Accepted Article
Received Date : 07-Sep-2014 Revised Date : 30-Jan-2015 Accepted Date : 26-Feb-2015 Article type
: Systematic Review
Systematic review and meta-analysis of the accuracy of MRI and ERUS in the restaging and response assessment of rectal cancer following neoadjuvant therapy
Sameer Memon, F.R.A.C.S.1*,
• Study conception and design • Acquisition of data • Analysis and interpretation of data • Writing manuscript
Andrew Craig Lynch M.Med.Sci., F.R.A.C.S.1 • Study conception and design • Writing manuscript Mathias Bressel M.Sc.2, • Study conception and design • Analysis and interpretation of data
Alan G Wise, F.R.A.N.Z.C.R, F.R.C.R, F.R.A.C.S. 3 • Interpretation of data • Writing manuscript
This article has been accepted for publication and undergone full peer review but has not been through the copyediting, typesetting, pagination and proofreading process which may lead to differences between this version and the Version of Record. Please cite this article as an 'Accepted Article', doi: 10.1111/codi.12976 This article is protected by copyright. All rights reserved.
Accepted Article
Alexander G Heriot, M.D., M.B.A., F.R.A.C.S. F.R.C.S.1 • Study conception • Analysis and interpretation of data • Writing manuscript
1
Colorectal Surgery Dept., Division of Cancer Surgery
2
Dept. of Biostatistics and Clinical Trials
3
Dept. of Cancer Imaging
Peter MacCallum Cancer Centre, East Melbourne Dept of Surgery, St Vincent’s Hospital, University of Melbourne
*Corresponding Author: Dr Sameer Memon Address: Dept. of Colorectal Surgery Peter MacCallum Cancer Centre Locked Bag 1 A’Beckett St VIC 8006 Australia Email: [email protected]
Key Words: Magnetic resonance imaging; Endoscopic ultrasonography; Rectal cancer; Staging; Meta-analysis
This article is protected by copyright. All rights reserved.
Accepted Article
Text Disclaimers: No funding has been used for the preparation of this manuscript. None of the authors have commercial associations that might pose a conflict of interest in connection with the submitted article.
ABSTRACT Aim: Restaging imaging by magnetic resonance imaging (MRI) or endorectal ultrasound (ERUS) following neoadjuvant chemoradiotherapy is not routinely performed, but the assessment of response is becoming increasingly important to facilitate individualisation of management. Method: A search of the Medline and Scopus databases was performed for studies which evaluated
the
accuracy
of
restaging
of
rectal
cancer
following
neoadjuvant
chemoradiotherapy with MRI or ERUS against the histopathological outcome. A systematic review of selected studies was performed. The methodological quality of studies that qualified for meta-analysis was critically assessed to identify studies suitable for inclusion in the meta-analysis. Results: 63 articles were included in the systematic review. Twelve restaging MRI studies and 18 restaging ERUS studies were eligible for meta-analysis of T-stage and N-status restaging accuracy. Overall, ERUS T-stage restaging accuracy (65% [56%-72%]) was nonsignificantly higher than MRI T-stage accuracy (52% [44%-59%]). Restaging MRI is accurate at excluding circumferential resection margin involvement. Restaging MRI and ERUS were equivalent at the prediction of nodal status: the accuracy of both investigations was 72% with over-staging and under-staging occurring in 10-15%. Conclusion: The heterogeneity amongst restaging studies is high, limiting conclusive findings regarding their accuracies. The accuracy of restaging imaging is different for different
This article is protected by copyright. All rights reserved.
Accepted Article
pathological T stages and highest for T3 tumours. Morphological assessment of T or N-stage by MRI or ERUS is currently not accurate or consistent enough for clinical application. Restaging MRI appears to have a role in excluding circumferential resection margin involvement.
What does this paper add to the literature? Response assessment is a very topical area of research in rectal cancer and has been assessed by restaging MRI and EUS in a large number of studies. MRI and EUS are imaging techniques which colorectal surgeons are familiar with as they are used for primary staging of rectal cancer and restaging is often performed by surgeons without a good understanding of the evidence base. The aim of this paper is to provide a comprehensive review of the restaging literature using these imaging techniques from a surgeon’s perspective and to approach metaanalysis of the data from a similar angle by assessing over-staging and under-staging of T-stage as opposed to performing an in-depth statistical analysis of the outcome which would not be as practically applicable. Such an approach to analysis of restaging imaging has not been previously performed. Although lengthy, this review covers all applications of restaging imaging with MRI and EUS and provides a concise summary of the evidence for situations, which we may encounter when performing restaging imaging.
INTRODUCTION Advances in the management of locally advanced rectal cancer (LARC) over the last decade have resulted in an increased application of multimodal therapy with the aim of tailoring neoadjuvant, surgical and adjuvant therapy. Central to this has been the finding that neoadjuvant radiotherapy decreases local recurrence (1) and results in a down-staging response in some patients (2). Key to individualisation of treatment is the ability to assess accurately and predict the response to neoadjuvant therapy, following which 20% of patients can be expected not to respond (3) and some may show progression of disease. These patients are at increased risk of an involved circumferential resection margin (CRM) (4) and they may also be at increased risk of recurrence due to the poor response phenotype of their
This article is protected by copyright. All rights reserved.
