William H. McBride and H. Rodney Withers
INTRODUCTION
Clinically Relevant Physicochemical Events
Ionizing radiation (IR) interacts with matter in many different ways; however, for high-energy photons, a two-step process dominates. Energetic ionizing electrons are produced that directly ionize atoms and break chemical bonds. Subsequent physicochemical reactions involve production of free radical and other reactive species that are heavily influenced by the intracellular milieu, such as free radical scavengers and the quaternary structure of the biologic target. For low linear energy transfer (LET) x- or γ-rays, 1 Gy represents about 1,000 ionization tracks, and most of the free radicals formed such as hydroxyl radicals, singlet oxygen, superoxide, and hydrogen peroxide are from ionization of water, a cell’s major (about 90%) constituent. These oxygen-containing molecules are collectively, although not totally accurately, often referred to as reactive oxygen species (ROS). They generate oxidative damage within a cell by virtue of their unpaired valence shell electrons and are the major cause of low LET radiation damage (perhaps 70%)—the radiation acts largely indirectly, which has numerous implications for the biologic effectiveness of radiation therapy (RT).
The importance of the indirect pathway varies with the type (quality) of radiation as the density at which energy is deposited determines the amount and type of damage caused. For all IRs, the density is especially high at the end of the electron tracks. For high LET radiations, such as α-particles, dense ionizing takes place along and close to tracks, which number about four for each Gy. As a result, more energy is deposited directly in biologic molecules, and more direct rather than indirect damage results.
Because free radicals are involved, the biologic microenvironment where these events take place contributes to the outcome. If oxygen, or other electron-affinic molecules, is present,1 it can participate in the free radical cascadic reactions, but its major effect is probably to “fix” chemical damage in biologic molecules and limit chemical repair. Oxygen is therefore a potent radiosensitizer, whereas hypoxia will limit radiation damage. Within tumors, hypoxia is a potential cause of failure of RT. Considerable effort has gone into the search for oxygen mimetic drugs that might penetrate into tumor hypoxic regions; some, such as the imidazoles, have been used clinically with effect, although they are not without toxicity.
Conversely, antioxidants such as glutathione, which is present in millimolar amounts in cells, scavenge radiation-induced free radicals to limit damage. The nuclear bomb era prompted searches for radioprotectors of normal tissue. The organic thiophosphates, such as WR2721, which is hydrolyzed in vivo by alkaline phosphatase to the thiol metabolite WR-1065, received much attention. Its commercial version, Amifostine, has been approved by the Food and Drug Administration (FDA) for reduction of xerostomia in patients receiving RT for head and neck cancer, although its use has not gained general acceptance.
Free radicals react with substrates in many different ways and at different rates; however, in general, IR has an oxidizing effect that lowers buffering power of the antioxidants in a cell and changes its redox status. The sensitivity of cells to IR depends on the levels and redox potential of many molecules, and this may be heavily influenced by the cell’s metabolic state. For example, quiescent stem cells appear to have high levels of free radical scavengers and antioxidants that may account for their relative radioresistance.2 Cells contain many molecular sensors for redox changes. These may make conformational changes to initiate rapid molecular responses through activation of transcriptional and translational pathways.
DNA as a Biologic Target of Radiation
Whereas the primary ionization and excitation events associated with IR exposure are over within a second or so, the biologic consequences may last for life. IR will damage all cellular organelles, but its biggest footprint is in DNA. Indeed, the efficiency with which IR causes complex DNA damage, as opposed to other lesions, in both cycling and noncycling cells is why it is such an effective cytotoxic agent. Complex or clustered damage (also referred to as multiply-damaged sites) are multiple lesions located within about one helical turn of the DNA that involve both DNA strands and about 15 to 20 base pairs. These are a direct consequence of the characteristic way that IR spatially deposits energy in dense “packets.” For sparse IR about 30% of events are of this form, whereas it is >90% for α-particles. Few complex lesions result from everyday oxidative and chemical damage. Instead, of the 10,000–20,000 DNA lesions produced per cell per day, the preponderance is single-strand breaks (SSBs) and base damage. Many agents produce DNA damage—even breathing; however, IR is exceptional in that it causes a relatively large number of complex lesions in DNA that are frequently lethal and all cell cycle phases are affected, although not equally when it comes to survival.
Safeguarding the integrity of DNA so as to minimize mutation is a biologic imperative. A complex group of highly efficient DNA repair mechanisms have evolved to achieve this, with each focused largely on a type of DNA lesion. SSBs and base damages formed by IR (1,000 per cell per Gray), as with everyday oxidative damage, are rapidly and faithfully repaired using processes such as base excision repair (BER), which uses the complementary undamaged DNA strand as a template.
DNA double-strand breaks (DSBs) at 15 to 20 per cell per Gray are more significant than SSBs for lethality and carcinogenesis following radiation exposure; however, many of these also can be repaired. The two major DSB repair pathways are nonhomologous end joining (NHEJ), which is error-prone but efficient, and homologous recombination (HR), which contributes in the late S/G2 cell cycle phase. HR is error free because it uses the sister chromatid as a template but has slower kinetics. These are canonical mechanisms that exist to deal with DSBs produced physiologically during meiosis (HR), DNA replication (HR), and for the generation of specific immune receptors (NHEJ), as well as by pathologic oxidative damage. The key proteins required for both NHEJ (Ku70, Ku80, DNA-PKcs, XRCC4, XLF, DNA ligase IV) and HR (Rad51, Rad52, Rad54, BRCA2, RPA) have been identified by a combination of genetic experiments and examination of the composition of ionizing radiation–induced foci (IRIF) that form at sites of DSBs. MRE11, Rad50, and NBS1 act as a complex (MRSN) to participate in both HR and NHEJ. Many other molecules play important roles in DSB repair, including the protein mutated in ataxia telangiectasia (ATM), whose loss gives rise to extreme radiosensitivity in humans. Analysis of IRIFs has shown that their composition and number vary depending on whether the DSB is in euchromatin or heterochromatin and on the phase of the cell cycle when irradiation occurs.
On the other hand, complex IR-induced DNA damage is highly diverse and presents an especially difficult challenge for repair mechanisms. Breaks resolve slowly with time, and non-DSB lesions can sometimes convert to DSBs during processing. For example, SSBs may be converted into DSBs at replication forks, making BER relevant to the irradiation outcome. One consequence of this complexity is that the time taken for DNA repair varies with the repair mechanism and site of damage. “Fast” and “slow” components are the minimum consideration, with the latter having estimated half-lives of up to 4 hours and the former 50 minutes to 1 hour. Estimates of “repair” rates within a tissue are therefore very uncertain, and when the times for repair during clinical fractionated treatments are discussed, this usually refers to tissue recovery rather than DNA repair rates.
The complexity of the different DNA DSB repair mechanisms and their involvement at different cell cycle phases has major ramifications for cancer treatments. Chemotherapy agents frequently target cells in the S phase and emphasize the importance of the HR mechanism especially when combined with RT, even though NHEJ is the predominant repair mechanism overall because most cells are in the G0-G1 phase. Tumors that carry BRCA1 or BRCA2 mutations that compromise their ability to perform HR are sensitive to PARP-1 inhibitors that block BER mechanisms through a process known as synthetic lethality, which is where a mutation in either of two genes is not lethal but mutations in both cause cell death. These drugs are in clinical trials in combination with RT. Finally, critical defects in DNA damage and repair are not always easy to detect because they may affect only one cell cycle phase or one mechanism. For example, ATM or BRCA1 mutated cells do not always show radiosensitivity in terms of DNA damage or rate of DNA repair. The detection of such mutations also indicates the lesser role of these genes in DNA repair compared to certain others where the mutations cause fetal lethality.
Our ability to interrogate DNA repair mechanisms has been greatly enhanced by the use of IRIF assays. One early IRIF event is phosphorylation of a histone 2A subtype, H2AX, by one of several phosphoinositol 3-kinase-related protein kinases (PIKKs), including the ATM checkpoint kinase, DNA-PK, and ATR. Gamma-H2AX molecules bind to DSB over several megabases of the flanking chromatin and can easily be detected using specific antibody. This sensitive assay has become a routine way to measure DSB formation and repair.3 More importantly, this histone provides interaction surfaces for other repair and checkpoint molecules and activates kinase-dependent pathways that lead to a coherent downstream DNA damage responses (DDRs). The classical DDR through ATM drives activation of p53, whose primordial function was transcriptional activation of stress responses. Cell cycle checkpoints are activated to allow time for repair, as are cell death pathways to remove highly damaged cells.
Other Biologic Targets of Radiation
RT targets cellular structures other than DNA to activate signaling pathways. As mentioned earlier, some of these are driven by changes in redox. In fact, although IR is an oxidative stress, most of the ROS and reactive nitrogen species (RNS) that are generated by IR can come from secondary sources such as damaged mitochondria, activation of cation membrane channels and activation of NADPH and other oxidases, alterations in cellular metabolism, induction of nitric oxide synthetase, and proinflammatory cytokines such as members of the tumor necrosis factor (TNF) family (e.g., TNF-α, fasL, Trail) that can be generated by radiation. Low levels of ROS and RNS are an integral part of metabolism and participate in most cellular processes, including cell signaling. Higher levels can precipitate cell death. It is not surprising that through inducing oxidative stress, IR can activate redox-sensitive transcription factor systems, such as NF-κB, AP-1, HIF-1α, Nrf2, PPARγ, p53, Sp1, c-abl, and STAT3, and other molecules including membrane receptors such as epidermal growth factor receptor (EGFR). As a result, primary (also known as immediate early response) genetic programs can be activated rapidly without de novo protein synthesis following IR exposure. These rapid and relatively promiscuous responses become consolidated where appropriate by the generation of more restricted secondary gene programs. Importantly, these pathways couple molecular damage to DNA repair, cell cycle arrest, phenotypic changes, and cell death. They also serve to signal “danger” to the body through tissue damage responses (TDRs).
Cellular Damage Responses After Irradiation
The ways that cells and tissues “perceive” radiation damage influence the final outcome. In other words, they depend on pre-existing internal “sensors” that are influenced by metabolic status and the external molecular signals from hypoxia, cytokines, cell–cell and cell extracellular matrix interactions, the signaling pathways that are activated, and so forth. Because tumorigenesis involves mutations in molecular pathways that govern DNA repair, cell cycle, and cell death, it follows that genetic alterations associated with cancer frequently alter the cellular response to RT. These pathways are activated by stresses other than IR, including chemotherapy, hyperthermia, and inflammation making these interactions complex. Given the complexity of the biologic system, clinical outcomes cannot be predicted from the amount of physical energy deposited, but only from the biologic context. This is why what is known as “biologic dose” differs considerably from “physical dose.”
Because IR causes a relatively large number of complex lesions in DNA that are frequently lethal, it is a highly effective cytotoxic agent, and loss of reproductive ability of tumor cells is the desired therapeutic outcome. This can occur in several different ways, depending on the dose and the cell type. In 1956, Puck and Marcus4 noted: “Cells in which the ability to reproduce has been destroyed by doses below 800r can still multiply several times. At higher doses, even a single cell division is precluded. The mitotic death that they observed is the most common form of cell death caused by IR.5 It is a slow process that contrasts with rapid interphase death that is over in about 4 to 6 hours in certain cell types, including many lymphocytes, endothelial cells, and some cells in the salivary gland, thyroid, intestinal crypt, and hair follicles. Interphase death is now known to be by rapid apoptosis.6–7,8–9 Mitotic death, however, can be caused by any of several mechanisms—for example, failure of spindle formation in the M phase, loss of the G2 checkpoint leading to “mitotic catastrophe,” or improper chromosome segregation from damage and loss of genetic material—and the molecular machinery that is employed can be apoptotic, necrotic, or involve other mechanisms.
Radiation-induced apoptosis in normal tissues is often, although not always, dependent on activation of the DDR through p53 and its downstream effectors Bax and Bak.10 These disrupt mitochondrial membranes, releasing factors that activate the caspase cascade of proteolytic enzymes and endonucleases that cleave DNA between nucleosomes to commit cellular “suicide.”10 Apoptosis can therefore be recognized morphologically by cell shrinkage or by labeling the 5’-ends of apoptotic DNA strand breaks using terminal deoxynucleotidyl transferase (TdT) (TUNEL technique). This intrinsic mitochondrial factor activates caspase 9 to lead to apoptosis and often overlaps mechanistically with an extrinsic pathway activated through tumor necrosis factor receptor (TNFR) family members in the plasma membrane that activates caspase 8. Most notably, both utilize the same final executioner caspase 3–dependent pathway.
Because radiation can induce expression of both TNF and TNFR family members, these pathways may form an additional indirect pathway leading to death or survival of some cell types following irradiation.11,12 For example, mice lacking TNFR2, which does not have a death domain and drives cell survival, are particularly sensitive to late effects of IR to the brain.13 The cytotoxicity of certain chemotherapeutic agents, such as 5-fluorouracil (5-FU) and cisplatin, can also involve TNF-α.14
Apoptosis is a form of programmed cell death that during development shapes organs. In adults, its role is in physiologic homeostatic control—for example, removing excess cells at sites of proliferation and self-reactive lymphocytes. Pathologically stressed cells may use also this form of cell death (e.g., after IR exposure). Interestingly, IR-induced apoptosis signals “danger” in tissues through inflammatory signaling, whereas cells that die by physiologic apoptosis are immunologically silent. Many cells are protected against apoptosis. Only cells that have their internal molecular “rheostat” on a proapoptosis setting undergo this form of death in response to IR. Therefore, IR increases the frequency of apoptosis, although only in cells that have the relevant molecular circuitry—for example, lymphomas die in this way, whereas glioma cells tend not to. Tumors formed of cells that are predisposed to apoptosis often are among the most clinically sensitive.
In contrast to apoptosis, necrosis is a purely pathologic tissue injury process that does not involve activation of cellular pathways. Membrane integrity is lost, cells increase in size, lysosomal enzymes are released, vasculature is damaged, and inflammatory responses are generated. DNA is cleaved in an indiscriminatory fashion.
Autophagy is another possible alternative death style following IR exposure. This is a primordial survival response in which cells internalize their cellular organelles within vacuoles and digest them; it is most often seen in nutrient deprivation. Cells can be rescued from autophagy, although they die if taken to excess. Another important alternative radiation-induced outcome is senescence, which involves activation of cell cycle checkpoint molecules such as p21 and p16 and can be promoted by cytokine transforming growth factor beta (TGF-β),15,16 which drives collagen production. IR-induced senescence is therefore highly relevant to IR-induced fibrosis.17
Under normal circumstances, cell death in tissues is balanced by cell production and survival. Signals in the cell’s environment, such as growth factors, cell–cell contact, and extracellular matrix, are critical for survival, and their loss may lead cells to exit the cell cycle and cell death by anoikis,18 or “homelessness,” which may occur by any mechanism.
Apoptosis, autophagy, and senescence all have the effect of removing damaged cells from the reproductive pool. This may limit the chances of radiocarcinogenesis, but cells in senescence and autophagy may retain function for some time. The long-term outcome may be IR-induced fibrosis through enhanced collagen production,17 or re-entry into the cell cycle, or in the case of tumor cells, re-entry into a stem cell phase and subsequent tumor regrowth.
Because cells need positive signals to proliferate and survive, these pathways often promote radioresistance, and blocking them often results in radiosensitization. This is an oversimplification of complex biology, but it may help to explain why blocking EGFR tyrosine kinase signaling with cetuximab or gefitinib (Iressa), or NF-κB activation with bortezomib, may radiosensitize tumor cells. Survival pathway signaling may underlie some of the phenomena ascribed to potentially lethal damage repair (PLDR) following irradiation. In PLDR, the cellular microenvironment determines the likelihood of cell death. Classically, PLDR occurs when cells are irradiated and maintained in a contact-inhibited, plateau-phase culture. If such contact-inhibited cells are trypsinized immediately, or soon after irradiation, survival is compromised. One interpretation is that the cells are rendered homeless by trypsinization and are more likely to die. It must be noted, however, that PLDR, like autophagy, has been invoked as a mechanism that increases survival under diverse sets of experimental conditions and that multiple mechanisms contribute to the final outcome.
TISSUE DAMAGE RESPONSES
Molecular and Cellular Aspects
In tissues, cellular apoptosis can often be observed within hours of IR exposure. This is generally restricted to specific sites within the tissue. Radiation-induced rapid apoptosis in the mouse small intestine is maximal around position 4 from the base of the crypt,19 which is the site of most proliferation and most spontaneous apoptosis. In contrast, apoptosis is not so marked in the colon and is not seen in the proliferative region. Indeed, the antiapoptotic molecule Bcl-2 is expressed by cells in this region.20 In the thymus, about 98% of the cells that are generated die (about 5 × 107 cells per day in a young adult mouse), and after irradiation, there is massive rapid, p53-dependent apoptosis of T cells in the cortex but less in the medulla, which contains more mature cells. Concordance with the p53-related DDR is obvious in some tissues but not others. In lymphoid tissues, small intestine, hair follicles, and ependyma, the position of radiation-induced apoptotic cells correlates with up-regulated p53.19,21,22 In subpopulations of cells in other tissues, p53 can be up-regulated with little evidence of rapid apoptosis, whereas most cells in liver, skeletal muscle, and brain show neither p53 nor much apoptosis.10,14
The role of apoptosis in normal tissue responses to IR is still controversial. It may depend on the physiologic role of the proapoptotic cells. If they are superfluous to needs, IR-induced cell death may have little impact; however, if they are critical to tissue function, the opposite will be true. Although little information is available in humans, acute parotitis can develop in the first 24 hours of treatment of patients receiving head and neck irradiation, which reflects apoptotic death of serous cells.23 There is no such acute death in mucous cells; hence, the mouth is dry and the saliva more viscous.
The role of the vasculature in radiation responses is a long-standing controversy. The arguments have recently been resurrected by the suggestion that radiation-induced endothelial cell apoptosis through ceramide activation is critical for both normal tissue and tumor damage.24,25–27 There is little doubt that IR causes vascular damage, and in particular microvascular “pruning,” and this is an important aspect of tissue responses to IR, as well as to any ensuing hypoxia; however, parenchymal cell responses are also obviously relevant. A balanced view is that what is most critical is the integrated tissue response.
Part of the TDR occurs through the generation of pro-oxidant conditions that signal “danger” in damaged tissues through elaboration of proinflammatory chemokines and cytokines such as TNF-α and IL-1, as well as proteases, cell adhesion molecules, and extracellular matrix materials.28,29 This regulated acute tissue reaction has its roots in fighting tissue infection by processes of cell death, inflammation, cell proliferation, and wound healing with tissue regeneration and remodeling. IR doses above approximately 6 to 7 Gy are especially proinflammatory.30,31 Indeed, 2 Gy may have become a popular fraction size in the early days of RT simply because it minimized skin inflammation, the power output of x-ray tubes being low, and radiation dose to the skin being high.
This danger signaling can extend beyond the radiation field and may generate tumor-specific immunity by similar mechanisms to those that generate antipathogen immunity. These may be observed as “abscopal” effects of IR.32,33Cytokine-mediated responses can play many roles that could influence the overall outcome of RT. For example, radiation-induced basic fibroblast growth factor (bFGF) may act through autocrine pathways to promote survival of endothelial cells.34 In vivo, antagonists of radiation-induced IL-1 and TNF-α increase the intrinsic sensitivity of mice to bone marrow death after irradiation,35,36 suggesting that such responses have an adaptive survival value. On the other hand, radiation-induced TNF-α can cause certain cells to apoptose11 and may trigger clinical symptoms that are not associated with cell death. Examples are nausea or vomiting that can occur within hours of RT involving the upper abdomen; acute erythema and edema associated with vascular leakage; fatigue in patients receiving RT to a large volume, especially within the abdomen; and somnolence that can develop within a few hours of cranial irradiation. Radiation-induced proliferative responses such as gliosis37 or certain forms of fibrosis could also cause symptoms unrelated to cell depletion.
In the longer term, danger signaling by IR can set in motion a train of events culminating in continued proinflammatory responses in tissues long after exposure. Dose-related oscillating waves of inflammatory responses have been observed for months after IR tissue exposure. Failure to control these can be very damaging. Just as IR-induced redox imbalances are countered in time by antioxidant responses (e.g., superoxide dismutase, catalase, glutathione peroxidase, thioredoxin reductase, peroxidase), inflammation is controlled by anti-inflammatory mechanisms that critically mediate wound healing and tissue repair, and continuing waves of proinflammatory cytokines presumably reflect continual failing attempts at tissue recovery and remodeling.38 Again, the inflammatory aspects of the TDR can extend outside the irradiated site to mediate bystander regional and systemic effects of local RT; because of their persistence, late radiation effects can be considered to have a chronic inflammatory component.28,33
Cytokines and growth factors are important mediators of late IR effects. For example, signaling through the TNFR2 protects mice from late effects of brain irradiation.13 Anscher39 has reported that lung cancer patients with elevated plasma levels of TGF-β prior to RT are more likely to develop radiation pneumonitis, illuminating the systemic influence on local radiation damage. Elevated TGF-β levels could be from the tumor or the stromal cells that invade it, or may be IR induced; the outcome may be the same. Not surprisingly, inhibition of TGF-β activation during RT is being investigated as a strategy to lower the risk of pneumonitis in patients with non–small cell lung cancer.40 In addition, dose escalation is being attempted in patients whose TGF-β levels normalize during a course of RT.
Kinetics of Normal Tissue Radiation Injury
Although these more pathologic aspects of RT are an important aspect of the TDR, the time for a given tissue to express complications (latency) is determined in large part by the physiologic cell turnover rates—that is, on the kinetics of cell differentiation, loss, and renewal. As a result, the latency time to a complication is quite similar between individuals—for example, mucosal reactions generally occur at similar time intervals after the start of RT.
The terms acute, subacute, and late are commonly used to describe the time to occurrence of functional inadequacy after RT and reflect the major kinetic differences between tissues. The terms are also often loosely used to describe the tissues in which such effects are seen, as in “acute effects tissue,” but are misleading because tissues and organs comprise more than one cell type, each with its own turnover rate characteristics. Any one tissue can therefore express both acute and late symptoms of radiation damage, depending on the cell type that is limiting function at that time. In addition, a severe acute radiation injury can lead to nonspecific late (consequential) changes such as fibrosis, atrophy, or ulceration (e.g., stenosis consequent to mucosal ulceration of the bowel, or fibrosis or necrosis of skin or oropharyngeal tissues consequent to desquamation and acute ulceration).
