Fax: (713) 745-1163
Lymphovascular invasion as a tool to further subclassify T1b esophageal adenocarcinoma
Article first published online: 18 JAN 2008
Copyright © 2008 American Cancer Society
Volume 112, Issue 5, pages 1020–1027, March 2008
How to Cite
Cen, P., Hofstetter, W. L., Correa, A. M., Wu, T.-T., Lee, J. H., Ross, W. A., Davilla, M., Swisher, S. G., Fukami, N., Rashid, A., Maru, D. and Ajani, J. A. (2008), Lymphovascular invasion as a tool to further subclassify T1b esophageal adenocarcinoma. Cancer, 112: 1020–1027. doi: 10.1002/cncr.23265
- Issue published online: 19 FEB 2008
- Article first published online: 18 JAN 2008
- Manuscript Accepted: 18 SEP 2007
- Manuscript Revised: 31 AUG 2007
- Manuscript Received: 16 JUL 2007
- Dallas, Cantu, Smith, and Park Families
- Rivercreek Foundation
- lymphovascular invasion;
- esophageal cancer;
Lymphovascular invasion (LVI) and/or lymph node metastases (LNM) adversely influence the overall survival (OS) of patients with T1 esophageal adenocarcinoma. Although endoscopic therapy may be adequate for patients with T1a cancer, patients with T1b cancer require esophagectomy/lymphadenectomy. The authors hypothesized that LVI status would subclassify T1b cancers and facilitate new therapeutic strategies.
Ninety-nine consecutive patients with T1 adenocarcinoma were analyzed after they underwent esophagectomy/lymphadenectomy. LNM was assessed in all patients, and LVI was assessed in 89 patients. OS was correlated with pathologic cancer stage in association with LVI and LNM.
The 5-year OS rate for patients with T1a tumors (88%) was superior to that for patients with T1b tumors (62%; P = .001). The 5-year OS rate for patients who had cancers without LVI (85%) was superior to the rate for patients who had cancers with LVI (36%; P = .0001). It is noteworthy that, for cancers without LVI, the 5-year OS rate for patients with T1b tumors (77%) was similar to the rate for patients with T1a tumors (90%; P = .08), but it was superior to the rate for patients with T1b tumors that had LVI (27%; P = .006). The presence of LVI and/or LNM resulted in worse 5-year OS (≤37%) compared with the lack of LVI and/or LNM (88%; P < .001). The rate of LNM for patients who had T1b tumors without LVI still was 19%, and the relapse rate was 16%.
The current results demonstrated that LVI distinguishes the biologic behavior of early esophageal cancer, and patients who have T1b cancer without LVI have a clinical biology similar to that of patients with T1a cancer. If LNM before surgery can be diagnosed with high sensitivity by better endoscopic techniques and/or molecular biomarkers, then a new therapeutic paradigm for T1b cancers could emerge. Further research is needed on patients with T1b esophageal adenocarcinoma. Cancer 2008. © 2008 American Cancer Society.
In 2007, approximately 15,560 individuals in the United States will be diagnosed with esophageal cancer, and approximately 13,940 will die of this disease.1 The incidence of adenocarcinoma at the gastroesophageal junction (GEJ) has risen over the last 15 years.2 Because Barrett metaplasia has been identified as a premalignant condition, this may have resulted in closer surveillance and frequent diagnosis of early stages3; however, this concept has not been accepted widely. For the most common stage (II or III) of esophageal adenocarcinoma, patients often receive preoperative chemoradiation4–10 or postoperative adjuvant chemoradiation.11
Endoscopic mucosal resection (EMR) of T1a adenocarcinoma is very appealing and remains an evolving concept. Longer follow-up and individualization of this approach is needed when considering the accompanying high-grade dysplasia, long-segment Barrett metaplasia, nodularity, and multifocality. The use of surgical resection for T1 lesions has been traditional and widely accepted. EMR also is being viewed as a diagnostic/therapeutic tool for early esophageal cancers. Whereas patients with stage T1a cancers may be amenable to endoscopic resections, lesions that are invasive into the submucosa (T1b) are at significant risk for lymphovascular invasion (LVI) or even lymph node metastasis (LNM). For this reason, these patients typically are treated primarily by resection, including a proper lymphadenectomy.12–15 Despite complete resection, 5-year survival rates of only 50% to 60% are attained.16–18 However, we have observed considerable heterogeneity in the survival of patients with T1b cancer, and we hypothesized that a detailed histopathologic evaluation that included LVI would further subclassify patients with T1b cancer. If there was a significant difference in outcome within subgroups of T1b-staged patients, and if this could be identified effectively, then there would be potential for individualization of therapy decisions. Here, we present our data on 99 consecutive patients with T1 adenocarcinoma who underwent esophagectomy as primary therapy.
