HPB
http://dx.doi.org/10.1016/j.hpb.2015.10.009
ORIGINAL ARTICLE
Colorectal liver metastases: disappearing lesions in the era of Eovist hepatobiliary magnetic resonance imaging Joseph W. Owen1, Kathryn J. Fowler2, Maria B. Doyle3, Nael E. Saad2, David C. Linehan4 & William C. Chapman3 1
University of Kentucky, 2Washington University, Department of Radiology, 3Washington University, and 4University of Rochester, Department of Surgery, United States
Abstract Background: Hepatobiliary contrast enhanced MRI is known to be the most sensitive imaging modality for detection of colorectal hepatic metastasis. To date no study has investigated the rate of disappearing lesions with gadoxetic acid MR (Eovist/Primovist), or characterized the pathologic response of lesions which disappear on gadoxetic acid MR. Methods: Retrospective review of hepatic resections for colorectal metastases between 01/2008 and 01/2014 was performed to evaluated the rate of disappearance of lesions on gadoxetic acid MR and the rate of complete pathologic response in the lesions that disappear. “Disappearing lesions” were lesions on baseline imaging that were not identifiable on pre-operative Eovist MRI. Complete pathologic response was defined as no viable tumor on pathology or by lack of recurrence within 1 year. Results: In 23 patients, 200 colorectal metastases were identified on baseline imaging. On preoperative Eovist MR 77 of the 200 lesions (38.5%) were “disappearing” lesions. At surgical pathology or 1 year follow-up imaging, 42 of 77 lesions (55%) demonstrated viable tumor (21) or recurrence (21). Thirty of 77 lesions (39%) were nonviable at pathology (10) or without evidence of recurrence at 1 year (20). 5 lesions were indeterminate. Discussion: Despite disappearance on Eovist MR imaging (the most sensitive available imaging modality), 38.5% of all colorectal metastases disappeared and of those, 55% were viable. Received 12 August 2015; accepted 31 August 2015
Correspondence Joseph W. Owen, 800 Rose St, Room HX-311, Lexington, KY 40536, United States. Tel.: +1 859 323 5069. Fax: +1 859 257 4457. E-mail: [email protected]
Introduction Colorectal cancer results in an estimated 600,000 deaths per year worldwide, with 1.2 million new cases annually, making it the 3rd most common cancer in men and the 2nd most common cancer in women.1 Hepatic metastatic disease has been found in 56% of patients with colorectal cancer at autopsy, with isolated hepatic metastases in 35% of patients.2 Metastatic colorectal cancer has a poor prognosis with overall 5 year survival of 12–17%.3,4 Patients who undergo intent to cure resection for hepatic metastatic disease have demonstrated improved 1 year survival of 93% and 5 year survival of 47%.5
This study was presented at the Annual Meeting of the AHPBA, 11-15 March 2015, Miami, Florida.
HPB 2016, 18, 296–303
Prior to undergoing hepatic resection, patients with hepatic metastatic disease often receive neoadjuvant chemotherapy, which can result in lesions “disappearing” or becoming radiologically occult (Fig. 1). Despite a complete response on imaging, many studies have shown persistent viable tumor at the site of “disappearing lesions”.6–12 With a lesion based analysis, the incidence of disappearing lesions ranges from 11%12 to 36%7 of initially detected liver metastases. With a patient based analysis, disappearing lesions have been shown to occur in 6%7 –23%13 of patients treated for colorectal metastatic disease. Reported complete pathologic response of disappearing lesions ranges from 15%10 to 73%.6 While the current surgical dogma mandates removal of all tumor if feasible, the role of disappearing lesions in pre-surgical planning remains controversial with no demonstrated survival benefit to resecting all sites of disappearing lesions.14
© 2016 International Hepato-Pancreato-Biliary Association Inc. Published by Elsevier Ltd. All rights reserved.
HPB
297
Figure 1 Coronal hepatobiliary 20 min delay MR image at baseline (a) demonstrating colorectal metastases in the liver (arrows). Coronal hepa-
tobiliary 20 min delay MR image at pre-surgical staging (b) demonstrating no lesion at the site of metastases on the prior exam (arrowheads).
Studies investigating disappearing lesions rely on a variety of imaging modalities, including PET/CT, US, contrast enhanced CT, MRI, and intra-operative US. Bischof et al. 2013 report a pooled sensitivity for hepatic metastases after chemotherapy of 85.7% for MRI and 69.9% for CT, and suggest MRI is the preferred modality for pre-operative evaluation of hepatic metastases. Over the past 7 years there has been an increasing reliance on a new MRI contrast agent gadoxetic acid (Eovist or Primovist; Bayer Healthcare Pharmaceuticals) for imaging colorectal hepatic metastases. Gadoxetic acid has increased sensitivity over previously available MR techniques,15 and shows improved sensitivity for small colorectal metastases over CT,16 especially in the setting of background hepatic steatosis. To date no study has investigated the rate of disappearing lesions with gadoxetic acid MR, or characterized the pathologic response of lesions which disappear on gadoxetic acid MR. Gadoxetic acid is characterized as a hepatobiliary contrast agent due to its high rate of uptake and excretion by functioning hepatocytes. On 20 min delayed imaging, (often referred to as hepatobiliary phase) gadoxetic acid uptake peaks in normal liver, with liver lesions becoming dark areas of non-enhancement (Fig. 2). The hepatobiliary phase allows for optimal conspicuity of lesions compared to background liver and is the reason for the improved sensitivity for lesion detection with gadoxetic acid compared to other MRI contrast agents. Ideally, patients with colorectal hepatic metastatic disease receive high quality quadruple phase CT or gadoxetic acid MR prior to undergoing neoadjuvant chemotherapy to serve as a roadmap for hepatic resection. In reality many of patients are
diagnosed and staged at outside institutions, and often receive chemotherapy prior to establishing care at tertiary care centers with experienced hepatobiliary surgeons. With outside imaging as a reference, Eovist MR may be performed prior to resection to assess treatment response and determine the extent of disease. The purpose of our study was to determine the incidence of disappearing colorectal metastases with state-of-the-art hepatobiliary agent MR imaging and assess the rate of complete pathologic response for disappearing lesions on hepatobiliary phase MR imaging. We hypothesize gadoxetic acid is more sensitive for viable colorectal metastases, and disappearance on gadoxetic acid imaging is more likely an indication of pathologic complete response compared to reported rates of viability on other less sensitive imaging modalities.
Methods Patients After IRB approval and waiver of consent, a database of surgery clinic patients was queried using the following ICD 9 codes – 197.7 (Secondary neoplasm of the liver) and 153.* (neoplasm of colon) or 154.0 (neoplasm rectosigmoid junction) or 154.1 (neoplasm rectum). The dates of inclusion were January 1st, 2008 to January 1st, 2014. Patients were then cross referenced to a database containing all patients imaged with gadoxetic acid at our institution. The individual medical records were then reviewed to determine eligibility based on the following inclusion criteria: diagnosis of colorectal cancer, evidence of metastatic colorectal cancer, treatment of colorectal metastatic disease with liver resection, pre-operative re-staging with
Figure 2 Arterial phase: hepatic artery and early portal vein enhancement (A). Portal Phase: peak portal vein enhancement (B). Transitional
phase: early parenchymal uptake (C). Transitional 5 min phase: gradual blood pool clearance (D). Hepatobiliary 20 min phase: blood pool clearance with peak normal liver parenchymal enhancement. Note the colorectal metastasis (arrow) and the arterial shunting (arrowhead).
HPB 2016, 18, 296–303
© 2016 International Hepato-Pancreato-Biliary Association Inc. Published by Elsevier Ltd. All rights reserved.
