Introduction
Esophageal cancer is the eighth leading cause of cancer with worldwide estimates of
more than 400,000 new cases and over 300,000 cancer related deaths annually [1]. Given the uniformly poor results achieved with surgical intervention alone, treatment
at many centers has shifted to a multimodality approach that incorporates pre-operative
neoadjuvant chemo-radiotherapy in patients with invasive disease who are potential
candidates for subsequent surgical resection [2]. There are several rationales for this treatment strategy. First, down-staging may
lead to improved surgical resectability. Second, the use of neoadjuvant therapy up
front may prevent the systemic spread of cancer more effectively than post-operative
chemo-radiotherapy, which is often deferred for several months to allow for surgical
recovery. Third, chemo-radiotherapy (CRT) may be better tolerated preoperatively.
Fourth, the tumor tissue may be better oxygenated prior to surgery leading to a higher
kill of cancerous cells. Finally, a rather pessimistic argument in favor of neoadjuvant
therapy is that sub-clinical metastatic disease may manifest itself in the preoperative
period during which the neoadjuvant therapy is being administered and these patients
may be spared extensive surgical resection [3]. Although neoadjuvant therapy is supported by sound clinical logic, results from
available studies evaluating the role of neoadjuvant therapy in this setting are promising
but still equivocal. While two large trials (the United Kingdom MAGIC trial [4] and the UK Medical research Council trial [5]) showed survival advantage with neoadjuvant chemotherapy, another large study from
the US (the US Intergroup trial [6]) failed to show any major benefit. A meta-analysis from the Cochrane review group
found improved 5-year survival in patients who underwent neoadjuvant chemotherapy.
However, the improved survival could not be explained on the basis of proportion of
patients undergoing resection or RO resection [7]. Similarly, large randomized trials that evaluated preoperative CRT with surgery
against surgery alone and a subsequent meta-analysis failed to show impressive survival
benefit with neoadjuvant CRT [8]
[9]
[10]
[11].
With such mixed results, it is clear that only a subgroup of patients with esophageal
cancer potentially benefit from neoadjuvant therapy. Therefore, a focus of active
research in this field is the identification of the subset of patients who respond
to neoadjuvant therapy. From a clinical perspective, one would like to know before
or early in the course of neoadjuvant therapy, which cancers are likely to respond.
Ideally, if there were biomarkers, histologic characteristics, or morphologic criteria
that directly allowed one to predict response to particular chemotherapeutic agents,
then appropriate patients could be selected for neoadjuvant therapy. In lieu of such
predictive markers, even if there was a method to detect response to neoadjuvant therapy
early in the course of treatment, patients unlikely to respond could avoid further
toxicity associated with the treatment. In the absence of a method for assessing response
early during neoadjuvant treatment, even a method for restaging the tumor after completion
of CRT could help guide decisions regarding subsequent curative surgery. Patients
with a complete pathological response to induction therapy appear to have the best
long term outcome [10]
[11]
[12]. Thus, a modality that accurately assessed response to treatment could allow stratification
of patients according to the likelihood of further surgical benefit and could optimize
the management of these patients.
Assessment of response to neoadjuvant therapy
It is known that simple clinical variables such as histological types and pre-treatment
tumor stage do not reliably predict response to neoadjuvant therapy. While the role
of biological markers such as level of expression of cycloxygenase2 mRNA, survivin
mRNA, tumor suppressor gene p53, proliferative marker Ki-67, epidermal growth factor
and gene and protein expression profiling in predicting response to neoadjuvant therapy
are being studied at an experimental level [13]
[14]
[15], in clinical practice the response to neoadjuvant therapy in these patients has
either been assessed by morphological imaging such as computed tomography and endoscopic
ultrasonography (EUS) or metabolic imaging such as fluorine 18 fluoro-deoxyglucose
positron emission tomography, or a combination of these imaging tests.
Computed tomography (CT) for restaging
Several studies of computed tomography for assessing response after neoadjuvant therapy
in patients with esophageal cancer have been published [16]
[17]
[18]. In a systematic review of published articles meeting strict inclusion criteria,
Westerterp et al., calculated the sensitivity and specificity of CT in predicting
pathological response to neoadjuvant therapy to range from 33 to 55 % and 50 % to
70 %, respectively [19]. This relatively poor discrimination between responders and non-responders is because
CT is unable to distinguish between viable tumor from reactive changes such as scar
tissue, necrosis, and edema. CT also cannot easily distinguish a hiatal hernia from
a mass at the esophagogastric junction and likely has limited ability to distinguish
tumor thickening from a collapsed or contracted esophagus. CT is particularly poor
in differentiating the various T stages. It should however be noted that most studies
have used single detector technology with 8 - 10 mm tomographic slices. In a recently
published study using third generation or higher CT machines with 5 mm columnated
cuts, the restaging accuracy of CT was reported to be 76 % for T stage and 78 % for
nodal disease [20].
