Presented in part at the 2008 American Society of Clinical Oncology Gastrointestinal Cancers Symposium; January 25-27, 2008; Orlando, Florida.
The aim of this study was to examine prospectively the utility of adding preoperative [18F]fluorodeoxyglucose positron emission tomography (FDG-PET)/computed tomography (CT) to routine CT, endoscopic ultrasound (EUS), and laparoscopic staging of localized gastric cancer.
Patients with locally advanced gastric/gastroesophageal cancer were screened for 2 institutional review board–approved Memorial Sloan-Kettering Cancer Center neoadjuvant chemotherapy protocols. Locally advanced disease was defined as T3 or T4, or lymph node–positive, based on EUS and high-resolution CT scan. All patients underwent both standard FDG-PET/CT and laparoscopy with cytological examination of washings. The sensitivity and specificity of FDG-PET/CT for the identification of metastatic disease not seen on CT was determined. An economic model using Medicare/Medicaid reimbursement charges was developed to assess the cost-effectiveness of these interventions.
A total of 113 patients were enrolled from 2003 to 2010. All patients were assessed as having locally advanced disease by CT/EUS. FDG uptake in the primary tumor was associated with male sex, proximal tumors, and nondiffuse Lauren's subtype. 31 (27%) patients had occult metastatic disease detected by PET/CT (n = 11, 10%) and/or laparoscopy (n = 21, 19%), with a single overlap. Economic modeling suggests that the addition of FDG-PET/CT to the standard staging evaluation of patients with locally advanced gastric cancer resulted in an estimated cost savings of ∼US $13,000 per patient.
Surgical resection is the only curative approach for patients with localized gastric carcinoma. However, despite modern staging evaluation, 65% to 75% of patients will still recur from occult micrometastatic disease following curative intent resection. Although the addition of perioperative chemotherapy or postoperative chemoradiation may result in an absolute gain in overall survival of 10% to 15%, most patients will still die of recurrent metastatic cancer.1, 2 Surgical morbidity from gastrectomy is significant and the mortality rate for surgical resection approaches 1%.3 Therefore, careful patient selection prior to surgical resection is essential in order to avoid the morbidity and mortality risk of unnecessary surgery in patients who will not benefit from this intervention. Current staging guidelines include the use of endoscopic biopsy and contrast-enhanced computed tomography (CT) of the chest, abdomen, and pelvis. Fluorodeoxyglucose (2-deoxy-2-[18F]fluoro-D-glucose) positron emission tomography (FDG-PET) is preferred but not mandated in the absence of M1 disease, whereas staging laparoscopy is recommended for patients who are eligible for local-regional therapy.4
The role of FDG-PET as an imaging modality in upper gastrointestinal malignancies has evolved substantially over the last decade. In esophageal cancer, FDG-PET has demonstrated both prognostic value and provides an early predictor of response to chemotherapy.5-8 FDG-PET may also upstage 13% to 15% of esophageal cancer patients by detecting occult metastatic disease that is not identified by conventional imaging,9 thus avoiding futile esophagectomy for a significant number of patients if suspicious metastatic disease identified on PET/CT is confirmed.10 Unlike esophageal carcinoma, in which virtually all locally advanced tumors are identified on FDG-PET functional imaging, primary gastric tumors may be less well-visualized by FDG-PET. Variable and occasionally intense physiological uptake can occur within the gastric wall, obscuring the primary tumor, and increased FDG uptake may also correspond with acute inflammation such as gastritis.11-13 In addition, distinct gastric cancer histological subtypes differ significantly in FDG avidity due to differences in glucose transport protein expression.14
FDG-PET and FDG-PET/CT have each been evaluated alone, and in comparison to conventional CT in gastric cancer staging. These initial studies both demonstrate decreased sensitivity but increased specificity when compared with CT for detection of regional and distant lymph node metastases, and poor resolution of peritoneal disease.15-17 There has been no prospective evaluation of the utility of FDG-PET/CT in staging gastric cancer within the context of good-quality contrast-enhanced CT, endoscopic ultrasound (EUS), and laparoscopy. Specifically, in this study, we sought to prospectively examine the potential added benefit of FDG-PET/CT to modern gastric cancer staging paradigms.
