Radiotherapy and Oncology xxx (2014) xxx–xxx
Contents lists available at ScienceDirect
Radiotherapy and Oncology journal homepage: www.thegreenjournal.com
Original article
Can perfusion MRI predict response to preoperative treatment in rectal cancer? Milou H. Martens a,b,c,⇑, Samina Subhani a, Luc A. Heijnen a,b,c, Doenja M.J. Lambregts a, Jeroen Buijsen c,d, Monique Maas a, Robert G. Riedl e, Cecile R.L.P.N. Jeukens a, Geerard L. Beets b, Ewelina Kluza a, Regina G.H. Beets-Tan a,c a Department of Radiology, Maastricht University Medical Center; b Department of Surgery, Maastricht University Medical Center; c GROW – School for Oncology and Developmental Biology, Maastricht University Medical Center; d Department of Radiation Oncology, Maastro Clinic; and e Department of Pathology, Maastricht University Medical Center, The Netherlands
a r t i c l e
i n f o
Article history: Received 22 September 2014 Received in revised form 24 November 2014 Accepted 24 November 2014 Available online xxxx Keywords: Rectal cancer Neoadjuvant chemoradiation Response prediction Preoperative treatment
a b s t r a c t Background and purpose: Dynamic contrast-enhanced MRI (DCE-MRI) provides information on perfusion and could identify good prognostic tumors. Aim of this study was to evaluate whether DCE-MRI using a novel blood pool contrast-agent can accurately predict the response to neoadjuvant chemoradiotherapy in locally advanced rectal cancer. Materials and methods: Thirty patients underwent DCE-MRI before and 7–10 weeks after chemoradiotherapy. Regions of interest were drawn on DCE-MRI with T2W-images as reference. DCE-MRI-based kinetic parameters (initial slope, initial peak, late slope, and AUC at 60, 90, and 120 s) determined preand post-CRT and their D were compared between good (TRG1–2) and poor (TRG3–5) responders. Optimal thresholds were determined and sensitivities, specificities, positive predictive values (PPV), and negative predictive values (NPV) were calculated. Results: Pre-therapy, the late slope was able to discriminate between good and poor responders (0.05 103 vs. 0.62 103, p < 0.001) with an AUC of 0.90, sensitivity 92%, specificity 82%, PPV 80%, and NPV 93%. Other pre-CRT parameters showed no significant differences, nor any post-CRT parameters or their D. Conclusions: The kinetic parameter ‘late slope’ derived from DCE-MRI could potentially be helpful to predict before the onset of neoadjuvant chemoradiotherapy which tumors are likely going to respond. This could allow for personalized treatment-options in rectal cancer patients. Ó 2014 Elsevier Ireland Ltd. All rights reserved. Radiotherapy and Oncology xxx (2014) xxx–xxx
After neoadjuvant chemoradiotherapy (CRT) of locally advanced rectal cancer (LARC) [1,2], a complete response is obtained in 8–24% [3–6]. There is now a growing awareness of an organpreserving approach (local excision of the remaining scar tissue or a non-operative ‘‘wait-and-see’’ approach) as an alternative curative treatment for these patients [7–9], given the substantial risk for morbidity and mortality associated with major resection [10,11]. While currently the decision for organ-preservation is made after completion of CRT, knowing upfront which patients have a high chance for a good response would be highly advantageous. Moreover, it is anticipated that the spectrum of rectal tumors manageable with organ-preservation may enlarge and ⇑ Corresponding author at: Maastricht University Medical Center, Department of Radiology, PO Box 5800, 6202 AZ Maastricht, The Netherlands. E-mail addresses: [email protected], [email protected] (M.H. Martens).
selection criteria may become more liberal. However, it is not justified to irradiate all patients knowing that not all tumors will respond and still require TME, resulting in worse functional outcome compared with TME without preoperative chemoradiation. Apart from this, there may be a subgroup of patients with LARC who may have a reasonably good chance for a complete response but require intensification of preoperative treatment to increase the chance of obtaining a complete response. Therefore noninvasive tools that can predict response before the onset of CRT may allow for a more individualized treatment. Currently no reliable conventional imaging modalities exist for response prediction, because they can only provide morphological information. Dynamic Contrast-Enhanced (DCE) MRI is a functional imaging technique, mapping angiogenic microvasculature by measuring the inflow and leakage of contrast into the extracellular space. This can be expressed in signal enhancement curves from which we can quantify perfusion parameters. These
http://dx.doi.org/10.1016/j.radonc.2014.11.044 0167-8140/Ó 2014 Elsevier Ireland Ltd. All rights reserved.