Accepted Article
16 Sizer B, Arulampalam T, Lacey N, Motson RW. Optimising the Timing of Surgery using Serial Magnetic Resonance Imaging after Long-Course Chemoradiotherapy for Locally Advanced Rectal Cancer. Clinical Oncology. 2009;21(10):787. 17 Baatrup G, Pfeiffer P, Svolgaard B, Jensen HA. Resectability of rectal cancers still fixed after radiochemotherapy: evaluation by digital rectal examination, MRI, and intraoperative examination. Int J Colorectal Dis. 2006 Jan;21(1):7-10. 18 Jacques AE, Rockall AG, Alijani M, et al. MRI demonstration of the effect of neoadjuvant radiotherapy on rectal carcinoma. Acta Oncol. 2007;46(7):989-95. 19 Cho YB, Chun HK, Kim MJ, et al. Accuracy of MRI and 18F-FDG PET/CT for restaging after preoperative concurrent chemoradiotherapy for rectal cancer. World J Surg. 2009 Dec;33(12):268894. 20 Sani F, Foresti M, Parmiggiani A, et al. 3-T MRI with phased-array surface coil in the local staging of rectal cancer. Radiol Med. 2011 Jan 12. 21 Kim YC, Lim JS, Keum KC, et al. Comparison of diffusion-weighted MRI and MR volumetry in the evaluation of early treatment outcomes after preoperative chemoradiotherapy for locally advanced rectal cancer. J Magn Reson Imaging. 2011;34(3):570-6. 22 Barbaro B, Fiorucci C, Tebala C, et al. Locally advanced rectal cancer: MR imaging in prediction of response after preoperative chemotherapy and radiation therapy. Radiology. 2009;250(3):730-9. 23 Chen CC, Lee RC, Lin JK, Wang LW, Yang SH. How accurate is magnetic resonance imaging in restaging rectal cancer in patients receiving preoperative combined chemoradiotherapy? Dis Colon Rectum. 2005 Apr;48(4):722-8. 24 Kim SH, Lee JM, Park HS, Eun HW, Han JK, Choi BI. Accuracy of MRI for predicting the circumferential resection margin, mesorectal fascia invasion, and tumor response to neoadjuvant chemoradiotherapy for locally advanced rectal cancer. J Magn Reson Imaging. 2009 May;29(5):1093101. 25 Yeo SG, Kim DY, Kim TH, et al. Tumor volume reduction rate measured by magnetic resonance volumetry correlated with pathologic tumor response of preoperative chemoradiotherapy for rectal cancer. International Journal of Radiation Oncology Biology Physics. 2010;78(1):164-71. 26 Young HK, Dae YK, Tae HK, et al. Usefulness of magnetic resonance volumetric evaluation in predicting response to preoperative concurrent chemoradiotherapy in patients with resectable rectal cancer. International Journal of Radiation Oncology Biology Physics. 2005;62(3):761-8. 27 Mezzi G, Arcidiacono PG, Carrara S, et al. Endoscopic ultrasound and magnetic resonance imaging for re-staging rectal cancer after radiotherapy. World Journal of Gastroenterology. 2009;15(44):5563-7. 28 Kuo LJ, Chern MC, Tsou MH, et al. Interpretation of magnetic resonance imaging for locally advanced rectal carcinoma after preoperative chemoradiation therapy. Diseases of the colon and rectum. 2005;48(1):23-8.
This article is protected by copyright. All rights reserved.
Accepted Article
Figure 2: The accuracy of restaging MRI for T-stage. The boxes represent the accuracy of each study, the size of the boxes represents its weight and the lines represent the 95% confidence interval. The diamond represents the average accuracy with 95% confidence interval.
W(random) - weighting of study .calculated using the random effects model
Figure 3: The accuracy of restaging MRI for N-status. The boxes represent the accuracy of each study, the size of the boxes represents its weight and the lines represent the 95% confidence interval. The diamond represents the average accuracy with 95% confidence interval.
This article is protected by copyright. All rights reserved.
Accepted Article
W(random) - weighting of study .calculated using the random effects model
Figure 4: The accuracy of restaging ERUS for T-stage. The boxes represent the accuracy of each study, the size of the boxes represents its weight and the lines represent the 95% confidence interval. The diamond represents the average accuracy with 95% confidence interval.
This article is protected by copyright. All rights reserved.
Accepted Article
Figure 5: The accuracy of restaging ERUS for N-status. The boxes represent the accuracy of each study, the size of the boxes represents its weight and the lines represent the 95% confidence interval. The diamond represents the average accuracy with 95% confidence interval.
This article is protected by copyright. All rights reserved.
Accepted Article
(22, 27) - 6 (30) weeks before surgery. The only pathological exclusion applied was in the series by Barbaro et al (22), who excluded mucinous tumours to avoid misinterpretation of mucin pools on restaging.
Meta-analysis of the accuracy of MRI T-Stage staging Fifteen studies assessed the T-stage restaging accuracy of MRI. In one study it was not possible to extract absolute T-stage numbers (32), in two some T stages were excluded (7, 33) and in one study the accuracy for only some T stages was calculated (22), leaving 12 studies eligible for meta-analysis (Table 1). The accuracy of restaging MRI prediction of T-stage ranged from 34% (34) to 82% (35) (Fig 2). The average accuracy was 52% [44-59%], (I2 = 58%). The majority of T0 and T1 tumours were over-staged by restaging MRI but this occurred less frequently for higher stage tumours (Table 2). Conversely 49% of T4 tumours were under-staged with this being less frequent for early stage tumours.
The prediction of down staging responses by restaging MRI Both surgically useful anatomical down-staging (eg down-staging to Stage 1 disease) as well as histological responses have been assessed by restaging MRI (Table 3). The study assessing morphological prediction of persistent T4 disease found a low positive predictive value (PPV) but high negative predictive value (NPV) (89%) for restaging MRI (33). Studies that morphologically assessed down-staging of Stage 1 disease showed poor accuracy owing to the combined error in assessing both T-stage and N-stage. As morphological assessment of over-staging can be difficult, alternative methods of the assessment of the response by MRI have been developed such as MRI volumetry, which quantifies MRI tumour volume regardless of its anatomical extent. A number of series found significant associations
This article is protected by copyright. All rights reserved.
