The maximum standardized uptake value (maxsuv) on

Transcription

1 Ratio of the Maximum Standardized Uptake Value on FDG-PET of the Mediastinal (N2) Lymph Nodes to the Primary Tumor May Be a Universal Predictor of Nodal Malignancy in Patients With Nonsmall- Cell Lung Cancer Robert James Cerfolio, MD, and Ayesha S. Bryant, MSPH, MD Department of Surgery, Division of Cardiothoracic Surgery, and Department of Epidemiology, University of Alabama at Birmingham School of Public Health, Birmingham, Alabama Background. The maximum standardized uptake value (maxsuv) on F-18 fluorodeoxyglucose positron emission tomography (FDG-PET) scan of mediastinal (N2) lymph nodes may predict pathology in patients with nonsmallcell lung cancer. However, the maxsuv varies among PET scanners. Thus, we evaluated the ratio of the max- SUV of the lymph node to the primary tumor at different centers to determine whether it was a universal predictor of lymph node malignancy. Methods. This is a retrospective review of a prospective database. Patients with nonsmall-cell lung cancer, a dedicated FDG-PET with the maxsuv of the primary lung tumor and FDG-avid mediastinal (N2) nodes reported (before therapy), and who underwent lymph node removal were eligible. Results. There were 239 patients with 335 FDG-PET positive N2 nodes at 14 different PET centers. The median ratio of the maxsuv of the lymph node to the maxsuv of the primary tumor of the pathologically proven malignant nodes was 0.58 (range, 0.32 to 1.61). Benign nodes had a median ratio of 0.40 (range, 0.21 to 1.10, p 0.02). The median value was similar for all centers except one. Receiver operating characteristics analysis determined the optimal value of the ratio that maximized sensitivity to be 0.56 or greater ( LR 6.6, sensitivity 94%, specificity 72%). Conclusions. The ratio of the maxsuv of the mediastinal (N2) lymph node to the maxsuv of the primary tumor in patients with nonsmall-cell lung cancer predicts mediastinal nodal pathology across different PET centers. When the ratio is 0.56 or greater, there is a 94% chance that the node is malignant. The ratio may take into account the different techniques used at different centers. (Ann Thorac Surg 2007;83: ) 2007 by The Society of Thoracic Surgeons The maximum standardized uptake value (maxsuv) on F-18 fluorodeoxyglucose positron emission tomography (FDG-PET) of a lung mass or of regional N2 mediastinal lymph nodes has been shown to correlate to pathology [1]. However, because the mean SUV and the maxsuv values are thought to differ from one center to another, some disagree as to the clinical value of an absolute value if it is not translatable from one center to another. Moreover, there are a few institutions in which the nuclear radiologists do not even report the maxsuv values. Despite these isolated exceptions, most have come to understand the clinical importance of the maximum or mean standardized uptake values (or peak SUV values) of the primary pulmonary mass and of the regional N2 mediastinal lymph nodes and now report them routinely. Recent data support the clinical importance of these values [2 4]. These reports Accepted for publication Dec 19, Presented at the Fifty-third Annual Meeting of the Southern Thoracic Surgical Association, Tucson, AZ, Nov 8 11, Address correspondence to Dr Cerfolio, Division of Cardiothoracic Surgery, University of Alabama at Birmingham, 1900 University Blvd, THT 712, Birmingham, AL 35294; robert.cerfolio@ccc.uab.edu. have shown that the maxsuv of a nonsmall-cell lung cancer (NSCLC) is a predictor of lymphovascular invasion and lymph node metastases, and others show the usefulness of the change in the maxsuv as a measure of the effectiveness of neoadjuvant therapy [5, 6]. The maxsuv has been shown to be a more reproducible and reliable number than the mean SUV, and thus is the preferred value to report [7]. Because the maxsuv of the tumor and the N2 lymph nodes is clinically significant and as there exists about a 10% to 15% difference in this value across centers, we sought a way to account for this variation and yet help clinicians who treat patients with NSCLC at different PET centers across the world. We hypothesized that a ratio between two SUV values may take this difference into account and control for it. We generated a ratio by dividing the maxsuv of the primary tumor by the maxsuv of the regional lymph node. Our hypothesis was that a ratio of these two maxsuv values, which were derived at the same time under the same FDG-PET scanning techniques, might be a universal predictor of lymph node pathology. We have called this ratio the 2007 by The Society of Thoracic Surgeons /07/$32.00 Published by Elsevier Inc doi: /j.athoracsur

