Specific author contributions are as follows. Conception: William A. Ross and Robert S. Bresalier; design: William A. Ross and Brian R. Weston; acquisition of data and data analysis: Brian R. Weston, William A. Ross, Revathy B. Iyer, and Wei Qiao; interpretation of results: Brian R. Weston, William A. Ross, and Jeffrey H. Lee; drafting of manuscript: William A. Ross and Brian R. Weston; critical revision of manuscript: Brian R. Weston, Revathy B. Iyer, Jeffrey H. Lee, Robert S Bresalier, and William A. Ross.
An oral presentation of the preliminary results of this study was presented during the 2008 Digestive Disease Week in San Diego, California, with the abstract appearing in Gastrointestinal Endoscopy, 2008; 67:AB82-AB83.
The ability of integrated positron emission tomography and computed axial tomography (PET-CT) to detect colonic pathology is not fully defined. The purpose of this study was to assess the ability of PET-CT to detect colonic pathology and to determine the significance of (18F)2-fluoro-2-deoxyglucose (18F-FDG) activity noted incidentally in the colon on PET-CT.
Records for all patients who underwent PET-CT and colonoscopy at our institution were reviewed. Patients with history of colonic malignancy or colon surgery were excluded.
Fifty-eight patients had incidental colonic 18F-FDG activity on PET (Group A) and 272 had none (Group B). In Group A, 65% of patients had pathologic findings detected on colonoscopy that corresponded to the site of PET activity. Standardized uptake value (SUV) readings were not helpful in distinguishing true-positives from false-positives. In Group B, 11.8% of patients were found to have significant colonic findings. Lesions not detected by PET-CT included 4 colon cancers, 7 advanced adenomas, and 10 patients with colonic lymphoma. Overall, the sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and accuracy of PET-CT for detecting significant pathology were 53%, 93%, 65%, 89%, and 85%, respectively. For detecting colon cancer and adenomas 10 mm or more, the sensitivity, specificity, PPV, NPV, and accuracy of PET-CT were 72%, 90%, 45%, 96%, and 88%, respectively.
Positron emission tomography (PET), which is used primarily for cancer staging and surveillance, relies on the uptake of the radiolabeled glucose analogue (18F)2-fluoro-2-deoxyglucose (18F-FDG) by metabolically active cells. However, cells can be metabolically active for a variety of reasons unrelated to malignancy, and these benign processes can lead to false-positive findings.1, 2 One way to reduce false-positive findings has been by combining PET with computed tomography (CT).3, 4 Combined PET-CT is better at localizing metabolically active foci and helps to distinguish physiological activity from pathological findings more readily than PET or CT alone.5 In addition to an increasing number of oncological indications, PET-CT has been proposed as a modality for cancer screening6 and as a tool for monitoring inflammatory conditions of the colon, such as Crohn's disease.7 Integrating PET with CT colonography has been suggested as a way to screen patients for colorectal neoplasia.8, 9
Occasionally, PET-CT will reveal intense metabolic activity at sites not considered to be involved with the malignant process under study. One common site for such incidental activity is the colon.10-13 Colonoscopic evaluation of incidental colonic PET activity frequently reveals abnormalities, but no lesion is found in a third of patients in most reports.11, 14 Furthermore, PET-CT does not detect all clinically relevant lesions.15, 16 Thus, although it has been proposed as a new way of diagnosing and monitoring colonic diseases, PET-CT has a significant false-positive rate and a poorly defined false-negative rate for colon pathology.
The frequent lack of a uniform gold standard test (providing histological diagnoses) to confirm PET-CT results in previous studies raises the potential for verification bias. This may lead to an overestimation of the utility of PET-CT.17 By evaluating all patients who underwent PET-CT and then colonoscopy, the false-negative rate can be better defined. To assess the significance of incidentally detected 18F-FDG activity in the colon and the number of colonic findings missed by PET-CT, we evaluated all patients without known colonic pathology who underwent PET-CT followed by colonoscopy at a tertiary cancer center.