Accepted Article
between volumetry and T down-staging (TDS) (22) (36) (30) (13) (25, 37), although others have not (38). The only study to assess early restaging MRI, two weeks after initiation of long course chemo-radiotherapy (LCCRT) found highly significant differences in tumour volume response ratios (TVRRs) between TDS response groups but not between TRG response groups (21). Three studies assessed complete regression, once by morphological assessment (39) and two by volumetry (25, 40). Curvo-Semedo et al. calculated the optimal cut-off for TVRR which resulted in high overall accuracy in their series for volumetry.
The prediction of CRM status by restaging MRI Nine studies assessed the CRM on restaging MRI. Five assessed the prediction of CRM involvement by restaging MRI (22, 24, 29, 41, 42) in patients with LARC (Table 4). As most patients had a non-involved CRM the overall accuracy was high accross the studies, but the PPVs and sensitivities were variable and as low as 30-50% respectively. . Over-staging was associated with mucinous degeneration (22) and cancers in the low rectum (24). Only the MERCURY group reported the quality of total mesorectal excision (41). The NPV for prediction of margin status improved on restaging MRI compared to primary staging MRI (98% vs 91%) although overall accuracy was similar (77% vs 81%) as a consequence of the majority of false positives (77%) occurring on restaging, also resulting in a poor PPV (45%). As morphological assessment of CRM involvement on restaging MRI may be inaccurate, Salerno et al. created the MRI-tumour response grade (mrTRG) to assess response (43). The mrTRG, which quantifies the extent of residual tumour signal intensity on restaging MRI and is scored similarly to Mandard's tumour regression grade (TRG), was significantly predictive of CRM involvement. Two studies investigated restaging MRI for the assessment of the CRM in LARC with radiologically threatened margins (44) (45). Both found an accuracy of 81% and high NPVs
This article is protected by copyright. All rights reserved.
Accepted Article
(91-100%) but other aspects of accuracy were variable (sensitivites 54%-100%, PPV 4457%).
Meta-analysis of the accuracy of restaging MRI in nodal restaging Fifteen articles assessed the per-patient accuracy of restaging MRI assessment of nodal involvement which were all eligible for meta-analysis (Table 1). There was variation between the studies in the criteria were used to define nodal involvement: five used morphologic changes only (19, 22, 32, 33, 46), four used size criteria only (23, 28, 29), four used both criteria to define involvement (7, 11, 44, 47) , and in two studies (27, 48) the criteria were not reported. The accuracy of restaging MRI assessment of N-stage ranged from 60% (33) - 88% (46) with an average accuracy of 72% [67-76%] (Fig. 3). Both over-staging and under-staging occurred in 16% [11-23%].
The prediction of Nodal Status by restaging MRI Two groups assessed novel MRI intravenous contrast agents. Lambregts et al found a significantly higher per-nodal accuracy for Gadofosvest-enhanced MRI compared with standard MRI (per node area under the curve (AUC) 0.88 vs 0.94) in 42 patients (47). Lahaye et al investigated two Ultra Small Particle Iron Oxidecontrast response parameters: Ratio A and estimated % white region of the node, in 43 patients (49). The AUC of the estimated % white region was 0.98 and the optimal cut-off resulted in 97% sensitivity and 98% specificity. The AUC for Ratio A was 0.99 and a cut-off of 0.34 resulted in 100% sensitivity and 96% specificity for predicting nodal involvement.
This article is protected by copyright. All rights reserved.
Accepted Article
RESTAGING ERUS Twenty-four studies assessing restaging ERUS were included in the systematic review. Surgery was performed between a mean of 2-3 (50) to 10-12 weeks (51) following neoadjuvant therapy. ERUS was performed within one to two weeks of surgery, with the exception of Pomerri et al (29) who performed it at four to five weeks prior to surgery. Four studies did not report the timing of ERUS (51-54).
Meta-analysis of ERUS T-stage Restaging Accuracy Twenty studies assessed the accuracy of restaging ERUS of T-stage. Two studies which excluded tumours of certain T-stages (55, 56) were ineligible for the meta-analysis leaving 18 studies (Table 5). All except one (29) were performed within two weeks of surgery. The accuracy of restaging ERUS of T-stage ranged from 26% (34) - 93% (22) with an average accuracy of 65% [56-72%] (Table 5). There was high heterogeneity amongst studies (I2 =82%) (Fig 4). Lower stage tumours showed a higher classification error rate, mostly due to under-staging, compared to T3 tumours (Table 2).
Meta-analysis of ERUS Nodal Restaging Accuracy Eighteen papers assessed the accuracy of restaging ERUS assessment of nodal status. One did not supply details on patient numbers (56) leaving 17 studies in the meta-analysis (Table 5). Definitions of involved nodes on restaging ERUS varied between studies: 4 studies assessed echogenicity only (53, 54, 57, 58), 4 studies assessed size only (29, 5961), 7 studies assessed size and echogenicity (22, 51, 62-66), in one study any node seen was defined as malignant (67) and in 1 study criteria were not defined (27). The accuracy of restaging ERUS in predicting of N-stage ranged from 57% (58) - 92% (66) with an
This article is protected by copyright. All rights reserved.
Accepted Article
average of 73% [67-78%] (Fig. 5). Over-staging of nodal status occurred in 12% [8 - 18%] and under-staging occurred in 14% [10 - 19%].
DISCUSSION The major limitation of many of the analyses in the present review and meta-analysis was the moderate to high heterogeneity between studies indicating that the accuracy was different among the studies. This limitation was also noted in the recent meta-analysis by Zhao et al despite their tighter inclusion criteria (68). We did not perform sensitivity analyses to asses for possible sources of heterogeneity as the quality of the data did not allow for useful subgroup analysis. However we have shown that the distribution of pathological tumour stages within studies may be a significant source of heterogeneity between studies as the accuracy of restaging imaging is different for different T-stages. The accuracy of ERUS for T3 disease was over 80% whilst it was 25% for T1 tumours. We also showed there was almost no heterogeneity between studies for the proportion of patients understaged for T-restaging by both modalities, giving more meaning to these meta-analysis results.
