| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1 Department of Radiology, University of Alabama at Birmingham, 619 S 19th St.
Birmingham, AL 35249.
2 Division of Gastroenterology, University of Alabama at Birmingham, Birmingham,
AL 35249.
3 Department of Biostatistics, University of Alabama at Birmingham, Birmingham,
AL 35249.
Received May 27, 2003;
accepted after revision June 17, 2003.
Address correspondence to C. L. Canon
(ccanon{at}uabmc.edu).
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MATERIALS AND METHODS. Independent double readings of 1,003 consecutive barium enemas (822 double- and 181 single-contrast examinations) were prospectively performed. From this pool of 1,003 examinations, 994 were included in our study. Examinations showing at least one polyp or carcinoma 5 mm or larger were considered to have positive results. For combined readings, results were considered positive if either of the two interpreters reported finding a polyp or carcinoma. A McNemar test was used to compare the first reader's results with the combined results of the first and second readers. Results were retrospectively correlated with endoscopic or surgical results in 360 patients, and agreement between first and combined readings and endoscopic results was determined.
RESULTS. Adding a second reader increased the number of positive results on examinations from 249 to 315 (p < 0.0001) and resulted in potential alteration of clinical treatment in 98 patients (9.9%). Sensitivity of the first and combined readings for detection of all lesions was identical, 76.3% (95% CI, 65.4–87.1%). Specificity decreased from 91.0% (95% CI, 87.9–94.3%) for the first reading to 86.4% (95% CI, 82.2–90.0%) for the combined reading. The overall measurement of agreement decreased from a kappa value of 61.8 (95% CI, 51.2–72.4%) for the first reading to 52.9 (95% CI, 42.2–63.6%) for the combined reading. The second reading required an average of 3.3 min. Sensitivity for the detection of adenocarcinomas was 100%.
CONCLUSION. Although feasible, double reading of barium enemas does not improve sensitivity for detection of polyps and produces a higher false-positive rate.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Single-contrast barium enema compares favorably with the double-contrast study. Sensitivity for polyps larger than 1 cm is similar for the two techniques, 94% and 96%, respectively [3]. For smaller polyps, the sensitivity of the single-contrast examination decreases to 72%, compared with 88% for the double-contrast enema. The accuracy in revealing colorectal cancer is not significantly different for the two methods [4].
Perception error is the most common cause of misdiagnosis of findings in the barium enema examination. Using multiple readers for such examinations has been proposed as a viable solution [5–9]. Retrospective studies have shown improved sensitivity with double reading of barium enemas [7, 8]. Despite these findings, most imaging centers have not implemented this practice, which is perceived as a time-consuming and labor-intensive routine. In the current environment of cost-consciousness, it is difficult to balance productivity with quality assurance. Double reading barium enema examinations would potentially increase sensitivity but would require additional work by the radiologists. Also to be considered is the added cost and risk from endoscopy that might result from an increase in false-positive results.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
All patients underwent a standardized colon preparation, including a clear liquid diet for 24 hr and 10 oz (283.5 g) of magnesium citrate and three bisacodyl tablets administered orally. Both double- and single-contrast barium enemas were included in the study. The radiologist performing the examination chose a single-versus a double-contrast technique on the basis of a patient's age and mobility. Studies using water-soluble contrast material were not included.
Double-contrast barium enemas were performed in a standardized protocol. Spot radiographs included prone and lateral images of the rectum and oblique images of the sigmoid colon, splenic and hepatic flexures, and cecum. Overhead radiographs were obtained with patients in the supine, prone, prone-angled, and both lateral decubitus positions. For the single-contrast examinations, spot radiographs included a supine image of the rectum and oblique images of the sigmoid colon, splenic and hepatic flexures, and cecum. Overhead radiographs were obtained with patients in the left lateral rectum, supine, supine left posterior oblique, and supine-angled positions. Abdominal radiologists, diagnostic radiology residents, and specially trained technologists performed double-contrast studies. Single-contrast barium enemas were performed only by radiologists and radiology residents. All studies were interpreted by two of seven attending radiologists who specialize in abdominal radiology.