MATERIALS AND METHODS
The study was approved by The University of Texas M. D. Anderson Cancer Center Institutional Review Board. The medical records of all adult patients who underwent PET-CT followed by colonoscopy within 90 days, between January 2005 and April 2007, were retrospectively reviewed for demographic data, medical history (including primary cancer), and indication for PET-CT. PET-CT reports were reviewed to identify those that were read as having incidental FDG activity in the colon. Endoscopy and pathology records were reviewed to obtain findings and indication for colonoscopy. Patients with a history of malignancy or disease involving the colon including prior colon resection were excluded from analysis. Patients with incomplete colonoscopy to the cecum or with inadequate bowel prep were also excluded.
18F-FDG avidity was measured by maximum standardized uptake value (SUV). A physiological finding was defined by an SUV ≤3.5 based on institutional experience and the literature.18 Incidental colon activity detected by PET-CT was defined by an SUV >3.5, which was not suspected to be related to the primary cancer under evaluation.12 To ensure consistent SUV determinations, one of the authors (RBI) with 6 years of experience with PET-CT reviewed the studies with incidental colonic activity identified on initial reading. These reviews were done while blinded to the colonoscopy findings. SUV max was measured at the most intense colonic element in those with focal activity and the entire element if activity was diffuse. Patients with incidental activity in the colon revealed by PET-CT were designated Group A, and patients without incidental colonic activity on PET-CT were designated Group B.
We defined a true-positive PET-CT finding as a focus of FDG avidity with an SUV >3.5 and a corresponding lesion on colonoscopy situated in the same segment. A false-positive finding was defined as a focus with an SUV >3.5 with no corresponding lesion on colonoscopy. A PET-CT was considered a false-negative if there was any clinically relevant colonoscopic finding that had no reported corresponding abnormal 18F-FDG uptake in the colon. Noteworthy colonic findings included cancers and adenomas ≥10 mm, which were referred to collectively as advanced colonic neoplasms. Other findings felt to be clinically relevant were histologically proven new lymphomatous involvement of the colon and inflammatory conditions whose onset of symptoms occurred before PET-CT. Lesions <10 mm were considered to be undetectable with PET-CT and were omitted from test performance calculation.19 Clinical course of patients with false-positive PET-CT studies were subsequently reviewed to determine if there was a diagnosis of colonic disease after initial negative colonoscopy.
PET-CT Imaging Protocol
All scans were obtained using an integrated PET-CT inline scanner (Discovery ST-8, GE Healthcare, Waukesha, Wis). The full width half maximal of this scanner is 6 mm. Patients fasted for ≥6 hours before undergoing PET-CT. Blood glucose was measured 1 hour before 18F-FDG injection; if the blood glucose level was ≥150 mg/dl, then PET-CT was deferred. Approximately 1 hour before PET-CT, each patient received an injection of 12-20 mCi (444-740 MBq) of 18F-FDG.
Noncontrast-enhanced CT was performed from the base of the skull to the upper thighs for attenuation correction and diagnosis (150 mAs, 120 kVp, table speed 13.5 mm/rotation, beam collimation 8 × 1.25 mm). Axial CT images were reconstructed using a soft reconstruction kernel with a 3.75-mm slice thickness and a 3.27-mm slice thickness to match PET images.
PET was conducted in the 2-dimensional or 3-dimensional mode for 3 minutes per bed position, and images were reconstructed with standard vendor-provided reconstruction algorithms that incorporated ordered subset expectation maximization. Attenuation correction of PET images was performed using attenuation data from CT. We used the manufacturer's software to correct emission data for scatter, random events, and dead-time losses. Full width half maximal on the scanner was 6 mm. PET-CT parameters were uniform for all studies and disease processes under evaluation, except in cases of melanoma when all extremities and the brain were included.
Fisher exact test was used to assess the association between SUV activity and colon cancer, as well as SUV activity and lymphoma. The test was also used to compare pathology between those with positive PET-CT and those without. Wilcoxon rank sum test was used to assess the difference in tumor size between patients with and without18 F-FDG avid lesions. The test was also used to test associations between SUV and various endoscopic and pathologic findings. In addition, Spearman correlation method was used to determine the correlation between18 F-FDG activity and tumor size.
All computations were carried out using SAS (SAS Institute, Cary NC). All reported P-values were 2-sided test and P-values ≤.05 were considered statistically significant.
