A current goal in medicine is to individualize treatment to eradicate disease while reducing toxicity and improving quality of life. Currently radiation doses are generalized with consensus statements for dose tolerance of normal tissues and local control of tumor based on histology and organ/location. Current radiation dose guidelines are based on laboratory studies of the general radiosensitivity/radioresistance of a particular tumor and modeling the tumor control probability (TCP) based on growth characteristics of a tumor. Recently molecular biomarkers allow for personalized treatment approaches and potentially adaptive radiotherapy and or radiation dose escalation/de-intensification. This chapter gives a review of known common biomarkers used in clinical practice today and is sectioned by organ site. The chapter also presents data on the more recent large scale analysis of a patient’s molecular profile, i.e.–omics profiling of tumors (genomics SNPs, proteomics, etc.) to predict radiation effects. The goal of biomarker research is to one-day tailor treatment based on an individual’s genetic and molecular profile. This book chapter aims to highlight various biomarkers in each cancer type by histology.

HPV (human papilloma virus)/p16: predictive biomarker. HPV is the most well-known and reproducible predictive marker in head and neck cancer to date (Wierzbicka et al. 2015). It is the most widely discussed marker researched in radiation oncology today. This section focuses on head and neck biomarkers with an emphasis on HPV. Radiation dose de-escalation trials are ongoing in patients with HPV/p16+ head and neck cancer based on the breadth of research on this virus (Wierzbicka et al. 2015).

Traditionally all squamous cell cancers of the head and neck were treated the same with dose guidelines based on tumor size and surgical lymph node drainage patterns. Historically risk factors for head and neck cancer included excessive smoking and alcohol history. More recently there has been an increase in non-smokers with head and neck cancer. The common thread in this subtype of patients has been the prevalence of HPV/p16 in the tumor cells (Chau et al. 2014). HPV is a DNA virus; there are many subtypes. The one subtype of HPV consistently found to correlate with head and neck cancer is HPV/p16. HPV positivity is confirmed by both presence of HPV DNA using PCR and protein overexpression of p16 on immunohistochemical stains (Lee et al. 2015).

In patients with cancer of the oropharynx patients that have HPV p16 positivity tend to present with more locally advanced disease (Ang et al. 2010). In spite of this HPV p16 expression imparts a better response to radiation both standard fractionation and altered fractionation (Lassen et al. 2011, 2014). Lau et al. (2011) demonstrated that p16+ head and neck squamous cell cancer patients had improved overall survival, disease-free survival, and less locoregional recurrence when compared to p16-patients (Lau et al. 2011). Hong et al. (2010) demonstrated that HPV-patients had 13-fold increased risk of locoregional failure and 4-fold increased risk of death as compared to HPV+ patients (Hong et al. 2010). Lassen et al. (2014) showed that the HPV/p16 expression only positively correlates with response in oropharynx patients; p16 expression did not affect outcome of non-oropharynx patients (larynx, hypopharynx) (Lassen et al. 2014). HPV/p16 seems to be more predictive of response to radiation over surgery; Quon et al. (2013) analyzed p16 expression in resectable oropharyngeal carcinoma and found no difference in surgical outcomes of p16+ and p16-patients treated with surgery first (Quon et al. 2013). Future trials underway in head and neck cancer use radiation dose de-escalation in HPV p16 positive oropharynx patients due to this correlation as a predictive biomarker (Ang and Sturgis 2012).

As HPV positivity is established as a predictive biomarker of response to radiation, there are now studies trying to delineate biomarkers that predict for failure in the HPV+ subset (Lee et al. 2015). Inflammatory cells have been evaluated as a marker for predicting treatment failure in HPV+ tonsil cancer and Lee et al. (2015) demonstrated that both overall survival and disease specific survival was affected by high CD68+ and low CD8/CD4 T lymphocyte ratio (Lee et al. 2015). Neither T stage nor N stage were related to outcomes in this HPV+ tonsil cohort (Lee et al. 2015). Extent of inflammation and response to radiation is a common theme and this paper attempts to start the further subtype characterization of HPV+ patients. Other studies have tried to identify a panel of biomarkers that will predict treatment failure, Thibodeau et al. (2015) found that upregulation of LCE3D (late cornified envelope 3D) and KRTDAP (keratinocyte differentiation-associated protein) and down regulation of KRT19 (keratin 19) was observed in posttreatment failures of HPV+ patients (Thibodeau et al. 2015). These biomarkers haven’t been extensively studied in radiation and so future studies will be needed to validate these results.