MATERIALS AND METHODS
We searched the Thoracic and Cardiovascular Department's esophageal database at the University of Texas M. D. Anderson Cancer Center between November 1988 and November 2005. We identified 99 consecutive patients with T1 adenocarcinoma of the esophagus or the GEJ who underwent primary esophagectomy. We retrospectively reviewed the medical records and pathology for this analysis. Inclusion criteria for this review were T1N0 or T1N1 cancer, adenocarcinoma, and the availability of a postoperative pathology specimen for reanalyses.
Exclusion criteria for patients were preoperative chemotherapy or radiation therapy, pathology other than adenocarcinoma, histology limited to high-grade dysplasia, invasion beyond the submucosa, or distant metastases (M1). The Institutional Review Board of the University of Texas M. D. Anderson Cancer Center approved this project.
Routine preoperative evaluation included upper gastrointestinal endoscopy with biopsy and endoscopic ultrasonography (EUS). Spiral computed tomography images of the chest and abdomen were obtained to exclude distant metastases. Positron emission tomography scans were obtained in 32 patients (32.3%).
All 99 patients underwent resection with lymph node dissection for curative intent. The type of surgeries included transhiatal esophagectomy (50 patients), transthoracic esophagectomy (Ivor-Lewis) in 41 patients, extended 3-field lymphadenectomy (6 patients), and minimally invasive esophagectomy (1 patient).
Resection specimens were processed according to a standardized protocol. Entire segments of Barrett esophagus and associated, distinct, lesion-like nodules, polyps, and ulcers were sampled for histopathologic examination. All resected specimens were sampled to detect periesophageal/perigastric lymph nodes. In addition, other regional lymph nodes were sampled separately and reviewed microscopically. Reports of the findings were reviewed, and the following data were obtained: size, depth, location of the cancer, LVI status, degree of differentiation, presence of Barrett mucosa, margin status, and the presence or absence of LNM. LVI was defined as the identification of tumor cells in the vessel lumen with attachment to the vessel wall.
All cancers were staged according to the tumor, lymph node, metastasis (TNM) staging system of the American Joint Committee on Cancer/International Union Against Cancer. Tumors that were classified as pathologic T1 tumors (pT1) were subclassified into 2 groups based on the depth of invasion: T1a (intramucosal; infiltration into the lamina propria or into but not beyond the muscularis mucosae) and pT1b (submucosal; infiltration into the submucosa without invasion of the muscularis propria). The submucosal invasion was identified by the presence of tumor cells in association with submucosal ducts or large, thick-walled blood vessels. In some tumors, duplicated muscularis mucosa was noted, and submucosal invasion was identified only if the tumor extended beyond this duplicated muscularis mucosae.
Follow-up and Survival
Follow-up data were obtained from review of the hospital records and from the University of Texas M. D. Anderson Cancer Center tumor registry. The time to relapse and location of relapse (local and distant), the date of death, and the cause of death were ascertained. Follow-up was calculated from the date of surgery to the date the events occurred or the date of last contact from the most recent clinical, endoscopic, or radiologic investigation and/or telephone communication.
