Scandinavian Journal of Gastroenterology. 2015; 50: 1045–1053
ORIGINAL ARTICLE
Who should undergo a colonoscopy among patients with incidental colon uptake on PET-CT?
SOO-YOUNG NA1, KYUNG-JO KIM2, SEUNGBONG HAN3, SOYOUNG JIN4, JAE SEUNG KIM4, DONG-HOON YANG2, KEE WOOK JUNG2, BYONG DUK YE2, JEONG-SIK BYEON2, SEUNG-JAE MYUNG2, SUK-KYUN YANG2 & JIN-HO KIM2 1
Department of Internal Medicine, Jeju National University Hospital, Jeju National University School of Medicine, Jeju, Korea, 2Department of Gastroenterology, Asan Medical Center, University of Ulsan College of Medicine, Seoul, Korea, 3 Department of Clinical Epidemiology and Biostatistics, Asan Medical Center, University of Ulsan College of Medicine, Seoul, Korea, and 4Department of Nuclear Medicine, Asan Medical Center, University of Ulsan College of Medicine, Seoul, Korea
Abstract Objectives. To investigate the optimal cut-off of the maximum standard uptake value (SUVmax) for the detection of colorectal neoplasms and to suggest those for whom further colonoscopy is recommended among patients with incidental colonic uptake on positron emission tomography-computed tomography (PET-CT). Materials and methods. In 306 patients who underwent colonoscopy within 3 months of receiving PET-CT between January and December 2009, measurements of the per-patient and per-lesion diagnostic performance of PET-CT for the detection of colonic neoplasms were obtained. Receiver operating characteristic (ROC) analysis was used to identify the SUVmax that provided a high probability of diagnosing malignancy and high-grade dysplasia. Results. The per-patient and per-lesion PET-CT detection sensitivities for malignancies were 93.3% (28/30; 95% confidence interval (CI) 76.5% to 98.9%) and 93.5% (29/31, 95% CI 77.2% to 98.9%), respectively; the sensitivities for high-grade dysplasia were both 90.0% (9/10; 95% CI 54.1% to 99.5%). As a criterion to specifically detect both malignancy and high-grade dysplasia on focal uptake, a SUVmax greater than 2.5 yielded a 92.3% per-lesion sensitivity and a 42.9% per-lesion positive predictive value (PPV). In the ROC curve analysis, a cut-off value of SUVmax = 5.8 was established, at which the sensitivity, PPV and positive likelihood ratio for diagnosing malignancy and high-grade dysplasia were 71.8% (28/39; 95% CI 54.9% to 84.5%), 84.8% (28/33; 95% CI 67.3% to 94.3%) and 6.9, respectively. Conclusion. The optimal cut-off value to identify a malignancy or high-grade dysplasia was SUVmax = 5.8. However, to avoid missing a malignancy or high-grade dysplasia, a colonoscopy should be performed above a SUVmax = 2.5.
Key Words: colonoscopy, colorectal neoplasm, maximum standard uptake value, PET-CT
Introduction The use of positron emission tomography (PET) with 18F-fluoro-2-deoxy-D-glucose (FDG) uptake for tumor staging and monitoring of cancer therapy effects has increased notably [1,2]. The combination of PET and computed tomography (CT) has the advantage of obtaining both metabolic and structural informations; PET-CT provides better discrimination
of the pathological metabolic foci from the physiologic form and localizes the pathologic foci [3–5], and it has already considerably reduced the incidence of falsepositive findings compared to FDG-PET [6,7]. Nevertheless, many patients who have indicated incidental colonic uptake (ICU) on PET-CT may unnecessarily undergo colonoscopy based on the maximum standardized uptake value (SUVmax). Approximately, 13–45% of patients with ICU positivity on
Correspondence: Kyung-Jo Kim, MD, Department of Gastroenterology, Asan Medical Center, University of Ulsan College of Medicine, 88, Olympic-ro 43-gil, Songpa-gu, Seoul 138-736, South Korea. Tel: +82 2 3010 3196. Fax: +82 2 476 6517. E-mail: [email protected]
(Received 10 September 2014; revised 20 November 2014; accepted 23 November 2014) ISSN 0036-5521 print/ISSN 1502-7708 online 2015 Informa Healthcare DOI: 10.3109/00365521.2014.992363
1046
S. -Y. Na et al.
Table I. The indications for PET-CT and colonoscopy in 691 consecutive patients who undergoing colonoscopy after PET-CT within 3 months. No. of patients (%) Reason for PET-CT Oncologic staging or surveillance Screening Cancer of unknown origin Differential diagnosis of malignancy from benign disease Others Reason for colonoscopy Screening Abnormal PET-CT or CT images Baseline study for gynecologic malignancies Gastrointestinal symptoms or signs Cancer of unknown origin Polypectomy surveillance Others
377 216 58 20
(54.6) (31.3) (8.4) (2.9)
20 (2.9) 334 122 78 53 49 44 11
(48.3) (17.7) (11.3) (7.7) (7.1) (6.3) (1.6)
Abbreviation: PET = Positron emission tomography; CT = Computed tomography.
PET-CT showed no demonstrable lesions on colonoscopy [8–11]. Therefore, more information is essential for the optimal use of colonoscopy after obtaining a positive PET-CT scan based on the risk of the patient having an actual advanced colorectal neoplasm. Although a few previous studies have proposed a SUVmax cut-off value for further examination in cases of ICUs [3,12,13], there is still a paucity of data on this topic and a lack of consensus regarding an optimal strategy. Therefore, this study was conducted to identify an optimal cut-off value for the SUVmax of PET-CT for detecting advanced colorectal neoplasms and to suggest a means of determining which patients with ICU on PET-CT would benefit from further colonoscopy. Materials and methods Patients This study was approved by the Institutional Review Board of Asan Medical Center (AMC) (No. 2011-0897). From January to December 2009, 15,907 PET-CTs were performed in 14,370 adult patients at AMC. Of these patients, 691 subsequently underwent colonoscopy within 3 months, including patients who had abnormal PET-CT findings. Data regarding patient demographics, medical histories, indications and results of colonoscopies were extracted. The indications for PET-CT were oncologic staging or surveillance in 377 (54.6%) patients; screening in 216 (31.3%); cancer of unknown primary (CUP) work-up in 58 (8.4%); differential diagnosis of malignancy from benign disease in 20 (2.9%); and
other conditions, such as work-up for ascites, elevated tumor marker, pleural effusion, fever of unknown origin, etc. in 20 (2.9%). The reasons for undergoing colonoscopy were screening in 334 patients (48.3%); further evaluation for abnormal PET-CT or CT images in 122 (17.7%); baseline study for gynecologic malignancies in 78 (11.3%); gastrointestinal symptoms or signs in 53 (7.7%); CUP work-up in 49 (7.1%); polypectomy surveillance in 44 (6.3%); and other reasons in 11 (1.6%). The indications for PET-CT and colonoscopy were simply summarized in Table I. Of those 691 patients, 312 patients were excluded from the study for the following reasons: patients with a history of colorectal cancer including prior colon resection (n = 300) because ICUs are defined as unexpected colorectal findings that are discovered on an imaging study unrelated to the large bowel and colonic resection could not allow an accurate topographic matching between possible PET-CT uptake sites and the colonoscopy findings; patients with an incomplete colonoscopy (n = 6); and patients with incomplete data (n = 6). The technical assessment of PET-CT included all 379 patients. The mean age of the 379 patients was 56.0 ± 11.8 years, and 49.3% (187/379) of the study subjects were male. The median interval from PET-CT to colonoscopy was 5 days (range, 0–81 days). Finally, based on the inclusion and exclusion criteria, 306 patients were included in the analysis of the colonic findings (Figure 1). We only analyzed significant colonic lesions, which included malignancies (primary or metastatic colorectal cancers, lymphomas, and neuroendocrine tumors), advanced adenomas (‡10 mm in diameter with any grade of dysplasia or any size with at least a high-grade or villous architecture) [14], and inflammatory conditions in which the onset of symptoms occurred before the PET-CT. The exclusion criteria, i.e., insignificant colonic lesions, were non-advanced adenomas, non-neoplastic polyps and inflammatory conditions in which the onset of symptoms occurred after the PET-CT. The presence of non-advanced adenomas was not considered a significant finding in this study for the following reasons: only 2.5 polyps per 1000 develop into cancer per year [15]; the size and shape are well-known as adequate predictors of malignancy [16]; and small polyps can be negative on PET due to both partial-volume artifacting and non-FDG-avid metabolism [17]. Protocol, definitions and blinded interpretation of PET-CT PET-CT scanning was performed using three PET-CT camera systems: the Discovery STe (GE Healthcare,
Colonoscopic evaluation for ICU on PET-CT
1047
Consecutive patients who undergoing colonoscopy after PET-CT within 3 months (n = 691) Excluded (n = 312) Colorectal cancer including prior colon rescction (n = 300) Incomplete colonoscopy (n = 6) Incomplete data (n = 6) Cohort for techincal assessment (n = 379) Insignificant lesions on colonoscopy (n = 73)
Cohort for colonic findings analysis (n = 306)
Non-advanced adenomas (n = 51) Non-neoplastic polyps (n = 18) Inflammations occurred after PET-CT (n = 4)
Figure 1. Flow diagram of study patients.