Acute Responses
Acute responses to RT are defined as occurring during a standard 6- to 8-week course of treatment and are seen in tissues with large populations of cells that turn over rapidly (gastrointestinal mucosa, bone marrow, skin, oropharyngeal and esophageal mucosa). Hierarchical organization exists in such tissues with a small number of relatively quiescent stem cells that proliferate slowly to produce a highly proliferative compartment of progenitor cells that differentiate into mature, nonproliferative, functional cells. IR generally depletes the more radiation-sensitive progenitor cell pool first. Nonproliferating, differentiated cells, however, maintain tissue function until they are lost through continuing physiologic cell turnover. In many circumstances, such as after lethal doses of whole body irradiation, depletion of the progenitor pool may be the critical event leading to death through loss of function, not loss of stem cells per se. The recent discovery of stem cell markers is currently allowing these responses to be evaluated in many different tissues.
After irradiation, there is homeostatic proliferation in most tissues in an attempt to re-equilibrate, in particular with respect to the generation of sufficient functional cells. Depleted stem and progenitor cells may, however, first reconstitute their own numbers before differentiating to restore function, although this seems to vary with the tissue. A useful model to consider is that under normal steady-state circumstances (i.e., not growing or involuting), tissues have, by definition, a cell loss factor (φ) of 1. The only requirement for tissue growth is a decrease in (φ) to <1, which is characteristic of embryos and fetuses, tissue regeneration, and malignancy. After IR exposure, some tissues (e.g., jejunal crypts) appear to reduce (φ) to zero and regenerate quickly; others (e.g., skin) may reduce it to about 0.5 and regenerate less quickly, continuously producing some functional cells; others such as seminiferous epithelium show little change in (φ) and mostly continue in steady state, producing sperm in numbers that are reduced for months or years in direct proportion to the extent of stem cell depletion. This, however, is simply a model that requires further confirmation using defined stem cell markers.
Because acute-responding tissues are organized in a hierarchical fashion, the severity of radiation injury depends on both the extent of stem/progenitor cell depletion and the length of the delay before new functional cells are generated. Severity of injury increases with dose; however, providing the proliferative pool does not fall below a critical value, symptoms are transient and recovery can be complete. Dose fractionation can lessen the severity of acute effects by allowing regeneration from the stem/progenitor cell compartment during the course of therapy. Unlike the extent of injury, the rate at which acute injury develops and the latent time to the appearance of symptoms is relatively, although not completely, independent of dose because latency is mainly determined by the rate of loss of differentiated cells. For example, in hematopoiesis, leukocyte and platelet numbers drop quickly after bone marrow irradiation because they have a fast turnover rate, whereas anemia is not an obvious acute effect because red cells turn over slowly. In the testis, each spermatogenic stem cell division ultimately produces more than 1,000 sperm through successive divisions of spermatogonia and spermatocytes—a process that in humans takes more than 60 days. Early differentiating spermatogonia are few in number and are more radiosensitive than mature stages of spermatogenesis. This is why sperm counts remain normal for several weeks after exposure, falling steeply only at the time when the progeny of the irradiated spermatogonia would normally have reached the seminal vesicles. In the mucosa of the small bowel, because crypt cells divide rapidly (an average of more than once daily in humans), they are lost within days if sterilized by radiation. The nonproliferative villus shows no immediate effect of irradiation, with shortening becoming evident only as differentiated cells are shed into the lumen in the absence of renewal from the crypts. This is why symptoms take about 2 weeks to appear in patients undergoing abdominal RT.
Subacute Responses
Certain tissues may display subacute reactions several months after RT. Symptoms are generally reversible, although in some instances they may be associated with severe damage and even death. Examples of transient effects are Lhermitte’s syndrome after spinal cord irradiation and subacute pneumonitis 2 to 3 months after the start of lung irradiation. The target population appears to have a longer turnover time, and transient symptoms are considered due to a lag between loss of functional populations and stem/progenitor cell recovery, although production of proinflammatory cytokines driven by continuing cell loss and hypoxia due to microvasculature pruning may play a considerable role, as will the genetically determined mechanisms that control these responses.
Late Responses
Late reactions to RT in normal tissues can be severe, and recovery is often limited. They are generally considered to be the result of the depletion of slowly proliferating “target” cells that are lost from the tissue at a slow rate—for example, from central (oligodendroglia) or peripheral (Schwann cells) nervous tissue, kidney (tubule epithelium), blood vessels (endothelium), dermis (fibroblasts), and bones (osteoblasts and chondroblasts). Some lesions, such as those associated with atherosclerosis and heart disease, may occur decades after RT and are an increasing problem as patients live longer following therapy. Pathologic findings associated with late effects can be quite variable. For instance, late demyelination after brain irradiation has often been ascribed to loss of oligodendrocytes, and subsequently of neurons, but proliferation of astrocytes and microglial cells can be observed,41 as can vascular lesions with edema, hemorrhage, or inflammatory infiltrates. Infiltrating cells may contribute to the pathogenesis of radiation injury as is illustrated by their involvement in radiation pneumonitis, or recovery from injury as in the slower healing of skin wounds in mice receiving total body irradiation compared with those irradiated only locally.
Unlike acute-responding hierarchical tissues, slowly proliferating tissues, in which late effects occur, contain cells that are usually both functional and able to proliferate on demand. In an operational sense, such tissues can be regarded as “flexible.” This does not deny the presence of stem cells with limited function or functional cells that do not proliferate; however, the roles of such cells are probably of lesser importance than in hierarchical tissues. The more chronic and debilitating nature of late reactions may be because of their relative inability to be repopulated from a stem cell pool. Alternatively, regeneration may be compromised by inflammation and fibrosis that is an alternative healing response for slow-responding tissues.
Because the proliferative cells may also be functional, dose has a greater apparent influence on latency in late reactions, with injury developing more quickly with increase in dose. This may be because the greater the dose, the fewer the number of division cycles the cells can successfully negotiate before death. Another reason may be that as cells die, residual (mostly lethally injured) target cells are increasingly recruited to the proliferative pool, causing a cascade or “avalanche” of cell death and functional tissue failure. A third possible explanation is that because late effects are complex and involve interactions between many cell types, the nature of the lesion may change with time depending on which cell type is critically limiting. The time course to development of injury can be accelerated and the severity increased by various insults such as surgery, chemotherapy, infection, or physical trauma. Indeed, such factors may play a major role in precipitating the onset of late effects in humans (e.g., necrotic, nonhealing ulcers after trauma). Conversely, slowing the proliferation process and decreasing stress may reduce their incidence and severity.
An issue of growing clinical importance is the extent to which late radiation effects can be reversed. It has been shown recently that certain agents given late after radiation can modify injury in tissues. For example, captopril, an angiotensin-converting enzyme inhibitor, slows the development of radiation-induced nephritis and lung fibrosis in rats. Steroids also can prevent death from radiation pneumonitis in animals, although their withdrawal before the end of the usual period of pneumonitis can result in accelerated mortality. Pentoxifylline, alone or in combination with vitamin E, protects against radiation-induced late effects in some experimental models. In a clinical study, the combination, but neither agent alone, reversed chronic radiation-induced fibrosis. It is not clear how these agents act, but such studies point to ways to improve the future management of late complications of RT.
Whereas a severe early response in a rapidly proliferating tissue permits adjustment of the dose schedule during the standard course of RT, this is not the case for late injuries because they occur after completion of treatment. Furthermore, because cell turnover kinetics determines the time to a normal tissue effect, latency is not an indicator of radiosensitivity.
Tolerance Doses and Functional Subunits
Normal tissue tolerance doses have not been precisely defined, even though generally accepted dose limits exist for various organs.
The tolerance of a tissue to IR is determined not only by its intrinsic radiosensitivity but also by the number of cells with regenerative potential and the way they are organized. Tissues can be thought of as being composed of functional subunits (FSU)—the minimum clonogenic entity required for regeneration of a structure. For example, epilation requires doses lower than those for desquamation, primarily because there is a smaller number of clonogenic cells in the FSU that produces a hair than in the sheet of basal cells that is capable of self-regeneration. Similarly, hair is depigmented by lower doses of radiation than the epidermis because each hair follicle contains a smaller number of melanocytes, sometimes only one.
In the kidney, each nephron is an FSU. If a tubule is completely de-epithelialized, it is lost permanently because it is not repopulated from adjacent nephrons. Therefore, the tolerance dose for the kidney is determined more by the number of tubule stem/progenitor cells per nephron than the number of nephrons. For example, if the kidney contained 1011 clonogenic tubule cells distributed as 104 cells in each of 107 nephrons, then most tubules should regenerate after a dose that reduced survival to 10−4. Because of the random nature of IR events, from Poisson distribution statistics, 37% of FSU nephrons would be eliminated. If on the other hand, 1011 clonogenic tubule cells were distributed as 107 cells in each of 104 nephrons, the dose required to eliminate 37% of the nephrons would have to reduce survival to 10−7. In a multifractionated dose regimen, during which a logarithmic decline in cell number occurs, this is 7/4(1.75) times higher. Tolerance doses can therefore vary greatly among tissues and organs, even if the target cells have the same intrinsic radiosensitivity. In mouse skin, the survival of about 10 out of approximately 106 basal stem/progenitor cells per cm2 is required to prevent overt desquamation. Therefore, the FSU would be about 1/10 cm2.
Organs with more tubular architecture (e.g., salivary glands, pancreas, sweat glands, testis, mammary epithelium, lung, and perhaps liver) may resemble the kidney in having better-defined FSUs organized “in series,” whereas the target cells in dermis, mucosa, gut epithelium, and epidermis may be considered to be organized “in parallel.” These lack restricting physical barriers to cell migration, which may assist tissue regeneration. Structurally well-defined organs appear to have small FSUs, whereas those in the spinal cord are intermediate and the dermis large. A tumor behaves as just one FSU, as one surviving stem/progenitor cell can lead to recurrence. The number of such cells will therefore play a big role in determining local tumor control. For metastatic deposits, in general a larger number of small metastatic deposits will be cured with lower radiation doses than a small number of larger metastases even if the total cell number is the same in the two cases.
FIGURE 2.1. Diagrammatic representation of the influence on the probability of a complication from increasing the treatment volume in a tissue where FSUs are arranged serially. The average survival of FSUs was 1 in 16, with sterilized FSUs being denoted by the black squares. With the small volumes (A), the probability of myelitis was 6% (1/16), whereas it would approach 100% if 16 FSUs in one patient were exposed (E). The actual probabilities can be calculated using the equation in the text. (From Withers HR, Taylor JMG, Maciejewski B. Treatment volume and tissue tolerance. Int J Radiat Oncol Biol Phys 1988;14:751.)
FIGURE 2.2. Curves illustrating how the probability of producing a complication increases with increase in the number of serially arranged FSUs included in the treatment volume. The curves were positioned by assuming that 58 Gy in 2-Gy fractions sterilized 10% of FSUs and that for a series of 2-Gy fractions, the effective D0 for the target cells was 4 Gy. The curves are shifted to the left and are steeper with increase in number of FSUs exposed; however, this effect becomes less obvious once large numbers of subunits are involved. (From Withers HR, Taylor JMG, Maciejewski B. Treatment volume and tissue tolerance. Int J Radiat Oncol Biol Phys 1988;14:751.)
Volume Effects
Traditionally, radiation oncologists have reduced the total dose when treating large volumes of normal tissue. In fact, the now widespread use of 1.8 rather than 2 Gy had its origin in a volume effect; the longer treatment duration enhanced mucosal tolerance in large head and neck treatment fields. In the orthovoltage era, a reduction in dose with increase in treatment volume was generally recommended but became less important with the advent of skin-sparing megavoltage beams. Modern intensity-modulated radiation therapy (IMRT) treatment planning gives a readout for dose-volume histograms for various normal tissues that are useful for avoiding excessive dose to too high a volume. However, they are limited because there is no spatial information and as a result do not predict toxicity well.
In reality, the concept of decreasing dose with increasing treatment volume has little cellular radiobiologic basis, except in specific circumstances. For example, if FSUs are arranged in series, as in tubular structures such as nerve tracts, the spinal cord, and the peritoneal sheath, the loss of one subunit may results in an overt expression of injury regardless of the state of the other subunits in the series. The probability of injury increases with volume (number of FSUs exposed) (Fig. 2.1). Such a volume effect has been demonstrated clinically for small bowel obstruction and experimentally for myelitis.
Experimental studies on volume effects of radiation in spinal cord of rats have indicated a steep volume effect for rat spinal cord at <10 mm of length and an interesting “bath-and-shower” effect42 for small fields with high tolerance for IR-induced paralysis. In such models, if an additional modest “bath” dose (about 4 Gy) is administered to 4-mm segments of spinal cord surrounding a targeted “shower” 2 mm in size, the effective dose for 50% paralysis (ED50) dramatically decreases from 88 to 61 Gy. The mechanism underlying this bath effect is not known, although inhibition of angiogenesis and cell migration are possible explanations.
Spinal cord has been a favorite model for volume effects. One recent study showed that IR-induced motor deficit in pigs appears to be independent of the irradiated volume in the lateral direction, with around 20 Gy being the effective dose for paralysis. This is of considerable clinical relevance because it suggests that even partial spinal cord irradiation can have deleterious effects. However, in general, preclinical spinal cord dose-volume studies indicate that dose distribution may be more critical than the volume irradiated, suggesting that neither dose-volume histogram analysis nor absolute volume constraints will be fully effective in predicting complications.
The relationship between the number of FSUs irradiated (n) and the probability of a complication (P) can be quantified by:
P = 1 – (1 – p)n,
where p is the probability of the loss of one FSU. This relationship is illustrated in Figure 2.2. When the average number of surviving cells per FSU is reduced to almost one, increasing the volume (number of FSUs exposed) reduces the dose necessary to produce a complication and increases the steepness of the dose-response curves. This may be less true for the small bowel than for spinal cord.
There is no evidence that cellular radiosensitivity is affected by an increase in treatment volume. The radiosensitivity of skin epithelium is constant over a 5,000-fold range of treatment area. In addition, no evidence exists for an increased role for vascular damage as volume increases. On the other hand, nonradiobiologic “volume effects” can be seen when:
1. A small area of injury (such as ulceration) is tolerated better than a large area of the same severity, even though the severity of the radiation response is independent of volume treated. In this case, pain, exudation, and infection may be worse; healing may be slower; and consequential contraction and scarring may be more of a problem.
2. As volume increases, so does dose heterogeneity across the field. A tumor dose prescribed at the 80% level may lead to a 25% higher dose at Dmax. If the threshold-sigmoid curve of the probability of normal tissue complications against dose is as steep as in experimental studies,20 a 25% increase in total dose could produce a marked change in the incidence of complications. Any increase is further compounded by the biologic effectiveness of each dose per fraction or the “double trouble” of increased physical and biologic dose that will depend on the size of the dose per fraction and the type of normal tissue, but it will be greatest in late-responding normal tissues for reasons that will be explained later. Because this additional augmentation of biologic doses is not evident from physical isodose contours, an increased biologic effect may be erroneously attributed to the large volume being treated per se. In addition, with large fields, large variations in contour may exist that could cause a high dose region where tissue thickness is less than that measured at the mid-plane; as, for example, in the spinal cord at the thoracic inlet in thoracic irradiation and in tangential fields for treatment of the breast.
3. If organ “reserve” is obliterated as volume is increased (e.g., lung, salivary gland). This is not a true volume effect because sequelae are determined by the volume and functional status of the tissue excludedfrom the treatment volume, not the volume irradiated.
FIGURE 2.3. Representation of the approximate kinetics of regeneration of irradiated normal tissues (solid lines, solid symbols) and tumors (dashed lines, open symbols). Curves are based on measurements or estimates of regeneration; symbols denote times at which an effect of regeneration has already appeared (left-pointing arrow) or has not yet appeared (right-pointing arrow). The logarithmic abscissa is for convenience of presentation only and has no biologic rationale. In general, the human data are displaced to the right of experimental animal data, reflecting a slower initiation of repopulation. Because human tissues proliferate more slowly than do their rodent counterparts, they were exposed to protracted dose regimens, and less-sensitive end points were used to detect onset of repopulation in humans. Numbers on the curve and symbols refer to different sources of data. These come from many sources. Further details are in the previous editions of this book.
Regeneration (Repopulation)
The time to onset of repopulation after RT and the rate at which it proceeds varies in different normal tissues. Both can be measured experimentally by a split-dose technique in which two doses are given separated in time. The size of the second dose required to produce a certain constant level of effect (isoeffect) increases with time after the first dose because of regeneration/repopulation.
In acute-responding tissues, repopulation starts early because cell loss is rapid. In the irradiated jejunal mucosa, the lag time may be <24 hours. In the colon and stomach, it is slightly longer. In mouse renal tubules, there is no histologic evidence of cell depletion for many months after irradiation; there is a long lag period, and it takes more than 12 months to reconstitute a tubule.43 The rate of repopulation has not been well quantified in tissues. In mice, some approximate doubling times for clonogenic cells are 8, 12, and 22 hours for jejunum,44 colon, and skin,45 respectively.
In humans, tissue turnover kinetics are slower than in mice. They have been approximated for oropharyngeal mucosa from consideration of responses to various dose fractionation regimens. Mucositis begins to appear 14 to 21 days after the start of a regimen of 2 Gy given five times per week, but repopulation begins at about 10 to 12 days.46 High initial doses may shorten the lag period, although only by 1 or 2 days. Repopulation can increase the tolerance of the mucosa to a conventional dose regimen by an average of at least 1 Gy per day, which is equivalent to approximately a doubling of clonogenic cell numbers every 2 days, and it may be significantly faster.47 If daily irradiation is suspended (e.g., during a 10- to 14-day break in a split-course accelerated regimen), clonogenic cells may repopulate at two or three times this rate.46–48 Figure 2.3 shows values for lengths of lag time and repopulation rates; however, they are, at best, estimates. The critical point is that there is a lag period followed by a phase of rapid exponential growth. In general, the lag period is shorter for chemotherapy, hyperthermia, and surgery because the cell depletion that stimulates regeneration occurs more rapidly than after irradiation.
The importance of repopulation is implicit in the history of RT. The current standard protracted overall treatment times confer a benefit by allowing regeneration of acute-responding tissues, which reduces toxicity. When attempts are made to deliver curative therapy more quickly, acute responses become more severe and dose limiting.
Growth factors may shorten the apparent lag phase and accelerate recovery in irradiated tissues. Hematopoietic growth factors such as G-CSF, GM-CSF, erythropoietin, and IL-11 can accelerate proliferation of hematopoietic cells.49 In doing so, they minimize the danger of infection. In epithelial tissues, keratinocyte growth factor (KGF), which is specific for epithelial cells, has similar potential. It protects the oral mucosa, small intestine, lung, and hair follicles against chemo- or radiation injury50–52 in preclinical models and has shown efficacy in clinical bone marrow transplantation trials. There is ongoing interest in the discovery of mitigators of radiation damage that can be given at least 24 hours after exposure of individuals in nuclear accidents or radiologic terrorist attacks.
”Remembered” Dose: Tolerance to Retreatment
Conventional wisdom in radiation oncology has been that a heavily irradiated tissue will not tolerate retreatment. The postulated reason was that the basis of late effects was irreversible vascular damage. Although irradiation may limit the tolerance of a tissue to retreatment, retreatment is often possible and may be better tolerated than previously expected.42 Factors that determine the extent to which residual injury will limit retreatment tolerance include the amount of cell depletion caused by prior treatment, the time elapsed since that treatment and therefore the extent of regeneration, and the tissue at risk. High prior doses, short intervals between treatment courses, and slow regeneration of target cells will reduce retreatment tolerance.
Some data for experimental radiation myelitis are shown in Figure 2.4. The plot shows the effect of size of the first dose on the dose required to produce myelitis in a second regimen. Recovery is complete after low doses but is progressively compromised as the initial dose approaches tissue tolerance.54 It should be remembered that clinical “tolerance” doses for the spinal cord of 45 to 50 Gy in 1.8- to 2-Gy fractions are low in terms of the injury evaluated in Figure 2.4 (50% incidence of myelitis). The time to recovery for the spinal cord is not accurately known; however, in rats, at 100 days it is about half of what it reaches by 200 days.55 In monkeys, there was extensive recovery from 44 Gy in 2.2-Gy fractions by 2 years, but a detailed profile of the time course could not be established.56
Not all tissues, or elements within tissues, recover at an equal rate or to an equal extent after irradiation. Acute-responding epithelial and hemopoietic tissues generally recover quickly and demonstrate a high tolerance to retreatment. However, the fibrovascular support in skin and mucosa and the stroma in bone marrow are less tolerant to retreatment because they respond more slowly. The kidney shows poor retreatment tolerance as assessed functionally in mice.57 Reirradiation tolerance in this organ is inversely related to the initial dose, although tolerance decreases significantly with increasing interval between treatments, suggesting progression rather than recovery from the initial damage.
Because different tissues show different levels of tolerance to retreatment, caution should be exercised in the application of these concepts to the clinic. In addition, the experimental studies deal with well-defined end points within a limited time scale. If different end points in the same tissue are examined or the time is extended, the same guidelines may not apply. It should also be noted that if slowly proliferating cells involved in late responses are extensively depleted, recovery may be permanently incomplete and the organ will be vulnerable to further injury, whether from radiation, trauma, cytotoxic drugs, or any other insult. For example, hyperthermia can precipitate myelitis in a patient who has had high but otherwise tolerable doses of x-irradiation, and trauma from dental intervention frequently precipitates mandibular necrosis.
Reproducible differences in tolerance of acute- and late-responding normal tissues and between different types of tumors to the same physical dose of IR are important because they define the basis of the biologic advantage of dose fractionation in conventional RT. This radiobiologic rationale has been encapsulated in the 4 “Rs” (repair, repopulation, redistribution, and reoxygenation).58
FIGURE 2.4. The dependence of remembered dose on size of priming dose (as a percentage of the ED50) is shown for a variety of animal species at long periods (6 months to 2 years) after the initial radiation treatment: O, adult rhesus monkey; 
, 12-week-old rat53; 
, 1-day-old guinea pig; •, young adult mouse; filled 
, 8-week-old guinea pig; filled 
, 8-week-old guinea pig; filled 
, 3-week-old weanling rat; filled diamond, young adult rat. (From Mason KA, Withers HR, Chiang CS. Late effects of radiation on the lumbar spinal cord of guinea pigs: re-treatment tolerance. Int J Radiat Oncol Biol Phys1993;26:643.)
