Predicting residual rectal adenocarcinoma in the surgical specimen after preoperative brachytherapy with endoscopic ultrasound

1Division of Gastroenterology and Department of Community Health Science, University of Calgary, Calgary, Alberta; 2Division of Gastroenterology, 3Department of Radiation Oncology, 4Department of Pathology and 5Department of Surgery, McGill University Health Centre, Montreal, Quebec Correspondence: Dr Joseph Romagnuolo, University of Calgary Medical Clinic, 3350 Hospital Drive Northwest, Room G-179, Calgary, Alberta T2N 4N1. Telephone 403-210-8575, fax 403-210-9368, e-mail j.romagnuolo@ucalgary.ca Received for publication October 16, 2003. Accepted April 6, 2004 J Romagnuolo, J Parent, T Vuong, et al. Predicting residual rectal adenocarcinoma in the surgical specimen after preoperative brachytherapy with endoscopic ultrasound. Can J Gastroenterol 2004;18(7):435-440.

Predicting residual rectal adenocarcinoma in the surgical specimen after preoperative brachytherapy with endoscopic ultrasound C ancer of the colon and rectum is diagnosed in over 100,000 new patients each year in the United States (1).The surgical therapy for rectal adenocarcinoma can be particularly disabling to patients, because it involves abdominoperineal resection (APR) for cases in which a 2 cm distal tumour-free margin is not felt to be possible using an anterior approach.Selective pre-or postoperative chemoradiotherapy has traditionally been regarded as adjuvant, in an attempt to reduce local recurrence and perhaps improve survival, but it has not been intended to achieve cure as primary therapy (2,3).Although endoluminal radiotherapy has been in use for over 20 years, this modality is still generally reserved for the small-sized stage T1 or perhaps T2 tumours and/or patients who are medically unfit for surgery (4)(5)(6)(7).Curative brachytherapy (BT) with avoidance of surgical morbidity, along with sphincter-preservation, is an attractive theoretical option for small tumours (8,9).Aggressive multimodality therapy for esophageal cancer, boasting high tumour sterilization (eg, absence of residual tumour) rates, has raised questions regarding the need for surgical therapy in patients with gastrointestinal cancer who have no residual tumour after chemoradiotherapy (10)(11)(12).The key, however, will be the accurate nonoperative identification of those patients.
Endoluminal ultrasound has been used since the early 1980s to stage rectal adenocarcinoma.In 14 studies reviewed by Heriot et al (13), many of which involved a nonoptical probe, this modality demonstrated a tumour (T)-staging accuracy of 75% to 93% (except for one study showing an accuracy of 67%), and a lymph node (N)-staging accuracy of 77% to 88% (except for one study showing an accuracy of 62%).Endoscopic ultrasound (EUS) also appears to be marginally more accurate than T-and N-staging by pelvic magnetic resonance imaging (MRI) (14).It appears, however, that the accuracy of EUS T-staging can be reduced from 86% to 47% by external beam radiotherapy (15).This phenomenon has also been noted in esophageal cancer (16,17).In these studies, although T-staging accuracy was reduced (17), sensitivity in predicting residual tumour was as high as 91%, with a false positive rate of 12% (16).
McGill University Health Centre's novel preoperative conformal high-dose rate endoluminal BT protocol, using threedimensional computed tomography (CT)-simulator planning, has been associated with a complete clinical response rate as high as 65% in stage T2 and T3 tumours (18).This technique's high preoperative sterilization rate shows promise in avoiding APR, or surgery altogether, in patients without residual carcinoma.This depends, however, on the accuracy of EUS to predict the presence or absence of residual tumour after BT, which has not been assessed previously.Thus, the aim of the present study was to evaluate the sensitivity and specificity of EUS performed following a full course of endorectal high-dose rate BT in diagnosing residual carcinoma.

METHODS
Between October 1998 and October 2000, 33 patients were recruited into the study protocol to assess the feasibility and effectiveness of a novel conformal high-dose rate BT protocol for locally invasive rectal adenocarcinoma.Local institutional review board approval was obtained and all patients gave informed consent.

Patient population
Patients with newly diagnosed, locally advanced (stage T2 or T3) rectal adenocarcinoma who presented to the McGill University Health Centre and fulfilled the entry criteria, were recruited for the study protocol.Preoperative, pre-BT staging was performed by pelvic MRI, helical CT and rectal EUS (GF UM 30; Olympus Co, USA).
The study exclusion criteria were the following: inability to undergo surgical resection; stage T1 or T4 disease by either EUS or pelvic MRI; and evidence of distant metastases on helical CT.Ten per cent of patients met the exclusion criteria.Eighteen of the 33 patients recruited into the trial were able to have a preoperative and post-BT EUS arranged and performed within two weeks of the planned surgical resection; a post-BT EUS was logistically not able to be arranged in the narrow window between four and eight weeks post-BT and the surgical date for the others.

