Prognostic Value of Metabolic Tumor Burden from 18 F-FDG PET in Surgical Patients with Nonesmall-cell Lung Cancer

Transcription

1 Prognostic Value of Metabolic Tumor Burden from 18 F-FDG PET in Surgical Patients with Nonesmall-cell Lung Cancer Hao Zhang, MD, PhD, Kristen Wroblewski, MS, Shengri Liao, MD, PhD, Rony Kampalath, MD, Bill C. Penney, PhD, Yi Zhang, MD, Yonglin Pu, MD, PhD Objective: To assess the prognostic value of metabolic tumor burden as measured with metabolic tumor volume (MTV) and total lesion glycolysis (TLG) on 2-deoxy-2-( 18 F)fluoro-D-glucose ( 18 F-FDG) positron emission tomography (PET)/computed tomography (CT), independent of current Union Internacional Contra la Cancrum/American Joint Committee on Cancer tumor, node, and metastasis (TNM) stage; in comparison with that of standardized uptake value (SUV) in surgical patients with nonesmall-cell lung cancer (NSCLC). Material and Methods: This study retrospectively reviewed 104 consecutive surgical patients (47 males, 57 females, median age at PET/CT scan of years) with diagnosed stage I to IV NSCLC who had baseline 18 F-FDG PET/CT scans. The 18 F-FDG PET/CT scans were performed in accordance with National Cancer Institute guidelines. The MTV of tumors in the whole body (MTV WB ), TLG of tumors in the whole body (TLG WB ), the maximum standardized uptake value of tumors in the whole body (SUV maxwb ) as well as the mean standardized uptake value of tumor in the whole body (SUV meanwb ) were measured. The median follow-up among 67 survivors was months from the PET/CT (range 2.82e80.95 months). Statistical methods included Kaplan-Meier curves, Cox regression, and C-statistics. The interobserver variability of SUV maxwb, SUV meanwb, MTV WB, and TLG WB between two observers was analyzed using concordance correlation coefficients (CCCs). Results: The interobserver variability of SUV maxwb, SUV meanwb, MTV WB and TLG WB was very low with CCCs greater than There was a statistically significant association of stage with overall survival (OS). The hazard ratio (HR) of stage III and stage IV as compared with stage I was 3.60 (P =.001) and 4.00 (P =.013), respectively. The MTV WB was significantly associated with OS with a HR for 1-unit increase of ln(mtv WB ) of 1.40/1.32 (P =.004/.039), before/after adjusting for stage and other prognostic factors including chemoradiation therapy, and surgical procedure, respectively. TLG WB had a statistically significant association with OS before and after adjusting for stage and the other prognostic factors. The HR for 1-unit increase in ln(tlg WB ) was 1.26 (P =.011) and 1.25 (P =.031), before and after the adjustment, respectively. Subjects with conditions that led to pneumonectomy (HR = 2.82, P =.035) or segmental resection (HR = 3.44, P =.044) had significantly worse survival than those needing lobectomy. There was no statistically significant association between OS and age, gender, tumor histology, ln(suv maxwb ), and ln(suv meanwb ) (all P >.05). There were 37 deaths during follow-up. Conclusion: Baseline whole-body metabolic tumor burden as measured with MTV WB and TLG WB on FDG PET is a prognostic measure independent of clinical stage and other prognostic factors including chemoradiation therapy and surgical procedure with low interobserver variability and may be used to further risk stratify surgical patients with NSCLC. This study also suggests that MTV and TLG are better prognostic measures than SUV max and SUV mean. These results will need to be validated in larger cohorts in a prospective study. Key Words: 18F-FDG; nonesmall-cell lung cancer; tumor burden; metabolic tumor volume; total lesion glycolysis. ªAUR, 2013 Lung cancer is the most common cause of cancer death in the world (1), the second most common cancer in both men and women, and the number one cause of Acad Radiol 2013; 20:32e40 From the Department of Radiology (H.Z.), First Hospital of Lanzhou University, Lanzhou, Gansu, China; Departments of Radiology (H.Z., S.L., R.K., B.C.P. Y.P.) and Health Studies (K.W.), MC 2026, University of Chicago, Chicago, IL, 60637; Key Laboratory of Carcinogenesis and Translational Research (Ministry of Education), Peking University Cancer Hospital & Institution, Beijing, China (S.L.); Department of Orthopedics, Third Affiliated Hospital of Beijing University of Chinese Medicine, Beijing, China (Y.Z.). Received December 11, 2011; accepted July 16, Address correspondence to: Y.P. ypu@radiology.bsd.uchicago.edu ªAUR, cancer-related deaths in the United States. Nonesmall-cell lung cancer (NSCLC) comprises 80%e85% of all lung cancer cases (2). The treatment and prognosis of NSCLC depend mainly on disease stage as defined by the Union Internacional Contra la Cancrum (UICC)/American Joint Committee on Cancer (AJCC) (3e5). The stage based on the evaluation of the T, N, and M components and the assignment to a stage grouping (I to IV) (3) is the single most prognostic factor in predicting the outcomes of both surgical and nonsurgical patients with NSCLC (3,6e9). A retrospective study from the Netherlands that examined outcomes of 2361 patients who underwent surgery for resectable NSCLC from 1970 to 1992 found that the 5-year survival rate ranges from 19% 32