Waukesha, WI, USA) (44.6%, 169/379); the Biograph 16 (Siemens Medical System/CTI, Erlangen, Germany) (23.5%, 89/379); and the Biograph Truepoint 40 (Siemens Medical System/CTI, Erlangen, Germany) (31.9%, 121/379). All patients fasted for more than 6 h prior to scanning. Whole-blood glucose concentration, measured immediately before FDG administration, was below 150 mg/dl for all patients. Image acquisition began approximately 60 min (59.4 ± 6.9 min) after the intravenous injection of 370– 555 MBq (10–15 mCi, 12.4 ± 1.8 mCi) of FDG. Non-contrast-enhanced low-dose spiral CT was performed from the base of the skull to the proximal thighs for attenuation correction and image fusion, followed by a three-dimensional caudocranial PET emission scan. The emission scan time per bed position was 2–3 min, and 6 or 7 bed positions were employed. The PET data were reconstructed iteratively both with and without attenuation correction based on CT data, and they were reoriented in axial, sagittal and coronal slices. The SUVmax reference values were between 0.9 and 2.5 (1.7 ± 0.3) in the liver and between 0.5 and 1.8 (1.2 ± 0.2) in the mediastinum. The bowel was completely unprepped, i.e., not cleansed or distended. To ensure the inhibition of bowel movement, 135 mg of duspatalin was orally administered to all patients. Lean body mass, instead of body weight, was used to assess the FDG activity by SUVmax, which was
measured at the most intense colonic element in patients with focal activity and at the entire element if the activity was diffuse or segmental. To ensure consistent SUVmax measurements, uptake patterns, and location determinations, one associate professor and board-certified nuclear physician (Kim JS) reviewed the PET-CT images with ICU identified on the initial reading while blinded to the colonoscopic findings. Physiologic FDG uptake of the colon was defined as contiguous diffuse uptake without colonic wall abnormality on PET-CT. Abnormal colonic uptake of FDG was defined as any wellcircumscribed foci of increased colonic uptake greater than what was measured in the normal hepatic parenchyma that was not suspected to be related to the primary cancer under evaluation. If abnormal uptake was considered to be present in the colon, the reviewer was required to assign the lesion to one of the five colonic segments: ascending, transverse, descending, sigmoid or rectum. Focal uptake was defined as abnormal nodular uptake, and segmental or diffuse uptake as abnormal colonic wall uptake that was shorter or longer than one colonic segment. Colonoscopy Colonoscopies were performed after bowel preparation with polyethylene glycol 4 l solution, all by
1048
S. -Y. Na et al.
experienced board-certified gastroenterologists, who had at least 1000-colonoscopy case experiences using a video colonoscope (CF 260; Olympus Optical Co., Tokyo, Japan). All significant colonic lesions were biopsied when possible. Each polyp was classified by an experienced pathologist as either an adenoma or another type of polyp. Dysplasia was defined as low-grade dysplasia, high-grade dysplasia or carcinoma according to the Vienna criteria [18]. Combination of the PET-CT and colonoscopic findings Colonoscopy is considered to be the gold standard colon diagnostic procedure. The PET-CT findings were classified on the basis of their colonoscopic findings as follows: a false-negative findings was defined if there was any clinically relevant colonic lesion on the colonoscopy without an abnormal colonic uptake of FDG; a true-negative finding was defined if both the PET-CT and colonoscopy were negative; a false-positive finding was defined if an abnormal colonic uptake had no relevant lesion on colonoscopy; and a true-positive PET-CT finding was defined if lesion-by-lesion matching could be performed between colonic uptake of FDG and colonoscopy in the same segment.
reported p-Values were obtained using two-sided tests and p-Values £ 0.05 were considered statistically significant. Results ICU and per-patient analysis of PET-CT Of the 306 patients who were enrolled in the cohort, both the PET-CT and colonoscopy showed negative findings in 211 patients. The remaining 95 patients had a total of 115 significant colonic lesions on PETCT and/or colonoscopy. Only 103 ICUs of 115 significant colonic lesions were observed in 88 patients; the ICU was 0.6% (88/14,370). The per-patient PET-CT sensitivity for detecting patients with significant colonic lesions was 86.8% (46/53; 95% CI 74.0% to 84.1%) (Table II). In the sub-group analysis, the sensitivities for detecting malignancy and high-grade dysplasia were 93.3% (28/30; 95% CI 76.5% to 98.9%) and 90.0% (9/10; 95% CI 54.1% to 99.5%), respectively. The overall per-patient NPV was 96.8% (211/218; 95% CI 93.2% to 98.6%). The NPVs for malignancy and high-grade dysplasia were 99.1% (216/218; 95% CI 96.4% to 99.8%) and 99.5% (217/218; 97.1% to 99.9%), respectively.
Statistical analysis The performance measures of PET-CT for diagnosing the colonic lesions included per-patient sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV), along with perlesion sensitivity and PPV. Regarding the analysis of the SUVmax on PET-CT suggestive of malignancy and high-grade dysplasia, the per-lesion sensitivities and PPVs of various SUVmax values for diagnosing malignancy and high-grade dysplasia were obtained. The 95% confidence intervals (CIs) were calculated for proportional data. In addition, a receiver operating characteristic (ROC) curve, displaying sensitivity along with the false-positive rate (1-specificity), was used to assess the best positive likelihood ratio (PLR) cut-point of SUVmax for distinguishing malignancy and high-grade dysplasia from low-grade dysplasia or inflammatory process. The associations of SUVmax with the PET-CT uptake pattern and colonic pathologic type were tested using the Mann–Whitney U test and Student’s t-test. The Mann–Whitney U test was also used to compare polyp sizes between the falsenegative group and the true-positive group. Fisher’s exact test was used to compare the sensitivity according to the colonic pathologic types. Statistical analysis was performed using a statistical software package (SPSS, Inc., version 18.0, Chicago, IL, USA). All
Per-lesion analysis of PET-CT according to colonic lesions A total of 115 lesions from either PET-CT or colonoscopy in 95 patients were divided into one of the three groups for analysis: 12 lesions in the falsenegative group, 54 lesions in the false-positive group and 49 lesions in the true-positive group. The perlesion sensitivities of PET-CT according to the type of colonic lesion are shown in Table III. The overall perlesion sensitivity of PET-CT was 80.3% (49/61; 95% CI 67.8% to 90.0%). Based on the final diagnosis, the sensitivities were 93.5% (29/31; 95% CI 77.2% to 98.9%) for malignant lesions and 90.0% (9/10; 95% CI 54.1% to 99.5%) for high-grade dysplasias. Sensitivities were highest in cases of malignancy and highgrade dysplasia, followed by low-grade dysplasia and the inflammatory process, which were significantly different (92.7% vs. 41.7% vs. 75.0%, p < 0.01). Although the size of the adenoma did not differ between the true-positive group and false-negative group (15.8 ± 6.5 mm vs. 17.5 ± 9.6 mm, p = 0.65), high-grade dysplasia showed a higher sensitivity than lower-grade dysplasia (90.0% vs. 41.7%, p = 0.03). The overall SUVmax was higher in the true-positive group than in the false-positive group (8.6 ± 5.0 vs.
Colonoscopic evaluation for ICU on PET-CT
1049
Table II. Per-patient analysis of PET-CT for diagnosing colonic lesions. Advanced adenomas Measures
Overall
Sensitivity, % (95% CI) 86.8 (74.0 No. of patients 46/53 Specificity, % (95% CI) 83.4 (78.1 No. of patients 211/253 PPV, % (95% CI) 52.3 (41.4 No. of patients 46/88 NPV, % (95% CI) 96.8 (93.2 No. of patients 211/218
to 84.1) to 87.7) to 62.9) to 98.6)
Malignancies
High-grade
Low-grade
Inflammatory processes
93.3 (76.5 28/30 78.3 (72.8 216/276 31.8 (22.5 28/88 99.1 (96.4 216/218
90.0 (54.1 to 99.5) 9/10 73.3 (67.8 to 78.2) 217/296 10.2 (5.1 to 20.0) 9/88 99.5 (97.1 to 99.9) 217/218
55.6 (22.7 to 84.7) 5/9 72.1 (66.5 to 77.0) 214/297 5.7 (2.1 to 13.4) 5/88 98.2 (95.1 to 99.4) 214/218
75.0 (35.6 to 95.5) 6/8 72.5 (67.0 to 77.4) 216/298 6.8 (2.8 to 14.8) 6/88 99.1 (96.4 to 99.8) 216/218
to 98.9) to 82.9) to 42.7) to 99.8)
Abbreviations: CI = Confidence interval; NPV = Negative predictive value; PET-CT = Positron emission tomography-computed tomography; PPV = Positive predictive value.
5.0 ± 2.1, p < 0.01). However, the difference in SUVmax was significant only for the focal uptake pattern between the two groups (p < 0.01) in the subanalysis. The final diagnosis in the true-positive group was added under Table IV.
