State of the Art ReviewRadiation Oncology

# New Paradigms and Future Challenges in Radiation Oncology: An Update of Biological Targets and Technology

See allHide authors and affiliations

Science Translational Medicine  20 Feb 2013:
Vol. 5, Issue 173, pp. 173sr2
DOI: 10.1126/scitranslmed.3005148

## Abstract

Radiation oncology exploits the biological interaction of radiation within tissue to promote tumor death while minimizing damage to surrounding normal tissue. The clinical delivery of radiation relies on principles of radiation physics that define how radiation energy is deposited in the body, as well as technology that facilitates accurate tumor targeting. This review will summarize the current landscape of recent biological and technological advances in radiation oncology, describe the challenges that exist, and offer potential avenues for improvement.

## Introduction

The field of radiation oncology—where ionizing radiation is used to treat a variety of cancers as well as benign conditions—was born shortly after the discovery of x-rays and their effects on tissue in 1895 (1). At first, treatments were typically delivered in single doses using low-energy cathode ray tubes or radium-filled glass tubes positioned close to tumors. Technological developments between 1920 and 1945 focused on improving beam output and energy. However, the low energy (200 to 500 kV) of x-rays used in that period was associated with skin toxicity because of poor penetration, thereby limiting the use of radiotherapy for deep tumors. The development of cobalt-60 units and linear accelerators in the 1950s to deliver “supervoltage” radiation energies (≥1 MeV) was a critical advance because these high-energy x-rays could penetrate further to reach deeper seated tumors. Fractionated therapy, which dates back to the 1920s, divides treatment into multiple small doses rather than one large radiation dose and has allowed for further improvement in tolerance of normal tissues to treatment. Advances in technology, imaging, and cancer biology over subsequent decades have pushed the field of radiation oncology closer toward the idealized goal of noninvasively achieving maximal local cancer control with minimal normal tissue toxicity.

## Basic Principles and Current Uses of Radiotherapy

Cancer may grow locally or spread systemically through lymphatic or hematogenous routes. Successful treatment requires therapy targeted toward all sites of involvement. The three major modalities of cancer therapy—surgery, radiation therapy (RT), and chemotherapy—can be used alone or in combination to address all sites at risk for harboring disease. In a curative-intent approach, surgery and/or RT are generally used to address local-regional areas of risk. Surgery remains the most commonly used modality to treat local disease. By its nature, surgery can be both therapeutic and diagnostic because tumor excision expeditiously provides tissue for histologic examination and staging. RT as a sole modality can sometimes offer a noninvasive alternative to the therapeutic role of surgery, with the possibility for organ preservation, such as with bladder and laryngeal cancer. As an adjuvant therapy, RT can facilitate resection when given before surgery, or treat microscopic residual disease when given after surgery, such as treatment after breast-conserving lumpectomy. On the other hand, chemotherapy is given to treat known metastatic disease or as an adjuvant to reduce the risk of potential micrometastasis. Chemotherapy is often also combined with RT to act as a radiosensitizer for the purpose of increasing local control. The optimal use of each modality of cancer therapy is tailored according to the cancer cell histology, anatomic location, stage of cancer, and other patient factors. Some common diseases that serve as examples of the role of radiotherapy in integrated multimodality cancer treatment are summarized in Table 1. It is crucial for treatment approaches to consider quality of life in survivorship because each modality carries a different set of risks that need to be balanced against one another to provide the optimal risk-benefit ratio for each individual patient.

Table 1

Multidisciplinary management of several common cancers treated with curative intent.

View this table:

Empiric clinical observations in the early 20th century demonstrated that daily radiation exposures can induce death in cancerous cells while allowing for normal tissue recovery, provided that the daily doses [expressed in units of gray (Gy)] are relatively small. This observation formed the basis for fractionated RT. Many mechanistic explanations for this effect have been proposed by radiobiologists. The ability of normal tissue to repopulate itself with healthy cells clearly represents one important component of normal tissue tolerance of fractionated radiotherapy. A common course of RT can average 6 to 8 weeks of treatment, with five to six daily treatments per week.

At the tissue level, the impact of RT on tissue function with increasing radiation dose can be graphically represented by a sigmoid-shaped curve (Fig. 1). The sigmoidal curve that describes tumor control probability is situated to the left of the sigmoidal curve that describes normal tissue complication probability. The degree of separation between these curves defines a therapeutic window, in which the dose of radiation is predicted to eradicate tumor while maintaining normal tissue tolerance. The sigmoidal relationship between dose and response implies that for any given tissue, there is a dose threshold above or below which incremental changes in dose yield little additional impact. However, within a critical range on the steep portion of the curve (such as cumulative doses of 40 to 100 Gy when given with “conventional” fractionation of 1.8 to 2.0 Gy per fraction), small increases in dose may result in large increases in clinical impact. Dose prescriptions in RT are determined with consideration to the unique relationship between the dose-response curves for the specific tumor and surrounding normal tissues, which vary widely for each clinical circumstance. The figure is an oversimplification because families of curves may exist for the tumor clones comprising a tumor and for the complex components of normal tissues intertwined with the tumor.

Standard chemotherapeutic agents are the most common agents used for increasing the local efficacy of radiotherapy; however, this review will focus on more recent strategies of identifying targeted radiosensitizers, particularly those that have been tested in the clinic. It is important to remember that inhibition of specific proteins can frequently generate sensitization of cells in culture, but ultimately, a drug must improve the therapeutic index of radiation to be clinically useful (see illustration of this concept in Fig. 1).

The molecular pathophysiology of radiotherapy: Sensors, transducers, and effectors of DNA damage. One gray generates about 105 ionization events per cell, producing about 1000 to 2000 single-strand DNA breaks (SSBs) and 40 double-strand DNA breaks (DSBs) per nucleus. A large and growing list of non-DNA repair/checkpoint–related mechanisms contributes to cellular responses to radiation. Although ionizing radiation is known to generate DNA base damage, SSBs, and DSBs, the DSBs are generally thought to represent the principal lethal events and the most critical lesions to radiotherapy. DSBs initiate a complex set of cellular responses including DNA damage recognition and transduction of the signal, resulting in many downstream effects including cell cycle checkpoint activation, induction and coordination of stress response genes, DNA repair, and/or activation of the apoptotic cascade (Fig. 2).

