A, Magnetic resonance axial image of a rectal tumor invading the muscularis propria. B, Gross pathologic findings of the same tumor sectioned according to the axial magnetic resonance images.
Gagliardi G, Bayar S, Smith R, Salem RR. Preoperative Staging of Rectal Cancer Using Magnetic Resonance Imaging With External Phase-Arrayed Coils. Arch Surg. 2002;137(4):447–451. doi:10.1001/archsurg.137.4.447
Rectal cancer can be accurately staged preoperatively by magnetic resonance imaging (MRI) with external phase-arrayed coils.
Comparison of MRIs with pathologic staging.
Twenty-eight consecutive patients with biopsy-proven rectal cancer who did not undergo irradiation.
Patients underwent imaging using a 1.5-T MRI scanner with external phase-arrayed surface coils. Streaking of the perirectal fat and disruption of the bowel wall margin were interpreted as transmural invasion. Lymph nodes were defined as metastatic when they had a diameter of at least 0.5 cm. Tumors were staged according to the TNM staging system (American Joint Committee on Cancer guidelines) as confined to the bowel wall (T1-T2) and invading through the bowel wall (T3-T4). Patients underwent anterior resection (n = 15), abdominoperineal resection (n = 11), or local excision (n = 2).
Main Outcome Measures
Calculation of sensitivity, specificity, and accuracy for invasion through the bowel wall and lymph node status.
Sensitivity of MRI in detecting invasion through the bowel wall was 89% (16/18), specificity was 80% (8/10), and accuracy was 86% (24/28). Sensitivity for malignant lymphadenopathy was 67% (8/12), specificity was 71% (10/14), and accuracy 69% (18/26).
Although more costly and not as accurate as endoscopic ultrasound, MRI with phase-arrayed coils had excellent sensitivity at detecting transmural penetration of rectal cancer.
THE PROGNOSIS and treatment of rectal carcinoma depend on the stage at presentation.1,2 Low-lying lesions confined to the rectal wall may be treated by local excision alone.3- 5 Lesions that extend beyond the rectal wall have an increased recurrence rate when the mesorectum is inadequately excised6,7 and are best treated through laparotomy. The use of preoperative irradiation may down-stage locally advanced disease8,9 and improve prognosis,10 but to date there are no established criteria to identify the patients who would benefit from this approach. For tumors known to invade adjacent organs, neoadjuvant chemotherapy and radiation therapy have been used to achieve tumor-free margins,11 and in selected cases resection of adjacent pelvic organs, such as the urinary bladder, with the appropriate reconstruction may be required. Accurate preoperative staging is therefore crucial to determine the most appropriate treatment modality.
Computed tomography (CT) and transrectal ultrasound (TRUS) offer a new dimension in the staging of rectal cancer because of their ability to evaluate extension of tumor into the perirectal fat, adjacent organs, and lymph nodes.12,13 In large series, CT was only 33% to 77% accurate for T staging and 22% to 73% for N staging.13 Transrectal ultrasound is better than CT in preoperatively staging rectal cancer, with accuracy between 67% and 93% for T staging and approximately 60% for N staging.13 The limitations of TRUS are operator dependency, the inability to stage upper rectal and obstructing lesions, and the limited depth of penetration. In recent years, magnetic resonance imaging (MRI) with external body coils or endorectal coils has been used in the diagnosis of pelvic pathologic conditions, including rectal cancer.12 Magnetic resonance imaging offers the theoretical advantages over TRUS of better tissue discrimination, higher penetration, and less operator dependency. Conventional body coils offer low resolution, whereas endorectal coils share some of the limitations of TRUS. The recent introduction of external phase-arrayed coils improved the signal-to-noise ratio, allowing high-resolution imaging.14 The purpose of the present study is to evaluate the accuracy of MRI with external phase-arrayed coils in the preoperative staging of rectal cancer by comparing MRIs with postoperative pathologic staging.
The biopsy-proven rectal cancers of 28 consecutive patients were staged with MRI. There were 18 men and 10 women; mean age was 63 years (range, 26-89 years). Patients who received neoadjuvant radiation therapy were excluded.
Fifteen patients underwent anterior resection; 11, abdominoperineal resection; 1, transanal excision, including full-thickness resection of the rectal wall; and 1, transsacral excision. Total mesorectal excision was performed in 13 of 14 anterior resections and 9 of 11 abdominoperineal resections. Vascular ligation was performed just distal to the first branch of the inferior mesenteric artery, other than for the patients undergoing transanal or transsacral excision.