SUVmax cut-off value for PET-CT to suggest malignancy and high-grade dysplasia We then analyzed the SUVmax values only for focal PET-CT uptakes to increase the sensitivity for the suggestion of malignancy or high-grade dysplasia on PET-CT. The per-lesion sensitivity and PPV of PETCT for diagnosing malignancy and high-grade dysplasia varied according to differing SUVmax values. As the cut-off value for SUVmax increased, the sensitivity decreased and the PPV increased. In our data, the lowest SUVmax to detect malignancy or highgrade dysplasia was 2.5. Therefore, a SUVmax higher than 2.5 on PET-CT yielded the highest sensitivity of 92.3% (36/39; 95% CI 78.0% to 98.0%) and a moderate PPV of 42.9% (36/84; 95% CI 32.3% to 54.1%) (Table V). The ROC curve for detecting malignancy and high-grade dysplasia in focal PET-CT uptakes
with the best PLR is shown in Figure 2. A cut-point of 5.8 was identified on the ROC curve, at which the area under the curve, sensitivity, PPV and PLR values were 0.76 (95% CI 0.63 to 0.88), 71.8% (28/39; 95% CI 54.9% to 84.5%), 84.8% (28/33; 95% CI 67.3% to 94.3%), and 6.9, respectively. A more illustrative scatter plot for this cut-off point is shown in Figure 3. The SUVmax for malignancies and adenomas with high-grade dysplasia was significantly higher than the values for false-positive findings, benign lesions and adenomas with low-grade dysplasia (p < 0.01). Discussion Incidental PET-CT uptake has been reported to be between 0.6% and 2.7% in the large bowel [3,12], 1.3% and 3.0% in the gastrointestinal tract [9,10] and 3.0% and 4.1% in the whole body [19,20]. In this study, the prevalence of ICU was 0.6%, which is similar to the values reported in other series; however, we included only those incidental findings evaluated by colonoscopy within 3 months, so our incidental uptake shows a somewhat low level.
Table III. Per-lesion sensitivities of PET-CT according to colonic lesions. Final diagnosis Malignancy Colorectal cancer Lymphoma Metastatic cancer Neuroendocrine tumor Advanced colonic adenoma Size* (cm) High-grade Low-grade Inflammatory process Overall
False-negative (No. of lesions)
True-positive (No. of lesions)
Total (No. of lesions)
2 0 0 0 2 8 15.8 ± 6.5 1 7 2 12
29 22 3 4 0 14 17.5 ± 9.6 9 5 6 49
31 22 3 4 2 22 10 12 8 61
Sensitivity, % (95% CI)
*Sizes (mean ± SE) between the false-negative and true-positive groups were not significantly different (p = 0.65). Abbreviations: CI = Confidence interval; PET-CT = Positron emission tomography-computed tomography.
93.5 100.0 100.0 100.0 0 63.6 90.0 41.7 75.0 80.3
(77.2 to 98.9) (81.5 to 100.0) (31.0 to 100.0) (39.6 to 100.0) (0 to 80.2) (40.8 to 82.0) (54.1 to 99.5) (16.5 to 71.4) (35.6 to 95.5) (67.8 to 90.0)
1050
S. -Y. Na et al.
Table IV. Per-lesion positive predictive values and SUVmax values of PET-CT for uptake patterns between the false-positive and true-positive groups. False-positive Uptake patterns
No. of lesions
Diffuse Segmental Focal Overall
True-positive
SUVmax (mean ± SE)
No. of lesions
± ± ± ±
3* 3† 43‡ 49
4 9 41 54
8.1 5.8 4.5 5.0
4.0 2.8 1.3 2.1
SUVmax (mean ± SE)
p-Value (SUVmax)
± ± ± ±
0.72 0.20 < 0.01 < 0.01
6.9 11.7 8.2 8.6
3.6 8.8 4.2 5.0
*Diffuse PET-CT uptake included cytomegalovirus colitis (1), pseudomembranous colitis (1) and non-specific colitis (1). Segmental PET-CT uptakes included lymphoma (2) and ulcerative colitis (1). ‡ Focal PET-CT uptakes included colorectal cancer (22), advanced colonic adenoma (14), lymphoma (1), metastatic cancer (4), large condyloma (1) and non-specific colitis (1). Abbreviations: PET-CT = Positron emission tomography-computed tomography; SUVmax = Maximum standardized uptake value. †
We found that malignancy and high-grade dysplasia were detected with very high sensitivity in both perpatient and per-lesion analyses. In addition, the NPVs for malignancy and high-grade dysplasia were more than 99% in per-patient analysis. Many previous studies have reported that PET-CT sensitivities ranged from 55% to 90% in colorectal cancer or advanced adenoma [3,8,12,17,21–23]. Our study demonstrated that all colorectal cancers, lymphomas and metastatic cancers had abnormal PET-CT uptakes, whereas two primary neuroendocrine tumors were not found, as in other studies [3,8,23]. It is well-known that neuroendocrine tumors are one of the malignancies that show FDG uptakes at the same rate as the normal surrounding tissue, leading to failures in their identification due to their low potential for malignancy [11]. Low-grade dysplasia and 1.0
Sensitivity
0.8
0.6
0.4
0.2
0.0 0.0
0.2
0.4
0.6
0.8
1.0
1-Specificity Figure 2. Receiver operating characteristic curve for determining the optimal cut-off value of the maximum standard uptake value (SUVmax). The cut-off is defined such that it minimizes the distance between the curve and the upper-left corner (dot). The cut-off point of the SUVmax for detecting malignancies and highgrade dysplasias in focal positron emission tomography-computed tomography uptake was 5.8.
inflammatory processes exhibited relatively low sensitivities but still fairly high NPVs. Adenoma sizes were not significantly different between the falsenegative group and the true-positive group. This finding is similar to that of Nakajo et al. [22], who demonstrated that the histologic grade was the strongest independent factor in the PET identification of colorectal polyps. Only the focal uptake pattern exhibited different SUVmax values between the true-positive and falsepositive groups. In addition, all malignancies, including colorectal cancer and advanced colorectal adenoma, excepting only two lymphomas, exhibited focal uptake patterns. The two lymphomas exhibited segmental uptake patterns. Therefore, we need to focus only the on focal uptake patterns to detect malignancy or high-grade dysplasia to increase the sensitivity. Previous studies [3,12,13,24,25] used a different SUVmax cut-off by choosing the absolute cut-off value for significant ICUs, which ranges from 2.5 to 5.0. We found that colorectal lesions, including malignancies (SUVmax = 2.5, 2.8, 4.1) and highgrade dysplasias (SUVmax = 2.9, 3.2, 4.1), could be detected at a lower level. The cut-off value for the SUVmax of PET-CT for diagnosing malignancies and high-grade dysplasias was 2.5 in our study. This value was the most sensitive cut-point to detect malignancies and high-grade dysplasias, with the exception of only two neuroendocrine tumors and one highgrade dysplasia that were not identified on PET-CT. However, reducing the SUVmax threshold to avoid misdiagnosing cancers with lower SUVmax values would rapidly and substantially decrease the PPV for the diagnosis of cancer or high-grade dysplasia by increasing the mischaracterization of PET-CTdetected lesions and false-positive lesions. This situation would generate a high rate of unnecessary colonoscopies. Accordingly, to obtain a high probability for diagnosing malignancies and high-grade dysplasias on focal PET-CT uptake, we analyzed
Colonoscopic evaluation for ICU on PET-CT
1051
p < 0.01
Maximum standardized uptake value (SUVmax)
p = 0.34 25
20 p = 0.69 15
10
5
8.1 ± 5.0
4.6 ± 1.3
9.75 ± 5.0
Cut-off value = 5.8
4.6 ± 1.1
0 False-positive and benign lesions
Low-grade dysplasias
High-grade dysplasias
Malignancies
Mean ± SE with 95% CI Figure 3. Illustrative scatter plot of positron emission tomography-computed tomography maximum standard uptake value (SUVmax), according to histology. The range of the 95% confidence interval (CI) for malignancy reached above the optimal SUVmax = 5.8 line. Alternatively, the 95% CI for false-positive results was located below the optimal SUVmax line.
the ROC curve. The results revealed that the SUVmax was 5.8, with a PLR of 6.9. The positive PET-CT probability for patients with malignancies or highgrade dysplasias increased as much as 6.9-fold, reflecting a good predictive level of detection [26]. When we used a cut-off value of 5.8, only 0.09% (incidence of ICU (1 - PPV)) [3] of patients who had PET-CT scans would undergo unnecessary colonoscopies for the diagnosis of malignancies or highgrade dysplasias. In addition, this SUVmax value may provide useful information for distinguishing malignancies or high-grade dysplasias from low-grade dysplasias or inflammatory processes. We acknowledge that this SUVmax value is limited to a reference value for patients with a high probability of malignancy or high-grade dysplasia for further colonoscopic evaluation, as its sensitivity may not be acceptable to many clinicians for use as a cut-off value. Therefore, in clinical settings, a multidisciplinary approach is necessary, and both the underlying disease state of patients and other examination findings should be considered before recommending a colonoscopy. This study had several limitations. First, our research was conducted retrospectively, and potential bias accrued from our interpretations of the PETCT data may exist. However, we attempted to reduce this bias by only considering patients who underwent colonoscopies after PET-CT examinations within a 3-month period. Second, although our research consisted of the largest comparable study population of
which we are aware, our final per-lesion analysis included only 115 lesions on PET-CT and/or colonoscopy. Therefore, the validity of the per-lesion analysis should be interpreted within the framework of this limited sample size. Third, we consider colonoscopy the gold standard. Although we focused only on malignancies and high-grade dysplasias, which are relatively large in size, a colonoscopy is not able to detect all polyps. Fourth, advanced adenomas include lesions with villous architecture, so we classified villous adenomas as advanced adenomas. However, villous adenomas were not included in our final per-lesion analysis due to the small sample size of advanced adenomas. Finally, the exclusion of nonadvanced adenomas could create significant problems integrating our findings into current evidence based recommendations for colonoscopy surveillance. Our findings only provide clinical implications for suggesting a SUVmax that may identify patients for whom colonoscopy might serve as a useful evaluation tool. Therefore, we recommend that asymptomatic patients over the age of 50 years with negative PET-CT cancer screenings follow the current recommendations for colonoscopy surveillance. In conclusion, it is clinically common to decide whether patients with ICU identified on PET-CT should undergo a colonoscopy. Although a cut-off value using a ROC curve has limited capability for screening for all malignancies and high-grade dysplasias, the SUVmax could be interpreted in two ways.
