Direct Comparison of ¹⁸F-FDG PET and PET/CT in Patients with Colorectal Carcinoma

PET/CT Acquisition

All consecutive clinical patients who were referred from June 2001 to mid-November 2001 for whole-body ¹⁸F-FDG PET/CT and had a known or suspected history of colorectal cancer were retrospectively included for analysis. Patients with suspected or documented active coexistent noncolorectal cancer were excluded. PET/CT was performed using a combined PET/CT scanner (Discovery LS; General Electric Medical Systems). Emission data were acquired for 5–7 bed positions, typically from the base of the skull to the mid thigh. Emission data were acquired for 5 min at each bed position. Each bed had 35 scanning planes with a 14.6-cm longitudinal field of view and a 1-slice overlap between scanning bed positions. ⁶⁸Ge transmission scans were used to generate a transmission map. The transmission time at each bed position was 3 min. A segmented attenuation map was then generated. PET images were reconstructed using ⁶⁸Ge for attenuation correction with the ordered-subsets expectation maximization algorithm (2 iterations, 28 subsets) and an 8-mm gaussian filter using a 128 × 128 matrix. Non–attenuation-corrected images were not reviewed for this analysis, nor were the CT-corrected emission images. The CT portion of the Discovery LS consisted of a multidetector helical CT scanner (LightSpeed Plus; General Electric Medical Systems). Imaging parameters were as follows for a 5-bed-position acquisition: 140 kVp, 80 mA, 0.8 s per CT rotation, a pitch of 6, a 22.5 mm/s table speed, 722.5-mm coverage, a 31.9-s acquisition time, and arms at the side of the torso. The CT slice thickness was 5 mm. The CT acquisition was done before the PET emission acquisition. CT data were resized from a 512 × 512 matrix to a 128 × 128 matrix to match the PET data in order to fuse the images. CT was also used for attenuation correction, but these images were not analyzed in this study. Patients fasted for at least 4 h before the PET acquisition. Blood glucose levels were checked before ¹⁸F-FDG injection, and if they were less than 11.1 mmol/L, the patients were injected intravenously with 555–740 MBq of ¹⁸F-FDG. A tracer uptake phase of about 60 min was implemented in which the patients were instructed to sit in a quiet room without talking. No oral or intravenous contrast agent was administered. No intravenous diuretics or intravenous hydration was given, and no bladder catheterization was performed.

Image Interpretation

Studies were interpreted using a computer workstation (eNTEGRA; General Electric Medical Systems). Studies were interpreted independently by a board-certified, moderately experienced nuclear medicine physician unaware of patient name and identification; study indication, except that the patient was referred for colorectal cancer evaluation; correlative imaging; and history. All PET studies were interpreted sequentially in randomized order. Subsequently, PET/CT studies were interpreted in a separate reading session in another randomized order, without availability of the previous PET-alone reading. We assessed ⁶⁸Ge attenuation–corrected PET images instead of CT-corrected PET images in order to evaluate the added value of CT in the interpretation of PET/CT images over the generally available ⁶⁸Ge-corrected PET images, not the impact of CT attenuation correction. Artifacts, such as “increased activity” in the presence of metallic objects, have been reported with CT attenuation correction (14) and represent confounding issues that could have rendered the evaluation of the impact of PET/CT more difficult. CT in the PET/CT reading session was used for interpretation and localization of the PET data, not for attenuation correction. Each study was reviewed for lesion identification. A lesion-by-lesion analysis was performed. Any discrepancies between lesion number found by the 2 techniques were then analyzed. Each lesion was evaluated regarding certainty of characterization and certainty of location, as explained below. Lesion location was assessed for each lesion identified and scored on a 3-point scale (0 = uncertain location, 1 = probable location, 2 = definite location). Each lesion identified was characterized using a 5-point score: 0 = definitively benign, 1 = probably benign, 2 = equivocal, 3 = probably malignant, 4 = definitively malignant.

Consensus Evaluation for Lesion Etiology

The impact of PET/CT over PET alone regarding lesion localization and characterization was analyzed on lesions identified by both techniques. All available patient records were reviewed. Lesions were scored using a consensus of 3 nuclear medicine physicians, based on all available data (correlative imaging findings, histologic proof, and clinical and radiologic follow-up). This consensus was scored on a 5-point scale: 0 = definitely benign with histologic proof; 1 = probably benign with CT and PET concordant findings or benign evolution on follow-up; 2 = equivocal with no histologic proof or follow-up; 3 = probably malignant with CT and PET concordant, with malignant evolution on follow-up, or with progression on therapy; and 4 = definitely malignant with histologic proof. Histologic proof was considered definitive if biopsy or surgery was performed less than 1 mo before and less than 2 mo after PET and there had been no intervening therapy. Concordant findings on PET and CT (or MRI) meant that both PET and CT were most consistent with malignancy (score 3 consensus).

Lesion-by-Lesion Analysis

For the accuracy analysis, equivocal (score 2) lesions by consensus were excluded. Score 3 and 4 consensus lesions were considered positive for malignancy, whereas score 0 and 1 lesions were considered negative for malignancy. For the PET and PET/CT readings, lesion characterization was dichotomized, with scores 2, 3, and 4 grouped as a positive test and scores 0 and 1 grouped as a negative test. When patients had multiple liver metastases, a maximum of 5 lesions were included for analysis.

