| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Ann Thorac Surg 2006;82:1009-1015
© 2006 The Society of Thoracic Surgeons
a Division of Thoracic Surgery, Department of Cardiothoracic Surgery, Weill Medical College of Cornell University, New York, New York
b Department of Genetic Medicine, Weill Medical College of Cornell University, New York, New York
Accepted for publication March 17, 2006.
* Address correspondence to Dr Korst, Department of Cardiothoracic Surgery, Suite M404, Weill Medical College of Cornell University, 525 East 68th St, New York, NY 10021 (Email: rjk2002{at}med.cornell.edu).
Presented at the Forty-second Annual Meeting of The Society of Thoracic Surgeons, Chicago, IL, Jan 30–Feb 1, 2006.
| General thoracic surgery:
The Annals of Thoracic Surgery CME Program is located online at http://cme.ctsnetjournals.org. To take the CME activity related to this article, you must have either an STS member or an individual non-member subscription to the journal.
|
 |
Abstract |
|---|
 Top Abstract IntroductionPatients and MethodsResultsCommentDiscussionReferences |
|---|
METHODS: A cohort of patients who had abnormal postoperative CT scans of the chest and upper abdomen in 2002 were followed up into 2005. Abnormalities consisted of pulmonary nodules, pleural effusions, or adenopathy. Data collected included recurrence patterns, the availability of previous scans for comparison, the interval between initial resection and the abnormal CT, nodule size, growth, and multiplicity, as well as progression of pleural effusions or adenopathy.
RESULTS: In all, 105 scans in 92 patients were read as abnormal in 2002 by the radiologist. After further investigation or follow-up, or both, for a mean of 3.2 years, 78% of patients who had recurrent NSCLC had their site of first recurrence inside the chest. The negative predictive value of CT when used by the thoracic surgeon was 99%; however, the positive predictive value was only 53%. Abnormalities that correlated with the diagnosis of recurrent cancer included pulmonary nodules that either grew or were larger than 1 cm and pleural effusions that developed after the first postoperative year.
CONCLUSIONS: Intrathoracic recurrent NSCLC was rarely missed by the surgeon utilizing surveillance CT, but a significant number of negative investigations were generated by its use. Characteristics of abnormal surveillance CT findings exist that correlate with the presence of malignancy.
Computed tomography (CT) of the chest is used by many clinicians to follow patients with previously resected nonsmall-cell lung cancer (NSCLC), despite a paucity of supporting data. As the majority of CT scans used in this capacity are read as abnormal by the radiologist [1], the potential exists for the generation of a large number of additional tests for these patients, which is both costly and may be harmful. Given this information, we sought to determine the outcome of all abnormalities detected by the radiologist on surveillance CT scans performed in a cohort of patients in a surveillance program conducted by their thoracic surgeon, and use these data to determine the accuracy of this follow-up technique in detecting recurrent lung cancer. In addition, radiographic characteristics of abnormalities on surveillance CT were evaluated to determine which abnormalities were predictive of malignancy.
|
Patients and Methods |
|---|
|
TopAbstractIntroductionPatients and MethodsResultsCommentDiscussionReferences |
|---|
Acquisition of Data and Definitions
Data were acquired from surveillance CT scan reports and office charts. When additional testing was performed to document disease recurrence, hospital records were reviewed. Approval for review of office charts and hospital records was obtained from the Institutional Review Board. Because this study is strictly retrospective, with no identification of individual patients, the need for informed consent was waived by the Institutional Review Board. Data collected included the stage of the initial lung cancer that was resected, the interval between the date of initial lung cancer resection and performance of surveillance CT, and the stability of findings when compared with previous scans. With regard to specific lesions, the following information was ascertained:
Pulmonary nodules
Nodule size was recorded as the size of the largest nodule reported on the scan. Nodules described as "tiny" were assigned a size of less than 1 cm for statistical analysis. The number of nodules reported on the surveillance scans was also recorded. A "growing" nodule refers to a lesion that had either demonstrated growth, or was a new finding when compared with previous scans.
Pleural fluid
Effusions were characterized as "new" only if they were not present on prior postoperative surveillance scans. It was also noted whether the effusion was detected on the initial postoperative surveillance scan. As an example, if a patient underwent an initial lung cancer resection in 2002, and the first postoperative surveillance scan was performed the same year and demonstrated the presence of pleural fluid, the effusion was not classified as "new."
Adenopathy
Lymph node size and stability were recorded.