HPB
298
gadoxetic acid enhanced magnetic resonance imaging, and available baseline imaging. Given the retrospective nature of the investigation, the neoadjuvant chemotherapy regimens used to treat patients were heterogeneous and often administered at outside institutions prior to establishment of care with our hepatobiliary surgeons. Imaging evaluation Baseline imaging was defined as the time point when the sum of the long axis dimension of all the hepatic colorectal metastases was greatest. If the patient had progressive disease, baseline imaging was defined as the first imaging study demonstrating liver metastatic disease. Baseline imaging was either MRI or contrast enhanced CT. Baseline imaging was reviewed with follow-up imaging, operative reports, and surgical pathology available to maximize sensitivity for lesions. The maximum transaxial dimension of all visible lesions was recorded. The liver was assessed for hepatic steatosis by imaging criteria on baseline imaging and at pre-surgical staging (liver spleen index on CT, or loss of signal on dual echo chemical shift MRI). Pre-operative restaging was performed with weight-based double dose (0.05 mmol/kg) gadoxetic acid (Bayer Healthcare Pharmaceuticals) intravenous contrast on 1.5 T magnetic resonance scanners (Siemens Healthcare). MR examinations included standard sequences plus diffusion weighted imaging (b values 50, 400, 800) and 20 min hepatobiliary phase contrast enhanced imaging. Pre-operative restaging MR examinations were reviewed with all prior and subsequent imaging studies, operative notes, and surgical pathology available for reference to maximize sensitivity for lesions. The maximum transaxial dimension of all visible lesions was recorded. Follow-up imaging consisted of standard of care MR, CT and PET/CT. Sites of disappearing lesions which were not resected were followed for at least 1 year to assess for recurrence. All follow-up imaging was reviewed to assess for disease recurrence to determine disease free survival. Surgical pathology Pathology reports available in the medical record were reviewed and correlated with the operative reports, intraoperative ultrasound and prior imaging for correlation of lesion location and viability. Standard protocol for partial liver resections involves the following: The cauterized surface is inked to define the resection margin. The specimen is sectioned perpendicular to the margin every 3–5 mm. All lesions are described with estimated viability and location relative to the inked margin. Reference standard for lesion characterization Lesions were defined as “disappearing” if they were identifiable on the baseline imaging study but not identifiable on gadoxetic acid enhanced MRI during pre-surgical restaging. For lesions resected at surgery, surgical pathology reports were correlated with the imaging location of disappearing lesions. Lesions were
HPB 2016, 18, 296–303
characterized as “complete response” if they were not detected at surgical pathology or were detected but no residual viable tumor was present. Lesions were characterized as “persistent disease” if they were detected on surgical pathology and demonstrated viable tumor. For disappearing lesions which were not resected, follow-up imaging was reviewed for 1 year with close attention to the location of the unresected disappearing lesions. A lesion was characterized as “complete response” if it did not recur within 1 year. A lesion was characterized as “persistent disease” if a lesion recurred within 1 year. Statistics Data distributions were tested for normality with the Shapiro– Wilk W test. Equality of variances was tested with the O’Brien, Brown-Forsythe, Levene, Bartlett, and 2-sided F Test. If data were normally distributed and variances were equal the t test was used; otherwise, the Wilcoxon rank sums test was used. For normally distributed data, means, standard deviations, and 95% confidence intervals were used for descriptive statistics. For normally distributed data, medians plus 25th and 75th quantiles (quartiles) were used for descriptive statistics. Regression analysis was used to determine the best model to describe the relationship between two continuous variables. Logistic regression (logit regression) was used to predict a categorical dependent variable from an independent continuous variable, and probability plots were created for each analysis. Statistical analyses were performed with JMP Pro Statistical Software Release 11.0.0 (SAS Institute, Inc., Cary, NC), and MedCalc Statistics for Biomedical Research Version 15.2.1.0 (MedCalc Software, Mariakerke, Belgium).
Results Patient population Search of the surgical patient database found 603 patients evaluated by surgery who had a secondary neoplasm of the liver and either colon cancer, colorectal cancer, or rectal cancer. Of the total patients, 23 had hepatic metastases and met inclusion criteria with available imaging for review (Table 1). Average age at presentation was 53 years (range 37–74). There were 16 men and 9 women. Six patients had hepatic steatosis at baseline. Tumor stage was T4 = 3, T3 = 16, T2 = 3, and T1 = 1. Nodal stage was N2 = 5, N1 = 8, and N0 = 10. Thirteen patients had metastatic disease at initial presentation (synchronous metastases) and 10 patients developed metastases after staging (metachronous metastases). Three had undergone prior regional therapy (resection or ablation) for prior metastatic disease and had redeveloped hepatic metastases. Neoadjuvant treatment 23 patients underwent some form of neoadjuvant therapy prior to hepatic resection. The majority of patients were treated with FOLFOX (n = 16) +/− bevacizumab (n = 9) as first line therapy,
© 2016 International Hepato-Pancreato-Biliary Association Inc. Published by Elsevier Ltd. All rights reserved.
HPB
299
Table 1 Patient variables
Table 2 Surgical therapy
Patient Variables
N [ 23
Surgical therapy
N [ 23
Mean age (range)
53 (37–74)
Right hemi-hepatectomy
6
Right hemi-hepatectomy + non-anatomic resection or ablation
2
Gender Male
15
Female
8
Diagnosis of metastasis Synchronous Metachronous
13 10
T-Stage T4/3
19
T2/1
4
N-Stage N2
5
N1
8
N0
10
Bilateral metastatic disease
16
Extrahepatic Disease (pulmonary nodule)
1
Extended right hepatectomy
1
Extended right hepatectomy + non-anatomic resection
1
Left hemi-hepatectomy
1
Left lateral sectionectomy + non-anatomic resection or ablation
5
Non-anatomic resections ± ablation
7
response to therapy had synchronous metastatic disease. Bilateral hepatic disease was present in 16 patients at the time of surgical resection. Extrahepatic disease was present in only 1 patient at the time of surgery (a pulmonary nodule which was subsequently resected). Two patients developed hepatic steatosis from baseline with a total of 8 patients with hepatic steatosis undergoing resection.
Baseline imaging MR
13
CT
9
PET/CT
1
Hepatic steatosis at baseline
5
Hepatic steatosis pre-surgery
7
Radiological response (RECIST) Progression
2
Stable
9
Partial response
12
Complete response
0
Average chemotherapy holiday in weeks
6 (2.5–16)
Patients with disappearing lesions
11
Median number of lesions Lesions/patient
6
Lesions/patient In patients with disappearing lesions
11 (4–52)
Disappearing lesions/patient In patients with disappearing lesions
4 (1–31)
with FOLFIRI + bevacizumab (n = 5) the most common second line therapy. Right portal vein embolization was performed in 4 of the patients to facilitate extended hepatectomy. Two patients received transarterial chemoembolization as neoadjuvant therapy prior to hepatic resection. Response to neoadjuvant therapy was determined using RECIST criteria with progression of disease = 2, stable disease = 7, partial response = 12, and complete response = 0. Patients with partial response to therapy had an average change in lesion size of 60% (range 35–85%) with average baseline lesion size of 12.3 mm (range 4.6–27.9 mm). Nine of the 12 patients showing a partial
HPB 2016, 18, 296–303
Surgical therapy Multiple types of surgical resections and combination resection and ablation were performed (Table 2). All surgeries were performed with intent to cure. Intra-operative ultrasound was used at the discretion of the surgeon and was not universally performed. Chemotherapy was held for an average of 6 weeks prior to surgery (median = 6, range 2.5–16). Disappearing lesions A total of 200 metastatic lesions were identified in 23 patients undergoing hepatic resection (Table 3). Disappearing lesions occurred in 11/23 patients, and 11/12 of patients demonstrating partial response to therapy. Patients with synchronous metastatic disease (Fig. 3) were more likely to have disappearing lesions (odds ratio 11.25, p = 0.015). The median size of all lesions at baseline was 9 mm, (range 3–151 mm). At follow-up imaging 38.5% (77/200) of lesions had disappeared. The median size of disappearing lesions at baseline was 6 mm (range 3–32 mm), and the median size of visible lesions at baseline was 15 mm
Table 3 Lesion Variables (median; range)
Lesion Variables (median; range)
N [ 200
Number of disappearing lesions
77 (38.5%)
Baseline average size in mm All lesions Disappearing lesions
13 (3–151) 8 (3–32)
Disappearing lesions With complete response
32 (39%)
With persistent disease
40 (55%)
Indeterminate
5 (6%)
© 2016 International Hepato-Pancreato-Biliary Association Inc. Published by Elsevier Ltd. All rights reserved.
