- Email: [email protected]
54. Proactive treatment of pelvic T4 locally advanced and recurrent colorectal cancer
ABSTRACTS Background: The indication for resection of the primary tumour in patients with asymptomatic metastatic colorectal cancer is under debate. So far, there are no results from RCTs; the CAIRO4 study is still enrolling patients and these results should be awaited. Several retrospective studies suggested a survival benefit in patients who underwent resection of the primary tumour, but bias due to confounding by indication is highly likely. Comparative Effectiveness Research, by using country as instrumental variable, could provide clues to the best treatment strategy for asymptomatic incurable metastatic colorectal cancer. Material and methods: Population-based cohorts (2007e2013) from the Netherlands and Norway including all patients with synchronous metastatic colorectal cancer (who had no surgery of metastatic disease) were compared on treatment strategy and overall survival. Using country as an instrumental variable (pseudo-randomisation), we assessed the effect of different treatment strategies on mortality within the first year. Analyses were adjusted for age, gender, localisation, year of diagnosis, and localisation of metastases. Results: Overall, 21,196 patients were included; 16,144 Dutch patients and 5,052 Norwegian patients. The proportion of patients who had surgery of the primary tumour was 38.6% in the Netherlands compared with 51.5% in Norway (p < 0.001). Of all Dutch patients, 58.4% received chemotherapy compared with 21.4% of Norwegian patients. Radiotherapy was given in 10.2% of Dutch patients compared with 11.2% of Norwegian patients. With the Netherlands as a reference category, the adjusted HR for overall survival was 0.98 (95% CI 0.93e1.03; p ¼ 0.36). Instrumental variable analysis showed an adjusted OR of 1.00 (95% CI 0.98e1.03; p ¼ 0.72). Conclusions: The present international comparison shows variations in treatment strategy between the Netherlands and Norway. However, we did not observe a difference in overall survival between these countries. Instrumental variable analysis showed no benefit of a treatment strategy with more surgery of the primary tumour on mortality within the first year. Conflict of interest: No conflict of interest. http://dx.doi.org/10.1016/j.ejso.2016.06.058
53. Optimal timing of surgery after neoadjuvant chemo-radiation therapy in locally advanced rectal cancer Z. Gad1, W. Gareer1, H. El Hossieny2, A. Naguib3, S. Abo Amer1 1 NCI e Cairo University, Surgical Oncology, Cairo, Egypt 2 NCI e Cairo University, Radiotherapy, Cairo, Egypt 3 NCI e Cairo University, Medical Oncology, Cairo, Egypt Background: Surgery is the corner stone for the management of rectal cancer. An unsolved aspect of neoadjuvant chemo-radiation is the appropriate timing of surgery after completion of neoadjuvant chemo-radiation. The purpose of this study was to demonstrate the optimal time of surgical resection after the completion of neoadjuvant chemo-radiotherapy in treatment of locally advanced rectal cancer. Material and methods: This study compared 2 groups of patients with locally advanced rectal cancer, treated with neoadjuvant chemo-radiotherapy followed by surgical resection either 6e8 weeks (group 1) or 9e14 (group 2) weeks after the completion of chemo-radiotherapy. The impact of delaying surgery was tested in comparison to early surgical resection after completion of chemo-radiotherapy. Results: There was no statistically significant difference between the 2 groups regarding treatment toxicity. The difference between the two groups regarding the type of surgery was statistically not significant (P ¼ 0.382). The total significant response rate that could result in functional preservation was estimated to be 3.85% in group I and 15.38% in group II. 9.62% of our patients had residual malignant cells at one cm surgical margin. All those patients with positive margins at one cm were in group I (19.23%). No patients in group II had positive margins at 1 cm (0%). There was less operative time in group II, but the difference between both groups was statistically insignificant (P ¼ 0.845). The difference between both groups regarding operative blood loss and intra operative blood transfusion was significantly less in group II (P ¼ 0.044). There was no
