Endoscopic ultrasound (EUS) uses a high frequency ultrasound transducer to obtain detailed images of the tumor mass and the relationship with the five-layered structure of the esophageal wall. EUS attempts to provide measurements of tumor thickness and is regularly used to estimate tumor extension in initial staging for esophageal cancer (Ribeiro et al. 2006). The discriminatory power for distinguishing between early stage tumors and those with deeper invasion may approach 80–90 %. Some have found this gross distinction prognostic in identifying those patients at risk for a positive circumferential resection margin (CRM), if treated with isolated surgery (Reid et al. 2012). However, its exact TNM accuracy is the least prognostic of available clinical information in predicting pre operative stage (Reid et al. 2012; van Vliet et al. 2008; Thosani et al. 2012).

While, EUS remains a frequently used clinical estimate of primary tumor staging, there are technical limitations to its ubiquitous use. Not all patients are capable of receiving a complete EUS due to esophageal stenosis. Additionally, the accuracy of EUS is operator dependent and is subject to a learning curve (Fockens et al. 1996). Ultimately, preoperative staging of lymph node status is challenging. A recently proposed tool from Gaur et al. (2010) for predicting pathologic lymph node involvement based on clinical information is discussed in further detail below in Nomograms and Predictive models section.

For response evaluation, EUS continues to be used as the primary diagnostic modality. However, the accuracy of EUS restaging varies significantly across several recent retrospective analyses (Ribeiro et al. 2006; Giovannini et al. 1997; Chak et al. 2000).

Different methods have been proposed for response assessment with EUS. The first method is to restage according to the TNM staging system (Fockens et al. 1996), second method is to measure the relative reduction in thickness of the tumor (Gaur et al. 2010; Giovannini et al. 1997) and a third method is to measure the relative tumor shrinkage at the maximum cross-sectional area (MCSA) (Gaur et al. 2010; Chak et al. 2000). However, with these different methods the accuracy is still poor and ranges from 17 to 59 % (Sloof 2006; Hirata et al. 1997; Zuccaro et al. 1999; Bowrey et al. 1999). Even when a EUS is combined with biopsy, the accuracy does not exceed 31 % to correctly predict a pCR (Sarkaria et al. 2009). Of patients that have a negative biopsy on restaging endoscopy (cCR) less than 30 % will have a pCR (Sarkaria et al. 2009).

A possible explanation for this discrepancy between endoscopic staging and subsequent pathologic staging is that EUS may not be able to differentiate between post-treatment inflammation or fibrosis and residual tumor (Jamil et al. 2008).

Computed tomography (CT) is usually one of the first steps in staging esophageal cancer patients and is used to evaluate the region of the primary tumor and evaluate for distant metastases. However, the accuracy for locoregional staging is limited. Accuracy for tumor staging has been reported with a range of 42–68 % (Lowe et al. 2005; Wu et al. 2003) and for regional lymph node metastases the pooled sensitivity and specificity is only 0.50 (95 % C.I. 0.41–0.6) (van Vliet et al. 2008). Also, when used for screening for distant metastases CT-scans have difficulty recognizing small distant metastases, while a PET-CT is more sensitive (van Vliet et al. 2008).

CT scan is the most commonly used diagnostic modality in monitoring response of nonsurgical therapy for solid tumors. For esophageal cancer restaging, however, its role remains ambiguous. CT gives good visualization of the tumor bulk in majority of the patients. However, when tumor shrinkage is correlated to pathological response following neoadjuvant treatment some have found a clear correlation (Swisher et al. 2004; Voncken et al. 2012), while others failed (Griffith et al. 1999; Jones et al. 1999).

This difference could be due to an overestimation of edema, inflammation and fibrosis for residual tumor (Westerterp et al. 2005). While, CT remains one of many modalities in re-staging after induction treatment, primary tumor response should be interpreted with caution.

FDG-PET can provide information in initial staging, especially for finding regional nodal metastases and silent distant metastases, where it has a role in selecting patients that will benefit from neoadjuvant CRT. In addition, FDG-PET can contribute to localization, size measurement and GTV definition of the primary tumor and lymph nodes (Katsoulis et al. 2007).

