ONCH-1971; No. of Pages 10
ARTICLE IN PRESS
Critical Reviews in Oncology/Hematology xxx (2015) xxx–xxx
Bevacizumab-based neoadjuvant chemotherapy for colorectal cancer liver metastases: Pitfalls and helpful tricks in a review for clinicians Filippo Pietrantonio a,∗ , Armando Orlandi b , Alessandro Inno c , Valentina Da Prat a , Daniele Spada d , Alessandro Iaculli e , Maria Di Bartolomeo a , Carlo Morosi f , Filippo de Braud a a
Medical Oncology Department, Fondazione IRCCS Istituto Nazionale dei Tumori, Milan, Italy b Medical Oncology Department, Università Cattolica del Sacro Cuore, Rome, Italy c Medical Oncology, Sacro Cuore Don Calabria Hospital, Negrar, Verona, Italy d Medical Oncology Department, Hospital of Urbino, Urbino, Italy e Medical Oncology, Bolognini Hospital, Seriate, Bergamo, Italy f Radiology Department, Fondazione IRCCS Istituto Nazionale dei Tumori, Milan, Italy Accepted 14 April 2015
Contents 1. 2.
3.
4.
Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Pre-surgical phase . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1. CT-SCAN: criteria of evaluation response . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1.1. PET-TC vs. CT-SCAN: what’s the better choice? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1.2. CT-SCAN behind evaluation of efficacy: Radiology predictor factor? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2. New imaging techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Post-surgical phase . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1. Pathological evaluation response . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1.1. Residual viable tumor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1.2. Tumor regression grade . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1.3. Modified tumor regression grade . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1.4. Tumor thickness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conflict of interest statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Reviewers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Biographies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
Abstract Bevacizumab added to chemotherapy has shown encouraging efficacy in the neoadjuvant therapy of colorectal cancer liver metastases. In absence of biological predictor factors of efficacy to bevacizumab-based treatment, the assessment of response may be a crucial point to select patients who may benefit the most from surgery. At the same time the pathological response after liver resection could represent a guide for the next therapeutic plan. In the pre-surgical phase, conventional computed tomography and response evaluation with RECIST criteria may underestimate the response to anti-angiogenic drugs. Modified computed tomography criteria of response, morphologic changes as well as
∗ Corresponding author at: Medical Oncology Department, Fondazione IRCCS Istituto Nazionale dei Tumori, Via Venezian, 1 – 20133 Milan, Italy. Tel.: +39 0223903807; fax: +39 0223902149. E-mail address: [email protected] (F. Pietrantonio).
http://dx.doi.org/10.1016/j.critrevonc.2015.04.008 1040-8428/© 2015 Elsevier Ireland Ltd. All rights reserved.
Please cite this article in press as: Pietrantonio F, et al. Bevacizumab-based neoadjuvant chemotherapy for colorectal cancer liver metastases: Pitfalls and helpful tricks in a review for clinicians. Crit Rev Oncol/Hematol (2015), http://dx.doi.org/10.1016/j.critrevonc.2015.04.008
ONCH-1971; No. of Pages 10
2
ARTICLE IN PRESS F. Pietrantonio et al. / Critical Reviews in Oncology/Hematology xxx (2015) xxx–xxx
novel imaging techniques and metabolic assessment by fluorodeoxyglucose positron emission tomography seem to be promising methods for the assessment of response and for leading the clinical choices. Pathological response at the time of surgery is an important prognostic factor and a surrogate of survival for resected patients. Different classification criteria to assess pathological response have been developed, residual viable tumor, tumor regression grade (TRG), modified TRG and tumor thickness at the tumor-normal interface, but to date a superiority of one approach over the others has not been clearly established. In this review, we evaluate the available data with the aim to help the clinicians in the pre- and post-surgical care of patient with colorectal cancer liver metastases treated with bevacizumab-based neoadjuvant strategy. © 2015 Elsevier Ireland Ltd. All rights reserved.
Keywords: Neoadjuvant; Bevacizumab; Colorectal cancer; Liver metastases; Response criteria
1. Introduction Worldwide, colorectal cancer (CRC) is a leading cause of tumor-related morbidity and mortality [1]. Most patients present with advanced disease or develop metastases within the first years after curative surgery, with the liver being the main site of metastatic involvement [2]. Strategies including hepatic resection are the only potentially curative treatment for colorectal cancer liver metastases (CLM). Therefore, patients are usually classified according to whether they have unresectable metastases, borderline resectable lesions or “upfront’ resectable disease [3]. Unfortunately, only 10%–20% of patients are diagnosed with initially resectable CLM [4]. Perioperative or neoadjuvant chemotherapy, with or without biological agents, is the standard of care for most patients with CLM, except those with low-risk profile [5]. In fact, neoadjuvant therapy is able to convert an initially unresectable lesion to resectability (conversion treatment) and to allow R0 resection in borderline resectable metastases. Moreover, perioperative chemotherapy has been shown to extend disease free survival (DFS) also in patients with initially resectable CLM [3]. In prospective trials, the addition of anti-epidermal growth factor (EGFR) or anti-vascular endothelial growth factor (VEGF) monoclonal antibodies to chemotherapy resulted in promisingly high resection rates. Anti-angiogenic agents have shown encouraging efficacy as neoadjuvant strategy in several studies. In the phase II BOXER trial, bevacizumab in combination with capecitabine and oxaliplatin (XELOX) lead to 40% of conversion of initially unresectable patients with poor risk features [6]. In the randomized phase II OLIVIA study, triplet chemotherapy with FOLFOXIRI plus bevacizumab – when compared to standard modified FOLFOX-6 plus bevacizumab – improved response rate, radical resections and progression-free survival (PFS) in patients with unresectable liver-only disease [7]. However, this trial investigated the role of bevacizumab-containing triplet chemotherapy versus doublets, rather than exploring the role of anti-angiogenic drugs in the neoadjuvant setting. In the planning of an integrated treatment of CLM with perioperative chemotherapy and surgery, two decision points seem to be crucial for clinicians: the pre-surgical and postsurgical phases. In the pre-surgical one, the assessment of
treatment activity is a mainstay step, in order to select patients who may benefit from liver resection with radical intent and to determine the best timing for surgery. On the opposite side, the post-surgical phase is conditioned by the pathological assessment of treatment activity, with the aim of planning the next therapeutic approach – and confirming whether or not the pre-surgical treatment had a sufficient level of efficacy. Despite the time elapsed since the introduction of bevacizumab into the clinical practice, currently the validation of predictive biomarkers of efficacy is still an unmet need [8]. In absence of biomarkers of response to Bevacizumab, in the presurgical phase a reproducible radiological assessment of response is essential, while histological evaluation of treatment activity plays a major role in the postsurgical phase. However, both radiological and pathological assessments of response need to be updated in the new era of targeted therapies. In fact, conventional imaging criteria do not perfectly apply to anti-angiogenic drugs such as bevacizumab, as well as standard pathological examination may be suboptimal for evaluation of response to this biological agent. In this review, we analyze and compare the tools for radiological and pathological assessment of response, to help clinicians in the difficult management of metastatic colorectal cancer patients with liver-limited disease who could receive a curative approach with bevacizumab-based therapy follwed by resection.
2. Pre-surgical phase A correct and early evaluation of efficacy of neoadjuvant treatments has a significant role in both the conversion setting and in initially resectable CLM. Indeed, in borderline or unresectable disease (the so-called “conversion setting”), in absence of a substantial modification of size of metastases, morphological changes within CLM can determine the continuation of treatment or a change of therapy to achieve the resectability. At the same time, in patients with resectable CLM, an early evidence of treatment refractoriness can lead the multidisciplinary team to shift to immediate surgery. Therefore, the early assessment of response with the most accurate tools is crucial to evaluate treatment activity and to drive treatment decision planning.