S87 statistically significant difference between both groups regarding the intra operative complications (P ¼ 0.609). The current study showed significantly less post-operative hospital stay period, and less post-operative wound infection in group II (P ¼ 0.012 and 0.017). The current study showed more tumor regression and necrosis in group II with a highly significant main effect of time F ¼ 61.7 (P < 0.001). Pathological TN stage indicated better pathological tumor response in group II (P ¼ 0.04). Pathological complete response rate was reported in only one patient in group I (3.85%), and in 7/26 patients (26.92%) in group II. The current study showed recurrence free survival for all cases at 18 months of 84.2%. In group I, survival rate at the same duration was 73.8%, however none of group II cases had local recurrence (censored). The difference was of statistical significance (P ¼ 0.031). Disease free survival (DFS) during the same duration (18 months) was 69.4% for patients in group I and 82.3% for group II. The difference was of no statistical significance (P ¼ 0.429). Conclusion: Surgical resection delay up to 9e14 weeks after chemoradiation was associated with better outcome and better recurrence free survival. Conflict of interest: No conflict of interest. http://dx.doi.org/10.1016/j.ejso.2016.06.059
54. Proactive treatment of pelvic T4 locally advanced and recurrent colorectal cancer C. Sassaroli1, A. Cassata2, O. Catalano3, E. Cardone1, F. Ruffolo1, P. Delrio1 1 National Cancer Institute of Naples, Colorectal Surgery, Naples, Italy 2 National Cancer Institute of Naples, Abdominal Oncology, Naples, Italy 3 National Cancer Institute of Naples, Radiology, Naples, Italy Introduction: Colorectal cancer with localized pelvic disease is amenable to a curative approach. Even more advanced cases, including T4 tumours and recurrent pelvic cancer can be managed with extensive surgery to achieve local control. Peritoneal involvement can also be dealed with and early referral or simultaneous HIPEC (Hyperthermic Perioperative Chemotherapy), together with appropriate surgery; it may be able to provide a complete clearance of the primary cancer and prevent intraabdominal diffusion of the disease. Patients and methods: From 2010 to 2015, among 50 consecutive pelvectomies performed both for pelvic recurrence or locally advanced rectosigmoid tumors, we performed simultaneous HIPEC (proactive treatment) in 13 patients. Total pelvectomy was performed in 3 patients: 2 for T4 tumors (1 mucinous) and one for pelvic recurrence (mucinous); posterior pelvectomy was performed in 10 patients: 3 for T4 tumors (all mucinous) and 7 for pelvic recurrence (4 mucinous). All patients received extended pelvic peritonectomy, intraperitoneal oxaliplatin with systemic 5 e Fluorouracil for a 30 min HIPEC treatment. Peritoneal dissection was started at the transverse umbilical line. Results: Among patients treated with HIPEC there were 4 Clavien Dindo grade IIIeIV complications. No postoperative death occurred. The median postoperative hospital stay was 17 days. Up to date 11 patients are disease free: respectively after 64, 52, 47, 40, 38, 26, 25, 17, 4 and 3 months. Of these, 1 patient developed liver disease and is now disease free after chemotherapy; 1 patient, pre-treated with chemotherapy for a lung metastasis before proactive HIPEC, underwent lung resection after abdominal surgery; 2 patients underwent an APR for recurrent cancer on rectal stump and colorectal anastomosis respectively. None of these patients developed a peritoneal recurrence. One patient died after 20 months with a huge pelvic recurrence of a G3 mucinous tumour and 1 patient is alive at 30 months with metastatic disease (liver and lung). Conclusions: In high risk patients (peritoneal involvement, ovarian metastases, perforated tumours, previous R1-2 resections or intraoperative tumour disruption, positive cytology, adjacent organs involvement, T3 mucinous tumor, T4 cancers) “proactive” HIPEC is a very promising treatment option to prevent intra-abdominal diffusion of pelvic disease. Our
S88
ABSTRACTS
results, in line with other series, show the feasibility and the oncological safety of this approach with acceptable morbidity and no mortality. Conflict of interest: No conflict of interest. http://dx.doi.org/10.1016/j.ejso.2016.06.060