Data were collected prospectively in accordance with a standardized protocol. All statistical analyses were performed using the SPSS Statistical Software Package version 15.0 (SPSS Inc., Chicago, Ill). Pearson chi-square tests or Fisher exact tests were used to compare categorical characteristics. A t test was used to compare continuous characteristics. A significance level of P < .05 was used for these tests. A univariate Cox regression analysis was performed using death as the outcome. Covariates that were significant at P < .25 were included in the multivariate logistic regression. OS functions and recurrence control functions were calculated according to the Kaplan–Meier method. Differences in OS and recurrence control were assessed by using the log-rank test.
Clinicopathologic Findings in the Esophagectomy Specimens
The clinicopathologic characteristics of the 99 patients with T1 esophageal adenocarcinoma are summarized in Table 1. Positive margins (microscopic residual disease) were identified in 1 patient's specimen (1%; 1 of 99 patients). LVI status was examined in 89 patients (89%), and LVI was observed in 17 patients (19%). Eleven resected specimens could not be retrieved for reassessment of the pathology evaluation and designation of LVI status. No patient received adjuvant chemotherapy or chemoradiation after surgery.
|No. of patients (%)|
|Total no. of cases||48||51||99|
|Men||41 (85.4)||46 (90.2)||87 (87.9)|
|Women||7 (14.6)||5 (9.8)||12 (12.1)|
|Type of esophagectomy||.368|
|Right transthoracic (Ivor-Lewis)||17 (35.4)||24 (47.1)||41 (41.4)|
|Transhiatal||27 (56.3)||23 (45.1)||50 (50.5)|
|Total (3-field technique)||2 (4.2)||4 (7.8)||6 (6.1)|
|Minimally invasive esophagectomy||1 (2.1)||0 (0)||1 (1)|
|N/A||1 (2.1)||0 (0)||1 (1)|
|Location of the primary cancer|
|Neck/upper/middle||4 (8.3)||4 (7.8)||8 (8.1)||1.000|
|Lower esophagus/GEJ||44 (91.7)||47 (92.2)||91 (91.9)|
|No||1 (2.1)||5 (9.8)||6 (6.1)|
|Yes||47 (97.9)||46 (90.2)||93 (93.9)|
|Postsurgery cancer length, cm||<.001|
|≤1.2||33 (70.2)||13 (25.5)||46 (46.9)|
|>1.2||14 (29.8)||38 (74.5)||52 (53.1)|
|Well differentiated||17 (36.2)||3 (5.9)||20 (20.4)|
|Moderately differentiated||26 (55.3)||28 (54.9)||54 (55.1)|
|Poorly differentiated||4 (8.5)||20 (39.2)||24 (24.5)|
|Absent||41 (93.2)||31 (68.9)||72 (80.9)|
|Present||3 (6.8)||14 (31.1)||17 (19.1)|
|Lymph node metastasis||.006|
|Absent||46 (95.8)||39 (76.5)||85 (85.9)|
|Present||2 (4.2)||12 (23.5)||14 (14.1)|
|Median [mean±SD]||65 [64±10.9]||68 [66±9.5]||68 [65±10.2]|
|No. of LNs resected||.666|
|Median [mean±SD]||10.5 [12.8±9.4]||11 [13.8±11.7]||11 [13.3±10.6]|
Among patients who had tumors with different depth of invasion, there were no significant differences in age, sex, location of primary tumor, type of esophagectomy procedure, incidence of Barrett esophagus, or numbers of lymph nodes examined in the resected specimen. However, LVI, the length of the cancer (>1.2 cm), grade of differentiation, and LNM were associated significantly with the depth of cancer invasion (T1a vs T1b) (Table 1). Among the patients who had pT1a cancer, only 2 patients (4%) had LNM; whereas, among the patients who had pT1b cancer, 12 patients (23%) had LNM.