Endoscopic ultrasonography (EUS) for restaging
EUS is currently the best modality for the initial regional, especially T, staging
of esophageal cancers. Of course, thoraco-abdominal CT is still a necessary part of
pre-treatment staging because it detects distant metastatic disease, particularly
in the liver and the lungs. It is not surprising that more than 10 published studies
have also evaluated the role of EUS, if any, in the restaging of patients with esophageal
cancer after neoadjuvant therapy [19]. The majority of these reports are single center experiences with limited, less
than 100, patient numbers. Therefore, the results have been subject to the institutional
expertise of the endosonographer. The more recent studies that attempted to correlate
traditional EUS based T staging with pathologic surgical specimens have shown very
poor accuracies, ranging from 29 to 59 % [20]
[21]
[22]
[23]
[24].
EUS suffers from the same disadvantage common to all morphological imaging. It can
not distinguish between viable tumor and post-necrosis residual inflammatory and scar
tissue. This limitation is supported by the finding that reported accuracy of restaging
with EUS after neoadjuvant therapy varies and is suboptimal. The wide variation in
accuracy of restaging by EUS is partly explained by variations in the design and definitions
and also, measured outcomes (pathological response vs. overall survival) used in some
of these studies.
Given the number of reported studies consistently demonstrating poor accuracy of traditional
EUS criteria for defining T stage following neoadjuvant therapy, some researchers
have attempted to assess changes in tumor size by measuring tumor thickness or maximal
cross sectional area before and after neoadjuvant therapy. These measurements are
obviously indirect measures of tumor volume. As such, they are subject to some measurement
error. Thickness is a single dimensional assessment and errors in measured thickness
will be compounded to the third power when used to assess tumor volume; errors in
measurements of area will be compounded by one order of magnitude when used as an
indirect assessment of tumor volume.
A few studies have examined reduction in tumor thickness as a method for assessing
neoadjuvant response. Bowrey et al. studied 17 esophageal cancer patients with EUS
before and after CRT [24]. Although the T stage using traditional EUS criteria was unchanged in almost all
patients, 10 of the 17 showed a reduction in the thickness of 2 mm or more. All four
patients who had a complete pathologic response at resection were in this group. Swisher
et al. obtained CT, EUS, as well as FDG-PET before and after CRT in 103 patients [25]. Their study found that standardized uptake value (SUV) of PET was the most accurate
test for predicting survival but also noted that tumor thickness as measured by EUS
was correlated with pathologic response. Both studies found that reduction in tumor
thickness correlated with response but the reduction was only a few millimeters. This
is because small changes in thickness can reflect large changes in volume.
Measurement of area is likely a more reliable method for assessing change in overall
size. Willis et al in a study of 41 patients with esophageal cancer showed that positive
response as assessed by measuring reduction in tumor area by EUS correlated well with
pathological response [26]. A positive response on EUS was defined as a greater than 50 % reduction in maximal
tumor cross-sectional area. The same group of investigators in a separate study showed
that positive response on EUS, as defined by the maximal tumor cross sectional area
also correlates with improved survival [27]. Other investigators have also shown that maximal cross sectional area correlated
with survival and histological evidence of response [28]. It is, however, fair to note that correlation between tumor regression even when
based on pathologic assessment has not always been found to correlate with survival.
Because patients with recalcitrant nodal disease after neoadjuvant therapy have markedly
poor survival [29], it is important to assess for nodal response after neoadjuvant therapy and many
surgeons will not consider curative resection in patients with persistent nodal disease.
Reported accuracy of N staging by EUS also varies widely. Studies that did not incorporate
EUS guided FNA for N staging reported suboptimal accuracy of EUS in N staging [21]
[22]. However, EUS guided FNA for confirmation of suspicious lymph nodes improves the
accuracy of EUS N staging [20]. In general, with restaging by EUS, accuracy of N staging is thought to be higher
than T staging. The overall sensitivity and specificity of EUS to predict response
to neoadjuvant therapy from selected studies of poor to moderate methodological quality
have been reported to range from 50 to 100 % and 36 % to 100 %, respectively [19].
