Colorectal cancer (CRC) is a common disease and belongs, with lung, breast, and prostate cancer, to the most frequently seen neoplasms in Western countries. In the United States, CRC frequency ranks third in men and second in women. CRC is the fourth leading cause of cancer mortality because it has a better prognosis than the other common cancers (1–3). During 1990–1995, the annual percent change in incidence was –2.3% and in mortality –1.1% compared to the previous half decade (2).
In 2004, a total of 146,940 new CRCs are likely to be diag- nosed, and 56,730 are expected to die of their disease (3).
Incidence of CRC is similar in men and women (M/F ratio
= 1.004), as is the mortality rate (M/F ratio = 0.997).
About 70% of the patients have resectable disease, but only two-thirds can be cured by resection. In the remain- ing one-third of these patients, recurrence is diagnosed in the first 2 years after resection.
A 5-year survival rate of 62% is reported in the United States compared to 41% in the European Union (EU). The lifetime risk of developing colorectal cancer in the United States appears to be 5.78% (1 in 17) in men and 5.55% (1 in 18) in women.
Advances in diagnostic imaging technology have been directed to (1) help establish the diagnosis, (2) stage the extent of disease, and (3) enhance the development of ac- cepted treatment protocols by monitoring response to therapy. Improved patient outcome may be expected if these goals are achieved. Various imaging modalities are available for this purpose, including the anatomic (e.g., radiography, computed tomography, sonography, mag- netic resonance imaging) and functional modalities (e.g., molecular imaging, radioimmuno- and receptor scintigra- phy, magnetic resonance spectroscopy). Positron emis- sion tomography (PET) is based on imaging of biochemical processes in vivo, and creates tomographic images similar to imaging modalities such as computed tomography (CT) and magnetic resonance imaging (MRI). PET is unique because it supplies an image repre- senting the metabolic activity of the underlying molecular processes (4, 5). The basics of PET are discussed elsewhere
in this volume. The place of PET in stratifying patients with colorectal cancer and in the workup of clinical prob- lems is discussed in this chapter. In this respect, it is ap- propriate to speak of correlative imaging, in which all imaging modalities have their specific contribution and are not seen as competitive modalities (6).
The declining mortality rate of CRC is related to screen- ing programs put in place, such as the hemoccult test and colonoscopy in asymptomatic individuals above age 50.
Lifestyle changes contributed in part to early detection with subsequent surgery and a decrease in mortality.
Sophisticated imaging such as PET is expensive and not a realistic option for screening purposes. An interesting study from Japan (7) reported an incidence of 2.1% neo- plasms found within 1 year after screening an asympto- matic group of more than 3,100 people with
18F-2-deoxy-D-glucose (FDG)-PET imaging. PET was true positive in 54% of tumors, and most of the false negatives were in the genitourinary (GU) tract. Obviously, PET cannot be cost effective under these circumstances.
The cost issue of PET for patient management in CRC has been addressed (8). The main advantage of PET in on- cology is that with one injection and imaging session the whole body can be imaged in a tomographic mode.
Generally, the more conventional approach with multiple CT or MR scans covering head, chest, abdomen, and pelvis are more expensive than a single whole-body PET.
The role of PET in colorectal cancer is a popular topic for review articles, in part because of the high incidence and mortality of this disease. Overviews emphasizing the clinical approach have been published by Arulampalam et al. (9) from a surgical point of view, by Akhurst and Larson (10) and by Anderson and Price (11) from the on- cologist’s perspective, and by Rohren et al. (12) from the imaging expert point of view.
The latest technologic development concerns dual- modality imaging, by combining a multislice CT and a PET scanner into one system. In addition, to providing high-resolution CT images in all planes and projections, CT-based attenuation correction can be performed (13).
147
10
PET and PET/CT Imaging in Colorectal Cancer
Christiaan Schiepers and Peter E.Valk
Historical Perspective
Despite the advancement of conventional diagnostic methods, such as computed tomography (CT), both cross- sectional and helical, magnetic resonance (MR) imaging, and external and endoultrasonography (US), early detec- tion of colorectal cancer remains problematic for pri- maries as well as recurrences. The addition of serum tumor markers or radioimmunoscintigraphy has not significantly improved detection at an early stage, ham- pering curative resection. Assessment of disease extent or tumor burden is necessary for proper patient selection of surgery with curative intent, or stratification to chemotherapy and/or radiation treatment for patients with advanced disease. Most distant metastases occur in the liver or lungs, and adequate staging is necessary to exclude patients with more extensive disease (14). Long- term survival after attempted curative resection of recur- rent disease is only 35%. Therefore, appropriate noninvasive staging plays a pivotal role in selecting pa- tients who would benefit from surgery and avoiding un- necessary surgery with major morbidity in those with unresectable disease.
Serum carcinoembryonic antigen (CEA) is used to monitor for possible recurrence. This technique has a re- ported sensitivity of 59% and specificity of 84% but cannot determine the location of recurrence (15). Lower gastrointestinal (GI) radiography with contrast has been used for detection of local recurrence with accuracy in the vicinity of 80%, but it is only 49% sensitive and 85%
specific for overall recurrence. CT has been the conven- tional imaging modality used to localize recurrent disease with an accuracy of 25% to 73%. CT cannot reliably dis- tinguish postsurgical changes from local recurrence and is often equivocal (16–18). CT of the abdomen misses hepatic metastases in about 7% of patients and underesti- mates the number of lobes involved in one-third of pa- tients. In addition, CT commonly misses metastases to the peritoneum, mesentery, and lymph nodes. Among the pa- tients with negative CT, half will have nonresectable lesions at the time of exploratory laparotomy. The results for MR imaging are comparable to CT (19).
FDG (2-18F-fluoro-2-deoxy-D-glucose) has been used most frequently as the tracer for metabolic imaging (20).
FDG is able to measure changes in glucose utilization, which is enhanced in cancer. After the introduction of PET for whole-body imaging (21), clinical studies have been aimed at staging CRC and evaluating the disease extent. In clinical practice, “PET whole-body imaging”
usually extends from the base of the brain to the upper thighs. The yield in imaging the brain for metastases is low (1.5%), as was reported in the UK (22), mainly because of the high FDG uptake in gray matter.