Finally, survival and site of first recurrence were recorded for all patients. Intrathoracic recurrence was proven by biopsy or resection in all patients, and was defined as recurrence anywhere in the lungs, mediastinum, and pleural spaces. All patients in the study cohort continued to be followed by the authors into 2005 according to the previously published follow-up regimen. For purposes of this discussion, "recurrent" refers to the reappearance of lung cancer after complete resection, with no attempt made to distinguish which cases represented a recurrence of the initially resected cancer versus those constituting a new primary cancer.
"Suspicious" scans were previously defined [1] as those for which the surgeon altered the follow-up regimen based on the scan findings, whereas "nonsuspicious" scans were those that, although read as abnormal by the radiologist, were not pursued further.
Statistical Analysis
Proportions were compared using 
2 contingency tables. Fisher's exact test was utilized when expected frequencies were less than 5. The difference between means was assessed using the two-tailed, independent samples t test. Values of p less than 0.05 were considered statistically significant.
|
Results |
|---|
|
TopAbstractIntroductionPatients and MethodsResultsCommentDiscussionReferences |
|---|
Specific Recurrence Patterns in Patients With Surveillance CT Scans
Figure 1
demonstrates the flow of these 92 patients through postoperative surveillance. Two had suspicious CT findings that were not classified as nodules, fluid, or adenopathy: 1 with a liver lesion and 1 with a chest wall mass, both of which were documented to be recurrent disease. These latter 2 patients were not considered further. Patients in whom recurrent or new primary lung cancer was not diagnosed in 2002 were followed into 2005, representing a mean follow-up period of 3.2 years. Table 1
demonstrates the sites of first recurrence found in all 90 patients in this study cohort. Seventy-eight percent of cases (21 of 27) of recurrent lung cancer were intrathoracic in this cohort undergoing surveillance CT.
|
|
|
|
|
Statements regarding the size of pulmonary nodules were present in all CT radiology reports. Specific nodule sizes were mentioned in 79 of 90 reports, with 11 reports referring to nodule size as "tiny." Figure 2 demonstrates the distribution of sizes of the largest nodules on these scans. Scans on which the largest nodule was greater than 1 cm in size were much more likely to reflect malignancy than those that contained nodules 1 cm or less in diameter (greater than 1 cm, 6 of 15 [40%], versus 1 cm or less, 3 of 75 [4%]; p < 0.0001). Despite this, 9 scans contained nodules greater than 1 cm in size that never proved to represent malignancy. In 6 of these patients, the surgeon thought the nodules were suspicious and pursued further workup, which was negative. In the remaining 3, the surgeon was not suspicious, and follow-up surveillance scans demonstrated nodule stability into 2005. Conversely, of the 3 scans containing nodules 1 cm or less in size that ultimately proved malignant, 2 had new lesions not seen on previous scans, 0.8 and 1.0 cm in size. The third patient had multiple "tiny reticulonodular densities suggestive of lymphatic spread."
|
Nodule multiplicity was addressed by grouping scans according to whether they contained a single or multiple nodules. Thirty-eight of 90 scans (42%) contained a solitary nodule, and 52 showed multiple nodules (58%). No statistically significant difference was detected in the prevalence of cancer according to the multiplicity of nodules (solitary, 6 of 38 [15.8%], versus multiple, 5 of 52 [9.6%]; p = 0.46).
The frequency of malignancy was then evaluated according to the stage of the initially resected NSCLC. Patients were arranged into three groups: stage IA, stage IB, and stages II through IV. The frequency of malignancy was not significantly different if the surveillance scan was performed in a patient with stage IA disease (4 of 40 patients) versus stage IB disease (4 of 18 patients) versus stages II through IV disease (1 of 21 patients; p = 0.28).
Pleural Fluid Reported on Surveillance CT
Pleural effusions were reported by the radiologist on 27 of the original 105 abnormal scans (25.7%) in 2002. These 27 scans were performed in 26 patients, with 1 patient having 2 scans. Nine patients had pleural effusions that were deemed suspicious by the surgeon in 2002, with 4 ultimately having malignant pleural disease in that hemithorax. Table 5
demonstrates the clinical characteristics of these patients. Scans showing fluid subsequent to the first postoperative year were much more likely to suggest malignancy in the pleural space than if the scan showed fluid in the first year after the initial lung cancer resection (first postoperative year, 0 of 20 [0%], versus after first postoperative year, 4 of 7 [57%]; p = 0.002). In addition, it should be noted that all 4 patients with malignant pleural disease had previous postoperative scans revealing the absence of pleural fluid ("new" effusions).