MATERIALS AND METHODS
Medically fit patients (Karnofsky performance status: 60%-100%) with locally advanced gastric cancer deemed to be candidates for surgical resection were eligible. Patients were prospectively screened for inclusion onto 1 of 2 sequential preoperative chemotherapy clinical trials at Memorial Sloan-Kettering Cancer Center (MSKCC), New York, NY. Both studies, MSKCC IRB 03-032 (NCT00062374) and MSKCC IRB 08-081 (NCT00737438), involved formal prospective pretreatment staging evaluation, including PET/CT and laparoscopy. All patients signed informed consent. Following completion of staging, patients without evidence of metastatic cancer underwent preoperative chemotherapy followed by surgical resection, and those with confirmed metastatic disease were referred for palliative chemotherapy. Both studies were approved by the MSKCC Institutional Review Board (IRB) prior to initiation.
Initial staging evaluation included contrast-enhanced CT of the chest, abdomen, and pelvis, and EUS. Patients with locally advanced but nonmetastatic disease on initial staging evaluation underwent laparoscopic staging and PET/CT. The order in which PET/CT and laparoscopy were performed was not prespecified. CT criteria for locally advanced gastric cancer include thickening of the wall of the stomach or gastroesophageal junction,18 or pathologically enlarged lymph nodes. By EUS criteria, locally advanced stage is defined as uT3-T4 or N positive. EUS T3 tumors are those which disrupt the serosa by penetrating the fifth endoscopic layer, or which have an irregular outer margin, whereas uT4 tumors invade adjacent structures or organs.19 Malignant nodes are defined as hypoechoic, well-demarcated, and round.20 Patients were staged according to the American Joint Committee on Cancer/International Union Against Cancer guidelines, sixth edition.21
After induction with general anesthesia, pneumoperitoneum was established via a Varess or Hassan approach using CO2 to a pressure of 15 mm Hg. A 5- or 10-mm port was placed in the midline at the umbilicus, with additional 5-mm ports in the right and left upper quadrants as required. The surfaces of the liver, stomach, and peritoneum, and ovaries (for female patients) were inspected, and enlarged lymph nodes were noted. Lesions suspicious for metastasis were biopsied. All patients underwent peritoneal lavage with 500 mL of normal saline in the right upper quadrant, left upper quadrant, and pelvis, with cytological examination of washings for malignant cells. Patients were discharged home the same day.
FDG-PET/CT was performed on Biograph (Siemens Healthcare) or Discovery LS (GE Medical Systems) machines. Prior to PET/CT scans, patients fasted for at least 6 hours, but liberal water intake was encouraged. Preprocedure mean blood glucose was 85 ± 15 mg/dL (range: 57-132 mg/dL). FDG was then injected intravenously. Imaging started after an approximately 60-minute (mean: 65 ± 10 minutes) uptake period, during which patients rested quietly in a chair, and also drank 450 mL of oral contrast (2% barium sulfate) to demarcate the bowel on CT. Following scout view, low dose (120-140 kV, 80 mA) CT and PET images were obtained from the skull base to the upper thigh. All PET, CT, and PET/CT fusion images were displayed on a dedicated workstation and were prospectively reviewed by the responsible study nuclear medicine physician in transaxial, coronal, and sagittal planes. Individual lesions were graded according to the following scale: 0 = normal, 1 = probably benign, 2 = equivocal, 3 = probably malignant, 4 = definitely malignant. Lesions with a certainty of 3 to 4 were considered FDG-avid. All sites of M1 disease were confirmed, either pathologically by outpatient radiology-guided fine-needle aspirate or core biopsy, or radiographically with additional imaging (magnetic resonance imaging or radionucleotide bone scan). Volumetric regions of interest were placed over FDG-avid lesions, and the standardized uptake value, normalized to body surface area, was calculated.
Estimated costs of FDG-PET/CT, biopsy, laparoscopy, and gastrectomy were obtained from the MSKCC finance department, and were based on mean Medicare/Medicaid billing charges for these procedures at MSKCC. Billing charges included both hospital and physician charges. The cost of the addition of PET/CT and laparoscopy to preoperative evaluation was calculated on a per-patient basis. Costs of all procedures were averaged over department physicians/surgeons and rounded to the nearest US $100. Two algorithms were examined, one in which PET/CT was performed as the initial investigation and another in which laparoscopy was performed prior to PET/CT.
Sensitivity and specificity were calculated for the different staging tools and exact 95% confidence intervals (CIs) for binomial proportions were provided. Associations were assessed using Fisher's exact test for categorical variables and using Cochran-Armitage test for trend for ordinal variables. McNemar's exact test was employed for binary paired variables. P values <.05 were considered statistically significant.
