Radiology
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

Published online before print July 23, 2004, 10.1148/radiol.2323031368
This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF)
Right arrow All Versions of this Article:
2323031368v1
232/3/773    most recent
Right arrow Submit a response
Right arrow Alert me when this article is cited
Right arrow Alert me when eLetters are posted
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Bipat, S.
Right arrow Articles by Stoker, J.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Bipat, S.
Right arrow Articles by Stoker, J.
(Radiology 2004;232:773-783.)
© RSNA, 2004

Evidence-based Practice

Rectal Cancer: Local Staging and Assessment of Lymph Node Involvement with Endoluminal US, CT, and MR Imaging—A Meta-Analysis1

Shandra Bipat, MSc, Afina S. Glas, MD, PhD, Frederik J. M. Slors, MD, PhD, Aeilko H. Zwinderman, PhD, Patrick M. M. Bossuyt, PhD and Jaap Stoker, MD, PhD

1 From the Departments of Radiology (S.B., J.S.), Epidemiology and Biostatistics (A.S.G., A.H.Z., P.M.M.B.), and Surgery (F.J.M.S.), Academic Medical Center, University of Amsterdam, Meibergdreef 9, 1105 AZ Amsterdam, the Netherlands. Received August 26, 2003; revision requested October 27; revision received January 6, 2004; accepted February 2. Address correspondence to S.B. (e-mail: s.bipat@amc.uva.nl).


   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
RESULTS
DISCUSSION
 APPENDIX
 REFERENCES
 
PURPOSE: To perform a meta-analysis to compare endoluminal ultrasonography (US), computed tomography (CT), and magnetic resonance (MR) imaging in rectal cancer staging.

MATERIALS AND METHODS: Relevant articles published between 1985 and 2002 were included if more than 20 patients were studied, histopathologic findings were the reference standard, and data were presented for 2 x 2 tables; articles were excluded if data were reported elsewhere in more detail. Two reviewers independently extracted data on study characteristics and results. Bivariate random-effects approach was used to obtain summary estimates of sensitivity and specificity for invasion of muscularis propria, perirectal tissue, and adjacent organs and for lymph node involvement. Summary receiver operating characteristic (ROC) curves were fitted for perirectal tissue invasion and lymph node involvement.

RESULTS: Ninety articles fulfilled all inclusion criteria. For muscularis propria invasion, US and MR imaging had similar sensitivities; specificity of US (86% [95% confidence interval {CI}: 80, 90]) was significantly higher than that of MR imaging (69% [95% CI: 52, 82]) (P = .02). For perirectal tissue invasion, sensitivity of US (90% [95% CI: 88, 92]) was significantly higher than that of CT (79% [95% CI: 74, 84]) (P < .001) and MR imaging (82% [95% CI: 74, 87]) (P = .003); specificities were comparable. For adjacent organ invasion and lymph node involvement, estimates for US, CT, and MR imaging were comparable. Summary ROC curve for US of perirectal tissue invasion showed better diagnostic accuracy than that of CT and MR imaging. Summary ROC curves for lymph node involvement showed no differences in accuracy.

CONCLUSION: For local invasion, endoluminal US was most accurate and can be helpful in screening patients for available therapeutic strategies.

© RSNA, 2004

Index terms: Lymphatic system, neoplasms • Rectum, neoplasms, 757.30


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
RESULTS
DISCUSSION
 APPENDIX
 REFERENCES
 
Rectal cancer is a common cancer and a major cause of mortality in Western countries. The diagnosis is usually established by means of clinical examination (rectal digital examination), endoscopy (sigmoidoscopy and colonoscopy), double-contrast enema examination, and histologic confirmation, supplemented by biochemistry (eg, blood carcinoembryonic antigen measurement). All of these techniques are poor indicators of the depth of invasion and lymph node involvement, which are both important features for prognosis (1–5).

Accurate preoperative assessment of these prognostic factors is an important first step in assigning patients to one of the available treatment strategies, which include transanal local excision, transanal endoscopic microsurgery, total mesorectal excision, preoperative irradiation, and preoperative chemotherapy. From the clinical point of view, it is important to select patients for local therapy, such as transanal local excision or transanal endoscopic microsurgery (mainly stage T1 or lower) (6–9); total mesorectal excision (mainly stages T2 and T3); and a long course of preoperative (chemotherapeutic) radiation therapy, aimed at downsizing and downstaging the tumor(s) (mainly stage T4) (10–12).

In patients considered suitable for total mesorectal excision, the spread of tumor to the mesorectal fascia is the second important feature that needs to be assessed (13–15). This relation determines if a patient can be treated directly with or without a short course of preoperative radiation therapy or whether the patient should be considered to have a locally advanced tumor necessitating a long course of chemotherapeutic radiation therapy. This is an important next step in the selection of patients for the proper treatment strategy. The identification and the role of mesorectal fascia are still under investigation, however; therefore, the assessment of the depth of cancer invasion (T stage) remains the primary and most important feature in the treatment of patients with rectal cancer. The presence of lymph node involvement is relevant for clinical decision making in two circumstances: (a) if local excision in the absence of lymphadenopathy is performed and (b) if lymph node metastases are present outside the endopelvic envelope, in which case the tumor is considered to be locally advanced.

Noninvasive radiologic modalities such as endoluminal ultrasonography (US), computed tomography (CT), and magnetic resonance (MR) imaging have proved to be important and have been widely used diagnostic tools in the assessment of depth of cancer invasion and/or lymph node involvement. Extensive research on the diagnostic performance of these modalities in the staging of rectal cancer has been performed (16–20), yet studies on the evaluation of all three imaging modalities within the same patient population are limited.

Furthermore, a wide variation in study design, patient population, imaging techniques, and results exists. These factors make it difficult for workers in this field to know the diagnostic performance of these imaging modalities. A meta-analysis of diagnostic tests represents a powerful tool to summarize findings in the literature by taking into account and enabling analysis of differences between studies (21,22). Thus, the purpose of our study was to perform a meta-analysis to compare endoluminal US, CT, and MR imaging in the staging of rectal cancer.


   MATERIALS AND METHODS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
DISCUSSION
 APPENDIX
 REFERENCES
 
Literature Search
A comprehensive computerized systematic literature search (23) was performed (S.B.) to identify abstracts of English-language articles from studies involving human subjects. Relevant studies on the diagnostic performance of endoluminal US, CT, and MR imaging in the staging of rectal cancer were identified. The MEDLINE database from January 1985 to December 2002 was searched with the following keywords: (a) "rectal neoplasms" (medical subject heading, or MeSH) and (b) "magnetic resonance imaging" (MeSH) or "tomography, x-ray computed" (MeSH) or "ultrasonography" (MeSH) and (c) "specificity" or "false negative" or "accuracy" as text words. The EMBASE, Cochrane, and CANCERLIT databases were also checked for relevant articles by using (a) "rectal cancer" and (b) "magnetic resonance imaging" or "computed tomography" or "ultrasonography" and (c) "specificity" or "false negative" or "accuracy" as text words. To identify additional relevant references, the reference lists of the articles retrieved were checked manually.

After reading the abstracts, one reviewer (S.B.) examined all potentially eligible articles in which endoluminal US, CT, and/or MR imaging were evaluated. Reviews, letters, comments, case reports, and articles that did not present raw data were excluded.