2 Ann Thorac Surg CERFOLIO AND BRYANT 2007;83: PREDICTOR OF NODAL MALIGNANCY IN NSCLC 1827 PET predictive ratio for mediastinal N2 lymph node pathology. Patients and Methods Patient Selection Patients who were more than 19 years of age, underwent FDG-PET and computed tomography (CT) scanning or dedicated FDG-PET scanning with a detectable maxsuv (2.5 or greater) for both the primary tumor and N2 lymph node were eligible for this trial. Patients who received any chemoradiotherapy before PET scan were excluded from this trial. Patients who had N2 lymph nodes that were all PET silent or negative were not included in this study. Patients had to have at least one N2 node that was PET positive. The University of Alabama at Birmingham s Institutional Review Board approved this study and also approved the prospective database used for this study. Patient consent was obtained for entry into the prospective database. Radiologic Imaging The FDG-PET/CT scans that were performed at the University of Alabama were done using an integrated PET/CT scanner (GE Discovery LS PET-CT Scanner; General Electric, Milwaukee, Wisconsin). Patients were asked to fast for 4 hours, and then subsequently received 555 MBq (15 mci) of FDG intravenously followed by PET after 1 hour. The scans were performed from the skull base to midthigh level. The CT examination was used for attenuation correction of PET images. The scanning time for emission PET was 5 minutes per bed position. Iterative reconstruction with CT attenuation correction was performed. The most recent CT scan of the chest was also available for visual correlation. The maxsuv of the primary and of each suspicious lymph node station was determined by drawing regions of interest on the attenuation-corrected FDG-PET images around it. It was then calculated by the software contained within the PET or PET/CT scanner by the formula [8]: C Ci/ml MaxSUV ID Ci w kg where C activity at a pixel within the tissue defined by a Table 1. Results From Individual PET Centers region of interest; and ID injected dose per kg of patient s body weight (w). The maxsuv within the selected regions of interest was used throughout this study exclusively. Scans from other PET centers were also eligible for this study. Entry criteria used to accept other PET scan reports mandated a dedicated PET center (this excludes coincidence detection systems) and a PET report that provides the maximum standardized uptake values of the primary tumor in the lung and of the PET-positive regional N2 lymph nodes. In addition, the report had to be signed by a board-certified nuclear radiologist. If the patient was considered N2 lymph node negative by the PET or if the lymph node maxsuv was not provided, the patient was excluded from this study. Only the maxsuv was used for this study; mean and peak SUV were not used. If the patient had more than one lung nodule, the highest maxsuv of the two lung nodules was used to calculate the ratio. We have called this new ratio, which has no units, the PET predictive ratio (PPR) for mediastinal N2 lymph node pathology. It is thus calculated by: PPR for mediastinal N2 lymph node pathology maxsuv of N2 lymph node maxsuv of primary lung tumor Procedures, Staging, and Surgery All patients were clinically staged using the T, N, M classification system [9]. A clinical stage was assigned for the patient based on the integrated FDG-PET/CT scan results by one physician (R.J.C.). Biopsy specimens were obtained from suspicious N2, N3, or M1 areas (maxsuv 2.5) before pulmonary resection. Mediastinoscopy was used to obtain biopsies of suspicious lymph nodes in the paratracheal area (stations 2R, 4R, 2L, 4L, and the superior part of the 7), and endoscopic transesophageal ultrasound fine-needle aspiration (EUS-FNA) was used to obtain biopsies of suspicious posterior aortopulmonary window lymph nodes (5), subcarinal (7), periesophageal (8), and inferior pulmonary ligament lymph nodes (9). Endoscopic ultrasonography was performed under conscious sedation, as previously described [10, 11]. Statistical Methods Data were imported from the prospective database (Excel; Microsoft Corp, Seattle, Washington) into an Access data- Center No. Number of Patients Median Ratio of Pathology Positive Nodes (number of nodes) Median Ratio of Pathology Negative Nodes (number of nodes) Positive Predictive Value (%) ROC Value (156) 0.21 (23) 88% (29) 0.14 (9) 76% (23) 0.23 (8) 64% (18) 0.37 (12) 42% a (36) 0.33 (21) 59% 0.48 a Too few patients from center 4 to compute a receiver operating characteristics (ROC) curve. PET positron emission tomography.