2 Academic Radiology, Vol 20, No 1, January 2013 WHOLE-BODY METABOLIC TUMOR BURDEN ON PET/CT in stage IIIA to 63% in stage IA (8). In a multicenter North American clinical trial (9), involving 458 patients with unresectable NSCLC cancers, but without metastases or significant weight loss, the patients were randomized to three groups: standard once-daily radiation therapy alone; chemotherapy, followed by standard radiation; and hyperfractionated radiation. Five-year survival rates were only 5%, 8%, and 6% in these three groups, respectively. These two studies highlight the dramatic difference in survival between surgical and nonsurgical patients with NSCLC. Recently several studies demonstrated that metabolic tumor burden as measured with metabolic tumor volume (MTV) or total lesion glycolysis (TLG) is a prognostic marker in nonsurgical patients with NSCLC independent of TNM tumor stage (10e12). However, none of the studies specifically studied its prognostic value in the surgical patients. Lee et al was the first who found that the baseline whole body MTV (MTV WB ) measured semiautomatically is a statistically significant prognostic index and better than SUV max and SUV mean in the prediction of patient outcome in 19 lung cancer patients treated with different modalities (13). In their recent study, they expanded to a cohort of 61 patients with NSCLC treated with different modalities and confirmed the significant association of high MTV WB with decreased overall and progressionfree survival in patients who received definitive therapy (14). In another recent study in 120 patients treated nonsurgically with advanced NSCLC by Yan et al, MTVof the primary tumor was found to be an important independent prognostic factor of survival and better than the SUV max (12) in this regard. More recent studies showed that baseline whole-body metabolic tumor burden as measured with MTV and TLG on positron emission tomography (PET) is a prognostic measure independent of clinical stage in nonsurgical patients (10) and in homogeneous stage IV NSCLC (11). However, based on our extensive literature search, there was no report in the literature about the prognostic value of the MTV and TLG in surgical patients with NSCLC independent of the clinical TNM stage. Because prognosis of surgical patients with NSCLC is much better than nonsurgical patients (8,9), the prognostic value of the metabolic tumor burden in the surgical and nonsurgical patients need to be studied separately. Here we measured the MTV WB and TLG WB semiautomatically with commercially available PET/CT software to determine the additive prognostic value of MTV and TLG independent of tumor stage in 104 surgical patients with stage I to IV NSCLC. The prognostic value of the MTV and TLG independent of the clinical TNM stage of the tumor was compared with that of SUV measurements. MATERIALS AND METHODS Patient Recruitment This study was approved by our hospital s Institutional Review Board and was compliant with the Health Insurance Portability and Accountability Act. We conducted a retrospective review of the medical records of patients with NSCLC. There were a total of 1023 cases with NSCLC who were diagnosed and treated in the authors institution between January 1, 2004, and December 22, We identified the 104 consecutive surgical patients with NSCLC for this study from this retrospective database based on the following inclusion criteria: 1) all patients had a pretherapy baseline PET/CT scan, 2) they had definitive surgery, 3) they had stage I to IV NSCLC, and 4) they had no history or concurrent diagnosis of another type of cancer. The purpose of the PET/CT scan for this group of patients was to stage the disease or for the diagnosis of lung lesions. They had been followed with chest x-rays and CT of the chest and abdomen at our hospital at irregular intervals. These patients had been also followed by our Cancer Registry semiannually. Their survival status was determined through clinical follow-up and the Social Security Death Index. Clinical follow-up and the Illinois State Death Inquiry System were used to determine the causes of death whenever possible. The UICC/AJCC