Study Selection
Studies were selected if they fulfilled all of the following inclusion criteria: (a) More than 20 patients had histologically proved rectal adenocarcinoma or carcinoma and were not treated with preoperative chemotherapy and/or radiation therapy. (b) Histopathologic findings (specimens obtained at surgery, laparoscopy, laparotomy, lymph node biopsy) were used as the reference standard. (c) Sufficient data were presented to construct a 2 x 2 contingency table (either raw 2 x 2 data or sensitivity and/or specificity with absolute numbers of positive and negative findings or the standard errors) of the imaging modalities compared with the reference standard for invasion of the submucosa, muscularis propria, perirectal tissue, or adjacent organs or lymph node involvement (perirectal or distant lymph nodes).

Studies were excluded if data were reported elsewhere in more detail. When data were published more than once, the study with the most details or the most patients was included.

Data Extraction
Two reviewers (S.B., F.J.M.S.) independently extracted relevant data from each article, including study characteristics and test results, by using a standardized data extraction form. The reviewers were not blinded to authors, journal name, or year of publication. Both reviewers extracted data from all articles. In cases of discrepancies, a third blinded reviewer assessed all discrepant items, and majority opinion was used for analysis.

Study Characteristics
The following study design characteristics were scored: (a) patient selection: consecutive or nonconsecutive; (b) interpretation of test results: blinded or not blinded; (c) verification: complete or partial—in cases in which more than 10% of the study group was not subjected to the reference test, the study was scored as applying partial verification; all other cases were scored as complete verification; (d) methods of data collection: prospective, retrospective, or unknown—data collection was categorized as either prospective or retrospective; in case of doubt, the method of data collection was scored as unknown; (e) reporting of study population: sufficient or insufficient—a description of the study population was judged to be sufficient if at least the age of participants and male-to-female ratio were included; (f) reporting of diagnostic test(s): sufficient or insufficient; and (g) reporting of reference test: sufficient or insufficient.

In a study of diagnostic accuracy, both the reference test and the diagnostic test(s) should be described with sufficient detail to allow for replication, validation, and generalization of the study. Descriptions of the tests were scored as sufficient if clear definitions of positive and negative test results were mentioned in the text.

Additionally, the following study characteristics were recorded for each article: year of publication, sample size (number of patients), and mean age of patients.

Examination Results
The following imaging techniques were recorded in the assessment of retrieved articles: for endoluminal US, type of probe and frequency of transducer; for CT, type of contrast material (oral, rectal, or intravenous), section thickness, and use of spiral mode; and for MR imaging, magnetic field strength, sequence, intravenous contrast material (used or not used), and type of coil used (body coil with or without additional coil [eg, phased-array or endorectal coil]).

For local staging, 2 x 2 tables were extracted or reconstructed from reported sensitivity and specificity values and absolute numbers of positive and negative findings as follows: (a) for invasion of muscularis propria, stage T2 or higher versus stage T1; (b) for invasion of perirectal tissue, stage T3 or higher versus stage T2 or lower; and (c) for invasion of adjacent organs, stage T4 versus stage T3 or lower.

For invasion of the submucosa, no 2 x 2 tables could be extracted or reconstructed because of the limited data on negative results (T0 and Tis).

We extracted or reconstructed 2 x 2 tables for lymph node involvement (perirectal, iliac, or mesenteric lymph nodes). Cutoff values for positive lymph nodes were also extracted.

To avoid selection of data sets, in articles in which investigators tabulated the results for different readers (interobserver), for multiple observations per reader (intraobserver), for multiple MR imaging systems, and for multiple MR imaging sequences, all tabulated results (2 x 2 tables) were considered separate data sets.

Statistical Analysis
A bivariate random-effects model (24,25) was used to obtain summary estimates of sensitivity and specificity and to fit summary receiver operating characteristic (ROC) curves that corresponded to the observed ranges of sensitivity and specificity values.

Bivariate Random-Effects Analysis
In this model, we assumed that the true values of sensitivity and specificity followed a bivariate normal distribution around some common mean value of logit-transformed sensitivity (logit-sens) and logit-transformed specificity (logit-spec) with a variance matrix {sum} ({sigma}A and {sigma}B describe the variance among studies in logit-sens and logit-spec, respectively, and {sigma}AB is the covariance between logit-sens and logit-spec).

{r04se20e01}

Logit-sens and logit-spec were calculated as follows: logit-sens = ln[sens/(1 – sens)] and logit-spec = ln[spec/(1 – spec)], where ln is the natural logarithm, sens is sensitivity, and spec is specificity. Because of this transformation of sensitivity and specificity into logit-sens and logit-spec, these values will be approximately normally distributed with squared standard error 1/[nx(1 – x)], where n is the number of cases or control subjects and x is sensitivity or specificity.

The random-effects model produces estimates of the mean logit-sens and logit-spec with their standard errors. Sensitivity and specificity estimates with their 95% confidence intervals (CIs) were calculated after anti–logit transformation of the mean logit-sens and logit-spec. The random-effects model also produces the associated variances ({sigma}A and {sigma}B) and the covariance ({sigma}AB).

To display summary ROC curves, we estimated the intercept ({alpha}) and slope (ß) of the linear regression line: logit-sens = {alpha} + (ßlogit-spec) (26). The slope (ß) of this regression line equals the covariance between logit-sens and logit-spec ({sigma}AB) divided by the variance of logit-spec ({sigma}B): ß = {sigma}AB/{sigma}B.

After calculation of the slope, the intercept ({alpha}) was calculated by solving the regression equation between the mean values of logit-sens and logit-spec. After anti–logit transformation of the regression line, a summary ROC curve was obtained.

Bivariate Analysis with Covariates
To determine whether results were significantly affected by heterogeneity between individual studies, the following covariates were added to the bivariate random-effects model for modality: year of publication (continuous variable: 2000 was set to 0, 1999 to –1, 1998 to –2, 1997 to –3, etc), sample size (>50 vs ≤50 patients), interpretation of results (blinded vs not blinded), verification (complete vs partial), patient selection (consecutive vs nonconsecutive), and method of data collection (prospective vs retrospective or unknown). Year 2000 was chosen as the reference year because of the low number of publications after 2000.

We considered variables to be explanatory if their regression coefficients were statistically significant (P < .05). Subsequently, we performed bivariate regression analysis with multiple covariates for each stage per modality. In this analysis, previously identified explanatory variables were analyzed with a backward elimination procedure, where the variable with the highest P value was excluded first. Variables were considered statistically significant if P < .1.

Summary ROC Curves
For each modality, a model was obtained that was adjusted for significant variables that were set to 1, indicating the ideal design versus 0, as appropriate. The intercept and slope were estimated for the regression line (logit-sens = {alpha} + [ßlogit-spec]), and a summary ROC curve was fitted after anti–logit transformation of this regression line. The position of the summary ROC curve indicates the difference in diagnostic performance among the imaging modalities. A summary ROC curve located near the upper left corner indicates the better diagnostic modality.