Targeting of DSB response and repair. The MRN protein complex (Mre11, Rad50, Nbs1), a principal sensor of DNA damage, accumulates at DSBs very rapidly after radiation and participates in activating ataxia-telangiectasia mutated (ATM) protein. Although ATM and ATR (ATM- and RAD3-related) perform partially overlapping functions, ATM preferentially recognizes DSBs, whereas ATR preferentially senses replication-blocking DNA lesions. Both ATM and ATR phosphorylate downstream targets that regulate cell cycle checkpoints and apoptosis, as well as other forms of cellular responses like senescence, autophagy, and DNA repair. One centrally important phosphorylation target is the chromatin protein histone H2AX because chromatin structure subsequently becomes less condensed and allows for the recruitment of repair proteins. Because H2AX phosphorylation is easily detectable with fluorescent microscopic methods, it has become a common marker for DSB induction and resolution (repair) in both experimental and clinical settings. After DNA damage recognition, phosphorylated H2AX promotes the recruitment of other sensor/effector proteins including 53BP1, MDC1, and BRCA1, which regulate the processing of damaged DNA ends in preparation for repair. This process is regulated by several ubiquitin and SUMO ligases (PIAS1, PIAS4, RNF4, and RNF8) and the PSMD4 proteasome.

Homologous recombination (HR) and nonhomologous end-joining (NHEJ) repair pathways are the two pathways that contribute to the repair of DSBs. Both repair pathways occur after the recruitment of the sensor/effector proteins discussed above. HR involves the identification of a stretch of homologous DNA and replication of the missing genetic information from this homologous DNA template (2). Alternatively, the NHEJ pathway processes the broken DNA ends and religates them, frequently by making use of a region of microhomology (3). These pathways appear to have somewhat overlapping and complementary roles.

Many components of the DSB repair pathways have been investigated as therapeutic targets, and some chemical inhibitors might be considered as possible lead compounds in oncology drug development (Fig. 3). For example, mirin is a chemical inhibitor of the MRN complex (5). Consistent with the upstream role of MRN in sensing and signaling DNA damage, mirin generates a broad range of cellular effects, including inhibition of ATM activation, loss of G2-M cell cycle checkpoint, down-regulation of HR, and down-regulation of NHEJ repair efficiency (5, 6). This lack of cancer cell specificity may limit the utility of mirin in the clinic. Similar upstream signaling functions have been targeted with inhibitors of the family of phosphatidylinositol 3-kinase (PI3K)–related kinases (PIKK), which include ATM, ATR, mammalian target of rapamycin (mTOR), human suppressor of morphogenesis in genitalia-1 (hSMG-1), DNA-PKcs, and transformation/transcription domain–associated protein (TRRAP) (7). Several broad-spectrum inhibitors of PIKKs have been developed, including wortmannin or LY294002; however, at least some have toxicities that limit their utility in the clinical setting. More specific inhibitors of ATM and/or ATR include KU-55933, CGK733, NU6027, and CP466722. KU-55933 has been the focus of particular attention because it is a specific adenosine triphosphate–competitive inhibitor of ATM that is capable of sensitizing cells to radiation and several chemotherapeutic drugs, including etoposide, doxorubicin, and camptothecin (8).

Several inhibitory compounds have also been developed to modulate specific DSB repair pathways. HR has been inhibited by specifically targeting RAD51 protein or RAD51 paralogs [see Budke et. al. and references therein (9)]. Other drugs have also been shown to lower HR efficiency by nonspecifically reducing RAD51 protein levels. These strategies are promising because RAD51 protein is highly expressed in many human cancers (10, 11) and because HR inhibition has been shown to promote preferential sensitization of tumor cells relative to normal cells (12, 13). These observations suggest that human tumors may develop “addictions” to abnormally high RAD51 levels that can be exploited pharmacologically. Likewise, compounds have been developed to target components of the NHEJ repair. Again, this can be accomplished by blocking PIKKs. However, more targeted NHEJ inhibitors have also been developed, including several different DNA-PK inhibitors that include NU7441, Vanillin, SU11752, IC87102, IC87361, NU7026, CC-115, and Salvicine.

Relatively few of these inhibitors of DSB repair have transitioned into clinical trials, perhaps because of a lack of commercial interest and the recognition of radiotherapy as fertile ground for pharmaceutical development. One interesting new strategy that is being developed clinically by DNA Therapeutics involves small DNA molecules that act as DNA bait. One such agent, termed Dbait or DT01, mimics DNA double-strand breaks and acts to disorganize damage signaling and DNA repair (14). DT01 is currently being studied in a phase 1 trial for patients with metastatic melanoma.

Poly(adenosine diphosphate–ribose) polymerase inhibitors. Poly[adenosine diphosphate (ADP)–ribose] polymerase inhibitors (PARPi) are currently being studied in numerous clinical trials with chemotherapy and a few trials with radiotherapy. The PARP family of proteins is defined by their capacity to modify target proteins by the covalent addition of poly(ADP-ribose) polymers. PARP1 is the most abundant of this protein family, and it accounts for about 80% of PARP activity in cells. PARP1 and PARP2 have DNA binding domains, and their catalytic function is activated when they bind sites of DNA damage. When PARP1 becomes activated, it generates long and branching poly(ADP-ribose) chains on histones and other proteins located near DNA breaks. These polymer scaffolds are important in recruiting other DNA repair proteins (for example, the base excision repair protein XRCC1) to the break site. PARPi compounds have generated intense interest by the oncology community since 2005, following the demonstration of synthetic lethality in BRCA-defective cells (15, 16). Synthetic lethality is a concept whereby a tumor is defective in one survival pathway, and inhibition of an escape pathway is an effective cytotoxic strategy. A common feature of these drugs is the exquisite hypersensitivity observed with HR-defective tumor cells, including triple-negative breast cancers, which exhibit epigenetic deregulation of HR.

Several studies have demonstrated that PARPi compounds can radiosensitize tumors in preclinical models (1719). Ionizing radiation, as discussed earlier, induces both SSBs and DSBs. Although the PARPi-mediated mechanism of radiosensitization remains somewhat unclear, PARP inhibitors are known to sensitize cells to agents that generate SSBs (20). Chalmers and colleagues have pointed out two key aspects of PARPi effects pertinent to this issue: (i) radiosensitization occurs primarily in replicating cells and (ii) PARPi compounds delay rather than abolish SSB repair (18). These factors suggest that components of replication machinery may collide with unrepaired PARP-bound SSB lesions, thereby generating more toxic lesions than the starting SSBs. Other possible mechanisms of PARPi-mediated radiosensitization may include reoxygenation of hypoxic tumors; PARPi compounds have demonstrated vasoactive properties and could potentially counteract the radioresistance associated with hypoxia. Finally, recent studies have shown that chronic hypoxia in tumors can generate a reduction in HR protein expression and function. Hence, hypoxia within tumors may generate specific anatomic compartments that behave like BRCA-defective tumors, wherein a contextual synthetic lethality occurs for PARPi (21).