staging system for NSCLC (3e5) was used to stage patients. The pathological stage of each case was assigned according to the sixth edition and clinical stage of each case was assigned according to both the sixth and seventh editions of the staging system. The clinical stage of the disease was based on patients infused CT and PET/CT. The pathological stage of the disease was based on the pathology reports from the definitive surgery in all cases. Mediastinoscopic findings were also used for staging in 23 of the 104 patients who had mediastinoscopy/bronchoscopy before the definitive surgical treatment. The other 81 of the 104 had needle or bronchoscopic biopsy before definitive surgery. Imaging Protocols PET/CT imaging. The baseline pretreatment 2-deoxy-2- ( 18 F)fluoro-D-glucose ( 18 F-FDG) PET/CT scans were performed in accordance with National Cancer Institute guidelines (15) in all 104 patients. The 18 F-FDG PET images were obtained using a PET/CT scanner (Reveal HD, CTI, Knoxville, TN) equipped with high-resolution bismuth germanate detectors and a dual-slice CT scanner. The patients fasted for at least 4 hours before intravenous administration of 370e555 MBq of 18 F-FDG. In addition, the serum glucose levels were tested via finger stick sampling before injection and found to be less than 200 mg/dl. A whole-body unenhanced CT scan with no intravenous contrast administration was performed first for PET attenuation correction. We used a standard protocol for the CT with 130 kvp, 70e80 mas, a transaxial field of view 50 cm in diameter, a tube rotation time of 0.8 seconds per rotation, and a pitch of 3.0. About 90 minutes 30 minutes after injection of the 18 F-FDG, a whole-body static PET scan was acquired for about 30e35 minutes, starting at the thighs and proceeding to the head. PET scans were obtained with an acquisition time of 3e5 minutes per cradle position, with slice overlap at the borders of the field of view to avoid artifacts. The PET camera has a 33

3 ZHANG ET AL Academic Radiology, Vol 20, No 1, January 2013 Figure 1. Axial, sagittal, and coronal images from a positron emission tomography scan of a 67-year-old female with a new diagnosis of nonesmall-cell lung cancer, showing the tumor contours for the measurement of the maximum standardized uptake value, mean standardized uptake value, metabolic tumor volume, and total lesion glycolysis of tumors with the MIMvista PETedge tool. bed position length of 14.6 cm and a transaxial field of view of 66.6 cm in diameter. PET images were reconstructed using the ordered-subsets expectation maximization iterative algorithm with 8 subsets, 2 iterations, and pixels. The slice thickness was 2.4 mm, with 5 mm full-width at half-maximum 3D Gaussian smoothing after reconstruction. We used the 3D imaging mode with Fourier rebinning and model-based scatter correction. Periodic concentration calibrations were conducted using a Ge-68 tub phantom or an 18 F-FDG tub phantom. Measurement of tumor volume on PET/CT. The MTV and TLG, as well as the SUV max and SUV mean, of whole-body tumors were measured with the PETedge tool of the MIMvista software (version 5.1.2, MIMvista Corp, Cleveland, OH) with manual adjustment (Fig 1). The software used the gradient tumor segmentation method (16). In comparison with manual and constant threshold methods in a phantom study, the software was the most accurate and consistent technique for the segmentation of tumor with less interobserver variation and was the most robust for varying imaging conditions (16). This was done by two boardcertified radiologists with PET/CT imaging experience as well as familiarity with the MIMvista software and our PACS system (isite, Philips, Cleveland, OH). With the MIMvista PET edge tool, the radiologists indicated the approximate center of the tumor. The volume of interest was drawn automatically after the radiologists had identified the major and minor axes of the tumor in one plane. The software then automatically measured the SUV max, SUV mean, MTV, and TLG of a tumor. Adjustment of the estimated tumor surface was sometimes needed to include the entire tumor within the margins of the volume of interest. This was done visually by the readers using the two-dimensional or three-dimensional (3D) ball tool in the MIMvista contouring software. The resultant SUV, MTV, and TLG values were noted (outputted to an Excel [Microsoft Corp., Redmond, WA] spreadsheet). The independent PET measurements of the whole-body SUV max (SUV maxwb ), whole-body SUVmean (SUVmeanWB), MTV WB, and TLG WB from two radiologists were used for interobserver variability analysis. One radiologist recorded the start and end time to complete the PET tumor