Summary Estimates of Sensitivity and Specificity
To compare the estimates for endoluminal US, CT, and MR imaging, a final model was obtained that was adjusted for variables that significantly affected the estimates of the imaging modalities (set to 1, indicating the ideal design vs 0), as appropriate. Since tabulated results for different readers (interobserver), for multiple observations per reader (intraobserver), for multiple MR imaging systems, and for multiple MR imaging sequences were considered to be separate data sets, correlations were taken into account. For this approach, the empirical standard error calculated by means of the "sandwich estimator" was used, which is possible with the SAS software (version 8.02; SAS Institute, Cary, NC) procedure mixed (27). This approach was also used for intramodality intrapatient correlation (in some studies, different modalities were compared in the same patient population).

Studying the histograms of the residuals and the random-effects estimates confirmed the goodness of fit of the model. To evaluate the difference between estimates for endoluminal US, CT, and MR imaging, we included in our model a factor that indicated type of diagnostic modality; a P value of less than .05 of the regression coefficient of this factor was considered to indicate a significant difference.

Subgroup Analysis
Only the groups of studies on MR imaging and endoluminal US contained sufficient numbers to allow subgroup analysis of technical differences, although the numbers did not allow an analysis of the effects of covariates. We compared the following MR imaging techniques: (a) use of body coil versus surface and/or phased-array and/or endorectal coil, (b) unenhanced MR imaging (T1- and T2-weighted imaging) versus gadolinium-enhanced T1-weighted MR imaging, and (c) magnetic field strength used (<1.5 T vs ≥1.5 T). In a subgroup analysis of the endoluminal US techniques, low-frequency probes (<7.5 Mhz) versus high-frequency probes (≥7.5 Mhz) were compared.

All analyses were performed by using Microsoft Excel 2000 (Microsoft, Seattle, Wash), SPSS 10.0 for Windows (SSPS, Chicago, Ill), and SAS statistical software (SAS Institute).


   RESULTS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
DISCUSSION
 APPENDIX
 REFERENCES
 
Search Results and Study Selection
After the computerized search was performed and reference lists were extensively cross-checked, 357 articles were identified. We found 146 articles to be potentially eligible after reading the abstract, of which 90 fulfilled the criteria for inclusion. Reasons for not including studies were the following: less than 20 patients studied (n = 31); lack of reference standard (histopathologic findings obtained at surgery, laparotomy, laparoscopy, or lymph node biopsy) (n = 1); incomplete or inconclusive data to allow calculation of true-positive, false-positive, true-negative, and false-negative findings (n = 19); and data reported elsewhere in more detail (n = 5).

Data Extraction
From the 90 articles included, 299 data sets were retrieved. Most data sets suffered from selective patient sampling (64%), suboptimal interpretation of results (77%), and poor description of the reference standard (73%). The other study design characteristics were distributed as follows: complete verification of results (90%), sufficient description of patient populations (66%), sufficient description of diagnostic tests (89%), and prospective col-lection of data (50%). For each cancer stage and imaging modality, included data sets with corresponding numbers of patients, years of publication, and references are presented in Table 1. A full list of all included articles with all relevant study characteristics and complete examination results (for each stage and imaging modality) is available from the authors upon request.


View this table:
[in this window]
[in a new window]

  		TABLE 1. Study and Patient Characteristics of Included Data Sets for Each Cancer Stage

 
Data Analysis
After backward stepwise regression analysis was performed, several variables were identified as significant predictors of the diagnostic performance of endoluminal US, CT, and MR imaging for evaluation of invasion of the muscularis propria, perirectal tissue, and adjacent organs and lymph node involvement from rectal cancer (Table 2).


View this table:
[in this window]
[in a new window]

  		TABLE 2. Predictors Identified by Means of Backward Regression Analysis for Each Cancer Stage

 
Summary ROC Curves
With the regression models, intercepts and slopes of the summary ROC curves for US, CT, and MR imaging were calculated. Summary ROC curves could be fitted for data on perirectal tissue invasion and lymph node involvement. For invasion in the muscularis propria and adjacent organs, no intercepts and slopes could be defined because of the homogeneity of either the sensitivity or specificity values. The summary ROC curves for perirectal tissue invasion have different positions in ROC space (Fig 1), indicating differences in diagnostic performance between the imaging modalities, with the summary ROC curve for endoluminal US located nearest to the upper left corner. The summary ROC curves for lymph node involvement have similar positioning in ROC space, indicating no differences in diagnostic performance of the imaging modalities.



View larger version (28K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 1. Summary ROC curves based on the final regression model for evaluation of perirectal tissue invasion and lymph node involvement with US, CT, and MR imaging and results of individual data sets. Left: Summary ROC curves for perirectal tissue invasion included consecutive patient selection for US, publication year for CT, and prospective data collection for MR imaging. Right: Summary ROC curves for lymph node involvement included publication year and prospective data collection for US, complete verification of results for CT, and publication year and blinded interpretation of results for MR imaging. Year 2000 was chosen as the reference year. A summary ROC curve located near the upper left corner indicates better diagnostic modality. Curves for perirectal tissue invasion indicate differences in diagnostic performance among imaging modalities, with summary ROC curve for US located nearest to the upper left corner. Curves for lymph node involvement indicate no differences in diagnostic performance of imaging modalities.

 
Summary Estimates of Sensitivity and Specificity
Muscularis propria invasion.—In the final model, publication year and sample size (>50 patients) were included as covariates for endoluminal US. Because of the few CT data sets available for evaluation, no data analysis was performed for this aspect. No significant variables were identified for MR imaging (Table 2). Endoluminal US and MR imaging have similar sensitivity estimates of 94%. The specificity value for US (86%) was significantly higher (P = .02) than that for MR imaging (69%), indicating overstaging of T1 (or lower) tumors with MR imaging (Table 3).


View this table:
[in this window]
[in a new window]

  		TABLE 3. Summary Estimates of Sensitivity and Specificity for Endoluminal US, CT, and MR Imaging in the Staging of Rectal Cancer

 
Perirectal tissue invasion.—The final model included consecutive patient selection for endoluminal US, publication year for CT, and prospective data collection for MR imaging as covariates. The sensitivity estimates for US, CT, and MR imaging were 90%, 79%, and 82%, respectively, with a significantly higher sensitivity estimate for US than for CT (P < .001) and MR imaging (P = .003), indicating understaging of T3 (or higher) tumors with CT and MR imaging. The specificity estimates for US, CT, and MR imaging were comparable: 75%, 78%, and 76%, respectively (Table 3).

Adjacent organ invasion.—The model included year of publication and sample size (>50 patients) for US and publication year for MR imaging as covariates. No significant predictors were found for the diagnostic performance of CT. Sensitivity estimates of all imaging modalities were comparable: 70% for US, 72% for CT, and 74% for MR imaging. Specificity estimates were also comparable: 97% for US, 96% for CT, and 96% for MR imaging (Table 3).

Lymph node involvement.—Year of publication and prospective data collection for US, complete verification for CT, and year of publication and blind interpretation of results for MR imaging were included as covariates in the final model. Sensitivity estimates for US, CT, and MR imaging were comparably low: 67%, 55%, and 66%, respectively. Specificity values were also comparable: 78% for endoluminal US, 74% for CT, and 76% for MR imaging (Table 3).

In all models in which publication year was included as the covariate, year 2000 was chosen as the reference year because in some data sets, only one data set was available for publications after 2000.