Several clinical trials are presently evaluating combinations of PARPi drugs plus radiotherapy, either with or without chemotherapeutic drugs for diseases including rectal, brain, and breast cancers (http://clinicaltrials.gov). Again, it is unknown whether these agents will improve the therapeutic index of radiotherapy, and specific disease site trials will be necessary to determine potential efficacy.

Histone deacetylase inhibitors. Histone deacetylases (HDACs) represent a family of at least 18 enzymes that remove acetyl groups from lysine residues in core histone proteins. Deacetylation exposes the positive charges on histones, thereby increasing their interaction with the negatively charged phosphate backbone of DNA. This results in more compacted chromatin that is relatively inaccessible to modulation by transcription factors. This pathway has relevance to radiation biology, given that HDAC inhibitors stimulate radiation-induced cell cycle arrest, apoptosis, and DSB formation (22, 23). The mechanism is unclear, but studies suggest that HDAC inhibitors suppress DNA repair efficiency.

Traditional HDAC-inhibitory drugs, like valproic acid, have a long history as mood stabilizers and antiepileptics. Several second-generation inhibitors have more recently been developed, including vorinostat, belinostat, and panobinostat. Ongoing clinical trials are evaluating these drugs plus radiotherapy in many tumor types and in a wide range of clinical settings, including fractionated radiotherapy, stereotactic radiosurgery (SRS), and together with an infused [131I]MIBG radiopharmaceutical. One recently completed phase 1 study showed that vorinostat (at 300 mg once daily) was tolerable in combination with a palliative radiotherapy course to the pelvis, consisting of 30 Gy over 2 weeks (24). Unfortunately, efficacy results from phase 2 trials are not available.

Cell cycle arrest after radiation. It has been known for decades that mammalian cells arrest in G1 and G2 after radiation exposure. The pathways that govern cell cycle checkpoints were first discovered in yeast by Hartwell, Nurse, and Hunt, for which they received the 2001 Nobel Prize. Briefly, radiation-induced activation of ATM and ATR leads to phosphorylation of the downstream effector kinases Chk1 and Chk2. These in turn phosphorylate the phosphatase CDC25A, which in turn becomes degraded and thus unable to dephosphorylate and activate CDK2; the net result is cell cycle arrest. Additionally, Chk2 phosphorylation of p53 activates expression of p21, which also induces cell cycle arrest.

Targeted inhibition of the cell cycle checkpoint machinery has been explored as a method to sensitize cells to DNA damage. Some of these strategies have been based on the tendency of tumor cells to have abnormal G1 checkpoint functions but intact G2 checkpoint mechanisms. Targeted inhibition of G2 checkpoints can promote progression of damaged cells to mitotic catastrophe, and this effect may be tumor-specific because normal tissues are preferentially protected by intact G1 checkpoints. This may also potentially counteract the radiation resistance that has been observed in some cancer stem cells, such as CD133-positive glioma stem cells (25).

A number of chemical inhibitors have been developed to target Chk1 and/or Chk2, including UCN-01, AZD7762 (AstraZeneca), XL844 (EXEL-9844, Exelixis), LY2606368 (Eli Lilly), and PF-00477736 (Pfizer). All of these compounds have been tested in early-phase clinical trials against solid tumors. Many of these are being tested in combination with various chemotherapeutic drugs. However, it remains an open question as to whether these drugs can be safely combined with radiotherapy, and whether they can improve the therapeutic index.

Modifiers of cellular death after radiation. Cells subsequently face a critical period, particularly if radiation damage is not completely repaired (26). These critically injured cells may progress unrepaired to mitotic catastrophe or simply lose proliferative capacity by undergoing senescence. Radiation-induced senescence is commonly associated with activation of the p16/RB and p53/p21 tumor suppressor pathways. Alternatively, cells may progress to apoptotic death, which classically involves p53-dependent activation of the caspase cascade. Apoptotic death can also occur after activation of cell surface “death receptors” [receptors of tumor necrosis factor (TNF), Fas, or TRAIL (TNF-related apoptosis-inducing ligand)], which themselves can be up-regulated by radiation. Death receptors can also generate cytoprotective effects by activating transcription factor nuclear factor κB (NFκB), which increases expression of genes that promote proliferation or oppose apoptosis.

Pathways of intracellular signaling after radiation. Most of the radiation-induced reactive oxygen species (ROS) interact with cellular contents other than DNA. ROS such as superoxide and hydroxyl radicals are also known to deplete cellular stores of antioxidants like glutathione (27). The resulting cellular stresses stimulate a complex set of signaling cascades [reviewed in detail by Dent and colleagues (28)].

Radiation can also activate signaling pathways that cells normally use to respond to mitogens, which in turn promote survival, antiapoptotic responses, and transcriptional changes. The net effect can be variable and cell-specific; however, a common theme includes activation of cell surface receptors, like the ErbB family that includes epidermal growth factor receptor (EGFR). Receptor activation subsequently signals downstream pathways, including the mitogen-activated protein kinase (MAPK) superfamily of cascades [extracellular signal–regulated kinase (ERK), c-Jun N-terminal kinase (JNK), and p38] and the PI3K pathways. These pathways deliver antiapoptotic signals via Akt and ERK signaling. Radiation can also activate these pathways via autocrine mechanisms, like through the production of transforming growth factor–α (TGF-α), which binds and activates EGFR (28). Radiation also activates the proinflammatory cytokines including TNF-α and interleukin-6 (IL-6) (29), which in part may account for bystander effect (discussed below).

Finally, radiation-induced damage of the plasma membrane induces the breakdown of sphingomyelin to ceramide, which is proapoptotic independent of DNA damage (30). Radiation also activates cytosolic phospholipase A2 (cPLA2), an enzyme that recognizes phospholipids on the cell membrane and degrades them into inflammatory products like arachidonic acid and eventually eicosanoids. Lysophosphatidylcholine is one such product formed by cPLA2, and its production leads to activation of Akt and enhanced cell death (31).

One interesting strategy for radiosensitization has been the blockade of progrowth signaling from receptor tyrosine kinases, such as insulin-like growth factor 1 receptor (IGF-1R) and EGFR. Numerous antibodies and chemicals have been developed to target different levels of this signaling cascade. One successful pharmacologic effort was the inhibition of EGFR during radiotherapy for head and neck cancers, using the chimeric (mouse/human) monoclonal antibody cetuximab. A randomized trial demonstrated benefit of weekly cetuximab in addition to 6 to 7 weeks of radiotherapy. The updated data show a 5-year overall survival of 45% with cetuximab/radiotherapy and 36% with radiotherapy alone (32). Many related strategies are being evaluated in this and other anatomic tumor sites, using inhibitory antibodies or small molecule–based tyrosine kinase inhibitors.

