Fax: (011) 49 8941406377
The clinical impact of histopathologic response assessment by residual tumor cell quantification in esophageal squamous cell carcinomas†
Version of Record online: 10 APR 2006
Copyright © 2006 American Cancer Society
Volume 106, Issue 10, pages 2119–2127, 15 May 2006
How to Cite
Brücher, B. L. D. M., Becker, K., Lordick, F., Fink, U., Sarbia, M., Stein, H., Busch, R., Zimmermann, F., Molls, M., Höfler, H. and Siewert, J. R. (2006), The clinical impact of histopathologic response assessment by residual tumor cell quantification in esophageal squamous cell carcinomas. Cancer, 106: 2119–2127. doi: 10.1002/cncr.21850
Presented in part at the American Society of Clinical Oncology—Gastrointestinal (ASCO-GI) Symposium, January 26–28, 2006, San Francisco, California; and at the Annual Meeting of the American Association for Cancer Research (AACR), April 1–5, 2006, Washington, D.C.
- Issue online: 27 APR 2006
- Version of Record online: 10 APR 2006
- Manuscript Accepted: 14 DEC 2005
- Manuscript Revised: 14 OCT 2005
- Manuscript Received: 18 JUL 2005
- esophageal carcinoma;
- multimodal therapy;
- neoadjuvant radiochemotherapy
The objectives of this study were to investigate histomorphologic features as a response classification after neoadjuvant radiochemotherapy (RTx/CTx) and to correlate the results with clinical outcome parameters (e.g., postoperative morbidity and mortality, recurrence, and survival) in patients with locally advanced esophageal squamous cell carcinoma (ESCC).
Three hundred eleven patients with histologically proven, locally advanced, intrathoracic ESCC (clinical T3 or T4, N0-N+, M0) located at or above the level of the tracheal bifurcation underwent preoperative, combined, simultaneous RTx/CTx followed by esophagectomy. Response to RTx/CTx was classified by the quantification of residual tumor cells. A histopathologic response was defined as <10% residual tumor cells found within the specimen compared with a histopathologic nonresponse, which was characterized by >10% residual tumor cells.
A histopathologic response was correlated significantly with complete tumor resection status (R0 resection) (P .0001), histopathologic tumor (ypT) category (P <.0001), lymph node involvement (P <.0001), lymphatic vessel invasion (P <.001), and survival (P <.0001). A multivariate Cox regression analysis revealed that histopathologic response classification according to the percentage of residual tumor cells was an independent prognostic factor (P <.0001). Nonresponders had greater postoperative pulmonary morbidity (P = .01), a greater 30-day mortality rate (P = .02), and a dismal survival rate compared to histopathologic responders (P <.0001).
Histopathologic response evaluation based on the quantification of residual tumor cells provided meaningful information for the assessment of outcomes among patients with ESCC who have underwent neoadjuvant RTx/CTx. The current results indicated that histopathologic responders may represent a subgroup of patients who benefit from neoadjuvant therapy followed by surgery. Cancer 2006. © 2006 American Cancer Society.
The prognosis for patients with esophageal squamous cell carcinoma (ESCC) is poor.1 In these patients, prognostic factors that are evaluated by major surgical centers are the achievement of complete tumor resection (R0), the depth of tumor infiltration (pT), the presence of lymph node metastasis (pN), and invasion of lymphatic vessels.2–5 Surgery potentially is curative if complete (R0) resection can be achieved.6 Multimodal strategies have been proposed to increase locoregional tumor control by virtue of an increase in the proportion of R0 resections, because R0 resection is difficult to accomplish in patients with locally advanced tumors.7 Some studies have shown improved long-term survival compared with patients who underwent surgery alone.8, 9 In other studies, the impact on survival was unclear.10–12 It seems, however, that the impact of neoadjuvant treatment on survival depends on whether a patient is a responder or a nonresponder to multimodal therapy.3
It has been shown that clinical response evaluation in patients with ESCC does not correlate with clinical outcome or survival.3, 13 Moreover, it has not been demonstrated that clinical response is reliable or is easily reproduced.3 Fluorodeoxyglucose positron emission tomography (FDG-PET) may be a valuable tool with which to discriminate responders from nonresponders during and after neoadjuvant therapy of esophageal cancer.14–19 The International Union Against Cancer (UICC)20 and American Joint Committee on Cancer21 classification systems are necessary to determine postoperative tumor categories, lymph node categories, lymphatic vessel invasion, grading, and residual tumor. However, in patients with ESCC, there is no generally accepted system for the determination of tumor response after neoadjuvant therapy. Histopathologic regression classifications have been suggested as a means of standardizing the description of histopathologic results after neoadjuvant treatment in patients with esophagogastric cancer.22, 23 An important basis for the analysis of response is a unique and simple classification of histopathologic regression in the resected specimen. It has been demonstrated that esophageal adenocarcinomas (EAC) and ESCC are tumor entities with different survival rates and prognosis,6 and it is assumed that each has a unique tumor biology.6, 12 Thus, a tumor-specific evaluation of the histopathologic response of ESCC alone to neoadjuvant treatment would be valuable. In addition, there is an ongoing controversy regarding whether responders or nonresponders to neoadjuvant radiochemotherapy (RTx/CTx) should undergo esophagectomy.3–5, 8, 10, 12, 13, 16, 24–30 A more precise definition of response requires valid data on histopathologically observed tumor regression after neoadjuvant RTx/CTx.13, 31 In the current study, we evaluated a histopathologic regression classification based on residual tumor cell quantification in patients with locally advanced ESCC after neoadjuvant treatment and focused on prognosis and surgical outcomes.