Patient and Tumor Characteristics
From June 2003 to August 2010, 113 patients were identified as having locally advanced gastric or gastroesophageal cancer by CT and/or EUS criteria, and proceeded with additional preoperative staging with FDG-PET/CT and laparoscopy. The majority of patients (60%) were men, and median age was 61 years (range: 25-83 years) (Table 1). Tumors of the proximal stomach and/or gastroesophageal junction were less common than those of the gastric body or antrum (37% vs 63%). A total of 45% of tumors were nondiffuse (pure intestinal or mixed) by Lauren's criteria, 46% were diffuse, and 9% were of no reported Lauren's subtype.
Table 1. Patient Characteristics
Characteristic (n = 113)
1 patient with T1N1 tumor on endoscopic ultrasound.
FDG uptake in the primary tumor was more frequently associated with male sex and with proximal/gastroesophageal junction tumors. Diffuse tumors were FDG-avid in only 44% of patients, compared with 97% of pure intestinal subtype tumors being FDG-avid (P < .0001). A total of 81% of proximal tumors were FDG-avid, compared to 59% of distal tumors (P = .022). Tumors which were moderately or moderate-to-poorly differentiated were more likely to demonstrate FDG uptake than poorly differentiated tumors (P = .002). There were no well-differentiated tumors (Table 2).
Table 2. Fluorodeoxyglucose Avidity of Primary and Tumor Characteristics
70% poorly differentiated tumors were Lauren's diffuse subtype.
Thirty-one patients (27%) had confirmed metastatic disease prior to surgery: 11 (10%) were identified by PET/CT and 21 (19%) were identified by laparoscopy (one patient was confirmed by both modalities). The sensitivity of PET/CT for the detection of metastatic disease in the entire patient population was 35% (95% CI = 19%-55%), and specificity was 99% (95% CI = 93%-100%) (Table 3). Among patients with FDG-avid primary tumors (n = 76), the sensitivity was 50% (95% CI = 28%-72%) and the specificity was 98% (95% CI = 90%-100%) (Table 4). Sites of occult metastatic disease detected on PET/CT included bone (n = 2), liver (n = 4), and distant lymphadenopathy (n = 6). No peritoneal metastases were identified by FDG-PET/CT. The sensitivity of PET/CT for the detection of metastatic disease in laparoscopy-negative patients was 100% (95% CI = 73%-100%) with a specificity of 99% (95% CI = 91%-99%) (Table 5). There was one false-positive on PET/CT (a peritoneal nodule found at laparoscopy to represent an exophytic primary tumor).
Table 3. PET/CT Sensitivity and Specificity, All Patients (n = 113)
Table 4. PET/CT Sensitivity and Specificity Among the Subset of Patients With Gastric Cancer Who Have Fluorodeoxyglucose-Avid Primary Tumors (n = 76)
Metastatic Cancer Confirmed
Positive Predictive Value
CI indicates confidence interval; PET/CT, positron emission tomography/computed tomography.
Sensitivity for M1 50% (95% CI: 28%-72%) Specificity for M1 98% (95% CI: 90%-100%)
91.7% (95% CI: 62%-99.8%)
Table 5. PET/CT Sensitivity and Specificity Among the Subset of Patients With Gastric Cancer Who Have a Negative Laparoscopy
Metastatic Cancer Confirmed
Positive Predictive Value
CI indicates confidence interval; PET/CT, positron emission tomography/computed tomography.
Sensitivity for M1 100% (95% CI: 73%-100%) Specificity for M1 99% (95% CI: 96%-99%)
90.9% (95% CI: 66%-90.9%)
Twenty (17%) patients had occult metastatic disease detected with laparoscopy only; 5 of these were detected using peritoneal cytology only. One patient had metastatic disease identified by both laparoscopy (peritoneal disease) and by PET/CT (liver metastasis). If laparoscopy is performed in all patients (n = 113) without prior PET/CT, 21 of 113 (19%) of patients will have a positive laparoscopy. If laparoscopy is performed only in patients who did not demonstrate metastases on PET/CT, 20 of 101 (20%) of these remaining patients will still have metastatic disease identified by laparoscopy, and 11 of 113 (10%) of these patients will avoid a futile laparoscopy.
Metastatic disease in patients with Lauren's diffuse gastric cancer was more likely to be detected by laparoscopy (87.5%) than by PET/CT (12.5%) (exact McNemar's P value = .004).
