only if present during a particular phase of the cell cycle.
b. Pharmacokinetics. Cycle-specific, phase-specific drugs reach a limit in cell-killing ability, but their effect is a function of both time and concentration. Above a certain dosage level, further increases in drug dose do not result in more cell killing. If the drug concentration is maintained over a period of time, however, more cells enter the specific lethal phase of the cycle and are killed.
B. Cancer treatment involves the exploitation of the biologic characteristics of cancer cells to make them susceptible to drug therapy. Although malignant cellular proliferation occurs in the absence of normal internal and external growth controls, cancer cells depend on the same mechanisms for cell division that are found in normal cells. Damage to those mechanisms leads to cell death in both normal and malignant tissues.
1. Therapeutic selectivity. Both radiation therapy and chemotherapy exert their initial effects greater in neoplastic cells than in normal host tissues because normal tissues have intact genetic machinery. Normal cells, for example, in the bone marrow and gut and in contrast to cancer cells, are able to repair DNA damage and to destroy cells with irreparable DNA, rather than allowing damaged cells to progress through the normal cell cycle and potentially replicate their damaged DNA. Loss of normal tissue cells as the result of DNA damage triggers proliferation of normal tissue cells and replacement of the lost cells in a self-limited manner.
2. Exploitation of apoptosis in cancer. Cancer cells with intact mechanisms for apoptosis can be forced to undergo apoptosis by irreversible damage to their DNA. Radiation therapy and most cytotoxic antineoplastic agents kill cancer cells by damaging the cell and inducing apoptosis. Ideally, when cancer stem cells are destroyed, the cellular “template” for the production of the malignant phenotype is diminished or destroyed; thereby, these cells would not be replaced by more of their kind.
3. Exploitation of proliferation control factors in cancer
a. Biologic response modifiers have been used primarily to stimulate selected immune system cells, which then demonstrate anticancer activity. These modifiers include interferons, interleukins, and several growth factors.
b. Activation of epidermal growth factor receptors (EGFRs) and its downstream signaling events play a key role in regulating tumor cell growth and proliferation, DNA repair, invasion, metastasis, and angiogenesis. Increased expression of EGFRs is observed in a broad range of solid tumors. A number of clinical studies have correlated the expression of EGFRs with disease progression, poor treatment outcome, and poor patient survival. Other studies have shown no such relationship with EGFRs expression as measured in the laboratory. Several chemotherapeutic agents, such as erlotinib and lapatinib, exert their antitumor effect by inhibiting the TKs of EGFRs.
4. Exploitation of maturation abnormalities in cancer cells
a. Directly acting maturation factors force incompletely differentiated cells to fully mature. This technique is exemplified by transretinoic acid for the treatment of acute promyelocytic leukemia. Other agents, such as vitamin D and cytosine arabinoside, can induce maturation of some types of leukemic stem cells in vitro.
b. Eradication of stem cells can leave behind a population of maturing cells, which then complete their differentiation into mature, nonmalignant tissues. This phenomenon is demonstrated by the finding of residual tumor masses of benign teratoma cells after successful treatment of germ cell tumors.
5. Angiogenesis inhibition: Exploitation of the dependence of cancer cells to induce the formation of their own blood supply to proliferate. Angiogenesis inhibition has potential in controlling tumor growth by limiting tumor blood supply, with only limited effects on normal revascularization. Counteraction of the antiapoptosis effect of VEGF may prevent accumulation of genetic defects that make cancers more aggressive with time. Many inhibitors of angiogenesis are known. Some of these agents are useful in cancer therapy, including pentostatin, interferons, glucocorticoids, and bevacizumab.
C. Mechanisms of drug resistance
1. Tumor cell heterogeneity. Spontaneous genetic mutations occur in subpopulations of cancer cells before their exposure to chemotherapy. Some of these subpopulations are drug resistant and grow to become the predominant cell type after chemotherapy has eliminated the sensitive cell lines. The Goldie-Coldman hypothesis indicates that the probability of a tumor population containing resistant cells is a function of the total number of cells present and its inherent mutation rate. This hypothesis asserts the high likelihood of the presence of drug-resistant mutants at the time of clinical presentation, even with small tumors. The Goldie-Coldman model also predicts that the maximal chance for cure occurs when all available effective drugs are given simultaneously (an impossibility to execute).
