- Email: [email protected]
OC-0454 BLOOD BIOMARKERS PREDICT RESPONSE TO CHEMORADIATION: A PROSPECTIVE STUDY ON 260 RECTAL CANCER PATIENTS
S181
ESTRO 31
Conclusions: This study has demonstrated a significant dose-response relationship for tumour regression after preoperative CRT for locally advanced rectal cancer for tumour dose levels in the range of 50.4 to 70 Gy, which are higher than those usually considered. The present analysis may inform the design of dose-escalation studies and watchand-wait protocols. OC-0453 POSITIONAL VARIATION OF INVOLVED MESORECTAL LYMPH NODES DURING RT OF RECTAL CANCER B. Hollmann1, J. Nijkamp1, J.J. Sonke1, C.A.M. Marijnen2 1 The Netherlands Cancer Institute - Antoni van Leeuwenhoek Hospital, Radiation Oncology, Amsterdam, The Netherlands 2 Leiden University Medical Center, Clinical Oncology, Leiden, The Netherlands Purpose/Objective: Pre-operative chemo-RT is generally given for locally advanced rectal cancer. To improve tumor down-staging and increase the chance on a pathological complete response, several boost studies have been initiated recently, in which the tumor and involved lymph nodes are treated to a higher dose. However, knowledge on positional variation of the tumor and involved lymph nodes for margin calculation is lacking. The aim of this study was to quantify the positional variation of pathologically enlarged lymph nodes during chemo-RT of locally advanced rectal cancer using repeat CT scans. Materials and Methods: Repeat CT (rCT) scans were acquired for 6 patients treated with 25x2 Gy, prone on a flat table. A total of 10 scans were acquired for each patient, 1 for treatment planning (pCT), 5 daily during the first week, followed by 4 weekly scans. Pathologically enlarged lymph nodes (>0.5 cm) were identified on the pCT, and subsequently delineated on all rCT scans after bony anatomy registration. The center of mass (COM) was calculated for each delineation, and distances were calculated between the pCT and rCT delineations. The average lymph node shift was calculated for each scan time-point to indicate time-trends, and the group mean, systematic and random errors were calculated in LR, CC and AP direction. Correlation in position variation was also calculated between lymph nodes within a patient. Subsequent PTVboost margins were calculated assuming a background dose of 50 Gy in 25 fractions and simultaneously integrated boost dose of 8 Gy also in 25 fractions. Results: A total of 24 lymph nodes were identified (range 3-6 per patient) and delineated on a dataset of 52 CT scans. Eight scans were missing due to withdrawal from the study in week 2, 3, and 4 of 3 patients. There was a clear time trend indicating larger positional deviations from the planning situation towards the end of the treatment (Fig). Correlation between lymph nodes varied from almost perfect positive correlation (r=1) to negative correlation (r=-1) and also no correlation at all (r2=0) (Table). The time-trend resulted in a caudal group mean error of 0.4 cm. Systematic and random errors were substantial, especially in AP and CC direction (Table). Required margins for the PTVboost were 0.7, 1.1 and 1.1 cm in LR, CC and AP direction, respectively, assuming full correlation of motion between different lymph nodes within each patient. In the margin calculation, group mean errors and target delineation errors were not taken into account.
Conclusions: The positional variation of pathologically enlarged lymph nodes within the mesorectum is substantial, and should be taken into account in a PTVboost margin. A significant time-trend towards the end of treatment should be exploited, by for example re-scanning and re-planning for a sequential boost delivery. Our future research will focus on increasing the sample size and incorporating motion correlation in the PTVboost margin.