MATERIALS AND METHODS
Three hundred eleven patients with histologically proven, locally advanced ESCC without distant metastases (clinical T3 [cT3]-cT4, N0-N+ M0) who were considered medically fit for surgery32 and who received RTx/CTx in consecutive Phase II trials between 1989 and 2004. Only patients with tumors located at or above the level of the tracheal bifurcation were chosen for the multimodal protocol (Table 1).3 In all patients, the staging procedures comprised endoscopy, endoluminal ultrasound (EUS), bronchoscopy (including brush cytology and biopsy), and computed tomography (CT) scans. Patients who had previously received CTx, RTx, laser therapy, or stent implantation were excluded from this study. Informed consent was obtained from all patients.
|Clinical TNM Status||No. of Patients (n = 31)||No. of Patients (%)||P|
|Responders (0-10% Residual Tumor Cells; n = 162)||Nonresponders (> 10% Residual Tumor Cells; n = 149)|
|T3N1||304||157 (96.9)||147 (98.7)|
The neoadjuvant regimen consisted of simultaneous RTx/CTx. External-beam RTx was delivered by a 2-opposed-field technique using 10-megavolt (MV) or 15-MV photons (Mevatron KD 2; Siemens,New York NY) with 2 grays (Gy) per fraction per day and 5 fractions per week until 2000. The total dose from 1989 until 1994 was 40 Gy; this was reduced to 30 Gy between 1994 and 1998. Coincidently, the clinical target volume, which initially included all locoregional lymph node stations along with the tumor, was modified and reduced to the direct juxtaregional lymph nodes (tumor ± 4 cm craniocaudal), and it was enlarged by 1.5 cm in all directions for the planning target volume, taking patient movements into account. The total dose was increased again to 40 Gy with single doses of 2.0 Gy between 1998 and 2000, and it was increased thereafter to 45 Gy using 1.8-Gy single doses. From that time on, 3-dimensional treatment planning with 3-field to 5-field conformal techniques was applied. Between 1989 and 1994, 5-fluorouracil (5-FU) (450 mg/m2 per day) was administered simultaneously as a continuous infusion for 21 days by means of a systemic venous port-system. Between 1994 and 2000, the dose of 5-FU was changed to 300 mg/m2 per day with the same schedule of application. Cisplatin was administered in addition to 5-FU between 2000 and 2003 5 times weekly in Weeks 1 and 5 at a dose of 20 mg/m2. From 2003 onward, patients were treated within a Phase I-II dose-finding study in which oxaliplatin (40-50 mg/m2 given once weekly) replaced cisplatin, and the dosage of 5-FU was reduced to 225 mg/m2 per day given as a continuous infusion. The composition of changes in neoadjuvant treatment between 1989 and 2004 is illustrated in Figure 1. Tumor resection was performed from 4 to 6 weeks after the completion of neoadjuvant treatment. Preoperatively, clinical tumor response was assessed as described previously.3, 16
Clinical Workup and Surgery
Clinical response evaluation by endoscopy and computed CT scanning, with its well known minor impact on prognostic value,3, 13 was obtained directly prior to surgery, as described previously. The patients underwent standardized transthoracic en bloc esophagectomy with 2-field lymphadenectomy.33 Gastrointestinal continuity was restored by gastric pull-up as a 2-stage procedure, as described previously. To avoid a high rate of paresis of the recurrent nerves, lymph node dissection in the neck was not performed. Cervical tumors were treated with a partial esophageal resection, and the reconstruction was made with a free jejunal graft.34