Of 10,031 patients who underwent PET-CT at M. D. Anderson between January 2005 and April 2007, 424 had subsequent colonoscopies within 90 days. Eighty-eight patients who had known colon pathology or prior colon resection were excluded. Six patients with incomplete colonoscopy were also excluded. Of the remaining 330 patients, 58 had 60 foci of incidental colonic activity on PET and were designated Group A. The other 272 patients had no incidental colonic 18F-FDG-avid foci and were designated Group B.
The mean age was 61 years in Group A and 58.7 years in Group B. Sixty-two percent of Group A patients and 53.7% of Group B patients were male (P = .31). In Group A, the median time between PET-CT and colonoscopy was 19 days (mean 23.4 days, standard deviation [SD], 16.7 days); in Group B, the median time was also 19 days (mean, 29.5 days; SD, 27.8 days). Six (10.3%) patients in Group A had undergone colonoscopies with no remarkable findings at a mean of 11 months before PET-CT. In Group B, 88 patients (32%) had undergone colonoscopies with no remarkable findings at an average of 27.0 months before PET-CT. The primary cancer diagnoses in both groups are detailed in Table 1. Lymphoma was the most common diagnosis in both groups. The difference in the proportion of lymphoma patients between the groups was not statistically significant (P = .24).
Table 1. Number of Primary Cancer Diagnoses in Patients With (Group A) and Without (Group B) Incidental Colonic Activity on Positron Emission Tomography
Group A patients were referred for colonoscopy primarily for further assessment of abnormal PET-CT findings. Indications for colonoscopy in Group B are listed in Table 2; the most common indications were for lymphoma staging or surveillance (31.6%), and evaluation of anemia or rectal bleeding (19.8%).
Table 2. Indications for and Significant Findings of Colonoscopy in Group B Patients
No. of Patients N=272
No. of Significant Findings N=32
ACA indicates adenocarcinoma of the colon; GVHD, graft-versus-host disease; CMV, cytomegalovirus; CT, computed tomography.
Table 3 outlines significant colonoscopic findings. Overall, 65% of Group A patients and 11.8% of Group B patients had significant colonoscopic or histological findings (P < .0001). The only reason for referral in Group B associated with a significantly higher rate of noteworthy colonoscopic findings was rectal bleeding (P = .03).
Table 3. Significant Colonoscopic Findings by Group
No. of Patients (%)
Group A n=58
Group B n=272
NS indicates not significant with P > .1.
Colon polyp ≥0 mm
Nonmucinous colon cancers were identified in 5 Group A patients (8.6%) and 4 Group B patients (1.5%, P = .008). The 4 colon cancers that were not detected by PET-CT were found in patients who underwent colonoscopy because of an unknown primary cancer, rectal bleeding, abnormal CT findings, and lymphoma staging, respectively. The mean size of colon cancers detected by PET-CT was 3.5 cm; the mean size of cancers not detected by PET-CT was 4.4 cm. There was no significant difference in the size of cancers between Groups A and B.
Twenty-five adenomas ≥10 mm were found in 23 patients (36%) in Group A, and 7 advanced adenomas were found in 6 patients (2.2%) in Group B (P < .0001). The mean size of advanced adenoma was 2.76 cm (SD, 1.04 cm) in Group A and 1.34 cm (SD, 0.54 cm) in Group B (P = .002). Among the colonic neoplasms detected by PET-CT, there was no correlation between size and SUV regardless of histology (Fig. 1 and Fig. 2). No trends in configuration or location of missed advanced colonic neoplasms were found compared with those detected by PET-CT. Fifty-six patients (20.6%) in Group B had adenomas <10 mm compared with 19 (32.8%) in Group A (P = .06).
Colonic lymphoma was found on colonoscopy in 4 (11.3%) of 30 lymphoma patients in Group A and in 10 (6.1%) of 165 lymphoma patients in Group B (P = .24). Additional significant colonoscopy findings not detected by PET-CT during initial lymphoma staging included 1 colon cancer and an advanced adenoma. Other clinically relevant colonoscopic findings in Group A patients were metastatic uterine cancer in the sigmoid colon (1 patient), diverticulitis (2 patients), and radiation colitis (3 patients).