2. Single-drug resistance
a. Catabolic enzymes. Exposure to a drug can induce the production of catabolic enzymes that result in drug resistance. The drug is catabolized more rapidly inside the cell by gene amplification of DNA for the specific catabolic enzymes. Examples include increased dihydrofolic reductase, which metabolizes methotrexate; deaminase, which deactivates cytarabine; and glutathione (GSH), which inactivates alkylating agents.
b. GSH is essential for the synthesis of DNA precursors. Increased levels of GSH enzymes have been found in various cancers and not in their surrounding normal tissues. GSH and its enzymes scavenge free radicals and appear to play some role in inactivation of alkylating agents through direct binding, increased metabolism, detoxification, or repairing DNA damage.
c. Resistance to topoisomerase inhibitors may develop with decreased drug access to the enzyme, alteration of the enzyme structure or activity, and increased rate of DNA repair, and as the result of the action of multidrug-resistance protein.
d. Transport proteins. Exposure to a drug can induce the production of transport proteins that lead to drug resistance. As a result, smaller amounts of the drug enter the cell or larger amounts are carried out because of adaptive changes in cell membrane transport. Examples include methotrexate transport and the multidrug-resistance gene.
D. Mechanisms of multidrug resistance. Resistance to many agents, particularly antimetabolites, may result from mutational changes unique to that agent. In other cases, however, a single mutational change after exposure to a single drug may lead to resistance to apparently unrelated chemotherapeutic agents.
1. P-170 and the mdr-1 gene. The process of multidrug resistance appears to occur as a result of induction or amplification of the mdr-1 gene. The gene product is a 170 Dalton membrane glycoprotein (P-170) that functions as a pump and rapidly exports hydrophobic chemicals out of the cell.
P-170 is a normal product of cells with inherent resistance to chemotherapy, including kidney, colon, and adrenal cells.
P-170 membrane glycoprotein can be induced by and mediates the efflux of vinca alkaloids, anthracyclines, dactinomycin, epipodophyllotoxins, and colchicine. When exposed to one of these drugs, the cells become resistant to the others but remain sensitive to drugs of other classes (e.g., alkylating agents or antimetabolites). Calcium-channel blockers (e.g., verapamil), amiodarone, quinidine, cyclosporine, phenothiazines, and other agents have been studied for their ability to reverse or block the effects of P-170.
2. Loss of apoptosis as a mechanism of drug resistance. All cells, including cancer cells, must have intact mechanisms for replication and repair to avoid loss of information necessary for survival. Loss of apoptosis is manifested by the increasing aneuploidy often seen as cancers become more aggressive and by the very high frequency of mutations in the p53 suppressor gene.
a. p53 is a tumor-suppressor protein and is a potent inducer of apoptosis within a cell in which DNA damage has occurred (see Section I.C.3.b). DNA-damaging agents cause increased levels of p53 in normal cells. Mutations in the p53 gene are present in >50% of all human tumors.
The wild-type p53 suppresses the promoter of the mdr-1 gene, while mutant p53 protein can stimulate the promoter. Various tumors expressing mutant p53 or deleted p53 are resistant to a wide range of anticancer agents. Dysregulation of the p53 pathway might well be a prominent mechanism of drug resistance due to the overproduction of gene products responsible for entry into S-phase and rapid cell growth. However, loss of p53 function is not always associated with chemoresistance.
b. Bcl-2 is a potent suppressor of apoptotic cell death. Permutations of Bcl-2 expression (or related genes) can result in either repression or promotion of apoptosis triggered by γ-irradiation or chemotherapeutic agents. Bcl-xL, a functional and structural homologue of Bcl-2, is also able to confer protection against apoptosis induced by radiation as well as by several anticancer agents, including bleomycin, cisplatin, etoposide, and vincristine.
c. NF-kB (nuclear factor-kappa B) activation results in potent suppression of the apoptotic potential of a number of external stimuli including various cytokines, TNF-α, and radiation. Activation of NF-κB expression in response to chemotherapy may represent an important mechanism of inducible tumor chemoresistance.
d. The relationships among p53 status, NF-κB, Bcl-2, the caspase cascades, and chemotherapy sensitivity and resistance are obviously complex.