OC-0454 BLOOD BIOMARKERS PREDICT RESPONSE TO CHEMORADIATION: A PROSPECTIVE STUDY ON 260 RECTAL CANCER PATIENTS J. Buijsen1, R.G. Van Stiphout1, P.P.C.A. Menheere2, G. Lammering1, P. Lambin1 1 Maastricht Radiation Oncology (MAASTRO), Radiotherapy, Maastricht, The Netherlands 2 Maastricht University Medical Center (MUMC), Clinical Chemistry, Maastricht, The Netherlands Purpose/Objective: Chemoradiation (CRT) has been shown to lead to downsizing in an important part of rectal cancers. In 15-20% of cases a pathological complete response (pCR) occurs. This subgroup can follow a 'non surgical - wait and see policy'. In order to tailor treatment at an earlier stage, predictive models are being developed. Adding blood biomarkers may be attractive for prediction, as they can be collected very easily in clinical practice. The hypothesis of this study was that blood biomarkers related to tumor load, hypoxia and inflammation can help to predict response to CRT in rectal cancer. Materials and Methods: Between 5-2004 and 12-2010 260 patients with locally advanced rectal cancer who were planned to undergo CRT were prospectively entered into a biobank protocol (NCT01067872). Blood samples were drawn before start of CRT and processed and stored in a standardized way. Clinical data, including ypTN stage, were collected from our electronic medical records. Nine biomarkers were selected, based on a previously defined hypothesis, and measured in a standardized way by a certified lab: CEA, CA19-9, LDH, CRP, IL-6, IL-8, CA IX, osteopontin and vit D25. Outcome was analyzed in two ways: pCR vs. non-pCR and responders (defined as ypT0-2N0) vs. non-responders (all other ypTN stages). Correlations between biomarkers and between biomarkers and outcome were tested using Spearman’s rho. For univariate and multivariate analysis the Wilcoxon signed-rank test and the logistic regression was used respectively. Results: 260 patients were included in the study. After CRT 55 patients (21.1%) developed a pCR and 126 patients (48.5%) were classified as responder. Fig. 1 shows the correlations between the different biomarkers and between biomarkers and response. Boxes marked with a circle indicate a significant correlation (p<0.05). In univariate analysis CEA and osteopontin turned out to be significant predictors for pCR (p-values resp. <0.001 and 0.013), while there was a trend for IL-8 (p=0.087). Taking response as outcome CEA, IL-8 and
S182
osteopontin were significant predictors (p-values resp. <0.001, <0.001 and 0.006), and CA19-9 showed a trend (p=0.065). In multivariate analysis osteopontin was the strongest predictor for pCR (p=0.028) and CEA and IL-8 predicted for response (p=0.029 resp. 0.021).
ESTRO 31
internal validation performances were consistent for all outcomes: 0.54 for LR (0.52-0.57), 0.60 for DM (0.58-0.62), 0.70 for OS (0.690.72). Tumor diameter, pT-stage and pN-stage were predictive for all outcomes. Microscopic presence of disease was predictive for both DM and OS. Figure 1 shows the nomogram for OS. Conclusions: The provided models are internally validated and are able to accurately predict long-term outcome for pancreatic cancer patients with various clinical characteristics and given treatments. The nomograms allow decision support in daily clinical practice, i.e. selection of patients who may benefit most from specific treatment options, and stimulate also generation of new clinical hypotheses.
Conclusions: CEA, IL-8 and osteopontin were identified as strong predictive biomarkers for tumor response and PCR after CRT in rectal cancer. Thus, these blood biomarkers could lead to an additional accuracy of earlier developed prediction models using clinical variables and PET-information. OC-0455 MULTIVARIATE PREDICTION OF FOLLOW-UP OUTCOME FOR PATIENTS WITH PANCREATIC CANCER: A POOLED DATA ANALYSIS G.C. Mattiucci1, R.G.P.M. van Stiphout2, S. Alfieri3, F.A. Calvo4, J.B. Dubois5, B.W. Maidment 1116, R.C. Miller7, M. Reni8, N. Sharma9, M. Falconi10 1 Università Cattolica del Sacro Cuore-Policlinico A. Gemelli, Radiotherapy, Rome, Italy 2 MAASTRO GROW University Medical Centre, Radiation Oncology, Maastricht, The Netherlands 3 Università Cattolica S. Cuore, Surgery, Rome, Italy 4 Hospital General Universitario Gregorio Marañón, Oncology, Madrid, Spain 5 CRLC, Radiotherapy, Montpellier, France; 6 University of Maryland School of Medicine, Radiotherapy, Baltimore, USA 7 Mayo Clinic, Radiation Oncology, Rochester, USA 8 S. Raffaele Scientific