Diveriticulitis versus Colon Cancer: Differentiation

Kedar N. Chintapalli, MD Shailendra Chopra, MD, MRCP, FRCR Abraham A. Ghiatas, MD Christine C. Esola, MD Steven F. Fields, MD Gerald D. Dodd III, MD

Index terms: Colon, diverticula, 75.273 Colon, neoplasms, 75.321 Computed tomography (CT), helical 75.12115, 75.12143 Radiology 1999; 210:429–435 Abbreviation: ROC ⫽ receiver operating characteristic 1

From the Department of Radiology, University of Texas Health Science Center at San Antonio, 7703 Floyd Curl Dr, San Antonio, TX 78284-7800. From the 1996 RSNA scientific assembly. Received February 25, 1998; revision requested May 4; revision received May 27; accepted August 10. Address reprint requests to K.N.C. r RSNA, 1999

Diveriticulitis versus Colon Cancer: Differentiation with Helical CT Findings1 PURPOSE: To retrospectively determine the computed tomographic (CT) signs associated with diverticulitis or colon cancer, and to prospectively apply these signs in the differentiation of these diseases. MATERIALS AND METHODS: Fifty-eight CT scans in cases of proved diverticulitis (n ⫽ 27) or colon cancer (n ⫽ 31) were evaluated retrospectively by radiologists (blinded to the proved diagnosis) for the presence or absence of previously reported CT findings. These findings were correlated with the pathologic or final clinical diagnosis. The retrospective findings were applied in a prospective evaluation with a new set of 72 CT scans. The same five radiologists rendered a diagnosis according to a five-point confidence scale from definitely benign to definitely malignant. Individual and consensus readings were correlated with the final diagnosis. RESULTS: Retrospectively, pericolonic inflammation (P ⬍ .01) and segment involvement greater than 10 cm (P ⬍ .012) were the most significant findings for diverticulitis; pericolonic lymph nodes (P ⬍ .0001) and luminal mass (P ⬍ .003) were the most significant findings for colon cancer. Prospectively, an unequivocal diagnosis was made correctly in 16 (40%) of 40 cases of diverticulitis and 21 (66%) of 32 cases of colon cancer. CONCLUSION: When there are no pericolonic lymph nodes adjacent to a segment of colonic wall thickening, with pericolonic inflammatory changes, the most likely diagnosis is diverticulitis. When pericolonic lymph nodes are present, with or without pericolonic edema, the most likely diagnosis is colon cancer.

Computed tomography (CT) is frequently used to evaluate colonic diverticulitis and colon cancer. However, the results of a few limited reports (1–3) have indicated that there is an overlap in the imaging appearance of these diseases in about 10% of cases. To date, to our knowledge, there has been only one limited blinded retrospective study (4) and no prospective study to evaluate the degree of overlap between these two conditions or to determine the CT signs that may help in distinguishing diverticulitis and colon cancer. We therefore undertook this study to determine the CT signs helpful in differentiating between diverticulitis and colon cancer and to determine the discriminatory power of these signs when applied prospectively.

MATERIALS AND METHODS Author contributions: Guarantors of integrity of entire study, all authors; study concepts and design, all authors; definition of intellectual content, all authors; literature research, K.N.C., S.C.; clinical studies, all authors; data acquisition, K.N.C., S.F.F.; data analysis, all authors; statistical analysis, K.N.C., S.C., G.D.D.; manuscript preparation, K.N.C., S.C.; manuscript editing, K.N.C., S.C., C.C.E., G.D.D.; manuscript review, all authors.

The study consisted of retrospective and prospective components.

Analysis of CT Findings in the Retrospective Component From our radiology database, 58 CT scans from 57 patients (27 cases of diverticulitis and 31 cases of colon cancer) were identified for inclusion in the study. There were 40 men and 17 women aged 17–71 years (mean age, 57 years). One patient with cancer in the sigmoid colon also had a metachronous cancer in the descending colon and therefore had two CT scans that were included in the study. In 27 of the 30 patients with colon cancer, the final diagnosis was reached by means of surgery or colonoscopy. The remaining three cases of 429