HPB
300
Figure 3 Mosaic plot – Contingency analysis demonstrates that the
odds for 1 disappearing lesions with in the “Synchronous” group is 11.25 times greater than in the “Metachronous” group (odds ratio – 11.25; p = 0.0064).
(3–151 mm). There was a statistically significant difference in the baseline size of lesions which disappeared compared with those that were still visible at presurgical staging (p < 0.0001). Of the disappearing lesions 39% (30/77) were found to have complete response and 55% (42/77) were found to have persistent disease based on pathology or imaging. Complete response or persistent disease could not be determined in 6% (5/77) of lesions (Fig. 4). In reviewing the MR imaging for all patients, several lesions were identified that were not accounted for in the intent to cure surgical plan. A total of 14 lesions (7% of all lesions) fell into this category and were not resected. In some instances these lesions were not individually identified by radiology report, in other cases they were not found at the time of surgery by intraoperative ultrasound and hence not resected. Thirteen lesions were positive for persistent disease with interval growth on follow-up imaging. These lesions for purposes of analysis were called “missed” lesions. The median size of the missed lesions on pre-surgical staging MR was 5 mm (range: 2–17 mm).
Figure 4 Diagram demonstrating the treatment and viability of disap-
pearing lesions.
HPB 2016, 18, 296–303
Survival and recurrence Average patient follow-up was 1285 days (median = 1205; range 212–2108). Five patients were lost to follow up at an average of 1135 days (median 1135; range 212–1637). One patient died during the follow-up interval at 921 days. The remaining patients were alive at the time of data censorship. Average disease free survival was 339 days (median 330; range 28–861). Patients undergoing intent to cure surgical resection had an average disease free survival of 470 days (median = 483; range 111–861). Patients with only disappearing lesions that were not resected had an average disease free survival of 329 days (median 360, range 109–519). No statistical difference was found, p = 0.49. The presence of missed lesions showed a statistically significant reduction in disease free survival with an average of 194 days (median 67; range 28–585; p = 0.015). Combining patients with missed lesions and disappearing lesions left behind also showed a statistically significant reduction in disease free survival with average of 232 days (median 109; range 28–585) (Fig. 5).
Discussion Despite reported improvements in imaging sensitivity with gadoxetic acid hepatobiliary MR contrast agent, 38.5% of colorectal metastases disappeared, and of the disappearing lesions 55% had surgical pathology or imaging evidence of persistent disease. This disproves the hypothesis that the increased sensitivity of gadoxetic acid for small hepatic metastases would reduce the incidence of disappearing lesions, and disproves the hypothesis that disappearing lesions with gadoxetic acid MR imaging would represent true complete response. (Fig. 6). Our incidence of disappearing lesions is greater than previous estimates in the literature. Ferrero et al., 2012 reported a 10.7% incidence of disappearing lesions (64/624), occurring in 19% of patients receiving preoperative chemotherapy. This difference may arise from the pre-chemotherapy staging by triphasic CT in all patients in Ferrero et al., 2012. In this investigation 13 of 23 patient’s baseline imaging with MR and 9 of those patients received a hepatobiliary MR agent. MR with hepatobiliary agents may be more sensitive to small lesions not detected by other imaging modalities, and small lesions may have an increased incidence of disappearance. Another reason for the increased incidence of disappearing metastases in our series may also be the interval improvements in chemotherapy with FOLFOX/ FOLFIRI and bevacizumab showing promising response with median progression free survival of 10.2 months and overall survival of 23.3 months.17 Literature reports of persistent disease associated with disappearing lesions is varied with persistent disease reported as high as 85% with PET/CT based imaging10 or as low as 27% with multimodality detection including US, CT, MR, intra-operative US, palpation and inspection.6 The sensitivity of PET/CT for small lesions is inferior to triphasic CT and MR and likely accounts for the high rate of persistent disease in Tan et al., 2007.
© 2016 International Hepato-Pancreato-Biliary Association Inc. Published by Elsevier Ltd. All rights reserved.
HPB
301
Figure 5 Kaplan Meier Survival Curve – Disease free survival was shorter for patients with “Lesion Left Behind” (Yellow) than for patients with all
lesions resected (Blue) with p=0.005.
Figure 6 Baseline contrast enhanced CT image showing extensive colorectal metastatic disease (arrows) in the liver (A). Presurgical staging MR
image with Gadoxetate Disodium in the hepatobiliary 20 min phase (B) showing resolution of previously seen lesions (arrowheads), and presurgical staging PET/CT (C) showing no increased FDG-uptake in the region of previously seen lesions (open arrowheads).
HPB 2016, 18, 296–303
© 2016 International Hepato-Pancreato-Biliary Association Inc. Published by Elsevier Ltd. All rights reserved.
HPB
302
The low rate of persistent disease in Elias et al., 2004 may be attributable to multimodality staging and the use of intra-arterial chemotherapy, which shows higher rates of lesional necrosis than standard chemotherapy. In addition to persistent viable tumor in disappeared lesions, there were a number of lesions, visible on MR, but not detected by intraoperative ultrasound which demonstrated persistent disease on follow-up imaging (previously described as ‘missed lesions’). These lesions were often not specifically identified in the radiology reports that guided surgical decision making, either due to a large number of lesions or small size. In some instances, the lesions were small and likely below the threshold for prospective detection, in other instances they were felt to be indeterminate by imaging. These lesions were visible on MR in retrospect, and were not found with intra-operative US, palpation and/or inspection, and were not resected. 14 lesions (2% of all lesions) were categorized as “missed lesions” and 13 of these lesions (93%) recurred. It may make sense for “disappearing lesions left behind” and “missed lesions left behind” to be grouped together since for practical purposes “missed lesions” were not able to be identified prospectively. In analyzing these groups together, the difference in disease free survival between patients with all lesions resected (460 days), and patients with lesions left behind (231 days) reaches statistical significance (p = 0.019). This brings up an important question about the true sensitivity of pre-operative imaging for detecting and facilitating the removal of all disease. This is especially difficult in the era of effective neoadjuvant chemotherapy and likely compounded by the difficulty of intraoperative ultrasound in finding small lesions. Reliance on intraoperative imaging to detect all lesions may be a pitfall resulting in a higher rate of residual disease. Our institution has placed of fiducial markers prior to neoadjuvant chemotherapy at the site of small metastases outside the planned anatomic resection in a small number of cases. In patients with synchronous bilateral metastatic disease who have lesions smaller than 10 mm outside of the planned resection margin a fiducial marker could be placed percutaneously to allow for nonanatomic resection or ablation of the site of disease at the time of surgery. It would not be practical to place fiducial markers at all sites of disease, but only at the location of small lesions outside the planned anatomic resection. Not taking into account “missed lesions” there was no statistical difference (p = 0.21) between the disease free survival of persons with disappearing lesions left behind (329 days), and those with complete resection (461 days), although that could reflect Type II error due to lack of power. Survival from hepatic resection and overall survival from initial diagnosis in this study was not meaningful as only 1 person is confirmed as deceased at the time of manuscript submission, (average follow-up was nearly 3.5 years, median 3.3 years; range 212–2108 days). This study is limited by its retrospective design, in which data was collected and determinations made from the best available
