2  Patients and methods

2.1  Patient selection
This prospective study was approved by the Institutional Review Board of Severance Hospital. Patients were enrolled between September 2011 and December 2013. Inclusion criteria were as follows: age >20 years; ECOG performance status ≥2; diagnosis of rectal adenocarcinoma via endoscopic biopsy, stage T3 or T4, or with clinically enlarged regional lymph nodes identified by CT or rectal magnetic resonance imaging (MRI) regardless of T stage; and having completed preoperative CRT followed by TME with curative intent. Exclusion criteria were as follows: stage IV disease; familial adenomatous polyposis or hereditary nonpolyposis colorectal cancer; benign disease; tumor located more than 10.0 cm from the anal verge; previous or concurrent malignant disease; and patient refusal. Of 122 consecutive patients originally recruited, 12 patients were initially excluded for the analysis: 8 with a mucinous component by preoperative rectal MRI, 2 with incomplete MRI protocols, 1 with stent insertion for an obstructive lesion, and 1 with incomplete data. The remaining 110 patients who fulfilled the inclusion criteria were finally included (Fig. 1).
Figure 1  Flowchart of patient selection and study protocol. MRI = magnetic resonance imaging.

2.2  Endoscopic assessment of tumor response before and after preoperative CRT
All endoscopic procedures were performed by gastrointestinal endoscopists with more than 5 years’ experience. A pre-CRT endoscopy, accompanied by tissue biopsy, was performed before initiating preoperative CRT. Four weeks after the completion of preoperative CRT, endoscopic tumor response was evaluated prior to surgery (Fig. 1). All reviewers were blinded for histology and each other's results. Under same categorical criteria, all reviewers conducted reassessment of tumor response considering the tumor size, morphology, and involved intraluminal circumference before surgery. Endoscopic tumor response was categorized as follows: cCR, no visualization of tumor, white or red scar; nearly-cCR, minimal residual nodularity or stenosis; non-cCR, any ulcer with a necrotic bed regardless size, a definite residual mass, or nodularity (Fig. 2A).
Figure 2  (A) Tumor on endoscopy before and after preoperative chemoradiation therapy (pre-CRT). (a–d) Primary tumor before pre-CRT. (e, f) Clinically complete response of primary tumor after completion of pre-CRT; a typical sign of clinical complete response (black arrows indicate a red scar and a white scar). (g) Minimal residual nodularity and (h) residual tumor with large necrotic area after pre-CRT (a typical sign of poor tumor response). (B) Magnetic resonance imaging (MRI) scans from patients before and after pre-CRT. (a–d) Primary tumor before pre-CRT. (e) Tumor regression grade (TRG) 1; the green arrow indicates radiologic complete response with a hypointense rectal wall. (f) TRG 2; the green arrow indicates a slightly thickened hypointense rectal wall. (g) TRG 3; the green arrow indicates a visible intermediate signal. (h) TRG 4; the green arrow indicates a small area of fibrosis.

2.3  MRI assessment of tumor response before and after preoperative CRT
All magnetic resonance images were obtained using a 3-T scanner (Magnetom Tim Trio, Siemens Medical Solutions, Erlangen, Germany) with a 6-element body phased-array coil on the anterior side of the patient and another 6 elements on the spine coil on the posterior side. For optimal rectal distension, 80 to 100 mL of sonography transmission gel was administered endorectally using an enema syringe. Two experienced gastrointestinal radiologists evaluated MR tumor response after preoperative CRT, and determined TRG before surgery. The mrTRG was assessed into 5 grades. mrTRG1: the absence of any residual tumor lesion; mrTRG2: a small residual tumor presenting as a predominant fibrotic low SI; mrTRG3: all lesions showing partial decrease in size upon comparison of pre-/post-CRT MRIs, but not meeting the criteria of mrTRG 1 or 2; and mrTRG 4 and 5 were each graded if the tumor did not decrease in size, or progressed, respectively.[19] MRI findings were finally categorized according to mrTRG: cCR, TRG1; nearly-cCR, TRG2; non-cCR, TRG3-5 (Fig. 2B). The median time to restaging was 4.1 weeks (interquartile range 3.9–4.3 weeks).

2.4  Preoperative chemoradiation therapy protocol and surgical treatments
We used a standard long-course regimen of 5-fluorouracil (5-FU)-based chemotherapy and a total dose of 50.4 Gy of external beam radiation. Preoperative chemotherapy of 425 mg/m2 5-FU per day and 20 mg/m2 leucovorin per day during weeks 1 and 5 of radiotherapy was administered intravenously. After a median interval of 7.8 weeks (interquartile range: 6.4–8.4 weeks) from completion of CRT, all patients underwent radical cancer resection, based on the principles of TME.

2.5  Histopathologic assessment
The surgical specimens were prepared and dissected according to the protocol described by Nagtegaal et al.[20] The resection surface of the mesorectum and the specimen was fixed in formalin for a minimum of 48 hours. Dissection consisted of serial 5- to 10-mm slicing of the whole tumor and the surrounding mesorectum in the transverse plane. Specimens were embedded in paraffin for histologic examination and stained with hematoxylin and eosin. For all tumors, the shortest distance from the outermost part of the tumor to the circumferential resection margin (CRM) was measured histologically. All tumors were staged according to the 7th American Joint Committee on Cancer (AJCC) TNM classification. At pathologic examination, the response was considered to be major for tumors classified ypT1 or ypT2, and complete (ypT0) when no viable tumor was present.

2.6  Data and statistical analysis
We calculated a sample size of 110 patients in this study, based on a rate of ypT0 of 20% observed in previous studies. The power was set at 80% and the 2-sided significance level set at 0.05, and the sample size was inflated by 10% to account for an ineligibility rate of 10%. Pathologic tumor response to CRT was divided into good and poor response groups using 3 different criteria as follows: ypT0 vs ypT1-4; ypT0-1 vs ypT2-4; or ypT0-2 vs ypT3-4. In the analysis of diagnostic performance of endoscopic and MR tumor response for predicting the good response group, post-CRT findings were scored on a 3-point scale: 3 for cCR, 2 for nearly-cCR, and 1 for non-cCR. For both endoscopic and MR tumor response, positive test results were defined as a score of 2 or higher and negative results were defined as a score of 1. For the combination of endoscopic and MR tumor response (combination modality), positive test results were defined as the sum of these 2 methods scores equaling 3 or higher and negative results were defined as the sum of the scores equaling 2. The sensitivities, specificities, positive predictive values (PPVs), negative predictive values (NPVs), and accuracy of endoscopic tumor response, MR tumor response, and the combination modality were calculated for differentiating the good response from the poor response groups, when applying each definition for the good response group (ypT0, ypT0-1, or ypT0-2). The area under the curves (AUCs) were also calculated from construction of the receiver operating characteristic (ROC) curve.[21] Pairwise comparison of AUCs of endoscopic tumor response, MR tumor response, and the combination modality were performed (endoscopy vs mrTRG, endoscopy vs combination modality, mrTRG vs combination modality). Generalized estimating equation (GEE) model was used to obtain statistically unbiased estimates. SPSS software version 20.0 for Windows (SPSS Corp, Chicago, IL) was used for analyses. Quantitative data were expressed as mean ± standard deviation and qualitative data as frequency and percent. All P-values were 2-sided, and P < .05 was considered statistically significant.