colon cancer were diagnosed with biopsy of liver metastases. In 23 of the 27 patients with diverticulitis, the diagnosis was confirmed with surgery or colonoscopy. The remaining four cases of diverticulitis were diagnosed with clinical follow-up. CT examinations were performed with a variety of CT scanners (HiSpeed Advantage, GE Medical Systems, Milwaukee, Wis; PQ Helical, Picker International, Highland Heights, Ohio; Tomoscan TX60, Phillips Medical Systems, Shelton, Conn; and model DR-3, Siemens, Erlangen, Germany). All patients had received oral contrast material (1,000 mL of 2%–4% meglumine and sodium diatrizoate [Gastrografin, Bracco Diagnostics, Princeton, NJ or MD-Gastroview, Mallinckrodt, St Louis, Mo]) 2–21⁄2 hours before the CT study, and 49 patients (85%) received 400–500 mL of 2%–4% rectal contrast material (meglumine and sodium diatrizoate). Scanning was performed in the craniocaudal direction, beginning at the dome of the right hemidiaphragm, 40–70 seconds after the administration of 125–150 mL of ionic or nonionic iodinated contrast medium (350 mg/mL iohexol [Omnipaque; Nycomed, Princeton, NJ], ioversol [Optiray 320; Mallinckrodt], or iothalamate meglumine [Conray 60; Mallinckrodt]), injected at rates varying between 2 and 3 mL/sec. Scanners were collimated to 8–10 mm, with contiguous scans obtained down to the symphysis pubis. The CT images depicting a colonic abnormality were selected for evaluation. The names and identification markers on the images were masked. The selected CT images were grouped together in a random fashion. The images were then evaluated individually by five board-certified radiologists (K.N.C., S.C., A.A.G., C.C.E., G.D.D.) (blinded to the proved diagnosis) for published (1,2,3–6) CT findings of diverticulitis and colon cancer. These included length of segment involved, pericolonic inflammation, pericolonic edema, luminal mass, pericolonic lymph nodes, and extraluminal fluid or air (Fig 1). The length of the involved segment of the colon was recorded as less than 5 cm, 5–10 cm, and greater than 10 cm. Pericolonic inflammation was considered to be present when strandlike areas of hyperattenuation were seen traversing the pericolonic fat. Pericolonic edema was considered to be present when the pericolonic fat had increased attenuation that ranged from mild haziness to a frankly fluid attenuation. A luminal mass was considered to be present if there was a soft-tissue mass in the lumen that arose from the bowel wall. Lymph nodes in the pericolonic fat were grouped by the short430 • Radiology • February 1999

Figure 1. Diagram shows the CT findings (arrows) evaluated in the retrospective and prospective parts of this study. Findings that are self-explanatory, such as length of segment and pericolonic lymph nodes, are not included.

axis diameter of the largest lymph node into those measuring less than 1 cm, those measuring 1–2 cm, and those measuring greater than 2 cm. Presence of fluid or air outside the colonic lumen was recorded. Each radiologist was asked to identify the presence or absence of the CT findings in each case. Results were recorded on standardized worksheets. The majority opinion in each case was tabulated. The data thus obtained were correlated with the proved diagnoses. The prevalence of all CT findings was calculated in each of the two conditions. The statistical significance was calculated by using the Fisher exact test. The most useful signs for differentiating between diverticulitis and

colon cancer were thus identified (Table 1). After the statistical tests were performed, all cases were reviewed jointly by the five readers as an educational exercise to maximize performance on the prospective portion of the study. At the time of the joint image review, additional findings not included in the individual readings were identified and discussed. These included shoulder formation, eccentric wall thickening, and fluid at the root of the mesentery. On the basis of the perceived potential diagnostic value, it was decided to include these findings in the prospective study. Shoulder formation was considered to be present when there was a soft-tissue mass in the wall similar to Chintapalli et al

TABLE 1 Prevalence of CT Findings in Colon Cancer and Diverticulitis: Retrospective and Prospective Comparisons Diverticulitis*

Colon Cancer*

P Values

CT Finding

Retrospective (n ⫽ 27)

Prospective (n ⫽ 40)

Retrospective (n ⫽ 31)

Prospective (n ⫽ 32)

Length of involved segment ⬍5 cm 5–10 cm ⬎10 cm Eccentric wall thickening Luminal mass Shoulder formation Pericolonic inflammation Pericolonic edema Pericolonic lymph nodes Extraluminal fluid or air Fluid in the root of mesentery

