Int J Colorectal Dis DOI 10.1007/s00384-015-2164-3
ORIGINAL ARTICLE
A clinical model for predicting lymph node metastasis in submucosal invasive (T1) colorectal cancer Fernando Macias-Garcia & Catuxa Celeiro-Muñoz & Lucia Lesquereux-Martinez & Francisco Gude-Sampedro & Laura Uribarri-Gonzalez & Ihab Abdulkader & Ana Alvarez-Castro & J. Enrique Dominguez-Muñoz
Accepted: 9 February 2015 # Springer-Verlag Berlin Heidelberg 2015
Abstract Background No single histopathological feature of submucosal invasive colorectal cancer (T1-CRC) can reliably predict the risk for lymph node metastasis (LNM). Aim The purpose of the study was to develop a prediction model of LNM in T1-CRC. Methods Ninety-seven surgically resected T1-CRC at our institution were retrospectively evaluated. Morphology, localization, grading, mode of growth, presence of background adenoma, lymphoid infiltration, angiolymphatic invasion, budding, and depth of invasion were assessed. Mortality and morbidity related to surgery were also evaluated. Benefit-risk balance was assessed according to the presence of severe complications and to the presence of LNM.
F. Macias-Garcia (*) : L. Uribarri-Gonzalez : A. Alvarez-Castro : J. E. Dominguez-Muñoz Gastroenterology Department, University Hospital of Santiago de Compostela, c/Choupana s/n, 15706 Santiago de Compostela, Spain e-mail: [email protected] C. Celeiro-Muñoz : I. Abdulkader Pathology Department, University Hospital of Santiago de Compostela, Santiago de Compostela, Spain L. Lesquereux-Martinez Digestive Surgery Department, University Hospital of Santiago de Compostela, Santiago de Compostela, Spain F. Gude-Sampedro Epidemiology Department, University Hospital of Santiago de Compostela, Santiago de Compostela, Spain F. Macias-Garcia : L. Uribarri-Gonzalez : A. Alvarez-Castro : J. E. Dominguez-Muñoz Foundation for Research in Digestive Diseases (FIENAD), Santiago de Compostela, Spain
Results Fourteen cases had LNM (14 %). Eight patients (8 %) presented severe surgical complications and there were two deaths (2 %). Infiltrative growth pattern (OR 31.91, 95 % CI 2.37–428.36; p=0.009) and the absence of lymphoid infiltrate (OR 28.75; 95 % CI 2.13–388.37; p=0.011) were the only variables independently associated with LNM in the multivariate analysis. Both variables were included in the prediction model together with sessile morphology (OR 4.88; 95 % CI 0.81–29.3; p=0.083) and poorly differentiated carcinoma (OR 11.77; 95 % CI 0.77–179.83; p=0.076). A 0–100 score was developed (infiltrative growth pattern: no=0, yes=33; lymphoid infiltrate: no=29, yes=0; sessile morphology: no=0, yes=15; poorly differentiated: no=0, yes=23). Cutoff point to indicate additional surgery was set in 35 points (i.e., 10 % risk LNM). Discrimination of the prediction model was excellent (AUC 0.90; 95 % CI 0.81–0.99). Conclusion Combined evaluation of infiltrative growth pattern, lymphoid infiltration, poorly differentiated carcinoma, and sessile appearance showed good performance for discriminating T1-CRC patients with LNM. The benefit-risk balance was in favor of surgery when at least two of these criteria were present. Keywords Early colorectal cancer (CRC) . Submucosal invasion (T1) . Lymph node metastasis (LNM) . Risk prediction model
Background and aims Early colorectal cancer (CRC) has been defined as carcinoma in which invasion is limited to the mucosa (Tis) or submucosa (T1), regardless of the presence or absence of lymph node metastases (LNM) [1]. While endoscopic resection of intramucosal carcinoma is accepted as curative therapy because there is no risk of LNM [2], nodal metastatic spread may occur in up 7–15 % of T1-CRC [3–6] so endoscopic
Int J Colorectal Dis