Relation Between Histopathologic Findings and Prognosis
The mean follow-up was 60 months (range, 0.3-198 months). Seventy-four patients (75%) patients were alive without recurrent disease, 1 patient was alive with distant metastases, and 24 patients (24%) died.
In the univariate analysis, age >65 years, cancer length >1.2 cm, poorly differentiated grade, pathologic T1b tumors, the presence of LVI, and LNM were identified as significant prognostic factors for poorer OS (Table 2). However, the multivariate analysis demonstrated that LNM (P = .03) and age >65 years (P = .04) were independent factors that were predictive of a poor OS. The presence of LVI was of borderline significant as an independent factor that was predictive of poor OS (P = .05) (Table 3).
|Characteristic||No.||P||HR||95% CI HR: lower-upper|
|Type of esophagectomy||.542|
|Right transthoracic (Ivor–Lewis) (reference)||41||1.000|
|Total (3–field technique)||6||.890||1.111||0.250–4.93|
|Location of the primary cancer|
|Op tumor size, cm|
|Pathologic tumor group|
|No. of LNs checked||99||.625||1.009||0.972–1.048|
|PreOp or PostOp Barrett|
|Pathologic lymph node group|
|Age at surgery date, y|
|Characteristic||No.||P||HR||95% CI HR: lower-upper|
|Location of the primary cancer|
|Postsurgery cancer length, cm|
|Well differentiated (reference)||18||1.000|
|Pathologic tumor group|
|Pathologic lymph node group|
|Age at surgery date, y|
OS in patients with T1a cancer versus T1b cancer
The patients who had T1b cancer (n = 51 patients) had a significantly worse 5-year OS rate compared with patients who had T1a cancer (n = 48 patients; 62% vs 88% at 5 years; P = .001) (Fig. 1).
OS in patients with T1 cancer with LVI versus without LVI
Patients who had cancer with LVI (n = 17 patients) had a significantly worse 5-year OS rate (36%) compared with patients who had cancer without LVI (n = 72 patients; 85%; P < .0001) (Fig. 2).
OS in patients with T1a versus T1b cancer without LVI versus T1b cancer with LVI
The 5-year OS rate of patients who had T1a cancer (n = 44 patients) versus T1b cancer (n = 31 patients) without LVI (90% vs 77%, respectively; P = .078) did not differ significantly (Fig. 3). However, the 5-year OS rate for patients who had T1b cancer (n = 14 patients) without LVI versus T1b cancer with LVI (77% vs 27%, respectively; P = .006), was statistically significantly different (Fig. 3).
Recurrence-free survival in patients with T1a versus T1b cancer without LVI versus T1b cancer with LVI
The 5-year recurrence-free survival rate was significantly different for patients in these 3 groups: 100% for patients who had T1a cancer, 74% for patients who had T1b cancer without LVI (P = .006), and only 35% for patients who had T1b cancer with LVI (P <.0001). All 4 deaths among the patients with T1a cancer were related to problems other than cancer. However, the recurrent cancer was attributed as a cause in 5 of 7 deaths among the patients who had T1b cancer without LVI and in 4 of 8 deaths among the patients who had T1b cancer with LVI. Patients who had T1b cancer with LVI had a 5-year recurrence rate (5 of 14 patients; 36%) that was twice as frequent as the rate for patients who had cancer without LVI (5of 31 patients; 16%). However, the difference wasnot statistically significant (P = .244). Among 25 patients who developed recurrent disease, 96% (24 of 25 patients) had distant metastases (Table 4).
|Variable||No. of patients T1b (n = 51)||Total|
|T1a (n = 48)||LVI (n = 14)||No LVI (n = 31)||Unknown LVI (n = 6)|
|Causes of death||5||8||8||3||24|
|Unrelated to cancer||5||4||3||3||15|
|Type of recurrence||10†|
|Distant metastasis only||1||2||3|
|Distant and locoregional||3||3||6|
Effect of LNM and LVI on OS
Patients without LVI and/or LNM fared much better than patients with either LVI or LNM (P < .001) (Fig. 4). The 5-year OS rate was only 37% for patients with LVI (n = 12 patients), 48% for patients with LNM (n = 7 patients), and 30% for patient with both LVI and LNM (n = 7 patients); however, these differences were not statistically significant. Patients without any of these risk factors (n = 65 patients) had a significantly higher 5-year OS rate (88%).