Three of the 18 Group B patients with abnormal CT findings but no activity on PET had significant colonoscopic findings: 1 had a cytomegalovirus ulcer, 1 had colon cancer, and 1 had radiation colitis. Remaining significant colonoscopic findings in Group B included metastatic disease (2 patients), radiation colitis or stricture (3 patients), a cytomegalovirus ulcer (1 patient), and acute colonic graft-versus-host disease (4 patients who were symptomatic at the time of PET-CT).
SUV and Colonoscopic Findings
Colonoscopic findings and corresponding mean maximum SUV values are presented in Table 4. Colon cancer patients had the highest SUVs, and the SUVs were significantly higher than in patients with no colonoscopic abnormalities (P = .03) and adenomas (P = .05). The difference in mean SUV between colon cancer and lymphoma or inflammatory processes did not reach statistical significance (P = .11). Malignancy did not necessarily result in high SUVs; in fact, the lymphoma group had the lowest mean SUVs. In addition, mean maximum SUVs were not useful in distinguishing false-positive examinations from true-positive findings (P = .18) (Table 5). The 20 patients in Group A who had no colonoscopic findings at sites of 18F-FDG avidity were followed for a median of 24 months. Patients with false-positive findings were more likely to have diffuse patterns of 18F-FDG activity (25%) than patients with true-positive findings (8%) (P = .11).
Table 4. Colonoscopic Findings and Corresponding Mean Maximum SUVs
No. of Patients
Mean SUV Maximum (SD)
SUVs indicates standardized uptake values; SD, standard deviation.
Table 5. Mean SUVs by Histological or Colonoscopic Findings
No. of Patients or Findings
Mean SUV (SD)
SUVs indicates standardized uptake values; SD, standard deviation; NS, P>.1; ACA, adenocarcinoma of the colon; TVA, tubulovillous adenoma; TA, tubular adenoma; HGD, high-grade dysplasia.
ACA, TVA, HGD
When Groups A and B were combined, PET-CT had a sensitivity of 72% for detecting advanced colonic neoplasms, with a specificity of 90%, PPV of 45%, NPV of 96%, and overall accuracy of 88%. When a significant pathology of any kind was considered, PET-CT had a sensitivity of 53%, specificity of 93%, PPV of 65%, and NPV of 89%, and overall accuracy of 85%.
Imaging using the ability of 18F-FDG to accumulate in metabolically active cells has found wide application in oncology. The recent integration of CT with PET has enhanced the ability to localize foci of 18F-FDG avidity. However, the experience with PET-CT is limited and consensus recommendations as to its proper clinical application are still in development.20 The use of “subsequent clinical course” without colonoscopic and/or pathologic confirmation as the standard of reference in many studies evaluating PET-CT raises the question of bias.21 To address this potential source of bias, we used colonoscopy as the reference standard and evaluated all patients without pre-existing colon conditions who had both imaging modalities and found a significant false-positive rate and false-negative rate with PET-CT. Despite negative PET-CT exams, colonoscopy found clinically significant lesions in 11.8% of patients including 5% with either colon cancer or lymphoma. By including patients with negative PET-CT findings who underwent subsequent colonoscopy, we address the verification bias present in other reports and, thus, provide a more complete picture of PET-CT's performance.17
Incidental colonic focal 18F-FDG uptake on PET-CT has been reported in 1.3%-2.7% of patients.11, 13, 14 Although subsequent evaluations revealed significant pathology in most of these patients with incidental colonic activity, false-positive rates of 16%-33% have been reported.11, 13, 14 The false-positive rate we found for incidental findings on PET-CT was at the high end of this range but likely reflects our selection process: all patients in our study had colonoscopies after PET-CT, whereas 30%-56% of patients with incidental colonic activity on PET-CT in other studies did not have subsequent colonoscopies.11, 13, 14 In our study, because we included only those incidental findings evaluated by colonoscopy within 90 days, our rate of incidental 18F-FDG colonic activity (0.6%) is lower than reported in earlier series. The 65% PPV of PET-CT to detect clinically significant pathology in our study is comparable to the 64%-67% PPVs reported in previous studies.11, 14 All but 8 patients in our study had focal 18F-FDG avidity, a pattern that has been shown to have a higher correlation with the presence of neoplasms than does diffuse 18F-FDG uptake.22 We also found that patients with focal 18F-FDG avidity had a lower false-positive rate than patients with diffuse 18F-FDG uptake.