3 1828 CERFOLIO AND BRYANT Ann Thorac Surg PREDICTOR OF NODAL MALIGNANCY IN NSCLC 2007;83: base (Microsoft Corp). Receiving operator characteristics (ROC) curves were generated using SAS 9.0 (SAS Institute, Cary, North Carolina). A p value of 0.05 or less was considered to indicate a statistically significant outcome that was unlikely due to chance. Results There were 239 patients (134 men and 105 women) with NSCLC who had 335 FDG-PET positive lymph nodes at 14 different PET centers. Median patient age was 68 years (range, 21 to 89). Table 1 displays the number of patients, median ratios of the pathologically positive and pathologically negative nodes, positive predictive value, and ROC curves by center. The values for centers that had more than 10 patients are shown individually, but centers with fewer than 10 patients had their data combined and are shown collectively. The overall median PPR for pathologically proven malignant N2 lymph nodes was 0.58 (range, 0.32 to 1.61). For the benign lymph nodes, the median PPR was 0.40 (range, 0.21 to 1.10; p 0.020). Table 1 also shows that the optimal cut-off ratio value as determined by ROC curves for each center. The value for the PPR that maximizes the sensitivity and specificity of an N2 mediastinal lymph node of being malignant for all five centers was similar, except for center 3. We also evaluated the lymph nodes that were removed that were deemed benign by PET scan. The vast majority (2,929) were pathologically negative (true negatives), but 73 lymph nodes were pathologically positive (falsely negative). Obviously, the median maxsuv of all of these nodes was 0. Figure 1 depicts the positive predictive value for various ranges of the PPR for mediastinal lymph nodes. This figure is powerful because it allows one to predict the probability that an N2 lymph node is malignant despite where the patient had their PET performed. Figure 2 shows the ROC curves for the optimal cut-off value of the PPR to predict metastatic in a mediastinal N2 lymph node for all centers. It found that a ratio of 0.56 or greater was predictive of lymph node malignancy ( LR 6.6, sensitivity 94%, specificity 72%). Fig 1. Positive predictive value (PPV [percent]) based on the positron emission tomography predictive ratio for mediastinal lymph N2 lymph node pathology. Fig 2. The receiver operating characteristics (ROC) curve for the optimal cut-off value of the positron emission tomography predictive ratio for mediastinal N2 lymph node pathology. Area under the curve 0.79; 95% confidence interval: 0.66 to 0.88; p Ratio 0.56 (sensitivity 94%, specificity 72%). Dashed lines are the 95% confidence intervals for the ROC curve (solid line). The diagonal gray line is the reference point for the area under the curve calculation. Comment The evolution from dedicated FDG-PET to integrated FDG- PET/CT scan has resulted in improved accuracy and improved location of lesions that are FDG avid. Integrated PET/CT allows one to pinpoint a lymph node location and determine if it is an N1 lymph node or an N2 node. Integrated PET/CT has become the standard of care for staging patients with NSCLC. Several studies have shown the superiority of integrated PET/CT scan over dedicated PET scan for staging of NSCLC [12, 13]. Recently, studies have shown that the SUVs that quantify the amount of FDG avidity a particular tumor has in a particular patient is an important piece of clinical information. Although SUVs are widely reported by most PET centers, there is an inherent variability in this measurement from one center to another. First, some centers report mean SUV instead of maxsuv. We prefer the max SUV because reports show that there is as much as 35% variability between different observers when the mean SUV value is used. This difference between observers when reading the same PET scan is reduced to only about 3% when the maximum SUV values are used [7]. However, the difference between maxsuvs at different PET centers remains high at 10% to 15%. Interestingly, a new value called the peak SUV has recently has been reported, but consensus on how to calculate it remains unresolved [14, 15]. The peak SUV may be determined by taking the top 5% or 10% of the hottest pixels and averaging them together. It may become the new preferred value over maxsuv in the future.