BT protocol
All patients were examined by flexible sigmoidoscopy and rectal EUS after completion of their MRI and CT; a pre-BT clinical staging was assigned using the American Joint Committee on Cancer TNM classification (19).During this examination, after the EUS was completed, endoclips (Olympus Co, USA) were endoscopically placed at the proximal and distal margins of the tumour mass to facilitate BT planning and imaging whenever technically possible.
Patients were treated with remote afterloading high-dose rate BT with a flexible multi-channel endorectal applicator (Nova applicator, Nucletron, The Netherlands), using a conformal technique involving three-dimensional CT-guided treatment planning, with the help of the endoscopically placed radiopaque endorectal clips.Dosimetry was tailored to the radial position of the bulk of the tumour, with differential loading of the delivery catheters.A dose of 26 Gy in four consecutive fractions (10 min to 20 min) of 6.5 Gy was delivered to the tumour radial margin.Conformal dosimetry was achievable for tumours up to 15 cm from the anal verge and at a maximum depth of 3 cm.Four to eight weeks were allowed for tumour downstaging before surgical resection.

EUS
Patients underwent a repeat EUS examination, performed by a single dedicated endosonographer, six weeks (range four to eight weeks) after completion of BT, and within two weeks before the planned surgical resection.The endosonographer was not blinded to the pre-BT EUS results.Patients with mild circumferential thickening of all wall layers, with layer blurring but without a focal hypoechoic mass, were considered to have inflammatory changes but no residual tumour; those thought to have residual tumour were re-staged according the American Joint Committee on Cancer (TNM) classification (19).Allowance was made for the frequent mildly irregular appearance of the serosal margin endosonographically, which was felt to represent radiation-related inflammatory change.Therefore, a diagnosis of a stage T3 lesion after BT was made only when the serosal margin appeared frankly invaded by a definite mass or obvious pseudopod projections.The maximal wall thickness (MWT) was also recorded for each patient.Nodal status was determined by accepted endosonographic criteria (node size greater than 1.0 cm, hypoechogenicity, round shape and discrete margins).

Surgery and pathological examination
All patients underwent surgical resection, based on pre-BT staging alone.APR was planned if it was felt by the surgeon that a 2 cm distal tumour-free margin was not achievable with low anterior resection.All of the resected surgical specimens were examined by one of the authors (RPM) and pathological staging was assessed using the American Joint Committee on Cancer TNM classification (19).Examination of the specimen was carried out without knowledge of the post-BT EUS results.
The resected colorectal specimens were fixed in 10% buffered formalin for one to three days, then described and sectioned according to a uniform protocol: multiple longitudinal 5 mmthick slices were made throughout the abnormal areas, either ulcerated or containing macroscopically visible tumour, and multiple tissue blocks were taken from all of these areas; in most cases, the entire abnormal-appearing regions were submitted for histological examination.The sections were embedded in paraffin, processed using standard histological techniques; 5 µm-thick sections were cut, stained with hematoxylin and eosin, and examined for residual tumour, radiotherapy-induced changes and other pathological features appropriate for the assessment of colorectal carcinoma specimens.
Following the surgery, the patients received adjuvant postoperative external beam radiotherapy and chemotherapy as per standard of care, if positive lymph nodes were found in the resected specimen.

Statistical analysis
Descriptive statistics were computed for demographic variables.Sensitivity, specificity, and positive and negative predictive values for the prediction of residual cancer were calculated, with 95% CIs.These test performance characteristics were also calculated for assessment of nodal involvement.Kappa statistics measuring agreement beyond chance were calculated.
Factors predicting residual tumour were assessed from both the pre-BT EUS (proportion with stage T3 disease, proportion with nodal involvement, MWT) and the post-BT EUS (proportion down-staged, absolute reduction in MWT).Unpaired Student's t tests were used for hypothesis testing involving continuous variables, and the χ 2 test (or Fisher's Exact test where appropriate) was used for comparisons of proportions.