Metastases below the inguinal regions are uncommon in many cancers. Traditionally, three general regions of metastatic spread of CRC in the body are distinguished:
local, hepatic, and extrahepatic. Hypermetabolic foci in these regions are suspicious for primary tumor, recur- rence, or metastasis and form the basis of metabolic imaging in oncology.
The role of PET for primary CRC has not been studied systematically. The study by Abdel-Nabi et al. (23) showed excellent sensitivity for PET but poor specificity (43%).
Their CT specificity of 37% is lower than reported in the literature, so that biased patient selection cannot be ex- cluded. Falk et al. (24) showed in a mixed group of 16 pa- tients that PET was superior to CT and had a positive predictive value (PPV) of 93% and a negative predictive value (NPV) of 50%. Ruhlmann et al. (25) directed their study to the clinical performance in a mixed group of pa- tients and did not find differences between primary and recurrent disease. From the sparse evidence, there is so far no established role for FDG-PET in primary CRC.
The majority of studies deal with recurrent disease. The first series of systematically studied patients came from Heidelberg and demonstrated the impact of PET versus CT in the differential diagnosis of a pelvic mass (26). Ito et al. (27) reported the first study of PET versus MR imaging.
Later, studies with more patients were published that confirmed these early findings. Table 10.1 provides the combined data for FDG-PET imaging in CRC as reported in the meta-analysis by Gambhir’s group (28). They re- viewed articles published during 1990–1999 that reported information on the use of new medical technology and se- lected 11 studies that fulfilled their strict guidelines for technology evaluation. Sensitivity and specificity in the meta-analysis were all 95% or higher, except for the specificity at extrahepatic sites, which revealed about 25%
false-positive results. The studies included in the meta- analysis come from institutions in the United States, Europe, and Australia and showed remarkable consis- tency of results. From the reported literature in the previ- ous decade, it is clear that PET has established a role in the evaluation and staging of patients with recurrent CRC.
Technique and Methodology
High rates of glycolysis are found in many malignant tumor cells (29) and high uptake of FDG is usually associ- ated with a high number of viable tumor cells and high ex- pression of glucose transporters (30). Increased FDG
Table 10.1. Clinical performance of 18F-2-deoxy-D-glucose (FDG)-positron emission tomography (PET) in recurrent colorectal cancer.
Body region Patients Sensitivity Specificity
Local 366 94.5% 97.7%
Hepatic 393 96.3% 99.0%
Whole body 281 97.0% 75.6%
Source: Adapted from the meta-analysis of Huebner et al. (28).
uptake is by no means specific for neoplasms (31). Inflam- matory processes may also have increased uptake, and false-positive results have been reported in presacral abscess, tuberculosis, fungal infections, acute postopera- tive and radiation changes, pancreatitis, or diverticulitis.
In the proper clinical context, many of these lesions will probably not be confused with local or regional disease, but all may contribute to false-positive distant findings.
The body imaging mode has become the standard for PET in oncology (21). After the uptake period, the patient is positioned in the scanner, and sequential acquisitions are performed along the length of the patient’s body. As the emitted photons pass through the tissues, varying degrees of attenuation and scatter affect the final number that reach the detectors. A separate transmission scan is performed to correct for attenuation effects. The trans- mission scan can be acquired after tracer injection, en- abling higher patient throughput. No difference in lesion detection was found between corrected and nonattenua- tion-corrected images for a variety of tumors (32, 33).
However, these studies used filtered back-projection to re- construct the images, instead of currently used iterative reconstruction methods. In the late 1990s, there was no consensus on the necessity of attenuation correction (34), but most facilities now use low noise attenuation correc- tion with CT and scatter correction to provide more true- to-life depiction of metabolic activity in the body.
The spatial resolution of a modern CT or MR system is better than that of a PET system. However, this is not the only determining factor in detecting abnormalities. The
“contrast resolution” or difference between the lesion and its surroundings helps determine the presence of disease.
The accuracy of anatomic images is limited by a lower lesion to background contrast. The sole exception is the lung, where anatomic contrast between solid lesions and air-filled lung is high. The target-to-background ratio is usually much higher for FDG-PET. Except for the lung, the high metabolic contrast predominates over anatomic contrast. Sensitivity of PET is also affected by lesion size.
Metabolically active lesions as small as 5 mm have been detected with FDG-PET.
True quantification of FDG metabolism is usually not performed for CRC because tumor kinetics are not known. The technique is more demanding of resources than imaging alone, and metabolic rate determination has been found to offer no diagnostic advantage. For therapy monitoring and prediction of prognosis, compartmental modeling may be able to discriminate responders from partial or nonresponders, but larger groups of patients need to be investigated (35).
Evaluation of static PET images can be performed visu- ally, or semiquantitatively using the standardized uptake value (SUV) or a lesion-to-background ratio. The SUV is the measured activity in the lesion in mCi/mL divided by injected dose, expressed as mCi/kg of body weight. Strauss et al. (26) reported SUVs of 1.1 to 4.2 for pelvic recurrences, and Takeuchi et al. (36) showed that an SUV cutoff of 2.8
diagnosed local recurrences with 100% accuracy. Abdel- Nabi et al. (23) found SUVs of 2.8 to 14.5 for primary bowel cancers. Semiquantitative evaluation offers a more-objec- tive way of reporting lesion uptake than visual image inter- pretation and is useful for comparing lesion activity in consecutive studies (treatment monitoring). However, visual interpretation appears sufficient for clinical needs and is equally effective for a one-time diagnosis.
Modifications that may improve the semiquantitative evaluation of FDG uptake include normalizing the dose to the body surface area (37) or the lean body weight (38) instead of the total weight of the patient. This SUV modification may be significant because the concentration of FDG is higher in muscle than in fat. For a concise comparison of the simplified quantitative analysis methods of FDG uptake, the article of Graham et al. (39) is recommended.
Patient Preparation and Diagnostic Protocol
Patients are studied in a prolonged fasting state to produce low insulin levels and induce low rates of glucose utilization by normal tissues, including voluntary muscles and myocardium. Malignant tissues are less dependent on hormone regulation, and thus will have higher uptake when compared to the surrounding normal tissues. The typical duration of a PET oncology protocol is about 2 h.