1052
S. -Y. Na et al.
Table V. Per-lesion sensitivities and PPVs of various SUVmax values of focal PET-CT uptake for diagnosing malignancy and high-grade dysplasia. Measures
SUVmax >2.5 SUVmax >3.0 SUVmax >3.5 SUVmax >4.0 SUVmax >4.5 SUVmax >5.0 SUV max >5.5 SUVmax >6.0
Sensitivity, 92.3 84.6 82.1 82.1 (65.9 % (95% CI) (78.0 to 98.0) (68.8 to 93.6) (65.9 to 91.9) to 91.9) No. of lesions 36/39 33/39 32/39 32/39 PPV, % 42.9 41.8 45.1 52.5 (95% CI) (32.3 to 54.1) (30.9 to 53.4) (33.4 to 57.3) (39.4 to 65.2) No. of lesions 36/84 33/79 32/71 32/61
74.4 74.4 71.8 66.7 (57.6 to 86.4) (57.6 to 86.4) (54.9 to 84.5) (49.7 to 80.4) 29/39 29/39 28/39 26/39 56.9 65.9 77.8 83.9 (42.3 to 70.4) (50.0 to 79.1) (60.4 to 89.3) (65.5 to 93.9) 29/51 29/44 28/36 26/31
CI = Confidence interval; PET-CT = Positron emission tomography-computed tomography; PPV = Positive predictive values; SUVmax = Maximum standard uptake value.
To avoid missing malignancies or high-grade dysplasias, a colonoscopy should be performed above SUVmax = 2.5. However, the optimal cut-off value to identify a malignancy or high-grade dysplasia was SUVmax = 5.8. Future studies that further validate the optimal SUVmax values among a diverse patient population in a prospective study are warranted. Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.
References [1] Ghotbi N, Iwanaga M, Ohtsuru A, Ogawa Y, Yamashita S. Cancer screening with whole-body PET/CT for healthy asymptomatic people in Japan: re-evaluation of its test validity and radiation exposure. Asian Pac J Cancer Prev 2007;8:93–7. [2] Czernin J, Allen-Auerbach M, Schelbert HR. Improvements in cancer staging with PET/CT: literature-based evidence as of September 2006. J Nucl Med 2007;48:78S–88S. [3] Luboldt W, Volker T, Wiedemann B, Zophel K, Wehrmann U, Koch A, et al. Detection of relevant colonic neoplasms with PET/CT: promising accuracy with minimal CT dose and a standardised PET cut-off. Eur Radiol 2010; 20:2274–85. [4] Delbeke D, Martin WH. PET and PET-CT for evaluation of colorectal carcinoma. Semin Nucl Med 2004;34:209–23. [5] Rosenbaum SJ, Lind T, Antoch G, Bockisch A. Falsepositive FDG PET uptake – the role of PET/CT. Eur Radiol 2006;16:1054–65. [6] Antoch G, Saoudi N, Kuehl H, Dahmen G, Mueller SP, Beyer T, et al. Accuracy of whole-body dual-modality fluorine-18-2-fluoro-2-deoxy-D-glucose positron emission tomography and computed tomography (FDG-PET/CT) for tumor staging in solid tumors: comparison with CT and PET. J Clin Oncol 2004;22:4357–68. [7] Yeung HW, Schoder H, Smith A, Gonen M, Larson SM. Clinical value of combined positron emission tomography/ computed tomography imaging in the interpretation of 2-deoxy-2-[F-18]fluoro-D-glucose-positron emission tomography studies in cancer patients. Mol Imaging Biol 2005;7: 229–35. [8] Gutman F, Alberini JL, Wartski M, Vilain D, Le Stanc E, Sarandi F, et al. Incidental colonic focal lesions detected by FDG PET/CT. AJR Am J Roentgenol 2005;185:495–500.
[9] Israel O, Yefremov N, Bar-Shalom R, Kagana O, Frenkel A, Keidar Z, et al. PET/CT detection of unexpected gastrointestinal foci of 18F-FDG uptake: incidence, localization patterns, and clinical significance. J Nucl Med 2005;46: 758–62. [10] Kamel EM, Thumshirn M, Truninger K, Schiesser M, Fried M, Padberg B, et al. Significance of incidental 18FFDG accumulations in the gastrointestinal tract in PET/CT: correlation with endoscopic and histopathologic results. J Nucl Med 2004;45:1804–10. [11] Abouzied MM, Crawford ES, Nabi HA. 18F-FDG imaging: pitfalls and artifacts. J Nucl Med Technol 2005;33:145–55. [12] Weston BR, Iyer RB, Qiao W, Lee JH, Bresalier RS, Ross WA. Ability of integrated positron emission and computed tomography to detect significant colonic pathology: the experience of a tertiary cancer center. Cancer 2010;116: 1454–61. [13] Keogan MT, Lowe VJ, Baker ME, McDermott VG, Lyerly HK, Coleman RE. Local recurrence of rectal cancer: evaluation with F-18 fluorodeoxyglucose PET imaging. Abdom Imaging 1997;22:332–7. [14] Winawer SJ, Zauber AG. The advanced adenoma as the primary target of screening. Gastrointest Endosc Clin N Am 2002;12:1–9. [15] Eide TJ. Natural history of adenomas. World J Surg 1991;15: 3–6. [16] Winawer SJ, Fletcher RH, Miller L, Godlee F, Stolar MH, Mulrow CD, et al. Colorectal cancer screening: clinical guidelines and rationale. Gastroenterology 1997;112: 594–642. [17] Yasuda S, Fujii H, Nakahara T, Nishiumi N, Takahashi W, Ide M, et al. 18F-FDG PET detection of colonic adenomas. J Nucl Med 2001;42:989–92. [18] Schlemper RJ, Riddell RH, Kato Y, Borchard F, Cooper HS, Dawsey SM, et al. The Vienna classification of gastrointestinal epithelial neoplasia. Gut 2000;47:251–5. [19] Ishimori T, Patel PV, Wahl RL. Detection of unexpected additional primary malignancies with PET/CT. J Nucl Med 2005;46:752–7. [20] Agress H, Cooper BZ. Detection of clinically unexpected malignant and premalignant tumors with whole-body FDG PET: histopathologic comparison. Radiology 2004;230: 417–22. [21] van Kouwen MC, Nagengast FM, Jansen JB, Oyen WJ, Drenth JP. 2-(18F)-fluoro-2-deoxy-D-glucose positron emission tomography detects clinical relevant adenomas of the colon: a prospective study. J Clin Oncol 2005;23: 3713–17. [22] Nakajo M, Jinnouchi S, Tashiro Y, Shirahama H, Sato E, Koriyama C. Effect of clinicopathologic factors on visibility of
Colonoscopic evaluation for ICU on PET-CT colorectal polyps with FDG PET. AJR Am J Roentgenol 2009;192:754–60. [23] Ravizza D, Bartolomei M, Santoro L, Tamayo D, Fiori G, Trovato C, et al. Positron emission tomography for the detection of colorectal adenomas. Dig Liver Dis 2010;42:185–90. [24] Patz EF, Lowe VJ, Hoffman JM, Paine SS, Burrowes P, Coleman RE, et al. Focal pulmonary abnormalities: evaluation with F-18 fluorodeoxyglucose PET scanning. Radiology 1993;188:487–90.
1053
[25] Beggs AD, Hain SF, Curran KM, O’Doherty MJ. FDG-PET as a "metabolic biopsy" tool in non-lung lesions with indeterminate biopsy. Eur J Nucl Med Mol Imaging 2002;29: 542–6. [26] Cucchetti A, Piscaglia F, Grigioni AD, Ravaioli M, Cescon M, Zanello M, et al. Preoperative prediction of hepatocellular carcinoma tumour grade and micro-vascular invasion by means of artificial neural network: a pilot study. J Hepatol 2010;52:880–8.
Copyright of Scandinavian Journal of Gastroenterology is the property of Taylor & Francis Ltd and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.
