A fundamental property of malignant tumors is increased glucose metabolism, which can be estimated by imaging the glucose analog fluorodeoxyglucose (FDG). The aim of this study was to determine whether FDG uptake in lung carcinoma, as measured on positron emission tomography (PET) imaging in patients with newly diagnosed bronchogenic carcinoma, has prognostic significance.
A retrospective review of all patients with a new diagnosis of nonsmall cell lung carcinoma and FDG-PET imaging was performed. Stage at presentation, cell type, tumor size, and survival data were recorded. For each patient, a standardized uptake ratio (SUR) was calculated for the primary lesion on PET and was correlated with clinical information to determine prognostic significance.
One hundred fifty-five patients, with a mean age of 56 years, were enrolled. One hundred eighteen patients (76%) had lesions with SUR values <10 and had a median survival of 24.6 months (95% confidence interval [CI]: 20.9-41.1). Thirty-seven patients (24%) had tumor lesions with SUR values >10 and had a median survival of 11.4 months (95% CI: 9.3-19.4). An SUR >10 correlated with poorer survival (P = 0.0049). Patients with primary lesions >3 cm and an SUR >10 had the worst prognosis, with a median survival of 5.7 months (95% CI: 4.0-13.1). Multivariate analysis demonstrated than an SUR >10 provided prognostic information independent of the clinical stage and lesion size.
An estimated 180,000 new cases of bronchogenic carcinoma will be diagnosed in the United States this year. Most patients present with advanced disease, as evidenced by an overall 5-year survival rate of 14%.1
In an effort to improve survival, a numerous prognostic factors, including stage at presentation, performance status, weight loss, and molecular markers, are used to stratify patients for optimal treatment regimens.2-11 Other features, such as tumor growth rates, reflected by the tumor doubling time, have also been shown to be closely related to prognosis.12-15 Most of these series, however, require sequential imaging and, realistically, have had limited clinical application. In a more recent study, a correlation between tumor doubling time and fluorine-18 fluorodeoxyglucose (FDG), a d-glucose analog, was demonstrated as measured by positron emission tomography (PET).16 To determine whether the amount of FDG uptake, a measure of glucose metabolism, has prognostic significance, we retrospectively reviewed all patients with a new diagnosis of lung carcinoma and an FDG-PET study and correlated this with survival.
MATERIALS AND METHODS
During a 4-year period (1992-1996), all patients with a new diagnosis of nonsmall cell lung carcinoma (NSCLC) and an FDG-PET study were considered eligible. Indications for PET imaging included a strongly suspected lung carcinoma based on chest radiographs and computed tomograph (CT) or a pathologically proven new diagnosis of lung carcinoma. Not all consecutive patients fulfilling these entrance criteria had a PET study due to scheduling limitations and patient consent.
One hundred fifty-five patients, 57 women and 98 men, with a mean age of 56 years (range 25-85 years) were enrolled. One hundred and forty-nine patients (96%) had a pathologic diagnosis of lung carcinoma made within 8 weeks of their PET scan. Six patients had a delay in diagnosis and, in one case, up to 10 months. Tumor characteristics, including stage at presentation, histologic cell type, and lesion size, were recorded from the patient's medical record. The tumor registry at our institution provided survival data. The median follow-up of the 70 patients who remained alive was 20.9 months (range 0.30-59.9 months).
Stage at Presentation
All patients were assigned a stage at presentation according to the new International Staging System.17 Clinical staging included radiologic studies (thoracic CT-lymph nodes were measured in the short axis and were considered abnormal if they exceed 1 cm in greatest dimension; bone scan; and head CT or magnetic resonance imaging) and physical examination. Pathologic confirmation by transthoracic needle biopsy, bronchoscopy, mediastinoscopy, thoracoscopy, thoracotomy, or a combination of procedures varied depending on the radiologic studies, patient medical condition, and desired treatment options. Patients were then assigned a pathologic stage at presentation.
Treatment options were based on these findings. Patients with Stage I and II were considered for primary surgical resection. Patients with Stage IIIA were potential surgical candidates if they responded to neoadjuvent therapy. Patients with Stage IIIB and IV were given combination chemotherapy and/or radiation therapy depending on a number of factors, including extent of disease and medical condition.