For invasion of muscularis propria and adjacent organs and for lymph node involvement, more significant variables were found for endoluminal US than for CT and MR imaging (Table 2). Adjustment of the models for US with more variables and the models for CT and MR imaging with fewer variables could lead to overestimation of the US estimates and underestimation of the CT and MR imaging estimates. To evaluate this, models were also studied in which all variables that significantly affected the estimates of US, CT, or MR imaging were included simultaneously. The results of these models were comparable to the results of the models in which adjustment per modality was performed. To avoid repetition, these results are not presented.

Subgroup Analysis
Since there was a sufficient number of data sets for perirectal tissue invasion, subgroup analysis could be performed for this stage. The results of the subgroup analysis for MR imaging techniques (use of a body coil vs a body coil with an additional coil, unenhanced MR imaging vs gadolinium-enhanced MR imaging, and low vs high magnetic field strength) and endoluminal US techniques (low vs high frequency) are presented in Table 4. Summary ROC curves are presented in Figure 2. No significant differences were found between the techniques, as shown by the summary estimates and the summary ROC curves.


View this table:
[in this window]
[in a new window]

  		TABLE 4. Subgroup Analysis on MR Imaging and Endoluminal US Techniques for Perirectal Tissue Invasion

 


View larger version (38K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 2. Summary ROC curves for different subgroups in the evaluation of perirectal tissue invasion and for results of individual data sets. Upper left: Summary ROC curves for MR imaging performed with body coil alone versus body coil with additional coil. Upper right: Summary ROC curves for unenhanced versus gadolinium-enhanced MR imaging. Lower left: Summary ROC curves for low magnetic field strength (<1.5 T) versus high magnetic field strength (≥1.5 T). Lower right: Summary ROC curves for endoluminal US: low frequency (<7.5 Mhz) versus high frequency (≥7.5 MHz). Summary ROC curve located near the upper left corner indicates the better diagnostic modality. Summary ROC curves for subgroups showed similar positioning in ROC space, indicating no differences between subgroups.

 

   DISCUSSION
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
DISCUSSION
APPENDIX
 REFERENCES
 
In this meta-analysis, we obtained summary estimates and summary ROC curves for the diagnostic accuracy of endoluminal US, CT, and MR imaging in the staging of patients with rectal cancer. Endoluminal US was found to be the most accurate modality when compared with CT and MR imaging for evaluation of local invasion of rectal cancer. For lymph node involvement, the results were comparable, with low sensitivity values. US was used to evaluate only perirectal or mesorectal lymph nodes, whereas CT and MR imaging were also used to evaluate iliac and mesenteric or retroperitoneal lymph nodes. However, these data were not presented separately. We found no significant differences between MR imaging and US techniques in a subgroup analysis.

In our meta-analysis of literature from a 16-year period, we attempted to minimize some of the well-known limitations of meta-analysis by applying (a) data extraction by two reviewers independently, since differences in interpretation and extraction of data can lead to biased results; (b) explicit inclusion criteria, such as use of histopathologic findings as the reference standard—differential verification has been shown to lead to overestimation of results (111); (c) exclusion of duplicate publications—positive results are more likely to be published more than once and could lead to overestimation of results; (d) extraction of study characteristics to study the effects on the diagnostic performance—Lijmer et al (111) showed that bias in study design characteristics led to either over- or underestimation of diagnostic performance; and (e) combination of results in a bivariate random-effects approach to account for variation in results. The outcomes of the bivariate approach are both (a) summary estimates of sensitivity and specificity, which are more familiar to clinicians, and (b) a covariance matrix to fit summary ROC curves.

The advantage of this regression analysis over regular summary ROC analysis (ln[DOR] = {alpha} + ßS, where DOR is the diagnostic odds ratio and S is sum) (112) (Appendix) is that this model accounts not only for the heterogeneity between studies due to different threshold settings but also for the error of estimation of the sensitivity and specificity values in each study. This random model also accounts for the residual heterogeneity that may remain even after adjusting for study characteristics and imaging techniques (113).

To avoid missing important articles, additional databases such as EMBASE, Cochrane, and CANCERLIT were checked. In addition, the reference lists of original articles and reviews, retrieved by means of electronic search in the MEDLINE database, were checked manually to identify relevant articles. To avoid exclusion of relevant articles, the literature search was performed for the years 1985–2002. The question arises whether techniques used in the earlier period represent outdated technology with inferior results. We therefore performed a subgroup analysis for different techniques. No significant differences could be observed. We performed covariate adjustment for publication year. The year of publication had different effects (increased or decreased) on the diagnostic accuracy. This may be explained not only by technical developments and different techniques employed over the years, but also by variation in training and expertise of the investigators. Until now, the latter has rarely been reported in the literature.

An important problem in the performance of meta-analysis is the possibility of publication bias. However, we attempted to study the aspect of publication bias by evaluating whether the size of studies affected the diagnostic accuracy. In particular, small studies with optimistic results may be published more easily than small studies with unfavorable results. Larger studies with optimistic results may also be published more easily than larger studies with unfavorable results, but this difference will be smaller. The evaluation of the effect of sample size did not show a better diagnostic performance of smaller studies compared with larger studies in all data sets. Funnel plot analysis was not performed, since the limited number of data points for some data sets could have decreased the power of detecting publication bias. We also did not quantify the number of unpublished studies, since authors are reluctant to provide information on unpublished studies.

One other limitation is the consideration of 2 x 2 tables for different readers, for multiple observations per reader, and for multiple MR imaging sequences as separate data sets. This has been performed to avoid selection bias. We are aware of the dependency inherent in data sets from the same patient population. Studying this dependency is not possible with our software, since SAS procedure mixed (SAS Institute) is only able to adjust for this potential dependency if the same amounts of data sets are available in each study. We studied this correlation by using the empirical standard error calculated by means of the "sandwich estimator," which is possible in SAS procedure mixed (SAS Institute) (27). We also used this approach to adjust for correlations between imaging modalities applied in the same patient population.

Another possible limitation of this meta-analysis is that a multiple backward stepwise regression analysis was performed with six covariates, and the final model was adjusted for significant variables. In our study, more data sets were available for US than for CT or MR imaging. For some stages (muscularis propria invasion, adjacent organ invasion, and lymph node involvement), more significant predictors were found for US; this can be explained by the larger number of data sets for US. By adjusting the CT and MR imaging models with fewer variables, the accuracy of CT and MR imaging could be underestimated. However, we also obtained models for each stage in which we simultaneously adjusted for all variables that significantly affected the estimates of three imaging modalities per stage. These results were comparable with the results of the models in which adjustment per modality was performed and are therefore not presented in this article. Moreover, because of the few MR imaging studies suitable for subgroup analysis, the sensitivity and specificity values had wide 95% CIs that overlapped completely or partially, indicating no significant differences between several techniques.

Patient characteristics (disease stage, age, or sex distribution) are also important for diagnostic accuracy, but variation in data presentation made it impossible to study the effect of these variables. In general, the time interval between performance of diagnostic tests and the reference test should be short. A longer period between performance of the diagnostic test and the reference test will lead to a greater change in the disease status and decrease in the discriminatory power of the diagnostic test. This implies that the comparison should ideally be performed on the same day. In most of the studies, this time period was not described or was diverse; therefore, this variable could not be analyzed. A large interval is not likely, however, given the disease under consideration.