Measurement bias was likely the most significant source of error due to a lack of standardised technique for the interpretation of restaging images and inter-observer error. Only a few series reported the criteria used to interpret T and N-stage responses. There is still no consensus on the optimal technique for interpretation of restaging MRI. Some radiologists consider areas of very low signal intensity as sterile fibrosis (69) whilst others consider these to be areas of residual tumour (70). Regardless of the method of interpretation, it is likely that the problem of differentiating between residual microscopic disease and sterile fibrosis will continue to limit the clinical value of T-restaging imaging. Part of the reason for the large variation in the MRI technique between studies is
This article is protected by copyright. All rights reserved.
Accepted Article
experimentation with different protocols to optimise the accuracy of restaging imaging and advances in technology over the 15 year period during which the studies were performed. We did not exclude series based on the MRI protocol used owing to the large variation in technique between studies. The other significant source of heterogeneity was variation within and between studies of the timing of restaging imaging before surgery. The accuracy of MRI may have been diminished by the variable and often long interval between restaging and surgery in most series. Delayed imaging may be more accurate because it allows time for radiation-induced inflammation to settle and there is less potential for ongoing downstaging to occur before resection. Zhao et al attempted to investigate this further by performing meta-regression of the studies in their analysis, but they were unable to find any significant difference in accuracy between studies in which patients underwent restaging within or more than three weeks after surgery (68). Although ERUS was performed within two weeks of surgery in all studies except one in the meta-analysis, the accuracy of studies assessing restaging ERUS was also more heterogeneous than for MRI (I2=82% vs 58%) which may suggest greater inter-observer error for ERUS than MRI, as the accuracy of ERUS is known to be highly operator dependent (71). Despite the heterogeneity of the included data, a recent large series of restaging MRI in 285 patients with delayed imaging and consensus reporting by two radiologists found accuracies similar to those from our study- 48% T-stage accuracy and 68% N-stage accuracy (72).
The overall restaging T-stage accuracy of ERUS was non-significantly higher than for MRI. The only study to compare MRI and ERUS in the same patients at the same time, which would allow the most accurate comparison of techniques, reported the lowest accuracies for both methods at T-staging and no significant difference in accuracy between techniques at T or N-staging (29). T0, T1 and T2 tumours were frequently over-staged on MRI and ERUS and was highest for T0 tumours (73% and 66%) indicating both methods are poor at
This article is protected by copyright. All rights reserved.
Accepted Article
assessing complete T-response. The most notable difference in accuracy between restaging ERUS and MRI appears to be in restaging T3 disease (accuracy 83% vs 65%) Twice as many patients with T3 disease were over-staged (14% vs 8%) and significantly more T3 tumours were under-staged on MRI compared with ERUS (21% vs 9%). The ability accurately to predict T3 disease on restaging imaging is important as over-staging results in planning for exenterative resection and if restaging is to be used for planning local resection, under-staging of T3 disease may result in positive margins. Zhao et al similarly concluded that ERUS was slightly superior to MRI for T-restaging (68). Both modalities had poor sensitivity for T0 tumours (37% and 15% respectively) but high specificity (95%). Our findings suggest that neither method of restaging is sufficiently accurate to confirm T4 disease for clinical use (sensitivity 64% for ERUS and 51% for MRI). The series which specifically examined restaging of T4b tumours found negative predictive values of over 90% for restaging which support the accuracy of restaging MRI at excluding T4 disease (33). These findings are consistent with our findings of low rates of T3 over-staging by MRI and ERUS .
Assessment of the mesorectal plane following chemoradiation may identify poor-responders who may benefit from further neoadjuvant treatment or it may provide a road-map to tailor extended resection outside the mesorectal plane. An involved CRM in patients following neoadjuvant treatment has been shown to be independently predictive of distal recurrence (73) and an even stronger predictor of local recurrence than a positive CRM in untreated patients (4) because it identifies a subgroup of patients with an adverse tumour phenotype. As CRM status is closely related to the plane of mesorectal excision, the accuracy of MRI at assessing involvement of the mesorectal plane can only be accurately assessed if the plane of mesorectal excision is also reported. As most patients with LARC will have an uninvolved mesorectal plane on primary staging, for which the NPV is known to be high (41) and the majority of patients will also have a downsizing response to chemoradiation, the NPV and
This article is protected by copyright. All rights reserved.
Accepted Article
specificity of MRI would be expected to be high on restaging imaging. Restaging MRI also appears to have a high NPV in patients with an involved mesorectal plane on primary staging MRI (44, 45). In rectal cancer managed without LCCRT, the PPV of MRI for CRM involvement was 74% (41) due to the desmoplastic and inflammatory reaction and the limits of MRI spatial resolution especially in the low rectum (42). Following LCCRT, radiationinduced inflammation and the down-staging effect would be expected to decrease the accuracy further as was shown in this review with overestimation of margin involvement on restaging MRI resulting in variable PPVs (30-92%) and sensitivity (50-100%) across the studies. This suggests that the PPV of an involved margin on restaging MRI is not accurate enough to warrant exposing patients to intensified neoadjuvant treatment for the purpose of down-staging.
The ability to assess nodal response with good accuracy is essential for the successful management of rectal cancer without radical surgery. Most important is the minimisation of the rate of nodal under-staging as this prevents under-treatment of node positive tumours with non-radical surgery.