Please cite this article in press as: Martens MH et al. Can perfusion MRI predict response to preoperative treatment in rectal cancer? Radiother Oncol (2014), http://dx.doi.org/10.1016/j.radonc.2014.11.044
2
Perfusion MRI for response prediction
parameters reflect the perfusion heterogeneity within the tumor. Therefore DCE-MRI could potentially differentiate between good and poor prognostic tumors. Few studies in literature investigated DCE-MRI for response prediction to CRT with conventional low molecular-weight contrast-agents and showed conflicting results [12–17]. We hypothesize that DCE-MRI based on a blood pool (e.g. high molecular-weight) contrast-agent is more accurate [18]. The aim of this study was to evaluate whether perfusion parameters derived from DCE-MRI using a blood pool contrast-agent measured both before and after neoadjuvant treatment can accurately differentiate between good and poor responders in locally advanced rectal cancer. Materials and methods Patients This study was part of a prospective study on rectal cancer MRI, which was approved by the local institutional review board. Between February 2011 and March 2014, 41 consecutive patients diagnosed with LARC (cT3–4 and/or cN+) were included. Inclusion criteria consisted of (1) biopsy proven rectal cancer (615 cm of the anorectal junction), (2) LARC without distant metastases, and (3) neoadjuvant CRT. All patients underwent a primary staging MRI and a second restaging MRI 7–10 weeks after completion of CRT, both including DCE-MRI. Neoadjuvant treatment routinely consisted of 50.4 Gy radiation (28 1.8 Gy) combined with capecitabine (825 mg/m2 twice daily). Eleven of the 41 eligible patients were excluded for the following reasons: delayed surgery of >3 months after restaging MRI (n = 7), artifacts on DCE-MRI hampering tumor delineation (n = 1), and mucinous tumors (n = 3) because they are known to exhibit different MR enhancing patterns compared to solid tumors. MR acquisition All MRI examinations were performed at 1.5T (Intera (Achieva) or Ingenia, Philips Medical Systems, Best, The Netherlands) using a phased array body coil. An intravenous bolus injection of 20 mg of butylscopolamine (BuscopanÒ, Boehringer Ingelheim bv, Ingelheim, Germany) was administered to reduce peristaltic movement, patients did not receive bowel preparation. The standard imaging protocol consisted of T2-weighted sequences in three orthogonal planes. DCE-MRI consisted of axial dynamic T1-weighted 3D fast field echo using the following parameters: 8 s temporal resolution, TR/TE 7.9/4.6 ms, 30° flip angle, 11 slices, 5 mm slice thickness, 5 mm interslice distance, 220 220 mm FOV, 256 228 matrix, 6 min total acquisition time. After 3 series of baseline measurements (24 s), 0.12 ml/kg bodyweight of the blood pool contrastagent gadofosveset (AblavarÒ, Lantheus Medical Imaging, North Billerica, Massachussets, USA) was administered at a rate of 0.70 ml/s into the brachial vein followed by a 20 ml saline flush using an MR compatible power injector (Spectris Solaris, MEDRAD, Warrendale, Pennsylvania, USA). The axial T2-weighted and DCEMRI sequences were angled in identical planes, perpendicular to the tumor axis.