1 (4) 20 (74) 6 (22) NE 1 (4) NE 22 (81) 20 (74) 4 (15) 7 (26) NE

2 (5) 30 (75) 8 (20) 10 (25) 3 (8) 1 (3) 38 (95) 31 (78) 3 (8) 12 (30) 20 (50)

8 (25) 22 (71) 1 (3) NE 11 (35) NE 15 (48) 17 (55) 22 (71) 1 (3) NE

10 (32) 21 (66) 1 (3) 16 (50) 15 (47) 16 (50) 21 (66) 11 (34) 23 (72) 2 (6) 3 (9)

Retrospective

Prospective

⬍.012

⬍.0025

NE ⬍.003 NE ⬍.01 .10 ⬍.0001 ⬍.02 NE

⬍.026 ⬍.0002 ⬍.0001 ⬍.0005 ⬍.0003 ⬍.0001 ⬍.011 ⬍.0002

Note.—NE ⫽ not evaluated in the retrospective study. * Data are the number of cases. Numbers in parentheses are percentages.

the apple core lesion described on contrast (barium or meglumine and sodium diatrizoate) enema studies. Wall thickening was considered to be eccentric when there was asymmetry in the thickening of the two walls of the colon. Whenever thickening or fluid in the root of the sigmoid mesentery was seen, it was recorded.

Application of CT Findings in the Prospective Component All consecutive patients with diverticulitis or colon cancer who underwent helical CT of the abdomen in our department between April 1996 and June 1997 were evaluated. There were a total of 83 CT scans in 82 patients. Eleven cases were excluded because of inadequate followup. The final diagnosis was established with surgery in 47 cases, colonoscopy in 19, biopsy of metastasis in two, and final clinical follow-up of more than 8 months in four cases. The four with clinical followup were cases of diverticulitis. The CT findings evaluated from our retrospective analysis were prospectively applied to 72 scans in 71 patients. There were 42 men and 29 women aged 24–90 (mean, 54 years). CT was performed with two types of CT scanners (HiSpeed Advantage, GE Medical Systems; and PQ Helical, Picker International). All patients received oral contrast material (1,000 mL of 2%–4% meglumine and sodium diatrizoate). Rectal contrast material (400–500 mL of 2%–4% meglumine and sodium diatrizoate) was administered in 65 patients (90%). The scanning was initiated at the dome of the right hemidiaphragm 65 seconds after the intravenous administraVolume 210 • Number 2

tion of 125–150 mL of ionic or nonionic iodinated contrast material (iohexol, ioversol, or iothalamate meglumine), with use of a mechanical injector (model CT 9000, Liebel-Flansheim, Cincinnati, Ohio, or Invision, Medrad, Pittsburgh Pa), at a rate of 2–5 mL/sec. Scanners were collimated to 7–8 mm, and a pitch of 1 was used throughout the abdomen and pelvis. Helical scans were obtained through the liver followed by axial scans through the lower abdomen and pelvis. The CT scans were individually read by the same five readers as in the retrospective part of the study. Each reader identified the CT signs present and rendered a diagnosis in each case. The evaluated signs included those of the retrospective study and the three additional signs identified in the joint review. Each reader indicated his or her confidence in each case by marking a five-point confidence scale: 1, definitely diverticulitis; 2, probably diverticulitis; 3, indeterminate; 4, probably malignant; and 5, definitely malignant. The final diagnoses in these patients were then identified. The data were analyzed in two ways. The individual CT diagnoses were correlated with the final diagnosis in each case (Table 2). There were thus a total of 360 CT readings (five readers ⫻ 72 studies) that were used to determine the overlap in each of the five confidence categories. Additionally, the majority (three or more readers) CT diagnosis in each patient was determined from the diagnoses rendered by the individual readers. This was then correlated against the final diagnosis in each patient. The majority opinion was also tabulated for individual findings evaluated in each case.

Statistical analysis consisted of calculating sensitivity, specificity, positive and negative predictive values, accuracy, and P values by using the Fisher exact test, and the Cronbach coefficient ␣ was used for interobserver agreement. Receiver operating characteristic (ROC) curves were also drawn for individual readers and for the group. Multivariate logistic regression was used to determine the combination of CT findings for which the accuracy of diagnosis was the highest.