ORIGINAL ARTICLE
Who should undergo a colonoscopy among patients with incidental colon uptake on PET-CT?
SOO-YOUNG NA1, KYUNG-JO KIM2, SEUNGBONG HAN3, SOYOUNG JIN4, JAE SEUNG KIM4, DONG-HOON YANG2, KEE WOOK JUNG2, BYONG DUK YE2, JEONG-SIK BYEON2, SEUNG-JAE MYUNG2, SUK-KYUN YANG2 & JIN-HO KIM2 1
Department of Internal Medicine, Jeju National University Hospital, Jeju National University School of Medicine, Jeju, Korea, 2Department of Gastroenterology, Asan Medical Center, University of Ulsan College of Medicine, Seoul, Korea, 3 Department of Clinical Epidemiology and Biostatistics, Asan Medical Center, University of Ulsan College of Medicine, Seoul, Korea, and 4Department of Nuclear Medicine, Asan Medical Center, University of Ulsan College of Medicine, Seoul, Korea
Abstract Objectives. To investigate the optimal cut-off of the maximum standard uptake value (SUVmax) for the detection of colorectal neoplasms and to suggest those for whom further colonoscopy is recommended among patients with incidental colonic uptake on positron emission tomography-computed tomography (PET-CT). Materials and methods. In 306 patients who underwent colonoscopy within 3 months of receiving PET-CT between January and December 2009, measurements of the per-patient and per-lesion diagnostic performance of PET-CT for the detection of colonic neoplasms were obtained. Receiver operating characteristic (ROC) analysis was used to identify the SUVmax that provided a high probability of diagnosing malignancy and high-grade dysplasia. Results. The per-patient and per-lesion PET-CT detection sensitivities for malignancies were 93.3% (28/30; 95% confidence interval (CI) 76.5% to 98.9%) and 93.5% (29/31, 95% CI 77.2% to 98.9%), respectively; the sensitivities for high-grade dysplasia were both 90.0% (9/10; 95% CI 54.1% to 99.5%). As a criterion to specifically detect both malignancy and high-grade dysplasia on focal uptake, a SUVmax greater than 2.5 yielded a 92.3% per-lesion sensitivity and a 42.9% per-lesion positive predictive value (PPV). In the ROC curve analysis, a cut-off value of SUVmax = 5.8 was established, at which the sensitivity, PPV and positive likelihood ratio for diagnosing malignancy and high-grade dysplasia were 71.8% (28/39; 95% CI 54.9% to 84.5%), 84.8% (28/33; 95% CI 67.3% to 94.3%) and 6.9, respectively. Conclusion. The optimal cut-off value to identify a malignancy or high-grade dysplasia was SUVmax = 5.8. However, to avoid missing a malignancy or high-grade dysplasia, a colonoscopy should be performed above a SUVmax = 2.5.
Key Words: colonoscopy, colorectal neoplasm, maximum standard uptake value, PET-CT
Introduction The use of positron emission tomography (PET) with 18F-fluoro-2-deoxy-D-glucose (FDG) uptake for tumor staging and monitoring of cancer therapy effects has increased notably [1,2]. The combination of PET and computed tomography (CT) has the advantage of obtaining both metabolic and structural informations; PET-CT provides better discrimination
of the pathological metabolic foci from the physiologic form and localizes the pathologic foci [3–5], and it has already considerably reduced the incidence of falsepositive findings compared to FDG-PET [6,7]. Nevertheless, many patients who have indicated incidental colonic uptake (ICU) on PET-CT may unnecessarily undergo colonoscopy based on the maximum standardized uptake value (SUVmax). Approximately, 13–45% of patients with ICU positivity on
Correspondence: Kyung-Jo Kim, MD, Department of Gastroenterology, Asan Medical Center, University of Ulsan College of Medicine, 88, Olympic-ro 43-gil, Songpa-gu, Seoul 138-736, South Korea. Tel: +82 2 3010 3196. Fax: +82 2 476 6517. E-mail: [email protected]
(Received 10 September 2014; revised 20 November 2014; accepted 23 November 2014) ISSN 0036-5521 print/ISSN 1502-7708 online 2015 Informa Healthcare DOI: 10.3109/00365521.2014.992363
1046
S. -Y. Na et al.
Table I. The indications for PET-CT and colonoscopy in 691 consecutive patients who undergoing colonoscopy after PET-CT within 3 months. No. of patients (%) Reason for PET-CT Oncologic staging or surveillance Screening Cancer of unknown origin Differential diagnosis of malignancy from benign disease Others Reason for colonoscopy Screening Abnormal PET-CT or CT images Baseline study for gynecologic malignancies Gastrointestinal symptoms or signs Cancer of unknown origin Polypectomy surveillance Others
377 216 58 20
(54.6) (31.3) (8.4) (2.9)
20 (2.9) 334 122 78 53 49 44 11
(48.3) (17.7) (11.3) (7.7) (7.1) (6.3) (1.6)
Abbreviation: PET = Positron emission tomography; CT = Computed tomography.
PET-CT showed no demonstrable lesions on colonoscopy [8–11]. Therefore, more information is essential for the optimal use of colonoscopy after obtaining a positive PET-CT scan based on the risk of the patient having an actual advanced colorectal neoplasm. Although a few previous studies have proposed a SUVmax cut-off value for further examination in cases of ICUs [3,12,13], there is still a paucity of data on this topic and a lack of consensus regarding an optimal strategy. Therefore, this study was conducted to identify an optimal cut-off value for the SUVmax of PET-CT for detecting advanced colorectal neoplasms and to suggest a means of determining which patients with ICU on PET-CT would benefit from further colonoscopy. Materials and methods Patients This study was approved by the Institutional Review Board of Asan Medical Center (AMC) (No. 2011-0897). From January to December 2009, 15,907 PET-CTs were performed in 14,370 adult patients at AMC. Of these patients, 691 subsequently underwent colonoscopy within 3 months, including patients who had abnormal PET-CT findings. Data regarding patient demographics, medical histories, indications and results of colonoscopies were extracted. The indications for PET-CT were oncologic staging or surveillance in 377 (54.6%) patients; screening in 216 (31.3%); cancer of unknown primary (CUP) work-up in 58 (8.4%); differential diagnosis of malignancy from benign disease in 20 (2.9%); and
other conditions, such as work-up for ascites, elevated tumor marker, pleural effusion, fever of unknown origin, etc. in 20 (2.9%). The reasons for undergoing colonoscopy were screening in 334 patients (48.3%); further evaluation for abnormal PET-CT or CT images in 122 (17.7%); baseline study for gynecologic malignancies in 78 (11.3%); gastrointestinal symptoms or signs in 53 (7.7%); CUP work-up in 49 (7.1%); polypectomy surveillance in 44 (6.3%); and other reasons in 11 (1.6%). The indications for PET-CT and colonoscopy were simply summarized in Table I. Of those 691 patients, 312 patients were excluded from the study for the following reasons: patients with a history of colorectal cancer including prior colon resection (n = 300) because ICUs are defined as unexpected colorectal findings that are discovered on an imaging study unrelated to the large bowel and colonic resection could not allow an accurate topographic matching between possible PET-CT uptake sites and the colonoscopy findings; patients with an incomplete colonoscopy (n = 6); and patients with incomplete data (n = 6). The technical assessment of PET-CT included all 379 patients. The mean age of the 379 patients was 56.0 ± 11.8 years, and 49.3% (187/379) of the study subjects were male. The median interval from PET-CT to colonoscopy was 5 days (range, 0–81 days). Finally, based on the inclusion and exclusion criteria, 306 patients were included in the analysis of the colonic findings (Figure 1). We only analyzed significant colonic lesions, which included malignancies (primary or metastatic colorectal cancers, lymphomas, and neuroendocrine tumors), advanced adenomas (‡10 mm in diameter with any grade of dysplasia or any size with at least a high-grade or villous architecture) [14], and inflammatory conditions in which the onset of symptoms occurred before the PET-CT. The exclusion criteria, i.e., insignificant colonic lesions, were non-advanced adenomas, non-neoplastic polyps and inflammatory conditions in which the onset of symptoms occurred after the PET-CT. The presence of non-advanced adenomas was not considered a significant finding in this study for the following reasons: only 2.5 polyps per 1000 develop into cancer per year [15]; the size and shape are well-known as adequate predictors of malignancy [16]; and small polyps can be negative on PET due to both partial-volume artifacting and non-FDG-avid metabolism [17]. Protocol, definitions and blinded interpretation of PET-CT PET-CT scanning was performed using three PET-CT camera systems: the Discovery STe (GE Healthcare,
Colonoscopic evaluation for ICU on PET-CT
1047
Consecutive patients who undergoing colonoscopy after PET-CT within 3 months (n = 691) Excluded (n = 312) Colorectal cancer including prior colon rescction (n = 300) Incomplete colonoscopy (n = 6) Incomplete data (n = 6) Cohort for techincal assessment (n = 379) Insignificant lesions on colonoscopy (n = 73)
Cohort for colonic findings analysis (n = 306)
Non-advanced adenomas (n = 51) Non-neoplastic polyps (n = 18) Inflammations occurred after PET-CT (n = 4)
Figure 1. Flow diagram of study patients.