measurement on all cases. The disagreements in the tumor detection and contours were reconciled by both radiologists looking at the cases together. Through discussion, the final consensus readings were decided. Therefore, the final tumor contours were determined based on reader consensus and they were determined using visual analysis. In this study, the SUV max was defined as SUV max = the maximum activity concentration in the tumor/(injected dose/body weight). The SUV mean was defined as the mean concentration of 18 F-FDG in the tumor/(injected dose/body weight). The output results included SUV max, SUV mean, MTV, and TLG of individual tumors. SUV maxwb was the maximum SUV max of all the tumors in the whole body. SUV meanwb was the mean SUV mean of all the tumors in the whole body. It was calculated as SUV meanwb = (TLG WB /MTV WB ). These values calculated from the consensus PET measurements were used in the survival analyses described herein. Statistical Analysis The distributions of MTV WB, TLG WB, SUV meanwb, and SUV maxwb were skewed, so a natural logarithm was applied and the transformed variables were used in all analyses. Interobserver variability in the PET measurements of the SUV maxwb, SUV meanwb, MTV WB, and TLG WB was assessed using concordance correlation coefficients (CCCs) (17,18). Univariate and multivariate analyses using Cox proportional hazards regression (19) were performed for assessment of the relationship between initial PET/CT measurements 34

4 Academic Radiology, Vol 20, No 1, January 2013 WHOLE-BODY METABOLIC TUMOR BURDEN ON PET/CT TABLE 1. Patient Characteristics and Overall Survival Based on Stage, Gender, Age, Tumor Histology, Treatment, and PET/CT Measurements in 104 Surgical Patients with Nonesmall-cell Lung Cancer Survival Rate Mean (SD) 1 year (%) 3 years (%) 5 years (%) Overall n = Stage I, n = II, n = III, n = IV, n = Gender Female, n = Male, n = Age at PET scan (y) (10.08) #67.92 (median), n = >67.92, n = Histology Adeno, n = Squamous, n = Large, n = Others, n = Chemoradiation None, n = Chemo, n = RT, n = N/A Chemo+RT, n = Surgery Lobectomy, n = Pneum, n = Segmental, n = Wedge, n = SUV maxwb 7.57 (5.79) #11.59, n = >11.59, n = SUV meanwb 3.30 (1.85) #2.14, n = >2.14, n = MTV WB (143.33) #37.34, n = >37.34, n = TLG WB (493.90) #407.48, n = >407.48, n = Adeno, adenocarcinoma; chemo, chemotherapy; chemo+rt, chemoradiation therapy; histology, tumor histology; large, large cell carcinoma; MTV, metabolic tumor volume; MTV WB, MTV of tumors in the whole body; none, no chemoradiation; pneum, pneumonectomy; RT, radiation therapy; SD, standard deviation; segmental, segmental resection; squamous, squamous cell carcinoma; others, other tumor histology; SUV max, maximum standardized uptake value; SUV maxwb, SUV max of tumors in the whole body; SUV mean, mean standardized uptake value; SUV meanwb, SUV mean of tumors in the whole body; TLG, total lesion glycolysis; TLG WB, TLG of tumors in the whole body; wedge, wedge resection; #, equal or less than a cutoff point (defined based on receiver operating characteristic curve analysis unless otherwise noted); >, greater than the cutoff point. and overall survival (OS). The multivariate models adjusted for stage (Stage I+II vs. III+IV) and other prognostic factors including chemoradiation therapy (none vs. chemotherapy only vs. radiation therapy+/-chemotherapy) and surgical procedure (lobectomy vs. other). Because of the relatively small number of deaths, a limited number of variables could be included in the multivariate models. The variables chosen were those with P <.10 in univariate analysis. C-statistics (20) were calculated (higher values indicating better discriminatory power). OS was calculated from the date of the initial PET/CT. The proportional hazards assumption was tested using Schoenfeld residuals (21). Kaplan-Meier 35

5 ZHANG ET AL Academic Radiology, Vol 20, No 1, January 2013 Figure 2. Kaplan-Meier curves of overall survival after baseline positron emission tomography/computed tomography based on clinical stage in 104 surgical patients with stage I to IV nonesmall-cell lung cancer. curves (22) were constructed after creating two groups using receiver-operating characteristic curves for the selection of the cutoff points. The value that maximized sensitivity and specificity for each of the PET/CT measurements was selected. The presurgical clinical TNM stage, according to the seventh edition of the AJCC manual (5), was used for survival analysis. The sixth edition of the postsurgical pathological stage was used as a reference standard to compare with the sixth edition of the presurgical clinical TNM stage in patients who had definitive surgery within 2 