III. CLINICAL USES OF CYTOTOXIC AGENTS
A. Indications. Chemotherapy is used in the following circumstances:
1. To cure certain malignancies
2. To palliate symptoms in patients with disseminated cancer when the potential benefits of treatment exceed the side effects of treatment
3. To treat asymptomatic patients in the following circumstances:
a. When the cancer is aggressive and treatable (e.g., acute leukemia, small cell lung cancer, intermediate/high grade lymphoma)
b. When treatment has been proved to decrease the rate of relapse and increase the disease-free interval or increase the absolute survival (stage III colon carcinoma, stages I or II breast carcinoma, osteogenic sarcoma)
4. To allow less-mutilating surgery by treating first with chemotherapy alone or in combination with RT (sarcomas and carcinomas of the anus, breast, esophagus, and larynx)
B. Contraindications. Chemotherapeutic agents are relatively or absolutely contraindicated in the following situations:
1. When facilities are inadequate to evaluate the patient’s response to therapy and to monitor and manage toxic reactions
2. When the patient is not likely to survive longer even if tumor shrinkage could be accomplished
3. When the patient is not likely to survive long enough to obtain benefits from the drugs (e.g., severely debilitated patients)
4. When the patient is asymptomatic with slow-growing, incurable tumors, in which case chemotherapy should be postponed until symptoms require palliation
C. Adjuvant chemotherapy is given to patients who have no evidence of residual disease but who are at high risk for relapse. The justifications for adjuvant chemotherapy are the high recurrence rate after surgery for apparently localized tumors, the inability to identify cured patients at the time of surgery, and the failure of therapy to cure these patients after recurrence of disease. The disadvantages of this therapy are the immediate patient discomfort and the short- and long-term risks associated with such treatment. To date, the only malignancies for which adjuvant chemotherapy has proved beneficial are breast cancer, colon cancer, and osteogenic sarcoma.
D. Sequential versus alternating cycles. The Norton-Day model indicates that the sequential use of combinations of chemotherapy is likely to outperform alternating cycles because no two combinations are likely to be strictly non–cross-resistant or have equal killing capacity.
IV. TERMINOLOGY USED IN CLINICAL TRIALS
A. Statistics as a tool of medicine. Statistics helps doctors to maximize the benefits and minimize the risks in recommending treatment. Statistics is a kind of ruler that measures the probability of what happens to a large number of people who participate in a study. A word of caution: Statistics cannot predict the future or the positive benefits or harm for any individual patient.
Imagine a universe that has two million inhabitants and every one of them has metastatic colon cancer. Aside from this malady in common, no one of these patients is exactly like another; they are of different ages, genders, races, levels of activity, extent of the tumor, and so on. Now, let us suppose that there is one lonely investigator who wants to try a particular treatment. The researcher cannot physically conduct the study on all two million people. The best the researcher can do is to try to determine how closely the results for the population studied looks like the whole two million. The researcher knows that the measurement will not be completely accurate, but it is possible to measure the probability of inaccuracy that the error will fall into some range of values. By studying the variation of results in the sample, the researcher can make a more consistent hypothesis of the risks and benefits of the treatment.
B. Definition of terms used for describing types of drug development trials
1. Phase I trials determine the optimal dose, schedule, and side effects of a new therapy.
Phase Ia: First in human (FIH) trial
Phase Ib: Drug combinations involving FIH drugs
2. Phase II trials determine what kinds of cancer respond to a particular treatment.
Phase IIa: Searching for activity in disease types
Phase IIb: Randomized phase II designs
3. Phase III trials are large-scale randomized trials that compare a treatment shown to have some effectiveness in a phase II trial with no treatment or with a treatment that has also shown effectiveness.
4. Meta-analysis is a retrospective study in which data from multiple randomized trials are pooled and analyzed. All patients, including those being treated but not entered into each study, must be accounted for. Meta-analysis is most useful for evaluating many small, randomized trials to look for an effect not evident in a single small study and for identifying subsets (by prognostic strata) of patients who benefit from treatment.
C. Definition of terms used in describing the design of a study
1. Sample space is the number of patients, tests, treatments, or other data points used to represent the entire “universe” of all such patients, tests, and outcomes.
2. Stratification of patients according to known prognostic factors (such as age, ethnicity, sex, performance status, and extent of disease) is essential if a clinical study is to be useful for the clinician making treatment decisions. Randomizing patients between a treated group and an untreated group is a technique used to deal with unknown factors that could affect prognosis.