Resected specimens were fixed with formaldehyde (4%) for 24 hours. The complete tumor was cut into slices that contained the entire esophageal wall, which was marked with ink at the circumferential resection margins. Lymph nodes were prepared from the remaining perimuscular tissue, and the oral and aboral resection margins were sampled. The tissue was embedded in paraffin, and 5-μm serial sections of each block were cut and stained with hematoxylin and eosin and van Gieson stains. All specimens were classified by 2 experienced pathologists who were blinded to the clinical data according the criteria of the UICC,20 including the R-classification and grading. In case of a disagreement about tumor grading, both pathologists reviewed the specimen on a double-headed microscope and reached a consensus diagnosis. The histopathologic response to RTx/CTx was classified on the basis of the percentage of residual tumor cells within the specimen. A previously published classification system was used to define 5 subtypes of histopathologic regression in resected specimens.23 Briefly, the percentage of viable residual tumor cells was assessed as follows: complete regression (ypCR) showed fibrosis with or without inflammation extending through the different layers of the esophageal wall, but with no viable residual tumor cells; subtotal regression (ypSR) was characterized by the presence of <10% viable residual tumor cells, partial regression (ypPR) was characterized by from 10% to 50% viable residual tumor cells, minimal regression (ypMR) was characterized by >50% viable residual tumor cells, and no change (ypNC) was defined by the absence of any regressive changes. This classification scheme was modified, because recent investigations indicated that a simplified classification with fewer subtypes is more useful and reliable.3, 16, 35 The accuracy and precision of regression coefficients and tests of their statistical significance are affected by the ratio of events (e.g. deaths) per variable.36 This influence is minimized by means of grouping, because the regression coefficients are affected less by a higher number of events and, thus, become less biased relative to the true values. Moreover, no statistically significant difference in survival between patients who achieved complete remission and subtotal remission could be observed. Accordingly, the analysis of 2 subgroups is more valid statistically. Histopathologic responders were defined as patients who had specimens with <10% residual tumor cells, compared with histopathologic nonresponders, who had specimens with >10% residual tumor cells.
Patients were examined every 3 months in the outpatient clinic for the first year and at 6-month intervals thereafter. Follow-up assessment included a physical examination, laboratory investigations, plain chest X-ray, endoscopy, abdominal ultrasound, and CT scans of the chest and abdomen. Immobilized patients who had low performance status were examined for disease progression and/or recurrence by their general practitioner. Follow-up was complete for all patients. The median follow-up was 4.6 ± 4.2 years (range, from 7 months to 15 years). Of 311 patients, 125 patients (40.2%) died during follow-up.
Quantitative data were expressed as the median and range. For the intraindividual analysis, the Wilcoxon test was used. Differences in the proportion of patients were analyzed by using a chi-square test. The death of a patient was chosen as an indisputable and objective endpoint for the multivariate analysis of prognostic factors. Survival rates were calculated according to Kaplan–Meier method,37 including the 95% confidence interval (95% CI), and were tested with the log-rank test.38 Statistical differences were determined by using the log-rank test. Independent predictive factors were determined by multiple proportional hazards (Cox) regression analysis.39 The following prognostic factors were entered as covariates: histopathologic response according to residual tumor cells within the specimen (responders with 0-10% residual tumor cells vs. nonresponders with >10% residual tumor cells), completeness of tumor resection (R0 resection vs. R1 or R2 resection], ypT category (ypT0-ypT2 vs. ypT3-ypT4), ypN category (ypN0 vs. ypN1), and lymphatic vessel invasion (positive vs. negative). The covariates were selected in a stepwise fashion by using the maximum likelihood ratio (equals variables in the equation). All tests were 2-sided, and a P value of .05 was considered significant. All analyses were performed using the statistical package SPSS for Windows (release 11.5; SPSS Inc., Chicago IL).