Waukesha, WI, USA) (44.6%, 169/379); the Biograph 16 (Siemens Medical System/CTI, Erlangen, Germany) (23.5%, 89/379); and the Biograph Truepoint 40 (Siemens Medical System/CTI, Erlangen, Germany) (31.9%, 121/379). All patients fasted for more than 6 h prior to scanning. Whole-blood glucose concentration, measured immediately before FDG administration, was below 150 mg/dl for all patients. Image acquisition began approximately 60 min (59.4 ± 6.9 min) after the intravenous injection of 370– 555 MBq (10–15 mCi, 12.4 ± 1.8 mCi) of FDG. Non-contrast-enhanced low-dose spiral CT was performed from the base of the skull to the proximal thighs for attenuation correction and image fusion, followed by a three-dimensional caudocranial PET emission scan. The emission scan time per bed position was 2–3 min, and 6 or 7 bed positions were employed. The PET data were reconstructed iteratively both with and without attenuation correction based on CT data, and they were reoriented in axial, sagittal and coronal slices. The SUVmax reference values were between 0.9 and 2.5 (1.7 ± 0.3) in the liver and between 0.5 and 1.8 (1.2 ± 0.2) in the mediastinum. The bowel was completely unprepped, i.e., not cleansed or distended. To ensure the inhibition of bowel movement, 135 mg of duspatalin was orally administered to all patients. Lean body mass, instead of body weight, was used to assess the FDG activity by SUVmax, which was
measured at the most intense colonic element in patients with focal activity and at the entire element if the activity was diffuse or segmental. To ensure consistent SUVmax measurements, uptake patterns, and location determinations, one associate professor and board-certified nuclear physician (Kim JS) reviewed the PET-CT images with ICU identified on the initial reading while blinded to the colonoscopic findings. Physiologic FDG uptake of the colon was defined as contiguous diffuse uptake without colonic wall abnormality on PET-CT. Abnormal colonic uptake of FDG was defined as any wellcircumscribed foci of increased colonic uptake greater than what was measured in the normal hepatic parenchyma that was not suspected to be related to the primary cancer under evaluation. If abnormal uptake was considered to be present in the colon, the reviewer was required to assign the lesion to one of the five colonic segments: ascending, transverse, descending, sigmoid or rectum. Focal uptake was defined as abnormal nodular uptake, and segmental or diffuse uptake as abnormal colonic wall uptake that was shorter or longer than one colonic segment. Colonoscopy Colonoscopies were performed after bowel preparation with polyethylene glycol 4 l solution, all by
1048
S. -Y. Na et al.
experienced board-certified gastroenterologists, who had at least 1000-colonoscopy case experiences using a video colonoscope (CF 260; Olympus Optical Co., Tokyo, Japan). All significant colonic lesions were biopsied when possible. Each polyp was classified by an experienced pathologist as either an adenoma or another type of polyp. Dysplasia was defined as low-grade dysplasia, high-grade dysplasia or carcinoma according to the Vienna criteria [18]. Combination of the PET-CT and colonoscopic findings Colonoscopy is considered to be the gold standard colon diagnostic procedure. The PET-CT findings were classified on the basis of their colonoscopic findings as follows: a false-negative findings was defined if there was any clinically relevant colonic lesion on the colonoscopy without an abnormal colonic uptake of FDG; a true-negative finding was defined if both the PET-CT and colonoscopy were negative; a false-positive finding was defined if an abnormal colonic uptake had no relevant lesion on colonoscopy; and a true-positive PET-CT finding was defined if lesion-by-lesion matching could be performed between colonic uptake of FDG and colonoscopy in the same segment.
reported p-Values were obtained using two-sided tests and p-Values £ 0.05 were considered statistically significant. Results ICU and per-patient analysis of PET-CT Of the 306 patients who were enrolled in the cohort, both the PET-CT and colonoscopy showed negative findings in 211 patients. The remaining 95 patients had a total of 115 significant colonic lesions on PETCT and/or colonoscopy. Only 103 ICUs of 115 significant colonic lesions were observed in 88 patients; the ICU was 0.6% (88/14,370). The per-patient PET-CT sensitivity for detecting patients with significant colonic lesions was 86.8% (46/53; 95% CI 74.0% to 84.1%) (Table II). In the sub-group analysis, the sensitivities for detecting malignancy and high-grade dysplasia were 93.3% (28/30; 95% CI 76.5% to 98.9%) and 90.0% (9/10; 95% CI 54.1% to 99.5%), respectively. The overall per-patient NPV was 96.8% (211/218; 95% CI 93.2% to 98.6%). The NPVs for malignancy and high-grade dysplasia were 99.1% (216/218; 95% CI 96.4% to 99.8%) and 99.5% (217/218; 97.1% to 99.9%), respectively.
Statistical analysis The performance measures of PET-CT for diagnosing the colonic lesions included per-patient sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV), along with perlesion sensitivity and PPV. Regarding the analysis of the SUVmax on PET-CT suggestive of malignancy and high-grade dysplasia, the per-lesion sensitivities and PPVs of various SUVmax values for diagnosing malignancy and high-grade dysplasia were obtained. The 95% confidence intervals (CIs) were calculated for proportional data. In addition, a receiver operating characteristic (ROC) curve, displaying sensitivity along with the false-positive rate (1-specificity), was used to assess the best positive likelihood ratio (PLR) cut-point of SUVmax for distinguishing malignancy and high-grade dysplasia from low-grade dysplasia or inflammatory process. The associations of SUVmax with the PET-CT uptake pattern and colonic pathologic type were tested using the Mann–Whitney U test and Student’s t-test. The Mann–Whitney U test was also used to compare polyp sizes between the falsenegative group and the true-positive group. Fisher’s exact test was used to compare the sensitivity according to the colonic pathologic types. Statistical analysis was performed using a statistical software package (SPSS, Inc., version 18.0, Chicago, IL, USA). All
Per-lesion analysis of PET-CT according to colonic lesions A total of 115 lesions from either PET-CT or colonoscopy in 95 patients were divided into one of the three groups for analysis: 12 lesions in the falsenegative group, 54 lesions in the false-positive group and 49 lesions in the true-positive group. The perlesion sensitivities of PET-CT according to the type of colonic lesion are shown in Table III. The overall perlesion sensitivity of PET-CT was 80.3% (49/61; 95% CI 67.8% to 90.0%). Based on the final diagnosis, the sensitivities were 93.5% (29/31; 95% CI 77.2% to 98.9%) for malignant lesions and 90.0% (9/10; 95% CI 54.1% to 99.5%) for high-grade dysplasias. Sensitivities were highest in cases of malignancy and highgrade dysplasia, followed by low-grade dysplasia and the inflammatory process, which were significantly different (92.7% vs. 41.7% vs. 75.0%, p < 0.01). Although the size of the adenoma did not differ between the true-positive group and false-negative group (15.8 ± 6.5 mm vs. 17.5 ± 9.6 mm, p = 0.65), high-grade dysplasia showed a higher sensitivity than lower-grade dysplasia (90.0% vs. 41.7%, p = 0.03). The overall SUVmax was higher in the true-positive group than in the false-positive group (8.6 ± 5.0 vs.
Colonoscopic evaluation for ICU on PET-CT
1049
Table II. Per-patient analysis of PET-CT for diagnosing colonic lesions. Advanced adenomas Measures
Overall
Sensitivity, % (95% CI) 86.8 (74.0 No. of patients 46/53 Specificity, % (95% CI) 83.4 (78.1 No. of patients 211/253 PPV, % (95% CI) 52.3 (41.4 No. of patients 46/88 NPV, % (95% CI) 96.8 (93.2 No. of patients 211/218
to 84.1) to 87.7) to 62.9) to 98.6)
Malignancies
High-grade
Low-grade
Inflammatory processes
93.3 (76.5 28/30 78.3 (72.8 216/276 31.8 (22.5 28/88 99.1 (96.4 216/218
90.0 (54.1 to 99.5) 9/10 73.3 (67.8 to 78.2) 217/296 10.2 (5.1 to 20.0) 9/88 99.5 (97.1 to 99.9) 217/218
55.6 (22.7 to 84.7) 5/9 72.1 (66.5 to 77.0) 214/297 5.7 (2.1 to 13.4) 5/88 98.2 (95.1 to 99.4) 214/218
75.0 (35.6 to 95.5) 6/8 72.5 (67.0 to 77.4) 216/298 6.8 (2.8 to 14.8) 6/88 99.1 (96.4 to 99.8) 216/218
to 98.9) to 82.9) to 42.7) to 99.8)
Abbreviations: CI = Confidence interval; NPV = Negative predictive value; PET-CT = Positron emission tomography-computed tomography; PPV = Positive predictive value.
5.0 ± 2.1, p < 0.01). However, the difference in SUVmax was significant only for the focal uptake pattern between the two groups (p < 0.01) in the subanalysis. The final diagnosis in the true-positive group was added under Table IV.
