Available online at www.sciencedirect.com
ScienceDirect EJSO xx (2014) 1e5
www.ejso.com
The value of pre-operative computed tomography scanning for the assessment of lymph node status in patients with colon cancer F.E.E. de Vries a, D.W. da Costa a, K. van der Mooren b, T.A. van Dorp b, B.C. Vrouenraets a,* a b
Department of Surgery, Sint Lucas Andreas Ziekenhuis, Amsterdam, The Netherlands Department of Radiology, Sint Lucas Andreas Ziekenhuis, Amsterdam, The Netherlands Accepted 13 August 2014 Available online - - -
Abstract Aim: Our aim was to determine the value of a pre-operative computed tomography (CT) scan for the assessment of lymph node status in patients diagnosed with colon cancer by comparing radiological N-stage to histopathological N-stage. Patients and methods: We performed a retrospective cohort study at the Sint Lucas Andreas Hospital in Amsterdam, the Netherlands. Between 2008 and 2010, two radiologists independently reviewed all pre-operative CT scans of patients diagnosed with colon cancer. The scans were examined for signs of regional lymphatic spread (Nþ), defined as lymph nodes exceeding 1 cm, clusters of 3 lymph nodes or a combination of the two. The results were compared with the histopathological N-stage. Inter-observer agreement, positive predictive value (PVV), negative predictive value (NPV), sensitivity, specificity, and accuracy were calculated. Results: We included 106 patients in our study. PVV, NPV, sensitivity, specificity, and accuracy of detecting regional lymph nodes metastases were 47%, 66%, 71%, 41% and 54%, respectively. Inter-observer agreement was 74.5% (l ¼ 0.48). Conclusion: Although our study group was relatively large and newer techniques were used in comparison to previous studies, our results demonstrated that the value of a pre-operative CT scan for the assessment of regional lymph nodes remained poor and unreliable. Therefore we question if a radiologist should assess regional lymph nodes on a pre-operative CT scan in colon cancer. Before treatment decisions are made on the appearance of lymph nodes in colon cancer patients, its diagnostic accuracy needs strong improvement. Ó 2014 Published by Elsevier Ltd.
Keywords: Colonic neoplasms; Lymph nodes; Neoplasm metastasis; Computed tomographic; Staging; Accuracy
Introduction Annually, more than 10,000 patients are diagnosed with colon cancer in the Netherlands. In men colorectal cancer has the third highest incidence after prostate- and lung cancer. For women, the incidence of colorectal cancer is even the second highest incidence after breast cancer.1 The preoperative work-up of colon cancer patients includes computed tomography (CT) scanning for the presence of distant metastases (especially liver metastases) and
* Corresponding author. Department of Surgery Sint Lucas Andreas Ziekenhuis. Jan Tooropstraat 164, 1061 AE Amsterdam, The Netherlands. Tel.: þ31 20 510 89 11. E-mail address: [email protected] (B.C. Vrouenraets).
assessment of the primary tumour. Theoretically, the preoperative CT scan can be used to predict lymph node involvement. Previous studies reported sensitivity for detecting malignant regional lymph nodes ranging from 13% to 92%.2e11 A meta-analysis of 19 studies of colorectal tumours in 2010 by Dighe et al.3 revealed relatively good results for assessment of tumour invasion grade (Tstage) but poor sensitivity, specificity, and diagnostic odds ratios for assessing nodal status (N-stage). Inter-observer agreement regarding the N-stage was poor. A systematic review by Leufkens et al.12 found a sample-size-weighted sensitivity, specificity, and accuracy of 76%, 55%, and 69%, respectively and considered CT scanning for N-status reasonable. This finding is in contrast with rectal cancer where CT scanning proves to be more accurate.13 Most
http://dx.doi.org/10.1016/j.ejso.2014.08.483 0748-7983/Ó 2014 Published by Elsevier Ltd. Please cite this article in press as: de Vries FEE, et al., The value of pre-operative computed tomography scanning for the assessment of lymph node status in patients with colon cancer, Eur J Surg Oncol (2014), http://dx.doi.org/10.1016/j.ejso.2014.08.483
2
F.E.E. de Vries et al. / EJSO xx (2014) 1e5
studies concerning this topic, however, are more than 10year-old and involve relatively small patient groups.2e11 Newer CT techniques and larger study populations might change these results. Our aim was to determine the value of the pre-operative CT scan in colon cancer patients for predicting the N-stage in terms of inter-observer agreement, positive predictive value (PVV), negative predictive value (NPV), sensitivity, specificity, and accuracy. Put differently, can CT scanning predict the histopathologically N-stage reliably?
Statistical analysis
Patients and methods
Study population
Retrospective cohort study
Between 2008 and 2010, 211 patients underwent a resection of a colorectal tumour in our hospital. Out of these 211 patients, after excluding those with rectal cancer (N ¼ 68), those who did not have a pre-operative CT scan (N ¼ 30), and those who had neo-adjuvant therapy (N ¼ 7), 106 patients remained who met our inclusion criteria. Of these, 56 (52.8%) were female and the median age was 70 years (ranges 29e87 years). Two patients did not receive intravenous contrast and nine patients did not receive oral contrast. Another two patients had not received contrast at all. In Table 1 we summarise patient and tumour characteristics.
This study was performed at the Departments of Surgery and Radiology of the Sint Lucas Andreas Hospital in Amsterdam, the Netherlands. Between 2008 and 2010, 211 patients underwent resectional curative surgery for colorectal cancer. The exclusion criteria for our study were rectal cancer, neo-adjuvant radiotherapy and/or chemotherapy and patients who were not scanned pre-operatively. Subsequently, two radiologists (both more than 15 years experience in reading abdominal CT scans) independently reviewed all the pre-operative CT scans. Apart from being told the endoscopic location of the primary tumour, the radiologists received no other information about the patients. The radiologists each filled in a case record form: the presence of either regional lymph nodes of >1 cm and/or clusters of 3 lymph nodes was scored as Nþ while the absence of enlarged or clustered lymph nodes was scored as N0, being the definition for a radiological positive nodal status in most previous studies.4e10,13,14 In addition, they recorded information about the use of either oral or intravenous contrast. In case of disagreement between the two radiologists, they discussed the case together until reaching agreement prior to statistical analysis. The histopathological analyses of the resected colon specimen were used as the reference standard. Identification of the lymph nodes was done thorough inspection, palpation and dissection of the specimen without any fat clearing techniques. Lymph nodes were found positive (Nþ) in case the pathologist found metastatic tumour in one or more lymph nodes. Distinction between pN1 and pN2 (for subgroup analysis) was done conform the 6th edition of TNM classification system. CT scanning CT scanning was performed with a 4-slice Toshiba Aquilion CT scanner (FOV 24 cm, slice thickness 0.5 mm, reconstruction matrix 512, pixel size 0.47, ideal voxel size 0.103, real voxel size 0.110, deviation 6.7%). All CT scans were viewed on 3 mm axial sliced images. Maximum short axis in the axial plane was measured. No reconstructions were done.
Statistical analysis was performed with SPSS. True positive (TP), true negative (TN), false positive (FP), false negative (FN), inter-observer agreement, positive predictive value (PVV), negative predictive value (NPV), sensitivity, specificity, and accuracy were calculated after the two radiologists had reached agreement on all cases. Results
Tumour After histopathological confirmation, one tumour was staged as T1, 14 tumours were staged as T2, 83 as T3, and eight as T4. Most tumours were located in sigmoid colon (39.6%). Out of the 106 patients, 10 had distant metastasis at the time of presentation. Table 1 Baseline characteristics. No. of patients (N ¼ 106) Female Age (median) Location Coecum Ascending colon Transverse colon Descending colon Sigmoid T-stage T1 T2 T3 T4 N-stage N0 N1 or N2 (¼ Nþ) N1 N2 M stage M0 M1
56 (52.8%) 70.5 (29e87) 23 21 11 9 42
(21.7%) (19.8%) (10.4%) (8.5%) (39.6%)
1 14 83 8
(0.9%) (13.2%) (78.3%) (7.5%)
61 45 29 16
(57.5%) (42.5%) (27.4%) (15.1%)
96 (90.6%) 10 (9.4%)
Please cite this article in press as: de Vries FEE, et al., The value of pre-operative computed tomography scanning for the assessment of lymph node status in patients with colon cancer, Eur J Surg Oncol (2014), http://dx.doi.org/10.1016/j.ejso.2014.08.483
F.E.E. de Vries et al. / EJSO xx (2014) 1e5
difference between the individual performances of the two radiologists.
Table 2 True positives, true negatives, false positives and false negatives.
Nþ (CT scan) Ne (CT scan)
3
pNþ
pNe
32 13
36 25
Discussion
We found 32 true positives (TP), 25 true negatives (TN), 36 false positives (FP), and 13 false negatives (FN) (Table 2). PPV, NPV, sensitivity, specificity, and accuracy for detecting malignant lymph nodes were 47% (95% CI 41e53%), 66% (95% CI 49e80%), 71% (95% CI 62e80%), 41% (95% CI 32e50%), and 54% (95% CI 49e59%), respectively. We performed a subgroup analysis in patients who did receive complete contrast preparation (95 patients). We found a PPV, NPV, sensitivity, specificity and accuracy of 45% (95% CI 32e58%), 77% (95% CI 59e90%), 80% (95% CI 63e92%), 41% (95% CI 29e55%) and 56% (95% CI 46e66%) respectively (Table 3). Subsequently, we did a subgroup analysis to reveal whether there was any difference between radiological distinction between histopathological stage N1 versus N0 and N2 versus N0 disease (Table 3). For the distinction between N1eN0 disease, PPV, NPV, sensitivity, specificity and accuracy were 36% (95% CI 23e50%), 74% (95% CI 56e87%), 69% (95% CI 49e85%), 41% (95% CI 29e54%) and 50% (95% CI 40e60%) respectively. For N2 versus N0 disease, we found a PPV, NPV, sensitivity, specificity and accuracy of 25% (95% CI 14e40%), 86% (95% CI 68e96%), 75% (95% CI 48e93%), 41% (95% CI 29e54%) and 48% (95% CI 37e59%) respectively.
Our study demonstrated that despite newer techniques and a considerable study group the value of CT scanning for the assessment of lymph nodes in patients with colon cancer remained poor with a diagnostic accuracy of only 54% for N-status. Improved conditions failed to improve the accuracy compared to studies in the past (Table 4). To the best of our knowledge, after the study by Smith et al.14 ours was the second largest study on this topic. The reliability of our study was increased considerably because all cases had histopathological confirmation (gold standard). This was in contrast to some previous studies, in which no histopathological confirmation was undertaken or it was not reported.12,13 Apart from knowing the primary endoscopic location of the tumour our two radiologists were blinded and all the scans were seen by each radiologist separately. We found a moderate inter-observer agreement of 74.5% (l ¼ 0.48). We selected our patients carefully and excluded rectal tumours because better results have been reported for rectal tumours.13 All scans had 3 mm slices, whereas previous studies had mostly 5, 8, or even 10 mm section thickness.2,4e9,11e14 In a subgroup analysis for nodal detection, the meta-analysis of Dighe et al.3 showed that a slice thickness of 5 mm or less has significantly better results than a slice thickness of more than 5 mm. Understaging colonic tumours is a risk if the surgical nodal harvest is inadequate.3 Previous studies did not mention the number of histopathological retrieved lymph nodes. We found a median of 13 lymph nodes. There are several limitations and difficulties in the assessment of lymph nodes with CT scanning. The criteria for defining lymph nodes as metastatic on a CT scan are inconsistent. The definition of positive lymph nodes varies from lymph nodes of >5 mm2, lymph nodes of >8 mm,13 lymph nodes of >1 cm,4e10,14 a cluster of three or more
Inter-observer agreement
Table 4 Previous studies on the subject.
Lymph nodes The median histopathological lymph node count was 13 (range 2e47) and 45 patients were found to have malignant spread in lymph nodes. Accuracy of CT staging for lymph nodes metastases
Out of the 106 CT scans, 27 needed to be discussed by the radiologists due to initial disagreement. Inter-observer agreement was moderate (74.5%, l ¼ 0.48). We found no Table 3 Subgroup analysis of patients excluded without contrast, N1eN0 en N2eN0.
PPV NPV Sensitivity Specificity Accuracy
Total (N ¼ 106)
Exclusion without contrast (N ¼ 93)
N1eN0 (N ¼ 90)
N2eN0 (N ¼ 70)
47% 66% 71% 41% 54%
45% 77% 80% 41% 56%
36% 74% 69% 41% 50%
25% 86% 75% 41% 48%
Author
Year
No. of patients
Accuracy
Sensitivity
Specificity
Dighe et al.2 Balthazar et al.4 Acunas¸ et al.5 Gazelle et al.6 Harvey et al.7 Gomille et al.8 Cademarati et al.9 Fillipone et al.10 Chamadol et al.11 Smith et al.14 De Vries et al. Dighe et al.*3 Leufkens et al.*12
2010 1988 1990 1995 1998 1998 2002 2004 2005 2007 2014 2010 2011
84 76 28 25 37 86 60 41 36 126 106 907 156
58%
64% 73% 75% 90% 55% 71% 55% 86% 92%
53% 58% 75% 85% 98% 57% 65% 75% 55%
71% 70% 55%
41% 78% 69%
62% 54% 69%
* ¼ review or meta-analysis.
Please cite this article in press as: de Vries FEE, et al., The value of pre-operative computed tomography scanning for the assessment of lymph node status in patients with colon cancer, Eur J Surg Oncol (2014), http://dx.doi.org/10.1016/j.ejso.2014.08.483
4
F.E.E. de Vries et al. / EJSO xx (2014) 1e5
nodes,4e7,9,10,13,14 enlarged lymph nodes with irregular borders,2 size not specified at all,8,11,12 combinations of these, to the criteria we used. Nevertheless, most studies use the definition we have used for our study.4e9 The existence of benign but enlarged lymph nodes due to inflammation is problematic as are micrometastasis in relatively small (non-clustered) lymph nodes. Hardly any literature is available addressing the correlation between lymph node size and malignancy of these lymph nodes. A study of Brown et al.15 concluded that regardless of cut-off, also the overall predictive value of magnetic resonance (MR) size is poor because of substantial overlap in size between benign and malignant nodes. Although their study was performed with MR instead of CT we think these results will probably be the same for CT. When assessing the lymph nodes, we only looked at the axial view for the criteria of a 1 cm or larger lymph node. The possibility exists that some lymph nodes might appear to be less than 1 cm in the axial view whereas they were more than 1 cm in the coronal or sagittal view. Kanamoto et al.16 reported better sensitivity, specificity, and accuracy with two-dimensional or even three-dimensional reconstructions, but these techniques are more time-consuming. In patients with little intra-abdominal fat, it is more difficult to distinguish between an invading tumour in pericolic fat or local lymph nodes. Also, acute patients with an obstructing tumour are sometimes difficult to assess because of the distended colon, which reduces view. A few of our patients did not receive either oral and/or intravenous contrast. These scans were probably also more challenging to assess, although a subgroup analysis of Dighe et al.3 did not find evidence that the use of oral contrast improved sensitivity and specificity. Also when 11 patients who did not receive either oral and/or intravenous contrast were excluded, we found an accuracy of 56% being comparable to the accuracy of the total group (54%) (Table 3). Neither could CT scanning accurately predict more extensive nodal disease (pN2) since accuracy dropped from 50% in N1eN0 to even 48% in N2 versus N0 disease (Table 3). We found that both radiologists identified a remarkable number of false positive lymph nodes. When we discussed the CT scans on which they had disagreed initially, we found that almost all discussion concerned the clusters of three or more lymph nodes. Both radiologists agreed that they were biased towards looking for these lymph nodes because of the definition. If they had found one or two small lymph nodes they started looking intensively for a third small one. This meant that probably a considerable number of CT scans were Nþ whereas they would probably not have been reported as Nþ initially. This might also explain the moderate inter-observer agreement as there was only one scan were they did not agree about lymph nodes larger than 1 cm. Another bias might exist in patients with extended liver metastasis or a large tumour, since one would expect positive lymph nodes in these patients more often than in patients who had a small tumour or patients without distant metastasis.
Recently, a discussion started about the use of neoadjuvant chemotherapy for colon cancer. A study of Nørgaard et al., in 201317 concluded that CT has a potential in the pre-operative selection of advanced tumours suitable for neo-adjuvant chemotherapy without overtreatment of low-risk patients. However, they found an accuracy of 53% for predicting N-stage. If we take a look at our results, we found a comparable accuracy but almost a third of our study group was found false positive. Next to this, almost 10% has been found false negative. Therefore, there is a great risk of overtreatment when choosing for neoadjuvant chemotherapy based solely on a pre-operative CT scan. We think that if future studies will result in different treatment options (for example neo-adjuvant chemotherapy) based on lymph node status on a CT scan, there needs to be improvement in universal definition and accuracy. There is hardly any literature available about the relevance of pre-operative lymph node status in relation to prognosis of the patient. A study of Dighe et al.2 in 2010 concluded that CT staging of local node status should not be used to assess the prognosis before operation. The same concluded Smith et al.14 in 2007 in a study were they tried to predict ‘good’ and ‘poor’ prognosis based on the T- and N-status on a pre-operative CT scan. They demonstrated that pre-operative CT scan predict clinical outcome for ‘good’ and ‘poor’ prognosis tumours with the same accuracy as histopathology for T-status, but found CT not particularly useful in the assessment of lymph nodes. We also found a remarkable number of false positives and false negatives and doubt if CT findings are reliable to predict patient prognosis. In conclusion, although we used improved techniques and the study group was larger compared to most previous studies, the value of CT scanning for the assessment of lymph nodes in patients with a colon cancer seems to remain poor. Prospective studies with two-dimensional and three-dimensional lymph node viewing, thinner slices, and large study populations are necessary to improve these results. In addition, a better and universal definition of radiological positive lymph nodes may help improve studies addressing the correlation between size and malignancy. In the meantime, we question whether radiologists should assess and report lymph node status in preoperative CT scans in patients with colon cancer. Conflict of interest The authors report no conflicts of interest.
References 1. Lemmens V. Clinical epidemiology of colorectal cancer in the Netherlands. Studies of variation and trends with the Eindhoven Cancer Registry 2007.
Please cite this article in press as: de Vries FEE, et al., The value of pre-operative computed tomography scanning for the assessment of lymph node status in patients with colon cancer, Eur J Surg Oncol (2014), http://dx.doi.org/10.1016/j.ejso.2014.08.483
F.E.E. de Vries et al. / EJSO xx (2014) 1e5 2. Dighe S, Blake H, Koh M, et al. Accuracy of multidetector computed tomography in identifying poor prognostic factors in colonic cancer. Br J Surg 2010;97:1407–15. 3. Dighe S, Purkayastha S, Swift I, et al. Diagnostic precision of CT in local staging of colon cancers: a meta-analysis. Clin Radiol 2010; 65:708–19. 4. Balthazar E, Megibow A, Hulnick D, Naidich D. Carcinoma of the colon: detection and preoperative staging by CT. AJR 1988 Feb;150: 301–6. 5. Acunas¸ B, Rozanes I, Acunas¸ G, C ¸ elik L, Sayi I, G€okmen E. Preoperative CT staging of colon carcinoma (excluding the recto-sigmoid region). Eur J Radiol 1990;11:150–3. 6. Gazelle S, Gaa J, Saini S, Shellito P. J Comput Assist Tomogr 1995 Jan/Feb;19(1):87–91. 7. Harvey C, Amin Z, Hare C, et al. Helical CT pneumocolon to assess colonic tumors: radiologic e pathologic correlation. AJR 1998 June1439–43. 8. Gomille T, Aleksic M, Ulrich B, Christ F. Significance of CT in the detection of regional lymph node metastases in colorectal carcinoma. Radiologe 1998;38:1077–82. 9. Cademartiri F, Luccichenti G, Rossi A, Pavone P. Spiral hydro-CT in the evaluation of colo-sigmoideal cancer. Radiol Med 2002;104:295– 306. 10. Filippone A, Ambrosini R, Fuschi M, Marinelli T, Genovesi D, Bonomo L. Preoperative T and N staging of colorectal cancer: accuracy of contrast-enhanced multi-detector row CT colonography e initial experience. Radiology 2004;231:83–90.
5
11. Chamadol N, Ninpiethoon T, Bhudhisawasd V, Pairojkul C. The role of CT scan in preoperative staging of colorectal carcinoma. J Med Assoc Thai 2005;88:1847–53. 12. Leufkens A, van den Bosch M, van Leeuwen M, Siersema P. Diagnostic accuracy of computed tomography for colon cancer staging: a systematic review. Scand J Gastroenterol 2011;46:887–94. 13. Chi Y, Zhang X, Li J, Sun Y. To be or not to be: significance of lymph nodes on pretreatment CT in predicting survival of rectal cancer patients. Eur J Radiol 2011;77:473–7. 14. Smith N, Bees N, Barnachano Y, Norman A, Swift R, Brown G. Preoperative computed tomography staging of nonmetastatic colon cancer predicts outcome: implications for clinical trials. Br J Cancer 2007; 96:1030–6. 15. Brown G, Richards C, Bourne M, et al. Morphologic predictors of lymph node status in rectal cancer with use of high-spatialresolution MR imaging with histopathologic comparison. Radiology 2003;227:371–7. 16. Kanamoto T, Matsuki M, Okuda J, et al. Preoperative evaluation of local invasion and metastatic lymph nodes of colorectal cancer and mesenteric vascular variations using multidetector-row computed tomography before laparoscopic surgery. J Comput Assist Tomogr 2007;31:831–9. 17. Nørgaard A, Dam C, Jakobsen A, Pløen J, Lindebjerg J, Rafaelsen SR. Selection of colon cancer patients for neoadjuvant chemotherapy by preoperative CT scan. Scand J Gastroenterol 2014 Feb;49(2):202–8.
Please cite this article in press as: de Vries FEE, et al., The value of pre-operative computed tomography scanning for the assessment of lymph node status in patients with colon cancer, Eur J Surg Oncol (2014), http://dx.doi.org/10.1016/j.ejso.2014.08.483
ScienceDirect EJSO xx (2014) 1e5
www.ejso.com
The value of pre-operative computed tomography scanning for the assessment of lymph node status in patients with colon cancer F.E.E. de Vries a, D.W. da Costa a, K. van der Mooren b, T.A. van Dorp b, B.C. Vrouenraets a,* a b
Department of Surgery, Sint Lucas Andreas Ziekenhuis, Amsterdam, The Netherlands Department of Radiology, Sint Lucas Andreas Ziekenhuis, Amsterdam, The Netherlands Accepted 13 August 2014 Available online - - -
Abstract Aim: Our aim was to determine the value of a pre-operative computed tomography (CT) scan for the assessment of lymph node status in patients diagnosed with colon cancer by comparing radiological N-stage to histopathological N-stage. Patients and methods: We performed a retrospective cohort study at the Sint Lucas Andreas Hospital in Amsterdam, the Netherlands. Between 2008 and 2010, two radiologists independently reviewed all pre-operative CT scans of patients diagnosed with colon cancer. The scans were examined for signs of regional lymphatic spread (Nþ), defined as lymph nodes exceeding 1 cm, clusters of 3 lymph nodes or a combination of the two. The results were compared with the histopathological N-stage. Inter-observer agreement, positive predictive value (PVV), negative predictive value (NPV), sensitivity, specificity, and accuracy were calculated. Results: We included 106 patients in our study. PVV, NPV, sensitivity, specificity, and accuracy of detecting regional lymph nodes metastases were 47%, 66%, 71%, 41% and 54%, respectively. Inter-observer agreement was 74.5% (l ¼ 0.48). Conclusion: Although our study group was relatively large and newer techniques were used in comparison to previous studies, our results demonstrated that the value of a pre-operative CT scan for the assessment of regional lymph nodes remained poor and unreliable. Therefore we question if a radiologist should assess regional lymph nodes on a pre-operative CT scan in colon cancer. Before treatment decisions are made on the appearance of lymph nodes in colon cancer patients, its diagnostic accuracy needs strong improvement. Ó 2014 Published by Elsevier Ltd.
Keywords: Colonic neoplasms; Lymph nodes; Neoplasm metastasis; Computed tomographic; Staging; Accuracy
Introduction Annually, more than 10,000 patients are diagnosed with colon cancer in the Netherlands. In men colorectal cancer has the third highest incidence after prostate- and lung cancer. For women, the incidence of colorectal cancer is even the second highest incidence after breast cancer.1 The preoperative work-up of colon cancer patients includes computed tomography (CT) scanning for the presence of distant metastases (especially liver metastases) and
* Corresponding author. Department of Surgery Sint Lucas Andreas Ziekenhuis. Jan Tooropstraat 164, 1061 AE Amsterdam, The Netherlands. Tel.: þ31 20 510 89 11. E-mail address: [email protected] (B.C. Vrouenraets).
assessment of the primary tumour. Theoretically, the preoperative CT scan can be used to predict lymph node involvement. Previous studies reported sensitivity for detecting malignant regional lymph nodes ranging from 13% to 92%.2e11 A meta-analysis of 19 studies of colorectal tumours in 2010 by Dighe et al.3 revealed relatively good results for assessment of tumour invasion grade (Tstage) but poor sensitivity, specificity, and diagnostic odds ratios for assessing nodal status (N-stage). Inter-observer agreement regarding the N-stage was poor. A systematic review by Leufkens et al.12 found a sample-size-weighted sensitivity, specificity, and accuracy of 76%, 55%, and 69%, respectively and considered CT scanning for N-status reasonable. This finding is in contrast with rectal cancer where CT scanning proves to be more accurate.13 Most
http://dx.doi.org/10.1016/j.ejso.2014.08.483 0748-7983/Ó 2014 Published by Elsevier Ltd. Please cite this article in press as: de Vries FEE, et al., The value of pre-operative computed tomography scanning for the assessment of lymph node status in patients with colon cancer, Eur J Surg Oncol (2014), http://dx.doi.org/10.1016/j.ejso.2014.08.483
2
F.E.E. de Vries et al. / EJSO xx (2014) 1e5
studies concerning this topic, however, are more than 10year-old and involve relatively small patient groups.2e11 Newer CT techniques and larger study populations might change these results. Our aim was to determine the value of the pre-operative CT scan in colon cancer patients for predicting the N-stage in terms of inter-observer agreement, positive predictive value (PVV), negative predictive value (NPV), sensitivity, specificity, and accuracy. Put differently, can CT scanning predict the histopathologically N-stage reliably?
Statistical analysis
Patients and methods
Study population
Retrospective cohort study
Between 2008 and 2010, 211 patients underwent a resection of a colorectal tumour in our hospital. Out of these 211 patients, after excluding those with rectal cancer (N ¼ 68), those who did not have a pre-operative CT scan (N ¼ 30), and those who had neo-adjuvant therapy (N ¼ 7), 106 patients remained who met our inclusion criteria. Of these, 56 (52.8%) were female and the median age was 70 years (ranges 29e87 years). Two patients did not receive intravenous contrast and nine patients did not receive oral contrast. Another two patients had not received contrast at all. In Table 1 we summarise patient and tumour characteristics.
This study was performed at the Departments of Surgery and Radiology of the Sint Lucas Andreas Hospital in Amsterdam, the Netherlands. Between 2008 and 2010, 211 patients underwent resectional curative surgery for colorectal cancer. The exclusion criteria for our study were rectal cancer, neo-adjuvant radiotherapy and/or chemotherapy and patients who were not scanned pre-operatively. Subsequently, two radiologists (both more than 15 years experience in reading abdominal CT scans) independently reviewed all the pre-operative CT scans. Apart from being told the endoscopic location of the primary tumour, the radiologists received no other information about the patients. The radiologists each filled in a case record form: the presence of either regional lymph nodes of >1 cm and/or clusters of 3 lymph nodes was scored as Nþ while the absence of enlarged or clustered lymph nodes was scored as N0, being the definition for a radiological positive nodal status in most previous studies.4e10,13,14 In addition, they recorded information about the use of either oral or intravenous contrast. In case of disagreement between the two radiologists, they discussed the case together until reaching agreement prior to statistical analysis. The histopathological analyses of the resected colon specimen were used as the reference standard. Identification of the lymph nodes was done thorough inspection, palpation and dissection of the specimen without any fat clearing techniques. Lymph nodes were found positive (Nþ) in case the pathologist found metastatic tumour in one or more lymph nodes. Distinction between pN1 and pN2 (for subgroup analysis) was done conform the 6th edition of TNM classification system. CT scanning CT scanning was performed with a 4-slice Toshiba Aquilion CT scanner (FOV 24 cm, slice thickness 0.5 mm, reconstruction matrix 512, pixel size 0.47, ideal voxel size 0.103, real voxel size 0.110, deviation 6.7%). All CT scans were viewed on 3 mm axial sliced images. Maximum short axis in the axial plane was measured. No reconstructions were done.
Statistical analysis was performed with SPSS. True positive (TP), true negative (TN), false positive (FP), false negative (FN), inter-observer agreement, positive predictive value (PVV), negative predictive value (NPV), sensitivity, specificity, and accuracy were calculated after the two radiologists had reached agreement on all cases. Results
Tumour After histopathological confirmation, one tumour was staged as T1, 14 tumours were staged as T2, 83 as T3, and eight as T4. Most tumours were located in sigmoid colon (39.6%). Out of the 106 patients, 10 had distant metastasis at the time of presentation. Table 1 Baseline characteristics. No. of patients (N ¼ 106) Female Age (median) Location Coecum Ascending colon Transverse colon Descending colon Sigmoid T-stage T1 T2 T3 T4 N-stage N0 N1 or N2 (¼ Nþ) N1 N2 M stage M0 M1
56 (52.8%) 70.5 (29e87) 23 21 11 9 42
(21.7%) (19.8%) (10.4%) (8.5%) (39.6%)
1 14 83 8
(0.9%) (13.2%) (78.3%) (7.5%)
61 45 29 16
(57.5%) (42.5%) (27.4%) (15.1%)
96 (90.6%) 10 (9.4%)
Please cite this article in press as: de Vries FEE, et al., The value of pre-operative computed tomography scanning for the assessment of lymph node status in patients with colon cancer, Eur J Surg Oncol (2014), http://dx.doi.org/10.1016/j.ejso.2014.08.483
F.E.E. de Vries et al. / EJSO xx (2014) 1e5
difference between the individual performances of the two radiologists.
Table 2 True positives, true negatives, false positives and false negatives.
Nþ (CT scan) Ne (CT scan)
3
pNþ
pNe
32 13
36 25
Discussion
We found 32 true positives (TP), 25 true negatives (TN), 36 false positives (FP), and 13 false negatives (FN) (Table 2). PPV, NPV, sensitivity, specificity, and accuracy for detecting malignant lymph nodes were 47% (95% CI 41e53%), 66% (95% CI 49e80%), 71% (95% CI 62e80%), 41% (95% CI 32e50%), and 54% (95% CI 49e59%), respectively. We performed a subgroup analysis in patients who did receive complete contrast preparation (95 patients). We found a PPV, NPV, sensitivity, specificity and accuracy of 45% (95% CI 32e58%), 77% (95% CI 59e90%), 80% (95% CI 63e92%), 41% (95% CI 29e55%) and 56% (95% CI 46e66%) respectively (Table 3). Subsequently, we did a subgroup analysis to reveal whether there was any difference between radiological distinction between histopathological stage N1 versus N0 and N2 versus N0 disease (Table 3). For the distinction between N1eN0 disease, PPV, NPV, sensitivity, specificity and accuracy were 36% (95% CI 23e50%), 74% (95% CI 56e87%), 69% (95% CI 49e85%), 41% (95% CI 29e54%) and 50% (95% CI 40e60%) respectively. For N2 versus N0 disease, we found a PPV, NPV, sensitivity, specificity and accuracy of 25% (95% CI 14e40%), 86% (95% CI 68e96%), 75% (95% CI 48e93%), 41% (95% CI 29e54%) and 48% (95% CI 37e59%) respectively.
Our study demonstrated that despite newer techniques and a considerable study group the value of CT scanning for the assessment of lymph nodes in patients with colon cancer remained poor with a diagnostic accuracy of only 54% for N-status. Improved conditions failed to improve the accuracy compared to studies in the past (Table 4). To the best of our knowledge, after the study by Smith et al.14 ours was the second largest study on this topic. The reliability of our study was increased considerably because all cases had histopathological confirmation (gold standard). This was in contrast to some previous studies, in which no histopathological confirmation was undertaken or it was not reported.12,13 Apart from knowing the primary endoscopic location of the tumour our two radiologists were blinded and all the scans were seen by each radiologist separately. We found a moderate inter-observer agreement of 74.5% (l ¼ 0.48). We selected our patients carefully and excluded rectal tumours because better results have been reported for rectal tumours.13 All scans had 3 mm slices, whereas previous studies had mostly 5, 8, or even 10 mm section thickness.2,4e9,11e14 In a subgroup analysis for nodal detection, the meta-analysis of Dighe et al.3 showed that a slice thickness of 5 mm or less has significantly better results than a slice thickness of more than 5 mm. Understaging colonic tumours is a risk if the surgical nodal harvest is inadequate.3 Previous studies did not mention the number of histopathological retrieved lymph nodes. We found a median of 13 lymph nodes. There are several limitations and difficulties in the assessment of lymph nodes with CT scanning. The criteria for defining lymph nodes as metastatic on a CT scan are inconsistent. The definition of positive lymph nodes varies from lymph nodes of >5 mm2, lymph nodes of >8 mm,13 lymph nodes of >1 cm,4e10,14 a cluster of three or more
Inter-observer agreement
Table 4 Previous studies on the subject.
Lymph nodes The median histopathological lymph node count was 13 (range 2e47) and 45 patients were found to have malignant spread in lymph nodes. Accuracy of CT staging for lymph nodes metastases
Out of the 106 CT scans, 27 needed to be discussed by the radiologists due to initial disagreement. Inter-observer agreement was moderate (74.5%, l ¼ 0.48). We found no Table 3 Subgroup analysis of patients excluded without contrast, N1eN0 en N2eN0.
PPV NPV Sensitivity Specificity Accuracy
Total (N ¼ 106)
Exclusion without contrast (N ¼ 93)
N1eN0 (N ¼ 90)
N2eN0 (N ¼ 70)
47% 66% 71% 41% 54%
45% 77% 80% 41% 56%
36% 74% 69% 41% 50%
25% 86% 75% 41% 48%
Author
Year
No. of patients
Accuracy
Sensitivity
Specificity
Dighe et al.2 Balthazar et al.4 Acunas¸ et al.5 Gazelle et al.6 Harvey et al.7 Gomille et al.8 Cademarati et al.9 Fillipone et al.10 Chamadol et al.11 Smith et al.14 De Vries et al. Dighe et al.*3 Leufkens et al.*12
2010 1988 1990 1995 1998 1998 2002 2004 2005 2007 2014 2010 2011
84 76 28 25 37 86 60 41 36 126 106 907 156
58%
64% 73% 75% 90% 55% 71% 55% 86% 92%
53% 58% 75% 85% 98% 57% 65% 75% 55%
71% 70% 55%
41% 78% 69%
62% 54% 69%
* ¼ review or meta-analysis.
Please cite this article in press as: de Vries FEE, et al., The value of pre-operative computed tomography scanning for the assessment of lymph node status in patients with colon cancer, Eur J Surg Oncol (2014), http://dx.doi.org/10.1016/j.ejso.2014.08.483
4
F.E.E. de Vries et al. / EJSO xx (2014) 1e5
nodes,4e7,9,10,13,14 enlarged lymph nodes with irregular borders,2 size not specified at all,8,11,12 combinations of these, to the criteria we used. Nevertheless, most studies use the definition we have used for our study.4e9 The existence of benign but enlarged lymph nodes due to inflammation is problematic as are micrometastasis in relatively small (non-clustered) lymph nodes. Hardly any literature is available addressing the correlation between lymph node size and malignancy of these lymph nodes. A study of Brown et al.15 concluded that regardless of cut-off, also the overall predictive value of magnetic resonance (MR) size is poor because of substantial overlap in size between benign and malignant nodes. Although their study was performed with MR instead of CT we think these results will probably be the same for CT. When assessing the lymph nodes, we only looked at the axial view for the criteria of a 1 cm or larger lymph node. The possibility exists that some lymph nodes might appear to be less than 1 cm in the axial view whereas they were more than 1 cm in the coronal or sagittal view. Kanamoto et al.16 reported better sensitivity, specificity, and accuracy with two-dimensional or even three-dimensional reconstructions, but these techniques are more time-consuming. In patients with little intra-abdominal fat, it is more difficult to distinguish between an invading tumour in pericolic fat or local lymph nodes. Also, acute patients with an obstructing tumour are sometimes difficult to assess because of the distended colon, which reduces view. A few of our patients did not receive either oral and/or intravenous contrast. These scans were probably also more challenging to assess, although a subgroup analysis of Dighe et al.3 did not find evidence that the use of oral contrast improved sensitivity and specificity. Also when 11 patients who did not receive either oral and/or intravenous contrast were excluded, we found an accuracy of 56% being comparable to the accuracy of the total group (54%) (Table 3). Neither could CT scanning accurately predict more extensive nodal disease (pN2) since accuracy dropped from 50% in N1eN0 to even 48% in N2 versus N0 disease (Table 3). We found that both radiologists identified a remarkable number of false positive lymph nodes. When we discussed the CT scans on which they had disagreed initially, we found that almost all discussion concerned the clusters of three or more lymph nodes. Both radiologists agreed that they were biased towards looking for these lymph nodes because of the definition. If they had found one or two small lymph nodes they started looking intensively for a third small one. This meant that probably a considerable number of CT scans were Nþ whereas they would probably not have been reported as Nþ initially. This might also explain the moderate inter-observer agreement as there was only one scan were they did not agree about lymph nodes larger than 1 cm. Another bias might exist in patients with extended liver metastasis or a large tumour, since one would expect positive lymph nodes in these patients more often than in patients who had a small tumour or patients without distant metastasis.
Recently, a discussion started about the use of neoadjuvant chemotherapy for colon cancer. A study of Nørgaard et al., in 201317 concluded that CT has a potential in the pre-operative selection of advanced tumours suitable for neo-adjuvant chemotherapy without overtreatment of low-risk patients. However, they found an accuracy of 53% for predicting N-stage. If we take a look at our results, we found a comparable accuracy but almost a third of our study group was found false positive. Next to this, almost 10% has been found false negative. Therefore, there is a great risk of overtreatment when choosing for neoadjuvant chemotherapy based solely on a pre-operative CT scan. We think that if future studies will result in different treatment options (for example neo-adjuvant chemotherapy) based on lymph node status on a CT scan, there needs to be improvement in universal definition and accuracy. There is hardly any literature available about the relevance of pre-operative lymph node status in relation to prognosis of the patient. A study of Dighe et al.2 in 2010 concluded that CT staging of local node status should not be used to assess the prognosis before operation. The same concluded Smith et al.14 in 2007 in a study were they tried to predict ‘good’ and ‘poor’ prognosis based on the T- and N-status on a pre-operative CT scan. They demonstrated that pre-operative CT scan predict clinical outcome for ‘good’ and ‘poor’ prognosis tumours with the same accuracy as histopathology for T-status, but found CT not particularly useful in the assessment of lymph nodes. We also found a remarkable number of false positives and false negatives and doubt if CT findings are reliable to predict patient prognosis. In conclusion, although we used improved techniques and the study group was larger compared to most previous studies, the value of CT scanning for the assessment of lymph nodes in patients with a colon cancer seems to remain poor. Prospective studies with two-dimensional and three-dimensional lymph node viewing, thinner slices, and large study populations are necessary to improve these results. In addition, a better and universal definition of radiological positive lymph nodes may help improve studies addressing the correlation between size and malignancy. In the meantime, we question whether radiologists should assess and report lymph node status in preoperative CT scans in patients with colon cancer. Conflict of interest The authors report no conflicts of interest.
References 1. Lemmens V. Clinical epidemiology of colorectal cancer in the Netherlands. Studies of variation and trends with the Eindhoven Cancer Registry 2007.
Please cite this article in press as: de Vries FEE, et al., The value of pre-operative computed tomography scanning for the assessment of lymph node status in patients with colon cancer, Eur J Surg Oncol (2014), http://dx.doi.org/10.1016/j.ejso.2014.08.483
F.E.E. de Vries et al. / EJSO xx (2014) 1e5 2. Dighe S, Blake H, Koh M, et al. Accuracy of multidetector computed tomography in identifying poor prognostic factors in colonic cancer. Br J Surg 2010;97:1407–15. 3. Dighe S, Purkayastha S, Swift I, et al. Diagnostic precision of CT in local staging of colon cancers: a meta-analysis. Clin Radiol 2010; 65:708–19. 4. Balthazar E, Megibow A, Hulnick D, Naidich D. Carcinoma of the colon: detection and preoperative staging by CT. AJR 1988 Feb;150: 301–6. 5. Acunas¸ B, Rozanes I, Acunas¸ G, C ¸ elik L, Sayi I, G€okmen E. Preoperative CT staging of colon carcinoma (excluding the recto-sigmoid region). Eur J Radiol 1990;11:150–3. 6. Gazelle S, Gaa J, Saini S, Shellito P. J Comput Assist Tomogr 1995 Jan/Feb;19(1):87–91. 7. Harvey C, Amin Z, Hare C, et al. Helical CT pneumocolon to assess colonic tumors: radiologic e pathologic correlation. AJR 1998 June1439–43. 8. Gomille T, Aleksic M, Ulrich B, Christ F. Significance of CT in the detection of regional lymph node metastases in colorectal carcinoma. Radiologe 1998;38:1077–82. 9. Cademartiri F, Luccichenti G, Rossi A, Pavone P. Spiral hydro-CT in the evaluation of colo-sigmoideal cancer. Radiol Med 2002;104:295– 306. 10. Filippone A, Ambrosini R, Fuschi M, Marinelli T, Genovesi D, Bonomo L. Preoperative T and N staging of colorectal cancer: accuracy of contrast-enhanced multi-detector row CT colonography e initial experience. Radiology 2004;231:83–90.
5
11. Chamadol N, Ninpiethoon T, Bhudhisawasd V, Pairojkul C. The role of CT scan in preoperative staging of colorectal carcinoma. J Med Assoc Thai 2005;88:1847–53. 12. Leufkens A, van den Bosch M, van Leeuwen M, Siersema P. Diagnostic accuracy of computed tomography for colon cancer staging: a systematic review. Scand J Gastroenterol 2011;46:887–94. 13. Chi Y, Zhang X, Li J, Sun Y. To be or not to be: significance of lymph nodes on pretreatment CT in predicting survival of rectal cancer patients. Eur J Radiol 2011;77:473–7. 14. Smith N, Bees N, Barnachano Y, Norman A, Swift R, Brown G. Preoperative computed tomography staging of nonmetastatic colon cancer predicts outcome: implications for clinical trials. Br J Cancer 2007; 96:1030–6. 15. Brown G, Richards C, Bourne M, et al. Morphologic predictors of lymph node status in rectal cancer with use of high-spatialresolution MR imaging with histopathologic comparison. Radiology 2003;227:371–7. 16. Kanamoto T, Matsuki M, Okuda J, et al. Preoperative evaluation of local invasion and metastatic lymph nodes of colorectal cancer and mesenteric vascular variations using multidetector-row computed tomography before laparoscopic surgery. J Comput Assist Tomogr 2007;31:831–9. 17. Nørgaard A, Dam C, Jakobsen A, Pløen J, Lindebjerg J, Rafaelsen SR. Selection of colon cancer patients for neoadjuvant chemotherapy by preoperative CT scan. Scand J Gastroenterol 2014 Feb;49(2):202–8.
Please cite this article in press as: de Vries FEE, et al., The value of pre-operative computed tomography scanning for the assessment of lymph node status in patients with colon cancer, Eur J Surg Oncol (2014), http://dx.doi.org/10.1016/j.ejso.2014.08.483