resection should be carefully performed as monotherapy only in selected cases. However, most of the T1-CRC cases are not associated with LNM and serious complications related to surgery could happen. The best treatment option should achieve the optimal combination between oncological radicalness and the highest quality of life [7]. The risk of LNM in T1-CRC has been previously related to several histopathological factors such as angiolymphatic invasion [3, 5, 6, 8–14], poorly differentiated carcinoma [8], and deep submucosal invasion [6, 15, 16]. Tumor budding has also been evaluated for its relationship with LNM [3, 5, 14, 17, 18]. The size and morphology of the tumor, its mode of growth, the absence of background adenoma [8], or lymphoid infiltration [9, 19] are other features also potentially related to LNM. However, it is well known that no single histopathological feature of CRC can reliably predict the risk of LNM [10] and actually, there are no universally accepted indications for surgical resection after the endoscopic resection of a submucosal invasive cancer. So, risk prediction models designed from a combination of risk factors for LNM should be proposed and evaluated to establish which would be the optimal combination of these predictive factors to better predict the real risk [11]. Therefore, our aim was to develop a prediction model of LNM in T1-CRC in order to define which patients should be submitted to additional colectomy after an initial endoscopic resection of the tumor. Secondary aim was to analyze the morbidity and mortality related to surgery in order to assess the risk-benefit balance of this procedure.
Patients and methods All patients who underwent a surgical resection because of a T1-CRC at the University Hospital of Santiago de Compostela, Spain, between January 2000 and December 2011 were retrospectively evaluated. Patients who were previously diagnosed with familial adenomatous polyposis or inflammatory bowel disease, those who underwent surgery after preoperative chemoradiation therapy, and those who underwent surgery because of synchronous advanced CRC were excluded. The surgical procedures were performed by laparotomy or laparoscopy depending on the surgeon’s preferences and the patient’s surgical history. Selected histopathological features of the primary tumor or at its front of invasion, which is considered to be an area that specially manifests the tumor aggressiveness, were assessed regarding to the presence or absence of LNM. Grading of the tumor was determined according to WHO criteria and categorized in two groups for the analysis: well or moderately differentiated (grade 1 or 2) and poorly differentiated (grade 3) [20].
Background adenoma was defined as the presence of an adenomatous component microscopically contiguous to resected cancer. Modes of growth at the front of invasion of the cancer were classified into three types: expansive type (submucosal invasion consisting of solid cancer nests without fibrous stroma), infiltrative type (characterized by widespread dissemination of tumor cells into normal tissue structures with loss of a clear boundary between tumor and host tissue) [21], and mixed type. The lymphoid infiltration at the submucosal invasive area of the cancer was classified as either negative (no or little infiltration) or positive (infiltration with follicular structure or more lymphoid cells than the number of cancer cells). Angiolymphatic invasion was defined as the presence of cancer cells within endothelial-lined channels. Tumor budding is considered to be a pathological characteristic corresponding to the initial phase of tumor invasion and it has been defined as an isolated cancer cell or a small cluster of fewer than five cancer cells at the front of invasion. Budding was divided into low and high grade (>10 budding foci viewed at 250-fold magnification), according to Ueno’s grading system [18]. The depth of submucosal invasion was measured according to the morphology of the tumor in two ways. Haggitt’s classification was evaluated only in pedunculated tumors [15]. Haggitt’s classification divides polyps with invasive carcinoma in four levels depending on the depth of invasion. Level 4 refers to those carcinomas that invade the submucosa below the stalk of a pedunculated tumor. Therefore, all sessile tumors with these invasive characteristics correspond to that level 4, so