A dose of 250 to 500 MBq (7–15 mCi) of FDG is the usual activity administered. After an uptake period of 45 to 75 min, the patient is asked to void and is positioned in the scanner. Interleaved emission (3–5 min per bed position) and transmission (2–3 min per bed) scans are acquired. A scan of the body has a total acquisition time of 35 to 50 min. Patients are usually scanned from feet to head, taking advantage of the low bladder uptake after voiding at the beginning of acquisition. The transmission scans can be acquired before, after, or interleaved with the emis- sion scans. The typical duration of a dedicated “PET- only” oncology protocol is 1.5 to 2 h, depending on the duration of the postinjection uptake period.
Dual-modality imaging with PET/CT is used in many institutions. The imaging protocols are variable depend- ing on the CT settings, that is, for attenuation correction only (low mAs sufficient) (40), diagnostic quality, and the use of contrast (intravenous and/or oral).
At University of California–Los Angeles (UCLA), oral contrast is given for the delineation of the bowel. A volume of 900 mL Ready-Cat (with 2% barium sulfate, but without glucose) is given orally in three portions during a period of 75 min before the acquisition starts. Our proto- col comprises one CT acquisition of the torso, that is, from the base of the skull to the midthigh level, followed by a PET scan of the same area, with the arms up.
Currently, we have a dual-slice CT that provides images
of diagnostic quality with the following settings: 130 kVp, 80 mAs, pitch 1, and reconstruction slice thickness 5 mm.
The helical CT takes about 80 sec to image the upper body. Intravenous contrast is given when ordered by the referring oncologist. Nonionic contrast (Omnipaque) in a volume of 100 to 130 mL at a rate of 1.5 mL/s is adminis- tered. The I.V. contrast may cause regions of high attenua- tion that lead to typical CT-based attenuation artifacts and pseudo-FDG uptake (41, 42). The FDG dose is 0.19 mCi/kg (7.0 MBq/kg) with a maximum of 15 mCi (555 MBq). The uptake interval between tracer administration and start of acquisition is 1 h. During the uptake period, the patient is comfortably resting in a chair with armrests in a dimly lit room without radio or television to minimize brain stimu- lation. The patient is covered with blankets to prevent shivering and activation of brown adipose tissue. Just before acquisition, the bladder is emptied. Our PET scanner has fast detectors (lutethium-ortho-silicate, LSO) allowing for imaging of 1 to 4 min per bed position to ac- cumulate the necessary counts. Body weight determines the imaging time: less than 75 kg, 1 min/bed; less than 100 kg, 2 min/bed, etc. (43). The overall PET acquisition takes 6 to 26 min. With this setup, a standard whole-body PET/CT can be completed within 45 min. An additional CT of the chest is acquired during deep inspiration.
The cortex of the brain uses glucose as its substrate;
therefore, FDG accumulation is high. Evaluation of metastatic disease to the brain with FDG is limited (22).
Diffuse thyroid uptake can be a normal variant and is seen in patients with thyroiditis and Graves’ disease. Metastatic cervical lymph nodes are occasionally seen in patients with colorectal carcinoma, and differentiation from thyroid uptake is important. In a typical fasting state, the myocardium primarily utilizes free fatty acids, but post- prandially or after a glucose load, it favors glucose. When the chest is evaluated with FDG to assess the presence of metastases, a 12-h fasting state may be preferable to achieve low myocardial FDG uptake. Despite these mea- sures, uptake in the myocardium is seen with high fre- quency in cancer patients.
FDG is filtered by the glomerulus and is only partly re- absorbed, unlike glucose. Thus, accumulation of FDG in the renal collecting system and urinary tract is normal.
Hydration to promote diuresis, or administration of di- uretics, can reduce the urinary activity seen on PET images. High bladder activity can result in positive and negative image artifacts if the images are not corrected for tissue attenuation. Correction for attenuation and itera- tive image reconstruction circumvent such artifacts, and bladder catheterization is to be avoided. Some centers use antiperistalsis and antimotility drugs to overcome promi- nent bowel uptake. The large bowel usually shows some FDG uptake, but this can be differentiated from abnormal activity by demonstrating a physiologic activity pattern along the bowel trajectory in the stack of 3D images. Many institutions in Europe utilize mild sedatives for patient comfort during the relatively lengthy acquisition.
To avoid misinterpretation of FDG images, it is critical to standardize the environment of the patient during the uptake period; to examine the patient for postoperative sites, tube placement, stoma, etc.; and to be aware of any invasive procedures or therapeutic interventions.
Accuracy of Metabolic and Anatomic Imaging in Recurrent Colorectal Cancer
Overall Accuracy of FDG-PET and Conventional Imaging
The aims and methods of studies evaluating FDG-PET in recurrent CRC have developed with time and experience.
The objectives in recurrent disease are quite different from primary staging of CRC. The early studies of Strauss et al. (26) and Ito et al. (27) were aimed at feasibility of the technique and general comparison to CT and MR imaging. Subsequent studies addressed tumor staging for better stratification of patients before contemplated surgery (24, 44–47). The accuracy of PET in CRC was eval- uated (47–49) as well as the impact of PET findings on management decisions in routine practice (25, 48–51). In presenting the findings of these studies, we included only reports of 50 or more patients that were published in peer-reviewed journals and that compared PET to CT. The referred patient populations vary from report to report, for instance, suspected versus diagnosed recurrence, single versus multiple sites of recurrence, first versus second recurrence, prospective versus retrospective study design, potentially resulting in referral bias. In addition, methods and PET systems used in the studies varied, for example, limited versus extended (i.e., upper body) field of view, acquisition duration, correction for attenuation, and scatter. Image reconstruction method, such as stan- dard filtered back-projection versus iterative methods, was not always the same in the reported series. Moreover, lesion-based instead of patient-based analysis was fre- quently used. A number of studies were reported sequen- tially from the same institution and included some of the same patients; in which case we selected only one repre- sentative study. Despite the multiple variations, different objectives, and inconsistencies involved, the end results of the studies that met the criteria are collated in the tables, and are quite similar.