Please cite this article in press as: Pietrantonio F, et al. Bevacizumab-based neoadjuvant chemotherapy for colorectal cancer liver metastases: Pitfalls and helpful tricks in a review for clinicians. Crit Rev Oncol/Hematol (2015), http://dx.doi.org/10.1016/j.critrevonc.2015.04.008
ARTICLE IN PRESS
ONCH-1971; No. of Pages 10
F. Pietrantonio et al. / Critical Reviews in Oncology/Hematology xxx (2015) xxx–xxx
3
Table 1 Assessment of target lesion response: conventional RECIST criteria (a) versus Choi modified criteria (b) in colorectal liver metastases. (a) RECIST
(b) Choi modified CT criteria
CR = Disappearance of all target lesions
CR = Disappearance of all target lesion
PR = At least a 30% decrease in the sum of diameters of target lesions, taking as reference the baseline sum of the diameters of target lesions
PR = ≥10% decrease in target lesion size or ≥15% decrease in tumor density (in Hounsfield units) of target lesion
SD = Any cases that do not qualify for either partial response or progressive disease
SD = Any cases that do not qualify for either partial response or progressive disease
PD = An increase of at least 20% in the sum of the diameters of target lesions, taking as reference the smallest sum of the diameters of target lesions recorded since treatment started
PD = ≥10% increase in target lesion size and does not meet criteria of PR by tumor density (in Hounsfield units) on CT
Table 2 Computed tomography morphology groups. Morphology group
Overall attenuation
Tumor-liver interface
Peripheral rim of enhancement
3 2 1
Heterogeneous Mixed Homogeneous and hypo-attenuating
Ill defined Variable Sharp
May be present If initially present, partially resolved If initially present, completely resolved
2.1. CT-SCAN: criteria of evaluation response The efficacy of chemotherapy is usually assessed through the Response Evaluation Criteria in Solid Tumors (RECIST) guideline version 1.1, according to size variation of the tumor lesions (Table 1a) [9]. RECIST criteria have important limitations in assessing the efficacy of anti-angiogenic therapy either as monotherapy or in combination with cytotoxic drugs, as evidenced by recent studies in various tumors [10–12]. In fact, antivascular targeted therapy have a marginal effect on tumor shrinkage, but cause necrosis, cavitation and bleeding of the tumor [13–15]. In addition, targeted therapies induce intra-tumor changes that may result in an apparent and paradox volumetric volume increase, simulating disease progression or stability. Several studies have confirmed the need for additional criteria for response evaluation in patients with metastatic CRC treated with bevacizumab [16–18]. In a recent study, the authors compared response in two groups of patients with CLM treated with either chemotherapy alone (group A) or chemotherapy combined with bevacizumab (group B). Response was assessed using both RECIST 1.1 and the modified CT criteria proposed by Choi et al. [19] (Table 1b). According to RECIST criteria, 9 out of 30 patients (30%) in group A and 12 out of 29 patients (41%) in group B were good responders. On the contrary, modified CT criteria revealed that 23 out of 30 patients in group A (77%) and 23 out of 29 patients in group B (79%) were good responders. Modified CT criteria evidenced that good responders in both groups had significantly longer time to tumor progression (TTP) than poor responders (p < 0.05). On the other side, RECIST 1.1 showed that good responders in group A had significantly longer TTP than poor responders in the same group (p = 0.0154), with no difference among patients in group B. Therefore, the authors suggested the
addition of tumor density changes to the new set of criteria for evaluating treatment response, particularly regarding bevacizumab-based chemotherapy. Two studies have associated pathological response of CLM with radiological changes in the CT-SCAN, leading to the introduction of morphologic response criteria [17,18]. As summarized in Table 2, these criteria divide the tumor lesions into three main groups. The first group is represented by lesions with a homogeneous structure and well-defined edges, whereas the third group encompasses metastases with non-homogeneous structure and poorly defined edges. Lesions with intermediate features fall into the second group. The morphologic response criteria distinguish between two kinds of response: “optimal” if metastases move from group 3 or 2 to group 1 and “incomplete” if metastases rise from group 3 to group 2; when metastases remain in the same group, there is no response (Table 3). Morphologic response criteria were compared to RECIST in order to assess their accuracy in predicting pathological response. A retrospective analysis was performed in CLM patients treated with neoadjuvant chemotherapy plus bevacizumab followed by surgery. The aim of this analysis was to propose morphologic CT criteria as newer biomarkers of the biological effects of anti-angiogenic treatments and of the pathological response – being the latter a surrogate marker of patients’ outcome. Complete or good pathological responses – as defined by Blazer et al. [20] – were found to match to “optimal” Table 3 Morphologic Response Criteria in colorectal liver metastases. Morphologic response criteria Optimal response Incomplete response No response
The metastases changed from group 3 or 2 to 1 The metastases changed from group 3 to 2. The metastases no changed in group 3 or 2
Please cite this article in press as: Pietrantonio F, et al. Bevacizumab-based neoadjuvant chemotherapy for colorectal cancer liver metastases: Pitfalls and helpful tricks in a review for clinicians. Crit Rev Oncol/Hematol (2015), http://dx.doi.org/10.1016/j.critrevonc.2015.04.008
ONCH-1971; No. of Pages 10
4
ARTICLE IN PRESS F. Pietrantonio et al. / Critical Reviews in Oncology/Hematology xxx (2015) xxx–xxx
morphologic responses in 22 out of 29 patients (76%), whereas the minor pathologic responses were associated with “incomplete” morphologic responses or no responses in 17 out of 21 patients (81%). Therefore, morphologic criteria response showed a superiority over size-based RECIST criteria. In addition, optimal morphologic response was associated with a significant survival benefit both in patients undergoing surgery and in patients with unresectable metastases. In the latter group, patients with optimal morphologic response had an overall survival of 31 months, whereas those with incomplete response had a survival of 19 months (p = 0.009). On the other hand, the response rate evaluated by RECIST criteria did not show any association with overall survival. In fact, patients with partial response had an overall survival of 28 months, while those with stable or progressive disease of 22 months (p = 0.45) [17]. A further study has confirmed the correlation between morphologic complete response and survival, both in terms of DFS (21.1 vs. 11.8 months, p = 0.004) and overall survival (114.2 vs. 49 months, p = 0.0009) in patients with resected CLM. This finding was statistically significant in a multivariate analysis [18]. Although the morphologic response criteria have shown superior diagnostic accuracy, they still have some limitations. First of all, the aforementioned studies were conducted retrospectively and need to be validated by larger prospective studies. A greater standardization of morphologic criteria is needed: in fact, they rely on subjective assessments, rather than on objective numeric values like the RECIST criteria. In addition, the morphologic evaluation of liver lesions can be performed with accuracy only for lesions with a diameter greater than 1–1.5 cm. Taken together all these considerations, the morphologic response criteria still cannot replace RECIST criteria in clinical practice, but they could be integrated in the evaluation of response for selected patients in the context of multidisciplinary management.
liver metastases after neoadjuvant treatment (65.3% vs. 49%, respectively; p < 0.0001). Moreover, sensitivity of the FDG-PET/CT, but not of CT, was lower in patients who received chemotherapy plus bevacizumab compared to the chemotherapy-only group, although the difference was not statistically significant (39% vs. 59%, p = 0.068). In this study pre-operative screening with FDG-PET/CT seemed to increase the accuracy of staging of patients who are candidates for liver resection, but its sensitivity significantly decreased after neoadjuvant chemotherapy. Similarly, Small et al. [22] highlighted that FDG-PET represents an optimal diagnostic modality with a high sensitivity and specificity for detecting hepatic and especially extra-hepatic disease. A meta-analysis by Wiering et al. [23] demonstrated that FDG-PET technique has a sensitivity of 88% and specificity of 91.5% for the detection of hepatic disease, and of 91.5% and 95.4% respectively for detecting extra-hepatic disease. However, this technique may fail to detect metastatic liver lesions smaller than 1 cm, has a poor accuracy for mucinous tumors and anatomic details, and has a limited spatial resolution. Therefore, FDG-PET/CT may underestimate the hepatic tumor burden and make tumor site localization difficult [24,25]. Finally, Nasti et al. presented a phase II trial to assess activity of neoadjuvant FOLFIRI plus bevacizumab in patients with CLM [26]. This study demonstrated that FDG-PET/CT response after 3 months of treatment was significantly predictive of long term outcome, with a higher accuracy than CT-SCAN evaluation based on RECIST criteria. However, this result needs further evidence to be confirmed and potentially used in the clinical practice. In conclusion, even in presence of a strong biological rationale for using FDG-PET/CT as imaging biomarker of early response to bevacizumab-based neoadjuvant therapy, in view of the limitations described above, this technique cannot be yet considered standard diagnostic modalities in clinical settings.