RESULTS Analysis of CT Findings in the Retrospective Component There were 27 cases of diverticulitis and 31 cases of colon cancer. The prevalence of the findings studied is shown in Table 1. The length of the involved segment was 5–10 cm in 20 cases (74%), longer than 10 cm in six cases (22%), and shorter than 5 cm in one (4%) of the 27 cases with diverticulitis. The length of the involved segment was 5–10 cm in 22 (71%), shorter than 5 cm in eight (25%), and longer than 10 cm in one (3%) of the 31 cases of colon cancer. Pericolonic inflammation and pericolonic edema were present in 22 (81%) and 20 (74%) of 27 cases of diverticulitis, respectively, and 15 (48%) and 17 (55%) of 31 cases of colon cancer, respectively. Presence of pericolonic lymph nodes was noted in 22 (71%) cases of colon cancer and four (15%) cases of diverticulitis. Luminal mass was seen in 11 (35%) cases of colon cancer and one (4%) case of diverticulitis. Applying the statistical analysis, the most specific findings for diverticulitis

Diveriticulitis versus Colon Cancer at Helical CT • 431

Figure 2. Typical acute sigmoid diverticulitis in a 54-year-old man. CT image of the pelvis shows colonic wall thickening (curved arrow), pericolonic inflammation (white arrow), and fluid at the root of the sigmoid mesentery (open arrows). These findings are characteristic of sigmoid diverticulitis.

Figure 3. Typical colon cancer in a 52-year-old man. CT image of the pelvis shows a circumferential colonic mass with shoulder formation (straight arrow) involving a short segment (⬍5 cm). A single enlarged lymph node (curved arrow) is present in the sigmoid mesocolon. Colon cancer with metastasis to lymph nodes was confirmed at surgery.

were presence of pericolonic inflammation (P ⬍ .01) and length of segment longer than 10 cm (P ⬍ .012). The most specific findings for colon cancer were presence of lymph nodes (P ⬍ .001) and luminal mass (P ⬍ .01). Pericolonic edema (P ⫽ .10) favored diverticulitis, but the numbers were not large enough to attain statistical significance (Table 1).

Confidence Category

Reader 1

Reader 2

Reader 3

Reader 4

Reader 5

Total

Group*

1, definitely diverticulitis 2, probably diverticulitis 3, indeterminate 4, probably malignant 5, definitely malignant

20/0 13/1 8/3 0/10 0/17

26/3 5/4 3/1 4/2 2/22

14/0 15/2 8/3 1/6 2/21

9/0 23/2 7/1 1/9 0/20

11/0 21/2 4/4 4/14 0/12

80/3 77/11 30/12 10/41 4/92

16/0 16/2 7/4 1/5 0/21

Application of CT Findings in the Prospective Component There were 40 cases of diverticulitis and 32 cases of colon cancer. Diverticulitis involved the sigmoid colon in 34 cases, descending colon in four, and transverse colon in two. Colon cancer involved the sigmoid in 23 cases, descending colon in two, transverse colon in two, and ascending colon in five. When all individual readings were taken into consideration, there were 360 readings (72 ⫻ 5 ⫽ 360). The correlation of these readings with the final diagnosis is shown in Table 2. There were 83 readings of definite diverticulitis. Of these, 80 (96%) were cases of diverticulitis and three (4%) were cases of colon cancer. All three readings of colon cancer in the definitely benign category were by a single reader. There were 96 readings of definite colon cancer. Of these, 92 (96%) were cases of colon cancer and four readings were cases of diverticulitis. Thus, for the individual readings, the accuracy of CT in making a 432 • Radiology • February 1999

TABLE 2 Categorization of Diverticulitis (n ⴝ 200) and Colon Cancer (n ⴝ 160) by Individual Readers: Prospective Component

Note.—Data are presented as number of diverticulitis cases/number of colon cancer cases. * Group refers to the majority opinion (three or more of the five readers).