months after PET/CT scans, because the sixth edition was in effect when the pathology report was being written. A P value <.05 was considered statistically significant. Analyses were performed using Stata Version 12.1 (StataCorp, College Station, TX). RESULTS There were 47 male and 57 female patients with a median age of 67.9 years. There were 66 cases with stage I, 18 with stage II, 14 with stage III, and 6 with stage IV NSCLC based on presurgical clinical TNM staging according to the seventh edition of the AJCC manual (5). All patients had definitive surgery treatment including lobectomy in 82 patients, pneumonectomy in nine patients, and segmental resection in 4 patients. Wedge resection was performed in the 7 cases with stage IA and 2 cases with stage IB NSCLC for curative purpose. Forty-five patients had chemotherapy and/or radiation therapy. Fifty-nine patients had no chemotherapy or radiation therapy. The mean time between the PET/CT scan and surgery was 11.4 weeks with a standard deviation of 15.4 weeks. There were 52 patients with adenocarcinoma, 32 patients with squamous cell carcinoma, 8 patients with large cell carcinoma, 5 patients with adenosquamous carcinoma, 3 patients with synchronous adenocarcinoma and squamous cell carcinoma, 1 patient with synchronous large cell carcinoma and squamous cell carcinoma, 1 patient with atypical carcinoid tumor, 1 with spindle cell carcinoma with focus of squamous differentiation, and 1 patient with NSCLC of a type that was not further specified. There were a total of 37 deaths during follow-up. There was good interobserver agreement between the two radiologists on the SUV maxwb, SUV meanwb, MTV WB, and TLG WB. SUV maxwb demonstrated the lowest interobserver variability (CCC = 1.000, 95% CI 0.999e1.000). The interobserver variability for SUV meanwb, MTV WB, and TLG WB was only slightly higher (CCC = 0.882, 95% CI 0.844e0.919; CCC = 0.984, 95% CI 0.978e0.990; CCC = 0.992, 95% CI 0.989e0.995, respectively). It took about 6 hours for a radiologist to complete the PET tumor measurements on the 104 patients in reference with clinical PET/CT and CT reports. On average, it took 3.5 minutes per patient. The median follow-up was months (range 2.82e80.95 months) among the 67 survivors. A summary of PET measurements and OS probabilities at 1, 3, and 5 years are shown in Table 1. The cutoff point chosen based on receiver-operating characteristic curve analysis was 11.59, 2.14, ml, and SUVmL, for the SUV maxwb, SUV meanwb, MTV WB, and TLG WB, respectively. Kaplan-Meier curves of overall survival are provided for descriptive purposes in Figures 2 and 3. There was evidence of a trend of worse survival with higher stage of disease. The 5-year survival rate was 66.92%, 74.29%, 20.95%, and 20.83% for Stage I, Stage II, Stage III, and Stage IV patients, respectively. Such a trend was also seen for higher values of age, SUV maxwb, SUV meanwb, MTV WB, and TLG WB.The 5-year survival rate was 67.37%, 63.34%, 70.76%, 67.26%, and 64.74% for patients with age, SUV maxwb, SUV meanwb, MTV WB, and TLG WB at or below the cutoff point, respectively. The 5-year survival rate was 47.39%, 41.91%, 52.06%, 31.59%, and 11.97% for patients with age, SUV maxwb, SUV meanwb, MTV WB, and TLG WB above the cutoff point, respectively. In univariate analyses (Table 2), there was a statistically significant association of OS with clinical stage with a hazard ratio (HR) of 1.04 for stage II (P =.946), 3.60 for stage III cancer (P =.001), and 4.00 for stage IV (P =.013) as compared with stage I. There was a statistically significant association of OS with surgical procedure. The HR was 2.82 (P =.035) in patients who had pneumonectomy, and HR was 3.44 (P =.044) in patients who had segmental resection, in comparison with those who had lobectomy. There was no statistically significant difference in OS between the patients who had wedge resection and those who had lobectomy (HR = 1.12, P =.856). Furthermore, those who received radiation therapy alone or in combination with chemotherapy (in addition to the surgery) had significantly worse survival than those who only had surgery. There was also a statistically significant association of OS with MTV WB and TLG WB with the HR for a 1-unit increase in ln(mtv WB ) and ln(tlg WB ) of 1.40 (P =.004) and 1.26 (P =.011), respectively. There was 36