HPB 2016, 18, 296–303
data in the medical record. As with retrospective reviews the data was heterogeneous and patient management was varied. Chemotherapy regimens and the timing of therapy differed between patients. Baseline imaging was varied with the majority of patients undergoing baseline and pre-surgical MRI, but often patients had suboptimal baseline imaging at outside institutions and underwent chemotherapy prior to establishing care at our institution. The results reflect a single center experience and a small patient series, with a small subset of those patients treated with hepatic resection. Early cases which used gadoxetic acid have selection biased as they were often complex cases. Gadoxetic acid was not used routinely for colorectal metastases imaging until 2012. A future prospective study in which all patients are uniformly imaged with gadoxetic acid MR before chemotherapy and prior to surgery could provide additional insight into the role of hepatobiliary MR contrast agents and disappearing lesions on surgical management of hepatic metastases. The high sensitivity of gadoxetic acid MR for any parenchymal liver abnormality also creates the potential for unnecessary resection to treat false positive lesions, as signal characteristics of lesions less than 5 mm may not allow for differentiation between benign and malignant lesions. Our study did not specifically address false positive lesions. In our experience, characterization of small lesions especially distinguishing hemangiomas from small mucinous mets can be challenging. Our institutional policy on imaging with Eovist is to use it in patients with prior contrast enhanced CT or MRI. Additionally, we may rely on a change in lesion size on preoperative imaging compared to baseline, either increase in size with disease progression or decrease in size with partial response, to help with differentiating small lesions, but this does not help in all instances. Further research is needed to investigate to rate of false positive lesions and develop criteria to characterize lesions less than 5 mm. This is the first study to confirm the phenomenon of disappearing lesions with gadoxetic acid MR imaging, with 1/3 of colorectal metastases disappearing, and approximately 90% of patients responding to therapy developing at least 1 disappearing lesion. Despite the improved sensitivity, the disappearance of lesions on gadoxetic acid MRI does not imply a complete pathologic response given that approximately 1/2 of these lesions had persistent disease. The implications of disappearing lesions on survival are not well evaluated in our small patient population. However, patients with lesions left behind at the time of surgery, either ‘missed lesions’ or ‘disappeared lesions’ did have a worse prognosis. Important questions remain regarding the approach to colorectal liver metastases in the era of highly active neoadjuvant therapy. Further studies should evaluate the role of neoadjuvant treatment in patients with small lesions given the high rate of disappearance which may impact the future ability to resect. Additionally, greater focus in clinical practice on meticulous liver staging with best radiologic approach and a
© 2016 International Hepato-Pancreato-Biliary Association Inc. Published by Elsevier Ltd. All rights reserved.
HPB
303
multidisciplinary review of imaging is likely essential to achieving optimal results.
disappearing liver metastases from colorectal cancer after chemotherapy. Cancer 116:1502–1509. 9. Elias D, Goere D, Boige V, Kohneh-Sharhi N, Malka D, Tomasic G et al.
Summary A large percentage of metastases will become radiologically occult following current neoadjuvant chemotherapy. Disappearance of metastases on hepatobiliary gadoxetic acid MR imaging does not equal a complete pathologic response.
(2007 Oct 15) Outcome of posthepatectomy-missing colorectal liver metastases after complete response to chemotherapy: impact of adjuvant intra-arterial hepatic oxaliplatin. Ann Surg Oncol 14: 3188–3194. 10. Tan MCB, Linehan DC, Hawkins WG, Siegel BA, Strasberg SM. (2007 Aug 22) Chemotherapy-induced normalization of FDG uptake by colorectal liver metastases does not usually indicate complete pathologic
Funding sources None.
response. J Gastrointest Surg 11:1112–1119. 11. Tanaka K, Takakura H, Takeda K, Matsuo K, Nagano Y, Endo I. (2009 Dec) Importance of complete pathologic response to prehepatectomy
Conflict of interest None declared
chemotherapy in treating colorectal cancer metastases. Ann Surg 250: 935–942. 12. Ferrero A, Langella S, Russolillo N, Vigano’ L, Lo Tesoriere R,
References
Capussotti L. (2012 Apr) Intraoperative detection of disappearing
1. Jemal A, Bray F, Center MM, Ferlay J, Ward E, Forman D. (2011 Mar) Global cancer statistics. CA Cancer J Clin 61:69–90.
colorectal liver metastases as a predictor of residual disease. J Gastrointest Surg 16:806–814.
2. Weiss L, Grundmann E, Torhorst J, Hartveit F, Moberg I, Eder M et al.
13. Van Vledder MG, de Jong MC, Pawlik TM, Schulick RD, Diaz LA,
(1986 Nov) Haematogenous metastatic patterns in colonic carcinoma:
Choti MA. (2010 Nov) Disappearing colorectal liver metastases after
an analysis of 1541 necropsies. J Pathol 150:195–203.
chemotherapy: should we be concerned? J Gastrointest Surg 14:
3. Saltz LB, Cox JV, Blanke C, Rosen LS, Fehrenbacher L, Moore MJ et al. (2000 Sep 28) Irinotecan plus fluorouracil and leucovorin for metastatic colorectal cancer. N Engl J Med 343:905–914. 4. De Gramont A, Figer A, Seymour M, Homerin M, Hmissi A, Cassidy J
1691–1700. 14. Bischof DA, Clary BM, Maithel SK, Pawlik TM. (2013 Oct) Surgical management of disappearing colorectal liver metastases: disappearing colorectal liver metastasis. Br J Surg 100:1414–1420.
et al. (2000 Aug) Leucovorin and fluorouracil with or without oxaliplatin
15. Löwenthal D, Zeile M, Lim WY, Wybranski C, Fischbach F, Wieners G
as first-line treatment in advanced colorectal cancer. J Clin Oncol Off J
et al. (2011 Apr) Detection and characterisation of focal liver lesions in
Am Soc Clin Oncol 18:2938–2947.
colorectal carcinoma patients: comparison of diffusion-weighted and
5. Wei AC, Greig PD, Grant D, Taylor B, Langer B, Gallinger S. (2006 May) Survival after hepatic resection for colorectal metastases: a 10-year experience. Ann Surg Oncol 13:668–676.
Gd-EOB-DTPA enhanced MR imaging. Eur Radiol 21:832–840. 16. Scharitzer M, Ba-Ssalamah A, Ringl H, Kölblinger C, Grünberger T, Weber M et al. (2013 Aug) Preoperative evaluation of colorectal liver
6. Elias D, Youssef O, Sideris L, Dromain C, Baton O, Boige V et al. (2004
metastases: comparison between gadoxetic acid-enhanced 3.0-T MRI
Apr 1) Evolution of missing colorectal liver metastases following
and contrast-enhanced MDCT with histopathological correlation. Eur
inductive chemotherapy and hepatectomy. J Surg Oncol 86:4–9.
Radiol 23:2187–2196.
7. Benoist S. (2006 Aug 20) Complete response of colorectal liver me-
17. Bendell JC, Bekaii-Saab TS, Cohn AL, Hurwitz HI, Kozloff M, Tezcan H
tastases after chemotherapy: does it mean cure? J Clin Oncol 24:
et al. (2012 Dec 1) Treatment patterns and clinical outcomes in patients
3939–3945.
with metastatic colorectal cancer initially treated with FOLFOX-
8. Auer RC, White RR, Kemeny NE, Schwartz LH, Shia J, Blumgart LH et al. (2010 Mar 15) Predictors of a true complete response among
HPB 2016, 18, 296–303
bevacizumab or FOLFIRI-bevacizumab: results from ARIES, a bevacizumab observational cohort study. Oncologist 17:1486–1495.