55. The diagnostic performance of CT imaging in detecting colorectal peritoneal metastases H. Van Eden1, M. Lahaye2, D. Lambregts2, N. Kok1, G. Beets1, R. Beets-Tan2, A. Aalbers1 1 Netherlands Cancer Institute e Antoni van Leeuwenhoek Hospital, Surgical Oncology, Amsterdam, Netherlands 2 Netherlands Cancer Institute e Antoni van Leeuwenhoek Hospital, Radiology, Amsterdam, Netherlands Background: Staging of tumor burden is essential in patients with colorectal peritoneal carcinomatosis (PC) who are eligible for cytoreductive surgery (CRS) and hyperthermic intraperitoneal chemotherapy (HIPEC). This study aims to compare the extent of PC on preoperative computed tomography (CT) with the intraoperative findings. Materials and methods: Preoperative abdominal CT scans of patients who underwent CRS-HIPEC, performed between 2005 and 2007, were evaluated by an experienced radiologist. The extent of PC was scored according to the Dutch region count, in which the peritoneal cavity is divided into seven regions. Gold standard was the intraoperative region count. When five or less abdominal regions were affected CRS-HIPEC was considered useful and thus executed. Diagnostic performance of CT was calculated and survival analyses were performed. Results/Discussion: Forty-nine patients were included. The accuracy for selection of colorectal PC patients with good and poor prognosis, based on a region count of five or less, was 86%. In 8% of the other patients the extent of PC was underestimated and in 6% the extent of PC was overestimated. Patients with more than five affected abdominal regions had poorer disease free survival and overall survival (median 11.3 months (interquartile range (IQR) 9.1e28.2) and 19.9 months (IQR 18.5e42.4)) compared to patients with a region count of five or less (median 21.8 months (IQR 15.6e39.7) and 39.5 months (25.0e84.7)). Conclusion: Abdominal CT has a high accuracy in selection of patients who could benefit from CRS-HIPEC provided examination by an experienced radiologist. Conflict of interest: No conflict of interest.
56. The effect of thrombocyte aggregation inhibitors on cancer survival: More evidence for a platelet-mediated effect of aspirin on cancer M. Frouws1, E. Rademaker1, E. Bastiaannet1, M. Van Herk-Sukel2, V. Lemmens3,4, C. Van de Velde1, J. Portielje5, G.J. Liefers1 1 Leiden University Medical Centre, Surgical Oncology, Leiden, Netherlands 2 PHARMO Institute, for Drug Outcomes Research, Utrecht, Netherlands 3 Comprehensive Cancer Organisation, the Netherlands, Eindhoven, Netherlands 4 Erasmus Medical Center, Public Health, Rotterdam, Netherlands 5 Haga hospital, Medical Oncology, The Hague, Netherlands Background: Several studies have suggested an association between aspirin (acetylsalicylate) use and improved cancer survival. The mechanism behind the beneficial effect is not completely clarified. One of the hypothesis is that circulating tumour cells (CTC’s) are guarded from the immune system by thrombocytes. When the aggregation of thrombocytes is inhibited, the protection of the CTC’s vanishes and the immune system will be able to clear the CTC’s. Aspirin is a thrombocyte aggregation inhibitor (TAI) and could in this way reduce the mortality in patients with colorectal cancer. The aim of this study was to provide epidemiological evidence for the hypothesised platelet-mediated mechanism by studying the effect of non-aspirin TAI on survival of patients with colorectal cancer. Material and methods: Patients with colorectal cancer, diagnosed between 1998 and 2011 in the southern region of the Netherlands (Eindhoven Cancer Registry) were linked to drug dispensing data from the PHARMO Database Network. Patients using TAI (mainly clopidogrel, dipyridamole) were compared to non-users. A sub analysis was performed in patients using solely TAI (not in combination with aspirin) versus non-users. Patients using only aspirin, using TAI before diagnosis and patients whose follow-up was less than 14 days were excluded from the analysis. The association between the use of TAI and overall survival was analysed using Cox regression models with use of TAI as time-varying exposure. Results: In total, 6184 patients were diagnosed with colorectal cancer. In this cohort, 5439 patients were non-users (88%) and 747 (12%) were users of TAI. The results of the survival analysis are shown in the table. The effect of TAI shows to be independent from the use of aspirin. Conclusions: The observation that TAI use is associated with improved overall survival in colorectal cancer patients makes the hypothesis that the mode of action of aspirin is platelet-mediated more plausible. Conflict of interest: No conflict of interest.