6 Academic Radiology, Vol 20, No 1, January 2013 WHOLE-BODY METABOLIC TUMOR BURDEN ON PET/CT Figure 3. Kaplan-Meier curves of overall survival after baseline positron emission tomography (PET)/CT grouped according to PET measurements in 104 surgical patients with stage I to IV nonesmall-cell lung cancer. (a) Whole-body tumor maximum standardized uptake value (SUV max ), (b) whole-body tumor mean standardized uptake value (SUV mean ), (c) whole-body metabolic tumor volume (MTV), (d) whole-body total lesion glycolysis (TLG). no statistically significant association of OS with age, gender, tumor histology, ln(suv maxwb ), or ln(suv meanwb )(P >.05). In multivariate analysis (Table 2), a statistically significant association of OS with MTV remained after adjusting for clinical stage of disease and other prognostic factors including chemoradiation therapy and surgical procedure. The adjusted HR for a 1-unit increase in ln(mtv WB ) was 1.32 (P =.039). There were also a statistically significant association of OS with TLG with the adjusted HR for a 1-unit increase in ln(tlg WB ) being 1.25 (P =.031). There was no statistically significant association of OS with ln(suv maxwb ) and ln(suv meanwb ) in multivariate analysis (P >.05). As further evidence of the prognostic value of the MTVand TLG variables, the C-statistics (a measure of discriminatory power) from the univariate models including these measures were all larger than the C-statistic of 0.59 for the model including stage only, whereas those for SUV mean and SUV max were smaller. A sensitivity analysis was performed by only including the 59 patients who were treated with surgery only, and the results also demonstrate the prognostic value of MTV and TLG (Table 3). The presurgical clinical and postsurgical pathological stages (sixth edition) were compared in 62 patients with definitive surgery within 60 days of the PET/CT scan. We used postsurgical pathological stage as a reference standard to compare the presurgical clinical stage. The overall stage, T-stage, and N-stage were the same on presurgical clinical staging and postsurgical pathology staging in 40 patients (64.52%), 34 patients (54.84%), and 44 patients (70.97%), respectively. The overall stage, T-stage, and N-stage were under-staged before surgery in 14 patients (22.58%), 20 patients (32.26%), and 10 patients (16.13%), respectively. The overall stage, T-stage, and N-stage were overstaged before surgery in 8 patients (12.90%), 8 patients (12.90%), and 7 patients (11.29%), respectively. As further comparison, survival analysis (starting from the date of surgery) was performed to compare the prognostic value of the presurgical clinical staging with that of postsurgical pathologic staging. 37

7 ZHANG ET AL Academic Radiology, Vol 20, No 1, January 2013 TABLE 2. Association of Overall Survival (OS) with Stage, Gender, Age, Tumor Histology, Treatment, and PET/CT Measurements in 104 Surgical Patients with Nonesmall-cell Lung Cancer Univariate Analysis Multivariate Analysis HR (95% CI) P Value HR* (95% CI) P Value* C-statistic Stage C-statistic = 0.59 I II 1.04 (0.38e2.79).946 III 3.60 (1.66e7.83).001 IV 4.00 (1.34e11.90).013 Gender Female Male 1.62 (0.85e3.11).146 Age 1.03 (0.99e1.06).163 Histology Adeno Squamous 0.89 (0.42e1.88).752 Large 0.31 (0.04e2.33).256 Others 2.06 (0.85e4.99).109 Chemoradiation None Chemo 0.55 (0.23e1.31).177 RT 4.93 (1.45e16.80).011 Chemo+RT 2.65 (1.11e6.30).027 Surgery Lobectomy Pneum 2.82 (1.08e7.39).035 Segmental 3.44 (1.03e11.48).044 Wedge 1.12 (0.34e3.72).856 SUV max C-statistic = 0.58 ln(suv maxwb ) 1.36 (0.93e1.99) (0.91e2.30) SUV mean C-statistic=0.57 ln(suv meanwb ) 1.43 (0.87e2.35) (0.91e3.29) MTV C-statistic = 0.63 ln(mtv WB ) 1.40 (1.11e1.76) (1.01e1.72) TLG C-statistic = 0.63 ln(tlg WB ) 1.26 (1.05e1.50) (1.02e1.54) CI, confidence interval; HR, hazard ratio; C-statistic, G onen and Heller s K concordance measure; ln, natural log; other abbreviations are as in Table 1. *Models adjusted for stage (I+II vs. III+IV), chemoradiation (none vs. chemo only vs. RT +/- chemo), and surgical procedure (lobectomy vs. other). Both staging methods were significantly associated with patients OS. The HR of stage III+IV was 3.61 with 95% CI of 1.79 to 7.29 (P <.001) and 2.74 with 95% CI of 1.16 to 6.51 (P =.022) in comparison with stage I on presurgical clinical staging and postsurgical pathological staging, respectively. The HR for stage II compared to stage I was not statistically different with both clinical and pathological staging methods. The C-statistics were 0.61 and 0.57 for the presurgical clinical and postsurgical pathological staging methods, respectively. DISCUSSION The present study is the first to show that the baseline wholebody MTV and TLG are prognostic indices independent of the AJCC/UICC TNM stage in surgical patients with stage I to IV NSCLC. The clinical implication of these results are that PET may give the ability to further stratify surgical patients with the same stage of NSCLC since the quantitative PET measures of whole-body MTV and whole-body TLG have prognostic value independent of clinical stage. Because accurate risk stratification based on whole-body MTV and whole-body TLG in NSCLC is important for determining treatment options and prognosis, incorporating whole-body MTV and whole-body TLG into the current TNM staging system should: 1) improve patient treatment selection, 2) aid medical and personal decision making by both clinicians and patients, and 3) in clinical trials, whole-body MTV and whole-body TLG may aid in the selection of comparable risk patients. On the last point, placing patients predicted to 38

