http://informahealthcare.com/sju ISSN: 2168-1805 (print), 2168-1813 (electronic) Scand J Urol, 2014; 49(2): 142–148 2014 Informa Healthcare. DOI: 10.3109/21681805.2014.969307
ORIGINAL ARTICLE
Diagnostic accuracy of preoperative computed tomography used alone to detect lymph-node involvement at radical nephrectomy Stephen S. Connolly1, Aditya Raja1, Helen Stunell2, Deepak Parashar3,4, Sara Upponi2, Anne Y. Warren5, Vincent J. Gnanapragasam1 and Tim Eisen6 1
Departments of Urology and, 2Radiology, 3Cambridge Cancer Trials Centre and, 4MRC Biostatistics Unit Hub for Trials Methodology Research, and, Departments of Histopathology, and 6Oncology, Addenbrooke’s Hospital and Cambridge University Health Partners, Cambridge, UK
5
Abstract
Keywords:
Objective. The aim of this study was to compare preoperative computed tomography (CT) with pathological findings in patients undergoing lymphadenectomy at the time of nephrectomy for renal cancer-associated lymphadenopathy. Materials and methods. Data from 515 consecutive nephrectomy surgeries (2004–2012) in a single university-affiliated centre were analysed to identify patients who had undergone lymph-node dissection concomitant with nephrectomy. Preoperative CT imaging was subjected to multiple repeated independent blinded reviews (two radiologists and one surgeon, each individually and on two separate occasions). Retroperitoneal lymph-node status was subjectively categorized (in a manner not based purely on size criteria) at each review as: 1 = unequivocally positive, 2 = equivocally positive, 3 = equivocally negative, or 4 = unequivocally negative. These findings were compared with pathological analysis, and interobserver and intraobserver agreement was assessed using non-weighted kappa () statistics. Results. In total, 71 patients were stratified as category 1 (n = 18), 2 (n = 14), 3 (n = 31) and 4 (n = 8); pathological lymph-node metastasis was present in 14 (78%), four (28%), four (13%) and zero patients, respectively. Sensitivity, specificity, positive and negative predictive values for preoperative CT were 82%, 71%, 56% and 90%, respectively. Intraobserver agreement was greater for the radiologists (values 0.490, 0.540) than for the surgeon (value 0.393). Interobserver agreement was strongest for radiological category 1 (unequivocally positive; value 0.75). Receiver operating characteristics curves did not reveal significant differences in any observer accuracy. Conclusion. Contrary to concerns about a high false-positive rate, metastasis within regional lymph nodes can be predicted with reasonable accuracy by preoperative CT imaging alone.
Kidney, lymphadenectomy, lymphadenopathy, renal carcinoma
Introduction Although renal cancer (renal cell carcinoma; RCC) is more renowned for haematogenous spread, lymph-node metastasis can occur without more disseminated disease. Deserved prominence is allocated to lymph nodes in the tumour, node, metastasis (TNM) staging system originally devised by Pierre Denoix [1] and now maintained by the Union for International Against Cancer [2]. More recently, numerous postnephrectomy prognostic nomograms specific to RCC have allocated considerable weight to lymph-node status [3–6]. But despite the clear lethality of RCC and the undisputed influence of lymph nodes, universal consensus on the contribution of lymph-node dissection (LND) for RCC has not yet been achieved. Although European Association of Urology (EAU) guidelines reasonably recommend that extended lymphadenectomy is not recommended and should be confined to staging purposes, this advice is largely based on a single prospective randomized trial [7]. The European Organisation for Research and Treatment of Cancer (EORTC) trial Correspondence: Stephen Connolly, Department of Urology, Addenbrooke’s Hospital, Box 43, Hills Road, Cambridge CB2 0QQ, UK. Tel: +44 1223 257261. Fax: +44 1223 216069. E-mail: [email protected]
History Received 11 February 2014 Revised 7 September 2014 Accepted 15 September 2014 Online 10 January 2015
30881 deliberately excluded patients in whom nodes were radiologically enlarged and notably the vast majority (96%) of patients who underwent LND had negative nodes [8]. Given these limitations, it seems reasonable to consider the lymphadenectomy debate with RCC unresolved, and it seems likely that some arguments for the discriminate use of LND will persist. The lack of popularity for LND with RCC is not likely to be attributable to any one single factor, but rather is complex and multifactorial. The single prospective trial of LND with RCC, the EORTC randomized phase 3 trial 30881, reported the frequency of positive nodes to be so low (4%) that routine regional LND cannot be justified [8]. Controversy exists as to whether LND provides improved staging alone or also offers a therapeutic survival advantage [9]. None of the major international guidelines currently promotes therapeutic LND and although the EAU guidelines (2010) state that “resection of affected lymph nodes should be performed”, it is clarified to be purely for the purpose of staging. It has been suggested that patients with lymph-node metastases and RCC fare worse than those with metastatic renal cancer in the absence of retroperitoneal lymphadenopathy [10]. Furthermore, while nephrectomy has proven to be superbly suited to laparoscopy, retroperitoneal LND indisputably represents a much
Lymphadenopathy and renal cancer
DOI: 10.3109/21681805.2014.969307
bigger challenge for the average-skilled laparoscopist [11–13]. As a result, perhaps “subtle” lymphadenopathy frequently remains ignored as the surgeon focuses on removal of the primary tumour with minimal morbidity, whereas overt lymphadenopathy may be considered by some to simply represent surgically incurable disease, even in the absence of other disseminated visceral metastases. Also of considerable relevance is the fact that there is substantial uncertainty regarding the true significance of lymphadenopathy with RCC, and node enlargement not attributable to renal cancer metastasis is frequently explained by “reactive change” including inflammation and follicular hyperplasia. In a frequently cited seminal publication, Studer et al. reported in 1990 that lymphadenopathy with RCC, based purely on size criteria (‡1 cm), was associated with metastasis in only 18 (42%) of 43 patients, a false-positive rate approximating 60% [14]. Others expanded this concern further, with Ming et al., in a demonstration of the utility of frozen section node examination, finding only 36 (31.6%) of 114 patients with enlarged nodes to harbour malignancy, a false-positive rate approaching 70% [15]. Most recently, in a subanalysis of patients with intraoperative lymph-node enlargement in the EORTC trial, resection of enlarged nodes in 84 patients found that 14 contained cancer, producing a false-positive rate in excess of 80% [8]. Currently, renal cancer is most commonly staged by computed tomography (CT), and locoregional lymph-node abnormality is evaluated purely on the basis of size [7]. Radiological accuracy for predicting metastasis within lymph nodes represents a considerable problem but is not one exclusive to renal cancer, nor is it a problem solved by other imaging modalities such as magnetic resonance imaging (MRI). A meta-analysis in the setting of prostate cancer found that CT and MRI demonstrate equally poor performance in the detection of lymph-node metastasis [16]. This study examines the correlation between CT-predicted retroperitoneal lymph-node metastasis with RCC and true pathological involvement of lymph nodes in a contemporary setting of patients undergoing lymphadenectomy at the time of nephrectomy due to lymphadenopathy. Furthermore, by comparison of multiple blinded reviews of preoperative imaging, the subjectivity that may surround CT-reported lymphadenopathy in this setting is addressed.
Materials and methods Study population and histopathological data A retrospective review of the departmental database of 515 consecutive nephrectomy surgeries performed between July 2004 and January 2012 was undertaken in compliance with international ethical standards. All patients with RCC and lymphadenectomy tissue submitted to pathology on the day of surgery were identified. The departmental database routinely includes primary renal tumour parameters such as size, renal vein involvement and the presence of synchronous visceral metastasis. All histopathology information for this study was taken from the formal histopathology reports issued in real time associated with each clinical episode, and
143
all reports were in accordance with the latest version of the Royal College of Pathologists’ histopathology reporting guidelines for renal cancer [2,17]. Of note, all macroscopically negative lymph nodes were routinely submitted for tissue processing and histological examination in their entirety, regardless of size. Lymphadenectomy surgery Operative records were reviewed. During the period under observation, the indication for LND in this unit was either enlarged lymph nodes identified preoperatively at conventional CT (>1 cm in short axis) or found to be enlarged intraoperatively, or suspicion of nodal metastasis based on advanced local stage. Routine LND has never been routinely performed at any time in this institution for “normal” lymph nodes. All patients with incidental retrieval of lymph nodes (e.g. hilar nodes frequently retrieved during laparoscopic nephrectomy) were excluded from analysis in this study. The type of LND performed in this series was not standardized (and cannot be considered universally to be a thorough retroperitoneal LND in each case), but an attempt was made in each case to remove all enlarged nodes. Radiology review For the specific purpose of this study, multiple retrospective reviews of the preoperative (CT) imaging available on the hospital picture archiving and communications system for each patient were undertaken. All preoperative imaging was required to be within 2 months of surgery, and although CT protocols were not uniform, all imaging included contrast enhancement from a multislice scanner. Reviews were undertaken in a blinded manner (to both the official radiology and histopathology reports) and were performed individually by a consultant surgeon with special interest in renal cancer, and also independently by two consultant radiologists with special interest in abdominal imaging. Each reviewer performed an independent review on two separate occasions at least 2 weeks apart. It was agreed by the team involved in this study that considerable subjectivity may exist in the definition of regional lymphadenopathy for patients with RCC. For this study, information provided to the reviewer immediately before review was confined to the fact that the patient had undergone nephrectomy for a renal cancer with removal of lymph nodes. Each reviewer was then asked to predict nodal involvement by choosing one of four options: category 1, “Unequivocal positive (for metastasis)”; 2, “Equivocal node enlargement – probably positive”; 3, “Equivocal node enlargement – probably negative”; or 4, “Unequivocal negative” nodes (with zero radiological concern for metastasis). Importantly, the criteria employed for the prediction of nodal involvement were not purely based on size, but are described in detail in Table 1. No specific weighting was attributed to any individual factor and no specific formula was devised. Notably, this constitutes a completely novel categorization system and has never been previously employed, and therefore has not been validated.
144 S. S. Connolly et al.
Scand J Urol, 2014; 49(2): 142–148
Table 1. Radiological criteria for the prediction of lymph-node metastasis. Radiological node category 1
2
3 4 a
Descriptiona
Implication
Overtly enlarged single or multiple nodes (usually ‡12 mm in short axis dimension) and considered very likely to harbour malignancy. One would reasonably expect this category to be readily commented upon by all reporting radiologists upon review of the same imaging Fulfilling criteria for clear concern based on size (>1 cm in short axis dimension) and appearance, but generally
ORIGINAL ARTICLE
Diagnostic accuracy of preoperative computed tomography used alone to detect lymph-node involvement at radical nephrectomy Stephen S. Connolly1, Aditya Raja1, Helen Stunell2, Deepak Parashar3,4, Sara Upponi2, Anne Y. Warren5, Vincent J. Gnanapragasam1 and Tim Eisen6 1
Departments of Urology and, 2Radiology, 3Cambridge Cancer Trials Centre and, 4MRC Biostatistics Unit Hub for Trials Methodology Research, and, Departments of Histopathology, and 6Oncology, Addenbrooke’s Hospital and Cambridge University Health Partners, Cambridge, UK
5
Abstract
Keywords:
Objective. The aim of this study was to compare preoperative computed tomography (CT) with pathological findings in patients undergoing lymphadenectomy at the time of nephrectomy for renal cancer-associated lymphadenopathy. Materials and methods. Data from 515 consecutive nephrectomy surgeries (2004–2012) in a single university-affiliated centre were analysed to identify patients who had undergone lymph-node dissection concomitant with nephrectomy. Preoperative CT imaging was subjected to multiple repeated independent blinded reviews (two radiologists and one surgeon, each individually and on two separate occasions). Retroperitoneal lymph-node status was subjectively categorized (in a manner not based purely on size criteria) at each review as: 1 = unequivocally positive, 2 = equivocally positive, 3 = equivocally negative, or 4 = unequivocally negative. These findings were compared with pathological analysis, and interobserver and intraobserver agreement was assessed using non-weighted kappa () statistics. Results. In total, 71 patients were stratified as category 1 (n = 18), 2 (n = 14), 3 (n = 31) and 4 (n = 8); pathological lymph-node metastasis was present in 14 (78%), four (28%), four (13%) and zero patients, respectively. Sensitivity, specificity, positive and negative predictive values for preoperative CT were 82%, 71%, 56% and 90%, respectively. Intraobserver agreement was greater for the radiologists (values 0.490, 0.540) than for the surgeon (value 0.393). Interobserver agreement was strongest for radiological category 1 (unequivocally positive; value 0.75). Receiver operating characteristics curves did not reveal significant differences in any observer accuracy. Conclusion. Contrary to concerns about a high false-positive rate, metastasis within regional lymph nodes can be predicted with reasonable accuracy by preoperative CT imaging alone.