Bystander effects and importance of tumor microenvironment. Nonirradiated cells often exhibit stress responses, after even low-dose (<0.1 Gy) radiation exposures to neighboring cells. This “bystander effect” has been well described in detail previously (33, 34). Bystander responses appear to be cell type–specific; in general, they consist of a broad range of effects including gene induction, genomic instability, differentiation, and changes in apoptotic potential. These processes are mediated, at least in part, by diffusible substances, given that the effects occur when bystander cells are physically separated from the irradiated cells. Some authors have additionally implicated cell-cell contacts (gap junctions) as contributing to this effect.

This illustrates an important concept—that one must consider radiation effects in the context of the entire tumor microenvironment, rather than simply the sensitivities of individual cancer cells. This notion is further demonstrated by the contribution of host stromal components within tumors to radiation responsiveness. Studies in mice, for example, show that host-derived blood vessels are a key determinant of tumor control with radiotherapy (35).

Impact of cancer stem cells (tumor-initiating cells) within irradiated tumors and normal tissues. Although many researchers disagree over various aspects of cancer stem cell biology and nomenclature, there clearly exists a population of tumor cells that exhibit an exclusive ability for self-renewal and differentiation into the heterogeneous lineages that promote tumor maintenance. These concepts have been supported by three recent important papers demonstrating the existence of cancer stem cells in mouse models of brain, skin, and intestinal tumors (3638). Results of these studies indicate that targeting cancer stem cells may improve therapeutic outcomes. In response to fractionated radiation, cell populations become enriched for cells expressing putative markers of “stemness.” Specific phenotypes vary based on tumor type and methodology, but common features of these cells after radiation often include increased survival, reduced apoptosis, and rapid resolution of H2AX foci (that is, fast repair of DSBs). For example, one study demonstrated an enrichment of glial cancer stem cells after radiation, and that these cells exhibited preferential activation of the DNA damage checkpoint responses and increases in DNA repair capacity (25).

The basic mechanisms of radiation resistance remain unclear in stem cells; however, aldehyde dehydrogenase 1 (ALDH1) may represent a partial explanation. High expression of ALDH1 protein in tumors is an adverse prognostic factor, perhaps because its aldehyde-catabolizing activity confers a survival advantage. Also, aldehyde-catabolizing enzymes are known to cooperate with Fanconi anemia genes, suggesting that ALDH activity might play a role in repairing adducts on DNA (39). Hence, because ALDH1 appears to be preferentially expressed in cancer stem cells, this may represent a therapeutic target to reverse radioresistance in stem cell clones. Another interesting feature of these cells is their propensity to repair DSBs using HR repair. Breast cancer–derived stem cells (CD24, ESA+) were “effectively sterilized” by inhibition of the HR pathway, whereas nonsorted cells from the same cell line (MDA-MB231) were unaffected (40). This finding suggests that HR inhibitors represent a promising therapeutic strategy for depleting cancer stem cell reservoirs in tumors. Also, screening of small-molecule libraries led to the identification of salinomycin and thioridazine, both of which appear to target cancer stem cells (41, 42) and may be used to improve radiotherapy.

Normal tissues also contain stem cell niches that likely influence their ability to tolerate radiotherapy. Thomas Helleday’s group recently reported a study, in which punch biopsies were taken from normal skin during the first and last weeks of a 5-week clinical radiotherapy course (43). The week 5 skin biopsies showed a significant enrichment for proliferating cells that were undergoing HR (Ki67+ and RAD51 focus+), as well as epidermal stem cells (β1 integrin+ and Ki67). These results suggest that tumor tissue and normal tissues share some features in terms of stem cell responses, which may pose challenges to targeting cancer stem cells. In some situations, the anatomic locations of normal stem cell niches are known and can perhaps be spared from radiation exposure. One example is the hippocampus, in which neural stem cells are known to reside. An ongoing trial by the Radiation Therapy Oncology Group (RTOG) is prospectively testing whether reduced exposure to the hippocampus can decrease neurocognitive toxicity associated with whole-brain radiotherapy. Another example comes from stem cell niches located within the walls of larger ducts in salivary glands (44), which could potentially be excluded from radiotherapy treatment target volumes.

Targeting tumor hypoxia and redox conditions. The investigation of tumor hypoxia as an effector and biomarker of tumor resistance has exploded in the past decade, and modalities that exploit tumor hypoxia have been widely tested. Tumors are known to outgrow their blood supply, thereby generating regions of necrosis that are surrounded by areas of hypoxia. Hypoxia promotes activation of the hypoxia-inducible transcription factor (HIF) family of proteins that regulate a variety of downstream genes that promote angiogenesis, cell survival, anaerobic energy metabolism, and treatment resistance (45). Hypoxia can also select for highly aggressive tumor cell clones. This topic is particularly relevant to radiotherapy because the presence of molecular oxygen during delivery of ionizing radiation enhances radiation-induced cell kill by 2.5- to 3.5-fold. The commonly accepted mechanism for this observation is that oxygen “fixes” free radical–induced DNA damage into a permanent state. Hence, hypoxic regions of tumor are generally considered radiation-resistant, and reversal of hypoxia has long been a goal for radiosensitization.

A newer class of radiation sensitizers has been developed to target hypoxia on the basis of reduction/oxidation conditions that predominate in anaerobic environments. Tirapazamine and porfiromycin are hypoxic cell cytotoxins/sensitizers, which act as bioreductive alkylating agents (53). Tirapazamine differs from oxygen mimetics in that it does not “fix” radiation damage, but instead, it is metabolized into a highly reactive radical species in anaerobic conditions. Preclinical data regarding tirapazamine were extremely promising; however, several phase 3 trials failed to demonstrate significant clinical benefit in combination with radiotherapy (54, 55). This disappointing outcome may reflect poor penetration of tirapazamine into hypoxic tumor regions. In an interesting related development, RAD51 and other HR-related DNA repair proteins have been shown to be transcriptionally down-regulated in response to chronic hypoxia (56). Contrary to traditional concepts of hypoxia, this down-regulation of DNA repair may actually render cells more sensitive to radiation.

Related efforts have used porphyrin-like macrocycles that form complexes with large metal cations and participate with the cellular redox cycle. Motexafin gadolinium (also known as MGd) is one such drug that inhibits antioxidant proteins such as thioredoxin reductase, and it exhibits preferential localization to tumor over normal tissue (45). In combination with whole-brain radiotherapy, MGd provided modest improvements for patients with brain metastases from lung cancer (57); however, these effects have not been impressive enough for Xcytrin to gain U.S. Food and Drug Administration (FDA) approval. MGd is also being evaluated in ongoing or completed clinical trials involving radiotherapy for other central nervous system tumors (glioblastoma multiforme and pediatric brainstem gliomas), as well as various carcinomas including pancreaticobiliary, non–small cell lung, and head and neck cancers.