RESULTS
Eighteen patients underwent a post-BT EUS assessment for the presence of residual tumour and none had endoclips remaining at this assessment.The mean age ± SD was 70±11 years (range 52 to 93 years) and 78% of the patients were male.Before BT, EUS showed two patients (11%) who were stage T2 (invading the muscularis propria but not into the serosa) and 16 (11%) who were T3 (through the muscularis propria into the serosa and/or perirectal fat).Mean MWT was 14±3.4 mm on EUS.Five patients (28%; 95% CI 7% to 49%) had regional lymph nodes that were sonographically suspicious for malignant involvement (N1); 72% did not have significant regional lymphadenopathy (N0).
In all 18 patients, resected specimens were evaluated according to the aforementioned pathology protocol.After BT, pathological examination revealed that seven patients (39%; 95% CI 16% to 61%; P=0.0006 versus the 0% expected if BT was not able to sterilize the tumour) had no evidence of residual tumour (Figures 1 and 2).One additional patient had only carcinoma in situ.Three of the 11 patients (27%) with residual tumour only had tiny foci of malignant cells, seen in selected blocks.One of the 18 (5.6%)treated had residual T1 disease, one (5.6%)had stage T2 disease and eight (44%) had stage T3 disease.Significant downstaging occurred with significantly fewer patients having stage T3 disease after BT, in the surgical specimen (P<0.001;95% CI for difference: 17% to 72%).Mean MWT on EUS decreased from 14 mm to 9.2 mm (P<0.001).In contrast, there was no significant downstaging in nodal status.Eight of 18 patients (44%) had positive nodes by pathology, after BT.
Eleven of 18 patients (61%) underwent an APR for their distal carcinomas, including four (36%; 95% CI 8% to 65%) who ultimately had no evidence of residual malignancy on pathological evaluation.Of the 11 with residual tumour, the degree of differentiation (grading) was reported as welldifferentiated in one, moderately differentiated in six and moderately to poorly differentiated in three patients; in one patient, the degree of differentiation could not be assessed reliably.One patient was found to have unexpected liver metastases at the time of surgery.Three patients (43%) without residual tumour at pathology had documented extramural spread: two with regional lymph node involvement and the other with unexpected liver metastases found at the time of operation.One of the three (33%) was detected to have nodes by EUS.
Median follow-up was 6.5 months from the time of recruitment (range two to 14 months).One patient died at nine months without local tumour recurrence; this patient had residual post-BT T3N1 disease by both EUS and pathology and had undergone an anterior resection with negative margins.No other recurrences have been documented.

Performance of post-BT EUS in predicting residual tumour
Nine of the 11 patients with residual tumour were detected by EUS (sensitivity 82%; 95% CI 59% to 100%) (Figure 2).One of the two false negative exams was in a patient with carcinoma in situ.In contrast, there were five false positive EUS examinations (specificity 29%; 95% CI 0% to 62%) (Figure 3).Two of the five false positives (40%) had mucin lakes and one had extensive necrosis.Most specimens without residual tumour had active inflammation.The corresponding positive and negative predictive values were 64% and 50%, respectively, and the overall accuracy for predicting residual tumour was 61% (95% CI 39% to 84%), with a corresponding poor kappa for agreement beyond chance (0.11).
In eight patients (44%; 95% CI 21% to 67%), the pathological T-stage was exactly predicted on the post-BT EUS.Of the remaining 10, seven (70%; 95% CI 42% to 98%) were overstaged, but four of these (40%) were still within one T-stage of the pathological stage.This comparison is summarized in Table 1, with exact matches bolded.Again, the corresponding kappa was poor for agreement beyond chance (0.11).
In the post-BT nodal assessments, there were four false negative and two false positive EUS examinations.These correspond to a sensitivity of 50% (95% CI 15% to 85%) and a specificity of 80% (95% CI 55% to 100%).

Predictors of residual tumour
A univariate comparison of baseline (eg, pretreatment) factors between patients with successful tumour 'sterilization' and those with residual tumour was performed.There was no statistically significant difference in age, prevalence of pretreatment stage T3 disease, prevalence of pretreatment nodal involvement or MWT (Table 2).Neither the downstaging rates nor the absolute reductions in MWT were significantly different in 'sterilized' patients versus those with residual tumour (Table 2).Patients with stage T3 disease and those who did not appear to be downstaged by EUS were more likely to ultimately have residual tumour after BT; the strength of these trends did not reach statistical significance.Also, proportionally, nodal disease was more prevalent in the sterilized group than in the group with residual disease; however, the CI of the difference was wide (Table 2).These negative exploratory analyses lack the statistical power to be definitive.