EGFR is another biomarker analyzed in head and neck patients. There have been mixed reports in its ability to predict locoregional control from radiation therapy (Lassen et al. 2013). While the signal for prediction is not as strong as HPV p16 expression, upregulation of EGFR has been shown to correlate with tumor growth and benefits from accelerated radiotherapy (Eriksen et al. 2004). In the DAHANCA 6 and 7 studies, low EGFR expression correlates with high HPV/p16 expression which seems reasonable given that HPV/p16 expression patients respond better to treatment (Lassen et al. 2013). However the signal for EGFR predicting head and neck cancer was not as strong as HPV/p16 expression and so is not routinely recommended for monitoring at this time (Lassen et al. 2013). Recently EGFR was reassessed in HPV+ and HPV-head and neck patients and again demonstrated that EGFR expression did not affect outcomes in HPV+ patients. In HPV-patients, EGFR expression correlated with worse locoregional failure but only in univariate analysis with T and N stage playing more prominent role (Vainshtein et al. 2014).

Similar to EGFR, p53 mutational status has also been analyzed in head and neck cancer patients. Alone p53 mutational status did not affect local control or overall survival but there was a suggestion that p53 mutant head and neck cancer patients may benefit from shortened treatment time similar to EGFR overexpressing patients (Eriksen et al. 2005). Future studies are underway examining EGFR expression, p53 mutational status, HPV/p16 expression, and smoking status to see if there are further subsets of head and neck cancer patients.

Hypoxia molecules have also been studied as predictive biomarkers of radiation resistance mostly because it is known that lack of oxygen makes tumor cells less sensitive to radiation (Overgaard et al. 2005). Osteopontin is one such biomarker associated with tumor hypoxia. In studies by the DAHANCA group, Overgaard et al. 2005demonstrated head and neck cancer patients with high levels of osteopontin (>167 ug/L) had poorer responses to radiation with higher levels of locoregional failure (Overgaard et al. 2005). In a parallel study at Stanford Petrik et al. (2006) demonstrated that high levels of osteopontin (>450 ng/ml) correlated with higher rates of locoregional failure (3 yr FFR was 72% for patients with osteopontin <450 ng/ml versus 48% for patients with >450 ng/ml (Petrik et al. 2006). Other markers of hypoxia being investigated as markers of radiation resistance include hypoxia inducible factor HIF-2 alpha (HIF-2) and carbonic anhydrase CA9; CA9 is actually one indicator of HIF-1alpha (HIF-1) function. HIF-1 and HIF-2 are thought to be two separate response pathways (Koukourakis et al. 2006). Using data from the CHART trial (continuous hyperfractionated accelerated radiotherapy), Koukourakis et al. (2006) demonstrated that head and neck cancer patients with high levels of HIF-2 and CA9 had worse locoregional control (Koukourakis et al. 2006). These studies haven’t led to routine measurement of hypoxic markers or use of hypoxia modifiable treatments such as nimorazole but overcoming hypoxia is still an active area of research in radiation resistance. Future patient samples may well be tested for these hypoxic markers.

From all these various markers only HPV is used routinely in radiation oncology clinical practice in head and neck cancers. Research studies are still underway with these other biomarkers and it is yet to be determined which will be of clinical use in the future.

Gynecologic and head and neck cancers share many similar biomarkers and thus this next section will highlight some studies of biomarkers in the gynecology literature.

Similar to head and neck cancer, HPV has been implicated in cervical cancer as well and is used as a biomarker. The high-risk HPV 16 and HPV 18 are the most common HPV strains implicated in cervical cancer (Song et al. 2011; Qin et al. 2014). Currently the standard treatment of cervical cancer is concurrent chemotherapy and radiation therapy (Qin et al. 2014). There has been suggestion that for a subset that is radioresistant treatment intensification is needed but finding that subset has remained elusive thus far. Some reports suggest that there is difference response to chemoradiation among the HPV strains (Ferdousi et al. 2010). In one small study of 113 cervical cancer patients, response to radiation was better in HPV-58 and HPV-31 versus HPV 16 and HPV-33 (Ferdousi et al. 2010). There have been reports that persistent HPV after definitive radiation for cervical cancer may predict worse local control of disease (Song et al. 2011). Song et al. (2011) showed that persistent HPV DNA 24 months after radiation predicted risk of local recurrence and HPV persistence at just 3 months alone was the earliest predictor of local recurrence (Song et al. 2011). Testing for HPV is routinely done in clinical practice and this data suggests that all patients treated for cervical cancer with radiation should have HPV testing after radiation is complete as well. Patients with persistent HPV may need treatment intensification either in form of altered radiation treatment regimens, or adjuvant chemotherapy. This data still needs to be validated in multi-institutional trials before becoming routine use in clinical practice.