2.1.1. PET-TC vs. CT-SCAN: what’s the better choice? Molecular imaging with fluorodeoxiglucose positron emission tomography and computed tomography (FDGPET/CT) is a useful technique in the management of CRC. FDG-PET/CT could be important in the preoperative staging of potentially resectable CLM. Nowadays, there is an increasing interest in its ability to predict response to biological agents. FDG-PET/CT could detect early metabolic changes in tumor cell metabolism, before any change in tumor size occurs. Lubezky et al. [21] reported the first experiences with FDG-PET/CT in the restaging of patients with CLM after neoadjuvant chemotherapy with or without bevacizumab, and in patients selected for upfront surgery. FDG-PET/CT demonstrated a higher sensitivity than contrast-enhanced CT for the detection of CLM in patients who did not receive chemotherapy rather than patients who received chemotherapy (FDG-PET/CT: 93.3 vs. 49%, p < 0.0001; CT: 87.5 vs. 65.3, p = 0.038). On the contrary, contrast-enhanced CT revealed a higher sensitivity than FDG-PET/CT for detecting
2.1.2. CT-SCAN behind evaluation of efficacy: Radiology predictor factor? A novel approach involves the enhancement gradient, defined as the difference in Hounsfield units between CTSCAN with and without iodinated contrast material. This gradient represents the density of vascularization within the neoplasia, as demonstrated in several studies [19,27–31]. A recent study showed that the enhancement gradient in a baseline CT-SCAN correlates with radiological responses to anti-angiogenic therapy [32]. This trial had several limitations, because of the retrospective conduction of the study and the unspecified response evaluation criteria; however, it represents a promising model for the identification of predictors of biological activity of bevacizumab. This approach could potentially change the role of CT-SCAN, which would become not only a tool to evaluate chemotherapy activity, but also a means to stratify patients who will likely benefit from an anti-angiogenic treatment.
Please cite this article in press as: Pietrantonio F, et al. Bevacizumab-based neoadjuvant chemotherapy for colorectal cancer liver metastases: Pitfalls and helpful tricks in a review for clinicians. Crit Rev Oncol/Hematol (2015), http://dx.doi.org/10.1016/j.critrevonc.2015.04.008
ARTICLE IN PRESS
ONCH-1971; No. of Pages 10
F. Pietrantonio et al. / Critical Reviews in Oncology/Hematology xxx (2015) xxx–xxx
2.2. New imaging techniques
3. Post-surgical phase
A novel technique to improve response evaluation to antiangiogenic agents represents contrast-enhanced ultrasound (CEUS) with the opportunity to detect changes of tumor perfusion at early time-points after administration of chemotherapy. In a little prospective study with 30 patients with unresectable CLM underwent to FOLFIRI+ Bevacizumab, Schirin-Sokhan et al. had demonstrate in liver metastases that baseline time to reach the maximum peak of contrast intensity (time to peak: TTP) and TTP quotient are significantly lower in the group of the responders compared to the non-responders. Furthermore, correlating to the antiangiogenic effect of bevacizumab we observed a strong increase in TTP and TTP quotient during chemotherapy, which was restricted to the group of the responders [33]. Despite these encouraging results, this study has some limitations for application in CLM setting both for the number of patients and for the absence of a comparison with the pathologic response being concerned patients with unresectable metastases. An interesting emerging role for this technique is the intraoperative use. Preoperative chemotherapy sometimes makes CLM disappear. Contrast-enhanced intraoperative ultrasound (CE-IOUS) using perflubutane may identify such metastases. A recent study had demonstrated that CE-IOUS detects 79% of the disappearing liver metastases (DLM) and sensitivity of CE-IOUS was superior to CT-SCAN (p < 0.04) [34]. The authors suggest CE-IOUS might be necessary after preoperative chemotherapy for disappeared CLM. Recently, a computerized tomography technology named dual-energy CT (DECT) entered the medical practice [35]. Thanks to two sources of radiation emission, DECT is able to reduce the patient global exposure to radiation, to obtain higher resolution images and to better assess the hemorrhagic and vascular changes in the tumor [35,36]. A small study in non-small-cell lung cancer analyzed the accuracy of DECT for response evaluation in anti-angiogenic therapy, showing an improvement in radiological assessment of tumor changes [36]. Therefore, this technique could confer more objectiveness to the morphologic response criteria to anti-angiogenic therapies, representing an important tool for clinical practice. However, these results need to be confirmed in larger prospective trials including other cancer types including CRC. Another promising technology is dynamic contrastenhanced magnetic resonance imaging (DCE-MRI). This technique assesses contrast inflow and egress from a region of interest in the neoplasia, and it is widely used for response evaluation in anti-angiogenetic therapies. In fact, the rapid acquisition of images before and after contrast infusion is able to reveal changes in tumor vascularization [37]. Furthermore, DCE-MRI is able to monitor structure changes and vascular permeability variations, thus representing an interesting biomarker to estimate bevacizumab effects [35]. These encouraging results need further studies to be applied in randomized clinical trials in order to become a predictive factor of histological response.
3.1. Pathological evaluation response
5
Due to possible misinterpretation and lack of reproducibility of radiological assessment, preoperative imaging may have a suboptimal accuracy in the assessment of chemotherapy-induced tumor changes, especially in the neoadjuvant setting [38]. On the contrary, histological evaluation should have high accuracy and reproducibility in the evaluation of residual microscopic tumor. The essential role of pathological response as a prognostic factor has been demonstrated in patients receiving preoperative therapy for breast, esophageal, gastric and rectal cancer [39–42]. Recently, histological analyses have been shown to be associated with survival in patients with CLM, with pathologic complete response being an independent prognostic factor of long-term outcome [20,43]. To date, there is a growing body of evidence suggesting that bevacizumab may improve pathologic response of CLM when added to preoperative chemotherapy (Table 4). Despite of the objective evaluation of the residual tumor after neoadjuvant therapy, the pathologic response can be assessed with different classification criteria, such as residual viable tumor, tumor regression grade or its modified version, and tumor thickness. Especially after neoadjuvant treatment with antiangiogenetic therapy, there is not unanimous consensus for the better classification to use. Indeed, an optimal quantification of pathological response may help to orient clinical choices, define patients prognosis and establish the subsequent therapeutic plan. 3.1.1. Residual viable tumor An intriguing method for pathological response assessment in CLM was described by Blazer et al. [20]. It consists of a semi-quantitative assessment of residual cancer cells referring to the total tumor area. Pathological response was categorized in 3 groups: complete (no residual cancer cells remaining), major (1–49%) or minor response (≥50%). In patients with multiple tumor nodules, the mean of the values for the various nodules was used to define the pathological response. In this study, treatment with oxaliplatin-based chemotherapy plus bevacizumab produced a higher rate of pathological complete or major response as compared to chemotherapy alone (63% versus 44%, p < 0.001). In multivariate analysis including age, lymph nodal status, disease-free interval, CEA level, tumor size, number of lesions and treatment type, fluoropyrimidine plus oxaliplatin and bevacizumab therapy was confirmed as an independent predictor of pathological response, which significantly correlated with OS (major response: HR 4.80, p = 0.34; minor response: HR 6.93, p = 0.007). Independently of treatment received, cumulative 5-year OS rates were 75% for a complete pathologic response, 56% for a major pathologic response and
Please cite this article in press as: Pietrantonio F, et al. Bevacizumab-based neoadjuvant chemotherapy for colorectal cancer liver metastases: Pitfalls and helpful tricks in a review for clinicians. Crit Rev Oncol/Hematol (2015), http://dx.doi.org/10.1016/j.critrevonc.2015.04.008
ONCH-1971; No. of Pages 10
6
ARTICLE IN PRESS F. Pietrantonio et al. / Critical Reviews in Oncology/Hematology xxx (2015) xxx–xxx
Table 4 Neoadjuvant anti-angiogenic therapy: recent studies correlating treatment with pathologic response. Study (year)
Treatment (n)
Pathological response, n (%) Complete
Ribero et al. (2007) [52]
5FU + oxaliplatin (43) 5FU + oxaliplatin + bevacizumab (62)
5 (12) 7 (11)
Blazer et al. (2008) [20]