Finally, positron emission tomography with fluorine 18 fluorodeoxyglucose is another imaging modality that can add incremental information to the preoperative assessment of patients with rectal cancer. However, data on this issue were limited (114–116) and were therefore not included in this meta-analysis.

We are aware of one other systematic review on the diagnostic performance of these imaging modalities in the staging of rectal cancer. Kwok and colleagues (117) reported summary sensitivity values of 93%, 78%, and 86% with specificity values of 78%, 63%, and 77%, respectively, for endoluminal US, CT, and MR imaging in the determination of wall penetration (stage T3). In the assessment of lymph node involvement, the sensitivity of US, CT, and MR imaging was found to be 71%, 52%, and 65% with specificity of 76%, 78%, and 80%, respectively.

Kwok et al (117) also found US to be the most accurate modality when compared with MR imaging and CT in the assessment of wall penetration. However, when evaluating studies of MR imaging that included use of an endorectal coil, this technique was found to be as effective as US in the assessment of wall penetration and was the most effective technique in the assessment of nodal involvement. The conclusion was that MR imaging with use of an endorectal coil offers the maximum amount of information by a single modality in the staging of rectal cancer. The latter is not in concordance with our findings, which is most likely caused by methodologic differences. Their analysis was based on a descriptive analysis by simply pooling data without accounting for (a) heterogeneity between studies due to different threshold settings (between-study variation) and (b) errors of estimation of sensitivity and specificity values in each study (within-study variation). Moreover, no 95% CIs or P values were reported to reflect the statistical precision of the observed differences.

Although US has proved to be a better imaging modality than CT and MR imaging in the present study, it has several limitations: operator dependency; limitation to tumors located 8–10 cm from the anal verge when a rigid probe is used; and no assessment of stenotic tumors. Selection of patients may therefore lead to biased results. Because of the lack of detailed information on patient selection, we could not study whether the selection of patients for US was different from that of CT and MR imaging; however, the prevalences per cancer stage were comparable (Table 1).

Moreover, endoluminal US is not able to depict lymph nodes that are outside the range of the transducer and cannot discriminate between lymph nodes inside or outside the mesorectal fascia, since the fascia is not identified at endoluminal US. The latter is also of importance in determining the spread of stage T3 tumors considered for total mesorectal excision. This may explain the more recent widespread use of MR imaging, since these limitations do not apply to MR imaging with external coils. To improve the sensitivity values of MR imaging for lymph node detection, newer techniques, such as use of new lymph node–specific MR imaging contrast agents (ultrasmall iron-based particles taken up by the lymphatic system), may provide a more sensitive MR imaging method to detect lymph node involvement (118–122). However, both the use of new lymph node–specific MR imaging contrast agents and the identification of the mesorectal fascia are still under investigation.

CT has limitations in differentiating and distinguishing the different layers of the rectal wall, demonstrating the mesorectal fascia, and depicting tumor invasion in surrounding pelvic structures. The introduction of multisection CT scanners may improve the diagnostic value of this modality, as the assessment of local disease with improved visualization of the mesorectal fascia can be combined with assessment of liver involvement and lung metastases (123,124).

On the basis of the results of this meta-analysis, endoluminal US seems to be a better diagnostic imaging test for local staging than are CT and MR imaging. Because of the limited information on the identification of the mesorectal fascia with MR imaging (125–127) and spiral CT, at present, endoluminal US might be helpful in selecting patients for available therapeutic strategies. The identification of lymph nodes with endoluminal US, CT, and MR imaging remains a major point of concern.


   APPENDIX
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
DISCUSSION
 APPENDIX
REFERENCES
 
Littenberg and Moses (112) suggested that D[logit(TPR) – logit(FPR)] is modulated as a linear function of S[logit(TPR) + logit(FPR)], where D is difference, FPR is false-positive rate (1 – specificity), TPR is true-positive rate (sensitivity), and S is sum.

Thus, they used the linear model D = {alpha} + ßS.

Suppose, in a 2 x 2 table, n1 is the number of patients with disease and n2 is the number of patients without disease.

{r04se20e02}
where DOR is the diagnostic odds ratio, logit x = ln[x/(1 – x)], x is true-positive rate or false-positive rate, and

{r04se20e03}


   FOOTNOTES
 
Authors stated no financial relationship to disclose.

Abbreviations: CI = confidence interval, ROC = receiver operating characteristic

Author contributions: Guarantor of integrity of entire study, S.B.; study concepts and design, S.B., P.M.M.B., J.S.; literature research, S.B.; data acquisition, S.B., F.J.M.S.; data analysis/interpretation, S.B., A.S.G., A.H.Z., P.M.M.B., J.S.; statistical analysis, S.B., A.S.G., A.H.Z., P.M.M.B.; manuscript preparation, S.B.; manuscript definition of intellectual content, all authors; manuscript editing, S.B., P.M.M.B., J.S.; manuscript revision/review and final version approval, all authors