Persistent nodal involvement following LCCRT has also been
shown to be a strong predictor of local recurrence and survival (74-76) and identification these patients may also allow selection of patients for aggressive neoadjuvant local and/or systemic treatment. We found similar accuracies for both modalities at nodal restaging with under-staging in approximately 15%. Neoadjuvant treatment decreases the size and number of benign and malignant appearing lymph nodes on restaging MRI and causes a greater percentage of morphologically benign and small (
Received Date : 07-Sep-2014 Revised Date : 30-Jan-2015 Accepted Date : 26-Feb-2015 Article type
: Systematic Review
Systematic review and meta-analysis of the accuracy of MRI and ERUS in the restaging and response assessment of rectal cancer following neoadjuvant therapy
Sameer Memon, F.R.A.C.S.1*,
• Study conception and design • Acquisition of data • Analysis and interpretation of data • Writing manuscript
Andrew Craig Lynch M.Med.Sci., F.R.A.C.S.1 • Study conception and design • Writing manuscript Mathias Bressel M.Sc.2, • Study conception and design • Analysis and interpretation of data
Alan G Wise, F.R.A.N.Z.C.R, F.R.C.R, F.R.A.C.S. 3 • Interpretation of data • Writing manuscript
This article has been accepted for publication and undergone full peer review but has not been through the copyediting, typesetting, pagination and proofreading process which may lead to differences between this version and the Version of Record. Please cite this article as an 'Accepted Article', doi: 10.1111/codi.12976 This article is protected by copyright. All rights reserved.
Accepted Article
Alexander G Heriot, M.D., M.B.A., F.R.A.C.S. F.R.C.S.1 • Study conception • Analysis and interpretation of data • Writing manuscript
1
Colorectal Surgery Dept., Division of Cancer Surgery
2
Dept. of Biostatistics and Clinical Trials
3
Dept. of Cancer Imaging
Peter MacCallum Cancer Centre, East Melbourne Dept of Surgery, St Vincent’s Hospital, University of Melbourne
*Corresponding Author: Dr Sameer Memon Address: Dept. of Colorectal Surgery Peter MacCallum Cancer Centre Locked Bag 1 A’Beckett St VIC 8006 Australia Email: [email protected]
Key Words: Magnetic resonance imaging; Endoscopic ultrasonography; Rectal cancer; Staging; Meta-analysis
This article is protected by copyright. All rights reserved.
Accepted Article
Text Disclaimers: No funding has been used for the preparation of this manuscript. None of the authors have commercial associations that might pose a conflict of interest in connection with the submitted article.
ABSTRACT Aim: Restaging imaging by magnetic resonance imaging (MRI) or endorectal ultrasound (ERUS) following neoadjuvant chemoradiotherapy is not routinely performed, but the assessment of response is becoming increasingly important to facilitate individualisation of management. Method: A search of the Medline and Scopus databases was performed for studies which evaluated
the
accuracy
of
restaging
of
rectal
cancer
following
neoadjuvant
chemoradiotherapy with MRI or ERUS against the histopathological outcome. A systematic review of selected studies was performed. The methodological quality of studies that qualified for meta-analysis was critically assessed to identify studies suitable for inclusion in the meta-analysis. Results: 63 articles were included in the systematic review. Twelve restaging MRI studies and 18 restaging ERUS studies were eligible for meta-analysis of T-stage and N-status restaging accuracy. Overall, ERUS T-stage restaging accuracy (65% [56%-72%]) was nonsignificantly higher than MRI T-stage accuracy (52% [44%-59%]). Restaging MRI is accurate at excluding circumferential resection margin involvement. Restaging MRI and ERUS were equivalent at the prediction of nodal status: the accuracy of both investigations was 72% with over-staging and under-staging occurring in 10-15%. Conclusion: The heterogeneity amongst restaging studies is high, limiting conclusive findings regarding their accuracies. The accuracy of restaging imaging is different for different
This article is protected by copyright. All rights reserved.
Accepted Article
pathological T stages and highest for T3 tumours. Morphological assessment of T or N-stage by MRI or ERUS is currently not accurate or consistent enough for clinical application. Restaging MRI appears to have a role in excluding circumferential resection margin involvement.
What does this paper add to the literature? Response assessment is a very topical area of research in rectal cancer and has been assessed by restaging MRI and EUS in a large number of studies. MRI and EUS are imaging techniques which colorectal surgeons are familiar with as they are used for primary staging of rectal cancer and restaging is often performed by surgeons without a good understanding of the evidence base. The aim of this paper is to provide a comprehensive review of the restaging literature using these imaging techniques from a surgeon’s perspective and to approach metaanalysis of the data from a similar angle by assessing over-staging and under-staging of T-stage as opposed to performing an in-depth statistical analysis of the outcome which would not be as practically applicable. Such an approach to analysis of restaging imaging has not been previously performed. Although lengthy, this review covers all applications of restaging imaging with MRI and EUS and provides a concise summary of the evidence for situations, which we may encounter when performing restaging imaging.
INTRODUCTION Advances in the management of locally advanced rectal cancer (LARC) over the last decade have resulted in an increased application of multimodal therapy with the aim of tailoring neoadjuvant, surgical and adjuvant therapy. Central to this has been the finding that neoadjuvant radiotherapy decreases local recurrence (1) and results in a down-staging response in some patients (2). Key to individualisation of treatment is the ability to assess accurately and predict the response to neoadjuvant therapy, following which 20% of patients can be expected not to respond (3) and some may show progression of disease. These patients are at increased risk of an involved circumferential resection margin (CRM) (4) and they may also be at increased risk of recurrence due to the poor response phenotype of their
This article is protected by copyright. All rights reserved.