5FU + oxaliplatin (81) 5FU + oxaliplatin + bevacizumab (50)
6 (12) 7 (9)
Kishi et al. (2010) [44]
FOLFOX (117) FOLFOX + bevacizumab (102)
Klinger et al. (2010) [45]
FOLFOX (50) FOLFOX + bevacizumb (50)
Bibeau et al. (2013) [49]
No treatment (29) Chemotherapy (31) Chemotherapy + bevacizumab (31)
Loupakis et al. (2013) [46]
FOLFOXIRI/XELOXIRI (18) FOLFOXIRI + bevacizumab (24)
33% for a minor pathologic response (complete vs. major response, p = 0.037; major vs. minor response, p = 0.28) [20]. Similarly, a further non-randomized observational study was conducted on 219 patients with CLM treated with FOLFOX with or without bevacizumab [44]. The frequency of the above mentioned complete or major pathologic response rate was higher after treatment with FOLFOX plus Bevacizumab versus chemotherapy alone (70% versus 45%, p < 0.001). These data suggest that Bazer et al. classification allows to highlight the difference in pathological response effectiveness between chemotherapy with or without bevacizumab. 3.1.2. Tumor regression grade Rubbia-Brandt et al. were the first to define histological criteria of response to preoperative chemotherapy in CLM based on the tumor regression grade (TRG) [43]. These criteria had previously been elaborated by Mandard to evaluate the effect of neoadjuvant treatment on esophageal carcinoma [40]. This method identifies five TRGs classes on the bases of residual tumor cells and the extent of fibrosis: TRG1, absence of residual cancer cells replaced with a large amount of fibrosis; TRG2, rare scattered residual cancer cells and abundant fibrosis; TRG3, more residual cancer cells and predominant fibrosis; TRG4, large amount of cancer cells predominating over fibrosis; and TRG5, most exclusively cancer cells without fibrosis. Due to intra-individual variability, patients with multiple liver metastases who showed different TRG were categorized according to the worst TRG. The TRG was then grouped in 3 classes: major histologic response
8 (7) 10 (10)
Major
Complete + Major
ILN
10 (23) 28 (45) p = 0.02 16 (32) 44 (54) p < 0.001
22 (44) 51 (63)
45 (38) 61 (60)
53 (45) 71 (70)
5 (10) 19 (38) p < 0.001 0 (0) 12 (39) 21 (68) p < 0.001 2 (11) 4 (16)
– –
5 (28) 15 (63) p = 0.033
4 (27) 19 (83) p = 0.0009
(MjHR; TRG1 and TRG2), partial histologic response (PHR; TRG3) and no histologic response (NHR; TRG4 and 5). In Rubbia-Brandt et al. study, MjHR was associated with better 3-year DFS (49% vs. 18%, log-rank p = 0.0014) and 5-year OS (41% vs. 9%, log-rank p = 0.0003) than NHR. Moreover, in a multivariate analysis including patients age, number and size of metastases, synchronous versus metachronous metastases and chemotherapy, TRG was an independent prognostic factor for 5-year DFS (hazard ratio, HR 0.71, p = 0.001) and OS (HR 0.55, p = 0.004) [43]. In addition, a retrospective analysis of two nonrandomized trials with preoperative fluoropyrimidines and oxaliplatin with or without bevacizumab showed that patients treated with bevacizumab achieved a significantly higher major histological response compared to those treated with chemotherapy only (38% versus 10%; p < 0.001), using the TRG criteria [45]. Finally, a retrospective study analyzed 18 patients treated with FOLFOXIRI or XELOXIRI and 24 patients treated with FOLFOXIRI plus bevacizumab who underwent secondary resection of CLM. In this study, major histological response according to TRG was 63% and 28% for patients receiving and not receiving the anti-VEGF antibody, respectively (p = 0.033). In the bevacizumab group, patients with a TRG 1, 2 or 3 in comparison to those with TRG 4 or 5 had an advantage in terms of PFS (HR 0.41, p = 0.031). OS data were not mature at the time of analysis [46]. Taken together these results, TRG classification allows to emphasize difference effectiveness in pathological response between chemotherapy with and without bevacizumab.
Please cite this article in press as: Pietrantonio F, et al. Bevacizumab-based neoadjuvant chemotherapy for colorectal cancer liver metastases: Pitfalls and helpful tricks in a review for clinicians. Crit Rev Oncol/Hematol (2015), http://dx.doi.org/10.1016/j.critrevonc.2015.04.008
ONCH-1971; No. of Pages 10
ARTICLE IN PRESS F. Pietrantonio et al. / Critical Reviews in Oncology/Hematology xxx (2015) xxx–xxx
3.1.3. Modified tumor regression grade According to the TRG, only fibrosis but not necrosis is considered as a marker of tumor response. The usual necrosis seen in CLM is the so-called “dirty” necrosis and is characterized by nuclear debris in a patchy distribution bordered by viable cells. It seems to be caused by spontaneous phenomena within the tumor, such as insufficient vascular supply, rather than by the treatment itself. In fact, a large amount of necrosis is seen in patients who respond poorly or in those not receiving preoperative treatment [47]. However, some other authors have interpreted the necrosis as a form of response to therapy. Chang et al. [48] identified a particular type of necrosis, the “infart-like” necrosis (ILN), which differs from the “dirty necrosis” because of the large confluent areas of eosinophilic cytoplasmic remnants, located centrally within a lesion and surrounded by a rim of fibrosis with foamy macrophages. The Chang et al. retrospective trial included 109 patients, of whom 46 patients received perioperative treatment. ILN was only observed in preoperatively treated patients and not in untreated patients, and it significantly correlated with improved DFS (log-rank p = 0.047) [48]. The association between bevacizumab and ILN was also confirmed in larger series retrospectively evaluated [49]. Based on this observation, ILN was incorporated by Chang et al. in a modified TRG (mTRG) which considers the presence of ILN a form of therapeutic effect equivalent to fibrosis. The inclusion of ILN into the definition of treatment effect altered the tumor regression score in about a quarter of patients. Modified TRG (mTRG) 1–2 scores were associated with significantly better DFS (log-rank p = 0.007) and OS (log-rank p = 0.020) compared with mTRG 3–5 scores, whereas statistical significance was not reached using the classic TRG score to detect survival differences. Even if this may partially be explained by the relatively small sample size, incorporating ILN in the modified grading system might enhance the prognostic utility of TRG after bevacizumab treatment [48]. 3.1.4. Tumor thickness Maru et al. introduced the tumor thickness at the tumornormal interface (TNI) as an indicator of pathologic response to preoperative chemotherapy. The maximum thickness of uninterrupted layers of tumor cells was measured perpendicular to the TNI by a ruler. Greater thickness predicted shorter recurrence-free survival (RFS), and its correlation with RFS remained in multivariate analysis (p = 0.015) [50]. Maru et al. found that TNI was also correlated with mRECIST and with the response criteria from Blazer et al. Additional observation of that study was that tumors from patients treated with bevacizumab had smaller tumor thickness at TNI than did tumors from patients who did not receive bevacizumab, and they found significantly thinner median TNI in patients who received oxaliplatin-based therapy than in patients who received irinotecan-based therapy. Nowadays there is only a study witch compare grade of residual viable tumor (Blazer et al.), tumor regression grade
7
(TRG) scoring system (Rubbia-Brandt et al.), modified tumor regression grade (mTRG) scoring system with the type of necrosis (Chang et al.), and the tumor thickness at the TNI (Maru et al.) in patients with CLM and underwent to neoadjuvant bevacizumab-based chemotherapy. Dede et al. analyzed if these histopathologic changes are specific to preoperative chemotherapy and if these methods have correlation with survival. In a retrospective study with 46 patients underwent to liver surgery after neodjuvant chemotherapy with bevacizumab, none of the analyzed pathological methods showed significant correlation with progression free survival (PFS) or with overall survival (OS). Residual tumor cell ratio and TRG showed positive but not significant correlation with OS and PFS. The PFS and OS curves showed a slight difference in the group of patients with TNI
ARTICLE IN PRESS
Critical Reviews in Oncology/Hematology xxx (2015) xxx–xxx
Bevacizumab-based neoadjuvant chemotherapy for colorectal cancer liver metastases: Pitfalls and helpful tricks in a review for clinicians Filippo Pietrantonio a,∗ , Armando Orlandi b , Alessandro Inno c , Valentina Da Prat a , Daniele Spada d , Alessandro Iaculli e , Maria Di Bartolomeo a , Carlo Morosi f , Filippo de Braud a a
Medical Oncology Department, Fondazione IRCCS Istituto Nazionale dei Tumori, Milan, Italy b Medical Oncology Department, Università Cattolica del Sacro Cuore, Rome, Italy c Medical Oncology, Sacro Cuore Don Calabria Hospital, Negrar, Verona, Italy d Medical Oncology Department, Hospital of Urbino, Urbino, Italy e Medical Oncology, Bolognini Hospital, Seriate, Bergamo, Italy f Radiology Department, Fondazione IRCCS Istituto Nazionale dei Tumori, Milan, Italy Accepted 14 April 2015
Contents 1. 2.