Patient-by-Patient Analysis

The accuracy of PET and PET/CT was evaluated regarding the following items: presence of active cancer in the patient, detection of the primary cancer, detection of intraabdominal disease (excluding primary cancer and liver metastasis), detection of liver metastasis, detection of metastasis (M status), and overall correct staging. If each liver, intraabdominal extrahepatic region, and extraabdominal region were correctly assessed by PET and PET/CT, the overall staging was considered to be correct. A true-positive or true-negative result had to correctly be characterized for primary disease, liver status, and intraabdominal extrahepatic and extraabdominal disease. Under- and overstaging according to these anatomic regions were considered false-negative and false-positive, respectively. To dichotomize the data, for the PET and PET/CT readings, lesion characterization scores 2, 3, and 4 were considered a positive test, whereas scores 0 and 1 were considered a negative test. Extraabdominal lesions suggestive of malignancy but classified as unrelated to colorectal cancer were analyzed separately and not included in the overall staging. Sensitivity, specificity, and accuracy were expressed along with an estimate of the 95% confidence interval. Differences in accuracy between PET and PET/CT for the by-lesion and by-patient analyses were tested with the McNemar test. P < 0.05 was considered significant.

Patient Population

Fifty-four ¹⁸F-FDG PET/CT examinations were performed on 52 consecutive clinical patients with colorectal carcinoma. Seven scans were excluded from analysis because ⁶⁸Ge-corrected images were not available. One patient with active metastatic thyroid and colorectal cancer was also excluded. One patient with active lung and colorectal cancer was also excluded from analysis. Thus, a total of 45 patients (45 scans) were included for analysis. Mean age was 60.8 ± 11.1 y (range, 36–83 y). There were 17 women and 28 men. One patient was referred for primary staging. Forty-four patients were referred for restaging (9 to rule out recurrence, 16 to stage a known recurrence, and 19 to assess therapy response). The histologic type of the primary or recurrent cancer was available from the history or medical records for 33 patients. There were 4 cases of mucinous carcinoma, 1 of well-differentiated carcinoma, 22 of moderately differentiated carcinoma, 4 of moderately to poorly differentiated carcinoma, and 2 of poorly differentiated carcinoma. The location of the primary cancer was not specified in the records of 1 patient. Primary cancer originated from the right colon (n = 12), transverse colon (n = 2), descending colon (n = 4), or rectosigmoid colon (n = 26). Two patients received intravenous insulin 60 min before ¹⁸F-FDG injection to lower glucose levels before ¹⁸F-FDG injection. The average serum glucose level for all patients at the time of injection was 6 ± 1.7 mmol/L (range, 4.3–9.8 mmol/L).

Location and Characterization Certainty

PET/CT identified 138 lesions; PET, 141; and both modalities, 134. Localization results are shown in Table 1. Figure 1 shows an example of improved location certainty with PET/CT. Lesion characterization results are shown in Table 2. Figure 2 shows an example of improved diagnostic certainty with PET/CT. Seven lesions were identified on PET only. Six of these lesions were additional liver metastases found in 5 patients with multiple liver metastases. The other lesion was an equivocal bone lesion. Four lesions were found on PET/CT only. One was a liver metastasis missed on PET alone, 1 was a liver metastasis not described on PET in a patient with multiple liver metastases missed on PET alone, and 1 was a liver metastasis not described on PET in a patient with multiple liver metastases. One tonsillar lesion with an equivocal characterization and 1 ureteral focus (definitely benign) were not described on PET.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 1.

¹⁸F-FDG PET images in anterior projection display (A) of patient with foci of increased tracer uptake near inferior border of liver (1), left paraaortic lymph nodes (2), and left upper lung nodule (3). Lung nodule and paraaortic lymph nodes were correctly localized and characterized with PET alone. However, precise location of lesion (1) was difficult to determine, and level 0 was assigned (uncertain; liver or porta hepatis). Transverse PET images (B) at level of lesion (1) clearly showed that activity (arrow) had corresponding lymph node on CT image (C) and fused image (D). Localization was thus improved from uncertain to definite (score 2) with PET/CT.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 2.

¹⁸F-FDG PET images in anterior projection display (A) of patient with multiple liver metastases (1). For small focus of moderate metabolic activity (2) in right mid abdomen, equivocal diagnostic impression (score 2) was given because focus could represent benign activity (ureter) or malignancy (metastatic lymph node). Transverse PET images (B) at level of lesion 2 clearly showed that activity (arrow) on PET image corresponded to ureter on CT image (C) and fused image (D). Lesion characterization was thus improved to definitely benign (score 0) with PET/CT.