correct unequivocal diagnosis of diverticulitis or colon cancer was 96%. However, an unequivocal diagnosis could be made in only 49% of all cases, whereas in the remaining 51%, there was overlap in the CT appearances of the two entities. The correlation of majority CT diagnoses with the final diagnoses is also shown in Table 2. Sixteen cases were read as definite diverticulitis and all had diverticulitis. Eighteen cases were read as probable diverticulitis; of these, 16 had diverticulitis and two had colon cancer. Eleven cases were read as indeterminate; of these, seven (64%) had diverticulitis and four (36%) had colon cancer. Six cases were read as probable colon cancer; of these, one (17%) had diverticulitis and five (83%) had colon cancer. Twenty-one cases were called definite colon cancer. All of these had colon cancer. From the above,

an unequivocal and correct diagnosis of diverticulitis or colon cancer was rendered in 37 (51%) of 72 cases. In the remaining 35 cases (49%), there was enough overlap in the findings to make the CT diagnosis equivocal. Representative cases are illustrated in Figures 2–7. The unequivocal diagnosis of diverticulitis and colon cancer had a high interobserver agreement (Cronbach coefficient ␣) of 95.6%. The area under ROC curves for the group and the five individual readers was 96% and 87%–98%, respectively (Fig 8). Statistical analysis of the data (Table 1) showed that fluid at the root of the mesentery (P ⬍ .0002), pericolonic edema (P ⬍ .0003), and pericolonic inflammation (P ⬍ .0005) had the highest specificity for diverticulitis. Pericolonic lymph nodes (P ⬍ .0001), shoulder formation Chintapalli et al

Figure 4. Atypical diverticulitis in a 20-year-old man. CT image of the pelvis demonstrates multiple sigmoid colon diverticula (straight arrows), eccentric wall thickening (curved arrows), and pericolonic inflammatory changes (arrowheads). Eccentric wall thickening can be seen in both colon cancer and diverticulitis, thus precluding an unequivocal diagnosis of diverticulitis.

Figure 5. Atypical colon cancer in a 53-year-old man. CT image of the pelvis shows a long segment of wall thickening (short arrows), pericolonic inflammatory changes (curved arrows), and a small lymph node (long arrow). The two oval areas on the right (open arrows) are arteries. The long segment of involvement and inflammatory changes suggest a diagnosis of diverticulitis over colon cancer.

Figure 6. Atypical diverticulitis in a 46-year-old woman. CT image of the midabdomen shows a short segment of narrowing with shoulder formation (black arrows) of transverse colon and minimal inflammatory changes in adjacent fat (curved arrow). The short segment and shoulder formation erroneously suggest cancer.

Figure 7. Atypical colon cancer in a 64-year-old man. CT image of the pelvis shows focal wall thickening (black arrows) and extraluminal fluid and air collection involving the left iliopsoas muscle (arrowheads) and extending to the abdominal wall (curved arrow). The inflammatory changes and pericolonic abscess suggest diverticulitis and mask the underlying cancer.

(P ⬍ .0001), and luminal mass (P ⬍ .0002) had the highest specificity for colon cancer. Presence of pericolonic edema, while not statistically significant in the retrospective review, became highly significant in the prospective study. Analysis of the frequency of the associated CT findings showed that, in diverticulitis, pericolonic inflammation or Volume 210 • Number 2

length of involvement of greater than 10 cm yielded a sensitivity of 95%, while fluid at the root of the mesentery or extraluminal air or fluid yielded a specificity of 90% and 94%, respectively (Table 3). However, the accuracy for diverticulitis was highest for pericolonic inflammation and fluid at the root of the mesentery (68%).

For colon cancer, the involvement of a short segment (⬍5 cm) of colon yielded a sensitivity of 68%, and the presence of pericolonic lymph nodes yielded a sensitivity of 63%. The presence of shoulder formation, pericolonic lymph nodes, and luminal mass had specificities of 97%, 92%, and 92%, respectively. The accuracy in colon cancer was the highest


for presence of pericolonic lymph nodes (79%). Multivariate analysis with use of logistic regression was performed. The predictors that significantly enhanced the diagnostic capability of presence or absence of lymph nodes are pericolonic inflammation (P ⬍ .0015) and edema (P ⬍ .0095). When the lymph nodes were absent and inflammation and/or edema was present, the sensitivity was 92% and accuracy was 86% for diagnosis of diverticulitis. When lymph nodes were present with or without pericolonic edema, the sensitivity was 90% and accuracy 79% for colon cancer. Specificity and accuracy for colon cancer were improved when pericolonic inflammation was absent in addition to the other two findings. The results are shown in Table 3.