3.
4.
Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Pre-surgical phase . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1. CT-SCAN: criteria of evaluation response . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1.1. PET-TC vs. CT-SCAN: what’s the better choice? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1.2. CT-SCAN behind evaluation of efficacy: Radiology predictor factor? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2. New imaging techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Post-surgical phase . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1. Pathological evaluation response . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1.1. Residual viable tumor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1.2. Tumor regression grade . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1.3. Modified tumor regression grade . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1.4. Tumor thickness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conflict of interest statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Reviewers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Biographies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
Abstract Bevacizumab added to chemotherapy has shown encouraging efficacy in the neoadjuvant therapy of colorectal cancer liver metastases. In absence of biological predictor factors of efficacy to bevacizumab-based treatment, the assessment of response may be a crucial point to select patients who may benefit the most from surgery. At the same time the pathological response after liver resection could represent a guide for the next therapeutic plan. In the pre-surgical phase, conventional computed tomography and response evaluation with RECIST criteria may underestimate the response to anti-angiogenic drugs. Modified computed tomography criteria of response, morphologic changes as well as
∗ Corresponding author at: Medical Oncology Department, Fondazione IRCCS Istituto Nazionale dei Tumori, Via Venezian, 1 – 20133 Milan, Italy. Tel.: +39 0223903807; fax: +39 0223902149. E-mail address: [email protected] (F. Pietrantonio).
http://dx.doi.org/10.1016/j.critrevonc.2015.04.008 1040-8428/© 2015 Elsevier Ireland Ltd. All rights reserved.
Please cite this article in press as: Pietrantonio F, et al. Bevacizumab-based neoadjuvant chemotherapy for colorectal cancer liver metastases: Pitfalls and helpful tricks in a review for clinicians. Crit Rev Oncol/Hematol (2015), http://dx.doi.org/10.1016/j.critrevonc.2015.04.008
ONCH-1971; No. of Pages 10
2
ARTICLE IN PRESS F. Pietrantonio et al. / Critical Reviews in Oncology/Hematology xxx (2015) xxx–xxx
novel imaging techniques and metabolic assessment by fluorodeoxyglucose positron emission tomography seem to be promising methods for the assessment of response and for leading the clinical choices. Pathological response at the time of surgery is an important prognostic factor and a surrogate of survival for resected patients. Different classification criteria to assess pathological response have been developed, residual viable tumor, tumor regression grade (TRG), modified TRG and tumor thickness at the tumor-normal interface, but to date a superiority of one approach over the others has not been clearly established. In this review, we evaluate the available data with the aim to help the clinicians in the pre- and post-surgical care of patient with colorectal cancer liver metastases treated with bevacizumab-based neoadjuvant strategy. © 2015 Elsevier Ireland Ltd. All rights reserved.
Keywords: Neoadjuvant; Bevacizumab; Colorectal cancer; Liver metastases; Response criteria
1. Introduction Worldwide, colorectal cancer (CRC) is a leading cause of tumor-related morbidity and mortality [1]. Most patients present with advanced disease or develop metastases within the first years after curative surgery, with the liver being the main site of metastatic involvement [2]. Strategies including hepatic resection are the only potentially curative treatment for colorectal cancer liver metastases (CLM). Therefore, patients are usually classified according to whether they have unresectable metastases, borderline resectable lesions or “upfront’ resectable disease [3]. Unfortunately, only 10%–20% of patients are diagnosed with initially resectable CLM [4]. Perioperative or neoadjuvant chemotherapy, with or without biological agents, is the standard of care for most patients with CLM, except those with low-risk profile [5]. In fact, neoadjuvant therapy is able to convert an initially unresectable lesion to resectability (conversion treatment) and to allow R0 resection in borderline resectable metastases. Moreover, perioperative chemotherapy has been shown to extend disease free survival (DFS) also in patients with initially resectable CLM [3]. In prospective trials, the addition of anti-epidermal growth factor (EGFR) or anti-vascular endothelial growth factor (VEGF) monoclonal antibodies to chemotherapy resulted in promisingly high resection rates. Anti-angiogenic agents have shown encouraging efficacy as neoadjuvant strategy in several studies. In the phase II BOXER trial, bevacizumab in combination with capecitabine and oxaliplatin (XELOX) lead to 40% of conversion of initially unresectable patients with poor risk features [6]. In the randomized phase II OLIVIA study, triplet chemotherapy with FOLFOXIRI plus bevacizumab – when compared to standard modified FOLFOX-6 plus bevacizumab – improved response rate, radical resections and progression-free survival (PFS) in patients with unresectable liver-only disease [7]. However, this trial investigated the role of bevacizumab-containing triplet chemotherapy versus doublets, rather than exploring the role of anti-angiogenic drugs in the neoadjuvant setting. In the planning of an integrated treatment of CLM with perioperative chemotherapy and surgery, two decision points seem to be crucial for clinicians: the pre-surgical and postsurgical phases. In the pre-surgical one, the assessment of
treatment activity is a mainstay step, in order to select patients who may benefit from liver resection with radical intent and to determine the best timing for surgery. On the opposite side, the post-surgical phase is conditioned by the pathological assessment of treatment activity, with the aim of planning the next therapeutic approach – and confirming whether or not the pre-surgical treatment had a sufficient level of efficacy. Despite the time elapsed since the introduction of bevacizumab into the clinical practice, currently the validation of predictive biomarkers of efficacy is still an unmet need [8]. In absence of biomarkers of response to Bevacizumab, in the presurgical phase a reproducible radiological assessment of response is essential, while histological evaluation of treatment activity plays a major role in the postsurgical phase. However, both radiological and pathological assessments of response need to be updated in the new era of targeted therapies. In fact, conventional imaging criteria do not perfectly apply to anti-angiogenic drugs such as bevacizumab, as well as standard pathological examination may be suboptimal for evaluation of response to this biological agent. In this review, we analyze and compare the tools for radiological and pathological assessment of response, to help clinicians in the difficult management of metastatic colorectal cancer patients with liver-limited disease who could receive a curative approach with bevacizumab-based therapy follwed by resection.
2. Pre-surgical phase A correct and early evaluation of efficacy of neoadjuvant treatments has a significant role in both the conversion setting and in initially resectable CLM. Indeed, in borderline or unresectable disease (the so-called “conversion setting”), in absence of a substantial modification of size of metastases, morphological changes within CLM can determine the continuation of treatment or a change of therapy to achieve the resectability. At the same time, in patients with resectable CLM, an early evidence of treatment refractoriness can lead the multidisciplinary team to shift to immediate surgery. Therefore, the early assessment of response with the most accurate tools is crucial to evaluate treatment activity and to drive treatment decision planning.