Table 10.1 shows the average sensitivity and specificity of PET as computed in the meta-analysis (28). The meta- analysis does not specify the results of conventional imaging, such as CT and ultrasound (US), and the com- parison standard is not provided. For this reason, we pooled the data of individual reports, that is, table 3 from Schiepers and Hoh (52) and table 2 of Delbeke and Valk (53), to calculate the overall diagnostic accuracy of PET and CT for whole-body imaging. For the CRC data with
direct comparison between modalities, 301 patients were accumulated between 1993 and 1997. The weighted average sensitivity was 95.5% for PET [95% confidence in- terval (CI), 93%–98%] and 71.4% for CT (95% CI, 66%–77%). The weighted average specificity was 88.3%
for PET (95% CI, 84%–92%) and 85.4% for CT (95% CI, 81%–90%).
In the late 1990s, the single largest study of the accu- racy of PET imaging with FDG was reported by Valk et al.
(51), who compared the sensitivity and specificity of PET and CT for specific anatomic locations. They found that PET was more sensitive than CT in all locations except the lung, where the two modalities were equivalent. The largest differences between PET and CT were found in the abdomen, pelvis, and retroperitoneum, where more than one-third of PET-positive lesions were negative by CT.
PET was also more specific than CT at all sites except the retroperitoneum, but the differences were smaller than the differences in sensitivity.
Local Recurrence
The results obtained by PET in detecting local recurrence that were reported in the early studies (26, 27) were cor- roborated in larger patient groups (Figure 10.1). Table 10.2 shows the results of the series evaluating local recur- rence that met the criteria. PET was 17% more accurate than CT for this indication. Keogan et al. (54) studied re- current rectal cancer, using pathology as comparison
standard. They compared visual versus quantitative inter- pretation of PET studies and found the two methods of analysis to be equivalent. Whiteford et al. (55) studied 101 patients, 70 of which were evaluated for locoregional re- currence. They found that PET sensitivity was about 20%
higher than for CT plus colonoscopy. They also found that PET sensitivity varied with histological tumor type, having a lower sensitivity for mucinous (58%) than non- mucinous CRC (92%).
Hepatic Metastasis
The aim of the presurgical workup is to distinguish iso- lated resectable disease, that is, local recurrence or soli-
Figure 10.1. Coronal FDG-PET images in a 73-year-old man who underwent re- section of colon cancer 4 years earlier. CT showed a mass in the lower lobe of the right lung. PET showed the pulmonary metastasis (b, ->) and also demon- strated local recurrence in the left lower quadrant (a, => ). Local recurrence was confirmed at surgery. Images are cor- rected for attenuation and scatter.
Table 10.2. Detection of local recurrence in colorectal cancer.
PET CT Reference (sensitivity– (sensitivity–
First author number Year Patients specificity) specificity)
Schiepers 48 1995 76 93%–97% 60%–72%
Valk 51 1999 115 97%–96% 68%–90%
Whiteford 55 2000 70 90%–90% 71%–85%
Weighted 261 94.0%–94.7% 66.5%–83.4%
average
Source: From Valk PE, Bailey DL, Townsend DW, Maisey MN. Positron Emission Tomography: Basic Science and Clinical Practice. Springer-Verlag London Ltd 2003, p. 562.
a b
tary liver metastasis, from advanced disease. By correct staging, patients with widespread metastasis may be identified in whom surgery is not an option, thus sparing the patient extensive surgery with its associated morbid- ity. The actual selection of patients with recurrent cancer results in 5-year survival rates of only 20% to 30% after secondary “curative” surgery (2).
Table 10.3 shows results obtained by PET and CT in di- agnosis of hepatic metastasis, showing smaller differences than in diagnosing local recurrence. For detection of hepatic metastasis, PET was 12% more sensitive than CT (Figure 10.2). PET specificity is about equal to CT, except for Delbeke’s study. Delbeke et al. (49) compared PET to conventional CT as well as CT portography, which is an invasive procedure. The sensitivity of PET (91%) was lower than CT portography (97%), but the specificity was 12% higher, particularly at postsurgical sites. Fong et al.
(56) and Delbeke et al. (49) found that the sensitivity of PET in detecting hepatic metastases varied with lesion size, as would be expected. Fong found that only 25% of lesions smaller than 1 cm were detected by PET compared to 85% of lesions larger than 1 cm. By excluding the lesions less than 1 cm, Delbeke found that the sensitivity increased about 8% for both PET and CT. This 25% versus 8% detectability change, found between two series re- ported only 3 years apart, highlights some of the
difficulties in interpreting small lesions and may be attrib- uted to the rapidly evolving technologic changes in multi- slice helical CT and high-resolution PET systems. Topal et al. (57) confirmed in a series of 99 patients that PET was a diagnostic tool complementary to CT and US in selecting patients for potentially curative liver resection. Similar findings were reported in a small series of 14 patients by Boykin et al. (58).
Fernandez et al. (59) investigated the 5-year survival after resection of metastasis from CRC. They established the 5-year survival of patients with conventional diagnos- tic imaging from the literature by pooling the data of 19 studies with a total of 6,090 patients. The 5-year survival rate was 30% and appeared not to have changed over time. These results were compared to their group of 100 patients with hepatic metastases, who were preoperatively staged for resection with curative intent. Addition of a preoperative FDG-PET study improved the 5-year survival rate to 58%, indicating that they were able to define a sub- group after conventional imaging that has a better prog- nosis (59). The main contribution was in detecting occult disease, leading to a reduction of futile surgeries.
Distant or Extrahepatic Metastasis
PET for staging of disease involvement may yield unex- pected lesions that often appear to be metastases (Figures 10.3, 10.4); this is a feature of the PET imaging technique in which the whole body can be examined after one injec- tion of an FDG dose. Table 10.4 provides results of detec- tion of unknown metastases. In about 25% of patients referred for restaging during their preoperative evalua- tion, occult disease was found that was not suspected clin- ically or detected with conventional imaging.
PET using FDG has a proven record of accomplishment for characterizing indeterminate pulmonary nodules that can be metastases from CRC (60). Lai et al. (46), in their study of 34 patients, found that FDG PET was especially useful for detecting retroperitoneal and pulmonary metastases. Schiepers et al. (48) found that half the chest lesions on PET were false-positives.