SUVmax cut-off value for PET-CT to suggest malignancy and high-grade dysplasia We then analyzed the SUVmax values only for focal PET-CT uptakes to increase the sensitivity for the suggestion of malignancy or high-grade dysplasia on PET-CT. The per-lesion sensitivity and PPV of PETCT for diagnosing malignancy and high-grade dysplasia varied according to differing SUVmax values. As the cut-off value for SUVmax increased, the sensitivity decreased and the PPV increased. In our data, the lowest SUVmax to detect malignancy or highgrade dysplasia was 2.5. Therefore, a SUVmax higher than 2.5 on PET-CT yielded the highest sensitivity of 92.3% (36/39; 95% CI 78.0% to 98.0%) and a moderate PPV of 42.9% (36/84; 95% CI 32.3% to 54.1%) (Table V). The ROC curve for detecting malignancy and high-grade dysplasia in focal PET-CT uptakes
with the best PLR is shown in Figure 2. A cut-point of 5.8 was identified on the ROC curve, at which the area under the curve, sensitivity, PPV and PLR values were 0.76 (95% CI 0.63 to 0.88), 71.8% (28/39; 95% CI 54.9% to 84.5%), 84.8% (28/33; 95% CI 67.3% to 94.3%), and 6.9, respectively. A more illustrative scatter plot for this cut-off point is shown in Figure 3. The SUVmax for malignancies and adenomas with high-grade dysplasia was significantly higher than the values for false-positive findings, benign lesions and adenomas with low-grade dysplasia (p < 0.01). Discussion Incidental PET-CT uptake has been reported to be between 0.6% and 2.7% in the large bowel [3,12], 1.3% and 3.0% in the gastrointestinal tract [9,10] and 3.0% and 4.1% in the whole body [19,20]. In this study, the prevalence of ICU was 0.6%, which is similar to the values reported in other series; however, we included only those incidental findings evaluated by colonoscopy within 3 months, so our incidental uptake shows a somewhat low level.
Table III. Per-lesion sensitivities of PET-CT according to colonic lesions. Final diagnosis Malignancy Colorectal cancer Lymphoma Metastatic cancer Neuroendocrine tumor Advanced colonic adenoma Size* (cm) High-grade Low-grade Inflammatory process Overall
False-negative (No. of lesions)
True-positive (No. of lesions)
Total (No. of lesions)
2 0 0 0 2 8 15.8 ± 6.5 1 7 2 12
29 22 3 4 0 14 17.5 ± 9.6 9 5 6 49
31 22 3 4 2 22 10 12 8 61
Sensitivity, % (95% CI)
*Sizes (mean ± SE) between the false-negative and true-positive groups were not significantly different (p = 0.65). Abbreviations: CI = Confidence interval; PET-CT = Positron emission tomography-computed tomography.
93.5 100.0 100.0 100.0 0 63.6 90.0 41.7 75.0 80.3
(77.2 to 98.9) (81.5 to 100.0) (31.0 to 100.0) (39.6 to 100.0) (0 to 80.2) (40.8 to 82.0) (54.1 to 99.5) (16.5 to 71.4) (35.6 to 95.5) (67.8 to 90.0)
1050
S. -Y. Na et al.
Table IV. Per-lesion positive predictive values and SUVmax values of PET-CT for uptake patterns between the false-positive and true-positive groups. False-positive Uptake patterns
No. of lesions
Diffuse Segmental Focal Overall
True-positive
SUVmax (mean ± SE)
No. of lesions
± ± ± ±
3* 3† 43‡ 49
4 9 41 54
8.1 5.8 4.5 5.0
4.0 2.8 1.3 2.1
SUVmax (mean ± SE)
p-Value (SUVmax)
± ± ± ±
0.72 0.20 < 0.01 < 0.01
6.9 11.7 8.2 8.6
3.6 8.8 4.2 5.0
*Diffuse PET-CT uptake included cytomegalovirus colitis (1), pseudomembranous colitis (1) and non-specific colitis (1). Segmental PET-CT uptakes included lymphoma (2) and ulcerative colitis (1). ‡ Focal PET-CT uptakes included colorectal cancer (22), advanced colonic adenoma (14), lymphoma (1), metastatic cancer (4), large condyloma (1) and non-specific colitis (1). Abbreviations: PET-CT = Positron emission tomography-computed tomography; SUVmax = Maximum standardized uptake value. †
We found that malignancy and high-grade dysplasia were detected with very high sensitivity in both perpatient and per-lesion analyses. In addition, the NPVs for malignancy and high-grade dysplasia were more than 99% in per-patient analysis. Many previous studies have reported that PET-CT sensitivities ranged from 55% to 90% in colorectal cancer or advanced adenoma [3,8,12,17,21–23]. Our study demonstrated that all colorectal cancers, lymphomas and metastatic cancers had abnormal PET-CT uptakes, whereas two primary neuroendocrine tumors were not found, as in other studies [3,8,23]. It is well-known that neuroendocrine tumors are one of the malignancies that show FDG uptakes at the same rate as the normal surrounding tissue, leading to failures in their identification due to their low potential for malignancy [11]. Low-grade dysplasia and 1.0
Sensitivity
0.8
0.6
0.4
0.2
0.0 0.0
0.2
0.4
0.6
0.8
1.0
1-Specificity Figure 2. Receiver operating characteristic curve for determining the optimal cut-off value of the maximum standard uptake value (SUVmax). The cut-off is defined such that it minimizes the distance between the curve and the upper-left corner (dot). The cut-off point of the SUVmax for detecting malignancies and highgrade dysplasias in focal positron emission tomography-computed tomography uptake was 5.8.
inflammatory processes exhibited relatively low sensitivities but still fairly high NPVs. Adenoma sizes were not significantly different between the falsenegative group and the true-positive group. This finding is similar to that of Nakajo et al. [22], who demonstrated that the histologic grade was the strongest independent factor in the PET identification of colorectal polyps. Only the focal uptake pattern exhibited different SUVmax values between the true-positive and falsepositive groups. In addition, all malignancies, including colorectal cancer and advanced colorectal adenoma, excepting only two lymphomas, exhibited focal uptake patterns. The two lymphomas exhibited segmental uptake patterns. Therefore, we need to focus only the on focal uptake patterns to detect malignancy or high-grade dysplasia to increase the sensitivity. Previous studies [3,12,13,24,25] used a different SUVmax cut-off by choosing the absolute cut-off value for significant ICUs, which ranges from 2.5 to 5.0. We found that colorectal lesions, including malignancies (SUVmax = 2.5, 2.8, 4.1) and highgrade dysplasias (SUVmax = 2.9, 3.2, 4.1), could be detected at a lower level. The cut-off value for the SUVmax of PET-CT for diagnosing malignancies and high-grade dysplasias was 2.5 in our study. This value was the most sensitive cut-point to detect malignancies and high-grade dysplasias, with the exception of only two neuroendocrine tumors and one highgrade dysplasia that were not identified on PET-CT. However, reducing the SUVmax threshold to avoid misdiagnosing cancers with lower SUVmax values would rapidly and substantially decrease the PPV for the diagnosis of cancer or high-grade dysplasia by increasing the mischaracterization of PET-CTdetected lesions and false-positive lesions. This situation would generate a high rate of unnecessary colonoscopies. Accordingly, to obtain a high probability for diagnosing malignancies and high-grade dysplasias on focal PET-CT uptake, we analyzed
Colonoscopic evaluation for ICU on PET-CT
1051
p < 0.01
Maximum standardized uptake value (SUVmax)
p = 0.34 25
20 p = 0.69 15
10
5
8.1 ± 5.0
4.6 ± 1.3
9.75 ± 5.0
Cut-off value = 5.8
4.6 ± 1.1
0 False-positive and benign lesions
Low-grade dysplasias
High-grade dysplasias
Malignancies
Mean ± SE with 95% CI Figure 3. Illustrative scatter plot of positron emission tomography-computed tomography maximum standard uptake value (SUVmax), according to histology. The range of the 95% confidence interval (CI) for malignancy reached above the optimal SUVmax = 5.8 line. Alternatively, the 95% CI for false-positive results was located below the optimal SUVmax line.
the ROC curve. The results revealed that the SUVmax was 5.8, with a PLR of 6.9. The positive PET-CT probability for patients with malignancies or highgrade dysplasias increased as much as 6.9-fold, reflecting a good predictive level of detection [26]. When we used a cut-off value of 5.8, only 0.09% (incidence of ICU (1 - PPV)) [3] of patients who had PET-CT scans would undergo unnecessary colonoscopies for the diagnosis of malignancies or highgrade dysplasias. In addition, this SUVmax value may provide useful information for distinguishing malignancies or high-grade dysplasias from low-grade dysplasias or inflammatory processes. We acknowledge that this SUVmax value is limited to a reference value for patients with a high probability of malignancy or high-grade dysplasia for further colonoscopic evaluation, as its sensitivity may not be acceptable to many clinicians for use as a cut-off value. Therefore, in clinical settings, a multidisciplinary approach is necessary, and both the underlying disease state of patients and other examination findings should be considered before recommending a colonoscopy. This study had several limitations. First, our research was conducted retrospectively, and potential bias accrued from our interpretations of the PETCT data may exist. However, we attempted to reduce this bias by only considering patients who underwent colonoscopies after PET-CT examinations within a 3-month period. Second, although our research consisted of the largest comparable study population of
which we are aware, our final per-lesion analysis included only 115 lesions on PET-CT and/or colonoscopy. Therefore, the validity of the per-lesion analysis should be interpreted within the framework of this limited sample size. Third, we consider colonoscopy the gold standard. Although we focused only on malignancies and high-grade dysplasias, which are relatively large in size, a colonoscopy is not able to detect all polyps. Fourth, advanced adenomas include lesions with villous architecture, so we classified villous adenomas as advanced adenomas. However, villous adenomas were not included in our final per-lesion analysis due to the small sample size of advanced adenomas. Finally, the exclusion of nonadvanced adenomas could create significant problems integrating our findings into current evidence based recommendations for colonoscopy surveillance. Our findings only provide clinical implications for suggesting a SUVmax that may identify patients for whom colonoscopy might serve as a useful evaluation tool. Therefore, we recommend that asymptomatic patients over the age of 50 years with negative PET-CT cancer screenings follow the current recommendations for colonoscopy surveillance. In conclusion, it is clinically common to decide whether patients with ICU identified on PET-CT should undergo a colonoscopy. Although a cut-off value using a ROC curve has limited capability for screening for all malignancies and high-grade dysplasias, the SUVmax could be interpreted in two ways.