Patient Characteristics and Surgical Approach
The 311 patients who were assessed in the current study had a median age of 56.4 ± 8.7 years (95% CI, 19.7-76). There were 251 males and 60 females. Important characteristics, including age, gender, pretherapeutic grading, and pretherapeutic clinical staging, did not differ significantly between histopathologic responders and nonresponders. Fifty-seven patients underwent cervical resection (18%) and 254 patients (82%) underwent transthoracic esophagectomy.
Postoperative Morbidity and Mortality Rates
Overall morbidity was 43.4% (135 of 311 patients), and there was no statistically significant difference between histopathologic responders and nonresponders. However, histopathologic nonresponders showed significantly higher pulmonary postoperative morbidity (39 of 149 patients; 26%) compared with responders (24 of 162 patients; 15%; P = .013). There were no significant differences between histopathologic responders and nonresponders with regard to other postoperative complications (e.g. anastomotic leakage, paralysis of the recurrent laryngeal nerves, chylothorax, bleeding, necrosis of the interponate, tracheal necrosis, aspiration, necessity for tracheotomy, cardiac events, thrombosis, neurologic complications, hepatic failure, sepsis, or wound infections). At 30 days, the mortality rate for all patients was 3.5% (11 of 311 patients). Histopathologic nonresponders with >10% residual tumor cells had a 30-day mortality rate of 6.0% compared with 1.2% in responders (Table 2). This difference was statistically significant (P = .02).
|Postoperative mortality||Histopathologic Response: No. of Patients (%)||P|
|Responders (< 10% Residual Tumor Cells: n = 162)||Nonresponders (> 10% Residual Tumor Cells; n = 149)|
|Hospital mortality||6 (4)||13 (9)||.06|
|30-Day mortality||2 (1)||9 (6)||.02|
A complete microscopic and macroscopic tumor resection (R0 resection) was achieved in 247 of 311 patients (79.4%) (Table 3). A microscopically incomplete tumor resection (R1 resection) was achieved in 56 patients (18%), and a macroscopically incomplete tumor resection (R2 resection) was achieved in 8 patients (2.6%). Histopathologic responders had a significantly higher R0 resection rate (152 of 162 patients; 94%)compared with nonresponders (95 of 149 patients; 64%; P <.0001) (Table 3). Furthermore, univariate analysis revealed that histopathologic responders had a significant shift toward more favorable T and N categories compared with nonresponders (P <.0001) (Table 3). The postoperative detection of lymphatic vessel invasion was less frequent in responders compared with nonresponders (P < .0001) (Table 3). An analysis that excluded patients who had ypT0 tumors (n = 78 patients) revealed that responding patients with <10% residual tumor cells still had a significantly a lower rate of positive lymph nodes than responders (P <.0001).
|UICC Category*||Histopathologic Response: No. of Patients (%)||P|
|Responders (< 10% Residual Tumor Cells; n = 162)||Nonresponders (> 10% Residual Tumor Cells; n = 149)|
|Completeness of tumor resection*|
|R0 resection||152 (94)||95 (64)||<.001|
|R1 resection||9 (5)||47 (31)|
|R2 resection||1 (1)||7 (5)|
|Histopathologic tumor status*|
|ypT1||32 (20)||7 (5)|
|ypT2||38 (23)||26 (17)|
|ypT3||14 (9)||102 (69)|
|Histopathologic lymph node status*|
|ypN0||127 (78)||71 (48)||<.0001|
|ypN1||35 (22)||78 (52)|
|Lymphatic vessel invasion*†|
|Negative||154 (95)||111 (74)||<.0001|
|Positive||8 (5)||38 (26)|
Correlations of Histopathologic Response, Patient Survival, and Recurrence
Of 311 patients, 186 patients (60%) still were alive at the date of last follow-up. The survival (median duration in years ± standard error) of all 311 patients was 2.2 ± 0.3 years (95% CI, 1.7-2.7 years). The estimated 2-year and 5-year survival rates (percentage of surviving patients ± standard error) according to the method of Kaplan and Meier37 were 55.4% (± 3.6%) and 35.7% (± 4.8%), respectively (Table 4, Fig. 2). Patients who responded in terms of T category, N category, and residual tumor showed a substantial benefit in survival, as shown in Table 4.