Pharmacologic approaches are being developed to target cellular signaling that occurs in response to hypoxia and redox status in tumors. Many of these have focused on the transcription factor HIF-1, using agents that modulate its transcription, stability, association with binding partners, or signal transduction (45). EZN-2968 (Enzon Pharmaceuticals), an antisense oligonucleotide to HIF-1α, demonstrated safety and potential activity in phase 1 testing; however, data from subsequent testing are still awaited.

A somewhat related strategy arose from a small interfering RNA screen, looking for tumor-selective radiosensitizing targets in head and neck cancers. Surprisingly, knockdown of uroporphyrinogen decarboxylase, a regulator of heme synthesis, sensitized head and neck cancer cells to radiation and some cytotoxic agents such as cisplatin, paclitaxel, and 5-fluorouracil (5-FU). This may occur due to alterations in iron homeostasis, which promotes an increased production of ROS (58). The role of hypoxia in tumor killing by radiotherapy remains to be resolved. Perhaps, biomarkers or noninvasive physical measures of hypoxia will determine which patients might benefit from modification of hypoxia.

Antiangiogenic drugs and radiotherapy. New blood vessel development is required for tumors to grow beyond 1 to 2 mm. Vascular endothelial growth factor (VEGF) is a key proangiogenic growth factor that is secreted by solid tumors and acts through one of three VEGF receptors (VEGFRs). The most widely studied VEGF inhibitor is a humanized monoclonal antibody (bevacizumab) that binds to and inhibits the activity of human VEGF. Preclinical data have demonstrated that blockade of VEGF signaling increases the antitumor effects of radiation. Additional studies have suggested that VEGF protein levels can become up-regulated in tumors in response to ionizing radiation, suggesting that VEGF might mediate the development of tumor endothelial cell radioresistance (59, 60). The therapeutic combination of radiotherapy plus VEGF inhibition could be considered counterintuitive because antagonism of tumor vasculature might be expected to increase tumor hypoxia. However, tumor angiogenesis has proven to be a dysregulated process that generates networks of tortuous and hyperpermeable vessels, resulting in spatial heterogeneity in tumor oxygenation and elevated interstitial fluid pressure. Current research in preclinical models has demonstrated that VEGF blockade can “normalize” the tumor vasculature, thereby reducing tumor hypoxia and interstitial pressure and improving the metabolic profile of the tumor microenvironment (61).

Gene therapy and radiotherapy. There has been extensive preclinical work over the past 15 years on combining gene therapy with radiotherapy. Replication-defective viruses have been used to deliver radiosensitizing prodrugs, cytokines, tumor suppressor genes, and immune-activating compounds. Other related strategies have used replication-competent viruses to generate oncolytic antitumor effects. The following section will focus on gene therapy modalities that have been tested with radiotherapy in clinical trials.

Other trials have used enzyme/prodrug strategies. For example, virus-directed expression of herpes simplex virus thymidine kinase (HSV-tk) can phosphorylate the prodrug ganciclovir into a toxic metabolite that interferes with DNA replication, leading to chain terminations and SSBs. A phase 3 randomized trial was unable to show a survival benefit with this strategy, when it was combined with surgery and radiation for glioblastoma multiforme (69), although the agent was delivered after therapy. Related strategies involve the use of an additional “suicide gene” to viral vectors. Freytag and colleagues have tested replication-competent adenovirus that carries both HSV-tk and cytosine deaminase, which converts the prodrug 5-fluorocytosine into 5-FU (70). This gene therapy strategy with prostate radiotherapy appears to be tolerable, but efficacy remains to be demonstrated in phase 2 to 3 trials.

Oncolytic viruses have also been combined with radiotherapy, based on interesting preclinical evidence that some viruses can cooperate with radiotherapy by preferentially infecting and lysing cancer cells. Replication-competent HSV-1 has been genetically modified to allow safe use with radiotherapy. For example, a modified HSV-1, called G207, has been safely administered via intratumoral injection immediately before radiotherapy in recurrent or progressive malignant gliomas. More recently, a genetically modified HSV-1 encoding granulocyte macrophage colony-stimulating factor was combined with chemoradiotherapy in a phase 1/2 trial for locally advanced head and neck cancer (71). Also, intratumoral injections of a reovirus were recently combined with palliative radiotherapy in a phase 1 trial (72).

Immune modulation and radiotherapy. Preclinical evidence shows that radiation can stimulate antitumor responses by the immune system, which has important potential implications clinically. Radiation generates an inflammatory microenvironment in tumors, whereby damaged cells increase antigen presentation and immune recognition. More specifically, radiation-induced immune effects include an elevation of major histocompatibility complex class I (MHC-I) expression, changes in antigenic peptide repertoire, and decreases in regulatory T cells. Radiation also up-regulates cytokines and adhesion molecules that recruit and activate CD8+ cytotoxic T lymphocytes and dendritic cells. The evolving understanding of these processes may allow for them to be exploited therapeutically (7375).

National Cancer Institute researchers reported a randomized phase 2 trial that treated prostate cancer with standard radiotherapy, with or without a gene therapy–based vaccine that encodes prostate-specific antigen (PSA) (76). The therapy consisted of a “priming” vaccination with recombinant vaccinia (rV) PSA plus rV containing a T cell costimulatory molecule, followed by monthly booster vaccines with recombinant fowlpox PSA. Most of the vaccinated patients did successfully develop increases in PSA-targeted T cells. Other clinical trials have explored related strategies of radiotherapy in combination with different vaccines, intratumoral injection of immature dendritic cells, or adoptive immunotherapy via infusion with expanded tumor-infiltrating lymphocytes [reviewed by Kamrava and colleagues (73)]. Many of these strategies have demonstrated an impressive induction of immune responses; however, it remains unclear whether these effects will translate to improved clinical outcomes. A recent phase 1 trial reported that patients with metastatic melanoma or kidney cancer treated with high-dose IL-2 demonstrated impressive tumor responses when one to three tumors were treated with stereotactic body radiotherapy compared with historical controls treated with high-dose IL-2 alone (77).

### Molecular predictors of outcome in radiotherapy

In the past decade, we have witnessed an explosion in new diagnostic technologies capable of subclassifying malignancies and predicting clinical outcomes, based on genetic and epigenetic tumor characteristics. Such studies have fueled a vision in the oncology community of “personalized medicine,” a system by which individualized treatment courses can be tailored to specific patient and tumor characteristics. These studies represent an expansive work that cannot be adequately encompassed by this review, and most of these studies have focused on predictors of overall prognosis.