TUMOR RADIOBIOLOGY
Kinetics
Tumor Cell Death and Survival
As is obvious from clinical practice, the doses of irradiation required for a certain control rate vary widely among human tumors. Tumor types that are traditionally radiocurable tend to show greater cellular radiosensitivity in vitro.59However, within one tumor type, there is a wide spectrum of radiosensitivities.60 Considerable effort has therefore been expended to develop molecular or cellular assays for molecules involved in cell death/survival, proliferation/arrest, and DNA repair so as to predict clinical responses to RT, as well as to identify potential targets for tumor radiosensitization.
Part of the heterogeneity between tumor radiation responses is because differences in mutations in oncogenes and tumor suppressor genes that affect pathways integral to cell death and proliferation alter intrinsic radiosensitivity. However, in addition, many cancers show an epithelial hierarchy similar to what is found in normal tissues. For example, in breast, undifferentiated estrogen receptor–negative mammary stem cells (MaSCs) maintain themselves through self-renewal and also differentiate into committed progenitors that ultimately give rise to mature ductal and alveolar cells, which belong to the luminal epithelial cell lineage that line the lumen of the mammary gland, and the mature myoepithelial cells, which surround the luminal epithelium.
Breast cancer is a heterogeneous disease with distinct molecular entities or “intrinsic” subtypes that have recently been linked to these developmental pathways by genomically defined tumor profiles and expression patterns of luminal, mesenchymal or claudin-low, and basal-like cells, with distinct blocks imposed by BRCA1 loss and HER2 amplification.61 These findings generally support the cancer stem cell (CSC) hypothesis that a cancer can arise from transformation of a normal stem or progenitor cell and give rise to a heterogeneous population of tumor cells. Alternatively, a differentiated cancer cell within a heterogeneous tumor may acquire stem cell–like features through self-renewal mechanisms prompted by oncogene expression, RT, chemotherapy, or hypoxia. Either way, the conclusion is that the bulk of hierarchically organized tumors is composed of more differentiated cells with limited proliferative potential, whereas the CSC compartment maintains the tumor and contributes to treatment resistance owing to its unique biologic properties. The importance of this concept is that tumor cure and regrowth is dependent on what may be a small minority of relatively quiescent cells.
No single marker is currently adequate to define CSCs of any histology; however, combinations of markers can be used, and in several tumor types, CSCs have been shown to be very radioresistant and chemoresistant.62 They are more tumorigenic and responsible for metastatic growths, and in general, the more primitive CSCs are more aggressive and more radioresistant. In vitro clonogenic assays are generally not good surrogates for CSC identification.
In the future, prediction of tumor behavior and response to therapy and the identification of biologic targets will require CSCs to be defined in molecular terms linked to normal developmental pathways and cancer-related mutations. A similar example of the importance of molecular profiling is the demonstration that patients with human papillomavirus (HPV)-positive oropharyngeal squamous cell carcinoma (OPSCC) have improved outcomes over those with HPV-negative tumors when treated with concurrent chemoradiation.63 Obviously, if CSCs are critical for cancer cure, then they have to be a prime target of any treatment. Therapies that fail to target them will be inadequate to affect cure even if they cause dramatic tumor shrinkage.
The mechanism of cell death may be important for both tumor cure and failure through stem cell reprogramming. Tumors of those histologic types that are traditionally radiocurable have a tendency to apoptose.8 Lymphocytic tumors generally apoptose more than carcinomas, whereas melanomas, sarcomas, and astrocytomas are relatively resistant.64 Importantly, apoptotic cells reappear between fractionated exposures.65 In addition, genetic modification of cells to introduce a proapoptotic phenotype frequently, although not always, radiosensitize.66 Despite these findings, studies specifically designed to find relationships between molecular markers of apoptosis and radiocurability of human tumors have yielded mixed and sometimes contradictory results.67 This may be because the CSCs do not express this phenotype. In contrast, cells that senesce or undergo autophagy or go through several divisions after RT may have a greater chance of surviving and undergoing CSC reprogramming.
In Vivo Kinetics of Tumor Responses
Most tumors regress during a course of RT and are considered analogous to acute-responding normal tissues in their radiation dose-fractionation responses. However, it has been known for decades that some tumors, such as melanoma, soft tissue sarcoma, and liposarcoma, behave similar to late-responding tissues and have a low α/β ratio (see later).68 More recently, prostate69 and breast70 cancers have been added to the list. Such tumors have a relatively slow turnover rate.
Regression of a tumor after RT reflects the rate of cell loss, which is determined by turnover kinetics, in the same way as for normal tissues. Thus, tumors with a small cell loss factor can respond slowly after RT, even though all CSCs have been sterilized, as evidenced by their failure to recur.
The kinetics of tumor responses to RT will depend on whether the cell of origin is an early or a more differentiated (progenitor) CSC as well as the effects of the cancer-related mutations.
Cell Loss Factors in Tumors
A cell loss factor (φ) of <1 is characteristic of tissue growth. In tumors—for example, head and neck carcinomas—φ is actually close to 1,71 which is why their growth rate is much slower (on average, doubling times of about 60 days)72 than could be predicted by their proliferative activity. Mitotic count, S-phase count, or labeling index (LI) would suggest a potential doubling time (Tpot) of about 3 to 7 days. It follows that a high proliferative index is not necessarily evidence of a tumor that will grow rapidly in size. A classic example is the slow-growing basal cell skin carcinoma, in which numerous mitotic figures are commonly visible. They have a high φ owing to extensive apoptosis.73,74
Tumor Regression After Irradiation
Tumors with a high rate of cell loss will regress rapidly during and after RT, regardless of pretreatment growth rate; providing the overall treatment duration is not unduly protracted, the prognosis is generally good.71,75 On the other hand, rapid regression would also be expected in a tumor with a low cell loss rate if a large proportion of its cells are actively cycling and it is growing quickly; however, in this case, the prognosis is poor. Rapid regression is therefore not a universal prognostic indicator,75 although it is usually a favorable prognostic sign.
Similar arguments can be applied to tumors that regress slowly, such as prostate carcinoma, some cases of nodular sclerosing Hodgkin’s disease, teratocarcinomas of testis, some soft tissue sarcomas, choroidal melanomas, meningiomas, pituitary adenomas, chordomas, or glomus tumors. Slow regression may reflect slow proliferation, low cell loss factor, residual stroma, or, sometimes, treatment failure.
Repopulation occurs as a homeostatic response to cell depletion caused by treatment. The rate of cell loss slows (the cell loss factor decreases), as in acute-responding normal tissues, during rapid repopulation. Tumors with a high rate of cell production and a large cell loss factor are likely to regress quickly but recur early and regrow rapidly after unsuccessful irradiation, chemotherapy, or surgery.76
A practical implication of the complex reasons for different rates of tumor response to RT is that it is not a good idea to reduce the total dose just because a tumor regresses rapidly.77 In addition, local control of tumors that grow slowly because of high cell loss factors may initiate an early repopulation response, and local control may be prejudiced by protraction of treatment time beyond normal, just as it is for fast-growing tumors that are initially fast growing.78,79 Well-differentiated tumors (which have a high cell loss factor) are more prone to an early reduction in cell loss factor and, consequently, an early repopulation response that prejudices local control, especially if overall duration of treatment is protracted.80
Tumor Regeneration After Irradiation
Potential Doubling Time
The potential regeneration rate of tumors after cytotoxic injury is better predicted by pre-RT proliferative activity than by pre-RT tumor growth rate. The Tpot,81–83 which is the time that would be required for the number of clonogenic cells to double if the cell loss factor were zero, is a measure of this and can be estimated from the average duration of the S phase (Ts) and the fraction of cells in S phase, measured by LI:
where l is a correction factor for the cell cycle distribution of the population.
Tpot is a logical predictor of the kinetics of a regenerative response, and attempts were made to use it to predict which tumors would benefit from acceleration of treatment that aims to minimize such repopulation.81,83,84 Early results indicated that it might be of some value, although later analyses indicate otherwise.85 This may be because the cell cycle time measured in an unperturbed tumor before treatment is different from that during or after treatment. More likely, without knowledge of the behavior of the CSC compartment, gross observations on whole tumor populations are likely to mislead.
FIGURE 2.5. Growth curves for a rat rhabdomyosarcoma and its constituent clonogenic cells after a dose that reduced survival to 1%. The upper curve (1) shows unperturbed growth of tumors; the middle curve (2) shows regression and regrowth of tumors irradiated on day 0 with a dose that reduced cell survival to 1%; the lower curve (B) traces the repopulation of the tumor by surviving clonogens. Exponential regrowth of the surviving clonogenic cells occurs while the gross tumor is regressing. (From Hermens AF, Barendsen GW. Changes of cell proliferation characteristics in a rat rhabdomyosarcoma before and after x-irradiation. Eur J Cancer 1969;5:173. © 1969 Pergamon Press, Ltd.)
Growth Fraction
The growth fraction86 is simply the fraction of tumor cells that are cycling. In solid tumors, this is usually a small proportion of the total (e.g., 20%). These are not the CSCs that are generally quiescent but more likely are progenitor cells. The growth fraction may decrease as tumors enlarge and grow more slowly (see control curve, Fig. 2.5). This changing growth rate can be approximated by a Gompertz equation.71 In contrast, after cytoreductive therapy, the growth fraction probably increases, CSCs can be induced to proliferate, and these contribute to accelerated tumor regrowth. Treatment may therefore accelerate tumor growth. An analogous response is found in some normal tissues (liver, dermis), which have only a small fraction of cells in cycle but can regenerate rapidly through recruiting resting, G0-phase cells into cycle.
Regeneration in Experimental Tumors
Hermens and Barendsen87 showed a rapid exponential increase of surviving clonogenic tumor cells in a rat rhabdomyosarcoma several days after irradiation. Because only 1% of the initial clonogens survived, tumors did not enlarge; rather, the tumor mass was still regressing when repopulation began (Fig. 2.5). This was later confirmed in other experiential tumors,88,89 and the same probably occurs during regression of clinical tumors. It would be important to revisit these experiments in light of the CSC hypothesis. In vitro fractionated daily irradiation can select for CSCs2 and also promote reprogramming from more differentiated tumor cells.
Regeneration in Human Tumors
Clonogen regeneration in human tumors can be assessed by an increase in dose required for tumor control as treatment duration is increased.90 Alternatively, if a constant dose has been used, the decrease in tumor control rate as treatment time is extended. These techniques have been used to derive evidence for accelerated regrowth during a standard RT regimen for head and neck cancer, although it is likely to occur in all tumor sites. The magnitude and timing of regeneration will vary from tumor to tumor of the same type and among different types.
The concept of accelerated repopulation by human tumor clonogens during and after a course of fractionated RT is supported by several observations and is of particular concern when delivery time is prolonged.
1. Time to Tumor Recurrence
If 104 tumor cells survived RT, they would have to undergo 15 doublings to present as a recurrence. Because most local recurrences of head and neck cancer are detectable within 12 months after RT, the average tumor volume doubling time would have to be about 2 weeks. Because the median volume doubling time for tumors at presentation is about 2 months,41 the growth rate of residual clonogens must accelerate following unsuccessful treatment. Similar rapid regrowth was seen in pulmonary metastases after subcurative RT.91
2. Split-Course Treatment
Split-course regimens for head and neck squamous cell carcinomas (SCCs) give lower local control rates than continuous regimens of the same total dose,92 suggesting that tumor regrowth occurred rapidly during the time extension. This does not happen for prostate cancer.92,93
3. Protracted Treatment
Protraction of treatment time decreased the rate of locoregional control for head and neck cancer in several retrospective analyses,79,94–97 which is consistent with accelerated tumor regeneration. The analyses were of three types:
a. Scattergram analysis. Protracting treatment time for OPSCC led to worse outcome95 (Fig. 2.6). For treatment durations of 30 to 55 days, each day’s extension required the total dose to be increased by about 0.6 Gy to achieve a constant rate of tumor control. Assuming that between 1.8 Gy and 2.4 Gy reduces cell survival by 50%, an increase of 0.6 Gy per day is consistent with clonogens doubling every 3 to 4 days. Because tumors at presentation have a doubling time of about 2 months,41 there must be a dramatic change in growth rate during RT. The same pattern (Fig. 2.6) was seen in 11 other subsets of patients with oropharyngeal cancers95 and carcinomas of the supraglottic larynx,98 as well as in reanalyses of earlier data,79 and for SCC of tonsil collected from nine centers in the United States, Canada, and England.97 The multicenter study of SCC of tonsil is important because differences in overall treatment duration predominantly reflect institutional policy; selection of longer treatments for worse tumors cannot explain the increase of tumor control dose 50 (TCD50) with extension of overall treatment time.
b. TCD50 analysis. Figure 2.7 presents TCD50 values for SCC of head and neck calculated from the literature.79 They are independent of treatment duration up to about 28 days, after which they increase rapidly (consistent with 0.6 Gy per day). The suggestion is that on average, head and neck SCCs exhibit a lag period of 3 to 4 weeks before beginning to repopulate, with an average doubling time of 3 to 4 days.
c. Analysis of primary tumor control rate. When a standard prescription (e.g., 50 Gy in 20 fractions in 4 weeks) is given but overall duration of therapy is extended for whatever reason, the control rate decreases, commonly by 1% to 2% per day for head and neck and cervix cancer (Fig. 2.6b).99,100
It should be noted that a lag period of up to 4 weeks and thereafter a 0.6 Gy per day increase in the “isocontrol” dose are not evidence that RT for head and neck cancer is best given in 4 weeks. Repopulation of mucosa begins at about 10 to 12 days and is more rapid than tumor, requiring thereafter an average daily dose increment of at least 1 Gy for a mucosal isoresponse.79 Thus, a therapeutic gain in mucosal tolerance relative to tumor control is still achieved by extending treatment beyond 4 weeks. It is only late-responding tissues, which do not benefit from repopulation, that lose out. The overall therapeutic differential will be greatest if the tolerance dose for the critical late-responding tissue is delivered in the shortest overall time consistent with an acceptable acute response,75,79 and without compromising the total dose delivered to the tumor.
4. Accelerated Treatment
If accelerated tumor growth contributes to treatment failure, it may be neutralized by acceleration of RT. In nonrandomized studies, shortening the overall duration of treatment improved the local control in inflammatory breast cancer,101 melanoma metastases to brain,102 and head and neck cancer.47,103,104 Randomized studies of accelerated treatment of head and neck cancer validated the benefit. Exceptions may be cancer of the prostate, which is slow growing,92,93 although hypofractionation with stereotactic body radiation therapy (SBRT) for prostate cancer has shown promise.105
Dose-intensity studies106,107,108 suggest that chemotherapy also accelerates tumor regrowth. Furthermore, the lack of benefit from neoadjuvant chemotherapy given for two or three cycles before the start of RT for head and neck cancer, despite shrinkage of the gross tumor mass, is consistent with accelerated regrowth of subclinical residual CSCs.
FIGURE 2.6. A: Scattergram with TCD50 and TCD90 (tumor control dose) curves for 3-year local control of SCC of the tonsil (1, local control; 2, recurrence; 3, persistence of detectable disease). For a given total dose, local control decreased with protraction of overall treatment time. For a given overall time, local control improved with increase in dose. The total doses were normalized to be equivalent to the total dose in 2.5-Gy fractions using the LQ isoeffect curve (Fig. 2.22). An α/β value of 2.5 Gy was used, being the best estimate from these and other data. (From Maciejewski B, Withers HR, Taylor JMG. Dose fractionation and regeneration in radiotherapy for cancer of the oral cavity and oropharynx. I. Tumor dose-response and repopulation. Int J Radiat Oncol Biol Phys 1989;16:831.) B: Pelvic control as a function of treatment time for 621 patients treated with a total dose of 85 Gy. (From Keane TJ, Fyles A, O’Sullivan B, et al. The effect of treatment duration on local control of squamous carcinoma of the tonsil and carcinoma of the cervix. Semin Radiat Oncol 1992;2:27.)
FIGURE 2.7. Estimated TCD50 (tumor control dose) values as a function of treatment duration from published results of radiation therapy for squamous carcinomas of the head and neck excluding nasopharynx and true vocal cord. TCD50 values are expressed as LQED2Gy (the equivalent dose given in 2-Gy fractions calculated using the LQ model). Total doses for an isoeffect increase steeply with protraction of treatment duration, implying accelerated repopulation by surviving tumor clonogens, consistent with the 3- to 4-day average doubling time calculated from scattergrams (Fig. 2.6A). The relatively constant TCD50 value for treatments lasting up to 4 weeks is consistent with an average time of onset of accelerated growth at about 4 weeks. Growth of clonogens at the average preirradiation doubling rate of 2 months would have little detectable effect on TCD50 values (about 2.5-Gy increase in 8 weeks). (From Withers HR, Taylor JMG, Maciejewski B. The hazard of accelerated tumor clonogen repopulation during radiotherapy. Acta Oncol 1988;27:131.)
Cell Cycle Redistribution After Radiation Therapy
Cells change in their radiosensitivity as they traverse the division cycle109 (Fig. 2.8). The difference in radiosensitivity between late S-phase and G2-M cells is greater than that between euoxic and hypoxic cells. After exposure of an asynchronous population of cells to 2 Gy, the survivors will be partially synchronized in relatively radioresistant cell cycle phases (because of the preferential killing of cells in sensitive phases). When these survivors resume their progression through the division cycle, they move into more sensitive phases. If they were to do so in a synchronized fashion, this could be exploited.110 Unfortunately, they do not. However, a greater proportion of the surviving population will be in sensitive phases of the division cycle than immediately after irradiation, which will produce a net “self-sensitization” effect. Dose fractionation will enhance the therapeutic ratio by permitting redistribution among tumor cells but not nonproliferating cells in late-responding normal tissues.111 The differential is greater the smaller the dose per fraction, and it is amplified as an exponential function of the number of fractions delivered. This amplification of small differentials between cycling tumor cells and the nonredistributing target cells in late-responding normal tissues was the initial rationale for clinical trials of hyperfractionation,111 before important intrinsic differences in response to low-dose fractionation between late-responding normal tissues and tumors were appreciated.112
The effect of cell cycle redistribution on tumor responses to multifraction irradiation is difficult to demonstrate.113 This may be because of intratumoral heterogeneity and timing. CSCs are believed to exist in the G0-phase of the cell cycle and to cycle slowly because of their intrinsic metabolic state influences from the niche in which they reside. Niches for most CSCs have yet to be convincingly demonstrated; however, glioma CSCs appear to reside in a perivascular location,114–116 although there is also evidence for hypoxic niches.117 Most CSCs are negative for the proliferation marker Ki67,118 but multiple fractions of IR may promote recruitment of CSCs from the niche and increase the proportion of cycling cells,118 with a possible concomitant increase in radiosensitivity. Therefore, for CSCs, redistribution following irradiation may be tied to their mobilization into the cell cycle and thus regeneration. This raises the interesting possibility that this may be strategy for increasing their radiosensitivity.
FIGURE 2.8. Radiation dose survival curves for one line of mammalian cells (V-79 Chinese hamster) synchronized in four positions in the division cycle. Significant differences occur in the survival of cells at different ages, with the differences relative to absolute survival being greatest at lower doses. The survival ratio between late S and G2M cells after 2 Gy is approximately 5. (From Withers HR, Peters LJ. Biological aspects of radiation therapy. In: Fletcher GH, ed. Textbook of radiotherapy, 3rd ed. Philadelphia: Lea and Febiger, 1980.)
FIGURE 2.9. Curve relating cellular radiation sensitivity to partial pressure of oxygen at the time of irradiation. Data were obtained by scoring anaphase aberrations in Ehrlich ascites tumor cells, although similar curves have been obtained for killing of bacteria. About 50% of the total sensitization by oxygen is seen at a partial pressure of approximately 4 mm Hg at 37°C. The inset shows survival curves for different levels of oxygenation.
The Oxygen Effect
In 1909, Schwarz reported that restricting the blood flow to a tissue decreased its radiation response.119 It was thought that this was a metabolic effect; however, in 1951, Read120 showed that oxygen sensitized cells through a radiochemical mechanism. Now, oxygen is recognized as a potent chemical modifier of radiosensitivity.121,122 The relationship between oxygen tension and radiosensitivity varies, but a radiosensitivity halfway between that of hypoxic and euoxic (aerobic) cells (the k value5) is achieved with oxygen concentrations ranging from about 3 to 10 mm Hg1,121,122 (Fig. 2.9). The curve relating metabolic activity to oxygen concentration is steeper and to the left of this. Therefore, hypoxic radioresistant cells can be metabolically normal and viable. The inset in Figure 2.9 shows how survival curves are modified by oxygen concentration. The ratio of doses required to produce the same level of effect (e.g., cell survival) in hypoxic, as in euoxia, conditions is called the oxygen enhancement ratio (OER). The OER is between 2.5 and 3 for cells exposed to high doses of x- or γ-rays and does not change much as cells progress through the cell cycle. At low doses or low dose rates,123 the OER is slightly lower. The reason may be that electrons scattered by photons may be regarded operationally as a mixture of low and high LET particles. At the ends of electron tracks, there is a densely ionizing component that will be relatively more important at low doses or low dose rates, and this may result in a lower OER. The OER for neutrons used in clinical trials is approximately 1.6 and 1 or close to it for α-particles or beams of stripped nuclei.
The property of oxygen critical to radiosensitization is its electron affinity. After this was realized, radiochemists developed oxygen-mimetic, electron-affinic radiosensitizers, such as metronidazole, misonidazole, etanidazole, nimorazole, and other nitroimidazoles. These radiosensitizers are easier to administer than oxygen, are less rapidly metabolized, and can therefore diffuse further from blood vessels into the hypoxic regions of the tumor. Toxicity has somewhat limited their use, but there is evidence of clinical efficacy. Recently, drugs have been developed, such as tirapazamine, that are selectively toxic to hypoxic cells and may both radiosensitize and chemosensitize them.124
Relevance of Hypoxia to Clinical Radiation Therapy
Tissue oxygenation is critically dependent on capillary blood flow. If blood is stagnant, as it can be in tumor capillaries, then arterial and venous oxygen tension are of secondary relevance.125,126,127 It is therefore not surprising that solid tumors contain hypoxic foci.90,126,128,129 These may be due to the tumor outgrowing the blood supply, necrosis, sluggish blood flow, shunts, or temporary occlusion.127 Temporary occlusion and blood shunting cause transient hypoxia that may be as, or more, important to the overall response as chronic hypoxia caused by limited diffusion.