|Category||Median Survival (Years)||2-Year Survival Rate (%)*||5-Year Survival Rate (%)*||P|
|Completeness of tumor resection†|
|Complete tumor resection (R0 resection)||2.5||64.7||40.2||<.0001|
|Incomplete tumor resection (R1 and R2 resection)||0.7||20.6||8.6|
|Histopathologic tumor status†|
|Histopathologic lymph node status†|
|Histopathologic response (percentage of residual tumor cells within the specimen)|
The median survival (duration in years ± standard error) of patients who had a histopathologic response was 5.3 ± 1.1 years (95% CI, 3.1-7.5 years), with estimated 2-year and 5-year survival rates of 75.4% (± 4.3%) and 55.2% (± 7%), respectively (Table 4, Fig. 2). In the presence of >10% residual tumor cells within the specimen, the median survival decreased to 1.3 ± 0.1 years (95% CI, 1.1-1.5 years; P <.0001) with estimated 2-year and 5-year survival rates of 33.6% (± 5.1%) and 16% (± 5.3%), respectively. The median recurrence-free interval in all patients who underwent R0 resection (n = 247 patients) was 2.5 ± 0.5 years (95% CI, 1.5-3.6 years). Among 247 patients who underwent R0 resection, 70 patients (28%) developed recurrences. Of 152 histopathologic responders who underwent R0 resection, 36 patients (24%) had recurrences found during follow-up, compared with 34 of 95 histologic nonresponders who underwent R0 resection (36%; P = .04). Histopathologic responders had a median recurrence-free survival of 5.3 ± 1.1 years (95% CI, 3.2-7.4 years) compared with 1.5 ± 0.1 years (95%CI, 1.3-1.7 years) among histopathologic nonresponders (P = .01). After patients with operative mortality were excluded from the analysis, the prognostic difference between responders and nonresponders remained at the same level of statistical significance (data not shown). A multiple proportional hazards (Cox) regression analysis included all univariately tested variables (Table 5) and identified both residual tumor and histopathologic response assessed by residual tumor cell quantification as independent prognostic factors (Table 6).
|Histopathologic responders according to residual tumor cells within the specimen*||37.275||1||<.0001|
|Completeness of tumor resection†‡||32.342||1||<.0001|
|Histopathologic tumor classification‡§||29.237||1||<.0001|
|Histopathologic lymph node classification‡∥||10.685||1||.001|
|Lymphatic vessel invasion‡¶||10.184||1||.001|
|Cox Multivariate Proportional Hazards Regression Analysis*||DF||Exp (B)||95% CI||P|
|Histopathologic responders according to residual tumor cells within the specimen*||1||2.580||1.733-3.842||<.0001|
|Completeness of tumor resection†||1||2.093||1.403-3.122||<.0001|
|Histopathologic tumor classification†||1||1.128||0.671-1.894||.650|
|Histopathologic lymph node classification†||1||1.133||0.770-1.668||.526|
|Lymphatic vessel invasion†||1||1.126||0.707-1.793||.616|
Mandard et al.23 investigated histopathologic regression in an inhomogeneous population of 93 patients with EAC, ESCC, undifferentiated esophageal carcinoma, and unspecified esophageal tumors who had received treatment with neoadjuvants, and those investigators subdivided tumor regression into 5 subtypes. Becker et al.22 investigated 36 patients with neoadjuvantly treated gastric carcinoma and subdivided tumor regression into the 4 subtypes described by Mandard et al.23 These scoring systems are based on the ratio of fibrosis to viable tumor. Pitfalls of this approach can be the differentiation of therapy-induced fibrosis from preexisting tumor desmoplasia, which may be difficult in individual tumors. Conversely, in patients with histopathologically complete tumor regression (ypCR), it cannot be excluded that single viable tumor cells have been overlooked, even when the complete tumor bed has been examined thoroughly. Recently published data40 indicated that the survival of patients with esophageal cancer who had complete tumor regression was better compared with the survival of patients who had subtotal regression (1-10% residual tumor cells). However, the survival difference was not statistically significant; and, in our own subgroup analysis, no significant survival differences between 0% and from 1% to 10% residual tumor cells could be demonstrated. Therefore, it seems justified that the cut-off value between responders and nonresponders may be defined at the 10% level. This classification correlates well with the strongest known prognostic factors, e.g. R0 resection, lymph node involvement, and lymphatic vessel invasion, and with survival, as shown previously.2 Multivariate testing confirmed that this very simple classification of tumor regression was an independent prognostic factor. This approach minimizes potential errors in histopathologic response evaluation that may be caused by posttherapeutic tumor heterogeneity. The histopathologic evaluation of the whole tumor bed and adjoining areas may serve as the gold standard for assessing response to neoadjuvant therapy in patients with ESCC.