Many studies have reported gene expression signatures that predict overall tumor behavior for different cancer types, and some of these signatures have been validated with clinical data. However, few have attempted to use expression array technology to predict radiocurability per se. One such gene expression signature, called the radiosensitivity index (RSI), consists of 10 genes (AR, cJun, STAT1, PKC, RelA, cABL, SUMO1, CDK1, HDAC1, and IRF1) that associate with radiosensitivity within a collection of human cancer cell lines (97). This signature has been clinically validated in five independent clinical data sets, which include rectal, esophageal, head and neck, and breast cancers treated with chemoradiation or radiotherapy alone (98). A related approach identified a chemotherapy/radiation resistance signature, using tumors generated in mice. This interferon-related gene signature analysis was evaluated retrospectively in clinical breast cancer data sets, and it successfully predicted the efficacy of adjuvant chemotherapy and local-regional control after radiation (99).

An alternative approach is to look for associations between radiosensitivity and genetic alterations, using methods like comparative genomic hybridization (CGH) arrays or next-generation sequencing. CGH and SNP (single-nucleotide polymorphism) arrays measure copy number variations and can quantify global levels of genomic instability. Ishkanian and co-workers have used CGH on prostate cancer biopsy specimens to identify deletions of chromosome regions that normally harbor DNA repair genes (including PARP1, ATM, DNA-PKcs, p53, Rb, and RAD17). Detection of these gene losses might help identify tumors susceptible to radiation or other specific treatments (100). Similar work showed NKX3.1 haploinsufficiency to predict risk for prostate cancer recurrence after radiotherapy, and it also predicted for recurrence after prostatectomy (101). These and other emerging methods will need to be validated more thoroughly in prospective clinical trials, but the findings are encouraging. Furthermore, next-generation sequencing of tumor genomes will presumably revolutionize these types of approaches as we move forward.

Although the excitement of these new methods has echoed throughout the entire oncology community, it remains unclear whether personalized medicine can live up to our collective expectations. These newer methods generally rely on “representative” tumor biopsies, which can underestimate tumor cell heterogeneity. A recent cautionary study demonstrated alarmingly high levels of intratumor heterogeneity in renal cell cancers, in terms of both gene expression levels and DNA sequence (102). More than 60% of all somatic mutations were not detectable across all regions of a given tumor. Hopefully, many of the “driver mutations” (which form the basis of many new targeted therapies) will prove to be better represented throughout tumors. These data suggest that our existing therapies, conventional chemotherapy drugs and radiotherapy, will continue to serve important roles.

Table 2

Modalities in RT. IMRT, intensity-modulated RT; SBRT, stereotactic body RT; SRS, stereotactic radiosurgery. Proton therapy illustration is courtesy of University of Florida Proton Therapy Institute.

View this table:

### Intensity-modulated RT

Standard radiation planning is performed with axial computed tomography (CT) imaging and three-dimensional (3D) planning software, in which multiple beams are arranged to pass through the target volume. With intensity-modulated RT (IMRT), the intensity of each beam is modulated to provide a tighter approximation of high-dose radiation to the intended target, a characteristic known as conformality. IMRT planning uses an iterative computer algorithm approach, which has input variables that specify dose coverage and normal tissue–sparing goals, and output that defines the shape of the multiple beam “segments” that are delivered. This high degree of conformality is especially useful for irregularly shaped targets or disease in close proximity to critical structures that are sensitive to high doses of radiation (103).

Clinical use of IMRT has become widespread and has enabled the escalation of dose and reduction of toxicity. IMRT has become fairly standard in the organ-sparing treatment of head and neck cancers with RT. The protection of parotid and submandibular salivary glands from radiation dose (sparing goal <26 Gy) has led to improvements in quality of life through a reduction in treatment-related xerostomia (104). IMRT improves the ability to treat irregularly shaped targets such as advanced nasopharynx tumors that can wrap around the brainstem or approximate the temporal lobes of the brain. Furthermore, because beam intensity can be modulated, IMRT can enable differential dosing within a singular volume to treat variable levels of disease risk. This technique of “dose painting” or “simultaneous integrated boost” planning offers the oncologist more latitude to individualize plans by tailoring dose to match the appropriate burden of disease (105). In tandem with concurrent chemotherapy (often including 5-FU or cisplatin) and altered fractionation (such as twice daily treatment with smaller fraction sizes), the therapeutic window can potentially be further widened (106).

Despite the many predicted advantages of IMRT, there are only a few instances in which IMRT has been proven to reduce morbidity compared to conventional RT in a randomized setting (107), and potential disadvantages exist. Related to the conformality of IMRT planning, accurate treatment setup including immobilization becomes more critical to avoid underdosing of the target volume. Target volumes that are subject to motion or deformation over the course of treatment require more advanced technology for daily target localization and, in some cases, may be more appropriately treated with a 3D plan. Additionally, IMRT requires more resources to plan and deliver treatment, and it results in significantly more low-dose exposure to tissues outside the target area because of the increased number of beam angles and cumulative “beam on” time required to deliver IMRT. There exists some concern that the exposure of more normal tissue to low doses of radiation might increase the risk of normal tissue complications, specifically radiation-induced malignancy years after RT (108). These concerns, along with additional costs, explain why IMRT has not entirely supplanted more traditional forms of RT for all disease sites.

### Image-guided RT, hypofractionated therapy, and ablative therapy (SRS, stereotactic body RT)

Although CT-based RT planning provides a more precise ability to target radiation to soft tissue areas, inaccuracies in the reproducibility of daily setup and organ motion (such as respiratory motion) force the radiation oncologist to treat a volume larger than the actual tumor. Advancements in technology have addressed both positional error and physiologic motion to allow this setup margin to be reduced. Linear accelerators are now commonly fitted with various on-board imaging devices, which can provide diagnostic quality kilovolt x-rays or cone beam CT. Soft tissue targets that are not easily visible on an x-ray can potentially be implanted with radio-opaque “fiducial” markers to better guide treatment delivery with pretreatment imaging (109). With the regular implementation of any of these imaging modalities, known as image-guided RT (IGRT), positioning setup error can be reduced, thereby enabling the use of small treatment volumes. Finally, commercially available software/hardware packages have been developed to account for physiologic motion such as respiratory motion.

Whereas the development of IMRT after 3D conformal RT could be considered evolutionary, in that IMRT was an incremental advancement in planning which arose from hardware (multileaf collimator) and software (inverse-planning) innovations, perhaps the implementation of SRS and SBRT could be considered more revolutionary. IMRT is typically used to achieve maximal conformality of the high-dose region, especially in cases where the target volume is irregularly shaped. On the other hand, SBRT typically prioritizes the steepness of the dose gradient between the target area and adjacent critical normal structures. The use of multiple beam angles, sometimes arranged with the treatment table rotated off axis, can help position this dose gradient in a favorable location. SBRT techniques allow for “hotspots” of increased dose within the target volume, which can significantly exceed the intended prescription dose; in certain cases, the hotspot within the target can be >50% the prescription dose. This dose inhomogeneity further contributes to the steepness of the dose gradient. Notably, IMRT and SBRT techniques are not mutually exclusive—IMRT techniques can be used to deliver more favorable and homogeneous dose distributions. However, a combined approach comes at the expense of added complexity of planning and delivery, and an increase in treatment time could compromise the reproducibility of setup because of intrafraction patient movement.