Hypoxia has been demonstrated by a variety of methods. When single high doses of radiation of varying magnitude are given to experimental tumors and clonogenic survival assessed in vitro, the dose survival curve has two components: an initial, relatively rapid decline as euoxic cells are killed, and a second slower decline resulting from the less efficient killing of radioresistant hypoxic cells.129 Polarographic oxygen electrodes have been used to measure oxygen tension (pO2) within human and experimental tumors,126 as have DNA strand break (Comet) assays,130,131 detection of binding of the 2-nitroimidazole by immunohistochemistry (EF5)132 or PET imaging, and immunohistochemistry for localized expression of hypoxia–inducible factor-1α (HIF-1α) or its downstream effectors. HIF-1 expression is stabilized under hypoxia, and it acts as a transcription factor to up-regulate several genes that promote cell and tissue survival. These include glycolysis enzymes and vascular endothelial growth factor (VEGF), which promotes angiogenesis. It is a metabolic switch mechanism that may also be produced in the presence of oxygen under inflammatory conditions under the direction of the transcription factor NF-κB (nuclear factor κB). HIF-1 may therefore be best regarded as a factor whose expression is dysregulated in cancer and has wider implications than promoting radioresistance.
Hypoxia is being increasingly linked with clinical outcome. Hyperbaric oxygen133,134 or correction of anemia133,135 has been reported to improve outcome, although the use of erythropoietin has been reported to have the opposite effect,136 most likely because it supports CSC proliferation and survival.137 High hypoxic fraction126 portends poor local control and survival rates in head and neck and cervix cancer treated with RT.95,138 However, outcome of patients with uterine cervix cancers treated with surgery only also correlated with hypoxia.139 In addition, hypoxia provides a growth advantage for cells with mutated p53140 and can select for, and be a marker for, more aggressive tumors.
Despite the probable existence of hypoxic cells within many, if not all, solid tumors of humans, the importance of hypoxia as a predictor of individual response to RT has yet to be established. CSCs have been shown to have a high level of antioxidants and respond less to IR with ROS production.2 This may make them resistant to oxygen-enhanced radiosensitization.
RT may also actually induce hypoxia, as may other treatments. Fuks and Kolesnick24 noted “rapid endothelial cell death in tumor displays an apparent threshold at 8–10 Gy and a maximal response at 20–25 Gy.” This apoptosis was observed within hours, and loss of microvasculature after RT has been reported after a few weeks.141 The failure of normal tissues irradiated with <20 Gy to support angiogenesis has been known for decades as the tumor bed effect.142 In light of this, tumor regrowth during or after RT may depend on vasculogenesis more than angiogenesis.143,144 Vasculogenesis is a relatively inefficient process, and an increase in hypoxia is a possible outcome.
How common these radiation-induced alterations in the tumor microenvironment are in clinical reality and their relationship to tumor cure has yet to be fully established. Dose may play a major role, as suggested by Fuks and Kolesnick,24 although Tsai et al.144 showed that modest fractionated protocols caused vascular loss similar to that of high single doses. Again, proinflammatory cytokines such as TNF-αmay be particularly important after high doses because they target vasculature.145 In fact, it is not definite that chronic hypoxia leads to radioresistance, because it has been shown to decrease DNA DSB repair, in particular RAD51-mediated HR.146 The nature of the hypoxia may therefore be critical, and transient acute hypoxia due to intermittent vessel closure that may reoxygenate rapidly could be more clinically relevant than chronic hypoxia, which is owing to the limitation of oxygen diffusion.
An important issue is whether reoxygenation occurs during a course of fractionated RT. If hypoxia is simply a marker of tumor aggression, then reoxygenation may not be important for outcome; however, if hypoxia is radiobiologically relevant, the rate and extent of reoxygenation will be critical.
TABLE 2.1 RATIOS OF DOSE FOR ISOSURVIVAL EQUIVALENT TO THAT FROM 2 GY IN OXIC CONDITIONSA
Tumor Reoxygenation
If 30% of tumor cells were hypoxic and 70% euoxic, and no change were to occur in the distribution of oxygenation during a course of conventional RT, most surviving cells would be hypoxic after the first few fractions and 70 Gy would reduce survival to <10−4. This is providing that the CSCs are sensitized by oxygen in vivo in the same way that many other cells are in vitro and are equivalent to clonogenic cells. These are large assumptions and probably incorrect; however, reoxygenation may still be relevant to tumor control in the clinic. The possible magnitude of the effect of reoxygenation on the response to multiple dose fractions can be appreciated from Table 2.1. For example, the dose to a tumor that repeatedly returns to an 80-to-20 euoxic-to-hypoxic cell mixture would need to be 15% higher than it would if all tumor cells were oxic. Reoxygenation may require reduction in total tumor cells with less loss of blood vessels, decreased interstitial tumor pressure, and better oxygen diffusion as a result of less temporary occlusions.125 It should be noted that free radicals generated during reoxygenation may be particularly toxic to cells. This may be why CSCs can exist in a perivascular niche, being more resistant to redox changes.62
Most studies on the kinetics of reoxygenation in animal tumors have used relatively large dose fractions and indicate considerable variation between tumors. In most cases, reoxygenation occurs rapidly and is completed within 6 to 24 hours.147 In human tumors, it is still a matter of controversy.148 The variation between tumors and tumor sites seems considerable.
It seems unlikely that the beneficial effects of reoxygenation are compromised much by changes in dose fractionation. In pure hyperfractionation, where the dose per fraction is small, responses will be little affected by 10% to 20% of the cells being hypoxic.149 The overall duration of the course of RT is not shortened; therefore, the time available for reoxygenation will be the same. In accelerated regimens, the theoretical problem of the proportion of surviving hypoxic cells increasing with time is of more concern. However, the rapid reoxygenation kinetics observed in experimental tumors, as well as high control rates achieved experimentally and clinically with brachytherapy and in centers using 3- or 4-week overall treatment durations, suggest that reoxygenation is adequate in most tumors, even with short courses of fractionated or low dose rate irradiation, or it is not a problem.
Where large dose fractions are given as a single exposure, as in intraoperative RT, single-dose stereotactic radiosurgery, and high dose rate brachytherapy, outcome may be more compromised by hypoxia. However, results show that local control is often achieved, suggesting that advantageous physical factors outweigh any lack of reoxygenation. In chemoradiotherapy or bioradiotherapy protocols that target tumor vasculature or angiogenesis, the extent of reoxygenation might also be an issue.
FIGURE 2.10. Random distribution in 100 equal-sized “targets” of 100, 200, or 300 “hits.” The probability that any one of the 100 targets will not be struck when 100 hits are delivered randomly is e−1, or 37%. The same probability of survival applies for each equal increment in the number of hits: 200 hits would result in a probability of survival of e−2, or 0.37 × 0.37. Even after 300 hits are delivered, there is still a chance of e−3 = 5% that any one target will survive. This proportionate, or geometric, decrement in survival rate may be plotted as a straight line on semilogarithmic coordinates. (From Withers HR, Peters LJ. Biological aspects of radiation therapy. In: Fletcher GH, ed. Textbook of radiotherapy, 3rd ed. Philadelphia: Lea and Febiger, 1980.)
QUANTITATIVE RADIOBIOLOGY AND DOSE FRACTIONATION
Random Nature of Cell Killing
Our earliest understanding of dose-response relationships for irradiated cells came from studies with bacteria.150 Bacterial cell survival decreases geometrically with dose. In other words, the dose that reduces the survival rate to 50% will, when doubled, reduce it to 25%, and if tripled, to 12.5%, and so forth. When such a relationship is plotted semilogarithmically, a straight line results. Such a dose-survival relationship reflects a random cell kill process, which means that if 100 lethal lesions are distributed randomly throughout 100 equally radiation-sensitive cells (mean lethal dose = 1), by Poisson statistics, 37 cells will be spared; 37 will have one lethal lesion; 18 will have two; 6 will have three, and the like (Fig. 2.10). It is immaterial whether a cell is killed by one or more lethal lesions; however, the survival rate of 37% recurs for each additional mean lethal dose, which ensures the semilogarithmic relationship.
The mathematic bent of early radiation biologists, many of whom were also physicists, caused them to describe the slope of survival curves in terms of the mean lethal dose (D37 or D0), which reduces survival by one natural logarithm (e−1), rather than D10, which reduces it by one common logarithm and is an easier term for biologists to think in. It is useful to remember that D10 is about 2.3 times D0.
Mammalian Cell Survival Curves
Puck and Marcus4 published the first survival curve for mammalian cells in 1956. Logarithmic decreases in cell survival with dose were found that fitted the model described for bacteria, although with two important differences. D0values for cells are generally between 0.75 and 2 Gy, which is less than one-tenth of those for bacteria, largely reflecting the latter’s smaller target size (less DNA). In addition, unlike those for bacteria, mammalian cell survival curves most often have a shoulder before the logarithmic decline (Fig. 2.11). The biologic basis for the shoulder is not firmly established; however, it is consistent with mechanistic view of Catcheside et al.150 that radiation-induced cell kill has a linear, single-hit, α-type killing term plus a quadratic, multihit, β-type killing term relating it to dose. At low clinical doses of around 2 Gy, most killing is single hit. At higher doses, there is additional accumulation of ionization multihit lesions from other electron tracks (intertrack). This “sublethal” injury can be converted into additional lethal injury.
An additional complexity in dose-response curves is that hyperradiosensitivity to very low radiation doses (<10 cGy) has been shown in some systems; however, as dose increases above about 30 cGy, radioresistance increases until about 1 Gy, when cell survival begins to follow the usual downward-bending curve with increasing dose.151,152 This phenomenon is not universal but has been seen in many human cell lines in vitro and in experimental studies in mouse skin, kidney, and lung. The precise mechanisms are still unclear, although low-dose hypersensitivity may be due to a failure to activate G2-phase cell cycle checkpoints.17
Other phenomena also challenge any simple relationship between DNA lesions and radiation response. One is radiation-induced genomic instability where the rate of genomic alterations increases with increasing number of divisions of irradiated cells, as manifested by chromosomal rearrangements, formation of micronuclei, gene amplification, or cell killing.153 Another is the nontargeted radiation effect in which cells that were not irradiated but were “bystanders” at the time of irradiation are affected. In some but not all cases, culture medium from irradiated cells is active.154,155 Again, chromosome-related alterations, mutations, gene induction, and cell killing are observed end points. The relevance of low-dose hypersensitivity, adaptive responses, induced genetic instability, and bystander effects to clinical radiotherapy has yet to be fully evaluated; however, they challenge classical radiobiologic paradigms. They are probably more relevant to radiation-induced carcinogenesis, where they can explain the higher than expected frequency of postradiation chromosomal aberrations; in the future, they may provide novel opportunities for therapeutic intervention. In any event, they do not significantly impact the mathematical models that have been proposed for RT, which were derived to fit existing data within the clinically relevant dose range.
FIGURE 2.11. A two-component survival curve for mammalian cells is characterized by a “shoulder” followed by a terminal exponential region, the slope of which is defined by a D0 value (slope = 1/D0). The position of the curve on the radiation dose axis can be fixed by the intercepts of the terminal exponential region extrapolated back to the zero dose axis (n) or to the 100% survival level (Dq): n is termed the extrapolation number, and Dq the quasi-threshold dose. Although n and Dq are parameters that define the width of the shoulder on the survival curve, they do not indicate its shape, which is of prime importance in RT. The survival curve shoulder can be considered to consist of an initial exponential region (the slope of which is defined by 1D0), followed by a downward-bending segment that merges asymptotically into the final exponential region of the survival curve. (From Withers HR, Peters LJ. Biological aspects of radiation therapy. In: Fletcher GH, ed. Textbook of radiotherapy, 3rd ed. Philadelphia: Lea and Febiger, 1980.)
FIGURE 2.12. Model dose survival curves for mammalian cells showing that the experimentally determined curve for acute exposures (lowest curve) is the product of two mechanisms: single-hit injury described by an exponential curve (e−αd), and multihit, or cumulative, injury described by a continuously bending curve related by a coefficient, β, to the square of the dose. At doses of clinical relevance, cell death from the single-hit mechanism predominates. The rate at which the survival curve bends from an initial, essentially exponential region depends on the ratio (α/β) of the coefficients for single-hit and multihit killing: the lower the value, the sooner and more steeply the curve bends. The value of α/β is the dose at which single- and multihit mechanisms contribute equally to cell killing. In these curves, α/β = 10 Gy, which is a value characteristic of acutely responding tissues. Target cells in late-responding normal tissues are characterized by low α/β values; hence, their survival curves are curvier. The flexure dose, Df, is the dose at which deviation from the initial exponential part of the curve is difficult to detect and, for available biologic assay systems, is about one-tenth of α/β. When doses in a multifraction regimen are less than Df, further dose fractionation does not produce detectable “sparing” (because cell killing is essentially all the result of single-hit events, the lesions potentially contributing to multievent killing being completely repaired during the fractionation intervals). The lower the α/β value, the lower the dose at which multihit mechanisms cause cell death, the lower the value of Df, and the lower the dose per fraction below which a sparing effect of dose fractionation is lost. The curve for single-hit killing (e−αd) can be measured experimentally using very small dose fractions or a continuous low dose rate exposure; however, the curves for multihit killing (e−βd2) can be determined only indirectly from a knowledge of the other two curves.
Linear Quadratic Formula
Lea150 and Read120,156 quantified biologic responses to irradiation in terms of a linear dose coefficient (α) and a coefficient (β) for the square of the dose so that effect is proportional to αD + βD2. This can be used to fit a continuously bending curve to cell survival data:
S.F. (survival fraction) = e−(αD+βD2)
The linear component (αD) is of major significance if RT is delivered in fractions of about 2 Gy (Fig. 2.12). Its importance was largely ignored in the 1960s but was “rediscovered” in the early 1970s when Dutreix et al.157demonstrated that reducing doses per fraction below 3 Gy did not result in additional sparing of acute effects in human skin. Now it is accepted that low-dose brachytherapy or standard fractionated RT could not eradicate cancer without single-lethal-hit damage.158,159
The survival curve at low doses is essentially linear because α-type lethality predominates and there is little opportunity for accumulated (β-type) injury. Likewise, exposure to low dose rate continuous irradiation results predominantly in α-type lethality because of continuous repair. Under these circumstances, the effective survival curve is linear and defined by α.
The dose range over which the linear component dominates depends on the relative values of α and β. The α/β ratio defines the dose at which cell killing by linear and quadratic components are equal. The higher the α/β ratio, the more linear and steeper is the dose-response curve and the less sensitive it is to dose fractionation. If the α/β coefficient is low, the survival curve will be “curvier,” bending down only after an initial linear region; there will also be a marked sparing effect of dose fractionation (Figs. 2.13 and 2.14). The parameters can be estimated from multifraction data if the reciprocal of the total dose 1/nd is plotted against dose per fraction (d) (Fe plot).160 The intercept on the ordinate is α/logeS and the slope is β/logeS. The ratio of the intercept to slope gives the α/β ratio. Late-responding tissues generally have a low α/β ratio and show a large fractionation effect. Acute-responding tissues generally have a large α/β ratio, as do many tumors, although they show considerable variation (Table 2.2) with melanoma, soft tissue sarcoma, and liposarcoma tending to have low α/β ratios,68 as do prostate69 and breast70 cancers. The considerable variation observed within any one tumor type may be caused by failure to take into account their CSC origin.61
FIGURE 2.13. Hypothetical survival curves for the target cells for acute and late effects in normal tissues exposed to X-rays or neutrons. The α/β ratio is lower for late effects than for acute effects in x-irradiated tissues, resulting in a greater change in effect in late-responding tissues with change in dose. At dose A, survival of target cells is higher in late-effects than in acute-effects tissues; at dose B, the reverse is true. Increasing the dose per fraction from A to B results in a relatively greater increase in late than acute injury. For neutrons, the α/β ratio is high, with no detectable influence of the quadratic function (βd2) over the first two decades of reduction in cell survival, implying that accumulation of sublethal injury plays a negligible role in cell killing by doses of neutrons of clinical interest. (From Withers HR, Thames HD, Peters LJ. Biological bases for high RBE values for late effects of neutron irradiation. Int J Radiat Oncol Biol Phys 1982;8:2071.)
FIGURE 2.14. Multifraction-dose survival curves compared with a single-dose curve. Effective survival curves for multifraction regimens that produce an equal (proportionate) decrement in survival from each dose are linear, with shallower slopes than the single-dose curve at the same dose. Slopes of the multifraction curves become less steep with a decrease in fraction size until the dose per fraction is so low that multihit killing contributes negligibly and the slope is the limiting one determined by single-hit killing (and eD0 = 1/α). The dose per fraction below which the effective survival curve becomes no shallower is a function of the curviness of the single-dose survival curve and is lower than the α/β value.
TABLE 2.2 α/β VALUES
Two-Component Model
Dose survival curves can be well fitted by equations other than the linear quadratic (LQ) model. The most common is the two-component (TC) model. This combines a single-hit component e−D/1D0, where 1D0 is the dose necessary to reduce survival to 0.37(e−1) in the initial region of the curve, with a multiple-event cell killing model e−(1−e−D/nDo)n, where nDo is the dose needed to reduce survival to e−1 in the final region of the curve and n is the extrapolation number (Fig. 2.11).
S.F. = e–D/1D0 · (1 – [e–Dn/D0]n)
Comparison of the Linear Quadratic and Two-Component Survival Models
Over a limited dose range (e.g., 2 to 8 Gy), including the region that matters in most clinical radiotherapy, both the TC and LQ models “fit” data indistinguishably (Fig. 2.15). At high doses, the LQ model fits some cell survival curves better because they appear to continue to bend, whereas the TC model better fits those that seem more linear. However, the experimental data suggest that isoeffective curves are not very linear over a wide range of dose per fraction for many normal tissue responses and that extrapolation from 2 Gy to more than 7-Gy fractions or single doses as in SBRT is unlikely to give realistic equivalent doses,161 although there are other opinions.162 In addition, the models differ significantly at predicting responses to doses <2 Gy using data from doses >2 Gy. The differences may appear small but would be amplified if a dose of 1.15 Gy, for example, were repeated 70 times or more, as could happen in a hyperfractionated RT regimen (Fig. 2.16).
The LQ model has gained popularity112,160,163–165 because it is simpler and because α/β ratios can be determined from in vivo multifraction experiments even though the absolute values of each coefficient are unknown. Adding to the utility of the LQ model is that responses of tissues to change in dose fractionation can be predicted from just the α/β ratio.112
FIGURE 2.15. Effective single-dose survival curves for clonogenic cells of jejunal crypts fitted to multifraction data using LQ (broken curves) and TC (solid lines) models. Numbered brackets illustrate that the data were from experiments using that number of fractions. Mean survival curve parameters are as follows: for LQ model, α = 0.23 Gy, β = 0.018 Gy−2; for TC model, 1D0 = 3.57 Gy, D0 = 1.43 Gy, nD0 = 2.37 Gy, and n = 20.4. (From Thames HD, Withers HR, Mason KA, et al. Dose-survival characteristics of mouse jejunal crypt cells. Int J Radiat Oncol Biol Phys 1981;7:1591.)
FIGURE 2.16. Effective multifraction-dose survival curves for cell populations, the survival of which from 2 Gy varies from 0.65 to 0.35. When survival from 2 Gy is 0.5, survival from 30 × 2 Gy is (0.5)30 = approximately 10−9. When this figure was constructed, this survival value was taken as an arbitrary standard against which the relative survival of other populations exposed to 30 × 2 Gy was plotted. The abscissa at the standard survival shows the total doses in 2-Gy fractions necessary to achieve that standard survival level in different cell populations. In all cases, an equal effect per dose fraction was assumed. The ratios of cell survival after a total dose of 60 Gy illustrate the exponential amplification of survival differences with increasing dose. Thus, dose fractionation can transform small differences in response at low doses (2 Gy) to large ultimate differences; measurements must be made accurately after a dose of 2 Gy to predict accurately the ultimate outcome of high-dose multifraction irradiation. (From Withers HR. Predicting late normal tissue responses. Int J Radiat Oncol Biol Phys1986;12:693.)
FIGURE 2.17. Recovery curves of the type first described by Elkind and Sutton.165 The repair of sublethal injury begins immediately and can be measured in terms of survival ratio (inset) or the increment in dose to achieve isosurvival.
FIGURE 2.18. Isoeffect curves in which the total dose necessary for a certain effect in various tissues is plotted as a function of dose per fraction (late effects, solid lines; acute effects, broken lines). Data were selected to exclude an influence on the total dose of regeneration during the multifraction experiments. The isodoses for late effects increase more rapidly with decrease in dose per fraction than is the case for acute effects. (From Withers HR. Biologic basis for altered fractionation schemes. Cancer 1985;55:2086.)
Multifraction Survival Curves
In 1959, Elkind and Sutton110 showed that sublethal damage (SLD) can be repaired, given a few hours of normal metabolic activity. The extent and rate of repair of SLD can be estimated by the change in survival fraction with increasing time between two dose fractions or by the increase in total dose necessary to achieve the same level of cell survival (D2-D1) (Fig. 2.17). If two doses are separated by enough time to permit complete repair of SLD, it is as though the survivors of the first dose had not been previously irradiated and harbor no residual injury. Net survival is the product of the survival after each exposure. In other words, if 2 Gy reduces survival to 50% (S.F.2Gy = 0.5), two doses of 2 Gy would reduce it to (0.5)2 and n doses to (0.5)n—that is, there is an equally proportionate decrease in the survival rate with each equal increment in dose. Although this and the absence of regeneration between doses are not universally accurate assumptions, they are sufficient for modeling purposes. Thus, the dose-survival relationship for a series of equal dose fractions can be considered to give a straight line when plotted on semilogarithmic coordinates (Fig. 2.18). The multifraction survival curve extrapolates to 1 (n = 1) from any dose level, unlike most single-dose low LET survival curves. The slope of such a linear multifraction curve is always less than that for the single-dose curve at an equivalent total dose and becomes shallower the smaller the dose per fraction (Fig. 2.14).