Accepted Article
16 Sizer B, Arulampalam T, Lacey N, Motson RW. Optimising the Timing of Surgery using Serial Magnetic Resonance Imaging after Long-Course Chemoradiotherapy for Locally Advanced Rectal Cancer. Clinical Oncology. 2009;21(10):787. 17 Baatrup G, Pfeiffer P, Svolgaard B, Jensen HA. Resectability of rectal cancers still fixed after radiochemotherapy: evaluation by digital rectal examination, MRI, and intraoperative examination. Int J Colorectal Dis. 2006 Jan;21(1):7-10. 18 Jacques AE, Rockall AG, Alijani M, et al. MRI demonstration of the effect of neoadjuvant radiotherapy on rectal carcinoma. Acta Oncol. 2007;46(7):989-95. 19 Cho YB, Chun HK, Kim MJ, et al. Accuracy of MRI and 18F-FDG PET/CT for restaging after preoperative concurrent chemoradiotherapy for rectal cancer. World J Surg. 2009 Dec;33(12):268894. 20 Sani F, Foresti M, Parmiggiani A, et al. 3-T MRI with phased-array surface coil in the local staging of rectal cancer. Radiol Med. 2011 Jan 12. 21 Kim YC, Lim JS, Keum KC, et al. Comparison of diffusion-weighted MRI and MR volumetry in the evaluation of early treatment outcomes after preoperative chemoradiotherapy for locally advanced rectal cancer. J Magn Reson Imaging. 2011;34(3):570-6. 22 Barbaro B, Fiorucci C, Tebala C, et al. Locally advanced rectal cancer: MR imaging in prediction of response after preoperative chemotherapy and radiation therapy. Radiology. 2009;250(3):730-9. 23 Chen CC, Lee RC, Lin JK, Wang LW, Yang SH. How accurate is magnetic resonance imaging in restaging rectal cancer in patients receiving preoperative combined chemoradiotherapy? Dis Colon Rectum. 2005 Apr;48(4):722-8. 24 Kim SH, Lee JM, Park HS, Eun HW, Han JK, Choi BI. Accuracy of MRI for predicting the circumferential resection margin, mesorectal fascia invasion, and tumor response to neoadjuvant chemoradiotherapy for locally advanced rectal cancer. J Magn Reson Imaging. 2009 May;29(5):1093101. 25 Yeo SG, Kim DY, Kim TH, et al. Tumor volume reduction rate measured by magnetic resonance volumetry correlated with pathologic tumor response of preoperative chemoradiotherapy for rectal cancer. International Journal of Radiation Oncology Biology Physics. 2010;78(1):164-71. 26 Young HK, Dae YK, Tae HK, et al. Usefulness of magnetic resonance volumetric evaluation in predicting response to preoperative concurrent chemoradiotherapy in patients with resectable rectal cancer. International Journal of Radiation Oncology Biology Physics. 2005;62(3):761-8. 27 Mezzi G, Arcidiacono PG, Carrara S, et al. Endoscopic ultrasound and magnetic resonance imaging for re-staging rectal cancer after radiotherapy. World Journal of Gastroenterology. 2009;15(44):5563-7. 28 Kuo LJ, Chern MC, Tsou MH, et al. Interpretation of magnetic resonance imaging for locally advanced rectal carcinoma after preoperative chemoradiation therapy. Diseases of the colon and rectum. 2005;48(1):23-8.
This article is protected by copyright. All rights reserved.
Accepted Article
Figure 2: The accuracy of restaging MRI for T-stage. The boxes represent the accuracy of each study, the size of the boxes represents its weight and the lines represent the 95% confidence interval. The diamond represents the average accuracy with 95% confidence interval.
W(random) - weighting of study .calculated using the random effects model
Figure 3: The accuracy of restaging MRI for N-status. The boxes represent the accuracy of each study, the size of the boxes represents its weight and the lines represent the 95% confidence interval. The diamond represents the average accuracy with 95% confidence interval.
This article is protected by copyright. All rights reserved.
Accepted Article
W(random) - weighting of study .calculated using the random effects model
Figure 4: The accuracy of restaging ERUS for T-stage. The boxes represent the accuracy of each study, the size of the boxes represents its weight and the lines represent the 95% confidence interval. The diamond represents the average accuracy with 95% confidence interval.
This article is protected by copyright. All rights reserved.
Accepted Article
Figure 5: The accuracy of restaging ERUS for N-status. The boxes represent the accuracy of each study, the size of the boxes represents its weight and the lines represent the 95% confidence interval. The diamond represents the average accuracy with 95% confidence interval.
This article is protected by copyright. All rights reserved.
Accepted Article
(22, 27) - 6 (30) weeks before surgery. The only pathological exclusion applied was in the series by Barbaro et al (22), who excluded mucinous tumours to avoid misinterpretation of mucin pools on restaging.
Meta-analysis of the accuracy of MRI T-Stage staging Fifteen studies assessed the T-stage restaging accuracy of MRI. In one study it was not possible to extract absolute T-stage numbers (32), in two some T stages were excluded (7, 33) and in one study the accuracy for only some T stages was calculated (22), leaving 12 studies eligible for meta-analysis (Table 1). The accuracy of restaging MRI prediction of T-stage ranged from 34% (34) to 82% (35) (Fig 2). The average accuracy was 52% [44-59%], (I2 = 58%). The majority of T0 and T1 tumours were over-staged by restaging MRI but this occurred less frequently for higher stage tumours (Table 2). Conversely 49% of T4 tumours were under-staged with this being less frequent for early stage tumours.
The prediction of down staging responses by restaging MRI Both surgically useful anatomical down-staging (eg down-staging to Stage 1 disease) as well as histological responses have been assessed by restaging MRI (Table 3). The study assessing morphological prediction of persistent T4 disease found a low positive predictive value (PPV) but high negative predictive value (NPV) (89%) for restaging MRI (33). Studies that morphologically assessed down-staging of Stage 1 disease showed poor accuracy owing to the combined error in assessing both T-stage and N-stage. As morphological assessment of over-staging can be difficult, alternative methods of the assessment of the response by MRI have been developed such as MRI volumetry, which quantifies MRI tumour volume regardless of its anatomical extent. A number of series found significant associations
This article is protected by copyright. All rights reserved.
Accepted Article
between volumetry and T down-staging (TDS) (22) (36) (30) (13) (25, 37), although others have not (38). The only study to assess early restaging MRI, two weeks after initiation of long course chemo-radiotherapy (LCCRT) found highly significant differences in tumour volume response ratios (TVRRs) between TDS response groups but not between TRG response groups (21). Three studies assessed complete regression, once by morphological assessment (39) and two by volumetry (25, 40). Curvo-Semedo et al. calculated the optimal cut-off for TVRR which resulted in high overall accuracy in their series for volumetry.