Institute, Oncology, Milan, Italy 9 University of Maryland School of Medicine, Radiation Oncology, Baltimore, USA 10 University of Verona, Surgery, Verona, Italy Purpose/Objective: To develop accurate nomograms to predict local recurrence (LR), distant metastases (DM) and overall survival (OS) within 5 years of follow-up for pancreatic cancer patients. This multivariate approach should allow treatment decision support for individual patients and generation of new evidences and hypotheses. Materials and Methods: Clinical data (N =1120) from 9 different institutes (Baltimore, Rochester, Montpellier, Madrid, Salzburg, Verona, Campobasso, Milano, and Rome) were pooled for this analysis. Patients with different combinations of treatments (preoperative RT, IORT, postoperative RT, concomitant chemo, neoadjuvant chemo, adjuvant chemo, surgery) were retrospectively included. The common clinical variables which were clinically relevant and eligible according to data constraints (not more than 15% missing data and heterogeneity in the data) were: age, gender, tumor location (head, body, tail), surgical procedure (DCP, distal, total), tumor grade (I-IV), microscopic residual disease (no/yes), lymphadenectomy (no/yes), pilorus preserving technique (no/yes), tumor diameter (mm), pathological tumor stage and pathological nodal stage. Variables were first selected by univariate analysis with the log-rank test for the corresponding Kaplan-Meier curves. These variables were used as an input for multivariate Cox’s proportional hazards regression model. Performance of the models was expressed by the c-index, which has similar properties as the area under the ROC curve (0.5, random prediction and 1, perfect prediction). Validation performance was tested with a 5-fold cross-validation scheme to avoid overfitting. Results: In the pooled database occurrences of events within 5 years of follow-up were 46.4% LR (outcome available: N = 690), 65.0% DM (N = 676) and 30.8% OS (N = 1050). The accuracy of the developed models was good with training accuracies of 0.62 for LR (95% CI: 0.590.65), 0.64 for DM (0.62-0.67) and 0.73 for OS (0.71-0.76). The
Figure 1. Nomogram of overall survival for pancreatic cancer OC-0456 NEOADJUVANT CHEMORADIOTHERAPY VS SURGERY FOR PANCREATIC CANCER. A MULTI-CENTRE RANDOMISED PHASE II TRIAL T. Brunner1, H. Golcher2, H. Witzigmann3, L. Marti4, W. Bechstein5, C. Bruns6, J. Hauss7, S. Merkel2, R. Fietkau8, W. Hohenberger2 1 University of Oxford, Oncology, Oxford, United Kingdom 2 University Hospitals Erlangen, Surgery, Erlangen, Germany 3 Krankenhaus Dresden Friedrichstadt, Surgery, Dresden, Germany 4 Kantonsspital St. Gallen, Surgery, St. Gallen, Switzerland 5 University Hospitals Frankfurt, Surgery, Frankfurt am Main, Germany 6 University Hospitals Munich Großhadern, Surgery, Munich, Germany 7 University Hospitals Leipzig, Surgery, Leipzig, Germany 8 University Hospitals Erlangen, Radiation Oncology, Erlangen, Germany Purpose/Objective: Primary surgery is standard treatment for resectable pancreatic carcinoma, and median overal survival time [mOS] for patients [pts] with resected pancreatic cancer is about 20 months. But still most pts die from local relapse or metastases. Therefore, we investigated in the first randomised controlled multicentre trial on this topic whether neoadjuvant chemoradiotherapy [CRT] in resectable pancreatic carcinoma is superior compared to primary resection. Materials and Methods: Pts with histologically proven ductal adenocarcinoma of the pancreatic head and <180° contact to peripancreatic vessels were randomised into Group A - primary surgery or Group B – neoadjuvant CRT (55.4 Gy; gemcitabine & cisplatin), followed by surgery 6 weeks later. In 2005 a protocol amendment suggested adjuvant chemotherapy (gemcitabine). The primary endpoint was mOS. The proposed sample size of 254 eligible patients was planned to provide at least 80% power to detect a 4.33 month difference in mOS with a significance level of 5% in a 2-sided log-rank test. Results: Between 06/2003 and 12/2009, 73 pts were randomised in 8 centres (A n=37, B n=36). Due to slow recruitment the trial was closed early. At submission data of 68 pts (A/B n=34) could be evaluated, 3 pts withdrew consent (A/B n=2/1), 1 centre sent no data (A/B n=1). The allocated therapy was received by 33 pts in Group A and 30 pts in Group B. One pt (A) developed sepsis and died before surgery. Four pts refused the allocated CRT and underwent primary surgery.