Accuracy Analysis

Seven patients were excluded from the accuracy analysis because no pathologic confirmation, no congruent correlative imaging, and no follow-up were available. Thirty-eight patients had follow-up, histologic confirmation of at least some of the lesions, or congruent correlative imaging findings. Two patients were excluded from the patient analysis because, after consensus evaluation, some lesions were scored as equivocal and the overall staging could not be determined. Thus, 36 patients were analyzed for the patient analysis. Twenty-five patients had a follow-up of at least 6 mo (mean, 8.6 ± 1.6 mo; range, 6–12 mo). Eleven patients had a follow-up of less than 6 mo. Four had previously documented liver metastases, and 1 had a brain metastasis. These 5 patients had congruent findings on correlative imaging.

Lesion-by-Lesion Analysis

The number of lesions defined by consensus was 140. By consensus, 18 lesions were classified as benign (6 with score 0 and 12 with score 1), 18 lesions were classified as equivocal, and 104 were classified as malignant (37 with score 4 and 67 with score 3). Thus, a total of 122 lesions were analyzed. Results are displayed in Table 3. With PET, there were 91 true-positives, 10 true-negatives, 13 false-negatives, and 8 false-positives. Of the 13 false-negatives with PET, PET/CT correctly identified 4 lesions as positive. Of the 8 false-positives with PET, PET/CT correctly identified 3 lesions as true-negative. With PET/CT, there were 89 true-positives, 12 true-negatives, 15 false-negatives, and 6 false-positives. Of the 15 false-negatives with PET/CT, PET correctly identified 6 lesions as positive. Of the 6 false-positives with PET/CT, PET correctly identified 1 lesion as negative.

PET and PET/CT both had an overall accuracy per lesion of 83% (P = 0.8) when equivocal lesions were characterized as positive. PET and PET/CT had an overall accuracy of 82% and 85%, respectively, when equivocal lesions were characterized as negative (P = 0.2).

Patient-by-Patient Analysis

Both PET and PET/CT correctly assessed the status of the primary lesion or local recurrence in all patients. The liver status was incorrectly assessed in 1 patient with PET/CT and in 2 with PET. In 1 patient, a focus of increased activity in the liver was incorrectly interpreted with PET and PET/CT as metastatic disease. Histologic examination of this lesion resected at surgery showed postsurgical inflammation. One patient with liver metastases from mucinous carcinoma was incorrectly classified as without active cancer with PET (Fig. 3), whereas with PET/CT, 1 liver metastasis was correctly identified.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 3.

¹⁸F-FDG PET images in anterior projection display (A) of patient evaluated for suspected recurrence did not clearly show abnormal foci of uptake in abdomen. Focus of moderate activity in left upper quadrant of abdomen was interpreted as splenic activity (arrow). Diagnostic impression was definitively benign. Transverse PET image (B) and fused image (D) at level of liver clearly showed small focus of mild activity at edge of a lesion on CT image (C). This lesion was interpreted as clearly malignant, and patient was classified as having metastatic liver disease. Metastasis from mucinous colorectal carcinoma was confirmed at surgery. With PET/CT, disease was correctly staged, unlike false-negative interpretation with PET alone.

The extrahepatic intraabdominal tumor status was incorrectly assessed for 5 patients with PET and 3 patients with PET/CT. One myelolipoma of the adrenal was falsely interpreted as malignant with both PET and PET/CT. One metastatic periportal lymph node was found at pathologic examination in a patient evaluated for liver metastasis resectability but was missed by both PET and PET/CT. Peritoneal tumor implants were incorrectly interpreted as bowel activity with PET/CT in another patient. In one patient, bowel activity was incorrectly interpreted as a pelvic lymph node with PET but correctly classified as benign with PET/CT. In another patient, a focus of pelvic activity was incorrectly classified as a metastatic lymph node with PET. It was shown to be localized in bone with PET/CT and thus was characterized correctly as a bone metastasis. In another patient, a porta hepatis lymph node was missed on PET but identified on PET/CT.

The distant metastasis status (extraabdominal) was incorrectly assessed for 2 patients with PET. In contrast, the status was correctly assessed for all patients with PET/CT. Degenerative spine changes were incorrectly interpreted as metastatic bone disease with PET for 1 patient. A bone metastasis was incorrectly interpreted as a metastatic pelvic lymph node with PET. The presence or absence of malignancy was correctly assessed for all patients with PET/CT, except for 1 with a false-positive result. PET had 2 false-positive and one false-negative results.

Overall staging was incorrectly assessed for 4 patients with PET/CT and 8 patients with PET (Table 4). PET/CT had 2 false-positives and 2 false-negatives, whereas PET had 4 false-positives and 4 false-negatives. Three of these incorrect stagings were congruent for both PET and PET/CT. PET and PET/CT had an accuracy for overall staging of 78% (95% confidence interval, 62%–89%) and 89% (95% confidence interval, 74%–96%), respectively (P = 0.4).

For lesions unrelated to the colorectal cancer but suspected to be malignant, there were 4 false-positives with PET/CT and 3 false-positives with PET. False-positive interpretations occurred with both PET and PET/CT for 3 patients (1 with an ¹⁸F-FDG–avid Warthin’s tumor, 1 with an ¹⁸F-FDG–avid benign thyroid nodule, and 1 with focal physiologic oropharyngeal uptake). Markedly asymmetric physiologic tonsillar activity was incorrectly interpreted as malignant with PET/CT for 1 patient.