3. Randomization is the assignment of a patient to a particular treatment by random chance. Randomization is done when a treatment is being compared with another treatment or with no treatment. Each of the treatments or no treatments is called an arm of the study.
4. Blinded studies are studies in which patients do not know to what treatment arm they have been assigned. In a double-blinded study, neither the patients nor the investigators know to what arm of the study any patient has been assigned. The data are codified; the study is stopped (“broken”) if one arm is significantly providing better or worse results than another study arm.
5. Matched populations. Treated and untreated patients must be closely matched for particular characteristics. For example, no information can be drawn about the effectiveness of a sarcoma therapy if the treated group consists of otherwise healthy, young Asian women with low-grade localized sarcomas and the untreated group consists of elderly diabetic white men with high-grade metastatic sarcomas.
6. Risks and benefits of a treatment are an essential part of statistical study design. For example, a study that conclusively demonstrates that a treatment results in 1% improvement in 5-month survival but increases time in the hospital for therapy of treatment-related complications in 50% of patients and causes death in 15% should probably not be recommended.
7. Sufficient duration in a clinical study is essential for determining the effectiveness of treatment. Inadequate duration of a study may result in an effective therapy being considered ineffective. Studies of a treatment for most of the common cancers require at least 5 to 10 years for any meaningful interpretation. Earlier data are often misleading.
D. Definitions of terms used to describe frequency of cancers
1. Incidence refers to the overall number of people developing cancer in a particular time frame, usually 1 year.
2. Incidence rate is the number of people developing cancer per 100,000 population per year.
3. Mortality rate refers to the number of people dying from cancer per 100,000 population per year.
4. Case fatality rate is the percentage of people with a particular cancer who die from that cancer.
5. Prevalence is the number of cases of cancer in a population at a specific point in time.
E. Definition of terms used in survival analysis
1. Cure is a statistical term that applies to groups of cancer patients rather than to individual patients; it describes those patients who are rendered clinically free of detectable cancer and who have the same survival expectancy as a healthy age-matched control group. A cure does not guarantee that the particular patient meeting these criteria will not eventually die from the original cancer.
2. Actuarial survival (or life table survival) is the life expectancy from a specified age of a group of patients with a particular cancer. These data are used to determine the chance that an individual patient will survive for a specified time. This parameter is useful in determining both the natural history of the cancer and the effectiveness of treatment by comparing patient survival with actuarial survival tables of a matched healthy population.
3. Observed survival rate is the percentage of patients alive at the end of a specified interval of observation from the time of diagnosis.
4. Relative survival rate corrects the survival rate for the “normal mortality expectation” in a matched population.
5. Adjusted survival rate corrects the survival rate by discounting deaths from causes other than cancer or cancer treatment in those patients who are free of cancer at the time of death.
6. Median survival is the time when 50% of patients are dead and 50% are still alive. Average or mean survival rates are meaningless because survival of patients with similar tumors may range from a few weeks to years. Median survival may be a useful index for comparison of clinical trials but can be misleading. In “mature studies,” a significant group of patients may survive for many months or years after the time that 50% of the patients have died.
7. Disease-free interval is the time from when the patient is rendered free of clinically detectable cancer until recurrent cancer is diagnosed.
8. Censored data. Data for patients who are still living and discontinue the trial therapy or whose fate is unknown are frequently excluded (censored) from statistical analysis. Censoring data can severely skew results and make a study uninterpretable. The larger the number of censored data in relationship to the overall study, the more likely it is that the study is not interpretable. Good reporting carefully defines the reasons for censoring data, what the data would look like if censored data were included, and the percentage of data points that were censored.
9. Overall 5-year survival rate is an arbitrary but convenient rate used to provide a short-range assessment of the value of therapy and the adverse effects of therapy. It is used for all cancers and cancer therapies because most of the frequent and long-term adverse effects of drugs are found within this time frame. Five-year survival does not represent cure or complete eradication of detectable tumor, nor does it predict future tumor recurrence in complete responders. However, the rate of future recurrences for many tumors declines significantly. This is especially true for rapidly growing, aggressive cancers.
a. Early detection. Much of the treatment-related improvement in survival has occurred as the result of early detection of common cancers; it is not yet clear whether some of these early cancers have the same lethal biologic potential as cancers discovered in a more advanced state. Early detection and diagnosis of cancers that appear histologically identical but are biologically less lethal can yield misleading survival figures.
b. Lead-time bias. Patients appear to live longer from the time of diagnosis only because the cancer was detected earlier rather than because of treatment.