The prediction of CRM status by restaging MRI Nine studies assessed the CRM on restaging MRI. Five assessed the prediction of CRM involvement by restaging MRI (22, 24, 29, 41, 42) in patients with LARC (Table 4). As most patients had a non-involved CRM the overall accuracy was high accross the studies, but the PPVs and sensitivities were variable and as low as 30-50% respectively. . Over-staging was associated with mucinous degeneration (22) and cancers in the low rectum (24). Only the MERCURY group reported the quality of total mesorectal excision (41). The NPV for prediction of margin status improved on restaging MRI compared to primary staging MRI (98% vs 91%) although overall accuracy was similar (77% vs 81%) as a consequence of the majority of false positives (77%) occurring on restaging, also resulting in a poor PPV (45%). As morphological assessment of CRM involvement on restaging MRI may be inaccurate, Salerno et al. created the MRI-tumour response grade (mrTRG) to assess response (43). The mrTRG, which quantifies the extent of residual tumour signal intensity on restaging MRI and is scored similarly to Mandard's tumour regression grade (TRG), was significantly predictive of CRM involvement. Two studies investigated restaging MRI for the assessment of the CRM in LARC with radiologically threatened margins (44) (45). Both found an accuracy of 81% and high NPVs
This article is protected by copyright. All rights reserved.
Accepted Article
(91-100%) but other aspects of accuracy were variable (sensitivites 54%-100%, PPV 4457%).
Meta-analysis of the accuracy of restaging MRI in nodal restaging Fifteen articles assessed the per-patient accuracy of restaging MRI assessment of nodal involvement which were all eligible for meta-analysis (Table 1). There was variation between the studies in the criteria were used to define nodal involvement: five used morphologic changes only (19, 22, 32, 33, 46), four used size criteria only (23, 28, 29), four used both criteria to define involvement (7, 11, 44, 47) , and in two studies (27, 48) the criteria were not reported. The accuracy of restaging MRI assessment of N-stage ranged from 60% (33) - 88% (46) with an average accuracy of 72% [67-76%] (Fig. 3). Both over-staging and under-staging occurred in 16% [11-23%].
The prediction of Nodal Status by restaging MRI Two groups assessed novel MRI intravenous contrast agents. Lambregts et al found a significantly higher per-nodal accuracy for Gadofosvest-enhanced MRI compared with standard MRI (per node area under the curve (AUC) 0.88 vs 0.94) in 42 patients (47). Lahaye et al investigated two Ultra Small Particle Iron Oxidecontrast response parameters: Ratio A and estimated % white region of the node, in 43 patients (49). The AUC of the estimated % white region was 0.98 and the optimal cut-off resulted in 97% sensitivity and 98% specificity. The AUC for Ratio A was 0.99 and a cut-off of 0.34 resulted in 100% sensitivity and 96% specificity for predicting nodal involvement.
This article is protected by copyright. All rights reserved.
Accepted Article
RESTAGING ERUS Twenty-four studies assessing restaging ERUS were included in the systematic review. Surgery was performed between a mean of 2-3 (50) to 10-12 weeks (51) following neoadjuvant therapy. ERUS was performed within one to two weeks of surgery, with the exception of Pomerri et al (29) who performed it at four to five weeks prior to surgery. Four studies did not report the timing of ERUS (51-54).
Meta-analysis of ERUS T-stage Restaging Accuracy Twenty studies assessed the accuracy of restaging ERUS of T-stage. Two studies which excluded tumours of certain T-stages (55, 56) were ineligible for the meta-analysis leaving 18 studies (Table 5). All except one (29) were performed within two weeks of surgery. The accuracy of restaging ERUS of T-stage ranged from 26% (34) - 93% (22) with an average accuracy of 65% [56-72%] (Table 5). There was high heterogeneity amongst studies (I2 =82%) (Fig 4). Lower stage tumours showed a higher classification error rate, mostly due to under-staging, compared to T3 tumours (Table 2).
Meta-analysis of ERUS Nodal Restaging Accuracy Eighteen papers assessed the accuracy of restaging ERUS assessment of nodal status. One did not supply details on patient numbers (56) leaving 17 studies in the meta-analysis (Table 5). Definitions of involved nodes on restaging ERUS varied between studies: 4 studies assessed echogenicity only (53, 54, 57, 58), 4 studies assessed size only (29, 5961), 7 studies assessed size and echogenicity (22, 51, 62-66), in one study any node seen was defined as malignant (67) and in 1 study criteria were not defined (27). The accuracy of restaging ERUS in predicting of N-stage ranged from 57% (58) - 92% (66) with an
This article is protected by copyright. All rights reserved.
Accepted Article
average of 73% [67-78%] (Fig. 5). Over-staging of nodal status occurred in 12% [8 - 18%] and under-staging occurred in 14% [10 - 19%].
DISCUSSION The major limitation of many of the analyses in the present review and meta-analysis was the moderate to high heterogeneity between studies indicating that the accuracy was different among the studies. This limitation was also noted in the recent meta-analysis by Zhao et al despite their tighter inclusion criteria (68). We did not perform sensitivity analyses to asses for possible sources of heterogeneity as the quality of the data did not allow for useful subgroup analysis. However we have shown that the distribution of pathological tumour stages within studies may be a significant source of heterogeneity between studies as the accuracy of restaging imaging is different for different T-stages. The accuracy of ERUS for T3 disease was over 80% whilst it was 25% for T1 tumours. We also showed there was almost no heterogeneity between studies for the proportion of patients understaged for T-restaging by both modalities, giving more meaning to these meta-analysis results.