4 Ann Thorac Surg CERFOLIO AND BRYANT 2007;83: PREDICTOR OF NODAL MALIGNANCY IN NSCLC 1829 Because the formula of the maxsuv, as shown in the Methods section, is based only on the activity at a pixel, the injected dose of FDG, and the patient s weight, one might think that there should be little to no variation across centers. However, several factors do conspire to lead to a 10% to 15% difference in this value. These factors include the well counter that is used to measure the does of FDG injected, the time between dose injection and imaging, attenuation correction variables, injected FDG that may extravasate into the subcutaneous tissue, patient muscular activity in between injection and scanning, and finally, glucose and insulin levels. These factors may vary some among institutions [16] and between one scan and another. We hypothesized that a ratio of maxsuv values (ie, one maxsuv value divided by another) may be one way to negate these inherent differences between PET centers and standardize the measurement. That would potentially generate clinical research on PET data that was applicable to physicians using PET centers all over the world. This new ratio, which we call the PET predictive ratio for mediastinal N2 lymph nodes (or, PPR for N2 nodes), does seem to be reproducible for four centers and for the combination of 11 centers, which served as a fifth center. These data, if corroborated from other reports, may be clinically useful. The histology of N2 nodes is crucial in the treatment strategy of patients with NSCLC. This ratio may help direct the type of N2 biopsy employed. For instance, if the ratio shows the paratracheal 4R nodes has a PPR of 0.9 and the no. 8 periesophageal lymph node has a PPR of 0.5, perhaps a mediastinoscopy may be preferred over EUS-FNA. However, if the no. 8 N2 lymph node station has a PPR of 0.7 and the other N2 nodes are less, then perhaps that patient should go first for EUS-FNA. Our future studies are analyzing other ratio values as well using the maxsuv data. For instance, perhaps the ratio of the maxsuv/cm of tumor (in the tumor s greatest dimension on the CT scan) or the maxsuv/cm 2 or cm 3 may be an important way to ensure that the SUV is not just a surrogate marker for size. This ratio value (and perhaps we should use peak SUV/cm or peaksuv/cm 2 in the future) may be another important value that is easily generated from PET data, and one that helps guide clinical decisions. More data are needed from a larger number of patients and from many different centers. This preliminary report suggests that the PET positive ratio for mediastinal N2 lymph nodes has promise. It may be a predictor of N2 mediastinal lymph node pathology for different PET centers for NSCLC patients who have the maximum SUV of both their lung cancer and of regional N2 lymph nodes reported. References 1. Bryant AS, Cerfolio RJ, Klemm KM, et al. Maximum standard uptake value of mediastinal lymph nodes on integrated FDG-PET-CT predicts pathology in patients with non-small cell lung cancer. Ann Thorac Surg 2006;82: Cerfolio RJ, Bryant AS. Maximum standardized uptake values on positron emission tomography of esophageal cancer predicts stage, tumor biology, and survival. Ann Thorac Surg 2006;82: Cerfolio RJ, Bryant AS, Ojha B, et al. The maximum standardized uptake values on positron emission tomography of a non-small cell lung cancer predict stage, recurrence, and survival. J Thorac Cardiovasc Surg 2005;130: Downey RJ, Akhurst T, Gonen M, et al. Preoperative F-18 fluorodeoxyglucose-positron emission tomography maximal standardized uptake value predicts survival after lung cancer resection. J Clin Oncol 2004;22: Cerfolio RJ, Bryant AS, Winokur TS, et al. Repeat FDG-PET after neoadjuvant therapy is a predictor of pathologic response in patients with non-small cell lung cancer. Ann Thorac Surg 2004;78: Ryu JS, Choi NC, Fischmann AJ, et al. FDG-PET in staging and restaging non-small cell lung cancer after neoadjuvant chemoradiotherapy: correlation with histopathology. Lung Cancer 2002;35: Lee JR, Madsen MT, Bushnel D, et al. A threshold method to improve standardized uptake value reproducibility. Nucl Med Commun 2000;21: Nabi HA, Zubeldia JM. Clinical applications of F18-FDG in oncology. J Nucl Med Technol 2002;30: Mountain CF. Revisions in the international system for staging lung cancer. Chest 1997;111: Hawes RH, Gress F, Kesler KA, Cummings OW, Conces DJ Jr. Endoscopic ultrasound versus computed tomography in the evaluation of the mediastinum in patients with nonsmall-cell lung cancer. Endoscopy 1994;26: Cerfolio RJ, Bryant AS, Ojha B, et al. Improving the inaccuracies of clinical staging of patients with NSCLC: a prospective trial. Ann Thorac Surg 2005;80: Cerfolio RJ, Buddhiwardhani O, Bryant AS, et al. The accuracy of integrated PET-CT compared with PET alone for the staging of patients with non-small cell lung cancer. Ann Thorac Surg 2004;78: Lardinois D, Weder W, Hany TF, et al. Staging of non smallcell lung cancer with integrated positron-emission tomography and computed tomography. N Engl J Med 2003;348: Salaun PY, Grewal RK, Dodamane I, et al. An analysis of the 18F-FDG uptake pattern in the stomach. J Nucl Med 2005;46: Vranjesevic D, Schiepers C, Silverman DH, et al. Relationships between 18F-FDG uptake and breast density in women with normal breast tissue. J Nucl Med 2003;44: Shankar LK, Hoffman JM, Bacharach S, et al. Consensus recommendations for the use of FDG-PET as an indicator for therapeutic response in patients in National Cancer Institute trials. J Nucl Med 2006;47: DISCUSSION DR JOSEPH I. MILLER (Atlanta, GA): Cerf, Dr Kern, Dr Moon. It is the collegiality of the Southern Thoracic Surgical Association and the size of our forum that allows a senior person such as I to poke a little fun at a younger member of whom I am very proud of what he has accomplished at UAB. So, Cerf, don t take this personally. I am glad that Cerf has moved on from air leaks, which he produced four papers, at least two presented before this Association, to the SUV, and now this is the fourth paper in the last two years on a similar subject, though it be either lung or esophagus. I think he is probably Professor of Nuclear Medicine at the University of Alabama. I know I need to be careful what I say because you remember yesterday in the pro-con debate when his opponent was already on the ground after being hit eight times, writhing in pain, pleading for mercy, Cerf says, I hate to put the dagger in, but he put

5 1830 CERFOLIO AND BRYANT Ann Thorac Surg PREDICTOR OF NODAL MALIGNANCY IN NSCLC 2007;83: it right through the heart, delivering the death knell. So I have to be careful what I say, but I will move on. But it is the collegiality of discussion at the Southern Thoracic that we are able to say such things. This is a very well thought out study and I appreciate the author sending me the manuscript before the meeting. First, he makes several well known points in his manuscripts that the SUV is a known predictor of disease virulence, predicts response to chemotherapy, and also the potential for predicting nodal spread. He then presents a new formula called a PET predictive ratio that would take a computer to compute. I really think he wants it to be known as the Cerf ratio. The criticism of the study is that it is carried out in 14 different centers all related to UAB. We all know that all CT-PET scanners are not the same. In addition, approximately 10 years ago, as I remember talking to Cerf, they did not even report the SUV out on PET scans at the University of Alabama. I am not sure that a PET predictive ratio at the present time really changes how we practice. Cerf, I have just a couple of questions for you. How does the PET predictive ratio change the way you practice at the current time? If lymph nodes are greater than 1.5 cm, then it is probably gong to be biopsied by either mediastinoscopy or the new technique of fiberoptic bronchoscopy with ultrasound FNA. In conclusion, he points to the ongoing future studies, so I know that we will have more. In his reference list, he has 18 references, and in these, nine are headed by Cerf. I know we are going to hear more from him in the future. I think he is correct in that a multi-institutional study is necessary to confirm the usefulness of the PET predictive ratio. Cerf, I think you are a great guy and you have done wonders at the University of Alabama, and I am very proud of you. DR CERFOLIO: First of all, and I mean this in all sincerity, it is an honor for me to have Dr Miller discuss my paper. He doesn t know it, but the only reason I am standing here today is because of Dr Miller. He took me under his wing when I first started at UAB and was a lone fish out of water, and Dr Miller introduced me to this society as well as others he got me into the Southern, and