this classification was not evaluated in these lesions. Depth of stalk invasion was measured as the vertical distance from the neck of the stalk (level 2 as defined by Haggitt’s classification) to the deepest portion of invasion. Advanced Haggitt’s stage was defined as depth of submucosal invasion ≥level 3. In sessile lesions, Kudo’s classification [22] was evaluated as follows to define the relative invasion depth of the submucosal layer: Sm1 (infiltration into the upper third of the submucosal layer and divided in Sm1a, Sm1b, and Sm1c), Sm2 (middle third), and Sm3 (lower third). Advanced Kudo’s stage was defined as depth of submucosal invasion ≥Sm2 (i.e., invasion greater than 300 μm from the muscularis mucosae). Mortality and morbidity were defined respectively as death or complications occurring either within 30 days following surgery or during the hospital stay. Complications were classified according to Clavien-Dindo’s classification [23]. Oncological benefit was measured by the LNM rate. The benefit-risk balance of this procedure was assessed with the assumption that the short-term risk (severe complications of grade 3–4 or death) was as serious as the long-term risk (presence of positive lymph nodes on the resected specimen) [24]. The flow chart of the 97 pT1-CRC cases included in the study is shown in Fig. 1. Primary surgical resection was
Int J Colorectal Dis
performed in 65 patients while 32 patients underwent primary endoscopic resection followed by additional surgical resection. Indications for subsequent surgery were based always on the presence of at least one of the following risk factors: positive or undetermined resection margins (including piecemeal resection), poorly differentiated adenocarcinoma, or, in most cases, deep submucosal invasion (Kudo’s level ≥Sm2 in sessile tumors and Haggitt’s level ≥3 in non-sessile tumors). Statistical analyses The chi-squared test and the Student’s t test were used to assess the relationship between histopathological factors and LNM. Logistic regression analysis was used to estimate the probability of LNM. The variables included, because they proved significant in the univariate analyses, were poorly differentiated carcinoma, infiltrative growth pattern, absence of lymphoid infiltration, lymphatic invasion, and high-grade budding; in addition, the inclusion of morphology was forced. Beginning with a model containing all these potential covariates, Akaike Information Criterion (AIC) was used to select the Bbest^ subset of predictor variables. Different aspects of the model performance were then studied, including calibration and discrimination. Calibration was assessed using the Brier score and the Hosmer-Lemeshow goodness-of-fit test [25]. The receiver operating characteristics (ROC) curves and the corresponding area under the ROC curve (AUC) were calculated to test for discrimination. Statistical analyses were performed using R [26]. Statistical significance was established at p
ORIGINAL ARTICLE
A clinical model for predicting lymph node metastasis in submucosal invasive (T1) colorectal cancer Fernando Macias-Garcia & Catuxa Celeiro-Muñoz & Lucia Lesquereux-Martinez & Francisco Gude-Sampedro & Laura Uribarri-Gonzalez & Ihab Abdulkader & Ana Alvarez-Castro & J. Enrique Dominguez-Muñoz
Accepted: 9 February 2015 # Springer-Verlag Berlin Heidelberg 2015
Abstract Background No single histopathological feature of submucosal invasive colorectal cancer (T1-CRC) can reliably predict the risk for lymph node metastasis (LNM). Aim The purpose of the study was to develop a prediction model of LNM in T1-CRC. Methods Ninety-seven surgically resected T1-CRC at our institution were retrospectively evaluated. Morphology, localization, grading, mode of growth, presence of background adenoma, lymphoid infiltration, angiolymphatic invasion, budding, and depth of invasion were assessed. Mortality and morbidity related to surgery were also evaluated. Benefit-risk balance was assessed according to the presence of severe complications and to the presence of LNM.