Fluorine 18 fluoro-deoxyglucose (FDG) positron emission
tomography (PET) for restaging
PET has rapidly become a non-invasive, whole body imaging method for the pre-operative
staging of a variety of cancers. The increased glucose metabolism of malignant cells
is imaged by the uptake of FDG a ”radiotracer” in oncological PET studies. Unlike
EUS, one of the basic advantages of FDG PET is its ability to image the whole body
and quantify uptake in the entire volume of the primary tumor, thus measuring response
of the primary tumor as well as metastatic disease in a single examination. Several
studies have assessed the role of FDG-PET in the restaging of patients with esophageal
cancer after neoadjuvant therapy.
Brucher et al reported the role of FDG-PET in 27 patients with squamous cell type
of esophageal cancer who underwent neoadjuvant CRT [30]. Quantitative measurement (standardized uptake value or SUV) of tumor FDG uptake
was correlated with histo-pathological response in the subsequently resected surgical
specimen as well as survival. An ROC analysis showed that a threshold of 52 % decrease
of FDG uptake compared with baseline had a sensitivity and specificity of 100 % and
55 %, respectively in predicting response to CRT. Also, patients with decrease in
the FGD uptake of less than 52 % had a significantly shorter median survival time
compared with patients with an SUV decrease of more than 52 % (8.8 vs. 22.5 months,
p < 0.0001). The findings of the study were subsequently confirmed by other investigators
[31]
[32]. While the study by Brucher et al assessed the PET response for the primary tumor
only, subsequent investigators considered both primary tumor sites and nodal disease
to define complete response to CRT. In a systematic review of published information,
the overall joint sensitivity-specificity (Q point) values for FDG-PET was reported
to be 85 % (95 % CI, 77 - 93 %) which was no different than that of EUS (86 %, 95
% CI, 80 - 93 %) and significantly better than that of CT (54 %, 95 % CI, 31 - 77
%) [19].
FGD PET vs. EUS in restaging
It is increasingly clear that for restaging of patients with esophageal cancer who
have undergone neoadjuvant treatment, EUS and FDG PET provide complementary information.
The relative advantages with FDG PET scanning is that it is noninvasive, is feasible
in most patients (unlike EUS where it is either not feasible or suboptimal in a significant
proportion of patients [28]), provides opportunity for whole body imaging with a single test and, when integrated
with CT as PET CT, provides both structural and metabolic information. FDG PET has
high reproducibility and internal validity [33], can be used to assess early response and almost all published studies of FDG PET
shows its ability to clearly differentiate responders and non-responders in terms
of long term survival [34]. Also, unlike EUS, pre-CRT FDG-PET has been shown to correlate with response to
CRT and is the only imaging modality that may identify a priori those patients who are likely to attain a major or even complete pathologic response
to neoadjuvant treatment [32]
[34]. The disadvantages with FDG-PET are that in absence of tissue diagnosis, both false
positive and false negative results are often misleading. The false positive results
which lead to underestimation of effect of CRT are related to post CRT reactive inflammation
with influx of leucocytes and scavenging macrophages in the tumor site. With application
of sophisticated analytic models to interpret tumor uptake of FDG in terms of SUV
measurements, the false positive rates may be lowered [34]. The false negativity of PET scans is due to the foci of residual cancer falling
below the detection of threshold of the PET device and is a major concern particularly
for assessment of nodal disease. Up to 20 % of the adenocarcinomas of the gastro-esophageal
junction, particularly those with poor differentiation, have low FDG-avidity. In these
patients FDG PET is not accurate for restaging after neoadjuvant therapy. FDG PET
has higher false negative rates with cervical and abdominal lymph nodes compared to
thoracic group of lymph nodes. Also, sensitivity for detecting small peritumoral lymph
node is often low due to limited resolution of the PET device to differentiate it
from the scatter effects seen in the area of primary tumor after CRT [34]
[35]. The unique advantage of EUS relative to other imaging modalities is its ability
to obtain tissue aspirate for pathological assessment. Also, EUS is very accurate
in identifying small peritumoral lymph nodes which are often below the threshold of
detection of FDG-PET. Only two published studies have directly compared the performance
characteristics of different modalities for restaging of esophageal cancer after neoadjuvant
therapy in the same patient population. In a prospective study reported by Cerfolio
et al., 41 consecutive patients of esophageal cancer from a single center underwent
restaging by CT, EUS with FNA and FDG-PET/CT after preoperative neoadjuvant CRT [19]. The accuracy of CT, EUS-FNA and FDG-PET/CT in distinguishing pathologic T4 disease
from T1-T 3 stage was 76 %, 80 % and 80 % respectively. The accuracy for nodal disease
was 78 %, 78 % and 93 %. The authors concluded that FDG-PET/CT was more accurate than
EUS-FNA and CT for predicting nodal status. PET was also better at identifying complete
responders after neoadjuvant therapy. It is to be noted that in this study an integrated
FDG-PET/CT scanner was used which may explain the relatively high negative predictive
value of FDG-PET reported in this study. Also, for restaging, the authors used only
traditional T staging by EUS and combined T1 to T3 disease in to one category on the
basis of their argument that in their practice all T1-T3 disease persistent after
CRT was treated by surgical resection if there was no nodal disease. They did not
assess tumor size by EUS, which is the only EUS measure that has been found to be
accurate for assessing neoadjuvant response. Swisher et al. retrospectively reviewed
103 patients who had PET, EUS, and CT performed before and after CRT [25]. Decreased wall thickness measured by EUS or CT and lower standardized uptake values
(SUV) on PET correlated with pathologic response but post-CRT PET SUV ≥ 4 was the
best predictor of response. The PET SUV was also an independent predictor of survival.