Abstract 55 Time dependent survival analysis No. at risk TAI-users versus non-user (n ¼ 6186) Non-users 5439 TAI-users 747 Solely TAI-user versus non-user (n ¼ 5534) Non-users 5439 Solely TAI-users 95 TAI-users versus non-users per tumour type Colon cancer (n ¼ 3904) Non-users 3398 TAI-users 506 Rectal cancer (n ¼ 2282) Non-users 2041 TAI-users 241
No. events
Crude Hazard Ratio
Adjusted Hazard Ratio*
2501 277
1 (Ref) 0.37 (95% CI 0.30e0.46)
1 (Ref) 0.49 (95% CI 0.38e0.62)
2501 54
1 (Ref) 0.58 (95% CI 0.35e0.95)
1 (Ref) 0.52 (95% CI 0.32e0.85)
1609 187
1 (Ref) 0.34 (95% C.I. 0.26e0.44)
1 (Ref) 0.46 (95% C.I. 0.34e0.62)
892 90
1 (Ref) 0.43 (95% CI 0.30e0.62)
1 (Ref) 0.55 (95% CI 036e0.85)
*Adjusted for use of aspirin, sex, age at incidence date, stage, surgery (yes/no), chemotherapy (yes/no), radiotherapy (yes/no), amount of comorbidities http://dx.doi.org/10.1016/j.ejso.2016.06.061
http://dx.doi.org/10.1016/j.ejso.2016.06.062
53. Optimal timing of surgery after neoadjuvant chemo-radiation therapy in locally advanced rectal cancer Z. Gad1, W. Gareer1, H. El Hossieny2, A. Naguib3, S. Abo Amer1 1 NCI e Cairo University, Surgical Oncology, Cairo, Egypt 2 NCI e Cairo University, Radiotherapy, Cairo, Egypt 3 NCI e Cairo University, Medical Oncology, Cairo, Egypt Background: Surgery is the corner stone for the management of rectal cancer. An unsolved aspect of neoadjuvant chemo-radiation is the appropriate timing of surgery after completion of neoadjuvant chemo-radiation. The purpose of this study was to demonstrate the optimal time of surgical resection after the completion of neoadjuvant chemo-radiotherapy in treatment of locally advanced rectal cancer. Material and methods: This study compared 2 groups of patients with locally advanced rectal cancer, treated with neoadjuvant chemo-radiotherapy followed by surgical resection either 6e8 weeks (group 1) or 9e14 (group 2) weeks after the completion of chemo-radiotherapy. The impact of delaying surgery was tested in comparison to early surgical resection after completion of chemo-radiotherapy. Results: There was no statistically significant difference between the 2 groups regarding treatment toxicity. The difference between the two groups regarding the type of surgery was statistically not significant (P ¼ 0.382). The total significant response rate that could result in functional preservation was estimated to be 3.85% in group I and 15.38% in group II. 9.62% of our patients had residual malignant cells at one cm surgical margin. All those patients with positive margins at one cm were in group I (19.23%). No patients in group II had positive margins at 1 cm (0%). There was less operative time in group II, but the difference between both groups was statistically insignificant (P ¼ 0.845). The difference between both groups regarding operative blood loss and intra operative blood transfusion was significantly less in group II (P ¼ 0.044). There was no
S87 statistically significant difference between both groups regarding the intra operative complications (P ¼ 0.609). The current study showed significantly less post-operative hospital stay period, and less post-operative wound infection in group II (P ¼ 0.012 and 0.017). The current study showed more tumor regression and necrosis in group II with a highly significant main effect of time F ¼ 61.7 (P < 0.001). Pathological TN stage indicated better pathological tumor response in group II (P ¼ 0.04). Pathological complete response rate was reported in only one patient in group I (3.85%), and in 7/26 patients (26.92%) in group II. The current study showed recurrence free survival for all cases at 18 months of 84.2%. In group I, survival rate at the same duration was 73.8%, however none of group II cases had local recurrence (censored). The difference was of statistical significance (P ¼ 0.031). Disease free survival (DFS) during the same duration (18 months) was 69.4% for patients in group I and 82.3% for group II. The difference was of no statistical significance (P ¼ 0.429). Conclusion: Surgical resection delay up to 9e14 weeks after chemoradiation was associated with better outcome and better recurrence free survival. Conflict of interest: No conflict of interest. http://dx.doi.org/10.1016/j.ejso.2016.06.059