ESTRO 31
Conclusions: This study has demonstrated a significant dose-response relationship for tumour regression after preoperative CRT for locally advanced rectal cancer for tumour dose levels in the range of 50.4 to 70 Gy, which are higher than those usually considered. The present analysis may inform the design of dose-escalation studies and watchand-wait protocols. OC-0453 POSITIONAL VARIATION OF INVOLVED MESORECTAL LYMPH NODES DURING RT OF RECTAL CANCER B. Hollmann1, J. Nijkamp1, J.J. Sonke1, C.A.M. Marijnen2 1 The Netherlands Cancer Institute - Antoni van Leeuwenhoek Hospital, Radiation Oncology, Amsterdam, The Netherlands 2 Leiden University Medical Center, Clinical Oncology, Leiden, The Netherlands Purpose/Objective: Pre-operative chemo-RT is generally given for locally advanced rectal cancer. To improve tumor down-staging and increase the chance on a pathological complete response, several boost studies have been initiated recently, in which the tumor and involved lymph nodes are treated to a higher dose. However, knowledge on positional variation of the tumor and involved lymph nodes for margin calculation is lacking. The aim of this study was to quantify the positional variation of pathologically enlarged lymph nodes during chemo-RT of locally advanced rectal cancer using repeat CT scans. Materials and Methods: Repeat CT (rCT) scans were acquired for 6 patients treated with 25x2 Gy, prone on a flat table. A total of 10 scans were acquired for each patient, 1 for treatment planning (pCT), 5 daily during the first week, followed by 4 weekly scans. Pathologically enlarged lymph nodes (>0.5 cm) were identified on the pCT, and subsequently delineated on all rCT scans after bony anatomy registration. The center of mass (COM) was calculated for each delineation, and distances were calculated between the pCT and rCT delineations. The average lymph node shift was calculated for each scan time-point to indicate time-trends, and the group mean, systematic and random errors were calculated in LR, CC and AP direction. Correlation in position variation was also calculated between lymph nodes within a patient. Subsequent PTVboost margins were calculated assuming a background dose of 50 Gy in 25 fractions and simultaneously integrated boost dose of 8 Gy also in 25 fractions. Results: A total of 24 lymph nodes were identified (range 3-6 per patient) and delineated on a dataset of 52 CT scans. Eight scans were missing due to withdrawal from the study in week 2, 3, and 4 of 3 patients. There was a clear time trend indicating larger positional deviations from the planning situation towards the end of the treatment (Fig). Correlation between lymph nodes varied from almost perfect positive correlation (r=1) to negative correlation (r=-1) and also no correlation at all (r2=0) (Table). The time-trend resulted in a caudal group mean error of 0.4 cm. Systematic and random errors were substantial, especially in AP and CC direction (Table). Required margins for the PTVboost were 0.7, 1.1 and 1.1 cm in LR, CC and AP direction, respectively, assuming full correlation of motion between different lymph nodes within each patient. In the margin calculation, group mean errors and target delineation errors were not taken into account.
Conclusions: The positional variation of pathologically enlarged lymph nodes within the mesorectum is substantial, and should be taken into account in a PTVboost margin. A significant time-trend towards the end of treatment should be exploited, by for example re-scanning and re-planning for a sequential boost delivery. Our future research will focus on increasing the sample size and incorporating motion correlation in the PTVboost margin.