F. Macias-Garcia (*) : L. Uribarri-Gonzalez : A. Alvarez-Castro : J. E. Dominguez-Muñoz Gastroenterology Department, University Hospital of Santiago de Compostela, c/Choupana s/n, 15706 Santiago de Compostela, Spain e-mail: [email protected] C. Celeiro-Muñoz : I. Abdulkader Pathology Department, University Hospital of Santiago de Compostela, Santiago de Compostela, Spain L. Lesquereux-Martinez Digestive Surgery Department, University Hospital of Santiago de Compostela, Santiago de Compostela, Spain F. Gude-Sampedro Epidemiology Department, University Hospital of Santiago de Compostela, Santiago de Compostela, Spain F. Macias-Garcia : L. Uribarri-Gonzalez : A. Alvarez-Castro : J. E. Dominguez-Muñoz Foundation for Research in Digestive Diseases (FIENAD), Santiago de Compostela, Spain
Results Fourteen cases had LNM (14 %). Eight patients (8 %) presented severe surgical complications and there were two deaths (2 %). Infiltrative growth pattern (OR 31.91, 95 % CI 2.37–428.36; p=0.009) and the absence of lymphoid infiltrate (OR 28.75; 95 % CI 2.13–388.37; p=0.011) were the only variables independently associated with LNM in the multivariate analysis. Both variables were included in the prediction model together with sessile morphology (OR 4.88; 95 % CI 0.81–29.3; p=0.083) and poorly differentiated carcinoma (OR 11.77; 95 % CI 0.77–179.83; p=0.076). A 0–100 score was developed (infiltrative growth pattern: no=0, yes=33; lymphoid infiltrate: no=29, yes=0; sessile morphology: no=0, yes=15; poorly differentiated: no=0, yes=23). Cutoff point to indicate additional surgery was set in 35 points (i.e., 10 % risk LNM). Discrimination of the prediction model was excellent (AUC 0.90; 95 % CI 0.81–0.99). Conclusion Combined evaluation of infiltrative growth pattern, lymphoid infiltration, poorly differentiated carcinoma, and sessile appearance showed good performance for discriminating T1-CRC patients with LNM. The benefit-risk balance was in favor of surgery when at least two of these criteria were present. Keywords Early colorectal cancer (CRC) . Submucosal invasion (T1) . Lymph node metastasis (LNM) . Risk prediction model
Background and aims Early colorectal cancer (CRC) has been defined as carcinoma in which invasion is limited to the mucosa (Tis) or submucosa (T1), regardless of the presence or absence of lymph node metastases (LNM) [1]. While endoscopic resection of intramucosal carcinoma is accepted as curative therapy because there is no risk of LNM [2], nodal metastatic spread may occur in up 7–15 % of T1-CRC [3–6] so endoscopic
Int J Colorectal Dis
resection should be carefully performed as monotherapy only in selected cases. However, most of the T1-CRC cases are not associated with LNM and serious complications related to surgery could happen. The best treatment option should achieve the optimal combination between oncological radicalness and the highest quality of life [7]. The risk of LNM in T1-CRC has been previously related to several histopathological factors such as angiolymphatic invasion [3, 5, 6, 8–14], poorly differentiated carcinoma [8], and deep submucosal invasion [6, 15, 16]. Tumor budding has also been evaluated for its relationship with LNM [3, 5, 14, 17, 18]. The size and morphology of the tumor, its mode of growth, the absence of background adenoma [8], or lymphoid infiltration [9, 19] are other features also potentially related to LNM. However, it is well known that no single histopathological feature of CRC can reliably predict the risk of LNM [10] and actually, there are no universally accepted indications for surgical resection after the endoscopic resection of a submucosal invasive cancer. So, risk prediction models designed from a combination of risk factors for LNM should be proposed and evaluated to establish which would be the optimal combination of these predictive factors to better predict the real risk [11]. Therefore, our aim was to develop a prediction model of LNM in T1-CRC in order to define which patients should be submitted to additional colectomy after an initial endoscopic resection of the tumor. Secondary aim was to analyze the morbidity and mortality related to surgery in order to assess the risk-benefit balance of this procedure.
Patients and methods All patients who underwent a surgical resection because of a T1-CRC at the University Hospital of Santiago de Compostela, Spain, between January 2000 and December 2011 were retrospectively evaluated. Patients who were previously diagnosed with familial adenomatous polyposis or inflammatory bowel disease, those who underwent surgery after preoperative chemoradiation therapy, and those who underwent surgery because of synchronous advanced CRC were excluded. The surgical procedures were performed by laparotomy or laparoscopy depending on the surgeon’s preferences and the patient’s surgical history. Selected histopathological features of the primary tumor or at its front of invasion, which is considered to be an area that specially manifests the tumor aggressiveness, were assessed regarding to the presence or absence of LNM. Grading of the tumor was determined according to WHO criteria and categorized in two groups for the analysis: well or moderately differentiated (grade 1 or 2) and poorly differentiated (grade 3) [20].