   REFERENCES
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
DISCUSSION
 APPENDIX
 REFERENCES
 

  1. Harrison JC, Dean PJ, el Zeky F, Vander ZR. From Dukes through Jass: pathological prognostic indicators in rectal cancer. Hum Pathol 1994; 25:498-505.[CrossRef][Medline]
  2. Jass JR, Love SB. Prognostic value of direct spread in Dukes’ C cases of rectal cancer. Dis Colon Rectum 1989; 32:477-480.[Medline]
  3. Tang R, Wang JY, Chen JS, et al. Survival impact of lymph node metastasis in TNM stage III carcinoma of the colon and rectum. J Am Coll Surg 1995; 180:705-712.[Medline]
  4. Willett CG, Badizadegan K, Ancukiewicz M, Shellito PC. Prognostic factors in stage T3N0 rectal cancer: do all patients require postoperative pelvic irradiation and chemotherapy? Dis Colon Rectum 1999; 42:167-173.[CrossRef][Medline]
  5. Wolmark N, Fisher B, Wieand HS. The prognostic value of the modifications of the Dukes’ C class of colorectal cancer. Ann Surg 1986; 203:115-122.[Medline]
  6. Akasu T, Kondo H, Moriya Y, et al. Endorectal ultrasonography and treatment of early stage rectal cancer. World J Surg 2000; 24:1061-1068.[CrossRef][Medline]
  7. Blair S, Ellenhorn JD. Transanal excision for low rectal cancers is curative in early-stage disease with favorable histology. Am Surg 2000; 66:817-820.[Medline]
  8. Gao JD, Shao YF, Bi JJ, Shi SS, Liang J, Hu YH. Local excision carcinoma in early stage. World J Gastroenterol 2003; 9:871-873.[Medline]
  9. Langer C, Liersch T, Markus P, et al. Transanal endoscopic microsurgery (TEM) for minimally invasive resection of rectal adenomas and "low-risk" carcinomas (uT1, G1–2). Z Gastroenterol 2002; 40:67-72.[CrossRef][Medline]
  10. Chen ET, Mohiuddin M, Brodovsky H, Fishbein G, Marks G. Downstaging of advanced rectal cancer following combined preoperative chemotherapy and high dose radiation. Int J Radiat Oncol Biol Phys 1994; 30:169-175.[Medline]
  11. Kaminsky-Forrett MC, Conroy T, Luporsi E, et al. Prognostic implications of downstaging following preoperative radiation therapy for operable T3–T4 rectal cancer. Int J Radiat Oncol Biol Phys 1998; 42:935-941.[CrossRef][Medline]
  12. Rau B, Wust P, Hohenberger P, et al. Preoperative hyperthermia combined with radiochemo therapy in locally advanced rectal cancer: a phase II clinical trial. Ann Surg 1998; 227:380-389.[CrossRef][Medline]
  13. Hermanek P, Hermanek P, Hohenberger W, Klimpfinger M, Kockerling F, Papadopoulos T. The pathological assessment of mesorectal excision: implications for further treatment and quality management. Int J Colorectal Dis 2003; 18:335-341.[Medline]
  14. Nagtegaal ID, van Krieken JH. The role of pathologists in the quality control of diagnosis and treatment of rectal cancer: an overview. Eur J Cancer 2002; 38:964-972.
  15. Wiggers T, van de Velde CJ. The circumferential margin in rectal cancer: recommendations based on the Dutch Total Mesorectal Excision Study. Eur J Cancer 2002; 38:973-976.
  16. Gualdi GF, Casciani E, Guadalaxara A, d’Orta C, Polettini E, Pappalardo G. Local staging of rectal cancer with transrectal ultrasound and endorectal magnetic resonance imaging: comparison with histologic findings. Dis Colon Rectum 2000; 43:338-345.[CrossRef][Medline]
  17. Hunerbein M, Pegios W, Rau B, Vogl TJ, Felix R, Schlag PM. Prospective comparison of endorectal ultrasound, three-dimensional endorectal ultrasound, and endorectal MRI in the preoperative evaluation of rectal tumors: preliminary results. Surg Endosc 2000; 14:1005-1009.[CrossRef][Medline]
  18. Chiesura-Corona M, Muzzio PC, Giust G, Zuliani M, Pucciarelli S, Toppan P. Rectal cancer: CT local staging with histopathologic correlation. Abdom Imaging 2001; 26:134-138.[CrossRef][Medline]
  19. Matsuoka H, Nakamura A, Masaki T, et al. Preoperative staging by multidetector-row computed tomography in patients with rectal carcinoma. Am J Surg 2002; 184:131-135.[CrossRef][Medline]
  20. Blomqvist L, Machado M, Rubio C, et al. Rectal tumour staging: MR imaging using pelvic phased-array and endorectal coils vs endoscopic ultrasonography. Eur Radiol 2000; 10:653-660.[CrossRef][Medline]
  21. Kinkel K, Lu Y, Both M, Warren RS, Thoeni RF. Detection of hepatic metastases from cancers of the gastrointestinal tract by using noninvasive imaging methods (US, CT, MR imaging, PET): a meta-analysis. Radiology 2002; 224:748-756.[Abstract/Free Full Text]
  22. Oei EH, Nikken JJ, Verstijnen AC, Ginai AZ, Myriam Hunink MG. MR imaging of the menisci and cruciate ligaments: a systematic review. Radiology 2003; 226:837-848.[Abstract/Free Full Text]
  23. Deville WL, Bezemer PD, Bouter LM. Publications on diagnostic test evaluation in family medicine journals: an optimal search strategy. J Clin Epidemiol 2000; 53:65-69.[CrossRef][Medline]
  24. Van Houwelingen HC, Zwinderman KH, Stijnen T. A bivariate approach to meta-analysis. Stat Med 1993; 12:2273-2284.[Medline]
  25. Van Houwelingen HC, Arends LR, Stijnen T. Advanced methods in meta-analysis: multivariate approach and meta-regression. Stat Med 2002; 21:589-624.[CrossRef][Medline]
  26. Rao CR. Linear statistical inference and its applications 2nd ed. New York, NY: Wiley, 1973; 417-420.
  27. Liang KY, Zeger SL. Regression analysis for correlated data. Annu Rev Public Health 1993; 14:43-68.[CrossRef][Medline]
  28. Hildebrandt U, Feifel G. Preoperative staging of rectal cancer by intrarectal ultrasound. Dis Colon Rectum 1985; 28:42-46.[Medline]
  29. Konishi F, Muto T, Takahashi H, Itoh K, Kanazawa K, Morioka Y. Transrectal ultrasonography for the assessment of invasion of rectal carcinoma. Dis Colon Rectum 1985; 28:889-894.[Medline]
  30. Beynon J, Foy DM, Roe AM, Temple LN, Mortensen NJ. Endoluminal ultrasound in the assessment of local invasion in rectal cancer. Br J Surg 1986; 73:474-477.[Medline]
  31. Kramann B, Hildebrandt U. Computed tomography versus endosonography in the staging of rectal carcinoma: a comparative study. Int J Colorectal Dis 1986; 1:216-218.[CrossRef][Medline]
  32. Feifel G, Hildebrandt U, Dhom G. Assessment of depth of invasion in rectal cancer by endosonography. Endoscopy 1987; 19:64-67.[Medline]
  33. Glaser F, Schlag P, Herfarth C. Endorectal ultrasonography for the assessment of invasion of rectal tumours and lymph node involvement. Br J Surg 1990; 77:883-887.[Medline]
  34. Orrom WJ, Wong WD, Rothenberger DA, Jensen LL, Goldberg SM. Endorectal ultrasound in the preoperative staging of rectal tumors: a learning experience. Dis Colon Rectum 1990; 33:654-659.[CrossRef][Medline]