PET-CT provides several pieces of clinically relevant information in restaging after induction treatment. A PET-CT can detect occult metastases after induction treatment and thus saving patients from undergoing a non-curative esophagectomy. A PET-CT after induction treatment detects metastases in 8 % of patients with a consequent adjustment in therapeutic plan (Bruzzi et al. 2007).

An additional advantage of response monitoring with PET-CT is the prognostic value in the decrease, or lack there of, in SUV of the primary tumor. While a PET-directed therapy does have the potential to change clinical practice and improve outcomes, it cannot currently be considered a standard approach in isolation. Standardization of quantitative results across facilities continues to be a technical roadblock. Several methods have been proposed as standard procedure for metabolic response analysis using the standard uptake value (SUV) as a semi quantitative measure of FDG uptake. Proposed methods are: SUV pretreatment, SUV after chemoradiation, percentage of decline of SUV, attainment of a metabolic complete response after chemoradiation and to show an early metabolic response 14 days after start of chemoradiation.

These different methods were analyzed in a systematic review of Omloo et al. evaluating 31 studies (Omloo et al. 2011). Fifteen of these studies tested the pretreatment FDG uptake as a predictive factor. On univariate analysis, SUV was a predictor of survival in 12 out of 15 studies and multivariate analysis showed only in two out of eight studies that SUV was an independent predictor of survival. SUV decrease after completion of neoadjuvant treatment was predictive in only two out of six studies. Finally, there were six studies looking at the SUV decrease and prognosis early during neoadjuvant therapy. SUV decrease was a predictor of response in all of these six studies and a predictor of survival in five of these six studies.

Comparative analysis across FDG-PET articles is challenging due the non-standardization of the image acquisition process and subsequent analytic thresholds. Since the methodology for image acquisition varies, the SUV threshold to predict prognosis varies significantly between analyses (from 3 to 10.5) and cutoff values for amount of SUV change differentiating responders from non-responders also varies depending on the publication (from −30 to −70 %) (Omloo et al. 2011). This illustrates the difficulty to translate these results to clinical practice, although the field of treatment stands to benefit from an adequately powered prospective trial evaluating the true relevance of early SUV decline during CRT. In conclusion, early SUV response assessment holds promise to potentially guide ongoing treatment, but the implementation and technical applicability have not yet developed to the extent required to find a clinical role for routine use.

Concurrent chemoradiation with or without surgery is commonly utilized as primary management of patients with non-metastatic disease. However, there is significant heterogeneity of response, suggesting that there are sub-populations that derive differential treatment benefit from RT. For example, approximately 30 % of patients experience a complete pathological response (Berger et al. 2005; Donahue et al. 2009). A molecular diagnostic that can identify these patients could be utilized clinically to avoid surgery for these patients. In contrast for patients that are predicted to be less responsive to RT, their management could be impacted by either offering RT dose intensification and/or prioritization of surgery (without RT).

At its most basic, a molecular signature is a collection of features that attempt to explain a complex phenotype. While a single predictive molecular marker would be ideal, such an isolated predictor of response to therapy in esophageal cancer has not been documented. In lieu of such a discovery, the technique of combining multiple analytes provides an opportunity to develop a predictive molecular assay; there continue to be a relatively small number of molecular signatures that are routinely part of clinical practice.

Developing a molecular signature typically involves two steps. In the first step, features (genes, proteins, microRNAs etc.) are selected that define the phenotype of interest (i.e. responders to radiation therapy). Once the features are selected then an algorithm is generated to predict the phenotype in an unknown sample. A classic approach is to use samples in a dataset as a “training set” to identify the features and develop the signature. Once the signature is developed, its predictive accuracy is tested on a validation set, ideally independent of the training set. A significant problem in the field of molecular signatures has been their inherent dependence on the “training set” and thus a lack of robust validation analysis (Watanabe et al. 2006; Dalton and Friend 2006).