All patients fasted for 4 hours prior their PET scan. PET imaging was performed on either a GE 4096 Plus or a GE Advance tomograph (GE Medical Systems, Milwaukee, WI). The GE 4096 produced 15 axial images with 6.5-mm-thick image planes. Full width at half-maximum is 5 mm, and the longitudinal field of view is 10.3 cm. The Advance system produced 35 axial images with 4.25-mm-thick image planes. The axial field of view is 15 cm, and the full width at half maximum is 5 mm. Image processing and reconstruction were performed on a VAX 4000-300 computer system and a VAX 3100 work station (Digital Equipment, Marlboro, MA) or on an HP735 (Hewlett Packard, Palo Alto, CA). Transmission scans using a rotating 68Ge pin source were performed prior to thoracic emission images through the nodule (one bed position) or entire chest (three bed positions, GE4096; two bed positions, Advance) by using 10-minute emission and 15-minute transmission acquisition at each bed position at least 30 minutes after intravenous injection of 10.0 mCi of 18FDG. FDG was synthesized for imaging by using standard methods.18 Cultures were obtained to assure sterility, and testing for pyrogens was performed to exclude endotoxins. Radiochemical purity was tested following each run by using standardized high-pressure liquid chromatography (HPLC).
A circular region of interest (ROI) was placed on the nodule on the emission image with maximum FDG uptake. ROI diameter was adjusted to 80% of peak counts in the nodule. The standardized uptake ratio (SUR) was calculated as follows:
Cox's proportional hazards model was used to assess the joint effects of the following variables on survival: SUR, clinical stage at presentation, histology, lesion size, and pairwise interactions of these covariables with SUR. Survival time was defined as the time between the PET study and death or the last follow-up date. Analysis of SUR initially examined survival within SUR subgroups that were defined by the quartiles of SUR distribution. A log rank test was used to determine a statistically significant SUR cut-off value of 10, which was used for survival analysis.
Kaplan-Meier product limit estimator was used to estimate survival probabilities for the patient subgroups defined by the covariables. Due to the small number of patients with Stage II NSCLC, they were combined with the Stage I patients for analysis.
Survival statistics for all patients categorized by SUR, lesion size, stage, and histology are shown in Table 1, and multivariate survival models are shown in Table 2.
Table 1. Survival Statistics Associated with Analyses of Patients with Nonsmall Cell Lung Carcinoma
Model 2: Lesion size, stage, and categorized SUR as predictor
SUR > 10
Lesion > 3 cm
Stage III vs. Stage I/II
Stage IV vs. Stage I/II
Model 3: Lesion size, stage, categorized SUR, and interaction between SUR and lesion size as predictors
SUR > 10
Lesion > 3 cm
Stage III vs. Stage I/II
Stage IV vs. Stage I/II
Lesion > 3 cm and SUR > 10
One hundred eighteen patients had lesions with SUR values ≤10 and had a median survival of 24.6 months (95% confidence interval [CI]: 20.9, 41.1). Thirty-seven patients had tumor lesions with SUR values >10 and had a median survival of 11.4 months (95% CI: 9.3, 19.4). Univariate analysis showed that SUR > 10 was correlated to poorer survival (P = 0.0049; Fig. 1 (22K)).
Stage at Presentation
Sixty-nine patients with NSCLC had Stage I-II disease, 55 patients had Stage III disease, and 31 patients had Stage IV disease. Median survival for each group was 41.1, 13.1, and 3.4 months, respectively. In univariate analysis, stage at presentation showed a statistically significant relationship with poorer survival (P < 0.0001).
Ninety-six patients had primary tumor lesion sizes ≤3 cm. Median survival in this group was 31.0 months (95% CI: 22.1, ∞). Fifty-nine patients had lesion sizes >3 cm, with a median survival of 12.7 months (95% CI: 10.2, 19.3). In univariate analysis, the effect of lesion size on prognosis was statistically significant (P = 0.0029), because patients with larger tumors did worse.
There were 57 patients (34%) with squamous cell carcinoma, 53 patients (31%) with adenocarcinoma, 11 patients (6%) with large cell carcinoma, and 35 patients (21%) labeled with NSCLC. There was no statistically significant relationship between survival and histology (P = 0.598).
The joint effect of SUR, stage, and lesion size on survival was examined with the expressed purpose of determining whether SUR had prognostic importance beyond that provided by lesion size and stage (Table 2). Model 1 examines the joint effect of lesion size and stage. After adjustment for lesion size, stage at presentation shows a statistically significant correlation with poorer survival (P = 0.0065 for Stage III versus Stage I-II; P < 0.0001 for Stage IV versus Stage I-II). The effect of lesion size on prognosis approached statistical significance (P = 0.069), because patients with larger tumors did worse.
Model 2 examines the joint effect of lesion size and stage as well as SUR. The test, which was conducted to determine whether SUR had prognostic importance beyond that provided by lesion size and stage, showed that patients with SUR > 10 had a significantly poorer prognosis (P = 0.026; Fig. 2 (36K)).
To further understand the effect of SUR on survival, an interaction between SUR and lesion size was added to the model (Table 2, Model 3). The statistically significant interaction (P = 0.041) showed that the effect of SUR on survival was not consistent across lesion sizes. Patients with large lesions and SUR >10 had significantly worse prognosis, with a median survival of 5.7 months (95% CI: 4.0, 13.1; Fig. 3 (29K)).