8 Academic Radiology, Vol 20, No 1, January 2013 WHOLE-BODY METABOLIC TUMOR BURDEN ON PET/CT TABLE 3. Association of Overall Survival (OS) with Stage, Gender, Age, and PET/CT Measurements in 59 Surgical Patients with Nonesmall cell Lung Cancer Treated with Surgery Only Univariate Analysis Multivariate Analysis HR (95% CI) P Value HR* (95% CI) P Value* C-statistic Stage C-statistic= 0.57 I+II, n = 52 III+IV, n = (2.04e14.82).001 Gender Female, n = 35 Male, n = (0.69e4.04).253 Age 1.01 (0.96e1.06).657 SUV max ln(suv maxwb ) 1.63 (0.98e2.71) (0.90e2.64) SUV mean ln(suv meanwb ) 1.92 (0.96e3.85) (0.91e4.17) MTV ln(mtv WB ) 1.65 (1.20e2.28) (1.04e2.08) TLG ln(tlg WB ) 1.43 (1.12e1.84) (1.03e1.73) CI, confidence interval; HR, hazard ratio; C-statistic, G onen and Heller s K concordance measure; ln, natural log; other abbreviations are as in Table 1. *Adjusted for stage (I+II vs. III+IV). have similar risk factors into the test and control groups using stratified randomization could help clinical trials obtain more comparable and meaningful results. Our study demonstrated that performing MTVand TLG in surgical patients is not excessively time consuming, because there are not many tumor lesions in the surgical patients with NSCLC. It only took 3.6 minutes per case, on average, to complete the PET tumor measurements. The exact time needed for an individual case is dependent on the total number of lesions in the case. Our interobserver variability study demonstrated it is highly reproducible to measure MTV and TLG on PET images with CCCs greater than 0.95 for both MTV and TLG. The prognostic value of SUV measurement for surgical patients with early-stage NSCLC is controversial in the literature. Some studies have reported the prognostic value of preoperative SUV in early stage NSCLC (23,24). However, recent studies have questioned the prognostic value of the SUV measurement in surgical patients (25,26). The study by Agarwal et al in 363 patients with stage I and II NSCLC demonstrated preoperative SUV max is not an independent prognosticator for overall survival (25). There was a trend toward higher SUV measurement being associated with worse survival. However, in the current study, we failed to find significant prognostic value for the SUV parameters including SUV max and SUV mean in surgical patients with NSCLC. The comparison of clinical and pathological staging methods according to the sixth edition of TNM stage system in the current study indicated there is a substantial difference in the assignments of the overall stage, T-stage, and N-stage using two staging methods. The overall stage, T-stage, and N-stage were the same by the two methods in only 40 patients (64.52%), 34 patients (54.84%), and 44 patients (70.97%), respectively. However, on the further comparison with survival analysis, both staging methods were significantly associated with patients overall survival. The C-statistics results suggest that preclinical staging method may be better than postsurgical pathological staging method. However, this finding needs to be confirmed with prospective studies with large patient population. There are some limitations in this study. First, this is a retrospective study. Second, the patient population is relatively small. Third, we could not perform disease specific survival analysis because some of the deaths of patients in the study group were of an unknown cause. However, NSCLC is a deadly cancer. The 5-year overall survival rate of patients with stage I to IIIA NSCLC on average is only about 19% to 63% depending on the stage of the disease (8). The vast majority of the surgical NSCLC patients should have died of the cancer within our follow-up time. Therefore, overall survival would be expected to be quite close to the diseasespecific survival. CONCLUSION Baseline whole-body metabolic tumor burden as measured by MTV WB and TLG WB on FDG PET are prognostic measures independent of clinical stage with low interobserver variability and may be useful to further stratify surgical patients with NSCLC. This study also suggests MTV and TLG have more prognostic value than SUV max and SUV mean. These results will need to be validated in larger cohorts in a prospective study. 39