54. Proactive treatment of pelvic T4 locally advanced and recurrent colorectal cancer C. Sassaroli1, A. Cassata2, O. Catalano3, E. Cardone1, F. Ruffolo1, P. Delrio1 1 National Cancer Institute of Naples, Colorectal Surgery, Naples, Italy 2 National Cancer Institute of Naples, Abdominal Oncology, Naples, Italy 3 National Cancer Institute of Naples, Radiology, Naples, Italy Introduction: Colorectal cancer with localized pelvic disease is amenable to a curative approach. Even more advanced cases, including T4 tumours and recurrent pelvic cancer can be managed with extensive surgery to achieve local control. Peritoneal involvement can also be dealed with and early referral or simultaneous HIPEC (Hyperthermic Perioperative Chemotherapy), together with appropriate surgery; it may be able to provide a complete clearance of the primary cancer and prevent intraabdominal diffusion of the disease. Patients and methods: From 2010 to 2015, among 50 consecutive pelvectomies performed both for pelvic recurrence or locally advanced rectosigmoid tumors, we performed simultaneous HIPEC (proactive treatment) in 13 patients. Total pelvectomy was performed in 3 patients: 2 for T4 tumors (1 mucinous) and one for pelvic recurrence (mucinous); posterior pelvectomy was performed in 10 patients: 3 for T4 tumors (all mucinous) and 7 for pelvic recurrence (4 mucinous). All patients received extended pelvic peritonectomy, intraperitoneal oxaliplatin with systemic 5 e Fluorouracil for a 30 min HIPEC treatment. Peritoneal dissection was started at the transverse umbilical line. Results: Among patients treated with HIPEC there were 4 Clavien Dindo grade IIIeIV complications. No postoperative death occurred. The median postoperative hospital stay was 17 days. Up to date 11 patients are disease free: respectively after 64, 52, 47, 40, 38, 26, 25, 17, 4 and 3 months. Of these, 1 patient developed liver disease and is now disease free after chemotherapy; 1 patient, pre-treated with chemotherapy for a lung metastasis before proactive HIPEC, underwent lung resection after abdominal surgery; 2 patients underwent an APR for recurrent cancer on rectal stump and colorectal anastomosis respectively. None of these patients developed a peritoneal recurrence. One patient died after 20 months with a huge pelvic recurrence of a G3 mucinous tumour and 1 patient is alive at 30 months with metastatic disease (liver and lung). Conclusions: In high risk patients (peritoneal involvement, ovarian metastases, perforated tumours, previous R1-2 resections or intraoperative tumour disruption, positive cytology, adjacent organs involvement, T3 mucinous tumor, T4 cancers) “proactive” HIPEC is a very promising treatment option to prevent intra-abdominal diffusion of pelvic disease. Our
S88
ABSTRACTS
results, in line with other series, show the feasibility and the oncological safety of this approach with acceptable morbidity and no mortality. Conflict of interest: No conflict of interest. http://dx.doi.org/10.1016/j.ejso.2016.06.060
55. The diagnostic performance of CT imaging in detecting colorectal peritoneal metastases H. Van Eden1, M. Lahaye2, D. Lambregts2, N. Kok1, G. Beets1, R. Beets-Tan2, A. Aalbers1 1 Netherlands Cancer Institute e Antoni van Leeuwenhoek Hospital, Surgical Oncology, Amsterdam, Netherlands 2 Netherlands Cancer Institute e Antoni van Leeuwenhoek Hospital, Radiology, Amsterdam, Netherlands Background: Staging of tumor burden is essential in patients with colorectal peritoneal carcinomatosis (PC) who are eligible for cytoreductive surgery (CRS) and hyperthermic intraperitoneal chemotherapy (HIPEC). This study aims to compare the extent of PC on preoperative computed tomography (CT) with the intraoperative findings. Materials and methods: Preoperative abdominal CT scans of patients who underwent CRS-HIPEC, performed between 2005 and 2007, were evaluated by an experienced radiologist. The extent of PC was scored according to the Dutch region count, in which the peritoneal cavity is divided into seven regions. Gold standard was the intraoperative region count. When five or less abdominal regions were affected CRS-HIPEC was considered useful and thus executed. Diagnostic performance of CT was calculated and survival analyses were performed. Results/Discussion: Forty-nine patients were included. The accuracy for selection of colorectal PC patients with good and poor prognosis, based on a region count of five or less, was 86%. In 8% of the other patients the extent of PC was underestimated and in 6% the extent of PC was overestimated. Patients with more than five affected abdominal regions had poorer disease free survival and overall survival (median 11.3 months (interquartile range (IQR) 9.1e28.2) and 19.9 months (IQR 18.5e42.4)) compared to patients with a region count of five or less (median 21.8 months (IQR 15.6e39.7) and 39.5 months (25.0e84.7)). Conclusion: Abdominal CT has a high accuracy in selection of patients who could benefit from CRS-HIPEC provided examination by an experienced radiologist. Conflict of interest: No conflict of interest.