The slope of a linear multifraction cell survival curve can be described by an “effective D0” (effD0, or eD0), which is the dose that reduces survival to e−1 for a particular fractionation regimen. In clinical RT, where most treatment involves multiple dose fractions, effD0 values are much more relevant than D0 values. If S.F.2Gy were 0.5, the corresponding eD0 value for a series of 2-Gy fractions would be 2.9 Gy (S.F. = e−D/effD0). If the S.F.2Gy values ranged from 0.45 to 0.67, the effD0 values would range from 2.5 to 5.0 Gy. In mice, S.F.2Gy values for jejunal crypt and spermatogenic stem cells are about 0.6 and for colonic cells about 0.65, giving effD0 values for 2-Gy fractions of about 3.9 and 5.0 Gy, respectively. It can be calculated, in two ways, that 30 fractions of 2 Gy would reduce survival of jejunal crypt cells to 2 × 10−7:
Small differences between cell types in their intrinsic radiosensitivity to dose fractions of 2 Gy could amplify into large differences after many fractions if there is equal effect per fraction. This could have a major influence on the outcome of therapy. For example, if S.F.2Gy in one tissue were 0.6 and in another 0.5, the ratio of cell survival after 30 doses of 2 Gy would be (
)30, or 237-fold. For 35 doses, the difference would increase to 590-fold. The effect of differences in S.F.2Gy ranging from 0.35 to 0.6 on outcome after 30 fractions is shown in Figure 2.16.166 These very large differences can be judged from the ratios of cell survival rate (on the ordinate), or the range of total doses needed to achieve the same level of cell survival (approximately 10−9) (abscissa).
Common Logarithms and e D10
The effective survival curve for a multifraction regimen is more easily considered in terms of eD10, which reduces survival to 10% (10−1), than in eD0. An approximate value for eD10 for 2-Gy fractions is 6.5 to 7 Gy, which corresponds with S.F.2Gy of about 0.5. Remembering that about 109 tumor cells tightly packed would have a volume of about 1 cm3 and that 1010 cells would form a sphere about 2.2 cm diameter, then an average T3 tumor would contain about 1010 clonogenic cells. Assuming an eD10 of 7 Gy, a T3 tumor treated with 70 Gy would have its surviving clonogen number reduced by 10−10 (10−70/7), or to an average of 1 clonogen per tumor. Because of the random nature of cell survival, 37% of such tumors would contain 0 clonogens and would be eliminated. If eD10 were 6.5 Gy, then 65 Gy would be sufficient to cure 37% of tumors containing 1010 cells, and a dose of (65 + 6.5) = 71.5 Gy would reduce cell survival to 10−11, resulting in a 90% local control rate.
Tumor Response and Dose Fractionation
As has already been mentioned, there is a wide spectrum of radiosensitivities within and between tumor types and variation in their α/β ratios (Table 2.2). There is also uncertainty in the weight that should be attributed to each of the four “Rs” affecting survival of tumor cells during a course of RT58: repair of SLD, repopulation, redistribution through the division cycle, and reoxygenation of hypoxic tumor clonogens. If the effects of such phenomena on the response to each dose were constant throughout the course of a multifraction regimen, the resulting survival curve would be linear and the slope would depend on the extent to which each phenomenon affected the response. More likely, their influence, and particularly the effect of reoxygenation and regeneration, varies with time as treatment progresses. In some normal tissues (e.g., oropharyngeal mucosa), regeneration of surviving clonogenic cells late in a course of 1.8- to 2-Gy fractions may outstrip the cytocidal effect of treatment and the net survival curve will have a positive slope. The same phenomenon may occasionally happen within a tumor during treatment. Despite the uncertainties, it is possible to model dose-response relationships for tumors and to use these models to guide treatment.
Tumor Control Probability
Tumor Control Probability for Clinically Detectable Disease
The probability of tumor control increases as radiation dose increases, although not linearly. Success or failure depends on killing the last surviving clonogen. Permanent tumor control (not palliation) is achieved abruptly as the last clonogen is sterilized. A plot of tumor control probability (TCP) versus dose for a single tumor therefore shows no response up to a certain dose and then an immediate increase to 100% at death of the last clonogen. For a series of patients, even if they have identical tumors, the shape of the TCP curve is different because cell killing is a random process. After a certain dose of irradiation, the numbers of surviving clonogens per tumor will begin to follow a Poisson distribution. For example, if cell survival is reduced to an average of one clonogen per tumor, there would be, on average, 37% with no survivors, 37% with one, 18.4% with two, 6.1% with three, and 1.5% with four or more surviving clonogens. Obviously, the local control rate would be 37%, not 0%. Poisson statistics correlate probability of tumor control with cell survival rate by:
where x is the average number of surviving clonogens per tumor, which in turn is the product of S.F. (fraction of cells surviving) and M (initial cell number). For example, if a tumor contains 1010 clonogens, doses that reduce survival to 10−10 would give an average cell survival of 1 and Pcure to e – (1010 . 10−10) = e−1 = 0.37, or 37%.
If the total dose were increased by two effD0 values, cell survival would be further reduced by two natural logarithms, from 1 to 1 × e−2, that is, to an average of 0.135 cells per tumor, and Pcure = e−0.135 = 0.87, or 87%. It can be calculated that an increase in dose by three effD0 values is sufficient to increase the probability of cure from 10% to 90%. If Pcure = 10% = 0.1 = e−x then x = 2.3. In other words, at 10% local control rate, there is an average of 2.3 clonogens per tumor. After an increase in dose by three effD0 values, Pcure = e−(2.3.e–3) = e−(0.115) = 0.89, or 89%.
FIGURE 2.19. Theoretic tumor control probability (TCP) curves for three sizes of spherical tumors (solid lines) and one that would result from a study incorporating into the dose-response analysis all three tumor sizes in equal proportions (broken line). They were calculated on the assumptions that the dose was given in 2-Gy fractions, that the eD0 value for 2 Gy per fraction was 3.5 Gy, and that a 1-cm diameter spherical tumor contained 107 clonogens. As tumor volume increases, so does the dose required for a certain probability of control. The cell number increases by 8 times and 64 times as the spherical tumor increases from 1-cm diameter to 2 and 4 cm, respectively. An exponential increase in clonogen number is related to a linear increase in dose for an isoeffect. Heterogeneity of even one factor, initial clonogen number, causes the TCP curve to be shallower (broken line). Retrospective clinical studies incorporate a large number of causes for heterogeneity of response.
FIGURE 2.20. Theoretical curves showing the probability of tumor control and normal tissue complications. Both curves have a threshold and are log sigmoid in nature. The art of radiotherapy is to increase the distance between these two curves—that is, to derive a therapeutic benefit. If the normal tissue damage curve is to the left of that for TCP, tumor control is unlikely without unacceptable normal tissue complications.
This relationship between probability of cure and dose, above a certain threshold, is described by a sigmoid curve (Fig. 2.19). It is obvious from the previous equations and calculations that the slope of the curve is a function of the effD0 values for the last few surviving tumor clonogens. It is steeper for neutrons or for single-dose x-ray treatments than for multifraction x-ray exposures and steeper (by the OER) for euoxic than for hypoxic cells.
The dose that yields a 50% control rate is known as the TCD50. A higher rate is usually sought in clinical practice; however, for experimental studies, the TCD50 is useful because it is in a steep part of the TCP curve and is sensitive to small changes in the effectiveness of therapy. (Actually, the curve is steepest at TCD37, where an average of one clonogenic cell survives per tumor.)
TCP curves from experimental animal data are steep, although those for human tumor control are shallower than would be predicted.53,168–176 This reflects heterogeneity in tumor characteristics and treatment prescriptions, which give a wider spread of responses to a given (nominal) dose. However, even in recent analyses where variation in tumor characteristics and treatment parameters has been minimized or adjusted for, and uniformity facilitated by analyzing a body of data large enough to allow relatively homogeneous stages of disease to be studied independently, the TCP curves have been fairly shallow, indicating a need for even better tumor profiling. Heterogeneity may come from molecular cancer-related differences, cancer origin, or numbers of CSCs that would impact their intrinsic radiosensitivity, redistribution and repopulation kinetics, reoxygenation rates, or physical parameters such as dose rate, dose calculation, inhomogeneities of dose distribution, inconsistency of methods of dose prescription, geographic misses, overall time, and more. Of course if RT were perfect, the TCP curve would be flat at a high level of control. As it is, treatment is often individualized (e.g., by prescribing higher doses for larger tumors), which tends to flatten TCP curves.
The effect of constructing TCP curves for a series of tumors nonhomogeneous in only one characteristic—volume (T-stages)—is illustrated in Figure 2.19. If tumors of three sizes varying in diameter by factors of 2 are stratified carefully, three distinct steep TCP curves would be obtained. If an equal number of tumors of all three sizes are analyzed together, the TCP curve obtained would be flat, as shown as a broken line. It would not be appropriate to include tumors with a 64-fold range of volumes in a clinical TCP analysis, although the range in CSC numbers in human cancers of similar T-stage may be of this magnitude. Further variation in the effective number of CSCs would also arise from differences in inherent growth and accelerated repopulation kinetics of surviving cells during RT.
Normal tissue dose-response curves for the incidence of a certain complication are also sigmoid above a certain threshold. Such curves are used for estimating LD50 or ED50 values, the doses that cause in 50% of cases lethality or any specified effect, respectively. Because normal tissues are more homogeneous than tumors in their composition and radiation responses, complication probability curves are steeper than those for tumor control.20
The art of RT can be quantified in a risk-benefit analysis as a balance between the TCP and the probability of complications, NTCP (both represented by sigmoid curves illustrated in Fig. 2.20), with many factors involved. Different points illustrate this:
1. If there is to be therapeutic gain, the biologic effectiveness of RT must be greater in the tumor than in normal tissues—that is, normal tissues must be preferentially spared.
2. In the steep midrange of TCP or complication frequency, a small change in the biologic effectiveness of therapy can give a substantial change in clinical outcome. Conversely, if a change in treatment produces a modest difference in TCP or frequency of complications, it should not be interpreted as being from a large change in biologic effectiveness. Effectiveness is expressed more appropriately as change in dose to achieve a certain isoeffect—that is, by quantifying the lateral shift of the TCP or NTCP curve. The ratio of isoeffect doses is the dose-modification factor.
3. At incidences of less than about 10% and greater than about 85%, changes in biologically effective dose will appear less than in the midrange. For example, if the TCP is 90%, minor therapeutic gain would be achieved from an increase in dose, and perhaps a therapeutic disadvantage if there was an associated incidence of severe complications that already lay on the bottom part of the NTCP curve.
4. Because TCP and NTCP curves are generally close together, it is usually inappropriate to produce no complications. Within the treatment volume, a certain incidence of injury to normal tissues sufficient to be defined as a complication is a prerequisite to good curative RT under most circumstances where the TCP is not close to 100%.
5. If the TCP is to the right of the NTCP curve, tumor control without a high incidence of complications is unlikely and other modalities should be used, either independently or as adjuvants. For example, if the TCP curves for large tumors lie to the right of those for complications, whereas those for smaller tumors are to the left, a therapeutic gain might be derived from excision of the large tumor or some form of chemotherapy that would add to the effect of RT without increasing normal tissue toxicity. However, a 50% or even 90% debulking is of modest value. A 90% reduction in tumor volume represents only a one-decade decrease in cell number, equivalent to about 6.5 to 7 Gy in 2-Gy fractions, other things being equal.
6. It defies biologic rationale to deny the existence of a potential benefit from maximizing the dose in a patient considered to have some finite chance of tumor control merely because retrospective studies show a shallow TCP curve.
Tumor Control Probability for Subclinical Disease
The threshold-sigmoid TCP curve for clinically detectable tumors is not appropriate for subclinical metastases. If 10n clonogens represent the upper limit of clinical undetectability of metastases, then patients who harbor subclinical metastases must have a tumor burden of between 1 and 10n cells. Given that micrometastases grow exponentially, it is reasonable (as a working hypothesis) to assume that an even distribution of the logarithm of metastatic clonogens (between 1 and 10n cells) exist within a series of patients.177,178
If a reasonable value for n is 9, then 11% of patients with subclinical metastases would have between 1 and 10 metastatic clonogens, another 11% between 10 and 100, another 11% of patients between 102 and 103, and so forth. Based on this simple model, the TCP curve would exhibit no threshold and would be shallow. The lack of a detectable threshold would reflect the existence of a very small number of metastatic cells in some patients, and the shallow slope would reflect the wide variation in cell burden within the population harboring subclinical metastases. The model in Figure 2.21A illustrates these concepts. Although this distribution could be modified by many factors (notably gompertzian growth of micrometastases), it is generally consistent with results of RT of subclinical metastases (Fig. 2.21B).
FIGURE 2.21. Percentage control rate as a function of dose for subclinical metastases. A: Modeling on the basis of a uniform distribution of the logarithm of numbers of metastatic tumor cells per patient ranging between 1 and 109. An SF2Gy value of 0.5 was used. The intercept of this theoretic curve is displaced slightly from zero because of the random statistical chance that any cell will survive any dose of radiation. In addition, at high doses, the probability of sterilizing all tumor cells approaches 100% asymptotically for the same reason. B: Percentage reductions in recurrence as a function of dose from reports in the literature for various tumor types. Solid symbols represent data from prospective randomized trials. Other data are retrospective comparisons between control rates with and without elective irradiation. (From Withers HR, Peters LJ, Taylor JMG. Dose-response relationship for radiation therapy of subclinical disease. Int J Radiat Oncol Biol Phys 1995;31:353.)
TIME-DOSE ISOEFFECT FORMULAS AND DOSE FRACTIONATION
History
Early isoeffect curves related the total dose required to produce certain skin reactions or to achieve a certain TCP to the treatment time over which the dose regimen was delivered.179 That work preceded the demonstration by Puck and Marcus4 in 1956 that mammalian cell survival curves had a shoulder. It also preceded the work of Elkind and Sutton180 demonstrating that sublethal injury could be repaired and therefore that the number of fractions, not just overall time, was important. Fowler and Stern181 varied the number of fractions (N) and overall time (T) experimentally and showed that they were independent variables. Ellis182 developed the nominal standard dose (NSD) formula to incorporate these two variables into an isoeffect curve that was thought to be more clinically relevant than the original Strandqvist curves.179
More recently, isoeffect curves that are based on parameters of dose survival curves only112,160,165,183 or that include other biologic parameters, such as regeneration, have been proposed.184,185 We now understand that there can be no single universally applicable isoeffect equation or curve because tissues (and tumors) differ in the characteristics that determine their fractionation responses and repopulation kinetics.
Acute- Versus Late-Responding Tissues
The most general biologic phenomenon influencing the fractionation response is repair of sublethal injury.110 This varies among tissues, with slow-responding tissues consistently showing a greater capacity than rapidly responding tissues.112,165,186–188 This may be because surviving cells in early-responding tissues redistribute through the division cycle during the interfraction interval and express unrepaired damage as they do so, or they may move into cell cycle phases that are more radiation sensitive. Alternatively, repair of sublethal injury may be more complete in late-responding than in early-responding tissues.189 Regardless of the mechanism, late-responding tissues are spared more than acute-responding tissues by dose fractionation—that is, the dose for an isoeffect increases more rapidly in late-responding tissues as dose per fraction is reduced.
The slopes of the isoeffect curves in Figures 2.18 and 2.22 reflect the shape of the dose-survival or dose-function curves for responses in various tissues. Those in Figure 2.22 are constructed using the LQ formula (see later discussion) to correct the total dose (normalized to 1) to that which would be given if the dose per fraction were a standard 2 Gy. Shown for comparison is a plot based on the NSD formula190 where the correction for time is ignored—that is, where D = NSD × N0.24.
In terms of the LQ cell survival model, late effects tissues have a low α/β ratio,112,188,191 describing a curvier curve than that for acute-responding tissues (Fig. 2.13). Thus, low dose per fraction spares late- more than acute-responding tissues. Examples of α/β values for tissues are shown in Table 2.2.
Some specific clinical implications of the differences in fractionation response between acute- and late-responding tissues are as follows:
1. Large dose fractions are relatively more harmful for late-responding tissues. If two different fractionation regimens are used—one with large and the other with small doses per fraction—that achieve the same acute reactions, late responses will be more severe from the large dose per fraction regimen. This has been observed in many clinical studies.112,165,192–194
2. Because the therapeutic differential between late-responding tissues and acute-responding tumors increases with decreasing dose per fraction, a therapeutic gain should result from use of the smallest practical dose per fraction.195For example, if two fractions of 1.15 Gy achieve the same effect in a late-responding tissue as one fraction of 2 Gy yet two fractions of only 1.05 Gy achieve the same tumor control rate, the therapeutic gain would be 1.15/1.05 = 1.1. Thus, hyperfractionation using two fractions of 1.15 Gy to replace one fraction of 2 Gy would increase the biologically effective tumor dose by 10% with no increase in late complications, although acute-responding normal tissues will also receive an increased biologic dose.
3. To maximize the potential therapeutic gain from hyperfractionation, repair of SLD in late-responding tissues must be complete, implying fractionation intervals of at least 6 hours.167,189 For the spinal cord, a longer interval seems needed.80,189,196,197–198
4. The relative biologic efficiency (RBE) for high LET radiations is greater at low doses than high doses for late effects because the sparing effect from fractionated x-ray doses is lost188 (Fig. 2.13).
FIGURE 2.22. Isoeffect curves relating total dose to dose per fraction, with total dose being expressed as a ratio of that necessary in 2-Gy fractions (i.e., the LQED2Gy). α/β ratios (in Gray) are shown on the curves. Although the α/β ratios are not yet established accurately or even precisely for most normal tissues, especially late-responding ones (Table 2.2), the likely order of fractionation sensitivities is as shown. The broken line traces the change in dose, as it would have been predicted by a factor N0.24 in the NSD formula. Phenomena other than repair of sublethal injury, specifically repopulation, are not accounted for by these curves. (From Withers HR. Contrarian concepts in the progress of radiotherapy. Radiat Res 1989;119:395.)
Linear Quadratic Formula for Calculating Isoeffect Relationships
It is possible to use the LQ response formula to change the size of dose per fraction (within limits)165:
where Dnew is the new total dose for a change in size of dose per fraction to dnew; Dref is the previous total dose in fractions of dref; and α/β is for the tissue in question, all doses being expressed in Gray. For example, if dnew were 4 Gy and dref were 2 Gy, the ratio of total doses to achieve the same effects in a tissue with an α/β value of 2 Gy would be 0.66. For a tissue with an α/β value of 10 Gy, it would be 0.85. This calculation was used to derive the data in Figure 2.22.
The following caveats apply to this use of the LQ isoeffect formula:
1. The LQ response model does not apply equally well at all dose levels. It fits experimental data well over a limited dose range (e.g., 2 to 8 Gy); however, its validity and precision above161 or below199 that range is in doubt.
2. Using uncertain values for α/β ratios have more of an effect in late-responding tissues and carry more risk. There is little difference in isoeffect curves for different dose per fraction regimens delivered to tissues with α/β values of 10 Gy or more, whereas the difference is much larger for tissues with α/β values of 1 and 2 Gy (Fig. 2.22).
3. Dose corrections are most affected at low doses per fraction, where the LQ response model is poorly validated. Using the isoeffect formula presented earlier, or the curves in Figure 2.22, compare the ratios of isoeffect doses for a tissue characterized by an α/β value of 2 Gy when the dose per fraction is changed from 2 to 1 Gy (a 33% increase) and from 2 to 3 Gy (a 20% decrease).
4. The basic LQ model has no time parameter. This is not so important for late-responding tissues that turnover slowly as it is for tumors and acute-responding tissues. Factors can be added for incomplete repair, reoxygenation, and redistribution184 and to account for the time to initiation of proliferation (kickoff time) and the impact of proliferation on overall response,200 although the values are uncertain.
For these reasons, the LQ formula or curve should be used cautiously.
Influence of Regeneration (Repopulation)
Normal Tissues
The influence of overall duration of therapy on fractionation responses comes largely from large differences in the time of onset and kinetics of regeneration (repopulation) among various normal and neoplastic tissues (Fig. 2.3). This is why a constant exponent for overall treatment time in an isoeffect formula without regard to tissue type is a dangerous simplification with no biologic foundation. For example, early regeneration of intestinal mucosa or bone marrow makes a major contribution to the net response if a treatment regimen is spread out over several weeks, whereas it provides little or no benefit to spinal cord, kidney, or dermis. A constant exponent of <1 also implies a greater effect on isoeffect dose when values for T are small, whereas the major contribution of regeneration occurs when T values are large (e.g., 10 to 60 days).
Tumors
Such as with acute-responding normal tissues, tumors accelerate their growth in response to injury. Some clinical implications of tumor regeneration for curative RT are as follows:
1. Protracting treatment longer than necessary will likely be a disadvantage. For example, using 1.8 Gy rather than 2-Gy fractions given five times per week extends overall treatment time by about 10% and should be reserved for situations in which acute responses are likely to limit the rate of dose accumulation or where there may be substantial inhomogeneities in dose distribution. Inhomogeneity can lead to double trouble, with areas receiving a high total dose also receiving high doses per fraction. The increase in physical dose has an added biologic component that will compromise later-responding tissues (Fig. 2.23).
2. If a break in treatment is necessary because of acute toxicity, it should be kept as short as is tolerable.
3. Planned split-course therapy is inadvisable unless it is part of an accelerated treatment protocol that ultimately shortens the overall treatment duration (discussed later).
4. Breaks in therapy for nonmedical reasons (machine breakdown, holidays) may merit “catch-up” treatments in patients being treated for cure—for example, by treating twice on some remaining day.
5. Obviously, rapidly growing tumors must be treated rapidly. However, it is reasonable to accelerate treatment of tumors with a high proliferative index, regardless of their growth rate, because they are likely to reduce their rate of cell loss and regenerate earlier and faster during treatment. In fact, treatment should never be unnecessarily protracted (consistent with other considerations), because it is difficult to predict the accelerated repopulation response of individual tumors.