DISCUSSION Clinically, colon cancer may mimic diverticulitis. CT is performed often as the initial imaging investigation in patients with clinical features suggestive of diverticulitis. The aim is not only to confirm the presence of diverticulitis and detect its complications but also to rule out colon cancer. Differentiation between colon cancer and diverticulitis is fundamental to correct treatment of such patients. Recently, there have been several studies describing the CT features of diverticulitis and colon cancer (1–9). Some of the studies have described an overlap in the CT features of these two diseases (1–4). In a nonblinded study, Balthazar et al (1) describe the atypical findings that mimic cancer in cases of diverticulitis. In that study, they found 16 (10%) of 150 cases of diverticulitis in which CT findings were deemed atypical and needed further evaluation. The points of overlap included wall thickening of more than 1 cm, associated soft-tissue mass, wall thickening with luminal narrowing, wall thickening without pericolonic inflammation, and short segment of wall thickening. In a blinded retrospective study, Padidar et al (4) addressed the CT differentiation of diverticulitis from colon cancer by using the CT findings of the mesenteric venous engorgement and fluid at the base (root) of the mesentery. In their study, fluid at the base of the mesentery had a sensitivity and specificity of 36% and 90%, respectively, for diverticulitis. Vascular engorgement had sensitivity and specificity of 29% and 100%, respectively, for diverticulitis. When fluid and/or engorgement were present, the sensitivity improved to 434 • Radiology • February 1999

Figure 8. ROC curves for the majority (C) and readers 1–5 (R1–R5) are shown on the graph.

TABLE 3 Statistical Analysis of CT Findings for Diverticulitis and Colon Cancer: Prospective Component Sensitivity Specificity PPV* NPV* Accuracy (%) (%) (%) (%) (%)

CT Finding

Diverticulitis Inflammation Fluid at the root Length ⬎10 cm Edema Extraluminal air or fluid Multivariate analysis Lymph nodes absent and edema present Lymph nodes absent and inflammation and/or edema present

95 50 95 76 30

34 90 31 66 94

64 87 63 74 86

84 54 83 70 52

68 68 67 67 58

70

90

90

71

79

92

78

84

89

86

Colon Cancer Luminal Mass Shoulder formation Lymph nodes Length ⬍5 cm Multivariate analysis Lymph nodes present with or without edema Lymph nodes present with or without edema and no inflammation

47 50 63 68

92 97 92 5

83 94 87 37

68 71 76 17

72 76 79 33

90

70

70

90

79

78

92

89

84

86

* PPV ⫽ positive predictive value, NPV ⫽ negative predictive value.

59%. Although these two studies (1,4) provided some useful insight, neither study addressed the complete spectrum of findings in these two diseases. To our knowledge, there are no studies that evaluate the relative values of different CT findings. Our aim in undertaking this study was to identify, through a retrospective review, the CT signs most useful in

differentiating diverticulitis from colon cancer and to test these signs in a prospective manner. From our retrospective review, we found that for the diagnosis of diverticulitis, the most specific findings were pericolonic stranding and length of the involved segment of more than 10 cm. The presence of pericolonic edema favored diverChintapalli et al

ticulitis but did not attain statistical significance. For colon cancer, the most specific findings were the presence of pericolonic lymph nodes and luminal mass. On analyzing our data from the prospective evaluation, the statistical significance for luminal mass for colon cancer and stranding for diverticulitis was even stronger (luminal mass, P ⬍ .0002; stranding, P ⬍ .0005). The P values for the presence of pericolonic lymph nodes (P ⬍ .0001) and shoulder formation (P ⬍ .0001) for colon cancer were highly significant. Despite statistically significant data, however, we found that an overlap of the findings allowed an unequivocal diagnosis in only 16 (40%) of 40 cases of diverticulitis and 21 (66%) of 32 cases of colon cancer. According to the majority (three or more readers) opinion, there was no case of cancer in the ‘‘definite diverticulitis’’ group and no case of diverticulitis in the ‘‘definite colon cancer’’ group, thus yielding an accuracy of 100%. In all other categories (2–4), however, there was an overlap in the two diseases. Analysis of the individual interpretations showed that there were a few interpretive mistakes in the definitely benign or malignant categories. Three cases of colon cancers were called definitely benign and four cases of diverticulitis were called definitely malignant. It is interesting to note that all three colon cancers were missed by one junior faculty in our group. However, ROC curves were similar for each of the five readers, and the diagnosis for the group showed good interobserver agreement (Cronbach coefficient ␣ ⫽ 95.6%). Thus, there was no marked difference in the abilities of the five different readers. Our data show that there is a significant overlap in the CT appearance of colon cancer and diverticulitis. The degree of overlap in our study is much higher than the 10% previously reported by Balthazar et al (1). However, their study was nonblinded, with the cases merely deemed atypical by the readers. The high degree of overlap between diverticulitis and colon cancer detected in our study can be explained on the basis of the