Magnetic resonance imaging was performed using a 1.5-T MRI scanner (Signa; GE Medical Systems, Milwaukee, Wis). An external phase-arrayed surface multicoil consisting of 2 anterior and 2 posterior coils was used for all cases. All patients underwent imaging while in the prone position following the placement of a small Foley catheter in the rectum and insufflation of approximately 200 to 300 cm3 of room air. No bowel preparation was used. A sagittal fast-spoiled gradient echo sequence was used to localize the lesion. This was followed by axial, conventional, spin echo T1-weighted images. Respiratory compensation and anterior in-field-of-view, superior, and inferior saturation pulses were used. The anterior saturation pulse was placed at the air-subcutaneous fat interface to reduce motion ghost artifact. Coronal and sagittal fast, spin echo T2-weighted images were then obtained. All images were interpreted by the same radiologist (R.S.). Specific comment was made regarding depth of invasion of the rectal wall, adjacent organ involvement, and the presence of lymphadenopathy.
Tumors confined to the bowel wall with smooth margins were considered as disease localized to the rectum (T1-T2). Soft tissue (ie, low-signal) extension into the high-signal perirectal fat and disruption of the outer margin of the bowel wall were considered as tumor extension into perirectal fat and therefore staged as T3, whereas adjacent organ involvement by tumor was considered T4. Soft tissue extension into perirectal fat was more apparent on fast T2-weighted images then on conventional T2-weighted images. Lesions were placed in 2 separate categories, either T1-T2 lesions (tumor confined to the bowel wall) or T3-T4 lesions (tumor invading through the bowel wall). The decision not to separate T1 and T2 tumors was based on the premise that both of these stages may be amenable to local excision and that decisions regarding adjuvant therapy would be determined based on postoperative pathologic staging. Although differentiation between T3 and T4 lesions has therapeutic implications, they were not separated in this study because of the limited number of patients with T4 lesions. Lymph nodes with axes more than 0.5-cm long were considered malignant and indicated to the surgeon to allow specific pathologic correlation as outlined herein.
Immediately after resection, the specimen was marked in the operating room with orienting sutures and then oriented for the pathologist. Lymph nodes that were identified preoperatively as being abnormal were specifically marked by the surgeon during the resection for pathologic correlation. Full pathologic examination and TNM staging according to the American Joint Committee on Cancer guidelines were carried out. Correlation was then made with the preoperative MRI findings.
Correlation of T staging was available for all 28 patients; correlation of N staging was available for 26 patients, because no nodes were removed in the 2 patients undergoing local excision.
Ten tumors were staged by pathologic findings as T1-T2; MRI correctly staged 8 of these and 2 were overstaged as T3. Eighteen tumors were staged by pathologic findings as T3-T4; MRI correctly staged 16 and 2 were understaged as T1-T2. Overall sensitivity for invasion through the bowel wall was 89%, specificity was 80%, and accuracy was 86%.
Two tumors were identified by pathologic findings as invading adjacent organs (T4); MRI correctly staged 1 and 1 was understaged as T3. The limited number of patients with T4 tumors precludes statistical analysis of these tumors as a separate group.
Twelve patients were identified by pathologic findings as having malignant lymphadenopathy; MRI correctly staged 8. Overall sensitivity for lymph node involvement was 67%, specificity was 71%, and accuracy was 69%.
Accurate preoperative staging of rectal cancer is often required to construct the most appropriate treatment strategy. Full-thickness transanal excision has been studied extensively during the last decade and may be indicated in selected patients with tumors confined to the submucosa or muscularis propria (T1-T2).3- 5 Distinction between stages T1 and T2 may have therapeutic implications, although most studies show that selection for adjuvant treatment or salvage resection can only be made based on the histologic characteristics of the resected specimen.15,16 Preoperative identification of malignant lymphadenopathy or invasion through the bowel wall allows patient selection for an abdominal approach and for possible neoadjuvant therapy.3,10,17 Regarding adjacent organ involvement (T4 disease), consideration must be given to the optimal approach whether it be preoperative irradiation to down-stage the disease11 or resection of adjacent organs with possible urologic and flap reconstruction.18
Our study was designed to determine the accuracy of MRI in this staging process. We excluded patients who underwent preoperative radiation therapy because this may confound interpretation of the MRI for the purpose of this study. Additionally, we have attempted to resolve a significant criticism of similar studies that correlated preoperative imaging with pathologic node positivity. This relates to the question of whether the lymph node identified as being positive by the preoperative study is in fact the involved node. Initially, we attempted to identify in the operating room or immediately thereafter the specific lymph node or nodes thought to be positive by the radiologist. Subsequently, we found that sectioning the specimen according to the axial images of the MRI allowed more specific correlation between MRI images and pathologic results (Figure 1).