Background adenoma was defined as the presence of an adenomatous component microscopically contiguous to resected cancer. Modes of growth at the front of invasion of the cancer were classified into three types: expansive type (submucosal invasion consisting of solid cancer nests without fibrous stroma), infiltrative type (characterized by widespread dissemination of tumor cells into normal tissue structures with loss of a clear boundary between tumor and host tissue) [21], and mixed type. The lymphoid infiltration at the submucosal invasive area of the cancer was classified as either negative (no or little infiltration) or positive (infiltration with follicular structure or more lymphoid cells than the number of cancer cells). Angiolymphatic invasion was defined as the presence of cancer cells within endothelial-lined channels. Tumor budding is considered to be a pathological characteristic corresponding to the initial phase of tumor invasion and it has been defined as an isolated cancer cell or a small cluster of fewer than five cancer cells at the front of invasion. Budding was divided into low and high grade (>10 budding foci viewed at 250-fold magnification), according to Ueno’s grading system [18]. The depth of submucosal invasion was measured according to the morphology of the tumor in two ways. Haggitt’s classification was evaluated only in pedunculated tumors [15]. Haggitt’s classification divides polyps with invasive carcinoma in four levels depending on the depth of invasion. Level 4 refers to those carcinomas that invade the submucosa below the stalk of a pedunculated tumor. Therefore, all sessile tumors with these invasive characteristics correspond to that level 4, so this classification was not evaluated in these lesions. Depth of stalk invasion was measured as the vertical distance from the neck of the stalk (level 2 as defined by Haggitt’s classification) to the deepest portion of invasion. Advanced Haggitt’s stage was defined as depth of submucosal invasion ≥level 3. In sessile lesions, Kudo’s classification [22] was evaluated as follows to define the relative invasion depth of the submucosal layer: Sm1 (infiltration into the upper third of the submucosal layer and divided in Sm1a, Sm1b, and Sm1c), Sm2 (middle third), and Sm3 (lower third). Advanced Kudo’s stage was defined as depth of submucosal invasion ≥Sm2 (i.e., invasion greater than 300 μm from the muscularis mucosae). Mortality and morbidity were defined respectively as death or complications occurring either within 30 days following surgery or during the hospital stay. Complications were classified according to Clavien-Dindo’s classification [23]. Oncological benefit was measured by the LNM rate. The benefit-risk balance of this procedure was assessed with the assumption that the short-term risk (severe complications of grade 3–4 or death) was as serious as the long-term risk (presence of positive lymph nodes on the resected specimen) [24]. The flow chart of the 97 pT1-CRC cases included in the study is shown in Fig. 1. Primary surgical resection was
Int J Colorectal Dis
performed in 65 patients while 32 patients underwent primary endoscopic resection followed by additional surgical resection. Indications for subsequent surgery were based always on the presence of at least one of the following risk factors: positive or undetermined resection margins (including piecemeal resection), poorly differentiated adenocarcinoma, or, in most cases, deep submucosal invasion (Kudo’s level ≥Sm2 in sessile tumors and Haggitt’s level ≥3 in non-sessile tumors). Statistical analyses The chi-squared test and the Student’s t test were used to assess the relationship between histopathological factors and LNM. Logistic regression analysis was used to estimate the probability of LNM. The variables included, because they proved significant in the univariate analyses, were poorly differentiated carcinoma, infiltrative growth pattern, absence of lymphoid infiltration, lymphatic invasion, and high-grade budding; in addition, the inclusion of morphology was forced. Beginning with a model containing all these potential covariates, Akaike Information Criterion (AIC) was used to select the Bbest^ subset of predictor variables. Different aspects of the model performance were then studied, including calibration and discrimination. Calibration was assessed using the Brier score and the Hosmer-Lemeshow goodness-of-fit test [25]. The receiver operating characteristics (ROC) curves and the corresponding area under the ROC curve (AUC) were calculated to test for discrimination. Statistical analyses were performed using R [26]. Statistical significance was established at p