The majority of molecular signatures in the literature have been developed to describe disease prognosis (independent of treatment), molecular subtypes and/or response prediction to chemotherapy. However, two independent groups have developed RT-specific signatures that have considerable clinical validation. Weichselbaum and colleagues developed an interferon-related gene signature for DNA damage, which was independently validated as a predictor of adjuvant chemotherapy efficacy and for local–regional control after RT in breast cancer (Weichselbaum et al. 2008). Separately, Eschrich and colleagues utilized a systems biology approach to identify a molecular signature of intrinsic tumor radiosensitivity (Eschrich et al. 2009a, b). Using ten specific genes they modeled a radiosensitivity index (RSI) that has been independently validated in multiple disease sites (rectal, esophagus, head and neck, breast) in over 1,000 patients. Of the two signatures, RSI has been validated in a small dataset of esophageal cancer patients (n = 12). The predicted RSI was significantly different in responders (R) vs. nonresponders (NR) in esophageal (RSI R vs. NR 0.37 vs. 0.50, p = 0.05). A low RSI value is consistent with a more radiosensitive tumor. The range of RSI values for 7 responders was 0.11–0.53 and for 5 nonresponders was 0.46–0.54. Therefore it is possible that this signature can be adjusted to support specific clinical decisions to improve clinical care for esophageal cancer patients.

A challenge to the development of a clinically relevant radiosensitivity molecular signature stems from RT’s broad applicability as a therapeutic agent in cancer. Since RT is used in different settings depending on disease site, the clinical utility of the signature would vary depending on the clinical application. A requirement for any signature that is to be applied routinely in the clinic is the development of a standardized and reliable process for tissue acquisition, processing, RNA isolation and gene expression measurement. Recently, the National Cancer Institute selected RSI for commercial development through the recently created Clinical Assay Development Program (CADP). The purpose of the project is the development of an analytically validated, commercial-grade diagnostic platform for RSI that will be ready for testing in clinical trials.

There remains significant opportunity to improve the clinical outcomes for esophageal cancer patients by identifying biological sub-populations that will derive differential treatment benefit from RT. Tailoring RT to fit a particular molecular RT profile will lead to the development of biology-based radiation oncology and result in better RT utilization.

Radiation induced acute esophagitis presents as dysphagia with resultant malnutrition and dehydration, requiring nutritional support (enteral or parenteral) in 17–35 % of the patients (Ahn et al. 2005).

The incidence of acute esophagitis (any grade) ranges from 19 to 79 % and grade ≥3 esophagitis is reported in 1–43 % of patients. This broad scale of the reported esophagitis depends, primarily, on the differences in toxicity reporting and definition. Other potential risk factors for esophagitis include dose schedules and treated volume of the esophagus. Finally, the incidence of esophagitis increases with the addition of chemotherapy to radiotherapy and may vary depending on the chemotherapy schedule given (Meluch et al. 2003; Urba et al. 2003; van Meerten et al. 2006; Cooper et al. 1999; Ajani et al. 2008).

Late esophagitis can present as dysphagia, stricture, necrosis or fistula of the esophagus. The incidence rates of this late toxicity are mainly based on definitive chemoradiation studies. Late esophageal toxicity of any grade occurs in about 35 % of the patients and a grade ≥3 late esophageal toxicity has been seen in 8–21 % of the patients (Cooper et al. 1999).

The strongest predictor for late esophageal toxicity is the severity of acute esophagitis, as a result of consequential late effects. Other predicting parameters for esophageal toxicity are dosimetric.

From lung cancer series we have learned several predicting parameters for acute and late esophageal toxicity. As described earlier, the strongest parameter is the severity of the acute toxicity, but other parameters are a combination of radiation dose and treated volume of the esophagus.

A number of dosimetric parameters have been developed in an effort to reduce the continuously distributed dose-volume histogram (DVH) to a few clinically relevant indices. These relevant indices include: the percent organ volume receiving at least a certain dose (V_20Gy, V_30Gy, V_40Gy); the surface area receiving at least a certain dose (SA_20Gy, SA_30Gy, SA_40Gy); the length of the esophagus included in the radiation field to a threshold dose (LETT_20Gy, LETT_30Gy, LETT_40Gy); the mean esophageal dose (MED), defined as the average dose to the esophagus; and the maximal dose, defined as the highest point-dose within the irradiated esophageal volume (Milano et al. 2007; Rose et al. 2009).

Rose et al. (2009

Stay updated, free articles. Join our Telegram channel

Join