Magnetic resonance imaging is a technique in continuous evolution. The use of external phase-arrayed coils has increased tissue discrimination and resolution14 and in our hands had an 89% sensitivity and an 80% specificity in detecting invasion through the bowel wall. Table 1 outlines the results of the studies evaluating different MRI techniques in preoperative staging of rectal cancer.4,19- 30 The difference in the methods used makes comparison difficult. The only study using a 4-element surface coil is by Brown et al,21 who used a combination of dedicated surface coils, thin slices, and optimal patient orientation with a small field of view, which is comparable to our technique of 4 phase-arrayed coils. By this technique, Brown and colleagues were able to accurately quantify the extent of extramural invasion, which correlates directly with tumor recurrence. Combination of phase-arrayed coils with endorectal coils offers no advantage.31- 33
Numerous methods have been used in preoperative evaluation of rectal cancer and have been compared with MRI. The value of clinical examination should not be underestimated, because with experienced examiners the accuracy is 72%.34
Although CT is unable to depict the various layers of the bowel wall, it can detect invasion into adjacent organs.35 Accuracy for nodes is poor and sensitivity is better in advanced stages (17% for T1-T3 and 81% for T4).13 Most studies22,24,25,27 comparing CT with MRI have observed greater accuracy with MRI. Endorectal ultrasound seems to be the current imaging modality of choice because it is sufficiently accurate, well tolerated, and inexpensive.13 The ultrasound technique we have adopted in our institution uses a combination of B-mode and color Doppler; accuracy rates in our hands are 86% for T stage and 80% for N stage.36 Advantages of MRI over TRUS include less interobserver variation23 and ability to stage bulky tumors and tumors high in the rectum.37 Transrectal ultrasound has 84% accuracy in distinguishing between T1 and T2 tumors,38 whereas MRI with external coils lacks this distinction.14,19,20 In addition, MRI is more accurate than TRUS in assessing invasion of adjacent organs22,24; this is possible with TRUS only at the expense of lower definition.38 Also, MRI may be better then TRUS in distinguishing between tumor tissue and reactive fibrosis.22 Neither MRI nor TRUS can distinguish between tumor tissue and inflammatory infiltrate, which occurs in 25% of rectal cancers39 and extends for only a few millimeters outside the bowel wall, thereby minimally affecting overstaging.22 Among the 4 studies that compared MRI with TRUS, 2 use endorectal coils and show comparable results.23- 25 One study28 comparing external body coil MRI with TRUS finds TRUS equivalent in T staging and better in N staging. A similar study29 reports better T staging with TRUS because of a clearer discrimination between T1 and T2 tumors, whereas results for N staging are similar. To date, no imaging modality has been demonstrated to be consistently accurate in determining lymph node involvement. Combination of different imaging modalities, such as CT or MRI and positron emission tomography, may be the next step together with the development of lymph node–specific contrast media.40
From our results, MRI with phase-arrayed multicoils identified 89% of tumors that invaded through the bowel wall with 86% accuracy. Sensitivity for lymph node involvement was 67% with 69% accuracy. These data and our review of the literature suggest that MRI with external phase-arrayed multicoils provides optimal assessment of tumor invasion through the bowel wall and invasion of adjacent organs. However, this particular MRI evaluation is a costly investigation, and because it does not yield higher accuracy rates than TRUS, it is unlikely to become a routine preoperative evaluation. In addition, MRI with external phase-arrayed multicoils has the advantage of low interobserver variation and may be of value in situations when the results of TRUS are not definitive and preoperative staging is critical.
This paper was presented at the 82nd Annual Meeting of the New England Surgical Society, Providence, RI, September 21, 2001.
Corresponding author and reprints: Ronald R. Salem, MD, Yale University, 333 Cedar St, New Haven, CT 06520 (e-mail: firstname.lastname@example.org).
Peter Cataldo, MD, South Burlington, Vt: I would like to congratulate the authors on some beautiful images of MRIs of rectal cancer and for emphasizing the importance of preoperative staging in the treatment of rectal cancer. As we know, preoperative selection is the key to appropriate therapy deciding between local excision, radical excision, and preoperative chemoradiation.
I have several questions for the authors today, but the main questions surround the utility of MRI vs endoanal ultrasound in the staging of rectal cancer. Several things are important here. Number one, in some individuals an MRI a full bowel preparation is necessary, whereas for ultrasound we generally require only a Fleets enema. For an MRI, most studies require 60 minutes to do the procedure with additional time for interpretation. Endoanal ultrasound generally requires only 10 to 15 minutes. In terms of logistics, we generally perform the endoanal ultrasound at the time of colonoscopic diagnosis of rectal cancer. The cost is the biggie there. At the University of Vermont, it costs $1495 to get an MRI, whereas the poor colorectal surgeons only charge $338 for an ultrasound. Our accuracies are relatively similar. Access may be greater to the MRI than to the ultrasound unit. We only have 1 endoanal ultrasound unit in the entire state of Vermont, and we have at least 2 or 3 MRIs.