Table 10.3. Detection of hepatic metastases.
PET CT Reference (sensitivity– (sensitivity–
First author number Year Patients specificity) specificity)
Schiepers 48 1995 76 94%–100% 85%–98%
Delbeke 49 1997 61 91%–95% 81%–60%
Valk 51 1999 115 95%–100% 84%–95%
Whiteford 55 2000 101 89%–98% 71%–92%
Weighted 353 92.4%–98.6 80.0%–88.7%
average
Source: From Valk PE, Bailey DL, Townsend DW, Maisey MN. Positron Emission Tomography: Basic Science and Clinical Practice. Springer-Verlag London Ltd 2003, p. 562.
Figure 10.2. Transverse FDG-PET images in a 55-year-old man with a history of colon cancer 15 months earlier and status postresection. CT demonstrated a large metastasis in the right lobe of the liver posteriorly. PET shows this lesion (a) and also demonstrates multiple small metastases that were not seen on CT (b, c).
a b c
Clinical Indications for PET Imaging in Colorectal Cancer
Evaluation of Elevated Serum CEA Level
Flanagan et al. (61) reported the use of FDG PET in 22 pa- tients with elevation of serum CEA level after resection of primary CRC and negative findings by conventional diag- nostic procedures (Figure 10.5). Sensitivity and specificity of PET for tumor recurrence were 100% and 71%, respec-
Figure 10.3. Coronal FDG-PET images in a 48-year-old man, with a history of resection of colon cancer 11 months earlier, who was found to have a lesion in the right lobe of the liver on CT exami- nation. The remainder of the CT images of the abdomen and pelvis were normal.
Needle biopsy confirmed the hepatic metastasis, and the PET study was per- formed for preoperative staging. PET images show the hepatic metastasis (b) with an extrahepatic focus of abdominal tumor, inferior to the right lobe of the liver (a). The PET findings were con- firmed at surgery.
Figure 10.4. Coronal (a) and sagittal (b) FDG-PET image in a 51-year-old man with a history of resection of rectal cancer 3 years earlier. CT demonstrated a lesion in the lower zone of the right lung, and biopsy confirmed recurrent rectal cancer. CT imaging showed no other abnormality, and PET imaging was performed for preoperative staging. PET showed high uptake in the lung metastasis (a) and also showed metastasis in a thoracic vertebra, thereby excluding surgical resection of the lung lesion. The patient was treated with chemotherapy and irradiation.
a b
a b
Table 10.4. Detection of occult metastases with FDG-PET.
Reference Unsuspected
First author number Year Patients metastases
Schiepers 48 1995 76 14 (18%)
Lai 46 1996 34 11 (32%)
Delbeke 49 1997 52 17 (33%)
Valk 51 1999 78 25 (32%)
Pooled average 240 67 (27.9%)
Source: Updated from Valk PE, Bailey DL, Townsend DW, Maisey MN. Positron Emission Tomography: Basic Science and Clinical Practice. Springer-Verlag London Ltd 2003, p. 563.
tively. Valk et al. (51) reported a sensitivity of 93% and specificity of 92% in a similar group of 32 patients, 18 of whom proved to have recurrence on surgical evaluation or clinical follow-up. Maldonado et al. (62) reported for PET a sensitivity of 94% and specificity of 83% in a group of 72 patients. Flamen et al. reported, in a retrospective series of 50 patients, 79% sensitivity and 89% positive pre- dictive value (63). From these studies, it is apparent that PET has a definite place in the workup of an unexplained CEA rise in CRC.
Preoperative Staging of Recurrent CRC
Table 10.4 shows results obtained by PET using FDG in detection of unsuspected distant metastases in patients undergoing preoperative evaluation, which is probably the most important function of PET in recurrent CRC.
Table 10.4 shows that FDG-PET discovered disease at un- suspected sites in 28% of patients referred for preopera- tive staging. Lai et al. (46), in their study of 34 patients, found that FDG-PET was especially useful for detecting retroperitoneal and pulmonary metastases.
Surgical findings in patients who underwent surgery following staging by PET were reported by Valk et al. (51).
Forty-two patients with PET evidence of localized recur- rence were submitted to surgery. Seven of these patients (17%) had nonresectable tumor at surgery. Two patients with liver metastasis also had diffuse peritoneal tumor and 1 had multiple small liver lesions that had not been
detected by PET. Three patients with focal pelvic recur- rence and 1 with focal abdominal recurrence on FDG-PET had undetected diffuse tumors that could not be resected.
Strassberg et al. (64) reported 43 patients with hepatic re- currence who were evaluated before surgery. PET findings contraindicated surgery in 6 of these patients (14%). Of the 37 patients who underwent laparotomy, liver resection was performed in all but 2 (5%). Median follow-up in the 35 patients who had resection was 24 months, and the Kaplan–Meyer estimate of overall 3-year survival was 77%, with a 3-year disease-free survival of 40%. These numbers were higher than had been reported previously (17). PET increased the rate of resectability at surgery and decreased the rate of rerecurrence following resection by detection of tumor preoperatively that had not been found by conventional imaging.
Three studies of preoperative staging by Beets et al. (44), Schiepers et al. (48), and Flamen et al. (50) reported results from the same institution over several years and contained overlapping groups of patients. In the series reported by Flamen et al. (50), the objective was to compare the number of lesions detected by PET and conventional imaging methods and study the discrepancies. All data were reprocessed with iterative reconstruction and reread by two nuclear medicine physicians. The accuracy of inter- pretation was similar to the earlier report of Schiepers et al. (48) indicating reproducibility of findings between dif- ferent reconstruction techniques and different interpreting nuclear medicine specialists. In 10% of lesions, PET and conventional imaging findings were discordant, and in the
Figure 10.5. Sagittal (a) and transverse (b) FDG-PET image in a 73-year-old man with a history of colon cancer resection 22 months earlier, followed by 6 months of chemotherapy. The patient was found to have carcinoembryonic antigen (CEA) elevation, negative CT examination of the abdomen, and negative ultrasound (US) examination of the liver. PET demonstrates a hypermetabolic focus in the anterior epigastrium (coronal slice), anterior to the left lobe of the liver (transverse slice). Metastasis was con- firmed by biopsy.
a b
other 90%, PET contributed additional diagnostic infor- mation. Almost half the discordant findings were locore- gional, near the site of the original primary (50).