The success of hypofractionated, image-guided RT to date has challenged conventional paradigms in radiation oncology. SBRT and SRS shift priority toward tumor ablation rather than normal tissue preservation, along the same lines of a surgical approach. Hence, the emphasis of hypofractionated RT is increasingly focused on local rather than local-regional therapy. Clinical data to date have provided reason for cautious optimism, given the high rates of local control and reasonably low rates of severe toxicity that have been reported (118). However, the best use of SBRT remains to be defined. Perhaps most importantly, appropriate dose limits to normal tissues need to be redefined with clinical outcome data, given that the bulk of the existing dose-volume analyses of normal tissue complications is derived from the conventional fractionation era (119). It is important to note that only a small subset of tumors is amenable to SBRT. Factors such as tumor volume (115), method of daily localization, and proximity and nature of adjacent normal tissue (120) may play critical roles in determining whether SBRT or traditional fractionated radiotherapy techniques are more appropriate.

### Brachytherapy

Brachytherapy represents a modality of treatment that could potentially outperform external beam RT for particular clinical situations. By varying the strength of the radioactive sources and manipulating the location and exposure time, it is possible to achieve highly conformal and dose-intense radiation distributions within tumors while generating relatively low exposures to adjacent normal organs. Source placement can be interstitial, in which sources are placed directly into the target tissue. Alternatively, in intracavitary brachytherapy, sources are placed in a space next to the target tissue, such as a body cavity or a body lumen. Because of the direct visualization of sources within the treatment site, concerns with setup uncertainty and organ motion are minimal.

Despite the dosimetric advantages of brachytherapy, this modality of radiotherapy is not used as commonly as external beam radiation. This is because only a small subset of cases present with well-defined areas of risk, which can be accessed in a relatively noninvasive fashion. Given this practical restriction, cancers of the cervix, prostate, breast, and skin are the most common sites to be treated with brachytherapy. In some situations, brachytherapy is combined with a course of external beam RT, which precedes, follows, or interdigitates with brachytherapy. Cervical cancer, for example, may include a course of conventional fractionation radiation and concurrent chemotherapy to treat the pelvic lymph nodes, cervical parametria, and cervix over 6 weeks of daily treatment. Additional intracavitary brachytherapy procedures are done to deliver dose to the primary cervical tumor. In prostate cancer, brachytherapy presents an alternative approach to the narrow therapeutic window. Some studies suggest that disease outcomes after brachytherapy are superior to those after dose-escalated external beam RT (121), although this hypothesis has not been formally tested. Note that many anatomic locations are not suitable for brachytherapy, such that the therapeutic ratio for external beam radiation exceeds that of brachytherapy. Additionally, the high doses delivered by brachytherapy carry the potential risk of increased morbidity if sources are misplaced or migrate (122), so a certain level of technical proficiency is required.

### Particle therapy

Particle therapy (123) is a form of external beam radiotherapy in which particles are delivered rather than x-rays. Beams of electrons, protons, carbon ions, or neutrons are generated by a particle accelerator and delivered to the target volume. The use of proton therapy in particular has grown in recent years because of its unique interaction in tissue. The underlying physics is characterized by a dose distribution known as the Bragg peak, whereby radiation is deposited immediately before the proton comes to rest in tissue, at a depth that can be manipulated by varying its energy. Photons, by comparison, deposit a trail of in-transit dose as the x-ray beams enter and exit the body, typically requiring multiple beams to be arranged so that their paths intersect over the target area. This exposure of tissue within each beam path can have negative clinical consequences for normal tissues. In clinical practice, to treat an entire tumor at a given depth, protons of different energies must be used. This “spread out Bragg peak” offers an improved ability to minimize radiation exposure to normal tissues situated adjacent to the target volume, especially for tissues behind or deeper to the tumor. These reductions in total body exposure (termed integral dose) can translate to reduced complication risks, including the risk of radiation-induced malignancies. Proton beam therapy has had a rapidly growing role in pediatric malignancies and particularly pediatric brain tumors. Tumors located near the spine, like chordomas or paraspinal chondrosarcomas, are a well-defined indication for proton beam therapy, because it allows these relatively radioresistant tumors to be aggressively treated while respecting the radiation tolerance of the spinal cord (124).

Another attribute that distinguishes particle therapies from x-ray beams is that some types of particle beams generate higher biologic effectiveness per unit of energy deposited. For example, the relative biologic effectiveness of protons is 1 to 1.2 times that of photons, whereas for neutrons or carbon ions it could be 4 to 10 times that of photons. The difference in biologic effect is related to the rate at which energy is deposited in tissue, which is a function of both mass and charge of the radiation. This radiobiological difference could decrease the dependency of treatment success on the oxygen levels and cell cycle phase of the tumor target. It is also possible that heavy ion particle therapy could have a more desirable effect on blood vessels that supply tumors. However, further research is necessary to determine whether these radiobiological differences will translate into a better therapeutic index relative to photon-based therapy. Neutron radiotherapy, for example, provided these theoretical advantages, but significant normal tissue damage has limited its clinical applicability.

A clear challenge to particle therapy is the high cost to build treatment centers, which require significantly more physical space and highly trained support staff than do photon-based centers. As of 2013, the International Particle Therapy Cooperative Group (125) reported that only 36 centers are using proton therapy and only 6 centers are using carbon ion therapy worldwide. Between 1969 and 2011, particle therapy has compromised <1% of external beam radiation treatment. This may change in the future, however, because there are currently plans to build 25 more particle facilities worldwide within the next 2 years.

## Conceptual and Therapeutic Bottlenecks and Future Directions to Improve Radiotherapy

### Radioresistance prevents local control of some tumor types

In tumors like glioblastoma, the probability of local tumor control remains miserably poor, despite the use of advanced radiotherapy technologies and very high doses of radiotherapy. This has been well illustrated by unsuccessful trials of dose escalation, including one study increasing doses to >200 Gy using brachytherapy (126). Many possible mechanisms have been proposed for this intrinsic treatment resistance, including prevalence of radioresistant tumor stem cell clones, hypoxia, and uncontrolled signaling by growth factor receptors and/or interferons. Although some of these hypotheses may prove to be correct, the current supporting data are not compelling and a complete explanation is likely to be much more complex. For example, unknown mechanisms may stimulate repopulation of malignant clones to brain tumors from sanctuary sites, which are located outside the irradiated volume. Another example of unyielding radioresistance is observed in very large tumors, like the typical presentation of locally advanced lung cancer. A recent trial for non–small cell lung cancer showed that radiation dose escalation by almost 25% above the standard dose offered no measurable improvement to patients (117). One probable explanation is that the tumor cell burden is simply too large to be controlled in these tumors, such that high enough radiation doses cannot be achieved without incurring major complications to adjacent normal tissues. Another likely explanation is that larger tumors have greater genomic heterogeneity and, hence, a greater likelihood for harboring treatment-resistant clones.