© 2016 International Hepato-Pancreato-Biliary Association Inc. Published by Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.hpb.2015.10.009
ORIGINAL ARTICLE
Colorectal liver metastases: disappearing lesions in the era of Eovist hepatobiliary magnetic resonance imaging Joseph W. Owen1, Kathryn J. Fowler2, Maria B. Doyle3, Nael E. Saad2, David C. Linehan4 & William C. Chapman3 1
University of Kentucky, 2Washington University, Department of Radiology, 3Washington University, and 4University of Rochester, Department of Surgery, United States
Abstract Background: Hepatobiliary contrast enhanced MRI is known to be the most sensitive imaging modality for detection of colorectal hepatic metastasis. To date no study has investigated the rate of disappearing lesions with gadoxetic acid MR (Eovist/Primovist), or characterized the pathologic response of lesions which disappear on gadoxetic acid MR. Methods: Retrospective review of hepatic resections for colorectal metastases between 01/2008 and 01/2014 was performed to evaluated the rate of disappearance of lesions on gadoxetic acid MR and the rate of complete pathologic response in the lesions that disappear. “Disappearing lesions” were lesions on baseline imaging that were not identifiable on pre-operative Eovist MRI. Complete pathologic response was defined as no viable tumor on pathology or by lack of recurrence within 1 year. Results: In 23 patients, 200 colorectal metastases were identified on baseline imaging. On preoperative Eovist MR 77 of the 200 lesions (38.5%) were “disappearing” lesions. At surgical pathology or 1 year follow-up imaging, 42 of 77 lesions (55%) demonstrated viable tumor (21) or recurrence (21). Thirty of 77 lesions (39%) were nonviable at pathology (10) or without evidence of recurrence at 1 year (20). 5 lesions were indeterminate. Discussion: Despite disappearance on Eovist MR imaging (the most sensitive available imaging modality), 38.5% of all colorectal metastases disappeared and of those, 55% were viable. Received 12 August 2015; accepted 31 August 2015
Correspondence Joseph W. Owen, 800 Rose St, Room HX-311, Lexington, KY 40536, United States. Tel.: +1 859 323 5069. Fax: +1 859 257 4457. E-mail: [email protected]
Introduction Colorectal cancer results in an estimated 600,000 deaths per year worldwide, with 1.2 million new cases annually, making it the 3rd most common cancer in men and the 2nd most common cancer in women.1 Hepatic metastatic disease has been found in 56% of patients with colorectal cancer at autopsy, with isolated hepatic metastases in 35% of patients.2 Metastatic colorectal cancer has a poor prognosis with overall 5 year survival of 12–17%.3,4 Patients who undergo intent to cure resection for hepatic metastatic disease have demonstrated improved 1 year survival of 93% and 5 year survival of 47%.5
This study was presented at the Annual Meeting of the AHPBA, 11-15 March 2015, Miami, Florida.
HPB 2016, 18, 296–303
Prior to undergoing hepatic resection, patients with hepatic metastatic disease often receive neoadjuvant chemotherapy, which can result in lesions “disappearing” or becoming radiologically occult (Fig. 1). Despite a complete response on imaging, many studies have shown persistent viable tumor at the site of “disappearing lesions”.6–12 With a lesion based analysis, the incidence of disappearing lesions ranges from 11%12 to 36%7 of initially detected liver metastases. With a patient based analysis, disappearing lesions have been shown to occur in 6%7 –23%13 of patients treated for colorectal metastatic disease. Reported complete pathologic response of disappearing lesions ranges from 15%10 to 73%.6 While the current surgical dogma mandates removal of all tumor if feasible, the role of disappearing lesions in pre-surgical planning remains controversial with no demonstrated survival benefit to resecting all sites of disappearing lesions.14
© 2016 International Hepato-Pancreato-Biliary Association Inc. Published by Elsevier Ltd. All rights reserved.
HPB
297
Figure 1 Coronal hepatobiliary 20 min delay MR image at baseline (a) demonstrating colorectal metastases in the liver (arrows). Coronal hepa-
tobiliary 20 min delay MR image at pre-surgical staging (b) demonstrating no lesion at the site of metastases on the prior exam (arrowheads).
Studies investigating disappearing lesions rely on a variety of imaging modalities, including PET/CT, US, contrast enhanced CT, MRI, and intra-operative US. Bischof et al. 2013 report a pooled sensitivity for hepatic metastases after chemotherapy of 85.7% for MRI and 69.9% for CT, and suggest MRI is the preferred modality for pre-operative evaluation of hepatic metastases. Over the past 7 years there has been an increasing reliance on a new MRI contrast agent gadoxetic acid (Eovist or Primovist; Bayer Healthcare Pharmaceuticals) for imaging colorectal hepatic metastases. Gadoxetic acid has increased sensitivity over previously available MR techniques,15 and shows improved sensitivity for small colorectal metastases over CT,16 especially in the setting of background hepatic steatosis. To date no study has investigated the rate of disappearing lesions with gadoxetic acid MR, or characterized the pathologic response of lesions which disappear on gadoxetic acid MR. Gadoxetic acid is characterized as a hepatobiliary contrast agent due to its high rate of uptake and excretion by functioning hepatocytes. On 20 min delayed imaging, (often referred to as hepatobiliary phase) gadoxetic acid uptake peaks in normal liver, with liver lesions becoming dark areas of non-enhancement (Fig. 2). The hepatobiliary phase allows for optimal conspicuity of lesions compared to background liver and is the reason for the improved sensitivity for lesion detection with gadoxetic acid compared to other MRI contrast agents. Ideally, patients with colorectal hepatic metastatic disease receive high quality quadruple phase CT or gadoxetic acid MR prior to undergoing neoadjuvant chemotherapy to serve as a roadmap for hepatic resection. In reality many of patients are
diagnosed and staged at outside institutions, and often receive chemotherapy prior to establishing care at tertiary care centers with experienced hepatobiliary surgeons. With outside imaging as a reference, Eovist MR may be performed prior to resection to assess treatment response and determine the extent of disease. The purpose of our study was to determine the incidence of disappearing colorectal metastases with state-of-the-art hepatobiliary agent MR imaging and assess the rate of complete pathologic response for disappearing lesions on hepatobiliary phase MR imaging. We hypothesize gadoxetic acid is more sensitive for viable colorectal metastases, and disappearance on gadoxetic acid imaging is more likely an indication of pathologic complete response compared to reported rates of viability on other less sensitive imaging modalities.
Methods Patients After IRB approval and waiver of consent, a database of surgery clinic patients was queried using the following ICD 9 codes – 197.7 (Secondary neoplasm of the liver) and 153.* (neoplasm of colon) or 154.0 (neoplasm rectosigmoid junction) or 154.1 (neoplasm rectum). The dates of inclusion were January 1st, 2008 to January 1st, 2014. Patients were then cross referenced to a database containing all patients imaged with gadoxetic acid at our institution. The individual medical records were then reviewed to determine eligibility based on the following inclusion criteria: diagnosis of colorectal cancer, evidence of metastatic colorectal cancer, treatment of colorectal metastatic disease with liver resection, pre-operative re-staging with
Figure 2 Arterial phase: hepatic artery and early portal vein enhancement (A). Portal Phase: peak portal vein enhancement (B). Transitional
phase: early parenchymal uptake (C). Transitional 5 min phase: gradual blood pool clearance (D). Hepatobiliary 20 min phase: blood pool clearance with peak normal liver parenchymal enhancement. Note the colorectal metastasis (arrow) and the arterial shunting (arrowhead).
HPB 2016, 18, 296–303
© 2016 International Hepato-Pancreato-Biliary Association Inc. Published by Elsevier Ltd. All rights reserved.