Please cite this article in press as: Pietrantonio F, et al. Bevacizumab-based neoadjuvant chemotherapy for colorectal cancer liver metastases: Pitfalls and helpful tricks in a review for clinicians. Crit Rev Oncol/Hematol (2015), http://dx.doi.org/10.1016/j.critrevonc.2015.04.008
ARTICLE IN PRESS
ONCH-1971; No. of Pages 10
F. Pietrantonio et al. / Critical Reviews in Oncology/Hematology xxx (2015) xxx–xxx
3
Table 1 Assessment of target lesion response: conventional RECIST criteria (a) versus Choi modified criteria (b) in colorectal liver metastases. (a) RECIST
(b) Choi modified CT criteria
CR = Disappearance of all target lesions
CR = Disappearance of all target lesion
PR = At least a 30% decrease in the sum of diameters of target lesions, taking as reference the baseline sum of the diameters of target lesions
PR = ≥10% decrease in target lesion size or ≥15% decrease in tumor density (in Hounsfield units) of target lesion
SD = Any cases that do not qualify for either partial response or progressive disease
SD = Any cases that do not qualify for either partial response or progressive disease
PD = An increase of at least 20% in the sum of the diameters of target lesions, taking as reference the smallest sum of the diameters of target lesions recorded since treatment started
PD = ≥10% increase in target lesion size and does not meet criteria of PR by tumor density (in Hounsfield units) on CT
Table 2 Computed tomography morphology groups. Morphology group
Overall attenuation
Tumor-liver interface
Peripheral rim of enhancement
3 2 1
Heterogeneous Mixed Homogeneous and hypo-attenuating
Ill defined Variable Sharp
May be present If initially present, partially resolved If initially present, completely resolved
2.1. CT-SCAN: criteria of evaluation response The efficacy of chemotherapy is usually assessed through the Response Evaluation Criteria in Solid Tumors (RECIST) guideline version 1.1, according to size variation of the tumor lesions (Table 1a) [9]. RECIST criteria have important limitations in assessing the efficacy of anti-angiogenic therapy either as monotherapy or in combination with cytotoxic drugs, as evidenced by recent studies in various tumors [10–12]. In fact, antivascular targeted therapy have a marginal effect on tumor shrinkage, but cause necrosis, cavitation and bleeding of the tumor [13–15]. In addition, targeted therapies induce intra-tumor changes that may result in an apparent and paradox volumetric volume increase, simulating disease progression or stability. Several studies have confirmed the need for additional criteria for response evaluation in patients with metastatic CRC treated with bevacizumab [16–18]. In a recent study, the authors compared response in two groups of patients with CLM treated with either chemotherapy alone (group A) or chemotherapy combined with bevacizumab (group B). Response was assessed using both RECIST 1.1 and the modified CT criteria proposed by Choi et al. [19] (Table 1b). According to RECIST criteria, 9 out of 30 patients (30%) in group A and 12 out of 29 patients (41%) in group B were good responders. On the contrary, modified CT criteria revealed that 23 out of 30 patients in group A (77%) and 23 out of 29 patients in group B (79%) were good responders. Modified CT criteria evidenced that good responders in both groups had significantly longer time to tumor progression (TTP) than poor responders (p < 0.05). On the other side, RECIST 1.1 showed that good responders in group A had significantly longer TTP than poor responders in the same group (p = 0.0154), with no difference among patients in group B. Therefore, the authors suggested the
addition of tumor density changes to the new set of criteria for evaluating treatment response, particularly regarding bevacizumab-based chemotherapy. Two studies have associated pathological response of CLM with radiological changes in the CT-SCAN, leading to the introduction of morphologic response criteria [17,18]. As summarized in Table 2, these criteria divide the tumor lesions into three main groups. The first group is represented by lesions with a homogeneous structure and well-defined edges, whereas the third group encompasses metastases with non-homogeneous structure and poorly defined edges. Lesions with intermediate features fall into the second group. The morphologic response criteria distinguish between two kinds of response: “optimal” if metastases move from group 3 or 2 to group 1 and “incomplete” if metastases rise from group 3 to group 2; when metastases remain in the same group, there is no response (Table 3). Morphologic response criteria were compared to RECIST in order to assess their accuracy in predicting pathological response. A retrospective analysis was performed in CLM patients treated with neoadjuvant chemotherapy plus bevacizumab followed by surgery. The aim of this analysis was to propose morphologic CT criteria as newer biomarkers of the biological effects of anti-angiogenic treatments and of the pathological response – being the latter a surrogate marker of patients’ outcome. Complete or good pathological responses – as defined by Blazer et al. [20] – were found to match to “optimal” Table 3 Morphologic Response Criteria in colorectal liver metastases. Morphologic response criteria Optimal response Incomplete response No response
The metastases changed from group 3 or 2 to 1 The metastases changed from group 3 to 2. The metastases no changed in group 3 or 2
Please cite this article in press as: Pietrantonio F, et al. Bevacizumab-based neoadjuvant chemotherapy for colorectal cancer liver metastases: Pitfalls and helpful tricks in a review for clinicians. Crit Rev Oncol/Hematol (2015), http://dx.doi.org/10.1016/j.critrevonc.2015.04.008
ONCH-1971; No. of Pages 10
4
ARTICLE IN PRESS F. Pietrantonio et al. / Critical Reviews in Oncology/Hematology xxx (2015) xxx–xxx
morphologic responses in 22 out of 29 patients (76%), whereas the minor pathologic responses were associated with “incomplete” morphologic responses or no responses in 17 out of 21 patients (81%). Therefore, morphologic criteria response showed a superiority over size-based RECIST criteria. In addition, optimal morphologic response was associated with a significant survival benefit both in patients undergoing surgery and in patients with unresectable metastases. In the latter group, patients with optimal morphologic response had an overall survival of 31 months, whereas those with incomplete response had a survival of 19 months (p = 0.009). On the other hand, the response rate evaluated by RECIST criteria did not show any association with overall survival. In fact, patients with partial response had an overall survival of 28 months, while those with stable or progressive disease of 22 months (p = 0.45) [17]. A further study has confirmed the correlation between morphologic complete response and survival, both in terms of DFS (21.1 vs. 11.8 months, p = 0.004) and overall survival (114.2 vs. 49 months, p = 0.0009) in patients with resected CLM. This finding was statistically significant in a multivariate analysis [18]. Although the morphologic response criteria have shown superior diagnostic accuracy, they still have some limitations. First of all, the aforementioned studies were conducted retrospectively and need to be validated by larger prospective studies. A greater standardization of morphologic criteria is needed: in fact, they rely on subjective assessments, rather than on objective numeric values like the RECIST criteria. In addition, the morphologic evaluation of liver lesions can be performed with accuracy only for lesions with a diameter greater than 1–1.5 cm. Taken together all these considerations, the morphologic response criteria still cannot replace RECIST criteria in clinical practice, but they could be integrated in the evaluation of response for selected patients in the context of multidisciplinary management.
liver metastases after neoadjuvant treatment (65.3% vs. 49%, respectively; p < 0.0001). Moreover, sensitivity of the FDG-PET/CT, but not of CT, was lower in patients who received chemotherapy plus bevacizumab compared to the chemotherapy-only group, although the difference was not statistically significant (39% vs. 59%, p = 0.068). In this study pre-operative screening with FDG-PET/CT seemed to increase the accuracy of staging of patients who are candidates for liver resection, but its sensitivity significantly decreased after neoadjuvant chemotherapy. Similarly, Small et al. [22] highlighted that FDG-PET represents an optimal diagnostic modality with a high sensitivity and specificity for detecting hepatic and especially extra-hepatic disease. A meta-analysis by Wiering et al. [23] demonstrated that FDG-PET technique has a sensitivity of 88% and specificity of 91.5% for the detection of hepatic disease, and of 91.5% and 95.4% respectively for detecting extra-hepatic disease. However, this technique may fail to detect metastatic liver lesions smaller than 1 cm, has a poor accuracy for mucinous tumors and anatomic details, and has a limited spatial resolution. Therefore, FDG-PET/CT may underestimate the hepatic tumor burden and make tumor site localization difficult [24,25]. Finally, Nasti et al. presented a phase II trial to assess activity of neoadjuvant FOLFIRI plus bevacizumab in patients with CLM [26]. This study demonstrated that FDG-PET/CT response after 3 months of treatment was significantly predictive of long term outcome, with a higher accuracy than CT-SCAN evaluation based on RECIST criteria. However, this result needs further evidence to be confirmed and potentially used in the clinical practice. In conclusion, even in presence of a strong biological rationale for using FDG-PET/CT as imaging biomarker of early response to bevacizumab-based neoadjuvant therapy, in view of the limitations described above, this technique cannot be yet considered standard diagnostic modalities in clinical settings.