EGFR has also been explored as a biomarker in cervical cancer in the same manner as it has been studied in head and neck cancer (Qin et al. 2014). Overexpression of EGFR has been shown to lead to more failures after definitive radiation suggesting it is a predictive biomarker of radiation resistance (Pérez-Regadera et al. 2011). Perez-Regadera et al. (2011) examined 112 cervical cancer biopsies and found that patients with high overexpression of EGFR on biopsy had more pelvic relapses and decreased disease free survival with hazard ratio of 2.31 (Pérez-Regadera et al. 2011). Cerciello et al. (2007) demonstrated that changing EGFR levels during radiotherapy administration did not have any correlation with response though they did not mention quantification of initial expression of EGFR (Cerciello et al. 2007). Thus, EGFR may be a biomarker only of inherent radiation resistance and from these studies it suggests that initial EGFR expression of tumor may be more significant in predicting radiation resistance. EGFR testing in cervical cancer is not routinely done currently but may be considered in future trials arguing for more intensive treatment of radioresistant tumors.

Other biomarkers being tested include the bcl2 apoptotic family members such as BAX, prostaglandin pathway molecules such as COX, and hypoxic markers such as HIF1alpha (Qin et al. 2014). Currently only HPV is routinely screened prior to radiation therapy in cervical cancer but these other biomarkers may become significant as we try to individualize treatment or argue for treatment intensification in radiation resistant cervical cancer subtypes.

Glioblastomas are the most aggressive brain tumor. Historically treatment was surgery and whole brain radiation. With the advent of chemotherapy and better imaging with MRI brain, radiation in glioblastoma is directed at the tumor. A landmark trial demonstrated the benefit of temozolomide with limited field radiation in improving overall survival (Stupp et al. 2009).

Current evidence has demonstrated that not all glioblastomas are the same and survival varies widely. New reports suggest specific molecular subtypes have better survival. In the landmark Stupp trial, subset analysis of this trial demonstrated that patients with MGMT methylation have double the survival at 5 years (Stupp et al. 2009). The MGMT (O6-methylguanine-DNA-methyltransferase) gene encodes a DNA repair protein. Methylation of the MGMT promoter silences the gene and prevents DNA repair namely of damage caused by alkylating agents. Thus, this gene is important for regulating the DNA integrity of the cell. MGMT methylation has been shown to sensitize glioblastoma cells to temozolomide, an alkylating agent and is thus a predictive biomarker of the chemotherapy response. Rivera et al. (2010) asked the question if MGMT methylation also sensitizes cells to radiation as radiation also works primarily through DNA damage (Rivera et al. 2010). 225 patients were analyzed in their study of glioblastoma patients who received radiation alone after maximal safe surgical resection (i.e. no chemotherapy such as temozolomide) (Rivera et al. 2010). They demonstrated that patients with MGMT methylated had better response to radiation and that unmethylated tumors were twice as likely to progress during radiation treatment. On multivariate analysis, methylation was independent of age, KPS, and extent of surgical resection (Rivera et al. 2010).

MGMT is now used routinely in clinical practice as both a predictive and prognostic biomarker for chemotherapy. New strategies in glioblastoma treatment involve using MGMT methylation to alter upfront therapy, i.e. adding other targeted agents such as everolimus, etc. that work through pathways different than temozolomide to sensitize glioblastoma cells to radiation. The goal is more tailored treatment for glioblastoma subtypes in an effort to more efficiently eradicate tumor cells. Recently a Phase III randomized control study (GLARIUS trial) was published showing that altering chemotherapy in MGMT methylated patients could improve progression free survival (Herrlinger et al. 2016). In this study, non-MGMT methylated (i.e. predicted temozolomide and radiation resistant) patients with newly diagnosed glioblastoma were randomized to standard of care temozolomide + radiation versus bevacizumab+irinotecan+radiation (Herrlinger et al. 2016). However, the study failed to show improved overall survival as the original Stupp trial so temozolomide and radiation is still the standard for glioblastoma patients (Stupp et al. 2009; Herrlinger et al. 2016). Future studies may target radiation dose escalation or altered radiation fractionation schedules such as hypofractionation or stereotactic body radiation doses.

Recent studies by Ahmed et al. (2015) have looked into generating a radiosensitivity index (RSI) for different cancer subtypes including glioblastoma (Ahmed et al. 2015). The RSI described previously by the same group uses gene expression patterns, tissue histology, and ras and p53 status when cells are treated with radiation (Ahmed et al. 2015; Eschrich et al. 2009). The RSI index directly correlates with tumor radioresistance (high RSI = radioresistance) (Ahmed et al. 2015; Eschrich et al. 2009). Ahmed et al. (2015) used the TCGA (the cancer genome atlas) which has large population data based on histology and centralized at the NIH to see if RSI could predict radiation response (Ahmed et al. 2015). RSI was a predictor of overall survival for the glioblastoma cohort (Ahmed et al. 2015). For radiosensitivity predictions RSI correlated with response in patients with high MGMT expression (Ahmed et al. 2015). MGMT is already known to predict radiation response so it will be interesting to see if the further information from RSI can help delineate earlier which patients will need treatment intensification.