First and second readers independently completed a study worksheet, on which they recorded the date of study, name of evaluator, type of study (single- or double-contrast), and findings. The size and location of each lesion were recorded on a diagram of the colon. A factor of two thirds was used to correct all measurements for radiographic magnification. Only polyps that were 5 mm (corrected for magnification) or larger were recorded. Diminutive polyps, those smaller than 5 mm, are considered by most clinicians as insignificant. Such polyps grow slowly and have a 0.01% incidence of invasive carcinoma. Annular carcinomas were also noted. Time required to perform the second reading was noted. All radiologists in the study served both as first and second readers, although never on the same examination.
The first readers analyzed the barium enema examinations, completing a worksheet for each study and placing it in a designated envelope. The images were left hanging on a multiviewer along with available clinical history of the patient. The assigned second readers then performed the second interpretation on the same day, completing a separate but similar worksheet. After each second reading, the second reader compared his or her findings with those of the first reader. Agreement or disagreement was noted on the second reader's worksheet. Cases in which there was disagreement were reviewed and discussed by the two readers, and a final agreement or continued disagreement was recorded. If a consensus interpretation agreed with the second reader's findings, an addendum to the radiology report was dictated, and the referring clinician was notified whenever the change potentially altered clinical treatment. Cases of continuing disagreement were brought to the attention of the referring clinician.
Results of flexible sigmoidoscopy, colonoscopy, and surgery were retrospectively reviewed. All endoscopic examinations were performed by an attending or fellow gastroenterologist, surgeon, or internist during the 6 months before or after performance of the barium enema. Colonoscopy was considered incomplete if the cecum was not reached. The distance reached from the anal verge was recorded. Radiologists were not aware of endoscopic findings when interpreting the barium enema examinations.
Initial and combined radiologic interpretations were evaluated for each patient. Results of a study were considered positive if at least one polyp 5 mm or larger was detected. For statistical purposes, annular tumors and masses were also considered polyps. First and combined readings were compared, and statistical significance was determined using the McNemar test. Using endoscopic or surgical findings as the reference standard, we calculated sensitivity and specificity for polyp detection for first and combined readings.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The average time required for the second reading was 3 min 19 sec: 3 min 31 sec for double-contrast examinations and 2 min 24 sec for single-contrast examinations. The second readers interpreted only the images and did not examine the patients fluoroscopically.
Of the 994 studies, 249 (25.1%) were interpreted as showing positive results on the initial reading. On the combined reading, 315 studies (31.7%) were interpreted as showing positive results, a 26.3% increase in number of positive studies (p < 0.0001). In 98 (9.9%) of the 994 studies, the combined reading resulted in a clinically significant change, either from negative to positive results or identification of a more proximal polyp, necessitating a change in the follow-up endoscopic examination from sigmoidoscopy to colonoscopy.
Of the 360 subjects with endoscopic or surgical follow-up, 344 had comparable findings on the first and combined readings, leaving only 16 cases in which disagreement between the first and combined readings held the possibility of improving the detection of polyps by adding a second reader (Table 1). Hence, with only 4.4% of patients whose findings on the first and combined readings were discordant, there was little opportunity to improve the polyp detection.
|
When we examined the specific combinations of outcomes for the first and combined readings, we found that no polyp was identified on the first and combined readings in 273 cases. In 260 of these cases, endoscopic results concurred with the barium enema results. In the remaining 13 cases, the endoscopist identified a polyp seen by neither the first nor second reader. For 71 of the cases, there was agreement of positive results on a study that was endoscopically confirmed in 44 cases. Therefore, there were 27 false-positive studies with the combined reading (Fig. 1).
|
Only in the 16 cases in which the first and combined readings disagree was there an opportunity for one of these measures to have a better predictive value than the other. In 15 of these 16 cases, the results of the first reading were negative, whereas the results of the combined reading were positive. Of these 15 cases, 14 had negative endoscopic results. In one case, an endoscopically confirmed polyp was identified on the combined reading that was not identified on the first. We did find one case with endoscopically confirmed positive results in which the results of the first reading indicated the presence of a polyp, but the interpretation of the first reader was overturned by the findings at the combined reading.
Traditional measures of agreement can also be made between the first and combined readings and endoscopic results. For the first readings, the sensitivity was 76.3% (95% CI, 65.4–87.1%), the specificity was 91.0% (95% CI, 87.9–94.3%), and the overall measure of agreement (kappa value) was 61.8 (95% CI, 51.2–72.4%). For the combined readings, no change in the sensitivity was noted, but there was a drop in the specificity to 86.4% (95% CI, 82.2–90.0%) and a substantial drop in the kappa value to 52.9% (95% CI, 42.2–63.6%).