FIGURE 2.23. Influence of dose heterogeneity on physical and biologic doses as a function of the isodose line chosen for defining the tumor dose and of α/β ratio of the tissue located at Dmax. The divergence of biologic and physical doses reflects the change in biologic dose that results from change in dose per fraction that derives from the heterogeneity of dose distribution. The lower the α/β ratio, the greater the divergence between biologic and physical doses. This “double trouble” is not reflected in physics isodose distributions and may explain not only a spurious volume effect but also may contribute to low values of tolerance doses that have appeared in the literature from time to time.
MODIFICATION OF DOSE FRACTIONATION PATTERNS
Standard RT regimens of about 2 Gy per day 5 days per week were developed empirically over the decades and serve as a good overall regimen for dose fractionation. However, it should be obvious that this may not be the best for all clinical situations and that it is certainly inappropriate in some (e.g., in obviously rapidly growing tumors75,78,99).
Some biologic factors relevant to modified dose fractionation are as follows:
1. Tissues differ in their response to multiple dose fractions (Figs. 2.14, 2.18, and 2.22) and tissue-related factors have to be taken into account.
2. Acute-responding normal tissues have an enormous capacity for repopulation (Fig. 2.3).
3. Tumors can show accelerated repopulation (Figs. 2.5, 2.6, and 2.7). Although there is probably great variability from tumor to tumor, on average, the lag time before its onset is longer and its rate slower than for acute-responding normal tissues.
4. Cell cycle redistribution to asynchrony between dose fractions produces a net sensitization in proliferative tissues but not in nonproliferative, late-responding tissues.
5. Hypoxia affects tumor responses to a lesser extent at low doses; however, the kinetics of reoxygenation are probably rapid in relation to the duration of most RT regimens, and it may be a factor only if single or a few large doses are used.
6. Slow-responding tissues “repair” better than acute-responding tissues.
7. Repopulation in normal tissues, and possibly in tumors, may occur rapidly during any treatment break.
8. The lag time to the onset of repopulation may be shortened to a limited extent by causing injury faster—for example, by increasing dose intensity or by concomitant chemotherapy.
9. Some tumors behave similar to late-responding tissues and have a low α/β ratio and proliferate slowly.
10. True “stem” cells and CSCs may be relatively quiescent and have different characteristics from the bulk of the normal tissue and tumor. For example, hematopoietic stem cells have a low α/β ratio201 even though most bone marrow responses including lethality have a high α/β ratio. As a result, the target population for an end point has to be carefully considered.
Hypofractionation
Many clinical and experimental animal studies have shown that increasing the size of dose fractions increases the severity of late responses in relation to acute responses. Classically trained radiation oncologists typically associate high dose fractions with increased vascular injury and chronic inflammation, and fear severe late effects that can drastically affect patient quality of life and even be deadly. However, there are advantages in hypofractionation that may outweigh the biologic disadvantages in some circumstances, although these should be carefully evaluated in advance of treatment.
Logistic considerations have driven attempts to hypofractionate since the early days of RT, with little success outside of palliation. These are still important today. The advances in physics, particularly those related to the use of IMRT to minimize the dose to normal tissues and reduce tumor margins, along with CT scanning and gating techniques for precision positioning and to account for body motion, make compelling arguments for revisiting hypofractionation for certain conditions. From the radiobiologic perspective, if the fractionation response of the tumor is similar to that of late-responding normal tissues, increasing the size of dose fractions (hypofractionation) should not provide much of a therapeutic disadvantage, and decreasing the overall treatment time may help counter accelerated tumor repopulation, although this would not be expected in tumors with low α/β ratios; therefore, these advantages are mutually exclusive. Disadvantages include an increased risk of geographic miss and “cold spots” that will be discussed later. In addition, it is an oversimplification to state that any tissue or tumor has an α/β ratio. Every tissue contains multiple structures and different cell types, each with their own response to changing size of dose fractions. Using IMRT to minimize dose to critical normal tissue structures that could be affected by high dose fractions or accelerated treatment may be critical for successful hypofractionated treatments.
The use of single and high dose fractions has had some clinical success, outside of the phase III clinical trials that tested modestly accelerated RT in head and neck SCC in the 1970s,202 which will be discussed later. Leksell203introduced sterotactic radiosurgery in the late 1960s at the Karolinska Institute to treat inaccessible cerebral lesions and in particular small arteriovenous malformations with single-dose treatments. This was followed by the use of linear accelerators to give single stereotactic radiosurgery doses or a small number of fractions as stereotactic radiotherapy (SRT), and these have proved clinically effective in treating a variety of benign and malignant brain diseases.204 SRT has been extended to extracranial sites in the form of SBRT and its extension, stereotactic ablative radiotherapy (SABR), which uses a small number of high dose fractions, such as three fractions of 15 to 20 Gy. Although SABR is relatively new, the finding that it dramatically improves outcome in medically inoperable early-stage non–small cell lung cancer patients5,205 and in patients with liver and other metastases18 has generated much interest in applying it more widely. These findings, together with reports of excellent outcomes from high dose rate afterloading brachytherapy,206 have prompted a re-evaluation of how traditional radiobiologic attitudes apply to these new clinical procedures.207
Stereotactic delivery of IR with IMRT is practically required for hypofractionation because it allows dose to a target to be chosen and delivered with a steep falloff to surrounding normal tissues. Tumor margins can be decreased so that less normal tissue receives high radiation doses, although the integral dose to the rest of the body is generally greater.208 This contrasts with the homogeneous fields that tend to be used in more classical treatments. Relatively nonhomogeneous dose distributions result from IMRT, although this may be positively exploited by generating “hot spots” within targets where they are thought of most value (dose painting), although this currently has more theoretical value than practical application.
The dose falloff and inhomogeneity may have a biologic advantage in generating gradients of cytokines and chemokines that spatially organize infiltrating cells. The more-focused beam and sharp dose falloff may generate more “danger” signals for tumor immunity and more abscopal effects, which could assist in combating micrometastatic disease.29,72,209,210 In fact, the preliminary data suggest that high single doses of ablative RT may not be optimal for the generation tumor-specific immunity and that moderately high hypofractionated doses may be superior in this respect72 [Schaue, in press, 2011]. Proof of this would require clinical studies to be performed with different radiation protocols. Another possible advantage of high dose fractions or single radiation doses may be that they are more cytotoxic for microvasculature than 2 Gy, as was suggested by Fuks and Kolesnick,24 although Tsai et al.144 in murine prostate tumors showed that fractionated protocols can cause vascular loss not dissimilar to that of high single doses a few weeks after RT.
It is important to consider the aim of the hypofractionated therapy. Conventional RT aims to maintain normal tissue function. Hypofractionation can be applied in situations where there is no advantage to be gained from conventional fractionation; however, the aim should be the same, and it must be remembered that acute toxicity may be increased and tumor cure compromised by less reoxygenation. It may therefore be best to limit fraction sizes to <6 Gy, total time to at least 1 week, and the total dose to that predicted by the LQ model to be safe. If there is a gain to be had by exploiting differences in α/β ratios between tumor and late-responding normal tissue, hypofractionation may be a mistake, even allowing for improved technology.
As an example, if we assume α/β ratios of 10 Gy and 3 Gy for tumor and critical late-responding normal tissue, respectively, and change a treatment of 66 Gy in 2-Gy fractions to a 4-Gy regimen, the isoeffective total dose for the late-responding tissue would be 0.71 of 70 Gy (49.7 Gy). However, to keep tumor control rate constant would require only 0.86 of 70 Gy (60.4 Gy). By giving 49.7 Gy, the tumor would be relatively underdosed, receiving only 83% (0.71/0.86) of the equivalent of 70 Gy in 2-Gy fractions. To maintain the biologic effect on the tumor, a total of 60.4 Gy should be given, and the biologic dose to the late-responding tissues would be 20% (0.86/0.71) too high. Obviously, the therapeutic ratio will be reduced regardless of whether the change in total dose was aimed at an isoeffect for late responses (49.7 Gy) or for isocontrol rate for the tumor (60.4 Gy).
SABR, by contrast, has a very different aim. Using the LQ formula, 3 × 20 Gy is equivalent to around 275 Gy in 2-Gy fractions to late-responding tissues with an α/β ratio of 3 Gy207; clearly, this is a dose that would not be given conventionally, and this calculation in part reflects the inappropriateness of using the LQ model above 7-Gy dose sizes. However, no calculation is needed to know that 3 × 20 Gy is an ablative regimen. SABR doses do appear to be well tolerated in the limited situations where they have been employed, with the caveat that there are not sufficient patients treated in this fashion to allow long-term effects to be properly assessed. As well, the radiobiologic underpinnings of this technique still have to be fully evaluated.
Under any circumstances, SABR will cause both vascular and parenchymal cell loss. Critical questions to be asked relate to the site irradiated and the volume involved and what can be tolerated. Radiobiologic advantages of high dose to a small volume may include allowing angiogenesis and stem cell migration from surrounding normal tissue to effect better normal tissue recovery and limit the development of normal tissue hypoxia. On the other hand, serially organized parenchymal structures such as nerves and bronchioles will be compromised more than structures where FSUs are organized in parallel, which is why SABR doses have to be moderated for tumors that are centrally located in the lung. In the lung, loss of peripheral tissue function in small areas is not clinically important because there is ample residual lung function. This is not the case for all tissues. Site and volume appear to be the main constraints that must be applied for ablative therapy, although the dose-volume relationships are not known for most sites, and it is unlikely that dose-volume histograms can be relied on for guidance.
Accelerated Treatment
Accelerated treatment may be defined as a shortening of the overall treatment duration without a comparable reduction in total dose. However, in practice, lower doses in fraction sizes >2 Gy are generally given (e.g., 50 to 55 Gy in 3 to 4 weeks), with the continuous hyperfractionated accelerated radiotherapy (CHART) regimen being an exception with 1.5 Gy three times per day, 7 days per week. The aim of accelerated treatment is to minimize tumor growth or regeneration during therapy. The importance of accelerated repopulation during tumor treatment can be visualized by thinking that two to three doublings of surviving clonogens should add about the same (four- to eightfold) increment in tumor cell burden as a one-step increase in T-stage. The difficulty is to avoid excess toxicity in acute-responding normal tissues by reducing the benefit normally derived from its repopulation. Using larger doses per fraction would accelerate treatment; however, this is not advised in curative therapy unless under specific circumstances, as discussed earlier. In practice, accelerated regimens should use conventional or even reduced doses per fraction given more frequently than usual (i.e., six or more times per week).
There are numerous ways to increase the intensity of dose accumulation from the “standard” of 2 Gy, five times per week (a regimen that is already accelerated 10% in relation to 1.8-Gy fractions):
1. Multiple standard 2-Gy fractions per day: This type of regimen has been given in a continuous course lasting <2 weeks with good local control rates but a high frequency of severe complications.211
2. Relative hypofractionation: About 50 Gy given in 15 fractions in 3 weeks or 20 fractions in 4 weeks134,172 are standard in a number of centers.
3. Concomitant boosting: The boost dose to a reduced volume is given “concomitantly” with the treatment of the initial larger volume rather than as a sequel, as would be standard in a shrinking-field procedure.103 The boost is given as a second dose in the 1 day, with an interfraction interval of at least 6 hours, on several days, preferably during the later part of treatment when normal tissue regeneration is in full progress.
4. CHART: With CHART, 51 to 54 Gy is given as 1.4- or 1.5-Gy fractions, three times daily at 6-hour intervals for 12 consecutive days.80,197 In a large prospective, randomized CHART trial, tumor control rates were increased with acceptable acute morbidity. Some permanent sequelae (e.g., xerostomia, fibrosis) were reduced, although myelopathy was more likely when the spinal cord was treated three times per day.
5. Split-course accelerated treatment: Head and neck tumors were given about 38 Gy over about 10 days as two fractions of 1.6 Gy per day; after a break of 12 to 14 days, an additional 28 Gy (approximately) was delivered, with the total treatment lasting about 6 weeks.47,104 The results were better than in historic controls.
6. Brachytherapy: This form of accelerated RT will be discussed separately.
7. Other: A meta-analysis of randomized trials treating mostly cancer of the oropharynx and larynx and comparing conventional RT with accelerated RT with or without total dose reduction showed increased survival benefit of 2% without dose reduction and 1.7% with dose reduction at 5 years (p = 0.02).202
Because of the lag time before its onset, accelerated tumor regrowth has its greatest effect late in a standard regimen. Therefore, even a 1-week shortening could be advantageous (Figs. 2.5 and 2.6). However, excessive shortening to less than the lag time (e.g., to <3 to 4 weeks in head and neck cancer) is unlikely to improve tumor control rates, especially if the total dose is reduced to maintain acceptable acute toxicity.
Patients most suited to accelerated regimens are those with tumors that are rapidly growing or have a high potential for rapid regrowth—for example, those with a high proliferative index or a high cell loss factor. It may become possible to predict which tumors will repopulate early and quickly.80,82,84 One sign may be rapid regression. In principle, all tumors should be treated in an overall time that is as short as possible and consistent with acceptable acute morbidity; however, caution should be taken to avoid self-defeating reductions in total dose below standard levels and gaps in treatment because of severe acute toxicity that can be counterproductive.
Although modifying and individualizing fractionation patterns may improve the outcome for some patients, accelerated tumor growth, repopulation in acute-responding tissues, and differences in response between late-responding normal tissues and tumors have important implications for everyday conventional treatment.79 For example, RT, at least for head and neck cancers, should not be completed on a Monday after a weekend break because this is a 3-day extension. Likewise, a course of curative therapy should not start on a Friday. In general, breaks in treatment (public holidays, patient demands, etc.) should be countered by delivering more than five fractions in at least one of the weeks of treatment without resorting to large fractions. Chemotherapy given over several weeks before the start of RT may also initiate accelerated repopulation and compromise the chance for local tumor control. Such neoadjuvant therapy might be more effective if given during or after RT, at a time when surviving tumor clonogens are actively proliferating. Obviously, treatment of metastases may affect these decisions.
Hyperfractionation
Hyperfractionation is defined as the use of smaller-than-standard doses per fraction. It can be achieved without extending the overall treatment duration by treating once a day for 6 or 7 days per week but is usually achieved by giving two fractions per day for 5 days per week. Its aim is to increase the therapeutic differential between late-responding normal tissues and acute-responding tumors. It does this primarily by exploiting differences in their response to dose fractionation, although historically it was introduced to exploit the self-sensitizing effect of cell cycle redistribution present in the tumor but absent in late-responding normal tissues. Another rationale is that the OER is lower at low doses. When two fractions are given per day, the interfraction interval should be as long as possible, although preferably not less than 6 hours, and longer if CNS tissue is involved because repair there may continue for more than 12 hours. Hyperfractionation may not be an advantage in the treatment of slowly proliferating tumors because, as with slowly proliferating normal tissues, their α/β ratio may be low.
Clinical evidence suggests that to achieve comparable toxicity in fibrovascular tissues, one fraction of 2 Gy per day should be replaced with two fractions of about 1.2 Gy. For comparison, the dose per fraction necessary for an isoeffect in acute-responding tissues (and most tumors) would be about 1.05 Gy. Assuming that the isoeffect dose for late-responding tissues was increased by 20% and that it was increased for tumor by only 5%, the therapeutic differential would be increased by 1.2/1.05 = 1.14. Therefore, if two fractions of 1.2 Gy per day replaced one fraction of 2 Gy per day, the acute responses of normal tissues and cytotoxicity for tumors would be increased as if the dose had been increased by 14%, whereas late responses would be unaltered. Coincidentally, if the overall treatment time were unchanged, the “biologic” rate of treatment of the tumor and acute-responding normal tissues would also be accelerated by 14%. Hyperfractionation has improved tumor control rates but also increases acute toxicity. A meta-analysis of randomized trials treating mostly cancer of the oropharynx and larynx and comparing conventional RT with hyperfractionated radiotherapy with or without total dose reduction showed increased survival benefit of 8% at 5 years.
Low Dose Rate Irradiation
Low dose rate continuous irradiation has the same biologic advantages as hyperfractionation. Additionally:
1. Proliferative cells may be delayed in their progression through the division cycle,123,138 in particular in late G2 phase. Such a skewed redistribution could self-sensitize proliferative tissues and tumors without affecting late-responding normal tissues.
2. The overall duration of therapy is shortened.
3. The high-dose regions near the radioactive sources have a high probability of being completely sterilized of tumor cells. However, when the logarithmic nature of cell killing is considered, this is not as great an advantage as it may seem. For example, radiation “cautery” of 50% of the tumor cells represents a gain that is equivalent to about 2 to 2.5 Gy of a standard multifraction regimen.
4. The volume of normal tissue receiving a high dose is minimized. Not only is the total dose beyond the treatment volume lower, the dose rate is as well, boosting further the sparing of late-responding tissues. This can be considered an inverse double-trouble effect, or a double advantage.
5. Potential biologic disadvantages are the relatively rapid falloff in dose beyond the treatment volume and the unintentional “cold spots” resulting from seed misplacement or movement. These could decrease the probability of tumor eradication if the tumor lay beyond the specified minimum tumor isodose (geographic miss).
In contrast to low dose rate brachytherapy, the advantages of high dose rate, high dose per fraction brachytherapy come from improved logistics and staff protection, as well as because normal tissues can be tolerably displaced from the high-dose field. Its biologic disadvantages may be loss of therapeutic differential between late effects tissues and tumors, reduced influence of cell cycle redistribution and delay, and increased influence of tumor hypoxia.
Permanent implants of low dose rate radioisotopes with long half-lives (e.g., 125I) also have radiobiologic disadvantages. Late-responding normal tissues may accumulate high doses, and if a set total dose is to be delivered over a relatively long time, the initial dose rate may have to be so low as to facilitate “escape” of CSCs. As with high dose rate brachytherapy, other considerations may overwhelm these radiobiologic disadvantages.
Optimal Dose Rate
A change in dose rate, even between relatively high dose rates such as 10 Gy per minute (600 Gy per hour) to 1 Gy per minute (60 Gy per hour), can affect the response of tissues by allowing more SLD repair. However, sparing of late-responding tissues is most pronounced with the lower dose rates that are more characteristic of brachytherapy, between about 10 and 0.1 Gy per hour. However, in acute-responding tissues, and presumably also in a proportion of tumors, repopulation may be more important than repair at very low dose rates.
In late-responding tissues, such as rat spinal cord and lung,212 a large dose rate effect is seen with change from 4 to 2 Gy per hour. Technical factors in such experiments make investigation of lower dose rates difficult; however, if 2 Gy per hour data are compared with multifraction data, the potential for substantial sparing with a further decrease in dose rate below 2 Gy per hour is indicated.189,198 Such a large sparing effect of the reduced dose rate is to be expected in late-responding tissues for the same reasons as hyperfractionation spares such tissues.
Low Dose Rate Total Body Irradiation
Because proliferative bone marrow populations and leukemia206 cells are characterized by a high α/β ratio (Table 2.2), it is reasonable to prepare patients for bone marrow transplantation using either multiple small dose fractions or continuous low dose rate exposure. Between about 1 and 7 Gy per hour has been chosen for various continuous total body irradiation regimens, which is where biologic effectiveness changes rapidly and even lower dose rates may provide better therapeutic differential.212,213 Because the time to deliver the dose at such low dose rates is so long, many transplant centers give multifraction exposures. In general, low doses per fraction and low dose rates provide the best therapeutic differentials, provided the overall treatment duration is kept short in relation to the growth rate of the leukemic and normal stem cells.
FIGURE 2.24. Effect of varying degrees of underdosage (in terms of D10 values) on tumor control probability (TCP) as a function of tumor volume underdosed. Note that the most important determinant is the magnitude of the underdosage.
Spatial Dose Considerations
Geographic Underdosage of Tumor
The impact of geographic underdosage (cold spots) of areas within tumor on the TCP depends on the number of CSCs in that area and the extent of the underdosage. The likely decline in TCP can be modeled with the assumption that the CSCs are distributed uniformly throughout the clinical tumor volume and are of uniform radiosensitivity. This is unlikely to be correct; however, the exercise is still instructive. Each tumor can be regarded as being composed of a large number of “tumorlets,” each receiving a specified, although variable, dose. The overall TCP can be estimated by summing the TCPs for all the tumorlets. As illustrated in Figure 2.24, this TCP will decrease more the greater the number of underdosed tumorlets (volume) and the larger the decrement in dose. Dose homogeneity and the radiosensitivity of the tumor clonogens will be important. (Note that physical dose inhomogeneities are amplified in “biologic” dose to an extent that will depend on the α/β ratio.)
Figure 2.24 traces the decline in TCP as a function of volume of tumor underdosed and the magnitude of the underdosage. The magnitude of the underdose is shown as multiples of D10, the dose that would reduce the number of surviving clonogens to 10% of the initial number—about 7 Gy for a standard regimen of 2 Gy fractions. Thus, from Figure 2.24 it can be seen that a 7 Gy (1 × D10) underdosage to 10% of the tumor would reduce TCP from 90% to 83%, whereas a 14 Gy underdosage to only 5% of the tumor would reduce TCP from 90% to 55%. The extent of underdosage is more important than the volume underdosed.
Theoretical dose-volume histograms are presented in Figure 2.25. The smallest deviation from 100% dose to 100% tumor is outlined by a-e, showing 5% of the tumor being underdosed by about 5% (0.5 × D10). Underdosage by 10% (1 × D10), 15% (1.5 × D10), or 20% (2 × D10) in a 70-Gy regimen is shown by b-e, c-e, and d-e, respectively. The impact on TCP of the four levels of underdosage depicted by the dose-volume histograms in Figure 2.25 can be read from Figure 2.24. Assuming that 70 Gy in 2-Gy fractions yields a TCP of 90%, then the effect of underdosing 5% of the tumor by 5% (0.5 × D10), 10% (1 × D10), 15% (1.5 × D10), or 20% (2 × D10) would be a decline in TCP by 1%, 3.5%, 12.5%, and 35%, respectively (Fig. 2.24).
It is obvious that most of the dose-volume histogram is irrelevant to TCP, but that small increments in the indentation along the x-axis in the upper right-hand corner may be critical, signifying that moderate to large reductions in dose to even small volumes are dangerous.
The dose-volume histograms b-e, b-f, and b-g in Figure 2.25 illustrate a 10% underdosage to 5%, 10%, and 20% of the tumor, respectively. From Figure 2.24, it can be seen that this 10% underdosage would decrease the calculated TCP by 3.5%, 7%, and 14%, respectively, much less of a loss than associated with the smaller indentations along the x-axis.