56. The effect of thrombocyte aggregation inhibitors on cancer survival: More evidence for a platelet-mediated effect of aspirin on cancer M. Frouws1, E. Rademaker1, E. Bastiaannet1, M. Van Herk-Sukel2, V. Lemmens3,4, C. Van de Velde1, J. Portielje5, G.J. Liefers1 1 Leiden University Medical Centre, Surgical Oncology, Leiden, Netherlands 2 PHARMO Institute, for Drug Outcomes Research, Utrecht, Netherlands 3 Comprehensive Cancer Organisation, the Netherlands, Eindhoven, Netherlands 4 Erasmus Medical Center, Public Health, Rotterdam, Netherlands 5 Haga hospital, Medical Oncology, The Hague, Netherlands Background: Several studies have suggested an association between aspirin (acetylsalicylate) use and improved cancer survival. The mechanism behind the beneficial effect is not completely clarified. One of the hypothesis is that circulating tumour cells (CTC’s) are guarded from the immune system by thrombocytes. When the aggregation of thrombocytes is inhibited, the protection of the CTC’s vanishes and the immune system will be able to clear the CTC’s. Aspirin is a thrombocyte aggregation inhibitor (TAI) and could in this way reduce the mortality in patients with colorectal cancer. The aim of this study was to provide epidemiological evidence for the hypothesised platelet-mediated mechanism by studying the effect of non-aspirin TAI on survival of patients with colorectal cancer. Material and methods: Patients with colorectal cancer, diagnosed between 1998 and 2011 in the southern region of the Netherlands (Eindhoven Cancer Registry) were linked to drug dispensing data from the PHARMO Database Network. Patients using TAI (mainly clopidogrel, dipyridamole) were compared to non-users. A sub analysis was performed in patients using solely TAI (not in combination with aspirin) versus non-users. Patients using only aspirin, using TAI before diagnosis and patients whose follow-up was less than 14 days were excluded from the analysis. The association between the use of TAI and overall survival was analysed using Cox regression models with use of TAI as time-varying exposure. Results: In total, 6184 patients were diagnosed with colorectal cancer. In this cohort, 5439 patients were non-users (88%) and 747 (12%) were users of TAI. The results of the survival analysis are shown in the table. The effect of TAI shows to be independent from the use of aspirin. Conclusions: The observation that TAI use is associated with improved overall survival in colorectal cancer patients makes the hypothesis that the mode of action of aspirin is platelet-mediated more plausible. Conflict of interest: No conflict of interest.
Abstract 55 Time dependent survival analysis No. at risk TAI-users versus non-user (n ¼ 6186) Non-users 5439 TAI-users 747 Solely TAI-user versus non-user (n ¼ 5534) Non-users 5439 Solely TAI-users 95 TAI-users versus non-users per tumour type Colon cancer (n ¼ 3904) Non-users 3398 TAI-users 506 Rectal cancer (n ¼ 2282) Non-users 2041 TAI-users 241
No. events
Crude Hazard Ratio
Adjusted Hazard Ratio*
2501 277
1 (Ref) 0.37 (95% CI 0.30e0.46)
1 (Ref) 0.49 (95% CI 0.38e0.62)
2501 54
1 (Ref) 0.58 (95% CI 0.35e0.95)
1 (Ref) 0.52 (95% CI 0.32e0.85)
1609 187
1 (Ref) 0.34 (95% C.I. 0.26e0.44)
1 (Ref) 0.46 (95% C.I. 0.34e0.62)
892 90
1 (Ref) 0.43 (95% CI 0.30e0.62)
1 (Ref) 0.55 (95% CI 036e0.85)
*Adjusted for use of aspirin, sex, age at incidence date, stage, surgery (yes/no), chemotherapy (yes/no), radiotherapy (yes/no), amount of comorbidities http://dx.doi.org/10.1016/j.ejso.2016.06.061
http://dx.doi.org/10.1016/j.ejso.2016.06.062