Impact of PET findings on Management Decisions
PET using FDG permits earlier evaluation for treatment by diagnosing recurrence when CT is still negative, and avoids unnecessary surgery by demonstrating nonre- sectable disease in some patients with known recurrence (65, 66). Whether PET will improve the resectability-rate by earlier tumor diagnosis or will reduce the rate of re-re- currence of undetected residual tumor at surgery remains to be established. Tabulating the studies that specifically addressed preoperative evaluation yields the results in Table 10.4. In about a quarter of patients, unsuspected metastases were detected. Although these findings may not always change the clinical stage of the disease, some cases will be reclassified to the nonsurgical category.
Surgical decision making in CRC has been reported fre- quently from different countries (50, 51, 65, 66).
The question of change in patient management was ad- dressed by several investigators, and the impact of PET varied from 23% in the series of Fong et al. (56) to 40% in the series of Beets et al. (44). Table 10.5 shows that, overall, a change in management was observed in about one-third of patients with recurrent CRC. A detailed analysis of the type of management changes can be found in the meta-analysis (25). Altogether, in 121 of 349 pa- tients (34.7%), a management change was reported.
Surgery was correctly avoided in more than half of the pa- tients (53.7%), and unintended surgery was correctly initi- ated in 20.7% of patients. PET incorrectly suggested a change in management in 9.1% of patients.
Management change was addressed in a collaborative study from UCLA and the Northern California PET
Imaging Center from the perspective of the referring physician. Questionnaires were sent to referring physi- cians after a PET study had been performed for diagnosis or staging of recurrent CRC determine whether the PET findings had had an impact on management decisions.
Clinical stage was changed in 42% of patients. A treatment modality switch (e.g., from surgery to chemotherapy, from medical treatment to surgery) was reported in 37%
of patients (67).
The great variability in study design of the published series evaluating management effect was addressed by Gambhir’s group in their meta-analysis (28). Using rigor- ous selection guidelines based on U.S. medical payer source criteria, only 10 studies between 1990 and 1999 listed on MEDLINE qualified for inclusion. One additional study was selected after direct contact with the primary author for clarification of the text and tables (25). These studies were performed in four different countries:
Australia, Belgium, Germany and the United States (four institutions), comprising seven different institutions in total. The reports from the same institution used overlap- ping patient groups. The results given in Table 10.1 were obtained from this meta-analysis.
In a prospective study, Kalff et al. (68) investigated the clinical impact in recurrent CRC. They found that FDG-PET directly influenced management in 59% of pa- tients. Their findings confirmed earlier retrospective reports (46, 51, 56, 66). Management changes in recur- rent and metastatic CRC was also studied by Desai et al.
(69) in 114 patients. They found that FDG-PET altered therapy in 40% of patients. They concluded that PET should be used in the management of patients with re- current CRC who are considered for potentially curative surgery.
The greater sensitivity of PET compared to CT in diag- nosis and staging of recurrent tumor is the direct result of two factors: (1) early detection of abnormal tumor metab- olism, before changes have become apparent on anatomic imaging, and (2) whole-body metabolic imaging, which permits diagnosis of tumor when it occurs in unusual and unexpected sites. PET imaging allows identification of areas with abnormal metabolism, which may guide subse- quent CT and possible biopsy of these lesions. Thus, exact anatomic location and potential resectability can be as- sessed. In the region of the primary, PET is helpful in de- tecting nodal involvement and differentiating local recurrence from postsurgical changes, indications for which CT has known limitations. For hepatic lesions, PET is helpful to assess the number of lesions and liver lobes involved. For distant, extrahepatic abnormalities, PET may be used to characterize abnormal lesions diagnosed with conventional imaging procedures or may identify occult metastasis.
Three indications for metabolic imaging have been es- tablished in patients with known or suspected recurrent
Table 10.5. Change in patient management introduced by FDG-PET.
Reference Changed
First author number Year Patients treatments
Beets 44 1994 35 14 (40%)
Lai 46 1996 34 10 (29%)
Valk 51 1999 78 24 (31%)
Fong 56 1999 40 9 (23%)
Whiteford 55 2000 101 26 (26%)
Pooled 288 83 (28.8%)
Source: Updated from Valk PE, Bailey DL, Townsend DW, Maisey MN. Positron Emission Tomography: Basic Science and Clinical Practice. Springer-Verlag London Ltd 2003, p. 566.
colorectal carcinoma: (1) rising serum CEA level with neg- ative conventional diagnostic imaging; (2) characteriza- tion of equivocal lesion(s) on conventional imaging; and (3) staging of recurrent disease. FDG-PET scans for staging and restaging of CRC are reimbursed by national insurers in the United States, U.K., Germany, Italy, Belgium, and Switzerland.
Therapy Monitoring
The use of FDG-PET for monitoring chemotherapy has been reported (51). Accurate information about the effec- tiveness of radiotherapy and/or chemotherapy for col- orectal cancer is important for continuation of the selected therapeutic regimen or switching to an alterna- tive regimen. The number of studies in which PET has been used successfully for monitoring therapy is small, but monitoring chemotherapy in advanced colorectal cancer appears promising. Strauss and collaborators (70) studied the response to therapy with PET for various tracers. In 1991, they reported that tumor perfusion (studied with 15O-water) is highly variable and does not correlate well with treatment response. Using labeled fluorouracil (FU) and FDG for measuring response to chemotherapy and/or radiotherapy, they found a correla- tion between uptake and outcome. High uptake of FU and low uptake of FDG correlated with good treatment re- sponse. The pharmacokinetics of FU suggest that FU may be superior to FDG in predicting the response to treat- ment on an individual patient basis (71). Strauss and col- laborators (72–74) studied PET images of FU metabolite concentration and compared these with growth rate before and during chemotherapy, determined by CT volume measurements. Lesions with low FU uptake had a significant increase in volume and no response to treat- ment. Findlay et al. (75) evaluated the metabolism of liver metastases from CRC before and after treatment in 18 pa- tients. The findings were compared to change in size on CT, and results were expressed as a tumor-to-liver uptake ratio. By this means they were able to differentiate even- tual responders from nonresponders on both lesion-by- lesion and patient-based analysis. Regional therapy of hepatic metastases by chemoembolization can also be monitored with FDG (76). FDG uptake decreased in re- sponding lesions and the presence of residual uptake was used to guide further regional therapy. The Sloan- Kettering group also found good correlation between the response of hepatic metastasis to chemotherapy and FDG uptake (10). Their most promising finding was the posi- tive correlation between PET findings at 4 weeks and CT findings at 12 weeks, whereas the MR imaging at 4 weeks did not show a change in tumor volume.