Clearly, the extent of underdosage is a more important determinant of TCP than the volume of tumor underdosed. Thus, the indentation along the x-axis of the upper right corner of the tumor dose-volume histogram in Figure 2.25 is more ominous than the indentation in the y-axis when considering TCP.
FIGURE 2.25. Multiple theoretical dose-volume histograms (DVH). By reference to Figure 2.24, the impact on TCP of indentations in the upper right corner of the DVH can be estimated (see text). The underdosage (a-e, b-e, c-e, d-e) is the most important determinant of TCP. The histogram h-j depicts a dangerous DVH for an organ with serially arranged FSUs such as spinal cord but benign for lung, liver, kidney, and so forth. Histogram k-m depicts danger for liver, lung, and kidney, although not for spinal cord.
FIGURE 2.26. Modeling of the effect on tumor control probability (TCP) of increasing dose to increasing proportions of the tumor. Small hot spots are not very useful, especially if TCP is already high. The closer to 100% of the tumor being “overdosed,” the steeper the TCP curve.
Geographic Overdosage of Tumor
Small areas of elevated dose, or hot spots (e.g., in ≤30% of the tumor) produce a negligible change in overall TCP, especially if the TCP from the homogeneous dose is already high (Fig. 2.26). Obviously, the larger the volume of tumor included in the overdosed region, the greater the potential for an increase in TCP, especially if the TCP from the homogeneously lower dose was already low (lower curves, Fig. 2.26). In general, raising the dose to the tumor by dose-painting subvolumes offers little advantage, whereas elevating the dose to the whole tumor could be of great value. For example, an escalation of dose to 30% of the tumor by 1 × D10 could raise the TCP from 10% to nearly 18%, whereas the same increment to the whole tumor could raise the TCP from 10% to 80%.
A significant advantage would only derive from introducing small hot spots in subvolumes within an otherwise homogeneous tumor dose distribution if the hot spots accurately targeted areas of substantially and consistently greater radioresistance (e.g., a nidus of hypoxic cells, or a CSC niche). A tumor with such a radioresistant nidus would not be cured by a relatively low homogeneous dose. At present, there are no proven ways of localizing tumor foci that are consistently radioresistant. Thus, with current levels of understanding and technical expertise, the largest and most certain benefits are likely to be achieved if the dose to the whole tumor is escalated. This is illustrated by the increasing slope of the curves in Figure 2.26 as the volume receiving the higher dose approaches 100%.
Effective Uniform Dose
The effects of inhomogeneous dose distribution on TCP can be quantified by the equivalent uniform dose (EUD).214 An EUD produces a constant probability of tumor control for different volumes of tumor under- or overdosed. In Figures 2.24 and 2.26, a horizontal line for any chosen TCP would join EUDs for various volumes exposed to doses differing by various multiples of D10, the relevant EUD value being the homogeneous dose that produced that TCP.
Dose-Volume Histograms for Normal Tissues
Heterogeneity of dose in the exposed volume will produce different levels of injury, although the net pathophysiologic outcome is dictated by the structure and function of the organ. An organ such as spinal cord with its FSUs arranged in series can be injured by a high dose to even a small volume (Fig. 2.25, h–j) but not by a low dose to a large volume (k-m). However, a large dose to a small volume (h-i) is of little consequence in an organ with a large “reserve” volume of FSUs. Conversely, a relatively low dose to a large volume (e.g., k-m), which may be of little consequence to spinal cord, could be devastating if applied to lung, liver, or kidney. Thus, dose-volume histogram configurations must be viewed against an understanding of the structure and physiology of the specific tissue, as discussed earlier.
QUALITY OF RADIATION
Linear Energy Transfer
The rate at which a charged particle, such as an electron or proton, deposits its energy along its track is described as its LET; the heavier the particle, the higher its LET. Thus, electrons have a predominantly low LET, protons a slightly higher LET, neutrons an even higher LET, and heavily charged particles the highest LET of clinically used radiations.
The rate of energy transfer increases as particles slow, which means that LET is only an average value that is little more than a useful guide to the radiation therapist. As the LET of a beam increases, so does its biologic efficiency, although the increase is most rapid and peaks around 100 to 150 kv/μm. This is thought to represent where ionization events are spaced so that they are most likely to hit both strands of DNA. It then decreases per unit of measured physical dose with further increase in LET (an “overkill” phenomenon). As LET increases, OER decreases inversely with biologic effectiveness, and the impact of variations in cell cycle-related radiosensitivity become less.113 At high LET, single-hit, nonrepairable cell killing increases relative to that from accumulation of sublethal injury; thus, the survival curve for neutrons or heavily charged particles is essentially linear over at least the first decade of cell killing, and there is little sparing from dose fractionation, reducing the differential in fractionation response between late-responding and acute-responding tissues.188
Relative Biologic Efficiency
RBE is a ratio of doses from two beams to produce the same effect:
RBE is usually used to compare high and low LET radiations; however, it has wider applicability for comparing the effectiveness of treatment approaches such as high- and low dose rate x-irradiation. Initially, the standard photon beam was 250 kVp x-rays, but now, at least from the viewpoint of RT, it is 60Co (250 kVp x-rays are about 15% more efficient than 60Co in killing mammalian cells). It is not widely appreciated that even at relatively high dose rates, photon beam effectiveness varies with dose rate and also that this is an important determinant of RBE. Furthermore, it can only be measured accurately if the same effect is achieved by both radiations. The conditions of RBE measurements must, therefore, be explicitly and precisely stated.
Neutrons
Neutrons deposit energy in a tissue through collisions with nuclei (mainly of hydrogen) rather than with electrons, as occurs with photon beams. Although neutrons are uncharged, they eject protons from the nucleus; therefore, the cellular injury they produce is through free radicals formed from ion pairs—the same basic mechanism as that for x-rays. The difference is that the column of ionization produced by the proton ejected from the nucleus by the neutron is much denser than that produced by an electron ejected by a photon. Because of the density of resulting free radicals, neutrons are more likely than x-rays to cause irreparable single-lethal-hit injury to DNA.
The high density of ionization from neutron irradiation also results in a reduced OER. If the OER is 2.6 for x-rays and 1.6 for neutrons, the ratio, which has been called the hypoxic gain factor, would be 2.6/1.6 = 1.6. In other words, if a course of neutrons is given that produces normal tissue sequelae equivalent to those from 66 Gy of x-rays given in 2-Gy fractions, the biologic effect of the neutrons on a completely hypoxic tumor would be equivalent to that from (1.6 × 66 Gy) = 106 Gy of x-rays. Of course, not all tumor cells are hypoxic; thus, the therapeutic gain factor is lower than 1.6 and will depend on the percentage of hypoxic cells and, in a multifraction regimen, the extent of reoxygenation.215
The response of cells to neutrons is less influenced by position in the cell cycle than is the case with x-rays. Therefore, the RBE is greater for cells in x-ray–resistant than x-ray–sensitive phases of the cell cycle. This may affect the therapeutic gain or loss from using neutrons. For example, if a tumor were to consist entirely of cells in an x-ray–resistant phase of the cell cycle, for which the RBE were 5, and if the critical normal tissue were to have an RBE of 3, then the therapeutic gain factor would be 5/3 = 1.66.
Because single-hit nonrepairable events are greater with neutrons, cell survival rate is more nearly exponential over a wider dose range, and steeper, than for x-rays. In terms of the LQ survival curve formula, neutrons have a very high α/β value (e.g., 30 to 100 Gy). Therefore, dose fractionation is of less significance in neutron therapy than in x-ray RT.
Neutrons may have a therapeutic advantage for poorly reoxygenating, poorly redistributing, intrinsically x-ray–resistant, rapidly growing, and rapidly repopulating tumors. Predictive assays would be needed to identify such tumors prospectively, which would also permit more selective modification of x-ray regimens.215
ACKNOWLEDGMENTS
Jan Haas and Natalia Mackenzie helped in the preparation of this manuscript.
REFERENCES
1. Alper T, Howard-Flanders P. Role of oxygen in modifying the radiosensitivity of E. coli B. Nature 1956;178:978.
2. Phillips TM, McBride WH, Pajonk F. The response of CD24(-/low)/CD44+ breast cancer-initiating cells to radiation. J Natl Cancer Inst 2006;98:1777.
3. Banuelos CA, Banath JP, Kim JY, et al. GammaH2AX expression in tumors exposed to cisplatin and fractionated irradiation. Clin Cancer Res 2009;15:3344.
4. Puck TT, Marcus PI. Action of x-rays on mammalian cells. J Exp Med 1956;103:653.
5. Timmerman RD, Park C, Kavanagh BD. The North American experience with stereotactic body radiation therapy in non-small cell lung cancer. J Thorac Oncol 2007;2:S101.
6. Dewey WC, Ling CC Meyn RE. Radiation-induced apoptosis: relevance to radiotherapy. Int J Radiat Oncol Biol Phys 1995;33:781.
7. Kerr JF, Wyllie AH, Currie AR. Apoptosis: a basic biological phenomenon with wide-ranging implications in tissue kinetics. Br J Cancer 1972;26:239.
8. Milas L, Stephens LC, Meyn RE. Relation of apoptosis to cancer therapy. In Vivo 1994;8:665.
9. Strasser A, O’Connor L, Dixit VM. Apoptosis signaling. Annu Rev Biochem 2000;69:217.
10. Kemp CJ, Sun S, Gurley KE. p53 induction and apoptosis in response to radio- and chemotherapy in vivo is tumor-type-dependent. Cancer Res 2001;61:327.
11. Hallahan DE, Haimovitz-Friedman A, Kufe DW, et al. The role of cytokines in radiation oncology. In: DeVita VT, Hellman S, Rosenberg SA, eds. Important advances in oncology. Philadelphia: Lippincott, 1993:71–80.
12. Hong JH, Chiang CS, Campbell IL, et al. Induction of acute phase gene expression by brain irradiation. Int J Radiat Oncol Biol Phys 1995;33:619.
13. Daigle JL, Hong JH, Chiang CS, et al. The role of tumor necrosis factor signaling pathways in the response of murine brain to irradiation. Cancer Res 2001;61:8859.
14. Okamoto M, Kasetani H, Kaji R, et al. cis-Diamminedichloroplatinum and 5-fluorouracil are potent inducers of the cytokines and natural killer cell activity in vivo and in vitro. Cancer Immunol Immunother 1998;47:233.
15. Rodemann HP, Binder A, Burger A, et al. The underlying cellular mechanism of fibrosis. Kidney Int Suppl 1996;54:S32.
16. Von Pfeil A, Hakenjos L, Herskind C, et al. Irradiated homozygous TGF-beta1 knockout fibroblasts show enhanced clonogenic survival as compared with TGF-beta1 wild-type fibroblasts. Int J Radiat Biol 2002;78:331.
17. Marples B, Collis SJ. Low-dose hyper-radiosensitivity: past, present, and future. Int J Radiat Oncol Biol Phys 2008;70:1310.
18. Kavanagh BD, Schefter TE, Cardenes HR, et al. Interim analysis of a prospective phase I/II trial of SBRT for liver metastases. Acta Oncologica 2006;45:848.
19. Potten CS, Merritt A, Hickman J, et al. Characterization of radiation-induced apoptosis in the small intestine and its biological implications. Int J Radiat Biol 1994;65:71.
20. Thames HD, Hendry JH, Moore JV, et al. The high steepness of dose-response curves for late-responding normal tissues. Radiother Oncol 1989;15:49.
21. Merritt AJ, Potten CS, Kemp CJ, et al. The role of p53 in spontaneous and radiation-induced apoptosis in the gastrointestinal tract of normal and p53-deficient mice. Cancer Res 1994;54:614.
22. Merritt AJ, Potten CS, Watson AJ, et al. Differential expression of bcl-2 in intestinal epithelia. Correlation with attenuation of apoptosis in colonic crypts and the incidence of colonic neoplasia. J Cell Sci1995;108(Pt 6):2261.
23. Stephens LC, Ang KK, Schultheiss TE, et al. Target cell and mode of radiation injury in rhesus salivary glands. Radiother Oncol 1986;7:165.
24. Fuks Z, Kolesnick R. Engaging the vascular component of the tumor response. Cancer Cell 2005;8:89.
25. Maj JG, Paris F, Haimovitz-Friedman A, et al. Microvascular function regulates intestinal crypt response to radiation. Cancer research 2003;63:4338.
26. Rotolo JA, Kolesnick R, Fuks Z. Timing of lethality from gastrointestinal syndrome in mice revisited. Int J Radiat Oncol Biol Phys 2009;73:6.
27. Rotolo JA, Maj JG, Feldman R, et al. Bax and Bak do not exhibit functional redundancy in mediating radiation-induced endothelial apoptosis in the intestinal mucosa. Int J Radiat Oncol Biol Phys2008;70:804.
28. McBride WH, Chiang CS, Olson JL, et al. A sense of danger from radiation. Radiat Res 2004;162:1.
29. Schaue D, McBride WH. Links between innate immunity and normal tissue radiobiology. Radiat Res 2010;173:406.
30. Hong JH, Chiang CS, Tsao CY, et al. Alterations of inflammatory cytokine mRNA levels in lungs of two murine strains with different radiation responses. Proceedings of the American Association for Cancer Research Annual Meeting 2000;708.
31. Hong JH, Chiang CS, Tsao CY, et al. Rapid induction of cytokine gene expression in the lung after single and fractionated doses of radiation. Int J Radiat Oncol 1999;75:1421.
32. Demaria S, Bhardwaj N, McBride WH, et al. Combining radiotherapy and immunotherapy: a revived partnership. Int J Radiat Oncol Biol Phys 2005;63:655.
33. Formenti SC, Demaria S. Systemic effects of local radiotherapy. Lancet Oncol 2009;10:718.
34. Fuks Z, Persaud RS, Alfieri A, et al. Basic fibroblast growth factor protects endothelial cells against radiation-induced programmed cell death in vitro and in vivo. Cancer Res 1994;54:2582.
35. Neta R, Okunieff P. Cytokine-induced radiation protection and sensitization. Semin Radiat Oncol 1996;6:306.
36. Neta R, Oppenheim JJ, Schreiber RD, et al. Role of cytokines (interleukin 1, tumor necrosis factor, and transforming growth factor beta) in natural and lipopolysaccharide-enhanced radioresistance. J Exp Med 1991;173:1177.
37. Chiang CS, McBride WH, Withers HR. Radiation-induced astrocytic and microglial responses in mouse brain. Radiother Oncol 1993;29:60.
38. Chiang CS, Hong JH, Stalder A, et al. Delayed molecular responses to brain irradiation. Int J Radiat Biol 1997;72:45.
39. Anscher MS, Kong FM, Jirtle RL. The relevance of transforming growth factor beta 1 in pulmonary injury after radiation therapy. Lung Cancer 1988;19:109.
40. Anscher MS. Targeting the TGF-beta1 pathway to prevent normal tissue injury after cancer therapy. Oncologist 2010;15:350.
41. Charbit A, Malaise EP, Tubiana M. Relation between the pathological nature and the growth rate of human tumors. Eur J Cancer 1971;7:307.
42. Laramore GE, Griffin TW, Parker RG, et al. The use of electron beams in treating local recurrence of breast cancer in previously irradiated fields. Cancer 1978; 41:991.
43. Withers HR, Mason KA, Thames HD Jr. Late radiation response of kidney assayed by tubule-cell survival. Br J Radiol 1986;59:587.
44. Withers HR. Regeneration of intestinal mucosa after irradiation. Cancer 1971;28:75.
45. Withers HR. Recovery and repopulation in vivo by mouse skin epithelial cells during fractionated irradiation. Radiat Res 1967;32:227.
46. Ang KK, Xu FX, Landuyt W, et al. The kinetics and capacity of repair of sublethal damage in mouse lip mucosa during fractionated irradiations. Int J Radiat Oncol Biol Phys 1985;11:1977.
47. Wang CC, Blitzer PH, Suit HD. Twice-a-day radiation therapy for cancer of the head and neck. Cancer 1985;55:2100.
48. Ang KK, Landuyt W, Xu FX, et al. The effect of small radiation doses per fraction on mouse lip mucosa assessed using the concept of partial tolerance. Radiother Oncol 1987;8:79.
49. Neta R. Modulation of radiation damage by cytokines. Stem Cells 1997;15(Suppl 2):87.
50. Dorr W, Noack R, Spekl K, et al. Modification of oral mucositis by keratinocyte growth factor: single radiation exposure. Int J Radiat Biol 2001;77:341.
51. Farrell CL, Bready JV, Rex KL, et al. Keratinocyte growth factor protects mice from chemotherapy and radiation-induced gastrointestinal injury and mortality. Cancer Res 1998;58:933.
52. Yi ES, Williams ST, Lee H, et al. Keratinocyte growth factor ameliorates radiation- and bleomycin-induced lung injury and mortality. Am J Pathol 1996;149:1963.
53. Meoz-Mendez RT, Fletcher GH, Guillamondegui OM, et al. Analysis of the results of irradiation in the treatment of squamous cell carcinomas of the pharyngeal walls. Int J Radiat Oncol Biol Phys1978;4:579.
54. Mason KA, Withers HR, Chiang CS. Late effects of radiation on the lumbar spinal cord of guinea pigs: re-treatment tolerance. Int J Radiat Oncol Biol Phys 1993;26:643.
55. Landuyt W, Fowler J, Ruifrok A, et al. Kinetics of repair in the spinal cord of the rat. Radiother Oncol 1997;45:55.
56. Ang KK, Jiang GL, Feng Y, et al. Extent and kinetics of recovery of occult spinal cord injury. Int J Radiat Oncol Biol Phys 2001;50:1013.
57. Stewart FA, Oussoren Y, Van Tinteren H, et al. Loss of reirradiation tolerance in the kidney with increasing time after single or fractionated partial tolerance doses. Int J Radiat Biol 1994;66:169.
58. Withers HR. The 4Rs of radiotherapy. In: Lett JT, Adler H, eds. Advances in radiation biology, vol. 5. New York: Academic Press, 1975:241.
59. Malaise EP, Fertil B, Chavaudra N, et al. Distribution of radiation sensitivities for human tumor cells of specific histological types: comparison of in vitro to in vivo data. Int J Radiat Oncol Biol Phys1986;12:617.
60. Peters LJ, Brock WA, Chapman JD, et al. Predictive assays of tumor radiocurability. Am J Clin Oncol 1988;11:275.
61. Lim E, Vaillant F, Wu D, et al. Aberrant luminal progenitors as the candidate target population for basal tumor development in BRCA1 mutation carriers. Nat Med 2009;15:907.
62. Pajonk F, Vlashi E, McBride WH. Radiation resistance of cancer stem cells: the 4 R’s of radiobiology revisited. Stem Cells 2010;28:639.
63. Nichols AC, Faquin WC, Westra WH, et al. HPV-16 infection predicts treatment outcome in oropharyngeal squamous cell carcinoma. Otolaryngol Head Neck Surg 2009;140:228.
64. Meyn RE, Stephens LC, Ang KK, et al. Heterogeneity in the development of apoptosis in irradiated murine tumours of different histologies. Int J Radiat Biol 1993;64:583.
65. Meyn RE, Stephens LC, Hunter NR, et al. Reemergence of apoptotic cells between fractionated doses in irradiated murine tumors. Int J Radiat Oncol Biol Phys 1994;30:619.
66. McBride WH, Dougherty GJ. Radiotherapy for genes that cause cancer. Nat Med 1995;1:1215.
67. Peltenburg LT. Radiosensitivity of tumor cells. Oncogenes and apoptosis. Q J Nucl Med 2000;44:355.
68. Thames HD, Bentzen SM, Turesson I, et al. Time-dose factors in radiotherapy: a review of the human data. Radiother Oncol 1990;19:219.
69. Brenner DJ, Hall EJ. Fractionation and protraction for radiotherapy of prostate carcinoma. Int J Radiat Oncol Biol Phys 1999;43:1095.
70. Bentzen SM, Agrawal RK, Aird EG, et al. The UK Standardisation of Breast Radiotherapy (START) Trial A of radiotherapy hypofractionation for treatment of early breast cancer: a randomised trial. Lancet 2008;9:331.
71. Steel GG. Basic clinical radiobiology. London, UK: Edward Arnold, 1993.
72. Dewan MZ, Galloway AE, Kawashima N, et al. Fractionated but not single-dose radiotherapy induces an immune-mediated abscopal effect when combined with anti-CTLA-4 antibody. Clin Cancer Res2009;15:5379.
73. Kerr JF, Searle J. A mode of cell loss in malignant neoplasms. J Pathol 1972;106:xi.
74. Kerr JF, Searle J. A suggested explanation for the paradoxically slow growth rate of basal-cell carcinomas that contain numerous mitotic figures. J Pathol 1972;107:41.
75. Withers HR. Biologic basis for altered fractionation schemes. Cancer 1985;55:2086.
76. Barkley HT, Fletcher GH. The significance of residual disease after external irradiation of squamous-cell carcinoma of the oropharynx. Radiology 1977;124:493.
77. Allen EP. A trial of radiation dose prescription based on dose-cell survival formula. Australas Radiol 1984;28:156.
78. Norin T, Onyango J. Radiotherapy in Burkitt’s lymphoma: conventional or superfractionated regime—early results. Int J Radiat Oncol Biol Phys 1977;2:399.
79. Withers HR, Taylor JM, Maciejewski B. The hazard of accelerated tumor clonogen repopulation during radiotherapy. Acta Oncol 1988;27:131.
80. Dische S, Saunders MI. Continuous, hyperfractionated, accelerated radiotherapy (CHART). Br J Cancer 1989;59:325.
81. Begg AC. Prediction of repopulation rates and radiosensitivity in human tumours. Int J Radiat Biol 1994;65:103.
82. Begg AC, McNally NJ, Shrieve DC, et al. A method to measure the duration of DNA synthesis and the potential doubling time from a single sample. Cytometry 1985;6:620.
83. Wilson GD, Dische S, Saunders MI. Studies with bromodeoxyuridine in head and neck cancer and accelerated radiotherapy. Radiother Oncol 1995;36:189.
84. Horiot JC, Begg AC, Le Fur R, et al. Present status of EORTC trials of hyperfractionated and accelerated radiotherapy on head and neck carcinoma. Recent Results Cancer Res 1994;134:111.