2.1.1. PET-TC vs. CT-SCAN: what’s the better choice? Molecular imaging with fluorodeoxiglucose positron emission tomography and computed tomography (FDGPET/CT) is a useful technique in the management of CRC. FDG-PET/CT could be important in the preoperative staging of potentially resectable CLM. Nowadays, there is an increasing interest in its ability to predict response to biological agents. FDG-PET/CT could detect early metabolic changes in tumor cell metabolism, before any change in tumor size occurs. Lubezky et al. [21] reported the first experiences with FDG-PET/CT in the restaging of patients with CLM after neoadjuvant chemotherapy with or without bevacizumab, and in patients selected for upfront surgery. FDG-PET/CT demonstrated a higher sensitivity than contrast-enhanced CT for the detection of CLM in patients who did not receive chemotherapy rather than patients who received chemotherapy (FDG-PET/CT: 93.3 vs. 49%, p < 0.0001; CT: 87.5 vs. 65.3, p = 0.038). On the contrary, contrast-enhanced CT revealed a higher sensitivity than FDG-PET/CT for detecting
2.1.2. CT-SCAN behind evaluation of efficacy: Radiology predictor factor? A novel approach involves the enhancement gradient, defined as the difference in Hounsfield units between CTSCAN with and without iodinated contrast material. This gradient represents the density of vascularization within the neoplasia, as demonstrated in several studies [19,27–31]. A recent study showed that the enhancement gradient in a baseline CT-SCAN correlates with radiological responses to anti-angiogenic therapy [32]. This trial had several limitations, because of the retrospective conduction of the study and the unspecified response evaluation criteria; however, it represents a promising model for the identification of predictors of biological activity of bevacizumab. This approach could potentially change the role of CT-SCAN, which would become not only a tool to evaluate chemotherapy activity, but also a means to stratify patients who will likely benefit from an anti-angiogenic treatment.
Please cite this article in press as: Pietrantonio F, et al. Bevacizumab-based neoadjuvant chemotherapy for colorectal cancer liver metastases: Pitfalls and helpful tricks in a review for clinicians. Crit Rev Oncol/Hematol (2015), http://dx.doi.org/10.1016/j.critrevonc.2015.04.008
ARTICLE IN PRESS
ONCH-1971; No. of Pages 10
F. Pietrantonio et al. / Critical Reviews in Oncology/Hematology xxx (2015) xxx–xxx
2.2. New imaging techniques
3. Post-surgical phase
A novel technique to improve response evaluation to antiangiogenic agents represents contrast-enhanced ultrasound (CEUS) with the opportunity to detect changes of tumor perfusion at early time-points after administration of chemotherapy. In a little prospective study with 30 patients with unresectable CLM underwent to FOLFIRI+ Bevacizumab, Schirin-Sokhan et al. had demonstrate in liver metastases that baseline time to reach the maximum peak of contrast intensity (time to peak: TTP) and TTP quotient are significantly lower in the group of the responders compared to the non-responders. Furthermore, correlating to the antiangiogenic effect of bevacizumab we observed a strong increase in TTP and TTP quotient during chemotherapy, which was restricted to the group of the responders [33]. Despite these encouraging results, this study has some limitations for application in CLM setting both for the number of patients and for the absence of a comparison with the pathologic response being concerned patients with unresectable metastases. An interesting emerging role for this technique is the intraoperative use. Preoperative chemotherapy sometimes makes CLM disappear. Contrast-enhanced intraoperative ultrasound (CE-IOUS) using perflubutane may identify such metastases. A recent study had demonstrated that CE-IOUS detects 79% of the disappearing liver metastases (DLM) and sensitivity of CE-IOUS was superior to CT-SCAN (p < 0.04) [34]. The authors suggest CE-IOUS might be necessary after preoperative chemotherapy for disappeared CLM. Recently, a computerized tomography technology named dual-energy CT (DECT) entered the medical practice [35]. Thanks to two sources of radiation emission, DECT is able to reduce the patient global exposure to radiation, to obtain higher resolution images and to better assess the hemorrhagic and vascular changes in the tumor [35,36]. A small study in non-small-cell lung cancer analyzed the accuracy of DECT for response evaluation in anti-angiogenic therapy, showing an improvement in radiological assessment of tumor changes [36]. Therefore, this technique could confer more objectiveness to the morphologic response criteria to anti-angiogenic therapies, representing an important tool for clinical practice. However, these results need to be confirmed in larger prospective trials including other cancer types including CRC. Another promising technology is dynamic contrastenhanced magnetic resonance imaging (DCE-MRI). This technique assesses contrast inflow and egress from a region of interest in the neoplasia, and it is widely used for response evaluation in anti-angiogenetic therapies. In fact, the rapid acquisition of images before and after contrast infusion is able to reveal changes in tumor vascularization [37]. Furthermore, DCE-MRI is able to monitor structure changes and vascular permeability variations, thus representing an interesting biomarker to estimate bevacizumab effects [35]. These encouraging results need further studies to be applied in randomized clinical trials in order to become a predictive factor of histological response.
3.1. Pathological evaluation response
5
Due to possible misinterpretation and lack of reproducibility of radiological assessment, preoperative imaging may have a suboptimal accuracy in the assessment of chemotherapy-induced tumor changes, especially in the neoadjuvant setting [38]. On the contrary, histological evaluation should have high accuracy and reproducibility in the evaluation of residual microscopic tumor. The essential role of pathological response as a prognostic factor has been demonstrated in patients receiving preoperative therapy for breast, esophageal, gastric and rectal cancer [39–42]. Recently, histological analyses have been shown to be associated with survival in patients with CLM, with pathologic complete response being an independent prognostic factor of long-term outcome [20,43]. To date, there is a growing body of evidence suggesting that bevacizumab may improve pathologic response of CLM when added to preoperative chemotherapy (Table 4). Despite of the objective evaluation of the residual tumor after neoadjuvant therapy, the pathologic response can be assessed with different classification criteria, such as residual viable tumor, tumor regression grade or its modified version, and tumor thickness. Especially after neoadjuvant treatment with antiangiogenetic therapy, there is not unanimous consensus for the better classification to use. Indeed, an optimal quantification of pathological response may help to orient clinical choices, define patients prognosis and establish the subsequent therapeutic plan. 3.1.1. Residual viable tumor An intriguing method for pathological response assessment in CLM was described by Blazer et al. [20]. It consists of a semi-quantitative assessment of residual cancer cells referring to the total tumor area. Pathological response was categorized in 3 groups: complete (no residual cancer cells remaining), major (1–49%) or minor response (≥50%). In patients with multiple tumor nodules, the mean of the values for the various nodules was used to define the pathological response. In this study, treatment with oxaliplatin-based chemotherapy plus bevacizumab produced a higher rate of pathological complete or major response as compared to chemotherapy alone (63% versus 44%, p < 0.001). In multivariate analysis including age, lymph nodal status, disease-free interval, CEA level, tumor size, number of lesions and treatment type, fluoropyrimidine plus oxaliplatin and bevacizumab therapy was confirmed as an independent predictor of pathological response, which significantly correlated with OS (major response: HR 4.80, p = 0.34; minor response: HR 6.93, p = 0.007). Independently of treatment received, cumulative 5-year OS rates were 75% for a complete pathologic response, 56% for a major pathologic response and
Please cite this article in press as: Pietrantonio F, et al. Bevacizumab-based neoadjuvant chemotherapy for colorectal cancer liver metastases: Pitfalls and helpful tricks in a review for clinicians. Crit Rev Oncol/Hematol (2015), http://dx.doi.org/10.1016/j.critrevonc.2015.04.008