In summary, I would like to ask the authors to again tell me the role of MRI in terms of preoperative staging of rectal cancer. If you have a disagreement between the MRI and the endoanal ultrasound, which are you going to believe?
Victor Pricolo, MD, Providence, RI: I have a couple of questions about the implications of grouping together T1 and T2 lesions for therapeutic purposes. I was wondering if you could comment on the association, for instance, between the incidence of positive nodes at the time of lymphadenectomy between, let's say, a T2 lesion with poor differentiation and lymphovascular invasion or whether or not you recommend that T1 and T2 lesions can similarly be treated with local excision. Most authors would agree that T1 lesions can be excised locally provided that there is no poor differentiation or lymphovascular invasion, but most T2 lesions are associated with over 10% or 20% incidence of regional lymphadenopathy and, of course, a local excision may not address that and so sometimes additional therapy is required.
I am somewhat biased because in our institution we do use endorectal ultrasound. There are a couple of advantages. One is that the colorectal surgeon can perform the procedure through a rigid proctoscope, placing the probe exactly at the accurate distance from the anal verge as measured by rigid scope. The reliability of both ultrasound or MRI, as far as lymph nodes staging is concerned, will probably never be greater than 60% or 70% in view of the fact that more and more we are detecting micrometastases through ultrastaging with cytokeratin staining and 20% to 30% of those positive nodes are of normal size. As long as we use size to predict lymph node involvement, we are probably not going to increase our reliability because we are getting better at diagnosing micrometastases pathologically.
John Russell, MD, New Britain, Conn: In your conclusion you suggested that there might be a role for MRI in patients in whom endorectal ultrasound was not feasible. I took that to mean patients with high lesions, bulky lesions, or those with stricture. However, I am a little confused, because those would be situations in which the option of local excision would generally not be considered. Could you perhaps tell us what group of patients might benefit by MRI as a means to determine appropriateness of local tumor excision vs the more traditional way of approaching this issue with either CT or endorectal ultrasound?
Neil Yeston, MD, Hartford, Conn: Final question. Do you think PET scanning might add to the sensitivity and specificity of your data?
Dr Gagliardi: Dr Cataldo, about the role of the MRI—this is also the same question of Dr Russell more or less—to be more specific about the indications, as I showed in my last slides, the indications are when transrectal ultrasound is not reliable or is difficult. This can happen either because the tumor is bulky or, as has been reported in high tumors and low tumors, transrectal ultrasound is less accurate. Fifteen percent of rectal cancers present with a complete stricture and this makes transrectal ultrasound impossible. The other role of MRI is when you have a locally very advanced cancer; you know that transrectal ultrasound has only a limited range, and as preoperative planning is essential in terms of preoperative radiation or planning additional procedures, maybe in those delicate cases MRI is indicated.
The second question was if there is a disagreement between MRI and transrectal ultrasound, and the way I can answer this question is, I have reviewed the literature very carefully and from our data—we also have published a paper on transrectal ultrasound—transrectal ultrasound seems to be a little more sensitive in T1 and T2, in the early rectal cancer, and MRI seems to be more sensitive in picking up invasion through the bowel wall and to adjacent organs, but this difference has never been proven to be statistically significant and is just a comment on the definition of the images that the authors of various papers make.
In answer to the third question of Dr Pricolo, yes, T1 and T2 are definitely important to distinguish, and this distinction can be made preoperatively, but if you do not have a biopsy of poorly differentiated cancer you usually, whether it is a T1 or a T2, locally excise it, and then based on the histology of the specimen, decide whether your local excision, as you mentioned, poorly differentiated specimen, or specimen with lymphatic and venous invasion are the ones that are at such a high risk of local recurrence that you should complete the procedure and go ahead and do your abdominoperineal or anterior resection.
The fourth question was about micrometastases in the lymph nodes. I can only say where the future may be, and this is in combinations of imaging techniques and functional imaging like PET scans or in discovery of new contrast media. There is now an MRI contrast media which is specific for lymphatics that has been experimented with in humans.
Yes, we will never be able to pick up things that are less than 3 mm and that is why we need something which looks at the function of the nodes instead of at size only. The way we increased our pickup rate of lymph nodes with transrectal ultrasound is by using the same machine that they use for prostate cancer, which has color Doppler and looks also at the vascularity of nodes; so some nodes that were less than 0.5 mm were considered positive because of increased vascularity.
Finally, the role of CT. Certainly, with the new CT scanners, CT images are more and more detailed, and I am sure that there is going to be a competition between CT and MRI, but at this point the studies that compare MRI with CT found MRI to be much more sensitive.