OS by T1aN0 versus T1bN0 cancer without LVI versus T1bN0 cancer with LVI
The 5-year OS rate was 89% for patients who had T1aN0 cancer (n = 43) and 85% (not significant) for patients who had T1bN0 cancer without LVI (n = 25 patients). However, both of these groups had better OS rates than the patients who had T1bN0 cancer with LVI (23%; n = 9 patients; P < .0001).
Future advances in the treatment of early esophageal cancer will come from efforts to individualize therapy based on information derived from a variety of sources. It is essential to identify low-risk cancers, because some patients with early cancer have the best chance of being treated successfully by complete local excision (or ablation), thus maximizing the preservation of organ function. The other side of this is the need to identify high-risk patients who are treated best by esophagectomy and, if necessary, adjuvant therapy.
In our study, we observed that LVI, cancer length (>1.2 cm), poorly differentiated cancer, and LNM were correlated significantly with the depth of penetration into the submucosa (T1b). In addition, the rate of LNM in patients with T1b cancer (23%) was 6-fold higher than that in patients with T1a cancer (4%). These findings are consistent with those published previously by others.12-15, 19 A multivariate analysis of our data demonstrated that LNM was an independent prognostic factor of poor OS in patients with T1 cancer (P = .03). These results also are agreement with previous reports by other authors.14, 19, 20 Age >65 years also predicted poor OS (P = .045). The implication of this finding is not clear and should be the subject of future investigation.
It is worth noting that our data indicated a borderline significance for the presence of LVI to predict poor OS (P = .05). Similar findings have been reported in patients with squamous cell carcinoma,21–23 gastric cancer,24 adenocarcinoma of the cardia, and subcardial gastric cancer.25 The current study was a retrospective analysis, and all patients were accrued before study entry and followed, some of them for 17 years. The hazard ratio was 5.16, and the hazard rates and attrition rates are shown in Tables 1 and 2. The sample size was 72 patients without LVI and 17 patients with LVI, and the α = .005 with a 2-tailed test. This resulted in a power of 22%, indicating that 22% of studies with this sample size and this population of patients are expected to yield a significant effect, rejecting the null hypothesis that the hazard ratio is 1.0. Because we used a P value <.001, our study falls easily into those 22% of studies that are likely to demonstrate a significant difference. This suggests that our data are highly reliable (because the P value is stronger than the value resulting from the assumption made), and the statistical difference observed holds up with a high degree of rigor.
The prevalence of LVI in our patients with T1 cancer was 19% (17 of 89 patients), which was similar to the rate reported by Stein et al (12%; 7 of 60 patients with T1 disease). However, unlike our data, which indicated borderline significance for LVI, the multivariate analysis reported by Stein et al did not identify LVI as a significant prognosticator in patients with esophageal adenocarcinoma (P = .9). Several factors may account for their reported outcome, including a short duration of follow-up (36 months), 18% of their patients had residual cancer (>R0 resection), and their study included only a few patients with T1 cancer (most patients had >T1 cancer; 66% had T2-T4 cancer). Thus, to our knowledge, we are reporting the influence of LVI in the largest T1 adenocarcinoma patient population studied to date.