ONCH-1971; No. of Pages 10
6
ARTICLE IN PRESS F. Pietrantonio et al. / Critical Reviews in Oncology/Hematology xxx (2015) xxx–xxx
Table 4 Neoadjuvant anti-angiogenic therapy: recent studies correlating treatment with pathologic response. Study (year)
Treatment (n)
Pathological response, n (%) Complete
Ribero et al. (2007) [52]
5FU + oxaliplatin (43) 5FU + oxaliplatin + bevacizumab (62)
5 (12) 7 (11)
Blazer et al. (2008) [20]
5FU + oxaliplatin (81) 5FU + oxaliplatin + bevacizumab (50)
6 (12) 7 (9)
Kishi et al. (2010) [44]
FOLFOX (117) FOLFOX + bevacizumab (102)
Klinger et al. (2010) [45]
FOLFOX (50) FOLFOX + bevacizumb (50)
Bibeau et al. (2013) [49]
No treatment (29) Chemotherapy (31) Chemotherapy + bevacizumab (31)
Loupakis et al. (2013) [46]
FOLFOXIRI/XELOXIRI (18) FOLFOXIRI + bevacizumab (24)
33% for a minor pathologic response (complete vs. major response, p = 0.037; major vs. minor response, p = 0.28) [20]. Similarly, a further non-randomized observational study was conducted on 219 patients with CLM treated with FOLFOX with or without bevacizumab [44]. The frequency of the above mentioned complete or major pathologic response rate was higher after treatment with FOLFOX plus Bevacizumab versus chemotherapy alone (70% versus 45%, p < 0.001). These data suggest that Bazer et al. classification allows to highlight the difference in pathological response effectiveness between chemotherapy with or without bevacizumab. 3.1.2. Tumor regression grade Rubbia-Brandt et al. were the first to define histological criteria of response to preoperative chemotherapy in CLM based on the tumor regression grade (TRG) [43]. These criteria had previously been elaborated by Mandard to evaluate the effect of neoadjuvant treatment on esophageal carcinoma [40]. This method identifies five TRGs classes on the bases of residual tumor cells and the extent of fibrosis: TRG1, absence of residual cancer cells replaced with a large amount of fibrosis; TRG2, rare scattered residual cancer cells and abundant fibrosis; TRG3, more residual cancer cells and predominant fibrosis; TRG4, large amount of cancer cells predominating over fibrosis; and TRG5, most exclusively cancer cells without fibrosis. Due to intra-individual variability, patients with multiple liver metastases who showed different TRG were categorized according to the worst TRG. The TRG was then grouped in 3 classes: major histologic response
8 (7) 10 (10)
Major
Complete + Major
ILN
10 (23) 28 (45) p = 0.02 16 (32) 44 (54) p < 0.001
22 (44) 51 (63)
45 (38) 61 (60)
53 (45) 71 (70)
5 (10) 19 (38) p < 0.001 0 (0) 12 (39) 21 (68) p < 0.001 2 (11) 4 (16)
– –
5 (28) 15 (63) p = 0.033
4 (27) 19 (83) p = 0.0009
(MjHR; TRG1 and TRG2), partial histologic response (PHR; TRG3) and no histologic response (NHR; TRG4 and 5). In Rubbia-Brandt et al. study, MjHR was associated with better 3-year DFS (49% vs. 18%, log-rank p = 0.0014) and 5-year OS (41% vs. 9%, log-rank p = 0.0003) than NHR. Moreover, in a multivariate analysis including patients age, number and size of metastases, synchronous versus metachronous metastases and chemotherapy, TRG was an independent prognostic factor for 5-year DFS (hazard ratio, HR 0.71, p = 0.001) and OS (HR 0.55, p = 0.004) [43]. In addition, a retrospective analysis of two nonrandomized trials with preoperative fluoropyrimidines and oxaliplatin with or without bevacizumab showed that patients treated with bevacizumab achieved a significantly higher major histological response compared to those treated with chemotherapy only (38% versus 10%; p < 0.001), using the TRG criteria [45]. Finally, a retrospective study analyzed 18 patients treated with FOLFOXIRI or XELOXIRI and 24 patients treated with FOLFOXIRI plus bevacizumab who underwent secondary resection of CLM. In this study, major histological response according to TRG was 63% and 28% for patients receiving and not receiving the anti-VEGF antibody, respectively (p = 0.033). In the bevacizumab group, patients with a TRG 1, 2 or 3 in comparison to those with TRG 4 or 5 had an advantage in terms of PFS (HR 0.41, p = 0.031). OS data were not mature at the time of analysis [46]. Taken together these results, TRG classification allows to emphasize difference effectiveness in pathological response between chemotherapy with and without bevacizumab.
Please cite this article in press as: Pietrantonio F, et al. Bevacizumab-based neoadjuvant chemotherapy for colorectal cancer liver metastases: Pitfalls and helpful tricks in a review for clinicians. Crit Rev Oncol/Hematol (2015), http://dx.doi.org/10.1016/j.critrevonc.2015.04.008
ONCH-1971; No. of Pages 10
ARTICLE IN PRESS F. Pietrantonio et al. / Critical Reviews in Oncology/Hematology xxx (2015) xxx–xxx
3.1.3. Modified tumor regression grade According to the TRG, only fibrosis but not necrosis is considered as a marker of tumor response. The usual necrosis seen in CLM is the so-called “dirty” necrosis and is characterized by nuclear debris in a patchy distribution bordered by viable cells. It seems to be caused by spontaneous phenomena within the tumor, such as insufficient vascular supply, rather than by the treatment itself. In fact, a large amount of necrosis is seen in patients who respond poorly or in those not receiving preoperative treatment [47]. However, some other authors have interpreted the necrosis as a form of response to therapy. Chang et al. [48] identified a particular type of necrosis, the “infart-like” necrosis (ILN), which differs from the “dirty necrosis” because of the large confluent areas of eosinophilic cytoplasmic remnants, located centrally within a lesion and surrounded by a rim of fibrosis with foamy macrophages. The Chang et al. retrospective trial included 109 patients, of whom 46 patients received perioperative treatment. ILN was only observed in preoperatively treated patients and not in untreated patients, and it significantly correlated with improved DFS (log-rank p = 0.047) [48]. The association between bevacizumab and ILN was also confirmed in larger series retrospectively evaluated [49]. Based on this observation, ILN was incorporated by Chang et al. in a modified TRG (mTRG) which considers the presence of ILN a form of therapeutic effect equivalent to fibrosis. The inclusion of ILN into the definition of treatment effect altered the tumor regression score in about a quarter of patients. Modified TRG (mTRG) 1–2 scores were associated with significantly better DFS (log-rank p = 0.007) and OS (log-rank p = 0.020) compared with mTRG 3–5 scores, whereas statistical significance was not reached using the classic TRG score to detect survival differences. Even if this may partially be explained by the relatively small sample size, incorporating ILN in the modified grading system might enhance the prognostic utility of TRG after bevacizumab treatment [48]. 3.1.4. Tumor thickness Maru et al. introduced the tumor thickness at the tumornormal interface (TNI) as an indicator of pathologic response to preoperative chemotherapy. The maximum thickness of uninterrupted layers of tumor cells was measured perpendicular to the TNI by a ruler. Greater thickness predicted shorter recurrence-free survival (RFS), and its correlation with RFS remained in multivariate analysis (p = 0.015) [50]. Maru et al. found that TNI was also correlated with mRECIST and with the response criteria from Blazer et al. Additional observation of that study was that tumors from patients treated with bevacizumab had smaller tumor thickness at TNI than did tumors from patients who did not receive bevacizumab, and they found significantly thinner median TNI in patients who received oxaliplatin-based therapy than in patients who received irinotecan-based therapy. Nowadays there is only a study witch compare grade of residual viable tumor (Blazer et al.), tumor regression grade
7
(TRG) scoring system (Rubbia-Brandt et al.), modified tumor regression grade (mTRG) scoring system with the type of necrosis (Chang et al.), and the tumor thickness at the TNI (Maru et al.) in patients with CLM and underwent to neoadjuvant bevacizumab-based chemotherapy. Dede et al. analyzed if these histopathologic changes are specific to preoperative chemotherapy and if these methods have correlation with survival. In a retrospective study with 46 patients underwent to liver surgery after neodjuvant chemotherapy with bevacizumab, none of the analyzed pathological methods showed significant correlation with progression free survival (PFS) or with overall survival (OS). Residual tumor cell ratio and TRG showed positive but not significant correlation with OS and PFS. The PFS and OS curves showed a slight difference in the group of patients with TNI