Eleven (3.1%) of the 360 polyps evaluated with endoscopy or surgery were proven histologically to be adenocarcinoma. All 11 were correctly identified by both readers. Seven of the 11 carcinomas were confirmed at endoscopic biopsy before surgery. One of the four patients who did not have preoperative endoscopic confirmation underwent colonoscopy that failed to reach a cecal carcinoma seen on barium enema. Two patients underwent flexible sigmoidoscopy that did not reveal a more proximal carcinoma, one in the cecum and another in the hepatic flexure. In the fourth patient, endoscopy was not performed because of a nearly obstructing mass in the descending colon. Five of the 11 cancers were annular; six were polypoid.
A 15-mm rectal polyp in one patient was identified by the first reader. After the second interpretation of the images and subsequent discussion between the two readers, this interpretation was thought to be a false-positive finding. However, at endoscopy, the lesion proved to be a rectal carcinoid. Both radiologic readers incorrectly interpreted one polyp as being in the sigmoid colon. At endoscopy, it was found to be located in the transverse colon, which overlapped the sigmoid colon on all images. This polyp contained a focus of high-grade dysplasia and was surgically resected. There was one false-positive interpretation of a large polyp. Both readers reported a 2.5-cm malignant-appearing rectal lesion (Fig. 2). The endoscopist did not see the lesion, and it was presumed to have been adherent stool.
|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Double reading failed to improve sensitivity in our study. Second readers identified only one polyp that had not been seen by the first reader. Our data show that agreement, as measured by the kappa statistic, between the radiographic endoscopic findings decreased with combined readings. Additionally, false-positive studies were generated, which could lead to increased risk for the patient from a repeated colonoscopy.
Brady et al. [6] discussed the models of double reading. The "believe the positive" model, as used by Markus et al. [7], assumes a positive finding by any one of the readers can result in a conclusion of a positive result. This is the method commonly used in screening studies. We used consensus agreement, and in cases of continued disagreement, we believed the positive. Other methods use arbitration, which is the most sound model. In arbitration, a third party, either aware or unaware of the other two readers' results, reviews the images on which the other two readers disagree. Arbitration would have required us to consult a third radiologist for each of the cases in which the findings of the two readers disagreed.
Although we measured the time it took to read each examination a second time (< 4 min), we did not take into account the second reader's travel time to the outpatient reading site. In some cases, this travel involved walking a considerable distance. It was often difficult for a radiologist to leave a work area to go to another to second-read the barium enema examinations performed that day without being interrupted or having to return to scheduled service. Unrecorded time was also spent by both readers reviewing the discrepancies and discussing disagreements. If it were of greater proven value, despite these considerations, double reading might still be feasible. Only clinically significant and discordant findings might warrant discussion.
Endoscopy is typically used as the reference standard for assessment of abnormalities of the colon; however, it is unfortunately not perfect [10–13]. In one study, 21 (41%) of 51 polyps larger than 1 cm identified on double-contrast barium enema but not at endoscopy proved to be neoplastic [10]. In another study, 12% of polyps larger than 1 cm detected on barium enema were not seen at endoscopy [11].
In addition to the flawed reference standard, our study had other limitations. Both symptomatic and screening patients were included in the study. We included both populations in an attempt to evaluate double reading in the setting of a routine outpatient radiology clinic. However, this inclusion does bias the positive results. Only about one third of the 994 patients underwent endoscopic evaluation, and the radiologic and endoscopic results were compared retrospectively. The endoscopists thought many of the smaller polyps diagnosed on the barium enema did not warrant colonoscopy at the time, and in some cases, they recommended that the patient return for a repeated study in 1 to 2 years, depending on the size of the polyp and clinical circumstances. If all the polyps identified by the second reader were endoscopically evaluated, the results might have been quite different. Ideally, every patient would have undergone endoscopy, and each potential polyp measuring 1 cm or larger not identified at endoscopy would have been addressed with an additional study if both the radiologist and endoscopist still had concerns. We included the single- and double-contrast studies in the data analysis together to increase the total number of cases and also to simulate the typical enema examinations encountered in an outpatient setting.
Double reading of barium enema examinations did not result in detection of a significant number of additional polyps and generated many false-positive findings. Routine double reading is not warranted for screenings using the barium enema examination.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