Colon adenoma size and dysplasia grade have been reported to correlate with likelihood of detection by PET-CT.23 However, a correlation between maximum SUV and colon neoplasm histology has not been demonstrated consistently. Although Gutman et al found that mean SUVs increased with higher grades of dysplasia with highest levels in patients with colon cancers, Israel et al did not observe such a progression in SUV values.11, 14 Such an association would be consistent with reported higher mitotic activity in colonic mucosa as the tissue progresses from normal to low-grade dysplasia and then high-grade dysplasia.24 We found there to be higher mean SUV in colon cancers compared with adenomas. Yet the difference in SUV readings between colon cancer plus polyps with high grade dysplasia versus all other adenomas did not reach significance. Although PET is more likely to detect larger colonic neoplasms,23, 25, 26 the correlation between tumor size and maximum SUV has been poor.27 We found that the mean size of colon cancers not detected by PET-CT was slightly larger than the mean size of cancers detected by PET-CT. Although adenomas detected were significantly larger than those missed by PET-CT, the correlation between the size of detected adenomas and SUVs was poor. Thus, PET-CT's ability to detect colon cancers and adenomas >10 mm in size, let alone predict histology, is imperfect.
Of the 4 colon cancers in Group B that were not detected by PET-CT in our study, none were mucinous (a feature associated with low 18F-FDG avidity).28 Why these cancers were not detected remains unclear; however, studies of PET for colon cancer staging report up to a 5%-6% incidence of non-18F-FDG-avid tumors.29 In the current study, the incidence of adenoma >9 mm was lower in Group B than in a population of comparable age undergoing screening colonoscopy, probably because a third of patients in Group B had a colonoscopy an average of 27 months before undergoing PET-CT.30 Still, the ability of PET-CT to detect colon cancer and adenomas >10 mm in both groups was low, with a sensitivity of 73%. This sensitivity calculation does not take into consideration the 20% of patients in Group B with adenomas ≤10 mm that were not detected by PET-CT. Combining PET with CT colonography may overcome some of the deficiencies of PET-CT in detecting colonic neoplasia. However, this approach is still in the initial phases of development,11, 31, 32 and preliminary studies have demonstrated that integrated PET and CT colonography has only a limited ability to detect adenomas <10 mm.8, 9, 33
The most frequent use of PET-CT in our study group was for staging or surveillance of patients with lymphoma. Our data showing 8 of 87 lymphoma patients with colonic involvement missed by PET-CT suggests a need to verify imaging results. In addition, recent reports of false positive PET findings in patients with lymphoma reinforce the need to histologically confirm PET results both positive and negative.34, 35 Limitations of this study include those inherent in any retrospective analysis of a single institution's experience. In addition, we cannot totally assess the utility of PET-CT as a screening tool for colon pathology as our patient population was a select one in that all patients had a known or suspected cancer that prompted the PET-CT. In addition, colonoscopy was scheduled to evaluate symptoms in most Group B patients, not for screening. Using a single dimension to describe colonic neoplasms does not take into account all physical attributes of a lesion that may influence its detectability by PET-CT, a modality dependent upon the concentration of metabolically active cells. We found no trends in configuration or location of missed advanced colonic neoplasms compared with those detected by PET-CT. Yet by including a large number of patients without a history of colonic disease and negative PET-CT for colonic abnormalities, we have provided a more complete perspective on the performance of PET-CT in detecting colonic pathology. However, the picture remains to be fully defined as the ideal study of having all patients without known colonic disease undergoing PET-CT then followed by colonoscopy has yet to be performed.
In conclusion, despite a high false-positive rate, incidental colonic uptake on PET-CT warrants further evaluation by colonoscopy with biopsies of the site. However, negative PET-CT findings do not necessarily rule out significant neoplastic or inflammatory conditions in the colon. Symptoms suggestive of colon disease should prompt colonoscopic verification of a negative PET-CT of the colon.