HPB
298
gadoxetic acid enhanced magnetic resonance imaging, and available baseline imaging. Given the retrospective nature of the investigation, the neoadjuvant chemotherapy regimens used to treat patients were heterogeneous and often administered at outside institutions prior to establishment of care with our hepatobiliary surgeons. Imaging evaluation Baseline imaging was defined as the time point when the sum of the long axis dimension of all the hepatic colorectal metastases was greatest. If the patient had progressive disease, baseline imaging was defined as the first imaging study demonstrating liver metastatic disease. Baseline imaging was either MRI or contrast enhanced CT. Baseline imaging was reviewed with follow-up imaging, operative reports, and surgical pathology available to maximize sensitivity for lesions. The maximum transaxial dimension of all visible lesions was recorded. The liver was assessed for hepatic steatosis by imaging criteria on baseline imaging and at pre-surgical staging (liver spleen index on CT, or loss of signal on dual echo chemical shift MRI). Pre-operative restaging was performed with weight-based double dose (0.05 mmol/kg) gadoxetic acid (Bayer Healthcare Pharmaceuticals) intravenous contrast on 1.5 T magnetic resonance scanners (Siemens Healthcare). MR examinations included standard sequences plus diffusion weighted imaging (b values 50, 400, 800) and 20 min hepatobiliary phase contrast enhanced imaging. Pre-operative restaging MR examinations were reviewed with all prior and subsequent imaging studies, operative notes, and surgical pathology available for reference to maximize sensitivity for lesions. The maximum transaxial dimension of all visible lesions was recorded. Follow-up imaging consisted of standard of care MR, CT and PET/CT. Sites of disappearing lesions which were not resected were followed for at least 1 year to assess for recurrence. All follow-up imaging was reviewed to assess for disease recurrence to determine disease free survival. Surgical pathology Pathology reports available in the medical record were reviewed and correlated with the operative reports, intraoperative ultrasound and prior imaging for correlation of lesion location and viability. Standard protocol for partial liver resections involves the following: The cauterized surface is inked to define the resection margin. The specimen is sectioned perpendicular to the margin every 3–5 mm. All lesions are described with estimated viability and location relative to the inked margin. Reference standard for lesion characterization Lesions were defined as “disappearing” if they were identifiable on the baseline imaging study but not identifiable on gadoxetic acid enhanced MRI during pre-surgical restaging. For lesions resected at surgery, surgical pathology reports were correlated with the imaging location of disappearing lesions. Lesions were
HPB 2016, 18, 296–303
characterized as “complete response” if they were not detected at surgical pathology or were detected but no residual viable tumor was present. Lesions were characterized as “persistent disease” if they were detected on surgical pathology and demonstrated viable tumor. For disappearing lesions which were not resected, follow-up imaging was reviewed for 1 year with close attention to the location of the unresected disappearing lesions. A lesion was characterized as “complete response” if it did not recur within 1 year. A lesion was characterized as “persistent disease” if a lesion recurred within 1 year. Statistics Data distributions were tested for normality with the Shapiro– Wilk W test. Equality of variances was tested with the O’Brien, Brown-Forsythe, Levene, Bartlett, and 2-sided F Test. If data were normally distributed and variances were equal the t test was used; otherwise, the Wilcoxon rank sums test was used. For normally distributed data, means, standard deviations, and 95% confidence intervals were used for descriptive statistics. For normally distributed data, medians plus 25th and 75th quantiles (quartiles) were used for descriptive statistics. Regression analysis was used to determine the best model to describe the relationship between two continuous variables. Logistic regression (logit regression) was used to predict a categorical dependent variable from an independent continuous variable, and probability plots were created for each analysis. Statistical analyses were performed with JMP Pro Statistical Software Release 11.0.0 (SAS Institute, Inc., Cary, NC), and MedCalc Statistics for Biomedical Research Version 15.2.1.0 (MedCalc Software, Mariakerke, Belgium).
Results Patient population Search of the surgical patient database found 603 patients evaluated by surgery who had a secondary neoplasm of the liver and either colon cancer, colorectal cancer, or rectal cancer. Of the total patients, 23 had hepatic metastases and met inclusion criteria with available imaging for review (Table 1). Average age at presentation was 53 years (range 37–74). There were 16 men and 9 women. Six patients had hepatic steatosis at baseline. Tumor stage was T4 = 3, T3 = 16, T2 = 3, and T1 = 1. Nodal stage was N2 = 5, N1 = 8, and N0 = 10. Thirteen patients had metastatic disease at initial presentation (synchronous metastases) and 10 patients developed metastases after staging (metachronous metastases). Three had undergone prior regional therapy (resection or ablation) for prior metastatic disease and had redeveloped hepatic metastases. Neoadjuvant treatment 23 patients underwent some form of neoadjuvant therapy prior to hepatic resection. The majority of patients were treated with FOLFOX (n = 16) +/− bevacizumab (n = 9) as first line therapy,
© 2016 International Hepato-Pancreato-Biliary Association Inc. Published by Elsevier Ltd. All rights reserved.
HPB
299
Table 1 Patient variables
Table 2 Surgical therapy
Patient Variables
N [ 23
Surgical therapy
N [ 23
Mean age (range)
53 (37–74)
Right hemi-hepatectomy
6
Right hemi-hepatectomy + non-anatomic resection or ablation
2
Gender Male
15
Female
8
Diagnosis of metastasis Synchronous Metachronous
13 10
T-Stage T4/3
19
T2/1
4
N-Stage N2
5
N1
8
N0
10
Bilateral metastatic disease
16
Extrahepatic Disease (pulmonary nodule)
1
Extended right hepatectomy
1
Extended right hepatectomy + non-anatomic resection
1
Left hemi-hepatectomy
1
Left lateral sectionectomy + non-anatomic resection or ablation
5
Non-anatomic resections ± ablation
7
response to therapy had synchronous metastatic disease. Bilateral hepatic disease was present in 16 patients at the time of surgical resection. Extrahepatic disease was present in only 1 patient at the time of surgery (a pulmonary nodule which was subsequently resected). Two patients developed hepatic steatosis from baseline with a total of 8 patients with hepatic steatosis undergoing resection.
Baseline imaging MR
13
CT
9
PET/CT
1
Hepatic steatosis at baseline
5
Hepatic steatosis pre-surgery
7
Radiological response (RECIST) Progression
2
Stable
9
Partial response
12
Complete response
0
Average chemotherapy holiday in weeks
6 (2.5–16)
Patients with disappearing lesions
11
Median number of lesions Lesions/patient
6
Lesions/patient In patients with disappearing lesions
11 (4–52)
Disappearing lesions/patient In patients with disappearing lesions
4 (1–31)
with FOLFIRI + bevacizumab (n = 5) the most common second line therapy. Right portal vein embolization was performed in 4 of the patients to facilitate extended hepatectomy. Two patients received transarterial chemoembolization as neoadjuvant therapy prior to hepatic resection. Response to neoadjuvant therapy was determined using RECIST criteria with progression of disease = 2, stable disease = 7, partial response = 12, and complete response = 0. Patients with partial response to therapy had an average change in lesion size of 60% (range 35–85%) with average baseline lesion size of 12.3 mm (range 4.6–27.9 mm). Nine of the 12 patients showing a partial
HPB 2016, 18, 296–303
Surgical therapy Multiple types of surgical resections and combination resection and ablation were performed (Table 2). All surgeries were performed with intent to cure. Intra-operative ultrasound was used at the discretion of the surgeon and was not universally performed. Chemotherapy was held for an average of 6 weeks prior to surgery (median = 6, range 2.5–16). Disappearing lesions A total of 200 metastatic lesions were identified in 23 patients undergoing hepatic resection (Table 3). Disappearing lesions occurred in 11/23 patients, and 11/12 of patients demonstrating partial response to therapy. Patients with synchronous metastatic disease (Fig. 3) were more likely to have disappearing lesions (odds ratio 11.25, p = 0.015). The median size of all lesions at baseline was 9 mm, (range 3–151 mm). At follow-up imaging 38.5% (77/200) of lesions had disappeared. The median size of disappearing lesions at baseline was 6 mm (range 3–32 mm), and the median size of visible lesions at baseline was 15 mm
Table 3 Lesion Variables (median; range)
Lesion Variables (median; range)
N [ 200
Number of disappearing lesions
77 (38.5%)
Baseline average size in mm All lesions Disappearing lesions
13 (3–151) 8 (3–32)
Disappearing lesions With complete response
32 (39%)
With persistent disease
40 (55%)
Indeterminate
5 (6%)
© 2016 International Hepato-Pancreato-Biliary Association Inc. Published by Elsevier Ltd. All rights reserved.