EUS has been highly reliable in correctly identifying T1a cancer (accuracy, >90%).26–29 For EUS-determined T1 cancers, EMR is emerging as a diagnostic tool (to distinguish reliably between T1a and T1b cancers and to assess LVI status) and as a therapeutic tool for T1a cancers.30–33 However, the sensitivity of EUS to detect LNM varies considerably (from 60% to 90%).26, 28, 34 Lymph node assessment by EUS has relied on criteria that are not embraced uniformly.35 In our study, irrespective of lymph node status, patients who had T1b cancer had a poorer 5-year OS rate than patients who had T1a cancer (62% vs 88%, respectively; P = .001). Patients who had cancer associated with LVI had a poorer 5-year OS rate than patients who had cancer without LVI (36% vs 85%, respectively; P < .0001).
Most important, we were able to use LVI status to subclassify T1b cancer into 2 groups. Patients who had T1b cancer with LVI had a worse 5-year OS rate (27%) than patients who had T1b cancer without LVI (70%; P = .006). Thus, patients who had T1b cancer with LVI had a worse 5-year OS rate, as reported by others, that patients who had resected T2N0 cancer.10, 17, 19, 36, 37 In contrast, patients who had T1b cancer without LVI had a 5-year OS rate comparable to the rate in patients who had T1a cancer (77% vs 90%, respectively; P = .078). The 5-year OS rate for our patients with T1a cancer was comparable to the 80% to 90% rate reported by others.17, 38 This observation may allow new possibilities in the treatment of patients with T1b esophageal cancer. Acknowledging that any paradigm that might result in the under treatment of patients with T1b cancer would be completely unacceptable, our data are provocative enough to propose further research with a new strategy (preoperative or postoperative therapy based on a 27% 5-year OS rate).
On the basis of our data, it might of interest to consider the following strategy: An EUS diagnosis of T1N0 (a or b) cancer is to be followed by a diagnostic/therapeutic EMR in conjunction with assessment of LVI; if the cancer is T1a with negative EMR margins, then the patient is to be surveyed closely; if the cancer is T1b with LVI, then an esophagectomy and lymphadenectomy will be recommended. However, if the cancer is T1b with negative EMR margins but also is lacking LVI, then sparing the esophagus should be considered, and this group of patients should be surveyed closely. We believe that this strategy for T1b cancer without LVI currently is not workable (and remains theoretical); because, in the current study, the rate of LNM in patients who had T1b cancer without LVI was 19%. This rate is too high to forego esophagectomy/lymphadenectomy, even though the rate of subsequent relapse in patients who had T1b cancer without LVI was 16% compared with patients who had T1b cancer with LVI (36%; P = .244). However, if a method with high sensitivity to detect LNM can be established, then such a strategy could be investigated. The other possibility is to identify molecular correlates that are highly predictive of LNM.
In our series, the presence of LVI, even without LNM, had a considerable influence on OS. Patients who had T1bN0 cancer with LVI had a 5-year OS rate of only 23%, which was similar to the OS rate among patients who had LNM (with or without LVI). The 5-year survival rate of patients with no adverse finding (LVI or LNM) was ≥85% (P < .001).
In conclusion, the current results demonstrate that LVI status subclassifies patients with T1b esophageal adenocarcinoma. However, patients who have T1b tumors without LVI still should be offered esophagectomy and lymphadenectomy because of a relatively high rate of LNM. Until LNM can be detected with high sensitivity before surgery or molecular markers for prognosis can subclassify such patients, a new treatment paradigm cannot be implemented. Patients who have T1b cancer with LVI still are treated best by esophagectomy, possibly with the addition of postoperative or preoperative systemic therapy. However, a prospective clinical trial is warranted. Further research to individualize therapy in patients with early esophageal adenocarcinoma is warranted.
- 9Superiority of trimodality therapy to surgery alone in esophageal cancer: results of CALGB 9781. ASCO Meeting Abstracts. 2006; 24. Abstract 4012., , , et al.
- 28The impact of endoscopic ultrasound and computed tomography on the TNM staging of early cancer in Barrett's esophagus. Am J Gastroenterol. 2006; 101: 2223–2229., , , , .Direct Link: