Comparison of Magnetic Resonance Imaging and Computed Tomography in the Preoperative Staging of Rectal Cancer Claude Guinet, MD, PhD; Jean-No\l=e"\lBuy, MD; Michel A. Ghossain, MD; Alain S\l=e'\zeur,MD; Alain Mallet, Jean-Michel Bigot, MD; Dominique Vadrot, MD; Jean Ecoiffier, MD
patients with middle and lower rectal carcinomas operated on, with abdominoperineal resection in 10 pa-
\s=b\ Nineteen were
tients, lower anterior resection with coloanal anastomosis in 6
patients, and colorectal anastomosis in 3 patients. The distance of the lower margin of the tumor to insertion of the levator ani on the rectal wall was correctly evaluated by computed tomography in 12(63%) of 19 patients and by magnetic resonance imaging in 13 (68%) of 19 patients, while digital examination correctly assessed the distance in 15(79%) of 19 patients. Computed tomography and magnetic resonance imaging were unable to assess extension through the rectal wall. No significant difference was observed between computed tomography and magnetic resonance imaging in assessing extension to the perivesical fat, adjacent organs, pelvic side wall, or lymph nodes. According to the TNM classification, magnetic resonance imaging correctly staged 74% (14/19) of carcinomas, while computed tomography correctly staged 68% (13/19). (Arch Surg. 1990;125:385-388)
MD;
The lower margins of all tumors were within 8 cm of the anal verge by digital examination (mean [SD], 5.4 [1.7] cm). Rectoscopy mea¬ surements were, in comparison, 5.8 (1.8) cm. Computed tomography was performed from 2 days prior to MRI to 6 days after MRI (mean, both performed on the same day). No preoperative radiotherapy was performed. All patients underwent surgery within a mean 10-day period. The treatment consisted of 10 abdominoperineal resections (APRs), 6 lower anterior resections (LARs) with coloanal anastomo¬ sis, and 3 LARs with colorectal anastomosis. Usual dissections of lymph nodes (including those along middle hemorrhoidal vessels) were performed in all cases. The mean height of the tumors was 4.1 cm (SD, 1.2 cm). The location of the tumor in the transverse plane was exacted using a clock dial representation with the posterior midline located at the 6-o'clock position (mean, 5.5 [SD, 3.5]). Histologie types were adenocarcinoma (18 cases) and undifferentiated (1 case). The TNM6 classification of the tumors was Tl, NO (n l), T2, NO (n 11) (7 limited to the rectal wall, 4 involving the perirectal fat), T2, Nl (n 6) (3 limited to the rectal wall, 3 involving the perirectal fat), and T4, Nl (n l). In 2 patients, the tumors staged T2, Nl were =
=
=
Recent staging
studies1"3 have used computed tomography (CT) and magnetic resonance imaging (MRI) in the preoper¬
ative of rectal carcinoma. However, it is too early to evaluate the role of MRI.4 To our knowledge, no precise comparison between these two techniques using statistical methods has been made. Therefore, we conducted a prospec¬ tive study to precisely compare CT and MRI in the preoper¬ ative staging of rectal carcinoma. This comparison was also performed in the determination of the precise relationship of the tumor with the levator ani muscle. These comparisons were quantified by using a statistical matched-pairs
technique.5
PATIENTS AND METHODS Nineteen patients with primary rectal carcinomas were studied with both MRI and CT from October 1985 to June 1988. There were 14 men and 5 women, 49 to 78 years old (mean, 66 years). All cases were confirmed by rectoscopic biopsy before imaging.
=
associated with liver métastases.
Magnetic resonance imaging was performed using a 0.5-T super¬ conducting magnet (Magniscan 5000, General Electric-CGR, Paris, France) and a 1.5-T superconducting magnet (Gyroscan, Philips, Best, the Netherlands). Patients were placed in a supine position and the bladder was kept as full as possible during the examination. In all patients, 12-mm-thick slices were obtained in the transverse plane using the multislice spin echo technique. Slices were contiguous or separated with a relative gap of 10%. Trweighted images were obtained using sequences repeated with repetition time ranging from
400 (echo time of 28 milliseconds) to 500 milliseconds (echo time of 20 milliseconds). Sequences repeated with repetition time ranging from 1200 (echo time of 40, 80, and 120 milliseconds) to 2000 millisec¬ onds (echo time of 50 and 100 milliseconds) were able to produce T2-weighted images on late echoes. According to the site of the lesion, additional slices were performed in coronal or sagittal planes and then Tj- and T2-weighted images were produced. Field of view ranged from 400 mm to 450 mm and the acquisition was made on a 256 256 matrix.
Computed tomography
was
performed using
1800, Elscint, Haifa, Israel) with
Accepted for publication May 23,1989. From the Department of Radiology, H\l=o^\tel-Dieude Paris (Drs Guinet, Buy, Ghossain, Vadrot, and Ecoiffier); Department of Surgery, H\l=o^\pitalRothschild (Dr S\l=e'\zeur);Department of Statistics, H\l=o^\pitalde la Piti\l=e'\(Dr Mallet); and Department of Radiology, H\l=o^\pitalTenon (Dr Bigot), Paris, France. Reprint requests to Department of Radiology, H\l=o^\tel-Dieude Paris, 1 Place du Parvis de Notre-Dame, 75004 Paris, France (Dr Guinet).
a scanner
(Excel
4-second scan time. Scan circle diameter was 350 mm, giving a resolution of 5.5 pixels/cm. A double bolus of 60 mL of intravenous contrast material was routinely in¬ jected. One liter of oral contrast medium was given to the patient 1 hour before examination, and 300 mL of rectal contrast was also administered. Ten-millimeter-thick slices were achieved from the inferior mesenteric artery to the pubic symphysis. Magnetic resonance imaging or CT scans were interpreted inde-
Downloaded From: http://archsurg.jamanetwork.com/ by a University of Michigan User on 05/20/2015
a
pendently by two of us (CG., J-N.B., M.A.G., or D.V.) without knowledge of the other imaging data. First, distance between the
lower part of the tumor and the levator ani muscle was measured. On transverse images, the plane demonstrating insertion of the levator ani muscle on the rectal wall was chosen at the boundary of the ischiorectal fossa and the subperitoneal space, more precisely, at the level where the perirectal fascia vanishes.7,8 Magnetic resonance imaging results were also correlated when possible with sagittal or coronal data. Patients having a tumor whose distance from the lower margin to the levator ani was less than 2 cm were considered candi¬ dates for APR, and those with a tumor whose distance was 2 cm were considered eligible for LAR. Second, MRI and CT staging were performed using TNM classification without knowledge of the patho¬ logic findings. Magnetic resonance imaging and CT results were correlated in all cases with the surgical report and pathological exami¬ nation of the surgical specimen. Three parameters were evaluated: (1) local extension to the perirectal fat, (2) extension to the adjacent pelvic organs or the pelvic wall, and (3) presence or absence of lymphadenopathy along the usual lymphatic spread. On MRI, mural involvement was assessed by localized rectal wall thickening (3=0.5 cm9) and by the contrast between tumor and normal rectal wall (especially on T2-weighted images). Invasion of the perir¬ ectal fat by tumor was considered when a soft-tissue mass having a signal intensity similar to that of the primary tumor extended in the perirectal fat. This finding was evaluated on both Tr and T2-weighted images. Association of the previous findings with streaky densities extending from the adjacent tumor12,10"13 was also recorded as a positive sign of perirectal fat invasion. In the same way, perirectal fascia thickening (halo sign) involved in rectal cancer2,7,8,1 was also noted. A mass having a signal intensity similar to the signal intensity of the primary tumor that extended into an adjacent organ and produced a change in the anatomic configuration of that organ was considered as evidence of involvement of that organ. This extension was assessed on both T,- and T2-weighted images. Computed tomo¬ graphic findings for extension beyond the rectal wall were analyzed '
according to previously published criteria.7,10"14
Pararectal nodes and extension to the normal lymphatic pathways recorded. Nodes with intermediate signal on short repetition time sequences with absence of rephasing effects on even echoes, and tissue densities without early contrast enhancement after bolus injec¬ tion on CT images were considered as evidence of presence of a lymph node. According to our experience, pararectal nodes greater than or equal to 6 mm were considered as invaded lymph nodes. In the same way, distant lymph nodes were considered abnormal when the short axis of the node was greater than or equal to 10 mm in diameter. Usual statistical indicators with confidence intervals (P .05) were calculated.15 Statistical analysis used data from the matched-pairs technique." Binomial probability associated with observed values of discordances under the hypothesis of equal performance (noted HO) of MRI and CT in the different items of TNM staging were also calculated. were
=
RESULTS Tumor Localization
Magnetic resonance imaging precisely localized the lower margin of 14 of 19 tumors, while CT nicely depicted the lower margin of 16 of 19 tumors. For 10 patients treated by APR, MRI localized the lower margin in 6 patients and CT in 7 patients. Magnetic resonance imaging and CT predictions were identical in 9 of the 10 patients (APR planed 5 of 9 patients) and different in 1 patient (APR was predicted by CT and no conclusion was given by MRI). Magnetic resonance imaging predicted APR in 5 of 10 patients and CT in 6 of 10 patients. The comparison on MRI and CT in the prediction of APR is shown below:
CT Results MRI Results APR LAR No conclusion
APR LAR 5 0 0 1 10
No Conclusion 0 0 3
Digital examination predicted APR in 6 of 10 patients.
Among nine patients treated by LAR, MRI localized the margin of the tumor in eight patients and CT in nine patients. Magnetic resonance imaging and CT findings were identical in seven patients (possibility of LAR planed in six of seven patients) and different in two patients (one planing by CT was true [no conclusion by MRI], and one MRI prediction was true [APR was predicted by CT]). Magnetic resonance imaging and CT predictions were true in seven patients. A comparison of MRI and CT in the prediction of LAR is shown lower
below:
CT Results MRI Results APR LAR No conclusion
APR LAR 6 1 0 1 10
No Conclusion 0 0 0
Digital examination predicted LAR in nine of nine patients. Perirectal Growth Perirectal growth was histologically proved in eight pa¬ tients (seven T2 and one T4). Magnetic resonance imaging and CT findings were identical in six patients (four true-positive results and two false-negative results). In these two falsenegative results, only microscopic extension through the muscularis propria was noticed. In one of these two cases, histologically proved neoplastic nodules were missed by CT and detected by MRI but were considered as normal lymph nodes because oftheir small diameter (5 mm). Magnetic reso¬ nance imaging and CT results were different in two of eight patients (one MRI false-negative and one CT false-negative). In the MRI false-negative case, the perirectal fat was slightly involved. In the CT false-negative case, a 10-mm nodule (connected with the tumor), representing extension to the pelvic side wall missed by CT, was correctly identified by MRI because of the very good contrast on T2-weighted im¬ ages. A comparison of MRI and CT in evaluation of perirectal extension for tumors extending beyond the rectal wall is shown below: CT Results MRI Results
True-positive False-negative
True-Positive 4 1
False-Negative 1 2
Finally, extension to perirectal fat was detected in five of eight patients by MRI and CT; MRI and CT sensitivities were equal to 0.63 (range, 0.25 to 0.91). The probability associated
with observed differences under HO is 1. Eleven patients had no histological extension beyond the muscularis propria. Magnetic resonance imaging and CT re¬ sults were identical in 10 of 11 patients (10 true-negative results). Magnetic resonance imaging and CT results were different in 1 case (1 CT false-positive result). A comparison of MRI and CT in evaluation of integrity of perirectal fat for tumors limited to the rectal wall is shown below: CT Results MRI Results
True-Negative
True-negative False-positive
4 1
False-Positive 1 2
In this case, the tumor located in the middle third of the rectum had an important inflammatory stroma. Integrity of perirectal fat was assessed in 11 of 11 patients by MRI (speci¬ ficity, 1 [range, 0.71 to 1]) and in 10 of 11 patients by CT (specificity, 0.91 [range, 0.59 to 1]). The probability associ¬
ated with observed discordances under HO is still 1, Streaky densities radiating from the tumor in the perirectal fat were seen by MRI in two of eight and by CT in four of eight cases of tumor histologically involving the perirectal fat. On
Downloaded From: http://archsurg.jamanetwork.com/ by a University of Michigan User on 05/20/2015
the other hand, no streaky density was seen by MRI or CT in the 11 cases of tumors limited to the rectal wall. Thickening of perirectal fascia was seen by both MRI and CT in two of eight and by CT only in one of eight cases of tumors involving histologically the perirectal fat. In the two cases, the thicken¬ ing was associated with the presence of streaky densities previously noted. No thickening was observed in the 11 tu¬ mors limited to the rectal wall. Extension to Adjacent Organs and to the Pelvic Wall
pelvic wall was proved histologically in Magnetic resonance imaging detected the invasion (one MRI true-positive), whereas CT did not (one CT falsenegative). A comparison of MRI and CT in evaluation of adjacent organ invasion is shown below: Extension to the
one case.
CT Results MRI Results
True-positive False-negative
True-Positive 0 0
False-Negative 1 0
Magnetic resonance imaging and CT results were identical for the 18 of 19 patients with no histologically proved invasion of adjacent organs. A comparison ofMRI and CT in evaluation of adjacent organ invasion is shown below: CT Results MRI Results
True-negative False-positive
True-Negative 18 0
False-Positive 0 0
Lymph Nodes As said elsewhere,3 we distinguished perirectal lymph node involvement and distant lymph node involvement. In seven patients with involvement of perirectal lymph node disease, MRI and CT findings were identical in six (three true-positive results, three false-negative results). False-negative results were due to a measured node diameter less than 6 mm. Magnetic resonance imaging and CT results were different in one of seven patients (one MRI false-negative due to the nonvisualization of the invaded node). A comparison of MRI and CT in evaluation of pararectal lymph node invasion is shown below: CT Results MRI Results
True-positive False-negative
True-Positive 3 1
False-Negative 0 3
Magnetic resonance imaging detected the extension in three of seven patients (sensitivity, 0.43 [range, 0.10 to 0.82]) and CT detected it in four of seven patients (sensitivity, 0.57 [range, 0.18 to 0.90]). The probability associated with ob¬ served discordances under HO is 1. In 12 patients without pathological finding of node extension, MRI and CT were equal in 11 of 12 patients (1 false-positive result). In this case, an 8-mm nodule was misinterpreted as an invaded lymph node, whereas a tumorous nodule was diagnosed by patholog¬ ic study. Magnetic resonance imaging and CT were different in 1 of 12 patients. In this case, a 6-mm nodule seen on CT images appeared to be 4 mm on MRI. A comparison of MRI and CT in evaluation of pararectal lymph node integrity is shown below:
CT Results MRI Results
True-negative False-positive
True-Negative 10 0
False-Positive 1 1
Magnetic resonance imaging was true in 11 of 12 patients (specificity, 0.92 [range, 0.61 to 1]) and CT was true in 10
(specificity, 0.83 [range, 0.52 to 0.98]). The probability associ¬
ated with this discordance is 1. In 1 patient with distant node involvement, MRI and CT were identical (1 true-positive result). In this case, diameter of the node was equal to 15 mm by MRI and 13 mm by CT. In the remaining 18 patients, MRI and CT results were also identical (18 true-negative results). TNM
Staging
Concerning TNM staging, MRI and CT were identical in 16 of 19 tumors (13 were correctly staged, 1 T2, NO was overstaged as T2, Nl, and 2 T2, Nl were understaged as T2, NO). Magnetic resonance imaging and CT were different in 3 of 19 patients. One Nl tumor was understaged by MRI and cor¬ rectly staged by CT, while CT understaged 1T4, Nl tumor as T2, Nl and overstaged 1 T2, NO tumor as T2, Nl (both were correctly staged by MRI). A comparison of MRI and CT in TNM classification is shown below: CT Results MRI Results
TlandT2,N0 TlandT2,Nl T4, Nl
TlandT2,N0
TlandT2,Nl
13 1 0
1 3 1
T4.N1 0 0 0
Finally, MRI correctly staged 14 of 19 tumors; 1 was overstaged (T2, Nl instead of T2, NO) and 4 were understaged (T2, NO instead of T2, Nl). Computed tomography correctly staged 13 of 19 tumors; 4 were understaged (1 was staged as T2, Nl instead of T4, Nl, and 3 were staged as T2, NO instead of T2, Nl), and 2 were overstaged (Tl, Nl instead of Tl, NO). COMMENT
Many articles have assessed the value of MRI1'3 and CT"4 in preoperative staging of rectal carcinoma. One of the main problems in the therapeutic approach is to define precisely the position of the lower margin of the tumor relative to the levator ani to plan a conservative surgery; thus, a preoper¬ ative evaluation is of great importance. Our study demon¬ strates that MRI in 12 (63%) of 19 patients and CT in 13 (68%) of 19 patients correctly predicted the type of intervention. Conclusions of MRI and CT were quite similar; however, imaging planing is inferior to digital examination predictions (15 of 19). Thinner slices (up to 2 mm) may improve accuracy of the measurement. Moreover, this study seems to prove that frontal or coronal sections are not usually necessary in evaluating the position of the lower margin of the tumor related to the levator ani. This is not amazing, as the perirectal fascia is usually very nicely depicted in transverse sections of MRI and CT and disappears in caudal sections at the level of the levator ani.8 Further sagittal sections are useful when the rectum is lying on the levator plane. No case of invasion of the levator ani muscle was displayed as happened in the series of Hodgman et al.2 Many studies evaluated the ability of MRI2 and CT9"14 sepa¬ rately and by comparison1-3 in the tumor extension into the perirectal fat. Although changes of fat were always better seen with MRI than with CT in the series of Butch et al,1 our results demonstrate that both methods are quite similar. This seems logical since contrast between fat and rectum wall is excellent with both MRI and CT.2 Presence of a halo sign can be helpful but must be interpreted in the absence of radiation or previous surgery.2,16 Streaky densities are very suggestive of perirectal invasion but were only seen on MRI in two of eight and on CT in four of eight cases of tumor involving the perirectal fat; here, MRI and CT seem similar. Microscopic extension was impossible to diagnose with both methods, as occurred in the series of Hodgman et al. The ability of MRI and CT to assess the invasion of adjacent
Downloaded From: http://archsurg.jamanetwork.com/ by a University of Michigan User on 05/20/2015
already been reported.1,3 However, the number of published is too limited. In our series, of the pelvic side wall was detected by only MRI on both Tr and T2-weighted images, while it was missed on CT scan very likely because of the small size and poor contrast between lesion and muscle. However, no definite conclusion can be made about the comparison ofboth methods in this field. Moreover, no false-positive result has been found by MRI or CT, and both methods seem identical. Perirectal lymph nodes are impossible to distinguish from tumorous nodules2,11 with CT and MRI. The necessity of defin¬ ing threshold values of the diameter of invaded pararectal lymph nodes and distant lymph nodes has already been sug¬ gested.3 However, these thresholds in different territories have been defined retrospectively in a sample of carcinomas of the rectum. To our knowledge, no study until recently has been performed in normal subjects, as was the study of Glazer et al17 about mediastinal lymph nodes. No significant differ¬ organs has
cases that have been a case of invasion
ence in
both methods has been observed in our series. Howev¬
rephasing on echoes can even be helpful in differentiating small lymph nodes from vessels. Butch et al1 described a case of a 15-mm lymph node showing benign reactive hyperplasia but did not mention the precise site of the node. Concerning TNM staging, MRI and CT were unable to evaluate extension through the rectal wall, and no difference between Tl and T2 lesions could be noted. Magnetic reso¬ nance imaging correctly staged 14 of 19 tumors, whereas CT correctly staged 13 of 19. These results are different from the results of Hodgman et al,2 where CT was declared superior to MRI in TNM staging, and are consistent with those of Butch et al,1 where MRI was declared equal to or better than CT. er,
CONCLUSION Our results do not weaken the hypothesis of equal perfor¬ of MRI and CT in both TNM staging of rectal carcino¬ ma and in measuring the relationship of the tumor with the pelvic floor. Moreover, the low number of observed discor¬ dances implies that to demonstrate a significant difference between MRI and CT would require a large number of pa¬ tients. In that case, the utility of such a study is questionable. mance
However, our study shows that the possibility of MRI and CT localizing the lower margin of the tumor is equally elusive, and that the accuracy of MRI and CT in TNM staging of rectal carcinoma is as low as 74% and 68%, respectively. Improve¬ ment of these two fields may occur by the use of thinner slices and by a better knowledge of normal rectal lymph nodes. References 1. Butch RJ, Stark DD, Wittenberg J, et al. Staging rectal cancer by MR and CT. AJR. 1986;146:1155-1160. 2. Hodgman CG, MacCarty RL, Wolff BG, et al. Preoperative staging of rectal carcinoma by computed tomography and 0.15T magnetic resonance imaging. Dis Colon Rectum. 1986;29:446-450. 3. Guinet C, Buy JN, S\l=e'\zeurA, et al. Preoperative assessment of the extension of rectal carcinoma: correlation of MR, surgical, and histopathologic findings. J Comput Assist Tomogr. 1988;12:209-214. 4. Margulis AR, Thoeni RF. The present status of the radiologic examination of the colon. Radiology. 1988;167:1-5. 5. Siegel S, Castellan NJ. Nonparametric Statistics for the Behavioral Sciences. New York, NY: McGraw-Hill International Book Co; 1988. 6. Spiessl B, Hermaneck P, Scheibe O, Wagner G. TNM Atlas UICC. New York, NY: Springer-Verlag NY Inc; 1985:108-113. 7. Grabbe E, Lierse W, Winkler R. The perirectal fascia: morphology and use in staging of rectal carcinoma. Radiology. 1983;149:241-246. 8. Guillaumin E, Jeffrey RB, Shea WJ, Asling CW, Goldberg HI. Perirectal inflammatory disease: CT findings. Radiology. 1986;161:153-157. 9. Thoeni RF, Moss AA, Schnyder P, Margulis AR. Detection and staging of primary rectal and rectosigmoid cancer by computed tomography. Radiology.
1981;141:135-138. 10. van Waes PFGM, Koehler PR, Feldberg MAM. Management of rectal carcinoma: impact of computed tomography. AJR. 1983;140:1137-1142. 11. Adalsteinsson B, Glimelius B, Graffman S, Hemmingson P\l=a"\hlmanL. Computed tomography in staging ofrectal carcinoma. Acta Radiol. 1985;26:45\x=req-\
55. 12. Freeny PC, Marks WM, Tyan JA, Bolen JW. Colorectal carcinoma evaluation with CT: preoperative staging and detection of postoperative recurrence.
Radiology. 1986;158:347-353.
13. Balthazar EJ, Megibow AJ, Hulnick D, Naidich DP. Carcinoma of the colon: detection and preoperative staging by CT. AJR. 1988;150:301-306. 14. Koehler PR, Feldberg MAM, van Waes PFGM. Preoperative staging of rectal cancer with computerized tomography. Cancer. 1984;54:512-516. 15. Beyer WH. Handbook of Tables for Probability and Statistics. Boca Raton, Fla: CRC Press Inc; 1981. 16. Doubleday LC, Bernardino ME. CT findings in the perirectal area following radiation therapy. J Comput Assist Tomogr. 1980;4:634-638. 17. Glazer GM, Gross BH, Quint LE, Francis IR, Bookstein FL, Orringer MB. Normal mediastinal lymph nodes: number and size according to American Thoracic Society mapping. AJR. 1985;144:261-265.
In Other AMA Journals JAMA The Use of Autologous Blood: The National Blood Resource Education Program Expert Panel The risk of transmitting disease through blood transfusions continues to fall as additional blood donor screening and testing measures are implemented. Nevertheless, when a blood transfusion is needed during the perioperative period, autologous blood is the safest option for eligible patients. Three methods for obtaining autologous blood to use during or after a planned surgical procedure are preoperative autologous blood donation, perioperative blood salvage, and acute normovolemic hemodilution. These techniques can be used alone or in combination to decrease or eliminate a patient's exposure to homologous blood. However, because all transfusions carry some health risk and blood administration costs, autologous blood should not be collected or reinfused indiscriminately. Autologous blood services should be used for eligible patients who are likely to require a transfusion but should not be employed for minor procedures in which transfusion is unlikely (JAMA. 1990;263:414-417).
Reprint requests
MD 20892 (Susan D.
to
Coordinator, NBREP, National Heart, Lung, and Blood Institute, Bldg 31, Room 4A05, Bethesda,
Rogus, RN, MS).
Downloaded From: http://archsurg.jamanetwork.com/ by a University of Michigan User on 05/20/2015
patients with middle and lower rectal carcinomas operated on, with abdominoperineal resection in 10 pa-
\s=b\ Nineteen were
tients, lower anterior resection with coloanal anastomosis in 6
patients, and colorectal anastomosis in 3 patients. The distance of the lower margin of the tumor to insertion of the levator ani on the rectal wall was correctly evaluated by computed tomography in 12(63%) of 19 patients and by magnetic resonance imaging in 13 (68%) of 19 patients, while digital examination correctly assessed the distance in 15(79%) of 19 patients. Computed tomography and magnetic resonance imaging were unable to assess extension through the rectal wall. No significant difference was observed between computed tomography and magnetic resonance imaging in assessing extension to the perivesical fat, adjacent organs, pelvic side wall, or lymph nodes. According to the TNM classification, magnetic resonance imaging correctly staged 74% (14/19) of carcinomas, while computed tomography correctly staged 68% (13/19). (Arch Surg. 1990;125:385-388)
MD;
The lower margins of all tumors were within 8 cm of the anal verge by digital examination (mean [SD], 5.4 [1.7] cm). Rectoscopy mea¬ surements were, in comparison, 5.8 (1.8) cm. Computed tomography was performed from 2 days prior to MRI to 6 days after MRI (mean, both performed on the same day). No preoperative radiotherapy was performed. All patients underwent surgery within a mean 10-day period. The treatment consisted of 10 abdominoperineal resections (APRs), 6 lower anterior resections (LARs) with coloanal anastomo¬ sis, and 3 LARs with colorectal anastomosis. Usual dissections of lymph nodes (including those along middle hemorrhoidal vessels) were performed in all cases. The mean height of the tumors was 4.1 cm (SD, 1.2 cm). The location of the tumor in the transverse plane was exacted using a clock dial representation with the posterior midline located at the 6-o'clock position (mean, 5.5 [SD, 3.5]). Histologie types were adenocarcinoma (18 cases) and undifferentiated (1 case). The TNM6 classification of the tumors was Tl, NO (n l), T2, NO (n 11) (7 limited to the rectal wall, 4 involving the perirectal fat), T2, Nl (n 6) (3 limited to the rectal wall, 3 involving the perirectal fat), and T4, Nl (n l). In 2 patients, the tumors staged T2, Nl were =
=
=
Recent staging
studies1"3 have used computed tomography (CT) and magnetic resonance imaging (MRI) in the preoper¬
ative of rectal carcinoma. However, it is too early to evaluate the role of MRI.4 To our knowledge, no precise comparison between these two techniques using statistical methods has been made. Therefore, we conducted a prospec¬ tive study to precisely compare CT and MRI in the preoper¬ ative staging of rectal carcinoma. This comparison was also performed in the determination of the precise relationship of the tumor with the levator ani muscle. These comparisons were quantified by using a statistical matched-pairs
technique.5
PATIENTS AND METHODS Nineteen patients with primary rectal carcinomas were studied with both MRI and CT from October 1985 to June 1988. There were 14 men and 5 women, 49 to 78 years old (mean, 66 years). All cases were confirmed by rectoscopic biopsy before imaging.
=
associated with liver métastases.
Magnetic resonance imaging was performed using a 0.5-T super¬ conducting magnet (Magniscan 5000, General Electric-CGR, Paris, France) and a 1.5-T superconducting magnet (Gyroscan, Philips, Best, the Netherlands). Patients were placed in a supine position and the bladder was kept as full as possible during the examination. In all patients, 12-mm-thick slices were obtained in the transverse plane using the multislice spin echo technique. Slices were contiguous or separated with a relative gap of 10%. Trweighted images were obtained using sequences repeated with repetition time ranging from
400 (echo time of 28 milliseconds) to 500 milliseconds (echo time of 20 milliseconds). Sequences repeated with repetition time ranging from 1200 (echo time of 40, 80, and 120 milliseconds) to 2000 millisec¬ onds (echo time of 50 and 100 milliseconds) were able to produce T2-weighted images on late echoes. According to the site of the lesion, additional slices were performed in coronal or sagittal planes and then Tj- and T2-weighted images were produced. Field of view ranged from 400 mm to 450 mm and the acquisition was made on a 256 256 matrix.
Computed tomography
was
performed using
1800, Elscint, Haifa, Israel) with
Accepted for publication May 23,1989. From the Department of Radiology, H\l=o^\tel-Dieude Paris (Drs Guinet, Buy, Ghossain, Vadrot, and Ecoiffier); Department of Surgery, H\l=o^\pitalRothschild (Dr S\l=e'\zeur);Department of Statistics, H\l=o^\pitalde la Piti\l=e'\(Dr Mallet); and Department of Radiology, H\l=o^\pitalTenon (Dr Bigot), Paris, France. Reprint requests to Department of Radiology, H\l=o^\tel-Dieude Paris, 1 Place du Parvis de Notre-Dame, 75004 Paris, France (Dr Guinet).
a scanner
(Excel
4-second scan time. Scan circle diameter was 350 mm, giving a resolution of 5.5 pixels/cm. A double bolus of 60 mL of intravenous contrast material was routinely in¬ jected. One liter of oral contrast medium was given to the patient 1 hour before examination, and 300 mL of rectal contrast was also administered. Ten-millimeter-thick slices were achieved from the inferior mesenteric artery to the pubic symphysis. Magnetic resonance imaging or CT scans were interpreted inde-
Downloaded From: http://archsurg.jamanetwork.com/ by a University of Michigan User on 05/20/2015
a
pendently by two of us (CG., J-N.B., M.A.G., or D.V.) without knowledge of the other imaging data. First, distance between the
lower part of the tumor and the levator ani muscle was measured. On transverse images, the plane demonstrating insertion of the levator ani muscle on the rectal wall was chosen at the boundary of the ischiorectal fossa and the subperitoneal space, more precisely, at the level where the perirectal fascia vanishes.7,8 Magnetic resonance imaging results were also correlated when possible with sagittal or coronal data. Patients having a tumor whose distance from the lower margin to the levator ani was less than 2 cm were considered candi¬ dates for APR, and those with a tumor whose distance was 2 cm were considered eligible for LAR. Second, MRI and CT staging were performed using TNM classification without knowledge of the patho¬ logic findings. Magnetic resonance imaging and CT results were correlated in all cases with the surgical report and pathological exami¬ nation of the surgical specimen. Three parameters were evaluated: (1) local extension to the perirectal fat, (2) extension to the adjacent pelvic organs or the pelvic wall, and (3) presence or absence of lymphadenopathy along the usual lymphatic spread. On MRI, mural involvement was assessed by localized rectal wall thickening (3=0.5 cm9) and by the contrast between tumor and normal rectal wall (especially on T2-weighted images). Invasion of the perir¬ ectal fat by tumor was considered when a soft-tissue mass having a signal intensity similar to that of the primary tumor extended in the perirectal fat. This finding was evaluated on both Tr and T2-weighted images. Association of the previous findings with streaky densities extending from the adjacent tumor12,10"13 was also recorded as a positive sign of perirectal fat invasion. In the same way, perirectal fascia thickening (halo sign) involved in rectal cancer2,7,8,1 was also noted. A mass having a signal intensity similar to the signal intensity of the primary tumor that extended into an adjacent organ and produced a change in the anatomic configuration of that organ was considered as evidence of involvement of that organ. This extension was assessed on both T,- and T2-weighted images. Computed tomo¬ graphic findings for extension beyond the rectal wall were analyzed '
according to previously published criteria.7,10"14
Pararectal nodes and extension to the normal lymphatic pathways recorded. Nodes with intermediate signal on short repetition time sequences with absence of rephasing effects on even echoes, and tissue densities without early contrast enhancement after bolus injec¬ tion on CT images were considered as evidence of presence of a lymph node. According to our experience, pararectal nodes greater than or equal to 6 mm were considered as invaded lymph nodes. In the same way, distant lymph nodes were considered abnormal when the short axis of the node was greater than or equal to 10 mm in diameter. Usual statistical indicators with confidence intervals (P .05) were calculated.15 Statistical analysis used data from the matched-pairs technique." Binomial probability associated with observed values of discordances under the hypothesis of equal performance (noted HO) of MRI and CT in the different items of TNM staging were also calculated. were
=
RESULTS Tumor Localization
Magnetic resonance imaging precisely localized the lower margin of 14 of 19 tumors, while CT nicely depicted the lower margin of 16 of 19 tumors. For 10 patients treated by APR, MRI localized the lower margin in 6 patients and CT in 7 patients. Magnetic resonance imaging and CT predictions were identical in 9 of the 10 patients (APR planed 5 of 9 patients) and different in 1 patient (APR was predicted by CT and no conclusion was given by MRI). Magnetic resonance imaging predicted APR in 5 of 10 patients and CT in 6 of 10 patients. The comparison on MRI and CT in the prediction of APR is shown below:
CT Results MRI Results APR LAR No conclusion
APR LAR 5 0 0 1 10
No Conclusion 0 0 3
Digital examination predicted APR in 6 of 10 patients.
Among nine patients treated by LAR, MRI localized the margin of the tumor in eight patients and CT in nine patients. Magnetic resonance imaging and CT findings were identical in seven patients (possibility of LAR planed in six of seven patients) and different in two patients (one planing by CT was true [no conclusion by MRI], and one MRI prediction was true [APR was predicted by CT]). Magnetic resonance imaging and CT predictions were true in seven patients. A comparison of MRI and CT in the prediction of LAR is shown lower
below:
CT Results MRI Results APR LAR No conclusion
APR LAR 6 1 0 1 10
No Conclusion 0 0 0
Digital examination predicted LAR in nine of nine patients. Perirectal Growth Perirectal growth was histologically proved in eight pa¬ tients (seven T2 and one T4). Magnetic resonance imaging and CT findings were identical in six patients (four true-positive results and two false-negative results). In these two falsenegative results, only microscopic extension through the muscularis propria was noticed. In one of these two cases, histologically proved neoplastic nodules were missed by CT and detected by MRI but were considered as normal lymph nodes because oftheir small diameter (5 mm). Magnetic reso¬ nance imaging and CT results were different in two of eight patients (one MRI false-negative and one CT false-negative). In the MRI false-negative case, the perirectal fat was slightly involved. In the CT false-negative case, a 10-mm nodule (connected with the tumor), representing extension to the pelvic side wall missed by CT, was correctly identified by MRI because of the very good contrast on T2-weighted im¬ ages. A comparison of MRI and CT in evaluation of perirectal extension for tumors extending beyond the rectal wall is shown below: CT Results MRI Results
True-positive False-negative
True-Positive 4 1
False-Negative 1 2
Finally, extension to perirectal fat was detected in five of eight patients by MRI and CT; MRI and CT sensitivities were equal to 0.63 (range, 0.25 to 0.91). The probability associated
with observed differences under HO is 1. Eleven patients had no histological extension beyond the muscularis propria. Magnetic resonance imaging and CT re¬ sults were identical in 10 of 11 patients (10 true-negative results). Magnetic resonance imaging and CT results were different in 1 case (1 CT false-positive result). A comparison of MRI and CT in evaluation of integrity of perirectal fat for tumors limited to the rectal wall is shown below: CT Results MRI Results
True-Negative
True-negative False-positive
4 1
False-Positive 1 2
In this case, the tumor located in the middle third of the rectum had an important inflammatory stroma. Integrity of perirectal fat was assessed in 11 of 11 patients by MRI (speci¬ ficity, 1 [range, 0.71 to 1]) and in 10 of 11 patients by CT (specificity, 0.91 [range, 0.59 to 1]). The probability associ¬
ated with observed discordances under HO is still 1, Streaky densities radiating from the tumor in the perirectal fat were seen by MRI in two of eight and by CT in four of eight cases of tumor histologically involving the perirectal fat. On
Downloaded From: http://archsurg.jamanetwork.com/ by a University of Michigan User on 05/20/2015
the other hand, no streaky density was seen by MRI or CT in the 11 cases of tumors limited to the rectal wall. Thickening of perirectal fascia was seen by both MRI and CT in two of eight and by CT only in one of eight cases of tumors involving histologically the perirectal fat. In the two cases, the thicken¬ ing was associated with the presence of streaky densities previously noted. No thickening was observed in the 11 tu¬ mors limited to the rectal wall. Extension to Adjacent Organs and to the Pelvic Wall
pelvic wall was proved histologically in Magnetic resonance imaging detected the invasion (one MRI true-positive), whereas CT did not (one CT falsenegative). A comparison of MRI and CT in evaluation of adjacent organ invasion is shown below: Extension to the
one case.
CT Results MRI Results
True-positive False-negative
True-Positive 0 0
False-Negative 1 0
Magnetic resonance imaging and CT results were identical for the 18 of 19 patients with no histologically proved invasion of adjacent organs. A comparison ofMRI and CT in evaluation of adjacent organ invasion is shown below: CT Results MRI Results
True-negative False-positive
True-Negative 18 0
False-Positive 0 0
Lymph Nodes As said elsewhere,3 we distinguished perirectal lymph node involvement and distant lymph node involvement. In seven patients with involvement of perirectal lymph node disease, MRI and CT findings were identical in six (three true-positive results, three false-negative results). False-negative results were due to a measured node diameter less than 6 mm. Magnetic resonance imaging and CT results were different in one of seven patients (one MRI false-negative due to the nonvisualization of the invaded node). A comparison of MRI and CT in evaluation of pararectal lymph node invasion is shown below: CT Results MRI Results
True-positive False-negative
True-Positive 3 1
False-Negative 0 3
Magnetic resonance imaging detected the extension in three of seven patients (sensitivity, 0.43 [range, 0.10 to 0.82]) and CT detected it in four of seven patients (sensitivity, 0.57 [range, 0.18 to 0.90]). The probability associated with ob¬ served discordances under HO is 1. In 12 patients without pathological finding of node extension, MRI and CT were equal in 11 of 12 patients (1 false-positive result). In this case, an 8-mm nodule was misinterpreted as an invaded lymph node, whereas a tumorous nodule was diagnosed by patholog¬ ic study. Magnetic resonance imaging and CT were different in 1 of 12 patients. In this case, a 6-mm nodule seen on CT images appeared to be 4 mm on MRI. A comparison of MRI and CT in evaluation of pararectal lymph node integrity is shown below:
CT Results MRI Results
True-negative False-positive
True-Negative 10 0
False-Positive 1 1
Magnetic resonance imaging was true in 11 of 12 patients (specificity, 0.92 [range, 0.61 to 1]) and CT was true in 10
(specificity, 0.83 [range, 0.52 to 0.98]). The probability associ¬
ated with this discordance is 1. In 1 patient with distant node involvement, MRI and CT were identical (1 true-positive result). In this case, diameter of the node was equal to 15 mm by MRI and 13 mm by CT. In the remaining 18 patients, MRI and CT results were also identical (18 true-negative results). TNM
Staging
Concerning TNM staging, MRI and CT were identical in 16 of 19 tumors (13 were correctly staged, 1 T2, NO was overstaged as T2, Nl, and 2 T2, Nl were understaged as T2, NO). Magnetic resonance imaging and CT were different in 3 of 19 patients. One Nl tumor was understaged by MRI and cor¬ rectly staged by CT, while CT understaged 1T4, Nl tumor as T2, Nl and overstaged 1 T2, NO tumor as T2, Nl (both were correctly staged by MRI). A comparison of MRI and CT in TNM classification is shown below: CT Results MRI Results
TlandT2,N0 TlandT2,Nl T4, Nl
TlandT2,N0
TlandT2,Nl
13 1 0
1 3 1
T4.N1 0 0 0
Finally, MRI correctly staged 14 of 19 tumors; 1 was overstaged (T2, Nl instead of T2, NO) and 4 were understaged (T2, NO instead of T2, Nl). Computed tomography correctly staged 13 of 19 tumors; 4 were understaged (1 was staged as T2, Nl instead of T4, Nl, and 3 were staged as T2, NO instead of T2, Nl), and 2 were overstaged (Tl, Nl instead of Tl, NO). COMMENT
Many articles have assessed the value of MRI1'3 and CT"4 in preoperative staging of rectal carcinoma. One of the main problems in the therapeutic approach is to define precisely the position of the lower margin of the tumor relative to the levator ani to plan a conservative surgery; thus, a preoper¬ ative evaluation is of great importance. Our study demon¬ strates that MRI in 12 (63%) of 19 patients and CT in 13 (68%) of 19 patients correctly predicted the type of intervention. Conclusions of MRI and CT were quite similar; however, imaging planing is inferior to digital examination predictions (15 of 19). Thinner slices (up to 2 mm) may improve accuracy of the measurement. Moreover, this study seems to prove that frontal or coronal sections are not usually necessary in evaluating the position of the lower margin of the tumor related to the levator ani. This is not amazing, as the perirectal fascia is usually very nicely depicted in transverse sections of MRI and CT and disappears in caudal sections at the level of the levator ani.8 Further sagittal sections are useful when the rectum is lying on the levator plane. No case of invasion of the levator ani muscle was displayed as happened in the series of Hodgman et al.2 Many studies evaluated the ability of MRI2 and CT9"14 sepa¬ rately and by comparison1-3 in the tumor extension into the perirectal fat. Although changes of fat were always better seen with MRI than with CT in the series of Butch et al,1 our results demonstrate that both methods are quite similar. This seems logical since contrast between fat and rectum wall is excellent with both MRI and CT.2 Presence of a halo sign can be helpful but must be interpreted in the absence of radiation or previous surgery.2,16 Streaky densities are very suggestive of perirectal invasion but were only seen on MRI in two of eight and on CT in four of eight cases of tumor involving the perirectal fat; here, MRI and CT seem similar. Microscopic extension was impossible to diagnose with both methods, as occurred in the series of Hodgman et al. The ability of MRI and CT to assess the invasion of adjacent
Downloaded From: http://archsurg.jamanetwork.com/ by a University of Michigan User on 05/20/2015
already been reported.1,3 However, the number of published is too limited. In our series, of the pelvic side wall was detected by only MRI on both Tr and T2-weighted images, while it was missed on CT scan very likely because of the small size and poor contrast between lesion and muscle. However, no definite conclusion can be made about the comparison ofboth methods in this field. Moreover, no false-positive result has been found by MRI or CT, and both methods seem identical. Perirectal lymph nodes are impossible to distinguish from tumorous nodules2,11 with CT and MRI. The necessity of defin¬ ing threshold values of the diameter of invaded pararectal lymph nodes and distant lymph nodes has already been sug¬ gested.3 However, these thresholds in different territories have been defined retrospectively in a sample of carcinomas of the rectum. To our knowledge, no study until recently has been performed in normal subjects, as was the study of Glazer et al17 about mediastinal lymph nodes. No significant differ¬ organs has
cases that have been a case of invasion
ence in
both methods has been observed in our series. Howev¬
rephasing on echoes can even be helpful in differentiating small lymph nodes from vessels. Butch et al1 described a case of a 15-mm lymph node showing benign reactive hyperplasia but did not mention the precise site of the node. Concerning TNM staging, MRI and CT were unable to evaluate extension through the rectal wall, and no difference between Tl and T2 lesions could be noted. Magnetic reso¬ nance imaging correctly staged 14 of 19 tumors, whereas CT correctly staged 13 of 19. These results are different from the results of Hodgman et al,2 where CT was declared superior to MRI in TNM staging, and are consistent with those of Butch et al,1 where MRI was declared equal to or better than CT. er,
CONCLUSION Our results do not weaken the hypothesis of equal perfor¬ of MRI and CT in both TNM staging of rectal carcino¬ ma and in measuring the relationship of the tumor with the pelvic floor. Moreover, the low number of observed discor¬ dances implies that to demonstrate a significant difference between MRI and CT would require a large number of pa¬ tients. In that case, the utility of such a study is questionable. mance
However, our study shows that the possibility of MRI and CT localizing the lower margin of the tumor is equally elusive, and that the accuracy of MRI and CT in TNM staging of rectal carcinoma is as low as 74% and 68%, respectively. Improve¬ ment of these two fields may occur by the use of thinner slices and by a better knowledge of normal rectal lymph nodes. References 1. Butch RJ, Stark DD, Wittenberg J, et al. Staging rectal cancer by MR and CT. AJR. 1986;146:1155-1160. 2. Hodgman CG, MacCarty RL, Wolff BG, et al. Preoperative staging of rectal carcinoma by computed tomography and 0.15T magnetic resonance imaging. Dis Colon Rectum. 1986;29:446-450. 3. Guinet C, Buy JN, S\l=e'\zeurA, et al. Preoperative assessment of the extension of rectal carcinoma: correlation of MR, surgical, and histopathologic findings. J Comput Assist Tomogr. 1988;12:209-214. 4. Margulis AR, Thoeni RF. The present status of the radiologic examination of the colon. Radiology. 1988;167:1-5. 5. Siegel S, Castellan NJ. Nonparametric Statistics for the Behavioral Sciences. New York, NY: McGraw-Hill International Book Co; 1988. 6. Spiessl B, Hermaneck P, Scheibe O, Wagner G. TNM Atlas UICC. New York, NY: Springer-Verlag NY Inc; 1985:108-113. 7. Grabbe E, Lierse W, Winkler R. The perirectal fascia: morphology and use in staging of rectal carcinoma. Radiology. 1983;149:241-246. 8. Guillaumin E, Jeffrey RB, Shea WJ, Asling CW, Goldberg HI. Perirectal inflammatory disease: CT findings. Radiology. 1986;161:153-157. 9. Thoeni RF, Moss AA, Schnyder P, Margulis AR. Detection and staging of primary rectal and rectosigmoid cancer by computed tomography. Radiology.
1981;141:135-138. 10. van Waes PFGM, Koehler PR, Feldberg MAM. Management of rectal carcinoma: impact of computed tomography. AJR. 1983;140:1137-1142. 11. Adalsteinsson B, Glimelius B, Graffman S, Hemmingson P\l=a"\hlmanL. Computed tomography in staging ofrectal carcinoma. Acta Radiol. 1985;26:45\x=req-\
55. 12. Freeny PC, Marks WM, Tyan JA, Bolen JW. Colorectal carcinoma evaluation with CT: preoperative staging and detection of postoperative recurrence.
Radiology. 1986;158:347-353.
13. Balthazar EJ, Megibow AJ, Hulnick D, Naidich DP. Carcinoma of the colon: detection and preoperative staging by CT. AJR. 1988;150:301-306. 14. Koehler PR, Feldberg MAM, van Waes PFGM. Preoperative staging of rectal cancer with computerized tomography. Cancer. 1984;54:512-516. 15. Beyer WH. Handbook of Tables for Probability and Statistics. Boca Raton, Fla: CRC Press Inc; 1981. 16. Doubleday LC, Bernardino ME. CT findings in the perirectal area following radiation therapy. J Comput Assist Tomogr. 1980;4:634-638. 17. Glazer GM, Gross BH, Quint LE, Francis IR, Bookstein FL, Orringer MB. Normal mediastinal lymph nodes: number and size according to American Thoracic Society mapping. AJR. 1985;144:261-265.
In Other AMA Journals JAMA The Use of Autologous Blood: The National Blood Resource Education Program Expert Panel The risk of transmitting disease through blood transfusions continues to fall as additional blood donor screening and testing measures are implemented. Nevertheless, when a blood transfusion is needed during the perioperative period, autologous blood is the safest option for eligible patients. Three methods for obtaining autologous blood to use during or after a planned surgical procedure are preoperative autologous blood donation, perioperative blood salvage, and acute normovolemic hemodilution. These techniques can be used alone or in combination to decrease or eliminate a patient's exposure to homologous blood. However, because all transfusions carry some health risk and blood administration costs, autologous blood should not be collected or reinfused indiscriminately. Autologous blood services should be used for eligible patients who are likely to require a transfusion but should not be employed for minor procedures in which transfusion is unlikely (JAMA. 1990;263:414-417).
Reprint requests
MD 20892 (Susan D.
to
Coordinator, NBREP, National Heart, Lung, and Blood Institute, Bldg 31, Room 4A05, Bethesda,
Rogus, RN, MS).
Downloaded From: http://archsurg.jamanetwork.com/ by a University of Michigan User on 05/20/2015