The axial T2W images were used as a reference to detect the primary tumor and the suspected residual tumor, defined as residual wall thickening with either an iso- or hypointense signal intensity at the former tumor location (Fig. 1). DCE-MRI analysis was performed on a voxel-by-voxel basis using Matlab (Matlab 2007, MathWorks, Natick, Massachussetts, USA). Semiquantitative data analysis was chosen in view of the complex pharmacokinetic properties of the protein-binding contrast-agent, which affects the conversion of signal enhancement to gadofosveset concentration and applicability of the current pharmacokinetic methods. For each voxel, a relative enhancement-time curve was obtained according to the equation: Srel ðtÞ ¼ ðSðtÞ S0 Þ=S0 , where S(t) represents the signal intensity at a time (t) and S0 the baseline signal intensity, i.e. the average of the three pre-contrast scans. Six semiquantitative kinetic parameters were calculated: (1) Initial slope, based on the initial 3 upslope data points of the enhancement-time curve. (2) Initial peak, represents the magnitude of enhancement at the end of the initial slope. (3) Areas under the first 60, 90, and 120 s of the enhancement curve (AUC60, AUC90, and AUC120), the integrals of the respective data. (4) Late slope, based on the late phase of the enhancementtime curve, i.e. from the initial peak until the end of acquisition (Fig. 2). The aforementioned parameters are dimensionless since they were calculated from relative enhancement curves. Histopathology Histopathological assessment of the resection specimen was performed according to the method of Quirke et al. [19]. Tumor regression grade (TRG) was assessed by a dedicated pathologist according to Mandard et al. [20]; TRG1: Complete regression, no residual tumor cells, TRG2: Predominant fibrosis with isolated tumor cells, TRG3: More residual tumor cells but fibrosis still dominant, TRG4: Residual tumor outgrowing fibrosis, TRG5: no regression. Therapeutic response was stratified into two groups: Good responders were defined as TRG1–2 and poor responders were defined as TRG3–5 [21]. Statistical analyses All statistical analyses were performed with Statistical Package for the Social Sciences (SPSS 20.0, Inc, Chicago, Illinois, USA). Normally distributed kinetic parameters were compared between good and poor responders using an independent sample T-test. The nonparametric Mann–Whitney U test was performed for non-normally distributed data. p-values
Contents lists available at ScienceDirect
Radiotherapy and Oncology journal homepage: www.thegreenjournal.com
Original article
Can perfusion MRI predict response to preoperative treatment in rectal cancer? Milou H. Martens a,b,c,⇑, Samina Subhani a, Luc A. Heijnen a,b,c, Doenja M.J. Lambregts a, Jeroen Buijsen c,d, Monique Maas a, Robert G. Riedl e, Cecile R.L.P.N. Jeukens a, Geerard L. Beets b, Ewelina Kluza a, Regina G.H. Beets-Tan a,c a Department of Radiology, Maastricht University Medical Center; b Department of Surgery, Maastricht University Medical Center; c GROW – School for Oncology and Developmental Biology, Maastricht University Medical Center; d Department of Radiation Oncology, Maastro Clinic; and e Department of Pathology, Maastricht University Medical Center, The Netherlands
a r t i c l e
i n f o
Article history: Received 22 September 2014 Received in revised form 24 November 2014 Accepted 24 November 2014 Available online xxxx Keywords: Rectal cancer Neoadjuvant chemoradiation Response prediction Preoperative treatment
a b s t r a c t Background and purpose: Dynamic contrast-enhanced MRI (DCE-MRI) provides information on perfusion and could identify good prognostic tumors. Aim of this study was to evaluate whether DCE-MRI using a novel blood pool contrast-agent can accurately predict the response to neoadjuvant chemoradiotherapy in locally advanced rectal cancer. Materials and methods: Thirty patients underwent DCE-MRI before and 7–10 weeks after chemoradiotherapy. Regions of interest were drawn on DCE-MRI with T2W-images as reference. DCE-MRI-based kinetic parameters (initial slope, initial peak, late slope, and AUC at 60, 90, and 120 s) determined preand post-CRT and their D were compared between good (TRG1–2) and poor (TRG3–5) responders. Optimal thresholds were determined