  35. Napoleon B, Pujol B, Berger F, Valette PJ, Gerard JP, Souquet JC. Accuracy of endosonography in the staging of rectal cancer treated by radiotherapy. Br J Surg 1991; 78:785-788.[Medline]
  36. Boyce GA, Sivak MV, Jr, Lavery IC, et al. Endoscopic ultrasound in the pre-operative staging of rectal carcinoma. Gastrointest Endosc 1992; 38:468-471.[Medline]
  37. Derksen EJ, Cuesta MA, Meijer S. Intraluminal ultrasound of rectal tumours: a prerequisite in decision making. Surg Oncol 1992; 1:193-198.[CrossRef][Medline]
  38. Herzog U, von Flue M, Tondelli P, Schuppisser JP. How accurate is endorectal ultrasound in the preoperative staging of rectal cancer? Dis Colon Rectum 1993; 36:127-134.[CrossRef][Medline]
  39. Houvenaeghel G, Delpero JR, Giovannini M, et al. Staging of rectal cancer: a prospective study of digital examination and endosonography before and after preoperative radiotherapy. Acta Chir Belg 1993; 93:164-168.[Medline]
  40. Hulsmans FJ, Tio TL, Fockens P, Bosma A, Tytgat GN. Assessment of tumor infiltration depth in rectal cancer with transrectal sonography: caution is necessary. Radiology 1994; 190:715-720.[Abstract/Free Full Text]
  41. Kim NK, Choi JS, Sohn SK, Min JS. Transrectal ultrasonography in preoperative staging of rectal cancer. Yonsei Med J 1994; 35:396-403.[Medline]
  42. Rafaelsen SR, Kronborg O, Fenger C. Digital rectal examination and transrectal ultrasonography in staging of rectal cancer: a prospective, blind study. Acta Radiol 1994; 35:300-304.[Medline]
  43. Adams WJ, Wong WD. Endorectal ultrasonic detection of malignancy within rectal villous lesions. Dis Colon Rectum 1995; 38:1093-1096.[CrossRef][Medline]
  44. Fedyaev EB, Volkova EA, Kuznetsova EE. Transrectal and transvaginal ultrasonography in the preoperative staging of rectal carcinoma. Eur J Radiol 1995; 20:35-38.[CrossRef][Medline]
  45. Starck M, Bohe M, Fork FT, Lindstrom C, Sjoberg S. Endoluminal ultrasound and low-field magnetic resonance imaging are superior to clinical examination in the preoperative staging of rectal cancer. Eur J Surg 1995; 161:841-845.[Medline]
  46. Kaneko K, Boku N, Hosokawa K, et al. Diagnostic utility of endoscopic ultrasonography for preoperative rectal cancer staging estimation. Jpn J Clin Oncol 1996; 26:30-35.[Abstract/Free Full Text]
  47. Nielsen MB, Qvitzau S, Pedersen JF, Christiansen J. Endosonography for preoperative staging of rectal tumours. Acta Radiol 1996; 37:799-803.[Medline]
  48. Akasu T, Sugihara K, Moriya Y, Fujita S. Limitations and pitfalls of transrectal ultrasonography for staging of rectal cancer. Dis Colon Rectum 1997; 40(suppl 10):S10-S15.[CrossRef][Medline]
  49. Hunerbein M, Schlag PM. Three-dimensional endosonography for staging of rectal cancer. Ann Surg 1997; 225:432-438.[CrossRef][Medline]
  50. Maier AG, Barton PP, Neuhold NR, Herbst F, Teleky BK, Lechner GL. Peritumoral tissue reaction at transrectal US as a possible cause of overstaging in rectal cancer: histopathologic correlation. Radiology 1997; 203:785-789.[Abstract/Free Full Text]
  51. Sailer M, Leppert R, Kraemer M, Fuchs KH, Thiede A. The value of endorectal ultrasound in the assessment of adenomas, T1- and T2-carcinomas. Int J Colorectal Dis 1997; 12:214-219.[CrossRef][Medline]
  52. Massari M, De Simone M, Cioffi U, Rosso L, Chiarelli M, Gabrielli F. Value and limits of endorectal ultrasonography for preoperative staging of rectal carcinoma. Surg Laparosc Endosc 1998; 8:438-444.[CrossRef][Medline]
  53. Akahoshi K, Kondoh A, Nagaie T, et al. Preoperative staging of rectal cancer using a 7.5 MHz front-loading US probe. Gastrointest Endosc 2000; 52:529-534.[CrossRef][Medline]
  54. Carmody BJ, Otchy DP. Learning curve of transrectal ultrasound. Dis Colon Rectum 2000; 43:193-197.[CrossRef][Medline]
  55. Kim JC, Cho YK, Kim SY, Park SK, Lee MG. Comparative study of three-dimensional and conventional endorectal ultrasonography used in rectal cancer staging. Surg Endosc 2002; 16:1280- 1285.[CrossRef][Medline]
  56. Steele SR, Martin MJ, Place RJ. Flexible endorectal ultrasound for predicting pathologic stage of rectal cancers. Am J Surg 2002; 184:126-130.[CrossRef][Medline]
  57. Garcia-Aguilar J, Pollack J, Lee SH, et al. Accuracy of endorectal ultrasonography in preoperative staging of rectal tumors. Dis Colon Rectum 2002; 45:10-15.[CrossRef][Medline]
  58. Okizuka H, Sugimura K, Ishida T. Preoperative local staging of rectal carcinoma with MR imaging and a rectal balloon. J Magn Reson Imaging 1993; 3:329-335.[Medline]
  59. Schnall MD, Furth EE, Rosato EF, Kressel HY. Rectal tumor stage: correlation of endorectal MR imaging and pathologic findings. Radiology 1994; 190:709-714.[Abstract/Free Full Text]
  60. Okizuka H, Sugimura K, Yoshizako T, Kaji Y, Wada A. Rectal carcinoma: prospective comparison of conventional and gadopentetate dimeglumine enhanced fat-suppressed MR imaging. J Magn Reson Imaging 1996; 6:465-471.[Medline]
  61. Pegios W, Vogl J, Mack MG, et al. MRI diagnosis and staging of rectal carcinoma. Abdom Imaging 1996; 21:211-218.[CrossRef][Medline]
  62. Hadfield MB, Nicholson AA, MacDonald AW, et al. Preoperative staging of rectal carcinoma by magnetic resonance imaging with a pelvic phased-array coil. Br J Surg 1997; 84:529-531.[CrossRef][Medline]
  63. Vogl TJ, Pegios W, Mack MG, et al. Accuracy of staging rectal tumors with contrast-enhanced transrectal MR imaging. AJR Am J Roentgenol 1997; 168:1427-1434.[Abstract/Free Full Text]
  64. Drew PJ, Farouk R, Turnbull LW, Ward SC, Hartley JE, Monson JR. Preoperative magnetic resonance staging of rectal cancer with an endorectal coil and dynamic gadolinium enhancement. Br J Surg 1999; 86:250-254.[CrossRef][Medline]
  65. Kim NK, Kim MJ, Park JK, Park SI, Min JS. Preoperative staging of rectal cancer with MRI: accuracy and clinical usefulness. Ann Surg Oncol 2000; 7:732-737.[Abstract]
  66. Beynon J, Mortensen NJ, Foy DM, Channer JL, Virjee J, Goddard P. Pre-operative assessment of local invasion in rectal cancer: digital examination, endoluminal sonography or computed tomography? Br J Surg 1986; 73:1015-1017.[Medline]
  67. Rifkin MD, Wechsler RJ. A comparison of computed tomography and endorectal ultrasound in staging rectal cancer. Int J Colorectal Dis 1986; 1:219-223.[CrossRef][Medline]
  68. Saitoh N, Okui K, Sarashina H, Suzuki M, Arai T, Nunomura M. Evaluation of echographic diagnosis of rectal cancer using intrarectal ultrasonic examination. Dis Colon Rectum 1986; 29:234-242.[Medline]
  69. Alessi G, Giuliani A, Risi D, Meli C, Schillaci A, Stipa S. Intrarectal ultrasonography in the preoperative staging of rectal cancer. Ital J Surg Sci 1987; 17:239-244.[Medline]