OC-0454 BLOOD BIOMARKERS PREDICT RESPONSE TO CHEMORADIATION: A PROSPECTIVE STUDY ON 260 RECTAL CANCER PATIENTS J. Buijsen1, R.G. Van Stiphout1, P.P.C.A. Menheere2, G. Lammering1, P. Lambin1 1 Maastricht Radiation Oncology (MAASTRO), Radiotherapy, Maastricht, The Netherlands 2 Maastricht University Medical Center (MUMC), Clinical Chemistry, Maastricht, The Netherlands Purpose/Objective: Chemoradiation (CRT) has been shown to lead to downsizing in an important part of rectal cancers. In 15-20% of cases a pathological complete response (pCR) occurs. This subgroup can follow a 'non surgical - wait and see policy'. In order to tailor treatment at an earlier stage, predictive models are being developed. Adding blood biomarkers may be attractive for prediction, as they can be collected very easily in clinical practice. The hypothesis of this study was that blood biomarkers related to tumor load, hypoxia and inflammation can help to predict response to CRT in rectal cancer. Materials and Methods: Between 5-2004 and 12-2010 260 patients with locally advanced rectal cancer who were planned to undergo CRT were prospectively entered into a biobank protocol (NCT01067872). Blood samples were drawn before start of CRT and processed and stored in a standardized way. Clinical data, including ypTN stage, were collected from our electronic medical records. Nine biomarkers were selected, based on a previously defined hypothesis, and measured in a standardized way by a certified lab: CEA, CA19-9, LDH, CRP, IL-6, IL-8, CA IX, osteopontin and vit D25. Outcome was analyzed in two ways: pCR vs. non-pCR and responders (defined as ypT0-2N0) vs. non-responders (all other ypTN stages). Correlations between biomarkers and between biomarkers and outcome were tested using Spearman’s rho. For univariate and multivariate analysis the Wilcoxon signed-rank test and the logistic regression was used respectively. Results: 260 patients were included in the study. After CRT 55 patients (21.1%) developed a pCR and 126 patients (48.5%) were classified as responder. Fig. 1 shows the correlations between the different biomarkers and between biomarkers and response. Boxes marked with a circle indicate a significant correlation (p<0.05). In univariate analysis CEA and osteopontin turned out to be significant predictors for pCR (p-values resp. <0.001 and 0.013), while there was a trend for IL-8 (p=0.087). Taking response as outcome CEA, IL-8 and
S182
osteopontin were significant predictors (p-values resp. <0.001, <0.001 and 0.006), and CA19-9 showed a trend (p=0.065). In multivariate analysis osteopontin was the strongest predictor for pCR (p=0.028) and CEA and IL-8 predicted for response (p=0.029 resp. 0.021).
ESTRO 31
internal validation performances were consistent for all outcomes: 0.54 for LR (0.52-0.57), 0.60 for DM (0.58-0.62), 0.70 for OS (0.690.72). Tumor diameter, pT-stage and pN-stage were predictive for all outcomes. Microscopic presence of disease was predictive for both DM and OS. Figure 1 shows the nomogram for OS. Conclusions: The provided models are internally validated and are able to accurately predict long-term outcome for pancreatic cancer patients with various clinical characteristics and given treatments. The nomograms allow decision support in daily clinical practice, i.e. selection of patients who may benefit most from specific treatment options, and stimulate also generation of new clinical hypotheses.
Conclusions: CEA, IL-8 and osteopontin were identified as strong predictive biomarkers for tumor response and PCR after CRT in rectal cancer. Thus, these blood biomarkers could lead to an additional accuracy of earlier developed prediction models using clinical variables and PET-information. OC-0455 MULTIVARIATE PREDICTION OF FOLLOW-UP OUTCOME FOR PATIENTS WITH PANCREATIC CANCER: A POOLED DATA ANALYSIS G.C. Mattiucci1, R.G.P.M. van Stiphout2, S. Alfieri3, F.A. Calvo4, J.B. Dubois5, B.W. Maidment 1116, R.C. Miller7, M. Reni8, N. Sharma9, M. Falconi10 1 Università Cattolica del Sacro Cuore-Policlinico A. Gemelli, Radiotherapy, Rome, Italy 2 MAASTRO GROW University Medical Centre, Radiation Oncology, Maastricht, The Netherlands 3 Università Cattolica S. Cuore, Surgery, Rome, Italy 4 Hospital General Universitario Gregorio Marañón, Oncology, Madrid, Spain 5 CRLC, Radiotherapy, Montpellier, France; 6 University of Maryland School of Medicine, Radiotherapy, Baltimore, USA 7 Mayo Clinic, Radiation Oncology, Rochester, USA 8 S. Raffaele Scientific