and sensitivities, specificities, positive predictive values (PPV), and negative predictive values (NPV) were calculated. Results: Pre-therapy, the late slope was able to discriminate between good and poor responders (0.05 103 vs. 0.62 103, p < 0.001) with an AUC of 0.90, sensitivity 92%, specificity 82%, PPV 80%, and NPV 93%. Other pre-CRT parameters showed no significant differences, nor any post-CRT parameters or their D. Conclusions: The kinetic parameter ‘late slope’ derived from DCE-MRI could potentially be helpful to predict before the onset of neoadjuvant chemoradiotherapy which tumors are likely going to respond. This could allow for personalized treatment-options in rectal cancer patients. Ó 2014 Elsevier Ireland Ltd. All rights reserved. Radiotherapy and Oncology xxx (2014) xxx–xxx
After neoadjuvant chemoradiotherapy (CRT) of locally advanced rectal cancer (LARC) [1,2], a complete response is obtained in 8–24% [3–6]. There is now a growing awareness of an organpreserving approach (local excision of the remaining scar tissue or a non-operative ‘‘wait-and-see’’ approach) as an alternative curative treatment for these patients [7–9], given the substantial risk for morbidity and mortality associated with major resection [10,11]. While currently the decision for organ-preservation is made after completion of CRT, knowing upfront which patients have a high chance for a good response would be highly advantageous. Moreover, it is anticipated that the spectrum of rectal tumors manageable with organ-preservation may enlarge and ⇑ Corresponding author at: Maastricht University Medical Center, Department of Radiology, PO Box 5800, 6202 AZ Maastricht, The Netherlands. E-mail addresses: [email protected], [email protected] (M.H. Martens).
selection criteria may become more liberal. However, it is not justified to irradiate all patients knowing that not all tumors will respond and still require TME, resulting in worse functional outcome compared with TME without preoperative chemoradiation. Apart from this, there may be a subgroup of patients with LARC who may have a reasonably good chance for a complete response but require intensification of preoperative treatment to increase the chance of obtaining a complete response. Therefore noninvasive tools that can predict response before the onset of CRT may allow for a more individualized treatment. Currently no reliable conventional imaging modalities exist for response prediction, because they can only provide morphological information. Dynamic Contrast-Enhanced (DCE) MRI is a functional imaging technique, mapping angiogenic microvasculature by measuring the inflow and leakage of contrast into the extracellular space. This can be expressed in signal enhancement curves from which we can quantify perfusion parameters. These
http://dx.doi.org/10.1016/j.radonc.2014.11.044 0167-8140/Ó 2014 Elsevier Ireland Ltd. All rights reserved.
Please cite this article in press as: Martens MH et al. Can perfusion MRI predict response to preoperative treatment in rectal cancer? Radiother Oncol (2014), http://dx.doi.org/10.1016/j.radonc.2014.11.044
2
Perfusion MRI for response prediction
parameters reflect the perfusion heterogeneity within the tumor. Therefore DCE-MRI could potentially differentiate between good and poor prognostic tumors. Few studies in literature investigated DCE-MRI for response prediction to CRT with conventional low molecular-weight contrast-agents and showed conflicting results [12–17]. We hypothesize that DCE-MRI based on a blood pool (e.g. high molecular-weight) contrast-agent is more accurate [18]. The aim of this study was to evaluate whether perfusion parameters derived from DCE-MRI using a blood pool contrast-agent measured both before and after neoadjuvant treatment can accurately differentiate between good and poor responders in locally advanced rectal cancer. Materials and methods Patients This study was part of a prospective study on rectal cancer MRI, which was approved by the local institutional review board. Between February 2011 and March 2014, 41 consecutive patients diagnosed with LARC (cT3–4 and/or cN+) were included. Inclusion criteria consisted of (1) biopsy proven rectal cancer (615 cm of the anorectal junction), (2) LARC without distant metastases, and (3) neoadjuvant