  70. Beynon J, Roe AM, Foy DM, Channer JL, Virjee J, Mortensen NJ. Preoperative staging of local invasion in rectal cancer using endoluminal ultrasound. J R Soc Med 1987; 80:23-24.[Abstract]
  71. Di Candio G, Mosca F, Campatelli A, Cei A, Ferrari M, Basolo F. Endosonographic staging of rectal carcinoma. Gastrointest Radiol 1987; 12:289-295.[CrossRef][Medline]
  72. Holdsworth PJ, Johnston D, Chalmers AG, et al. Endoluminal ultrasound and computed tomography in the staging of rectal cancer. Br J Surg 1988; 75:1019-1022.[Medline]
  73. Salo JO, Mecklin JP, Jarvinen H, Lehtonen T. Endo-ultrasonography for preoperative staging of rectal cancer. Ann Chir Gynaecol 1988; 77:57-60.[Medline]
  74. Rifkin MD, Ehrlich SM, Marks G. Staging of rectal carcinoma: prospective comparison of endorectal US and CT. Radiology 1989; 170:319-322.[Abstract/Free Full Text]
  75. Rotte KH, Kluhs L, Kleinau H, Kriedemann E. Computed tomography and endosonography in the preoperative staging of rectal carcinoma. Eur J Radiol 1989; 9:187-190.[Medline]
  76. Dershaw DD, Enker WE, Cohen AM, Sigurdson ER. Transrectal ultrasonography of rectal carcinoma. Cancer 1990; 66:2336-2340.[CrossRef][Medline]
  77. Jochem RJ, Reading CC, Dozois RR, Carpenter HA, Wolff BG, Charboneau JW. Endorectal ultrasonographic staging of rectal carcinoma. Mayo Clin Proc 1990; 65:1571-1577.[Medline]
  78. Goldman S, Arvidsson H, Norming U, Lagerstedt U, Magnusson I, Frisell J. Transrectal ultrasound and computed tomography in preoperative staging of lower rectal adenocarcinoma. Gastrointest Radiol 1991; 16:259-263.[CrossRef][Medline]
  79. Lindmark G, Elvin A, Pahlman L, Glimelius B. The value of endosonography in preoperative staging of rectal cancer. Int J Colorectal Dis 1992; 7:162-166.[CrossRef][Medline]
  80. Scialpi M, Andreatta R, Agugiaro S, Zottele F, Niccolini M, Dalla PF. Rectal carcinoma: preoperative staging and detection of postoperative local recurrence with transrectal and transvaginal ultrasound. Abdom Imaging 1993; 18:381-389.[CrossRef][Medline]
  81. Nielsen MB, Pedersen JF, Christiansen J. Rectal endosonography in the evaluation of stenotic rectal tumors. Dis Colon Rectum 1993; 36:275-279.[CrossRef][Medline]
  82. Thaler W, Watzka S, Martin F, et al. Preoperative staging of rectal cancer by endoluminal ultrasound vs. magnetic resonance imaging: preliminary results of a prospective, comparative study. Dis Colon Rectum 1994; 37:1189-1193.
  83. Barbaro B, Valentini V, Manfredi R. Combined modality staging of high risk rectal cancer. Rays 1995; 20:165-181.[Medline]
  84. Osti MF, Padovan FS, Pirolli C, et al. Comparison between transrectal ultrasonography and computed tomography with rectal inflation of gas in preoperative staging of lower rectal cancer. Eur Radiol 1997; 7:26-30.[CrossRef][Medline]
  85. Heneghan JP, Salem RR, Lange RC, Taylor KJ, Hammers LW. Transrectal sonography in staging rectal carcinoma: the role of gray-scale, color-flow, and Doppler imaging analysis. AJR Am J Roentgenol 1997; 169:1247-1252.[Abstract/Free Full Text]
  86. Lindmark GE, Kraaz WG, Elvin PA, Glimelius BL. Rectal cancer: evaluation of staging with endosonography. Radiology 1997; 204:533-538.[Abstract/Free Full Text]
  87. Barbaro B, Schulsinger A, Valentini V, Marano P, Rotman M. The accuracy of transrectal ultrasound in predicting the pathological stage of low-lying rectal cancer after preoperative chemoradiation therapy. Int J Radiat Oncol Biol Phys 1999; 43:1043-1047.[CrossRef][Medline]
  88. Adams DR, Blatchford GJ, Lin KM, Ternent CA, Thorson AG, Christensen MA. Use of preoperative ultrasound staging for treatment of rectal cancer. Dis Colon Rectum 1999; 42:159-166.[CrossRef][Medline]
  89. Harewood GC, Wiersema MJ, Nelson H, et al. A prospective, blinded assessment of the impact of preoperative staging on the management of rectal cancer. Gastroenterology 2002; 123:24-32.[CrossRef][Medline]
  90. Netri G, Coco C, Valentini V, et al. Clinical staging of rectal cancer: results of a prospective continuing study. Ital J Surg Sci 1985; 15:169-174.[Medline]
  91. Thompson WM, Halvorsen RA, Foster WL, Roberts L, Gibbons R. Preoperative and postoperative CT staging of rectosigmoid carcinoma. AJR Am J Roentgenol 1986; 146:703-710.[Abstract/Free Full Text]
  92. Cellini N, Coco C, Maresca G, et al. Clinical staging of rectal cancer: a study on 126 patients. Rays 1986; 11:69-79.
  93. Angelelli G, Macarini L, Lupo L, Caputi-Jambrenghi O, Pannarale O, Memeo V. Rectal carcinoma: CT staging with water as contrast medium. Radiology 1990; 177:511-514.[Abstract/Free Full Text]
  94. Civelli EM, Gallino G, Mariani L, et al. Double-contrast barium enema and computerised tomography in the pre-operative evaluation of rectal carcinoma: are they still useful diagnostic procedures? Tumori 2000; 86:389-392.[Medline]
  95. 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; 137:447-451.[Abstract/Free Full Text]
  96. Murano A, Sasaki F, Kido C, et al. Endoscopic MRI using 3D-spoiled GRASS (SPGR) sequence for local staging of rectal carcinoma. J Comput Assist Tomogr 1995; 19:586-591.[Medline]
  97. Scialpi M, Rotondo A, Angelelli G. Water enema transvaginal ultrasound for local staging of stenotic rectal carcinoma. Abdom Imaging 1999; 24:132-136.[CrossRef][Medline]
  98. Lupo L, Angelelli G, Pannarale O, Altomare D, Macarini L, Memeo V. Improved accuracy of computed tomography in local staging of rectal cancer using water enema. Int J Colorectal Dis 1996; 11:60-64.[Medline]
  99. Zainea GG, Lee F, McLeary RD, Siders DB, Thieme ET. Transrectal ultrasonography in the evaluation of rectal and extrarectal disease. Surg Gynecol Obstet 1989; 169:153-156.[Medline]
  100. Beynon J, Mortensen NJ, Foy DM, Channer JL, Rigby H, Virjee J. Preoperative assessment of mesorectal lymph node involvement in rectal cancer. Br J Surg 1989; 76:276-279.[Medline]
  101. Hildebrandt U, Klein T, Feifel G, Schwarz HP, Koch B, Schmitt RM. Endosonography of pararectal lymph nodes: in vitro and in vivo evaluation. Dis Colon Rectum 1990; 33:863-868.[CrossRef][Medline]
  102. Milsom JW, Graffner H. Intrarectal ultrasonography in rectal cancer staging and in the evaluation of pelvic disease: clinical uses of intrarectal ultrasound. Ann Surg 1990; 212:602-606.[Medline]
  103. Konishi F, Ugajin H, Ito K, Kanazawa K. Endorectal ultrasonography with a 7.5 MHz linear array scanner for the assessment of invasion of rectal carcinoma. Int J Colorectal Dis 1990; 5:15-20.[CrossRef][Medline]
  104. Milsom JW, Lavery IC, Stolfi VM, et al. The expanding utility of endoluminal ultrasonography in the management of rectal cancer. Surgery 1992;