Institute, Oncology, Milan, Italy 9 University of Maryland School of Medicine, Radiation Oncology, Baltimore, USA 10 University of Verona, Surgery, Verona, Italy Purpose/Objective: To develop accurate nomograms to predict local recurrence (LR), distant metastases (DM) and overall survival (OS) within 5 years of follow-up for pancreatic cancer patients. This multivariate approach should allow treatment decision support for individual patients and generation of new evidences and hypotheses. Materials and Methods: Clinical data (N =1120) from 9 different institutes (Baltimore, Rochester, Montpellier, Madrid, Salzburg, Verona, Campobasso, Milano, and Rome) were pooled for this analysis. Patients with different combinations of treatments (preoperative RT, IORT, postoperative RT, concomitant chemo, neoadjuvant chemo, adjuvant chemo, surgery) were retrospectively included. The common clinical variables which were clinically relevant and eligible according to data constraints (not more than 15% missing data and heterogeneity in the data) were: age, gender, tumor location (head, body, tail), surgical procedure (DCP, distal, total), tumor grade (I-IV), microscopic residual disease (no/yes), lymphadenectomy (no/yes), pilorus preserving technique (no/yes), tumor diameter (mm), pathological tumor stage and pathological nodal stage. Variables were first selected by univariate analysis with the log-rank test for the corresponding Kaplan-Meier curves. These variables were used as an input for multivariate Cox’s proportional hazards regression model. Performance of the models was expressed by the c-index, which has similar properties as the area under the ROC curve (0.5, random prediction and 1, perfect prediction). Validation performance was tested with a 5-fold cross-validation scheme to avoid overfitting. Results: In the pooled database occurrences of events within 5 years of follow-up were 46.4% LR (outcome available: N = 690), 65.0% DM (N = 676) and 30.8% OS (N = 1050). The accuracy of the developed models was good with training accuracies of 0.62 for LR (95% CI: 0.590.65), 0.64 for DM (0.62-0.67) and 0.73 for OS (0.71-0.76). The
Figure 1. Nomogram of overall survival for pancreatic cancer OC-0456 NEOADJUVANT CHEMORADIOTHERAPY VS SURGERY FOR PANCREATIC CANCER. A MULTI-CENTRE RANDOMISED PHASE II TRIAL T. Brunner1, H. Golcher2, H. Witzigmann3, L. Marti4, W. Bechstein5, C. Bruns6, J. Hauss7, S. Merkel2, R. Fietkau8, W. Hohenberger2 1 University of Oxford, Oncology, Oxford, United Kingdom 2 University Hospitals Erlangen, Surgery, Erlangen, Germany 3 Krankenhaus Dresden Friedrichstadt, Surgery, Dresden, Germany 4 Kantonsspital St. Gallen, Surgery, St. Gallen, Switzerland 5 University Hospitals Frankfurt, Surgery, Frankfurt am Main, Germany 6 University Hospitals Munich Großhadern, Surgery, Munich, Germany 7 University Hospitals Leipzig, Surgery, Leipzig, Germany 8 University Hospitals Erlangen, Radiation Oncology, Erlangen, Germany Purpose/Objective: Primary surgery is standard treatment for resectable pancreatic carcinoma, and median overal survival time [mOS] for patients [pts] with resected pancreatic cancer is about 20 months. But still most pts die from local relapse or metastases. Therefore, we investigated in the first randomised controlled multicentre trial on this topic whether neoadjuvant chemoradiotherapy [CRT] in resectable pancreatic carcinoma is superior compared to primary resection. Materials and Methods: Pts with histologically proven ductal adenocarcinoma of the pancreatic head and <180° contact to peripancreatic vessels were randomised into Group A - primary surgery or Group B – neoadjuvant CRT (55.4 Gy; gemcitabine & cisplatin), followed by surgery 6 weeks later. In 2005 a protocol amendment suggested adjuvant chemotherapy (gemcitabine). The primary endpoint was mOS. The proposed sample size of 254 eligible patients was planned to provide at least 80% power to detect a 4.33 month difference in mOS with a significance level of 5% in a 2-sided log-rank test. Results: Between 06/2003 and 12/2009, 73 pts were randomised in 8 centres (A n=37, B n=36). Due to slow recruitment the trial was closed early. At submission data of 68 pts (A/B n=34) could be evaluated, 3 pts withdrew consent (A/B n=2/1), 1 centre sent no data (A/B n=1). The allocated therapy was received by 33 pts in Group A and 30 pts in Group B. One pt (A) developed sepsis and died before surgery. Four pts refused the allocated CRT and underwent primary surgery.