We estimated the economic impact of including FDG-PET/CT in the preoperative staging algorithm for locally advanced gastric cancer. In this model, patients without confirmed metastatic disease proceeded to resection of their gastric primary tumor. In a model where only laparoscopy is performed preoperatively for patients with locally advanced gastric cancer, using our study population, 10 futile gastrectomies would have been performed; that is, of the 12 PET/CT scans that identified possible metastatic gastric cancer, 10 PET/CT scans identified occult metastatic disease not identified by laparoscopy, 1 PET/CT identified metastatic disease that was also identified at laparoscopy, and there was 1 false-positive PET/CT scan. In our group of 113 patients, this would lead to total charges of $16,073,000 and a unit charge per patient of $142,239 (Fig. 1, Algorithm C). In this model, the addition of PET/CT (plus biopsy if necessary) to preoperative staging results in significant savings for our population. When PET/CTs are performed in all patients who do not have metastatic disease by laparoscopy, 92 patients would have a PET/CT and 82 would proceed to gastrectomy (Fig. 1, Algorithm B), yielding a cost savings of $1,291,000 for the total population, and a savings of $11,425 per patient. Conversely, if a laparoscopy is performed only in patients who are M0 as determined by PET/CT, the total cost savings is $1,533,500, or $13,571 per patient (Fig. 1, Algorithm A). Importantly, the savings in this preliminary economic evaluation is primarily driven by avoiding 10 futile gastrectomies in patients who have occult metastatic gastric cancer.
The benefits of staging laparoscopy are well described; this procedure detects occult peritoneal metastatic disease in approximately one-third of patients with locally advanced gastric cancer.22-24 Patient selection for both perioperative chemotherapy and surgery with curative intent are enhanced, and patients with metastatic gastric cancer avoid the morbidity of an unnecessary laparotomy. In this study, we asked whether there is an additional benefit of adding FDG-PET/CT to the current standard diagnostic evaluation of patients with locally advanced gastric cancer who undergo preoperative laparoscopy.
Our study is the largest prospective evaluation of PET/CT in this patient population, but still must be considered preliminary because only 12 patients (10%) had metastatic disease identified by PET/CT in this study cohort. On the other hand, 19% had peritoneal disease detected by laparoscopy, for a total of 27% of patients with metastatic disease not detected by conventional CT. We found minimal overlap between the groups of patients with metastases identified by FDG-PET/CT or laparoscopy, suggesting that FDG-PET/CT and laparoscopy are almost mutually exclusive and that both are required in order to fully stage patients with locally advanced gastric cancer. Without laparoscopy, 19% of patients would undergo futile laparotomy; if FDG-PET/CT is omitted, 10% of patients would undergo unnecessary surgery. However, given the risk of false-positive PET/CT results, we do not advocate its use as a stand-alone study to rule out patients from receiving potentially curative surgery. Histological confirmation of metastases should be obtained if feasible. This study had too few events (ie, positive PET/CT scans) to precisely define the false-positive rate of PET/CT. Because of the nonspecificity of glucose uptake, a finite false-positive rate is expected, thus further supporting confirmation of M1 disease prior to a decision not to operate.
The sensitivity of FDG-PET/CT for the detection of metastatic peritoneal disease in our study was remarkably low; FDG-PET/CT failed to detect a single case of peritoneal metastases. Wang et al recently reported PET/CT to have a pooled sensitivity of only 28% (95% CI = 17%-44%) for detection of peritoneal metastases.25 Patients entering our study had already been screened with a high-quality contrast-enhanced CT, and those with overt peritoneal metastases had been excluded. The sensitivity of CT to detect peritoneal spread may be high as 77% when thin slices are used.17 Our data suggests therefore that FDG-PET/CT does not add to high-quality contrast CT for identifying gastric cancer peritoneal metastases. Accordingly, patients in our study had generally low volumes of peritoneal metastases found only at laparoscopy; 5 patients had positive peritoneal cytology only. This may also explain why our figure of 19% of patients upstaged by laparoscopy is lower than that in the older literature. Second, tumor subtype biological characteristics may affect the ability of FDG-PET to detect peritoneal metastases. In our study, 46% of tumors were of the diffuse type; of these, only 44% had an FDG-avid primary. Lauren's diffuse gastric cancer is more likely to spread to the peritoneal cavity,26 and less likely to be detected by FDG-PET when the metastases do.27 In our study, only 12% of patients with metastatic diffuse type gastric cancer had metastases detected by FDG-PET/CT, whereas 88% of metastatic diffuse gastric cancers were detected at laparoscopy. Although the value of FDG-PET/CT appears to be greater in gastric cancer with intestinal histology, given the difficulty of confirming Lauren's classification on small endoscopic biopsy samples, and that nearly half of diffuse gastric cancers are FDG-avid, it may not be prudent to limit the use of FDG-PET/CT to intestinal tumors alone.