onto the program committee so I am honored to have him as my discussant. Dr Miller, you sir can poke fun at me all you want. To answer your questions, all of these centers were not related to UAB. These are centers all throughout Alabama, Tennessee, Florida, and other places. So I just want to make sure that is clear. These are 14 different centers. The center with the most patients however is the one integrated PET/CT center at UAB. Second, our centers have always reported the maxsuvs when we first got it you may be confusing us with Duke that has not provided SUV data on PET scans over the past several years. We didn t have an integrated PET/CT scanner at UAB until I think We were using one in private practice across the street and they were recording maxsuv as well. Finally, your third question is, how does the PPR change my practice? Well, it really does change my practice. If a lymph node is enlarged on CT scan, you are right, we are going to biopsy it irrespective of the PET, but most enlarged lymph nodes are PET positive, assuming the patient has non-small cell lung cancer. So the PPR gives you a target as to what to go after first because it is most likely to be malignant. For instance, if the 4R has a PPR of 0.4, but the subcarinal #7 LN s PPR is 0.7, then I would go to an EUS-FNA, because we have a higher accuracy of hitting the subcarinal nodes with an EUS-FNA than a mediastinoscopy and it is more likely positive than the 4R. We would go to the EUS. So it has changed my practice. Thank you. DR DANIEL L. MILLER (Atlanta, GA): Cerf, I enjoyed your paper. It was actually the shortest presentation I think I have ever heard you give in your academic life, so I don t know how valid the data is in this series. The big question I have is that if you have a tumor that has an endobronchial component resulting in postobstruction pneumonia or if the tumor has a necrotic center, then you may have a falsely elevated maxsuv, which is going to drive your ratio down. Did you look at that at all, because I think you have to include the anatomical location of the tumor and its secondary effects in analyzing the maxsuv? DR CERFOLIO: That is an important point and is true for mean SUV not maxsuv. Your comments are correct if you look at the mean SUV value of a tumor with a necrotic center, but if you look at the maxsuv, the entire middle part of that tumor could have a maxsuv of zero but you will have a pixel with a maxsuv of 11, and so you report it as an 11. So this is why radiologists have used the max over the mean, because it is more reproducible, less objective. So the maxsuv won t be affected. Soon it will be peak SUV as I discussed during the talk. DR BRYAN FITCH MEYERS (St. Louis, MO): Dr Cerfolio, I have one question for you. You used a ratio, but you didn t show us that the ratio gave you any additional benefit over just using the maxsuv of the mediastinal lymph nodes. Do you have any comparison of using the ratio system that you have reported versus just a simpler maxsuv system on the same patients to show that it discriminates better? DR CERFOLIO: We have shown in previous reports that the maxsuv of lymph nodes, and I didn t want to re-bore anybody too much with the other stuff that we have presented nationally, that there is a very large overlap between falsely positive and truly positive mediastinal lymph nodes. The idea of a ratio was to try to decrease the false positives rates when compared to using just the absolute value and to produce a number that could be applied to patients on various PET s all over the world. Now it is true that we did not set out to do this in this manuscript the goal of this study was to identify a ratio and see if it was consistent between different Pet centers and it was. But we did not show the fall in the false positive rate is in this report, you are correct. One of the criticisms I have received in the past, as Dr Miller talked about me doing a lot of SUV papers, was that, well, maybe the maxsuv values I talk about are good for me, but how does it help you in Atlanta or how does it help you in St. Louis? So we wanted to do something that was going to be more applicable to people everywhere and not just isolated to the UAB PET scanner. And so that is why we did this ratio study. We could go back and look at the maxsuv of each lymph node and calculate the false positive rate compared to the false positive rate for the ratio.