F. Definition of terms used to describe response to treatment
1. Complete remission (CR). No clinically detectable cancer is found after treatment.
2. Partial response (PR). Measurable tumor mass decreases by 50% after treatment, no new areas of tumor develop, and no area of tumor shows progression. The approximation of the mass of an individual area of tumor is usually given as the product of two diameters of the lesion; the measurable tumor mass is the sum of the masses of all measurable lesions.
3. Minimal response (MR) is the same as PR, but the response does not meet the criteria of 50% reduction.
4. Progression. The mass (product of diameters) of one or more sites of tumor increases >25%, new lesions appear, or the patient dies as a result of the tumor.
5. Stable disease. Measurable tumor does not meet the criteria for CR, PR, MR, or progression.
G. Statistical measurements used to assess clinical research results
1. P Value (the false-positive value) is the probability that a measured difference between results of two arms of a study, or a confidence interval (CI) difference, would disappear if the entire universe of identical patients could be tested. When the P value is <0.05, the probability that measured differences occurred by chance is <5%. The choice of a 5% false-positive error is somewhat arbitrary and is used to give a value that is two standard deviations from the average value that test results are “the true value.” The P value has no significance in studies of small numbers of patients evaluated over a short time.
2. β Error (the false-negative value) is the probability that two treatments that appear to have identical results would be different if the entire universe of patients were tested. This error is not usually reported in clinical research reports because it is a part of study design and is used in advance to determine the number of patients who must be entered into the study to meaningfully interpret a P value.
3. Power of a statistical test. The power of a test is calculated as “1 – β.” The power of a statistical test takes into account the number of patients enrolled. Higher powers give increasing confidence that any observed differences between arms are less likely to be by chance, no matter how many patients are enrolled.
4. Kaplan-Maier graphs and “acceptable error.” When we try to estimate the effectiveness or ineffectiveness of a treatment based on a clinical trial or meta-analysis, we can estimate the probability that the sample of patients, stratified by various parameters, is representative of the whole universe of patients with the same disease and the same treatment. On survival curves, the x-axis is marked off in equal time divisions and the estimated probability of survival of patients who are alive from the time they started participating in the study is plotted on the y-axis. A stepladder-shaped “curve” called a Kaplan-Maier plot results when the dots are connected. This curve can be made smoother by decreasing the time intervals.
Each curve is composed only of samples of the “universe” of all untreated and all treated patients. Consequently, we have the probability of several errors that can be estimated by statistically measuring the differences between curves. We must define an “acceptable” percentage error in deciding whether curves are different. Do we want to be 95% certain (two standard deviations from the mean) or 99.7% certain (three standard deviations from the mean)?
5. Confidence interval (CI) is the range of values around a measured result that would be 95% certain (with two standard deviations) to contain that result if the entire universe of patients were tested. The more patients included in a trial, the more likely the measured value would be closer to the “true” value.
6. Hazard ratios. The hazard rate is the number of people in a control or treatment group who experience a specific event (e.g., death in a survival analysis) during a specific time frame. This rate identifies those patients who are alive at the beginning of that interval and who die later in that time frame. During each time interval, the hazard rate of the treatment group is compared with the hazard rate of the control group, resulting in the calculated hazard ratio for the particular time frame. This so-called future survival is a hazard rate.
Hazard rates in survival analysis are used to determine the probable future survival time of a person who has lived for a known period of time after some event (e.g., the start of chemotherapy). The hazard ratio is used to evaluate the overall improvement, if any, of a treatment group compared to a control group and is associated with a P-value and a CI. A hazard ratio of 1 would mean that the hazard rates of compared groups are not different.