Concerns regarding exposure to excessive ionizing radiation from radiological imaging are increasing; up to 2% of cancers in the United States may be due to CT-related radiation exposure.28 Could a single definitive investigation be performed instead of diagnostic CT followed by PET/CT? A diagnostic CT (with or without intravenous contrast) in conjunction with a PET scan may be performed, thereby eliminating the radiation dose of a CT scan performed routinely in gastric cancer staging. Although some gastric cancers are not FDG-avid, diagnostic information could still be derived from the diagnostic CT component of the PET/CT. However, such practice would require an a priori decision that all patients with gastric cancer undergo PET/CT imaging for staging. The cost-effectiveness of this staging algorithm was beyond the scope of this analysis.
In our exploratory economic model, we found that the addition of PET/CT to CT, EUS, and laparoscopy allowed us to avoid futile gastrectomy in almost 10% of patients and could lead to potential savings of more than $10,000 per patient using standard Medicare/Medicaid billing charges. Although the surgical morbidity of laparoscopy is low, there is a small risk of death associated with general anesthesia which could be avoided by performing PET/CT first. If PET/CT was performed prior to laparoscopy, this could lead to an additional savings of $2000 per patient by reducing the need for laparoscopy in those with biopsy-proven metastatic disease. Finally, with respect to our illustrative example, although suggested by National Comprehensive Cancer Network guidelines, we appreciate that preoperative laparoscopy is not a universal standard of care for locally advanced gastric cancer. If both laparoscopy and PET/CT were adopted as part of routine preoperative staging where this is not currently the case, then these potential savings would be multiplied. Our cost analysis illustrates the important point that, by virtue of avoiding a futile gastrectomy, under several imaging and surgery staging algorithms, the addition of a PET/CT is associated with a significant per-patient savings. However, we acknowledge that the figures we provide are meant to be illustrative, and that formal prospective analyses including measures of cost–benefit and patient outcome such as quality-adjusted life year are necessary to quantify the full economic and social benefit of this technology. In addition, although rarely indicated, palliative gastrectomy may still be required in a small proportion of patients with metastatic gastric cancer due to intractable bleeding or obstruction.29
Even from this initial analysis, can we propose a diagnostic algorithm for locally advanced gastric cancer in order to minimize the morbidity of unwarranted surgical resection? We suggest that following CT and/or EUS, patients with locally advanced gastric cancer first undergo PET/CT, where up to 10% patients will have metastatic lesions detected (Fig. 1, Algorithm A). Based on our findings, for each 10 PET/CTs performed, one patient may be spared laparoscopy. As more FDG-PET/CT scans are performed routinely, we may be able to gather additional data to provide greater confidence in these numbers. We note that patients who do not have M1 disease as determined by PET/CT still require a laparoscopy, because up to 20% of the remaining patients will still have peritoneal disease detected. This is particularly pertinent for patients with diffuse gastric cancers which often fail to demonstrate FDG-avidity on PET. Alternatively, patients may undergo laparoscopy first. Following this, between 5% (diffuse) and 15% (nondiffuse) of patients will have occult metastatic disease detected by FDG-PET/CT. However, this will lead to 10% unnecessary laparoscopies and additional cost per patient. For this reason, we propose PET/CT as the initial staging investigation in patients who have a routine CT scan that does not demonstrate metastatic disease, prior to laparoscopy.
We acknowledge that FDG-PET/CT does not identify patients who have truly occult micrometastatic disease who will undergo surgery and recur at a later date, which is a limitation of currently available imaging technology. In the future, novel developments in detection of small-volume metastatic disease using intraoperative imaging with molecular contrast agents, reverse transcriptase polymerase chain reaction on peritoneal washings, biomarkers, or circulating tumor cells may improve the identification of occult micrometastatic disease.30-33 At this time, however, we believe that sparing 27% of patients from futile surgery by using currently available technology remains a laudable goal. In this study, we demonstrate that the use of FDG-PET/CT appears to add significantly to our ability to correctly stage locally advanced gastric cancer prior to surgery, and recommend its use in staging all potentially operable gastric cancer patients.
No specific funding was disclosed.
CONFLICT OF INTEREST DISCLOSURE
Dr. Shah has received an American Society of Clinical Oncology Career Development Award. The other authors made no disclosure.