Avenues to overcome this resistance will require a better understanding of the underlying biological problems. They will likely also require a better arsenal of drugs that can specifically sensitize tumor cells or protect normal tissues. It is critical to widen the therapeutic window for radiotherapy at the biological level, particularly in situations where the physical and technical advances could be nearing a plateau. Some of the newer drug approaches discussed earlier may hold promise in this respect, particularly if issues related to drug pharmacokinetics and timing of administration relative to radiation are solved. However, we are left with the impression that there are many uncoordinated and competing research efforts. This probably occurs in part because radiation oncology is a relatively small field, relative to the much larger field of cancer biology research. Relatively few radiotherapy-modifying agents have progressed to testing in clinical trials, and the interest by large drug companies to invest in this class of agents appears somewhat underwhelming. Most of the current trials testing radiosensitizers involve HDAC and PARP inhibitors, neither of which was developed with this specific use in mind.

### Uncontrolled systemic disease can overshadow the importance of local control

For a malignancy to be cured, it must be controlled at both the local and systemic levels. However, the ability to prevent distant metastases remains an unmet challenge in many disease types, such as lung and pancreatic cancers. As systemic chemotherapies continue to improve outcomes in these diseases, the previously unnoticed importance of local control can become more obvious. Recent studies in breast cancer demonstrate this interdependent relationship well. Older studies showed that chest wall radiotherapy improved the regional control of locally advanced breast cancers after mastectomy; these improvements were often clinically dismissed because they did not translate to improvements in overall survival. However, more recent studies done in the context of more effective systemic chemotherapy have shown that postmastectomy chest wall RT does improve survival (127, 128). The magnitude of survival benefit from radiotherapy is similar or greater than that of chemotherapy. This highlights the fact that uncontrolled systemic disease can overshadow the importance of local control, such that oncologists may view local management as secondary. This, in turn, can reduce enthusiasm for improving local/regional modalities like radiotherapy.

An exciting new development is the use of SBRT to ablate sites of metastatic disease, which represents a 180° shift in paradigm. It is based on the observation that some cancers present with only a few sites of metastatic disease, a condition termed oligometastases (129, 130). It remains unclear which patients will benefit the most from SBRT to metastases. However, there is evidence for the concept, based on surgical series that achieved cures after resection of limited liver or lung metastases in patients with colon cancer and sarcomas, respectively. It is probably unrealistic to believe that local therapies will ever be able to address micrometastases, so SBRT for metastases and systemic chemotherapies should continue to evolve together. Nonetheless, impressive outcomes have been reported for SBRT in this metastatic setting. One unmet challenge is the ability to determine which tumors will behave in an oligometastatic manner and which are likely to develop widespread metastases. Advances in imaging would be helpful in this regard, so that oligometastases can be detected and treated when the disease burden is lowest. Recent microRNA signatures have begun to identify patients with oligometastasis who remain oligometastatic, and this might be useful to select optimal candidates for curative surgery or radiotherapy (131). Another exciting idea is the use of SBRT to augment antitumor immunotherapy. Although clinical data are limited (79) and this approach might be limited to particular cancers, this relatively unexplored strategy has a huge potential payoff if it is successful.

### A more individualized approach is needed in radiotherapy

Treatment guidelines are available for most types of cancer, such as those put forth by the National Comprehensive Cancer Network (132). In most cases, the decision to deliver radiation and the key parameters of radiation (for example, dose and volume) are primarily based on the stage of disease at presentation. Management decisions tend not to be personalized toward individual biology or sensitivity to treatment. As an example, locally advanced rectal cancer is typically treated with radiosensitizing chemotherapy and 50.4 Gy, followed by total mesorectal excision, and then adjuvant chemotherapy. Although 10 to 20% of tumors are essentially sterilized by the initial preoperative chemoradiotherapy, almost all patients complete the preplanned course of therapy including surgical excision and full-dose adjuvant chemotherapy. If, however, the treatment-sensitive tumors could be identified early, this information could be used to guide subsequent management. For example, systemic therapy could possibly be deintensified, and perhaps, surgical resection could be eliminated for some patients. By contrast, patients found to have treatment-refractory tumors could be offered even more intensive therapies.

The definition of such predictive markers remains yet another unmet challenge for many cancer types, but some promising leads are beginning to emerge. One example is carcinoma of the head and neck treated with chemoradiotherapy, where the presence of human papillomavirus carries a favorable prognosis (133). Several ongoing trials are testing whether the intensity of chemoradiotherapy can be safely reduced in this subpopulation. To move forward as a field, it will be important to start focusing on biomarkers that predict responsiveness to specific therapies, as opposed to simple prognostic markers. This concept has already been realized for some chemotherapies, like KRAS mutation status for predicting EGFR inhibitor sensitivity (134) or MGMT gene silencing for predicting temozolomide sensitivity (135). Similar markers that predict sensitivity to radiation are being developed, but these have not yet found widespread clinical use. Such innovations will help transition radiotherapy away from a “one-size fits all” mentality and allow us to start the inevitable transition to personalized medicine.

### The allocation of radiotherapy resources needs to be evidence-based

1. Acknowledgments: We apologize to colleagues whose contributions could not be included in this review due to space constraints. Funding: This work was supported by funding from the NIH [CA142642-02 2010-2015 (P.P.C.) and P50 CA090386 2001-2014 (R.R.W. and S.L.L.)], P01 CA071933 (R.R.W.), the Ludwig Center for Metastasis Research, the Center for Radiation Therapy, the Chicago Tumor Institute, a Young Investigator Award from the Prostate Cancer Foundation (S.L.L.), and The Foglia Foundation. Competing interests: S.L.L. and P.P.C. declare no competing interests. R.R.W. has a commercial organizational interest in Catherex, Magi, RefleXion, and OncoSenescence. He has consulted for MedImmune ($1500 one-time consultant fee) and Bristol-Myers ($4900 one-time consultant fee) and has a grant for \$100,000 to study the effect of immune modulators with radiotherapy. R.R.W. has patents in gene therapy and the use of tyrosine kinase inhibitors with DNA damaging agents.