Assessing response to radiation therapy (RT) was eval- uated in 21 patients with recurrent disease by Haberkorn et al. (77). Reduction in FDG uptake was found in half the
patients and correlated with the palliative benefit. PET was more accurate than serum CEA level in assessing re- sponse. These investigators found that FDG uptake imme- diately following radiation may be caused by inflammatory changes without residual tumor (77). They recommended a postradiation interval of 6 months before evaluating response. These findings mirror the results of an early report of Abe et al. (78). The time course of FDG uptake postradiation has not been studied systematically;
but an interval of 6 months between RT and PET is rec- ommended to assess presence of complete response.
Although the inflammatory response is real and may hamper image interpretation, early PET scanning may have a role to determine whether the tumor is sensitive to treatment. For assessing complete response, sufficient time needs to have elapsed, otherwise remaining FDG uptake may be incorrectly interpreted as residual or re- current tumor. Schiepers et al. (79) investigated the effect of radiation after 10 fractionated doses of 3 Gy on primary rectal cancer (total dose, 30 Gy). In this pilot experiment, all cancers showed an effect of irradiation 2 weeks after finishing RT (Figure 10.6), and one tumor was no longer detectable in the resected, irradiated specimen. More re- search is needed to sort out the exact timing of cell damage, cell death, and cleanup of necrotic cells and its effect on metabolic activity of the cells mediating in the inflammatory response cascade. As a rule of thumb, we recommend assessing tumor response with FDG-PET about 3 months after completion of RT and 1 month after chemotherapy.
The European Organization for Research and Treatment of Cancer (EORTC) published a position paper on the use of FDG-PET for the measurement of tumor re- sponse (80). The EORTC-PET study group formulated standard criteria for reporting alterations in FDG uptake to assess clinical and subclinical response to therapy.
Although there was wide variation in the methodology between PET centers surveyed, assessment of FDG uptake was thought to be a satisfactory method to monitor re- sponse to treatment. The group made initial recommen- dations on (1) patient preparation, (2) timing of PET scans,( 3) methods to measure FDG uptake with SUV, and (4) definition of FDG tumor response.
Costs
PET using FDG has been shown to be cost effective for di- agnosis and staging recurrent CRC in a study using clini- cal evaluation of effectiveness with modeling of costs (51) and a study using decision analysis (81). In both studies, all costs calculations were based on U.S. Medicare reim- bursement rates for conventional diagnostic and thera- peutic procedures and $1,800 cost for PET. The cost analysis by Valk et al. (51) assessed the use of PET in pa- tients who had been diagnosed with recurrence and were
considered to have resectable tumor on the basis of con- ventional diagnostic imaging. In a management algorithm where recurrence at more than one site was treated as nonresectable, they found a cost saving of $3,003 per patient resulting from diagnosis of nonresectable tumor by PET.
The group of Gambhir used a decision analysis model to determine the cost-effectiveness of PET in management of recurrent CRC (82). In patients with an elevated serum CEA, who were evaluated for hepatic resection, the CT plus PET strategy was superior to the conventional CT- only strategy. The model showed an increased cost of $429 per patient, which resulted in an increased mean life ex- pectancy of 9.5 days per patient. The assumptions that were used in this liver metastasis model (increased serum CEA with negative conventional imaging) included preva- lence of recurrent CRC, 32%; CT liver sensitivity, 79%; CT liver specificity, 88%; PET-liver sensitivity, 96%; and PET- liver specificity, 99%. Frequency of liver metastasis was assumed to be 28.5%, and 81% of these were estimated to have extrahepatic involvement. This model showed that the CT plus PET strategy was cost effective for PET costs less than $1,200 and hepatic recurrence frequencies greater than 46%.
Based on the available data, all patients should undergo a PET scan with FDG for preoperative staging of recurrent CRC. In case of distant, extrahepatic hypermetabolic foci, a CT of the corresponding region should be performed for anatomic correlation. This approach allows more-accu- rate selection of patients who will benefit from surgery
than conventional diagnostic methods and, more impor- tantly, excludes patients who will not benefit from laparo- tomy and liver resection because of unrecognized distant disease.
Dual-Modality Imaging with PET/CT
PET using FDG has come of age for staging CRC and has emerged as a molecular imaging modality. Several limita- tions of dedicated PET, such as prolonged imaging times, lengthy transmission scans, and lack of anatomic land- marks, prompted for integration of imaging information.
Traditionally, all imaging data was merged “inside the brain” of the imaging specialist. Subsequently, informa- tion from the different modalities was fused in computer memory. Patient position (e.g., arms up or down, tilt of image device relative to patient axis), breathing state (deep inspiration, breath-hold, or shallow breathing), res- olution, etc., differ significantly among the various imaging modalities and imaging centers, and image fusion post hoc is, therefore, prone to errors. This limita- tion gave rise to the development of dual-modality systems such as PET/CT, in which the patient is posi- tioned in a combined gantry and imaged during one single session. After the prototype PET/CT was introduced by the Pittsburgh group, dual-modality imaging has rapidly evolved as the imaging method of choice for on- cology (13, 83). The combination of a PET and CT scanner
Figure 10.6. Transverse FDG-PET images in a 48-year-old man with primary rectum cancer, clinical stage cT3 N0 M0. Note the tumor focus posterior to the bladder (upper row, arrowhead). The bottom row shows the findings after 10 fractionated doses of 3-Gy external radiation (total, 30 Gy). The second PET scan was performed 2 weeks after completion of radiation. Note the marked decrease in FDG uptake after therapy (arrow).
permits “mechanical” fusion of the images, greatly facili- tating the localization of hypermetabolic foci as well as increasing specificity while characterizing lesions (Figures 10.7, 10.8, 10.9). In addition, the CT can be used for atten- uation correction (see Chapter 1), eliminating the trans- mission scans and reducing acquisition time. In the dual imaging scenario, the purpose of the CT is to localize lesions in anatomic terms and of PET to categorize lesions in pathologic terms (benign versus malignant).