9 ZHANG ET AL Academic Radiology, Vol 20, No 1, January 2013 ACKNOWLEDGMENTS We gratefully acknowledge the generous assistance and valuable information provided to us by Ms. Cassie A. Simon. Special thanks to our chest oncological team in the University of Chicago for taking care of our study patients. REFERENCES 1. Parkin DM, Bray FI, Devesa SS. Cancer burden in the year 2000: the global picture. Eur J Cancer 2001; 37(Suppl. 8):S4e Traynor AM, Schiller JH. Systemic treatment of advanced non-small cell lung cancer. Drugs of Today 2004; 40:697e UyBico SJ, Wu CC, Suh RD, et al. Lung cancer staging essentials: the new TNM staging system and potential imaging pitfalls. Radiographics 2010; 30:1163e American Joint Committee on Cancer. AJCC Cancer Staging Manual. 6th ed. New York, NY: Springer, Lung. In: Edge SB, Byrd DR, Compton CC, et al., eds. AJCC Cancer Staging Manual. 7th ed. New York, NY: Springer, 2010; 253e Mountain CF. Revisions in the International System for Staging Lung Cancer. Chest 1997; 111:1710e Adebonojo SA, Bowser AN, Moritz DM, et al. Impact of revised stage classification of lung cancer on survival: a military experience. Chest 1999; 115:1507e van Rens MT, de la Riviere AB, Elbers HR, et al. Prognostic assessment of 2,361 patients who underwent pulmonary resection for nonesmall cell lung cancer, stage I, II, and IIIA. Chest 2000; 117:374e Sause W, Kolesar P, Taylor IVS, et al. Final results of phase III trial in regionally advanced unresectable non-small cell lung cancer: Radiation Therapy Oncology Group, Eastern Cooperative Oncology Group, and Southwest Oncology Group. Chest 2000; 117:358e Liao S, Penney BC, Wroblewski K, et al. Prognostic value of metabolic tumor burden on 18F-FDG PET in nonsurgical patients with non-small cell lung cancer. Eur J Nucl Med Mol Imaging 2012; 39:27e Liao S, Penney BC, Zhang H, et al. Prognostic value of the quantitative metabolic volumetric measurement on 18F-FDG PET/CT in stage IV non-surgical small cell lung cancer. Acad Radiol 2012; 19: 69e Yan H, Wang R, Zhao F, et al. Measurement of tumor volume by PET to evaluate prognosis in patients with advanced non-small cell lung cancer treated by non-surgical therapy. Acta Radiol 2011; 52:646e Lee P, Weerasuriya DK, Lavori PW, et al. Metabolic tumor burden predicts for disease progression and death in lung cancer. Int J Radiat Oncol Biol Phys 2007; 69:328e Lee P, Bazan JG, Lavori PW, et al. Metabolic tumor volume is an independent prognostic factor in patients treated definitively for non-small-cell lung cancer. Clin Lung Cancer 2012; 13:52e Shankar LK, HoffmanJM, Bacharach S, et al. Consensusrecommendations for the use of 18F-FDG PET as an indicator of therapeutic response in patients in National Cancer Institute Trials. J Nucl Med 2006; 47:1059e Werner-Wasik M, Nelson AD, Choi W, et al. What is the best way to contour lung tumors on PET scans? Multiobserver validation of a gradient-based method using a NSCLC digital PET phantom. Int J Radiat Oncol Biol Phys 2012; 82:1164e Lin LI. A concordance correlation coefficient to evaluate reproducibility. Biometrics 1989; 45:255e Lin LI. A note on the concordance correlation coefficient. Biometrics 2000; 56:324e Cox DR. Regression models and life-tables (with discussion). J Royal Stat Soc Series B 1972; 34:187e G onen M, Heller G. Concordance probability and discriminatory power in proportional hazards regression. Biometrika 2005; 92:965e Grambsch PM, Therneau TM. Proportional hazards tests and diagnostics based on weighted residuals. Biometrika 1994; 81:515e Kaplan EL, Meier P. Nonparametric estimation from incomplete observations. J Am Stat Assoc 1958; 53:457e van Baardwijk A, Dooms C, van Suylen RJ, et al. The maximum uptake of (18)F-deoxyglucose on positron emission tomography scan correlates with survival, hypoxia inducible factor-1alpha and GLUT-1 in non-small cell lung cancer. Eur J Cancer 2007; 43:1392e Downey RJ, Akhurst T, Gonen M, et al. Preoperative F-18 fluorodeoxyglucosepositron emission tomography maximal standardized uptake value predicts survival after lung cancer resection. J Clin Oncol 2004; 22:3255e Agarwal M, Brahmanday G, Bajaj SK, et al. Revisiting the prognostic value of preoperative (18)F-fluoro-2-deoxyglucose ((18)F-FDG) positron emission tomography (PET) in early-stage (I & II) non-small cell lung cancers (NSCLC). Eur J Nucl Med Mol Imaging 2010; 37:691e Vesselle H, Freeman JD, Wiens L, et al. Fluorodeoxyglucose uptake of primary non-small cell lung cancer at positron emission tomography: new contrary data on prognostic role. Clin Cancer Res 2007; 13:3255e