A correlation analysis of response, as defined by histopathologic classification in relation to postoperative outcomes, e.g. postoperative morbidity and mortality, to the best of our knowledge, has not been reported previously. Our analysis revealed a significant disadvantage in postoperative pulmonary morbidity for histopathologic nonresponders. Pulmonary complications account for most of the postoperative deaths after esophagectomy.8, 41, 42 In so-called “high-volume-centers”43 for upper gastrointestinal surgery, the postoperative mortality rates vary from 2% to 8.6%.6, 8, 3, 42–45 It has been shown elsewhere that more postoperative deaths occurred in patients who had received RTx/CTx.10 An unexpected finding from the current results was that nonresponders showed a significantly higher 30-day mortality rate than responders.
The overall survival in the current patient cohort was 2.2 years. This relatively favorable outcome reflects the monocentric character of this study, which was performed in a center with extensive experience in esophageal surgery,2, 3, 6, 16, 32, 33, 35, 46 and is consistent with the results achieved at other high-volume centers.8, 13 Patients who underwent R0 resection had a median survival that corresponded well with previously published studies.8 Histopathologic responders had a 4-fold better survival rate than nonresponders. Ajani et al.13 reported a median survival of 25.6 months in responders; however, in that analysis, histopathologic responders were classified according to whether they achieved complete remission (ypCR) or partial remission (ypPR). Otherwise, the median follow-up in our investigation was 4.6 years, compared with to from 24 months to 30 months.13, 47 Jin et al.48 observed an overall survival rate of 46% in all patients; and, in clinical responders, Adelstein et al.47 reported only 46% overall survival compared with the 55% rate in our current investigation.
Our results indicate that histopathologic responders to neoadjuvant RTx/CTx for ESCC have a significantly more favorable complete (R0) tumor resection rate. This result confirms recently published data from other groups.8, 12 In the current study, the rate of complete resection in responding patients was 94% compared with 64% in nonresponding patients. Results from other groups suggest that, after primary resection, an R0 resection rate of only 60% can be expected.25 The local disease recurrence rate in the current study was comparable with previously published data25; however, histopathologic responders had a lower rate of recurrence after R0 resection than nonresponders.
Although our study was retrospective, we conclude that histopathologic response evaluation based on quantification of residual tumor cells provides significant prognostic information for patients with esophageal squamous cell carcinoma after neoadjuvant RTx/CTx. Our data suggest that histopathologic responders are a subpopulation of patients who benefit from neoadjuvant therapy followed by surgery. Nonresponders had markedly less favorable survival than patients in whom definitive RTx/CTx49, 50 had been performed.
The clear limitation of tailored treatment based on the individual probability of response to RTx/CTx is that, to date, there is no reliable means of response prediction. Positron emission tomography may have the greatest potential to predict histopathologic response.15, 16 Metabolic response evaluation by measuring decreased standardized uptake values may be a valuable surrogate marker of a significant tumor cell kill in the future and merits further investigation. Molecular biologic markers also may prove to be helpful.
- 8Only pathologic complete response to neoadjuvant chemotherapy improves significantly the long term survival of patients with resectable esophageal squamous cell carcinoma: final report of a randomized, controlled trial of preoperative chemotherapy versus surgery alone. Cancer. 2001; 91: 2165–2174., , , et al.
- 11A randomized Phase III trial of preoperative chemoradiation followed by surgery (CR-S) versus surgery alone (S) for localized resectable cancer of the esophagus [abstract]. Proc Am Soc Clin Oncol. 2002; 21: 130..
- 28Pre-operative radiotherapy prolongs survival in operable esophageal carcinoma: a randomized, multicenter study of pre-operative radiotherapy and chemotherapy. The second Scandinavian trial in esophageal cancer. World J Surg. 1992; 16: 1104–1109; discussion, 1110., , , et al.
- 34Reconstruction of the Cervical Esophagus by Microsurgical Transfer of Intestinal Segment. New York: Springer; 1988., , , , .
- 35Cyclin D1 as a predictive marker in patients with esophageal squamous cell carcinoma and neoadjuvant radiochemotherapy [abstract]. Proc Am Soc Clin Oncol. 2003; 22: 289., , , , , .
- 39Regression models and life tables. J R Stat Soc. 1972; 34: 1987–2001..