CRT. All patients underwent a primary staging MRI and a second restaging MRI 7–10 weeks after completion of CRT, both including DCE-MRI. Neoadjuvant treatment routinely consisted of 50.4 Gy radiation (28 1.8 Gy) combined with capecitabine (825 mg/m2 twice daily). Eleven of the 41 eligible patients were excluded for the following reasons: delayed surgery of >3 months after restaging MRI (n = 7), artifacts on DCE-MRI hampering tumor delineation (n = 1), and mucinous tumors (n = 3) because they are known to exhibit different MR enhancing patterns compared to solid tumors. MR acquisition All MRI examinations were performed at 1.5T (Intera (Achieva) or Ingenia, Philips Medical Systems, Best, The Netherlands) using a phased array body coil. An intravenous bolus injection of 20 mg of butylscopolamine (BuscopanÒ, Boehringer Ingelheim bv, Ingelheim, Germany) was administered to reduce peristaltic movement, patients did not receive bowel preparation. The standard imaging protocol consisted of T2-weighted sequences in three orthogonal planes. DCE-MRI consisted of axial dynamic T1-weighted 3D fast field echo using the following parameters: 8 s temporal resolution, TR/TE 7.9/4.6 ms, 30° flip angle, 11 slices, 5 mm slice thickness, 5 mm interslice distance, 220 220 mm FOV, 256 228 matrix, 6 min total acquisition time. After 3 series of baseline measurements (24 s), 0.12 ml/kg bodyweight of the blood pool contrastagent gadofosveset (AblavarÒ, Lantheus Medical Imaging, North Billerica, Massachussets, USA) was administered at a rate of 0.70 ml/s into the brachial vein followed by a 20 ml saline flush using an MR compatible power injector (Spectris Solaris, MEDRAD, Warrendale, Pennsylvania, USA). The axial T2-weighted and DCEMRI sequences were angled in identical planes, perpendicular to the tumor axis.
The axial T2W images were used as a reference to detect the primary tumor and the suspected residual tumor, defined as residual wall thickening with either an iso- or hypointense signal intensity at the former tumor location (Fig. 1). DCE-MRI analysis was performed on a voxel-by-voxel basis using Matlab (Matlab 2007, MathWorks, Natick, Massachussetts, USA). Semiquantitative data analysis was chosen in view of the complex pharmacokinetic properties of the protein-binding contrast-agent, which affects the conversion of signal enhancement to gadofosveset concentration and applicability of the current pharmacokinetic methods. For each voxel, a relative enhancement-time curve was obtained according to the equation: Srel ðtÞ ¼ ðSðtÞ S0 Þ=S0 , where S(t) represents the signal intensity at a time (t) and S0 the baseline signal intensity, i.e. the average of the three pre-contrast scans. Six semiquantitative kinetic parameters were calculated: (1) Initial slope, based on the initial 3 upslope data points of the enhancement-time curve. (2) Initial peak, represents the magnitude of enhancement at the end of the initial slope. (3) Areas under the first 60, 90, and 120 s of the enhancement curve (AUC60, AUC90, and AUC120), the integrals of the respective data. (4) Late slope, based on the late phase of the enhancementtime curve, i.e. from the initial peak until the end of acquisition (Fig. 2). The aforementioned parameters are dimensionless since they were calculated from relative enhancement curves. Histopathology Histopathological assessment of the resection specimen was performed according to the method of Quirke et al. [19]. Tumor regression grade (TRG) was assessed by a dedicated pathologist according to Mandard et al. [20]; TRG1: Complete regression, no residual tumor cells, TRG2: Predominant fibrosis with isolated tumor cells, TRG3: More residual tumor cells but fibrosis still dominant, TRG4: Residual tumor outgrowing fibrosis, TRG5: no regression. Therapeutic response was stratified into two groups: Good responders were defined as TRG1–2 and poor responders were defined as TRG3–5 [21]. Statistical analyses All statistical analyses were performed with Statistical Package for the Social Sciences (SPSS 20.0, Inc, Chicago, Illinois, USA). Normally distributed kinetic parameters were compared between good and poor responders using an independent sample T-test. The nonparametric Mann–Whitney U test was performed for non-normally distributed data. p-values