HPB
300
Figure 3 Mosaic plot – Contingency analysis demonstrates that the
odds for 1 disappearing lesions with in the “Synchronous” group is 11.25 times greater than in the “Metachronous” group (odds ratio – 11.25; p = 0.0064).
(3–151 mm). There was a statistically significant difference in the baseline size of lesions which disappeared compared with those that were still visible at presurgical staging (p < 0.0001). Of the disappearing lesions 39% (30/77) were found to have complete response and 55% (42/77) were found to have persistent disease based on pathology or imaging. Complete response or persistent disease could not be determined in 6% (5/77) of lesions (Fig. 4). In reviewing the MR imaging for all patients, several lesions were identified that were not accounted for in the intent to cure surgical plan. A total of 14 lesions (7% of all lesions) fell into this category and were not resected. In some instances these lesions were not individually identified by radiology report, in other cases they were not found at the time of surgery by intraoperative ultrasound and hence not resected. Thirteen lesions were positive for persistent disease with interval growth on follow-up imaging. These lesions for purposes of analysis were called “missed” lesions. The median size of the missed lesions on pre-surgical staging MR was 5 mm (range: 2–17 mm).
Figure 4 Diagram demonstrating the treatment and viability of disap-
pearing lesions.
HPB 2016, 18, 296–303
Survival and recurrence Average patient follow-up was 1285 days (median = 1205; range 212–2108). Five patients were lost to follow up at an average of 1135 days (median 1135; range 212–1637). One patient died during the follow-up interval at 921 days. The remaining patients were alive at the time of data censorship. Average disease free survival was 339 days (median 330; range 28–861). Patients undergoing intent to cure surgical resection had an average disease free survival of 470 days (median = 483; range 111–861). Patients with only disappearing lesions that were not resected had an average disease free survival of 329 days (median 360, range 109–519). No statistical difference was found, p = 0.49. The presence of missed lesions showed a statistically significant reduction in disease free survival with an average of 194 days (median 67; range 28–585; p = 0.015). Combining patients with missed lesions and disappearing lesions left behind also showed a statistically significant reduction in disease free survival with average of 232 days (median 109; range 28–585) (Fig. 5).
Discussion Despite reported improvements in imaging sensitivity with gadoxetic acid hepatobiliary MR contrast agent, 38.5% of colorectal metastases disappeared, and of the disappearing lesions 55% had surgical pathology or imaging evidence of persistent disease. This disproves the hypothesis that the increased sensitivity of gadoxetic acid for small hepatic metastases would reduce the incidence of disappearing lesions, and disproves the hypothesis that disappearing lesions with gadoxetic acid MR imaging would represent true complete response. (Fig. 6). Our incidence of disappearing lesions is greater than previous estimates in the literature. Ferrero et al., 2012 reported a 10.7% incidence of disappearing lesions (64/624), occurring in 19% of patients receiving preoperative chemotherapy. This difference may arise from the pre-chemotherapy staging by triphasic CT in all patients in Ferrero et al., 2012. In this investigation 13 of 23 patient’s baseline imaging with MR and 9 of those patients received a hepatobiliary MR agent. MR with hepatobiliary agents may be more sensitive to small lesions not detected by other imaging modalities, and small lesions may have an increased incidence of disappearance. Another reason for the increased incidence of disappearing metastases in our series may also be the interval improvements in chemotherapy with FOLFOX/ FOLFIRI and bevacizumab showing promising response with median progression free survival of 10.2 months and overall survival of 23.3 months.17 Literature reports of persistent disease associated with disappearing lesions is varied with persistent disease reported as high as 85% with PET/CT based imaging10 or as low as 27% with multimodality detection including US, CT, MR, intra-operative US, palpation and inspection.6 The sensitivity of PET/CT for small lesions is inferior to triphasic CT and MR and likely accounts for the high rate of persistent disease in Tan et al., 2007.
© 2016 International Hepato-Pancreato-Biliary Association Inc. Published by Elsevier Ltd. All rights reserved.
HPB
301
Figure 5 Kaplan Meier Survival Curve – Disease free survival was shorter for patients with “Lesion Left Behind” (Yellow) than for patients with all
lesions resected (Blue) with p=0.005.
Figure 6 Baseline contrast enhanced CT image showing extensive colorectal metastatic disease (arrows) in the liver (A). Presurgical staging MR
image with Gadoxetate Disodium in the hepatobiliary 20 min phase (B) showing resolution of previously seen lesions (arrowheads), and presurgical staging PET/CT (C) showing no increased FDG-uptake in the region of previously seen lesions (open arrowheads).
HPB 2016, 18, 296–303
© 2016 International Hepato-Pancreato-Biliary Association Inc. Published by Elsevier Ltd. All rights reserved.
HPB
302
The low rate of persistent disease in Elias et al., 2004 may be attributable to multimodality staging and the use of intra-arterial chemotherapy, which shows higher rates of lesional necrosis than standard chemotherapy. In addition to persistent viable tumor in disappeared lesions, there were a number of lesions, visible on MR, but not detected by intraoperative ultrasound which demonstrated persistent disease on follow-up imaging (previously described as ‘missed lesions’). These lesions were often not specifically identified in the radiology reports that guided surgical decision making, either due to a large number of lesions or small size. In some instances, the lesions were small and likely below the threshold for prospective detection, in other instances they were felt to be indeterminate by imaging. These lesions were visible on MR in retrospect, and were not found with intra-operative US, palpation and/or inspection, and were not resected. 14 lesions (2% of all lesions) were categorized as “missed lesions” and 13 of these lesions (93%) recurred. It may make sense for “disappearing lesions left behind” and “missed lesions left behind” to be grouped together since for practical purposes “missed lesions” were not able to be identified prospectively. In analyzing these groups together, the difference in disease free survival between patients with all lesions resected (460 days), and patients with lesions left behind (231 days) reaches statistical significance (p = 0.019). This brings up an important question about the true sensitivity of pre-operative imaging for detecting and facilitating the removal of all disease. This is especially difficult in the era of effective neoadjuvant chemotherapy and likely compounded by the difficulty of intraoperative ultrasound in finding small lesions. Reliance on intraoperative imaging to detect all lesions may be a pitfall resulting in a higher rate of residual disease. Our institution has placed of fiducial markers prior to neoadjuvant chemotherapy at the site of small metastases outside the planned anatomic resection in a small number of cases. In patients with synchronous bilateral metastatic disease who have lesions smaller than 10 mm outside of the planned resection margin a fiducial marker could be placed percutaneously to allow for nonanatomic resection or ablation of the site of disease at the time of surgery. It would not be practical to place fiducial markers at all sites of disease, but only at the location of small lesions outside the planned anatomic resection. Not taking into account “missed lesions” there was no statistical difference (p = 0.21) between the disease free survival of persons with disappearing lesions left behind (329 days), and those with complete resection (461 days), although that could reflect Type II error due to lack of power. Survival from hepatic resection and overall survival from initial diagnosis in this study was not meaningful as only 1 person is confirmed as deceased at the time of manuscript submission, (average follow-up was nearly 3.5 years, median 3.3 years; range 212–2108 days). This study is limited by its retrospective design, in which data was collected and determinations made from the best available