No study to date has examined whether the combination of EUS measurement of cross
sectional area, EUS-FNA, and follow-up PET measurements can improve the accuracy for
identifying response to CRT.
Assessment of early response
Approximately half of the patients with esophageal cancer will have no or minimal
response to neoadjuvant CRT. Futile neoadjuvant treatment with its toxicity, morbidity
and costs, both direct and indirect, could be avoided if these patients could be identified
early after initiating CRT. Both CT and EUS are insensitive in this respect since
structurally they can not distinguish between viable and non-viable tumor tissue.
In our experience, the EUS accuracy for restaging using maximal cross sectional area
is worse when the time interval between completion of CRT and restaging EUS is less
than two weeks. On the other hand, FDG-PET has a unique advantage in this situation
because of its ability to selectively identify only viable tumor tissue, which is
metabolically intact. In an interesting study of 40 patients with esophageal cancer,
Weber et al showed that performing FDG-PET scan at 2 weeks after initiation of cisplatin
based poly-chemotherapy, a metabolic response (defined as 35 % or more reduction in
initial SUV) could be identified, which predicted eventual clinical response after
completion of CRT with more than 90 % sensitivity and specificity [36]. Moreover, patients without a metabolic response also had significantly shorter
time to progression/recurrence and shorter overall survival. Similar data on the ability
of FDG-PET to predict early response was also reported by Kroep at al in a smaller
number of patients [37].
Summary
Currently there is no single optimal imaging modality for assessing response in patients
with loco-regional and potentially resectable esophageal cancer who undergo neoadjuvant
CRT. There are also no clear guidelines for making clinical decisions based on the
results of re-staging imaging tests. If post-neoadjuvant assessment of response is
to be used to guide subsequent therapeutic decisions, a combination of different imaging
modalities may be needed. The choice will obviously vary with the local availability
of resources and expertise. It is fairly clear that a conventional CT, particularly
the single detector variety, has no role in restaging and should not be obtained.
Ideally, if a FDG-PET integrated with CT is available, then a FDG-PET/ CT may provide
useful information. If PET/CT shows suspected M1a or M1b disease, which is accessible
to EUS FNA, then a EUS FNA could help establish the diagnosis of metastatic disease.
If the PET/CT shows no evidence of metastatic disease but there is suggestion of residual
tumor, particularly nodal disease, then EUS with FNA of suspicious lymph nodes could
also be helpful. It is not known if a repeat EUS examination should be performed in
patients with a completely negative FDG-PET. Prospective studies are still needed
to determine the optimal time interval for performing FDG-PET after initiating CRT.
For assessment of response of primary tumor to treatment by EUS, traditional T staging
is unacceptably inaccurate. A surrogate measure of tumor volume, maximal cross sectional
area, is a better parameter for assessing response. For assessment of nodal response,
use of more liberal criteria for performing EUS guided FNA could improve the nodal
staging accuracy of EUS compared to that of FDG-PET. Finally, promising development
in the field of multi-detector CT scanning and also, innovative use of 3 D technologies
in the area of both CT and EUS imaging may revitalize the role of these modalities
in assessment of response after CRT in patients with esophageal cancer but are still
likely to be subject to difficulties in differentiating inflammation and scar from
viable cancer. Use of biomarkers and alternative forms of imaging, such as endoscopic
magnetic resonance imaging are at a rudimentary stage of development and are not ready
for clinical applications in the near future.
Fig. 1 EUS of patient with esophageal cancer (a responder) before (a) and after (b) chemoradiation
with tracing of maximal cross-sectional tumor area shown.
Fig. 2 EUS of patient with esophageal cancer (a nonresponder) before (a) and after (b) chemoradiation
with tracing of maximal cross-sectional tumor area shown.
Fig. 3 EUS guided fine needle aspiration (EUS FNA) of a suspicious peri-esophageal lymph
node performed using a linear echoendoscope.