There are now known to be three subclasses of glioma: proneural, proliferative, and mesenchymal. By using these subtypes, known biomarkers such as MGMT and new genomic profiles such as RSI we can begin to tailor treatment for glioblastoma.

Prostate cancer is the second leading cause of cancer deaths. Although prognostic factors such as clinical T stage, Gleason score and pretreatment PSA aid in prognosis of prostate cancer there are still many outliers. Early risk prostate cancer can fail localized therapy such as radiation earlier than planned and high risk prostate cancer can remain seemingly indolent for years. Other prognostic and predictors of radiation response in prostate radiation therapy are needed.

PSA is the single most used test in prostate cancer. It is used to screen men though its use as a screening tool has come into question given its high sensitivity and over-diagnosis of slow growing prostate cancers. Elevated PSA (generally >4) prompts urological consult and prostate biopsy. PSA response after definitive treatment (i.e. either surgery or radiation) is the most sensitive test and predicts progression free survival long before patient develops any recurrent tumor or metastases.

Kabarriti et al. (2014) and colleagues wanted to test if PSA can be used during radiation treatment to predict response. Such a marker would give patients confidence in radiation alone as salvage treatment and less worry about earlier need for additional salvage treatment such as hormonal treatment (i.e. lupron) or chemotherapy (i.e. docetaxel). Kabarriti et al. (2014) demonstrated that PSA response during radiation is a predictive biomarker of outcome of salvage prostatectomy patients (Kabarriti et al. 2014). 5 year biochemical control rate for PSA responders was 81% compared to 37% for non-responders (Kabarriti et al. 2014). This suggests that PSA should be used during radiation treatment to give an earlier predictor of patient outcome. If PSA is not responding adequately during radiation dose escalation could be considered or earlier use of additional chemotherapy may be warranted.

Another more recent biomarker is the genome prostate cancer classifier (GC) (Den et al. 2014). The GC score developed by Den et al. (2014) utilizes-omics data, specifically gene expression patterns with microarrays (Den et al. 2014). This GC score helps to predict which men would benefit from earlier adjuvant radiation versus delayed salvage radiation when frequently PSA is higher and radiation may be of less benefit (Den et al. 2014). The A high GC score predicted increased biochemical failure and metastases thus suggesting these men need more aggressive systemic therapy (i.e. long term hormones) (Den et al. 2014).

In lines with the GC score, research has focused on pretreatment molecular characteristics of the prostate cancer to determine if radiation as localized treatment should even be attempted or if patient should go to surgery. P53 accumulation and high expression in prostate cancer cells seems to predict radiation treatment failure in many prostate studies reported by independent research groups (Ritter et al. 2002; Scherr et al. 1999). Abnormal p53 expression was then analyzed in a multi-institutional RTOG trial, RTOG 8610 (Grignon et al. 1997). In this trial, all patients received radiation as the local treatment for prostate cancer and the phase III randomization was for ± addition of androgen deprivation (i.e. zoladex and flutamide) (Grignon et al. 1997). Abnormal p53 expression led to decreased time to development of distant metastases and increased incidence of distant metastases though these results must be taken with caution as they only demonstrated that p53 expression only affected response in patients who received both androgen deprivation and radiation and this was in an era without prostate radiation dose escalation which is standard today (Grignon et al. 1997).

Although in general prostate cancer is thought to be slow growing and with low proliferation index, there is a rare subtype of prostate cancer with a high proliferation index as measured by Ki-67 (Pollack et al. 2004). Pollack et al. (2004) analyzed Ki-67 expression in prostate cancer biopsies of men enrolled in a multi-institutional phase III randomized trial RTOG 92-02 (Pollack et al. 2004). In this trial, prostate cancer patients with locally advanced prostate cancer (intermediate and high risk) where randomized to long-term or short-term androgen deprivation concurrent with radiation therapy (Pollack et al. 2004). Pollack et al. (2004) demonstrated that high Ki-67 (cutpoint 7.1%) in prostate predicts poor response to treatment and these patients had higher biochemical failure, distant metastases and cause-specific death (Pollack et al. 2004). Future studies would need to aim at better initial treatment for this aggressive subtype of prostate cancer maybe with upfront plan for trimodality treatment versus trying one localized treatment and watching/waiting.