HPB 2016, 18, 296–303
data in the medical record. As with retrospective reviews the data was heterogeneous and patient management was varied. Chemotherapy regimens and the timing of therapy differed between patients. Baseline imaging was varied with the majority of patients undergoing baseline and pre-surgical MRI, but often patients had suboptimal baseline imaging at outside institutions and underwent chemotherapy prior to establishing care at our institution. The results reflect a single center experience and a small patient series, with a small subset of those patients treated with hepatic resection. Early cases which used gadoxetic acid have selection biased as they were often complex cases. Gadoxetic acid was not used routinely for colorectal metastases imaging until 2012. A future prospective study in which all patients are uniformly imaged with gadoxetic acid MR before chemotherapy and prior to surgery could provide additional insight into the role of hepatobiliary MR contrast agents and disappearing lesions on surgical management of hepatic metastases. The high sensitivity of gadoxetic acid MR for any parenchymal liver abnormality also creates the potential for unnecessary resection to treat false positive lesions, as signal characteristics of lesions less than 5 mm may not allow for differentiation between benign and malignant lesions. Our study did not specifically address false positive lesions. In our experience, characterization of small lesions especially distinguishing hemangiomas from small mucinous mets can be challenging. Our institutional policy on imaging with Eovist is to use it in patients with prior contrast enhanced CT or MRI. Additionally, we may rely on a change in lesion size on preoperative imaging compared to baseline, either increase in size with disease progression or decrease in size with partial response, to help with differentiating small lesions, but this does not help in all instances. Further research is needed to investigate to rate of false positive lesions and develop criteria to characterize lesions less than 5 mm. This is the first study to confirm the phenomenon of disappearing lesions with gadoxetic acid MR imaging, with 1/3 of colorectal metastases disappearing, and approximately 90% of patients responding to therapy developing at least 1 disappearing lesion. Despite the improved sensitivity, the disappearance of lesions on gadoxetic acid MRI does not imply a complete pathologic response given that approximately 1/2 of these lesions had persistent disease. The implications of disappearing lesions on survival are not well evaluated in our small patient population. However, patients with lesions left behind at the time of surgery, either ‘missed lesions’ or ‘disappeared lesions’ did have a worse prognosis. Important questions remain regarding the approach to colorectal liver metastases in the era of highly active neoadjuvant therapy. Further studies should evaluate the role of neoadjuvant treatment in patients with small lesions given the high rate of disappearance which may impact the future ability to resect. Additionally, greater focus in clinical practice on meticulous liver staging with best radiologic approach and a
© 2016 International Hepato-Pancreato-Biliary Association Inc. Published by Elsevier Ltd. All rights reserved.
HPB
303
multidisciplinary review of imaging is likely essential to achieving optimal results.
disappearing liver metastases from colorectal cancer after chemotherapy. Cancer 116:1502–1509. 9. Elias D, Goere D, Boige V, Kohneh-Sharhi N, Malka D, Tomasic G et al.
Summary A large percentage of metastases will become radiologically occult following current neoadjuvant chemotherapy. Disappearance of metastases on hepatobiliary gadoxetic acid MR imaging does not equal a complete pathologic response.
(2007 Oct 15) Outcome of posthepatectomy-missing colorectal liver metastases after complete response to chemotherapy: impact of adjuvant intra-arterial hepatic oxaliplatin. Ann Surg Oncol 14: 3188–3194. 10. Tan MCB, Linehan DC, Hawkins WG, Siegel BA, Strasberg SM. (2007 Aug 22) Chemotherapy-induced normalization of FDG uptake by colorectal liver metastases does not usually indicate complete pathologic
Funding sources None.
response. J Gastrointest Surg 11:1112–1119. 11. Tanaka K, Takakura H, Takeda K, Matsuo K, Nagano Y, Endo I. (2009 Dec) Importance of complete pathologic response to prehepatectomy
Conflict of interest None declared
chemotherapy in treating colorectal cancer metastases. Ann Surg 250: 935–942. 12. Ferrero A, Langella S, Russolillo N, Vigano’ L, Lo Tesoriere R,
References
Capussotti L. (2012 Apr) Intraoperative detection of disappearing
1. Jemal A, Bray F, Center MM, Ferlay J, Ward E, Forman D. (2011 Mar) Global cancer statistics. CA Cancer J Clin 61:69–90.
colorectal liver metastases as a predictor of residual disease. J Gastrointest Surg 16:806–814.
2. Weiss L, Grundmann E, Torhorst J, Hartveit F, Moberg I, Eder M et al.
13. Van Vledder MG, de Jong MC, Pawlik TM, Schulick RD, Diaz LA,
(1986 Nov) Haematogenous metastatic patterns in colonic carcinoma:
Choti MA. (2010 Nov) Disappearing colorectal liver metastases after
an analysis of 1541 necropsies. J Pathol 150:195–203.
chemotherapy: should we be concerned? J Gastrointest Surg 14:
3. Saltz LB, Cox JV, Blanke C, Rosen LS, Fehrenbacher L, Moore MJ et al. (2000 Sep 28) Irinotecan plus fluorouracil and leucovorin for metastatic colorectal cancer. N Engl J Med 343:905–914. 4. De Gramont A, Figer A, Seymour M, Homerin M, Hmissi A, Cassidy J
1691–1700. 14. Bischof DA, Clary BM, Maithel SK, Pawlik TM. (2013 Oct) Surgical management of disappearing colorectal liver metastases: disappearing colorectal liver metastasis. Br J Surg 100:1414–1420.
et al. (2000 Aug) Leucovorin and fluorouracil with or without oxaliplatin
15. Löwenthal D, Zeile M, Lim WY, Wybranski C, Fischbach F, Wieners G
as first-line treatment in advanced colorectal cancer. J Clin Oncol Off J
et al. (2011 Apr) Detection and characterisation of focal liver lesions in
Am Soc Clin Oncol 18:2938–2947.
colorectal carcinoma patients: comparison of diffusion-weighted and
5. Wei AC, Greig PD, Grant D, Taylor B, Langer B, Gallinger S. (2006 May) Survival after hepatic resection for colorectal metastases: a 10-year experience. Ann Surg Oncol 13:668–676.
Gd-EOB-DTPA enhanced MR imaging. Eur Radiol 21:832–840. 16. Scharitzer M, Ba-Ssalamah A, Ringl H, Kölblinger C, Grünberger T, Weber M et al. (2013 Aug) Preoperative evaluation of colorectal liver
6. Elias D, Youssef O, Sideris L, Dromain C, Baton O, Boige V et al. (2004
metastases: comparison between gadoxetic acid-enhanced 3.0-T MRI
Apr 1) Evolution of missing colorectal liver metastases following
and contrast-enhanced MDCT with histopathological correlation. Eur
inductive chemotherapy and hepatectomy. J Surg Oncol 86:4–9.
Radiol 23:2187–2196.
7. Benoist S. (2006 Aug 20) Complete response of colorectal liver me-
17. Bendell JC, Bekaii-Saab TS, Cohn AL, Hurwitz HI, Kozloff M, Tezcan H
tastases after chemotherapy: does it mean cure? J Clin Oncol 24:
et al. (2012 Dec 1) Treatment patterns and clinical outcomes in patients
3939–3945.
with metastatic colorectal cancer initially treated with FOLFOX-
8. Auer RC, White RR, Kemeny NE, Schwartz LH, Shia J, Blumgart LH et al. (2010 Mar 15) Predictors of a true complete response among
HPB 2016, 18, 296–303
bevacizumab or FOLFIRI-bevacizumab: results from ARIES, a bevacizumab observational cohort study. Oncologist 17:1486–1495.
© 2016 International Hepato-Pancreato-Biliary Association Inc. Published by Elsevier Ltd. All rights reserved.
![[Surgical management of disappearing colorectal liver metastases].](/assets/img/hold.png)