Kidney, lymphadenectomy, lymphadenopathy, renal carcinoma
Introduction Although renal cancer (renal cell carcinoma; RCC) is more renowned for haematogenous spread, lymph-node metastasis can occur without more disseminated disease. Deserved prominence is allocated to lymph nodes in the tumour, node, metastasis (TNM) staging system originally devised by Pierre Denoix [1] and now maintained by the Union for International Against Cancer [2]. More recently, numerous postnephrectomy prognostic nomograms specific to RCC have allocated considerable weight to lymph-node status [3–6]. But despite the clear lethality of RCC and the undisputed influence of lymph nodes, universal consensus on the contribution of lymph-node dissection (LND) for RCC has not yet been achieved. Although European Association of Urology (EAU) guidelines reasonably recommend that extended lymphadenectomy is not recommended and should be confined to staging purposes, this advice is largely based on a single prospective randomized trial [7]. The European Organisation for Research and Treatment of Cancer (EORTC) trial Correspondence: Stephen Connolly, Department of Urology, Addenbrooke’s Hospital, Box 43, Hills Road, Cambridge CB2 0QQ, UK. Tel: +44 1223 257261. Fax: +44 1223 216069. E-mail: [email protected]
History Received 11 February 2014 Revised 7 September 2014 Accepted 15 September 2014 Online 10 January 2015
30881 deliberately excluded patients in whom nodes were radiologically enlarged and notably the vast majority (96%) of patients who underwent LND had negative nodes [8]. Given these limitations, it seems reasonable to consider the lymphadenectomy debate with RCC unresolved, and it seems likely that some arguments for the discriminate use of LND will persist. The lack of popularity for LND with RCC is not likely to be attributable to any one single factor, but rather is complex and multifactorial. The single prospective trial of LND with RCC, the EORTC randomized phase 3 trial 30881, reported the frequency of positive nodes to be so low (4%) that routine regional LND cannot be justified [8]. Controversy exists as to whether LND provides improved staging alone or also offers a therapeutic survival advantage [9]. None of the major international guidelines currently promotes therapeutic LND and although the EAU guidelines (2010) state that “resection of affected lymph nodes should be performed”, it is clarified to be purely for the purpose of staging. It has been suggested that patients with lymph-node metastases and RCC fare worse than those with metastatic renal cancer in the absence of retroperitoneal lymphadenopathy [10]. Furthermore, while nephrectomy has proven to be superbly suited to laparoscopy, retroperitoneal LND indisputably represents a much
Lymphadenopathy and renal cancer
DOI: 10.3109/21681805.2014.969307
bigger challenge for the average-skilled laparoscopist [11–13]. As a result, perhaps “subtle” lymphadenopathy frequently remains ignored as the surgeon focuses on removal of the primary tumour with minimal morbidity, whereas overt lymphadenopathy may be considered by some to simply represent surgically incurable disease, even in the absence of other disseminated visceral metastases. Also of considerable relevance is the fact that there is substantial uncertainty regarding the true significance of lymphadenopathy with RCC, and node enlargement not attributable to renal cancer metastasis is frequently explained by “reactive change” including inflammation and follicular hyperplasia. In a frequently cited seminal publication, Studer et al. reported in 1990 that lymphadenopathy with RCC, based purely on size criteria (‡1 cm), was associated with metastasis in only 18 (42%) of 43 patients, a false-positive rate approximating 60% [14]. Others expanded this concern further, with Ming et al., in a demonstration of the utility of frozen section node examination, finding only 36 (31.6%) of 114 patients with enlarged nodes to harbour malignancy, a false-positive rate approaching 70% [15]. Most recently, in a subanalysis of patients with intraoperative lymph-node enlargement in the EORTC trial, resection of enlarged nodes in 84 patients found that 14 contained cancer, producing a false-positive rate in excess of 80% [8]. Currently, renal cancer is most commonly staged by computed tomography (CT), and locoregional lymph-node abnormality is evaluated purely on the basis of size [7]. Radiological accuracy for predicting metastasis within lymph nodes represents a considerable problem but is not one exclusive to renal cancer, nor is it a problem solved by other imaging modalities such as magnetic resonance imaging (MRI). A meta-analysis in the setting of prostate cancer found that CT and MRI demonstrate equally poor performance in the detection of lymph-node metastasis [16]. This study examines the correlation between CT-predicted retroperitoneal lymph-node metastasis with RCC and true pathological involvement of lymph nodes in a contemporary setting of patients undergoing lymphadenectomy at the time of nephrectomy due to lymphadenopathy. Furthermore, by comparison of multiple blinded reviews of preoperative imaging, the subjectivity that may surround CT-reported lymphadenopathy in this setting is addressed.