Measurement bias was likely the most significant source of error due to a lack of standardised technique for the interpretation of restaging images and inter-observer error. Only a few series reported the criteria used to interpret T and N-stage responses. There is still no consensus on the optimal technique for interpretation of restaging MRI. Some radiologists consider areas of very low signal intensity as sterile fibrosis (69) whilst others consider these to be areas of residual tumour (70). Regardless of the method of interpretation, it is likely that the problem of differentiating between residual microscopic disease and sterile fibrosis will continue to limit the clinical value of T-restaging imaging. Part of the reason for the large variation in the MRI technique between studies is
This article is protected by copyright. All rights reserved.
Accepted Article
experimentation with different protocols to optimise the accuracy of restaging imaging and advances in technology over the 15 year period during which the studies were performed. We did not exclude series based on the MRI protocol used owing to the large variation in technique between studies. The other significant source of heterogeneity was variation within and between studies of the timing of restaging imaging before surgery. The accuracy of MRI may have been diminished by the variable and often long interval between restaging and surgery in most series. Delayed imaging may be more accurate because it allows time for radiation-induced inflammation to settle and there is less potential for ongoing downstaging to occur before resection. Zhao et al attempted to investigate this further by performing meta-regression of the studies in their analysis, but they were unable to find any significant difference in accuracy between studies in which patients underwent restaging within or more than three weeks after surgery (68). Although ERUS was performed within two weeks of surgery in all studies except one in the meta-analysis, the accuracy of studies assessing restaging ERUS was also more heterogeneous than for MRI (I2=82% vs 58%) which may suggest greater inter-observer error for ERUS than MRI, as the accuracy of ERUS is known to be highly operator dependent (71). Despite the heterogeneity of the included data, a recent large series of restaging MRI in 285 patients with delayed imaging and consensus reporting by two radiologists found accuracies similar to those from our study- 48% T-stage accuracy and 68% N-stage accuracy (72).
The overall restaging T-stage accuracy of ERUS was non-significantly higher than for MRI. The only study to compare MRI and ERUS in the same patients at the same time, which would allow the most accurate comparison of techniques, reported the lowest accuracies for both methods at T-staging and no significant difference in accuracy between techniques at T or N-staging (29). T0, T1 and T2 tumours were frequently over-staged on MRI and ERUS and was highest for T0 tumours (73% and 66%) indicating both methods are poor at
This article is protected by copyright. All rights reserved.
Accepted Article
assessing complete T-response. The most notable difference in accuracy between restaging ERUS and MRI appears to be in restaging T3 disease (accuracy 83% vs 65%) Twice as many patients with T3 disease were over-staged (14% vs 8%) and significantly more T3 tumours were under-staged on MRI compared with ERUS (21% vs 9%). The ability accurately to predict T3 disease on restaging imaging is important as over-staging results in planning for exenterative resection and if restaging is to be used for planning local resection, under-staging of T3 disease may result in positive margins. Zhao et al similarly concluded that ERUS was slightly superior to MRI for T-restaging (68). Both modalities had poor sensitivity for T0 tumours (37% and 15% respectively) but high specificity (95%). Our findings suggest that neither method of restaging is sufficiently accurate to confirm T4 disease for clinical use (sensitivity 64% for ERUS and 51% for MRI). The series which specifically examined restaging of T4b tumours found negative predictive values of over 90% for restaging which support the accuracy of restaging MRI at excluding T4 disease (33). These findings are consistent with our findings of low rates of T3 over-staging by MRI and ERUS .
Assessment of the mesorectal plane following chemoradiation may identify poor-responders who may benefit from further neoadjuvant treatment or it may provide a road-map to tailor extended resection outside the mesorectal plane. An involved CRM in patients following neoadjuvant treatment has been shown to be independently predictive of distal recurrence (73) and an even stronger predictor of local recurrence than a positive CRM in untreated patients (4) because it identifies a subgroup of patients with an adverse tumour phenotype. As CRM status is closely related to the plane of mesorectal excision, the accuracy of MRI at assessing involvement of the mesorectal plane can only be accurately assessed if the plane of mesorectal excision is also reported. As most patients with LARC will have an uninvolved mesorectal plane on primary staging, for which the NPV is known to be high (41) and the majority of patients will also have a downsizing response to chemoradiation, the NPV and
This article is protected by copyright. All rights reserved.
Accepted Article
specificity of MRI would be expected to be high on restaging imaging. Restaging MRI also appears to have a high NPV in patients with an involved mesorectal plane on primary staging MRI (44, 45). In rectal cancer managed without LCCRT, the PPV of MRI for CRM involvement was 74% (41) due to the desmoplastic and inflammatory reaction and the limits of MRI spatial resolution especially in the low rectum (42). Following LCCRT, radiationinduced inflammation and the down-staging effect would be expected to decrease the accuracy further as was shown in this review with overestimation of margin involvement on restaging MRI resulting in variable PPVs (30-92%) and sensitivity (50-100%) across the studies. This suggests that the PPV of an involved margin on restaging MRI is not accurate enough to warrant exposing patients to intensified neoadjuvant treatment for the purpose of down-staging.
The ability to assess nodal response with good accuracy is essential for the successful management of rectal cancer without radical surgery. Most important is the minimisation of the rate of nodal under-staging as this prevents under-treatment of node positive tumours with non-radical surgery.
Persistent nodal involvement following LCCRT has also been
shown to be a strong predictor of local recurrence and survival (74-76) and identification these patients may also allow selection of patients for aggressive neoadjuvant local and/or systemic treatment. We found similar accuracies for both modalities at nodal restaging with under-staging in approximately 15%. Neoadjuvant treatment decreases the size and number of benign and malignant appearing lymph nodes on restaging MRI and causes a greater percentage of morphologically benign and small (
![[Restaging of rectal cancer with MRI and endoscopic ultrasound after neoadjuvant chemoradiotherapy].](/assets/img/hold.png)