DISCUSSION
The accurate identification of patients with and without residual tumour after a BT protocol that promises a high tumour sterilization rate in rectal adenocarcinoma would be ideal and could, in theory, lead to less extensive surgical intervention.EUS is the logical test for this purpose, although external beam radiation-induced inflammation decreases the accuracy of the EUS assessments.In the present study, the inflammatory changes caused by BT were also a significant limitation, leading to a high false positive rate for EUS in predicting post-BT residual tumour.The sensitivity for predicting residual tumour was reasonable but the negative predictive value was still poor.
One potential limitation was our inability to arrange preoperative post-BT assessments on a significant number of patients enrolled in the BT study.There is no reason to believe that patients selected by these scheduling difficulties would bias the apparent EUS accuracy in either direction.
Although higher stage tumours, in theory, could have been more urgently operated on and consequently missed their post-BT EUS, the high prevalence (89%) of stage T3 lesions in our cohort suggests this did not occur.
Overstaging by EUS after BT was a significant problem (Table 1): the inflammation that extends to the serosal edge, producing an irregular margin with small pseudopod-like projections, will likely remain indistinguishable from serosal invasion.Disappointingly, even though these previously described artifacts (15) were taken into account in our EUS evaluation of these patients after BT, and despite a six-week average lag time between BT and EUS, there remained a significant number of false positive assessments.Clinically, however, false positives are preferable to false negatives.Some post-BT EUS examinations in patients with complete tumour sterilization revealed focal hypoechoic masses, as demonstrated in Figure 3.The focal nature of this tailored conformal technique, which targets high doses of radiotherapy to those specific regions of the wall most involved by tumour, may lead to these more focal changes rather than the typical symmetrical diffuse changes seen with external beam radiotherapy.These masses mimicking recurrent tumour may represent focal edema from inflammation, or perhaps alternatively, necrosis or mucin lakes (Figures 4-6) as was seen pathologically in some specimens.The addition of deep tissue needle biopsy of endosonographically suspicious residual intramural masses, using curvilinear EUS and fine-needle aspiration (FNA) technique, is a possible future solution to this problem, with or without polymerase chain reaction to look for micrometastases.However, a negative cytology would still not completely exclude the presence of residual tumour.
One limitation of this study is the poor sensitivity of EUS for N-staging.Nodal re-staging was not a primary endpoint of this particular study; we did not expect BT to alter lymph node status because of its limited depth of penetration and its lack of effect on N-staging, as shown in previous studies (20).Furthermore, some of the missed lymph nodes are beyond the reach of the echoendoscope.The addition of EUS-guided FNA may have improved our results in terms of N-staging.
In our clinical context, a high negative predictive value (true negatives as a proportion of all negative assessments) would be ideal to 'rule out' residual tumour.Because the residual tumour   prevalence was relatively high (61%), the negative predictive value was correspondingly poor (50%), despite a reasonable sensitivity.Improvements in the specificity, and perhaps even the sensitivity, by mural FNA via EUS, possibly with polymerase chain reaction, as mentioned above, may improve the negative predictive value in the future.

CONCLUSIONS
Radial EUS assessment of residual rectal adenocarcinoma after high-dose rate conformal BT is inaccurate in terms of re-staging.
Although it appears sensitive for detecting residual disease, the false positive rate remains high and the poor predictive values corresponding to this moderate prevalence of residual disease limit its utility in this setting.Further studies are needed to examine if these patients with complete histologic sterilization of their low-lying tumours could be candidates for sphinctersparing operations.

Figure 1 )Figure 2 )
Figure1) Endoscopic views of rectal adenocarcinoma pre-and postbrachytherapy in a patient with pathologically documented tumour sterilization (ie, no residual tumour), in the left and right panels, respectively.Significant radiation proctitis is seen in the right panel

TABLE 1 18 *Figure 3 )
Figure 3) Radial endoscopic ultrasound (EUS) (7.5 MHz) images of rectal adenocarcinoma which was successfully 'sterilized' by brachytherapy (BT), but interpreted as being unsuccessful by post-BT EUS (ie, false positive post-BT EUS).Both images were taken from the same patient and at approximately the same distance from the anal verge.The left and right panels represent pre-and post-BT EUS examinations, respectively.The EUS represented by the right panel was interpreted incorrectly as having residual tumour (MASS [Hypoechoic mass]), with invasion into the serosa (T3) as evidenced by pseudopodlike projections, illustrating the difficulty in these assessments.P Prostate; S Seminal vesicles

TABLE 2
Comparison of baseline characteristics and changes in endoscopic ultrasound (EUS) appearance as a result of brachytherapy (BT) in patients with and without post-BT residual tumour