Lesions with abnormal metabolism can be assigned to normal or abnormal structures on CT and classified cor- rectly as true or false positives. Thus far, a few well-con- trolled studies have been reported with PET/CT imaging and its clinical impact in diagnostic oncology (84, 85). The
reports show a 10% to 20% increase in both sensitivity and specificity.
Cohade et al. (85) reported their experience with an in-line PET/CT device in colorectal cancer. They focused on two aspects of the identified lesions: (1) type, that is, malignant or benign and (2) anatomic location. In their study, the “conventional” technique of attenuation correction was applied, that is, measured with a positron-emitting transmission source (68Ge). The duration of the transmission scan was 3 min/bed, and a segmentation algorithm was used to calculate the atten- uation map.
The purpose of the CT was to help the physician in in- terpreting the study, localize the lesion in anatomic
Figure 10.7. Body PET/CT scan in a 72- year-old man with recurrent colorectal cancer. Patient was 6 months past ante- rior resection of an adenocarcinoma of the rectum. He returned with abdominal pain and constipation. Proctoscopy did not reveal stenosis. Coronal planes (left column) and sagittal cuts (right column) are shown. The top row displays the CT images (soft tissue window), the middle row the PET images, and the bottom row the fusion images (CT in black and white with PET superimposed in color). Local recurrence was diagnosed (green arrows), and a regional lymph node metastasis was identified (blue arrow).
Note the hot bladder, containing ex- creted FDG in the urine.
a b
Figure 10.8. Axial images of the pelvis for PET (top), CT (bottom), and fusion (middle) of the same patient as in Figure 10.7. (a) Local recurrence near the anastomosis (green arrow) and rectal lymph node with increased FDG uptake (blue arrow). (b) Additional affected rectal lymph node at a lower level than (a).
terms, and categorize the lesion as benign or malignant.
The results of PET/CT were compared to the interpreta- tion of PET alone. Cohade et al. (85) found that the un- certainty in lesion localization decreased by 55% and that in lesion type by 50%. These differences did not translate into a different performance of PET alone versus PET/CT.
There were no statistical differences in sensitivity, specificity, or accuracy. In evaluating the stage of CRC, an 11% increase in accuracy of staging was found. As Cohade et al. (85) pointed out, the gain by PET/CT is not
“tremendously high” for CRC (85). However, one has to keep in mind that sensitivity and specificity of FDG-PET for staging of CRC is already quite high, as was estab- lished earlier (see Table 10.1).
Although a significant advantage of PET-alone over CT- alone has been amply documented (51, 86), PET has size limitations and usually misses tumors less than 5 mm.
The Cohade paper demonstrates a benefit from registra- tion and fusion of PET to CT. The certainty that a focus of FDG accumulation constitutes a lesion is clearly in- creased. In a group of 169 patients referred for staging of a mixture of neoplasms, we found a 12% improvement by PET/CT (87).
One of the main areas of contribution for PET/CT is the precise location of the bowel and lymph nodes, and asso- ciation of FDG uptake with GI mucosa, which is quite variable. This consideration is even more important for restaging and for therapy monitoring after surgery when the anatomy has been changed. The effect of breathing on abdominal organs is less important than for the chest (88). Even artifacts in the liver, caused by breathing during the transport through the CT gantry, do not seem to pose clinical problems in staging of CRC. Non-attenua- tion-corrected tomograms and 2D projection images are always available to view images without artifacts and cir- cumvent interpretation problems. To reduce bowel uptake, some centers use pharmacologic interventions to inhibit secretion and motility, but this step does not seem necessary on a routine basis.
Evaluation of the effects of intravenous and oral con- trast used for CT on the PET images has just started (89, 90). PET/CT has already influenced the way we read stan- dard FDG-PET scans. Accurate localization of muscle and brown fat uptake by PET/CT has been demonstrated (91).
These patterns are found in young, tense, skinny, or shiv- ering patients and are physiologic variants, which are now recognized while interpreting PET alone.
Therapy monitoring will become increasingly impor- tant and will have a major impact. In the near future, PET/CT will play an important role in planning of radio- therapy, which will be relevant for all types of cancer, in- cluding CRC (92).
As was recently outlined by Wahl (93), imaging of ab- dominal and pelvic cancer in the future will be almost ex- clusively done by PET/CT. This projection was based on the experience of the Johns Hopkins PET Center, where more than 2,700 studies were performed in a 2-year period.
Conclusion
FDG-PET is indicated as the initial test for diagnosis and staging of recurrent CRC. In addition, it is indicated for preoperative staging of known recurrence that is consid- ered to be surgically resectable.
PET imaging is valuable for distinguishing posttreat- ment changes from recurrent tumor, differentiation of benign from malignant lymph nodes, and monitoring therapy. Addition of FDG-PET to the evaluation of these patients reduces overall treatment cost by accurately dis- tinguishing patients who will benefit from those who will not benefit from surgery. PET is evolving as a molecular imaging modality and will soon enter the realm of clinical gene imaging and gene therapy monitoring (5, 94).
Acknowledgments
The previous version of this chapter was co-authored by Peter Valk, who passed away in 2003. Tables were maintained and the old text incorpo- rated in the current updated version. His advancement of the PET tech- nology in the clinical scenario is well known. This text was prepared to honor his legacy.
a
Figure 10.9. (a) Coronal plane with a liver metastasis (blue arrow) and left lower quad- rant mesenteric metastasis (green arrow), in same patient as shown in Figure 10.7.
b c
Figure 10.9. Contd. (b) Axial images of the liver metastasis shown on the coronal cut (same layout as in Figure 10.8). (c) Left lower quadrant lesion seen on the coronal cut (A) (same layout as in Figure 10.8).