The following is an example for doctor-friendly interpretation of hazard ratio and CI data. In a randomized study of treatment A and treatment B in patients with metastatic pancreatic cancer, the median overall survival times were found to be 11.1 months in group A and 6.8 months in group B. The results were expressed as “hazard ratio 0.57; 95% CI 0.45 to 0.73; P < 0.001.” This means that the risk of death in group A is 57% compared to the risk of death in group B at any point in time. There is a 95% probability that the true hazard ratio is within the CI (0.45 to 0.73). Note that all the ratios, including the calculated hazard ratio, lie between 0.45 and 0.73 and are <1. In this example, group A always has a death rate lower than the death rate of group B. The probability that this is a false-positive result is <0.1%
Common uses of information derived from hazard rates and ratios include
a. Determining the probability that treated patients would have a significantly higher probability of surviving than untreated patients
b. Determining the probability of benefit from second-line therapy and the probable duration of that benefit
c. Determining whether additional courses of therapy improve probable survival times by measuring the effects on survival duration starting 6 months after the first course of one or more courses of therapy
7. The single most important statistical concept for the clinician. Many studies that show initial “promising results” ultimately prove the ineffectiveness of such therapy. Most initially exciting “state-of-the-art” therapies provide no improvement. It is essential that the clinician avoids one-article quotes, “early data,” and data reported only in abstract form as a basis for making clinical decisions.
H. Recommendations for the clinical application of the literature to patient care decisions. The status of ongoing clinical trials must be reported before the data are “mature” to allow coordination of trials among different institutions and investigators. Thus, critical evaluation of research reports is essential for clinicians who are deciding whether to recommend a new treatment for a patient. No trial result, however, is a substitute for sound clinical judgment or for individualization of treatment for patients not enrolled in a study. The clinician should ask the following questions when reading the literature:
1. Are patients who benefited from the study treatment substantially like your patient with regard to age, sex, stage of disease, performance status, and other prognostic factors?
2. Does the study exclude certain patients? If so, what were the reasons for exclusion?
3. Did the study treatment produce toxicity that is unacceptable in view of the potential benefits? On the other hand, was improvement in survival so superior that the risk for toxicity and drug-related death is warranted?
4. Did the study stratify and randomize patients in a manner that allows clear interpretation of the data?
5. Was the study large enough to provide sufficient confidence that observed differences were not by random chance?
V. A PHILOSOPHY OF MEDICAL PRACTICE AND ITS TOOLS
A. The principal goal of medical care is to provide the most beneficial treatment at the lowest risk. Although considered to be self-evident for medical care, this axiom is no simple matter to implement. Every person is different from every other—even for one of a pair of identical twins. The meaning of “benefit” and “optimal” is a highly personal matter. A very large number of interacting parameters make it impossible to have a preconceived idea of the meaning of benefit for each patient.
We tend to think of a benefit as solely a physical result of a medical decision. The benefit may be partial or complete eradication of a disease, optimization of the patient’s physical ability to function, or the psychological comfort of the patient and his or her significant others. Similarly, “acceptable risk” is also a special idea. The optimal care in each situation requires the judgment of the physician on a very personal basis.
Too often, the tools of medicine (science, technology, and statistics) are mistaken for the art of care. Most people think of science as a mathematical description of cause and effect—“if you do A, then B will occur.” People are so complex that cause-and-effect thinking is all but useless. If we recommend active treatment, the best we can do is to guess what will happen. But, we can often estimate the probabilities of obtaining a certain result and the probability of the risk and degree of side effects.
Like all arts, there is no satisfactory “one size fits all” approach to patient care. Attempts to reduce medical practice into a fixed set of algorithms that are “evidence-based” creates “one size fits nobody” programs and can damage optimal care for individual patients. The tools of our art are necessary, but alone do not determine the quality of the art. “Quality care” requires experience, thoroughness, attention to detail, good judgment, and, most of all, caring about the quality of the resulting product.
B. Lowitzisms for oncology. Barry B. Lowitz, M.D., is a superb teacher and cofounder of the Manual of Clinical Oncology. He fashioned the following aphorisms, which reflect his humor and good sense and our philosophy in caring for patients with cancer:
The decision of one is more useful than the opinions of many.
Life is most valuable when there is little of it left.
Palliative care should reduce symptoms and not cause them.
Avoid the cutting edge of oncology because it slices up too many people.
Use technology to confirm your diagnostic impression, not to rule it out.
The withholding of technology requires as much skill and judgment as its employment. Do not use chemicals when time and words are indicated.
In medical practice, as in most of life, a lot can be done, a little should be done, and a vanishingly small amount must be done.
An art of medicine is the ability to make decisions in the face of insufficient information. Obtaining sufficient information would usually alter the host to the point that a different diagnosis and therapy are required.
Quality medical care almost always involves just being thorough and meticulous. On rare occasions, however, the physician may actually have to resort to thinking.