1052
S. -Y. Na et al.
Table V. Per-lesion sensitivities and PPVs of various SUVmax values of focal PET-CT uptake for diagnosing malignancy and high-grade dysplasia. Measures
SUVmax >2.5 SUVmax >3.0 SUVmax >3.5 SUVmax >4.0 SUVmax >4.5 SUVmax >5.0 SUV max >5.5 SUVmax >6.0
Sensitivity, 92.3 84.6 82.1 82.1 (65.9 % (95% CI) (78.0 to 98.0) (68.8 to 93.6) (65.9 to 91.9) to 91.9) No. of lesions 36/39 33/39 32/39 32/39 PPV, % 42.9 41.8 45.1 52.5 (95% CI) (32.3 to 54.1) (30.9 to 53.4) (33.4 to 57.3) (39.4 to 65.2) No. of lesions 36/84 33/79 32/71 32/61
74.4 74.4 71.8 66.7 (57.6 to 86.4) (57.6 to 86.4) (54.9 to 84.5) (49.7 to 80.4) 29/39 29/39 28/39 26/39 56.9 65.9 77.8 83.9 (42.3 to 70.4) (50.0 to 79.1) (60.4 to 89.3) (65.5 to 93.9) 29/51 29/44 28/36 26/31
CI = Confidence interval; PET-CT = Positron emission tomography-computed tomography; PPV = Positive predictive values; SUVmax = Maximum standard uptake value.
To avoid missing malignancies or high-grade dysplasias, a colonoscopy should be performed above SUVmax = 2.5. However, the optimal cut-off value to identify a malignancy or high-grade dysplasia was SUVmax = 5.8. Future studies that further validate the optimal SUVmax values among a diverse patient population in a prospective study are warranted. Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.
References [1] Ghotbi N, Iwanaga M, Ohtsuru A, Ogawa Y, Yamashita S. Cancer screening with whole-body PET/CT for healthy asymptomatic people in Japan: re-evaluation of its test validity and radiation exposure. Asian Pac J Cancer Prev 2007;8:93–7. [2] Czernin J, Allen-Auerbach M, Schelbert HR. Improvements in cancer staging with PET/CT: literature-based evidence as of September 2006. J Nucl Med 2007;48:78S–88S. [3] Luboldt W, Volker T, Wiedemann B, Zophel K, Wehrmann U, Koch A, et al. Detection of relevant colonic neoplasms with PET/CT: promising accuracy with minimal CT dose and a standardised PET cut-off. Eur Radiol 2010; 20:2274–85. [4] Delbeke D, Martin WH. PET and PET-CT for evaluation of colorectal carcinoma. Semin Nucl Med 2004;34:209–23. [5] Rosenbaum SJ, Lind T, Antoch G, Bockisch A. Falsepositive FDG PET uptake – the role of PET/CT. Eur Radiol 2006;16:1054–65. [6] Antoch G, Saoudi N, Kuehl H, Dahmen G, Mueller SP, Beyer T, et al. Accuracy of whole-body dual-modality fluorine-18-2-fluoro-2-deoxy-D-glucose positron emission tomography and computed tomography (FDG-PET/CT) for tumor staging in solid tumors: comparison with CT and PET. J Clin Oncol 2004;22:4357–68. [7] Yeung HW, Schoder H, Smith A, Gonen M, Larson SM. Clinical value of combined positron emission tomography/ computed tomography imaging in the interpretation of 2-deoxy-2-[F-18]fluoro-D-glucose-positron emission tomography studies in cancer patients. Mol Imaging Biol 2005;7: 229–35. [8] Gutman F, Alberini JL, Wartski M, Vilain D, Le Stanc E, Sarandi F, et al. Incidental colonic focal lesions detected by FDG PET/CT. AJR Am J Roentgenol 2005;185:495–500.
[9] Israel O, Yefremov N, Bar-Shalom R, Kagana O, Frenkel A, Keidar Z, et al. PET/CT detection of unexpected gastrointestinal foci of 18F-FDG uptake: incidence, localization patterns, and clinical significance. J Nucl Med 2005;46: 758–62. [10] Kamel EM, Thumshirn M, Truninger K, Schiesser M, Fried M, Padberg B, et al. Significance of incidental 18FFDG accumulations in the gastrointestinal tract in PET/CT: correlation with endoscopic and histopathologic results. J Nucl Med 2004;45:1804–10. [11] Abouzied MM, Crawford ES, Nabi HA. 18F-FDG imaging: pitfalls and artifacts. J Nucl Med Technol 2005;33:145–55. [12] Weston BR, Iyer RB, Qiao W, Lee JH, Bresalier RS, Ross WA. Ability of integrated positron emission and computed tomography to detect significant colonic pathology: the experience of a tertiary cancer center. Cancer 2010;116: 1454–61. [13] Keogan MT, Lowe VJ, Baker ME, McDermott VG, Lyerly HK, Coleman RE. Local recurrence of rectal cancer: evaluation with F-18 fluorodeoxyglucose PET imaging. Abdom Imaging 1997;22:332–7. [14] Winawer SJ, Zauber AG. The advanced adenoma as the primary target of screening. Gastrointest Endosc Clin N Am 2002;12:1–9. [15] Eide TJ. Natural history of adenomas. World J Surg 1991;15: 3–6. [16] Winawer SJ, Fletcher RH, Miller L, Godlee F, Stolar MH, Mulrow CD, et al. Colorectal cancer screening: clinical guidelines and rationale. Gastroenterology 1997;112: 594–642. [17] Yasuda S, Fujii H, Nakahara T, Nishiumi N, Takahashi W, Ide M, et al. 18F-FDG PET detection of colonic adenomas. J Nucl Med 2001;42:989–92. [18] Schlemper RJ, Riddell RH, Kato Y, Borchard F, Cooper HS, Dawsey SM, et al. The Vienna classification of gastrointestinal epithelial neoplasia. Gut 2000;47:251–5. [19] Ishimori T, Patel PV, Wahl RL. Detection of unexpected additional primary malignancies with PET/CT. J Nucl Med 2005;46:752–7. [20] Agress H, Cooper BZ. Detection of clinically unexpected malignant and premalignant tumors with whole-body FDG PET: histopathologic comparison. Radiology 2004;230: 417–22. [21] van Kouwen MC, Nagengast FM, Jansen JB, Oyen WJ, Drenth JP. 2-(18F)-fluoro-2-deoxy-D-glucose positron emission tomography detects clinical relevant adenomas of the colon: a prospective study. J Clin Oncol 2005;23: 3713–17. [22] Nakajo M, Jinnouchi S, Tashiro Y, Shirahama H, Sato E, Koriyama C. Effect of clinicopathologic factors on visibility of
Colonoscopic evaluation for ICU on PET-CT colorectal polyps with FDG PET. AJR Am J Roentgenol 2009;192:754–60. [23] Ravizza D, Bartolomei M, Santoro L, Tamayo D, Fiori G, Trovato C, et al. Positron emission tomography for the detection of colorectal adenomas. Dig Liver Dis 2010;42:185–90. [24] Patz EF, Lowe VJ, Hoffman JM, Paine SS, Burrowes P, Coleman RE, et al. Focal pulmonary abnormalities: evaluation with F-18 fluorodeoxyglucose PET scanning. Radiology 1993;188:487–90.
1053
[25] Beggs AD, Hain SF, Curran KM, O’Doherty MJ. FDG-PET as a "metabolic biopsy" tool in non-lung lesions with indeterminate biopsy. Eur J Nucl Med Mol Imaging 2002;29: 542–6. [26] Cucchetti A, Piscaglia F, Grigioni AD, Ravaioli M, Cescon M, Zanello M, et al. Preoperative prediction of hepatocellular carcinoma tumour grade and micro-vascular invasion by means of artificial neural network: a pilot study. J Hepatol 2010;52:880–8.
Copyright of Scandinavian Journal of Gastroenterology is the property of Taylor & Francis Ltd and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.