85. Begg AC, Haustermans K, Hart AA, et al. The value of pretreatment cell kinetic parameters as predictors for radiotherapy outcome in head and neck cancer: a multicenter analysis. Radiother Oncol1999;50:13.
86. Mendelsohn ML. Autoradiographic analysis of cell proliferation in spontaneous breast cancer of C3H mouse. I. Growth and survival of cells labeled with tritiated thymidine. J Natl Cancer Inst1960;25:485.
87. Hermens AF, Barendsen GW. Changes of cell proliferation characteristics in a rat rhabdomyosarcoma before and after x-irradiation. Eur J Cancer 1969;5:173.
88. Lindenberger J, Hermeking H, Kummermehr J, et al. Response of human tumour xenografts to fractionated X-irradiation. Radiother Oncol 1986;6:15.
89. Malinen E, Freykowski S, Kummermehr J. Regrowth delay analysis of irradiated tumors in the curative dose region. Int J Radiat Oncol Biol Phys 2000;46:173.
90. Suit HD, Howes AE, Hunter N. Dependence of response of a C3H mammary carcinoma to fractionated irradiation on fractionation number and intertreatment interval. Radiat Res 1977;72:440.
91. Van Peperzeel HA. Effects of single doses of radiation on lung metastases in man and experimental animals. Eur J Cancer 1972;8:665.
92. Parsons JT, Thar TL, Bova FJ, et al. An evaluation of split-course irradiation for pelvic malignancies. Int J Radiat Oncol Biol Phys 1980;6:175.
93. Pino y Torres JL, Lee DJ, Leibel SA, et al. Local control and reduced complications in split course irradiation of prostatic cancer. Int J Radiat Oncol Biol Phys 1981;7:43.
94. Barton MB, Keane TJ, Gadalla T, et al. The effect of treatment time and treatment interruption on tumour control following radical radiotherapy of laryngeal cancer. Radiother Oncol 1992;23:137.
95. Maciejewski B, Withers HR, Taylor JM, et al. Dose fractionation and regeneration in radiotherapy for cancer of the oral cavity and oropharynx: tumor dose-response and repopulation. Int J Radiat Oncol Biol Phys 1989;16:831.
96. Overgaard M. Spontaneous radiation-induced rib fractures in breast cancer patients treated with postmastectomy irradiation. A clinical radiobiological analysis of the influence of fraction size and dose-response relationships on late bone damage. Acta Oncologica 1988;27:117.
97. Withers HR, Peters LJ, Taylor JM, et al. Late normal tissue sequelae from radiation therapy for carcinoma of the tonsil: patterns of fractionation study of radiobiology. Int J Radiat Oncol Biol Phys1995;33:563.
98. Rudat V, Stadler P, Becker A, et al. Predictive value of the tumor oxygenation by means of pO2 histography in patients with advanced head and neck cancer. Strahlenther Onkol 2001;177:462.
99. Fowler JF, Lindstrom MJ. Loss of local control with prolongation in radiotherapy. Int J Radiat Oncol Biol Phys 1992;23:457.
100. Fyles A, Keane TJ, Barton M, et al. The effect of treatment duration in the local control of cervix cancer. Radiother Oncol 1992;25:273.
101. Barker JL, Montague ED, Peters LJ. Clinical experience with irradiation of inflammatory carcinoma of the breast with and without elective chemotherapy. Cancer 1980;45:625.
102. Choi KN, Withers HR, Rotman M. Metastatic melanoma in brain. Rapid treatment or large dose fractions. Cancer 1985;56:10.
103. Knee R, Fields RS, Peters LJ. Concomitant boost radiotherapy for advanced squamous cell carcinoma of the head and neck. Radiother Oncol 1985;4:1.
104. Wang CC. Accelerated hyperfractionation radiation therapy for carcinoma of the nasopharynx. Techniques and results. Cancer 1989;63:2461.
105. King C. Stereotactic body radiotherapy for prostate cancer: current results of a phase II trial. Front Radiat Ther Oncol 2011;43:428.
106. Hryniuk W. Will increases in dose intensity improve outcome. Proc Am J Med 1995;99:69S.
107. Hryniuk WM. The importance of dose intensity in the outcome of chemotherapy. Important Adv Oncol 1988:121.
108. Kim JJ, Tannock IF. Repopulation of cancer cells during therapy: an important cause of treatment failure. Nat Rev Cancer 2005;5:516.
109. Terasima T, Tolmach LJ. Changes in the x-ray sensitivity of HeLa cells during the division cycle. Nature 1961;190:1210.
110. Elkind MM, Sutton H. X-ray damage and recovery in mammalian cells in culture. Nature 1959;184:1293.
111. Withers HR. Cell cycle redistribution as a factor in multifraction irradiation. Radiology 1975;114:199.
112. Thames HD Jr, Withers HR, Peters LJ, et al. Changes in early and late radiation responses with altered dose fractionation: implications for dose-survival relationships. Int J Radiat Oncol Biol Phys1982;8:219.
113. Withers HR, Mason K, Reid BO, et al. Response of mouse intestine to neutrons and gamma rays in relation to dose fractionation and division cycle. Cancer 1974;34:39.
114. Calabrese C, Poppleton H, Kocak M, et al. A perivascular niche for brain tumor stem cells. Cancer Cell 2007;11:69.
115. Gilbertson RJ, Rich JN. Making a tumour’s bed: glioblastoma stem cells and the vascular niche. Nat Rev Cancer 2007;7:733.
116. Vlashi E, Kim K, Lagadec C, et al. In vivo imaging, tracking, and targeting of cancer stem cells. J Natl Cancer Inst 2009;101:350.
117. Diabira S, Morandi X. Gliomagenesis and neural stem cells: key role of hypoxia and concept of tumor “neo-niche.” Med Hypotheses 2008;70:96.
118. Vlashi E, McBride WH, Pajonk F. Radiation responses of cancer stem cells. J Cell Biochem 2009;108:339.
119. Schwarz G. Oxygen dependency of irradiation. Muenchener Med Wschr 1909;24:1–2.
120. Read J. The effect of ionizing radiations on the broad bean root. Br J Radiol 1952;25(Pt X):154.
121. Deschner EE, Gray LH. Influence of oxygen tension on x-ray-induced chromosomal damage in Ehrlich ascites tumor cells irradiated in vitro and in vivo. Radiat Res 1959;11:115.
122. Littbrand B, Revesz L. The effect of oxygen on cellular survival and recovery after radiation. Br J Radiol 1969;42:914.
123. Hall EJ. Radiation dose-rate: a factor of importance in radiobiology and radiotherapy. Br J Radiol 1972;45:81.
124. Brown JM, Giaccia AJ. Tumour hypoxia: the picture has changed in the 1990s. Int J Radiat Biol 1994;65:95.
125. Jain RK. Barriers to drug delivery in solid tumors. Sci Am 1994;271:58.
126. Vaupel P, Schlenger K, Hoeckel M. Blood flow and tissue oxygenation of human tumors: an update. Adv Exp Med Biol 1992;317:139.
127. Vogel AW. Intratumoral vascular changes with increased size of a mammary adenocarcinoma: new method and results. J Natl Cancer Inst 1965;34:571.
128. Moulder JE, Rockwell S. Hypoxic fractions of solid tumors: experimental techniques, methods of analysis, and a survey of existing data. Int J Radiat Oncol Biol Phys 1984;10:695.
129. Powers WE, Tolmach LJ. A multicomponent x-ray survival curve for mouse lymphosarcoma cells irradiated in vivo. Nature 1963;197:710.
130. Olive PL. DNA damage and repair in individual cells: applications of the comet assay in radiobiology. Int J Radiat Biol 1999;75:395.
131. Olive PL, Trotter T, Banáth JP, et al. Heterogeneity in human tumour hypoxic fraction using the comet assay. Br J Cancer Suppl 1996;27:S191.
132. Wang J, Foehrenbacher A, Su J, et al. The 2-nitroimidazole EF5 is a biomarker for oxidoreductases that activate the bioreductive prodrug CEN-209 under hypoxia. Clin Cancer Res 2012;18:1684.
133. Dische S, Anderson PJ, Sealy R, et al. Carcinoma of the cervix—anaemia, radiotherapy and hyperbaric oxygen. Br J Radiol 1983;56:251.
134. Henk JM, Kunkler PB, Smith CW. Radiotherapy and hyperbaric oxygen in head and neck cancer. Final report of first controlled clinical trial. Lancet 1977;2:101.
135. Bush RS, Jenkin RD, Allt WE, et al. Definitive evidence for hypoxic cells influencing cure in cancer therapy. Br J Cancer Suppl 1978;3:302.
136. Henke M, Laszig R, Rube C, et al. Erythropoietin to treat head and neck cancer patients with anaemia undergoing radiotherapy: randomised, double-blind, placebo-controlled trial. Lancet 2003;362:1255.
137. Phillips TM, Kim K, Vlashi E, et al. Effects of recombinant erythropoietin on breast cancer-initiating cells. Neoplasia 2007;9:1122.
138. Bedford JS, Mitchell JB. Mitotic accumulation of HeLa cells during continuous irradiation. Observations using time-lapse cinemicrography. Radiat Res 1977;70:173.
139. Hockel M, Schlenger K, Hockel S, et al. Hypoxic cervical cancers with low apoptotic index are highly aggressive. Cancer Res 1999;59:4525.
140. Graeber TG, Osmanian C, Jacks T, et al. Hypoxia-mediated selection of cells with diminished apoptotic potential in solid tumours. Nature 1996;379:88.
141. Chen FH, Chiang CS, Wang CC, et al. Radiotherapy decreases vascular density and causes hypoxia with macrophage aggregation in TRAMP-C1 prostate tumors. Clin Cancer Res 2009;15:1721.
142. Milas L. Tumor bed effect in murine tumors: relationship to tumor take and tumor macrophage content. Radiat Res 1990;123:232.
143. Ahn GO, Tseng D, Liao CH, et al. Inhibition of Mac-1 (CD11b/CD18) enhances tumor response to radiation by reducing myeloid cell recruitment. Proc Natl Acad Sci U S A. 2010;107:8363.
144. Tsai CS, Chen FH, Wang CC, et al. Macrophages from irradiated tumors express higher levels of iNOS, arginase-I and COX-2, and promote tumor growth. Int J Radiat Oncol Biol Phys 2007;68:499.
145. Ten Hagen TL, Eggermont AM. Changing the pathophysiology of solid tumours: the potential of TNF and other vasoactive agents. Int J Hyperthermia 2006;22:241.
146. Chan N, Koritzinsky M, Zhao H, et al. Chronic hypoxia decreases synthesis of homologous recombination proteins to offset chemoresistance and radioresistance. Cancer Res 2008;68:605.
147. Howes AE, Suit HD. The effect of time between fractions on the response of tumors to irradiation. Radiat Res 1974;57:342.
148. Lyng H, Sundfor K, Rofstad EK. Changes in tumor oxygen tension during radiotherapy of uterine cervical cancer: relationships to changes in vascular density, cell density, and frequency of mitosis and apoptosis. Int J Radiat Oncol Biol Phys 2000;46:935.
149. Palcic B, Skarsgard LD. Reduced oxygen enhancement ratio at low doses of ionizing radiation. Radiat Res 1984;100:328.
150. Catcheside DG, Lea DE, Thoday JM. Types of chromosome structural change induced by the irradiation of Tradescantia microspores. J Genet 1946;47:113.
151. Joiner MC, Lambin P, Malaise EP, et al. Hypersensitivity to very-low single radiation doses: its relationship to the adaptive response and induced radioresistance. Mutat Res 1996;358:171.
152. Joiner MC, Marples B, Lambin P, et al. Low-dose hypersensitivity: current status and possible mechanisms. Int J Radiat Oncol Biol Phys 2001;49:379.
153. Baverstock K. Radiation-induced genomic instability: a paradigm-breaking phenomenon and its relevance to environmentally induced cancer. Mutat Res 2000;454:89.
154. Little JB. Repair of potentially-lethal radiation damage in mammalian cells: enhancement by conditioned medium from stationary cultures. Int J Radiat Biol Relat Stud Phys Chem Med 1971;20:87.
155. Mothersill C, Seymour C. Radiation-induced bystander effects: past history and future directions. Radiat Res 2001;155:759.
156. Read J. The effect of ionizing irradiations on the broad bean root: X. The dependence on the x-ray sensitivity of dissolved oxygen. Br J Radiol 1952;25:89.
157. Dutreix J, Wambersie A, Bounik C. Cellular recovery in human skin reactions: application to dose fraction number overall time relationship in radiotherapy. Eur J Cancer 1973;9:159.
158. Brown BW, Thompson JR, Barkley T, et al. Theoretical considerations of dose rate factors influencing radiation strategy. Radiology 1974;110:197.
159. Withers HR, Chen KY. Poor man’s neutrons? Br J Radiol 1971;44:818.
160. Douglas BG, Fowler JF. The effect of multiple small doses of x rays on skin reactions in the mouse and a basic interpretation. Radiat Res 1976; 66:401.
161. McBride WH, Schaue D. Radiation biology of SBRT: Is there a new biology involved? In: Ahmed MM, Pollack A, eds. Hypofractionation. Scientific concepts and clinical experiences. Ellicott City, MD: LumiText, 2011:3.
162. Brenner DJ. The linear-quadratic model is an appropriate methodology for determining isoeffective doses at large doses per fraction. Semin Radiat Oncol 2008;18:234.
163. Leith JT, DeWyngaert JK, Glicksman AS. Radiation myelopathy in the rat: an interpretation of dose effect relationships. Int J Radiat Oncol Biol Phys 1981;7:1673.
164. Rossi HH. Letter: a note on the effects of fractionation of high LET radiations. Radiat Res 1976;66:170.
165. Withers HR, Thames HD Jr, Peters LJ. A new isoeffect curve for change in dose per fraction. Radiother Oncol 1983;1:187.
166. Elkind MM, Sutton H. X-ray damage and recovery in mammalian cells in culture. Nature 1959;184:1293.
167. Thames HD, Bentzen SM, Turesson I, et al. Fractionation parameters for human tissues and tumors. Int J Radiat Biol 1989;56:701.
168. Agren Cronqvist AK, Kallman P, Turesson I, et al. Volume and heterogeneity dependence of the dose-response relationship for head and neck tumours. Acta Oncol 1995;34:851.
169. Bataini P, Brugere J, Bernier J, et al. Results of radical radiotherapeutic treatment of carcinoma of the pyriform sinus: experience of the Institut Curie. Int J Radiat Oncol Biol Phys 1982;8:1277.
170. Fletcher GH. Clinical dose-response curves of human malignant epithelial tumours. Br J Radiol 1973;46:1.
171. Ghossein NA, Bataini JP, Ennuyer A, et al. Local control and site of failure in radically irradiated supraglottic laryngeal cancer. Radiology 1973;112:187.
172. Harwood AR, Beale FA, Cummings BJ, et al. Supraglottic laryngeal carcinoma: an analysis of dose-time-volume factors in 410 patients. Int J Radiat Oncol Biol Phys 1983;9:311.
173. Peters LJ, Thames HD Jr. Dose-response relationship for supraglottic laryngeal carcinoma. Int J Radiat Oncol Biol Phys 1983;9:421.
174. Shukovsky LJ. Dose, time, volume relationships in squamous cell carcinoma of the supraglottic larynx. Am J Roentgenol Radium Ther Nucl Med 1970;108:27.
175. Shukovsky LJ, Fletcher GH. Time-dose and tumor volume relationships in the irradiation of squamous cell carcinoma of the tonsillar fossa. Radiology 1973;107:621.
176. Thames HD Jr, Peters LJ, Spanos W Jr, et al. Dose response of squamous cell carcinomas of the upper respiratory and digestive tracts. Br J Cancer Suppl 1980;4:35.
177. McBride WH, Withers HR. Radiobiology of subclinical disease. Front Radiat Ther Oncol 1994;28:46.
178. Withers HR, Peters LJ, Taylor JM. Dose-response relationship for radiation therapy of subclinical disease. Int J Radiat Oncol Biol Phys 1995;31:353.
179. Strandquist M. A study of the cumulative effects of fractionated x-ray treatment based on the experience at the Radiumhemmet with the treatment of 280 cases of carcinoma of the skin and lip. Acta Radiol 1944;55(Suppl):300.
180. Elkind MM, Sutton H. Radiation response of mammalian cells grown in culture. Part 1. Repair of x-ray damage in surviving Chinese hamster cells. Radiat Res 1960;13:556.
181. Fowler JF, Stern BE. Dose-time relationships in radiotherapy and the validity of cell survival curve models. Br J Radiol 1963;36:163.
182. Ellis F. Dose, time and fractionation: a clinical hypothesis. Clin Radiol 1969;20:1.
183. Cohen L. Theoretical “iso-survival” formulae for fractionated radiation therapy. Br J Radiol 1968;41:522.
184. Brenner DJ, Hlatky LR, Hahnfeldt PJ, et al. A convenient extension of the linear-quadratic model to include redistribution and reoxygenation. Int J Radiat Oncol Biol Phys 1995;32:379.
185. Travis EL, Tucker SL. Isoeffect models and fractionated radiation therapy. Int J Radiat Oncol Biol Phys 1987;13:283.
186. Jirtle RL, McLain JR, Strom SC, et al. Repair of radiation damage in noncycling parenchymal hepatocytes. Br J Radiol 1982;55:847.
187. Lauk S, Ruth S, Trott KR. The effects of dose-fractionation on radiation-induced heart disease in rats. Radiother Oncol 1987;8:363.
188. Withers HR, Thames HD Jr, Peters LJ. Biological bases for high RBE values for late effects of neutron irradiation. Int J Radiat Oncol Biol Phys 1982;8:2071.
189. Ang KK, Jiang GL, Guttenberger R, et al. Impact of spinal cord repair kinetics on the practice of altered fractionation schedules. Radiother Oncol 1992;25:287.
190. Ellis F. Nominal standard dose and the ret. Br J Radiol 1971;44:101.
191. Withers HR, Thames HD. Dose fractionation and volume effects in normal tissues and tumors. Am J Clin Oncol 1988;11:313.
192. Andrews JR. Dose-time relationships in cancer radiotherapy. A clinical radiobiology study of extremes of dose and time. Am J Roentgenol Radium Ther Nucl Med 1965;93:6.
193. Cosset JM, Henry-Amar M, Girinski T, et al. Late toxicity of radiotherapy in Hodgkin’s disease. The role of fraction size. Acta Oncol 1988;27:123.
194. Maciejewski B, Withers HR, Taylor JM, et al. Dose fractionation and regeneration in radiotherapy for cancer of the oral cavity and oropharynx. Part 2. Normal tissue responses: acute and late effects. Int J Radiat Oncol Biol Phys 1990;18:101.
195. Horiot JC, Le Fur R, N’Guyen T, et al. Hyperfractionation versus conventional fractionation in oropharyngeal carcinoma: final analysis of a randomized trial of the EORTC cooperative group of radiotherapy. Radiother Oncol1992;25:231.
196. Ang KK, Price RE, Stephens LC, et al. The tolerance of primate spinal cord to re-irradiation. Int J Radiat Oncol Biol Phys 1993;25:459.
197. Saunders MI, Dische S, Grosch EJ, et al. Experience with CHART. Int J Radiat Oncol Biol Phys 1991;21:871.
198. Van der Kogel AJ. Radiation tolerance of the rat spinal cord: time-dose relationships. Radiology 1977;122:505.
199. Marples B, Lambin P, Skov KA, et al. Low dose hyper-radiosensitivity and increased radioresistance in mammalian cells. Int J Radiat Biol 1997;71:721.
200. Fowler JF. The eighteenth Douglas Lea lecture. 40 years of radiobiology: its impact on radiotherapy. Phys Med Biol 1984;29:97.
201. Down JD, Boudewijn A, van Os R, et al. Variations in radiation sensitivity and repair among different hematopoietic stem cell subsets following fractionated irradiation. Blood 1995;86:122.
202. Bourhis J, Overgaard J, Audry H, et al. Hyperfractionated or accelerated radiotherapy in head and neck cancer: a meta-analysis. Lancet 2006;368:843.
203. Leksell L. Stereotactic radiosurgery. J Neurol Neurosurg Psych 1983;46:797.
204. Hazard LJ, Jensen RL, Shrieve DC. Role of stereotactic radiosurgery in the treatment of brain metastases. Am J Clin Oncol 2005;28:403.
205. Fakiris AJ, McGarry RC, Yiannoutsos CT, et al. Stereotactic body radiation therapy for early-stage non-small-cell lung carcinoma: four-year results of a prospective phase II study. Int J Radiat Oncol Biol Phys 2009;75:677.
206. Martinez AA, Demanes J, Vargas C, et al. High-dose-rate prostate brachytherapy: an excellent accelerated-hypofractionated treatment for favorable prostate cancer. Am J Clin Oncol 2010;33:481.
207. Fowler JF, Tome WA, Fenwick JD, et al. A challenge to traditional radiation oncology. Int J Radiat Oncol Biol Phys 2004;60:1241.
208. Verellen D, Vanhavere F. Risk assessment of radiation-induced malignancies based on whole-body equivalent dose estimates for IMRT treatment in the head and neck region. Radiother Oncol1999;53:199.
209. Demaria S, Formenti SC. Sensors of ionizing radiation effects on the immunological microenvironment of cancer. Int J Radiat Biol 2007;83:819.
210. Liao YP, Wang CC, Schaue D, et al. Local irradiation of murine melanoma affects the development of tumour-specific immunity. Immunology 2009;128:e797.
211. Nguyen TD, Panis X, Froissart D, et al. Analysis of late complications after rapid hyperfractionated radiotherapy in advanced head and neck cancers. Int J Radiat Oncol Biol Phys 1988;14:23.
212. Down JD, Easton DF, Steel GG. Repair in the mouse lung during low dose-rate irradiation. Radiother Oncol 1986;6:29.
213. Travis EL, Peters LJ, McNeill J, et al. Effect of dose-rate on total body irradiation: lethality and pathologic findings. Radiother Oncol 1985;4:341.
214. Niemierko A. Reporting and analyzing dose distributions: a concept of equivalent uniform dose. Med Phys 1997;24:103.
215. Withers HR. Neutrons and other clinical trials: impossible dreams? Int J Radiat Oncol Biol Phys 1987;13:1967.