Bronchogenic carcinoma continues to represent a major health problem in the United States and is the leading cause of cancer death for both men and women. Screening trials have failed to demonstrate clearly a survival advantage, although recent reexamination of these studies suggests that there may be some benefit in high risk patients.19 A number of clinical, radiologic, and pathologic features are used typically in determining treatment options and prognosis, although stage at presentation appears to be the most influential predictor of outcome. Tumor markers, including blood group antigens, DNA content, ras, p53, and c-erB-2, have become interesting targets for analysis, with the hope of understanding some of the molecular changes in tumor cells, and in determining which markers carry prognostic significance.2, 2-10, 20-28 Surface antigens or receptor overexpression may also help stratify patients into the most appropriate treatment protocols and, thus, improve survival rates.
Tumor growth rates and doubling times, as documented by sequential imaging studies, have also been shown to be related prognosis, although following lesions with repeat films is of limited practical value.12-15, 29-31 More recently, the amount of FDG-PET uptake in primary lung lesions has been shown to have a direct relation to tumor growth,16 and immunohistochemical analysis of cell membrane glucose transporter (Glut1 and Glut3) protein overexpression was correlated with a poor outcome.32 These data suggest that, presumably, tumors with increased glucose uptake are more metabolically active and more biologically aggressive.
Although the utility of FDG-PET imaging for diagnosing and staging lung carcinoma is well established,33-39 its potential as a prognostic marker has yet to be determined. FDG-PET takes advantage of one of the basic properties of tumor cells, increased glucose metabolism, and, theoretically, uses this feature in differentiating benign from malignant abnormalities. Although the sensitivity for detecting malignant lung lesions is high, the specificity is less than optimal. A spectrum of benign abnormalities has shown increased FDG uptake.33-36 It has become clear that FDG not only accumulates in tumor cells, but there is a component from the surrounding inflammatory immune response. A recent study demonstrated FDG uptake in a breast carcinoma animal tumor model and indicated that cancer cells were the main site of uptake, with only 20% uptake in nonneoplastic components.40 Other studies, however, have shown that FDG uptake is in predominately nonneoplastic cells. Kubota et al.41 indicated in their studies that the highest concentration of FDG uptake in FMA3A tumors in mice was found in newly formed granulation tissue around the tumor and in macrophages. In addition, Yamada et al.42 showed FDG accumulation in experimental turpentine-induced inflammatory tissue in rats, with highest concentration in areas with fibroblasts, endothelial cells from blood vessels, macrophages, and neutrophils. Although the specific cells of FDG accumulation within tumors remains a complex issue, our initial experience suggested that the more metabolically active the tumor, the worse the outcome.
This study addressed the role of FDG-PET in providing prognostic information and its potential to serve as a guide for therapeutic options. Univariate analyses showed that patients with tumors having an SUR >10 had a significant decrease in survival by approximately 13 months compared with those with an SUR < 10. Multivariate analyses revealed that an SUR > 10 provided prognostic information in addition to the clinical stage and lesion size. Thus, patients with tumors that are more active metabolically, as demonstrated by FDG-PET studies, should be considered to be at high risk for relapse regardless of clinical stage at presentation.
In addition (and more statistically significant), multivariate analysis demonstrated that the combination of increased SUR and large lesion size identified a subgroup of patients with the worst prognosis and a median survival of less than 6 months. These data may prove useful in several clinical scenarios. Patients with early Stage I or II disease and a large hypermetabolic lesion may benefit from chemotherapy and/or radiotherapy following surgery. There are currently no clear data that support the possibility that additional treatment in this setting improves survival. If a high risk group of these patients who realistically have microscopic metastasis at presentation could be identified, then improved outcomes may be possible. Those patients with more advanced Stage III or IV disease and a large hypermetabolic lesion may require a different treatment regimen than those with smaller, less metabolically active lesions. At this time, however, it is not clear whether a more aggressive treatment will improve survival, because the biology of these tumors may preclude successful tumor reduction or irradication with current therapeutic options.
Although, in some series, cell type and degree of differentiation may have prognostic value, there was no correlation with FDG uptake. This confirms a basic property of tumors, increased glucose metabolism, but it does not suggest that specific cell types or anaplastic tumors have more glucose utilization as demonstrated by FDG-PET.
In conclusion, these data indicate that FDG uptake in the primary lesion on PET studies in patients with lung carcinoma can provide prognostic information. We suggest that this may be important information when determining treatment options, particularly in patients with larger lesions, and further studies assessing patient survival following various therapeutic protocols are recommended.