Materials and methods Study population and histopathological data A retrospective review of the departmental database of 515 consecutive nephrectomy surgeries performed between July 2004 and January 2012 was undertaken in compliance with international ethical standards. All patients with RCC and lymphadenectomy tissue submitted to pathology on the day of surgery were identified. The departmental database routinely includes primary renal tumour parameters such as size, renal vein involvement and the presence of synchronous visceral metastasis. All histopathology information for this study was taken from the formal histopathology reports issued in real time associated with each clinical episode, and
143
all reports were in accordance with the latest version of the Royal College of Pathologists’ histopathology reporting guidelines for renal cancer [2,17]. Of note, all macroscopically negative lymph nodes were routinely submitted for tissue processing and histological examination in their entirety, regardless of size. Lymphadenectomy surgery Operative records were reviewed. During the period under observation, the indication for LND in this unit was either enlarged lymph nodes identified preoperatively at conventional CT (>1 cm in short axis) or found to be enlarged intraoperatively, or suspicion of nodal metastasis based on advanced local stage. Routine LND has never been routinely performed at any time in this institution for “normal” lymph nodes. All patients with incidental retrieval of lymph nodes (e.g. hilar nodes frequently retrieved during laparoscopic nephrectomy) were excluded from analysis in this study. The type of LND performed in this series was not standardized (and cannot be considered universally to be a thorough retroperitoneal LND in each case), but an attempt was made in each case to remove all enlarged nodes. Radiology review For the specific purpose of this study, multiple retrospective reviews of the preoperative (CT) imaging available on the hospital picture archiving and communications system for each patient were undertaken. All preoperative imaging was required to be within 2 months of surgery, and although CT protocols were not uniform, all imaging included contrast enhancement from a multislice scanner. Reviews were undertaken in a blinded manner (to both the official radiology and histopathology reports) and were performed individually by a consultant surgeon with special interest in renal cancer, and also independently by two consultant radiologists with special interest in abdominal imaging. Each reviewer performed an independent review on two separate occasions at least 2 weeks apart. It was agreed by the team involved in this study that considerable subjectivity may exist in the definition of regional lymphadenopathy for patients with RCC. For this study, information provided to the reviewer immediately before review was confined to the fact that the patient had undergone nephrectomy for a renal cancer with removal of lymph nodes. Each reviewer was then asked to predict nodal involvement by choosing one of four options: category 1, “Unequivocal positive (for metastasis)”; 2, “Equivocal node enlargement – probably positive”; 3, “Equivocal node enlargement – probably negative”; or 4, “Unequivocal negative” nodes (with zero radiological concern for metastasis). Importantly, the criteria employed for the prediction of nodal involvement were not purely based on size, but are described in detail in Table 1. No specific weighting was attributed to any individual factor and no specific formula was devised. Notably, this constitutes a completely novel categorization system and has never been previously employed, and therefore has not been validated.
144 S. S. Connolly et al.
Scand J Urol, 2014; 49(2): 142–148
Table 1. Radiological criteria for the prediction of lymph-node metastasis. Radiological node category 1
2
3 4 a
Descriptiona
Implication
Overtly enlarged single or multiple nodes (usually ‡12 mm in short axis dimension) and considered very likely to harbour malignancy. One would reasonably expect this category to be readily commented upon by all reporting radiologists upon review of the same imaging Fulfilling criteria for clear concern based on size (>1 cm in short axis dimension) and appearance, but generally