Volume 210 • Number 2

low sensitivity of the CT findings. Fluid at the root of the mesentery and extraluminal air or fluid had sensitivities for diverticulitis of 50% and 30%, respectively. Similarly, luminal mass, shoulder formation, and lymph nodes had sensitivities for colon cancer of 47%, 50%, and 63%, respectively. Thus, the number of cases in which these findings were present were few. When these findings were present, however, the specificity and accuracy of diagnosis were very high. Multivariate logistic regression showed that the presence of the findings of one disease in the absence of the findings of the other had an even higher specificity and added to the confidence of diagnosis. When pericolonic lymph nodes were absent and inflammation and/or edema were present, the specificity for diverticulitis was 90%. When lymph nodes were present and/or no edema and no inflammation were present, the specificity for colon cancer was 92%. The implications of the above findings become clear by looking at the aim of imaging in the context of differentiating between diverticulitis and colon cancer. Imaging is performed to rule out colon cancer in a patient with symptoms of diverticulitis. In this, CT has a low accuracy, with unequivocal diagnosis possible only in 49% of cases. The remaining 51% must undergo colonoscopy or follow-up imaging to clarify the diagnosis. In the context of ruling out cancer, however, it is important that a diagnosis of benign disease have high accuracy. The specificity of the CT findings of diverticulitis and colon cancer is so high that, depending on the reader’s experience, an unequivocal diagnosis carries an accuracy of 100% for both diverticulitis and colon cancer; in such patients, no further diagnostic work-up is required. However, further evaluation may be necessary to plan for surgery. In conclusion, we found statistically significant differences in the frequency of different CT findings in patients with colon cancer and those with diverticulitis. Although there is good statistical cor-

relation, there was considerable overlap (51%) in the CT diagnosis of these two conditions. By using strict criteria, however, one can make a correct unequivocal diagnosis of diverticulitis or cancer in approximately 50% of cases. In those cases, the patients need not undergo further diagnostic evaluation, and further evaluation may be done for surgical planning. When there are no pericolonic lymph nodes adjacent to a segment of colonic thickening, with pericolonic inflammatory changes, diverticulitis is the most likely diagnosis. When there are pericolonic lymph nodes adjacent to a segment of colonic thickening, colon cancer is the most likely diagnosis. References 1. Balthazar EJ, Megibow A, Schinella RA, Gordon R. Limitations in the CT diagnosis of acute diverticulitis: comparison of CT, contrast enema, and pathologic findings in 16 patients. AJR 1990; 154:281–285. 2. Neff CC, vanSonnenberg E. CT of diverticulitis diagnosis and treatment. Radiol Clin North Am 1989; 27:743–752. 3. Balthazar EJ. Diverticular disease. In: Textbook of GI radiology. Gore RM, Levine MS, Laufer I, eds. Philadelphia, Pa: Saunders, 1994; 1072–1095. 4. Padidar AM, Jeffrey RB, Mindelzun RE, Dolph JF. Differentiating sigmoid diverticulitis from carcinoma on CT scans: mesenteric inflammation suggests diverticulitis. AJR 1994; 163:81–83. 5. Freeny PC, Marks WM, Ryan JA, Bolen JW. Colorectal carcinoma evaluation with CT: preoperative staging and detection of postoperative recurrence. Radiology 1986; 158: 347–353. 6. Balthazar EJ, Megibow AJ, Hulnick D, Naidich DP. Carcinoma of the colon: detection and preoperative staging by CT. AJR 1988; 150:301–306. 7. Scharling ES, Wolfman NT, Bechtold RE. Computed tomography evaluation of colorectal carcinoma. Semin Roentgenol 1996; 31:142–153. 8. Gazelle GS, Gaa J, Saini S, Shillito P. Staging of colon carcinoma using water enema CT. J Comput Assist Tomogr 1995; 19:87–91. 9. Chintapalli KN, Esola CC, Chopra S, Ghiatas AA, Dodd GD. Pericolic mesenteric lymph nodes: an aid in distinguishing diverticulitis from cancer of the colon. AJR 1997; 169:1253–1255.