|
|
Comment |
|---|
|
TopAbstractIntroductionPatients and MethodsResultsCommentDiscussionReferences |
|---|
Site of First Recurrence in Patients Undergoing Surveillance CT
In the present report, the site of first recurrence was inside the thoracic cavity in approximately 80% of patients, with only 2 patients having brain lesions as the first site of recurrence. Further, nearly 40% of intrathoracic recurrences were treated as new primary NSCLC. Although this cohort represents a select group of patients undergoing surveillance CT with abnormalities found by the radiologist, the paucity of distant metastatic recurrence, most significantly the brain, is worth noting, especially since historical studies have suggested that the brain is the most common site of recurrence in resected lung cancer patients [3–7]. This is especially important given that the majority of patients undergoing surveillance CT have abnormal scans as interpreted by the radiologist [1]. One of the limitations of the present study is the lack of follow-up data on patients with negative surveillance scans in 2002, data that are needed to definitively assess recurrence patterns in resected patients. Prospective studies will need to be conducted to determine whether recurrence patterns are indeed altered by the performance of surveillance CT in patients undergoing complete resection for NSCLC.
Accuracy of the Surgeon's Interpretation of Surveillance CT
Given that the majority of postoperative surveillance CT scans are read as abnormal by the radiologist with regard to pulmonary nodules, pleural fluid, and adenopathy, the clinician's ability to distinguish benign from malignant findings becomes of paramount importance, since the vast majority of these abnormal findings do not represent cancer. The nondiscriminative further workup of all abnormalities that are read on CT scans will unquestionably result in high costs, with potential danger to patients. In the present study, patients undergoing surveillance CT were all part of a postoperative follow-up regimen conducted by the thoracic surgeon who performed the initial lung cancer resection. Each individual surgeon based the decision to perform further workup of CT abnormalities on not only the radiologist's report, but also the surgeon's own interpretation of the films while taking into account his or her existing knowledge regarding an individual patient's disease process.
The overall accuracy of the surgeons' initial impressions of the 105 abnormal scans was nearly 88%, implying that the combination of the surgeon's CT interpretation as well as the accompanying clinical information fared reasonably well in the detection of recurrent lung cancer inside the chest. With regard to the negative predictive value, it was unusual for a surgeon to "miss" a patient with recurrence when the surgeon thought the surveillance scan was not suspicious for metachronous lung cancer. That occurred only once in 60 patients (73 scans), in a patient with resected stage IIIA disease who had 1.5 cm mediastinal lymph nodes on his 2002 surveillance CT. Although not considered suspicious at the time by the surgeon, the patient underwent a PET scan ordered by an outside physician and it was positive in the mediastinum, prompting a positive biopsy.
The positive predictive value of the surgeons' impressions of the 105 abnormal scans was significantly lower, however. Although the surgeon found the abnormalities on 30 surveillance scans to be suspicious for metachronous lung cancer, only 16 of these 30 lesions deemed suspicious by the surgeon proved to be malignant (either by further workup in 2002 or during the additional 3 years of follow-up). Importantly, as Table 4 reveals, this relatively low positive predictive value applied to all three defined abnormalities (nodules, fluid, adenopathy). The clinical significance of the positive predictive value is that 14 of the 92 patients in the follow-up regimen who had scans read as abnormal by the radiologist (15%) underwent unnecessary further testing, some of which was invasive.
Differentiating Benign From Malignant Findings on Postoperative Surveillance CT
With an additional mean follow-up of 3.2 years, it became possible to specifically identify which lesions on the 2002 surveillance scans actually represented cancer. Using this information, we sought to determine the specific characteristics of the nodules, pleural fluid or adenopathy that were most likely to represent malignancy, with the aim of improving the positive predictive value of the clinician's impression of surveillance CT.
Pulmonary nodules were evaluated according to the size of the largest nodule on an individual scan, their multiplicity, whether or not they represented new nodules or grew in size when compared with prior scans, as well as the patient's initial cancer stage. Other characteristics, including nodule shape and density, were not routinely addressed by the radiologists in their reports and, therefore, could not be addressed in this study. The size of the pulmonary nodules as read by the radiologist was an important predictor of malignancy. Although lesions greater than 1 cm in diameter were significantly more likely to be malignant, it is important to note that the majority of these larger lesions were still benign after further investigation and follow-up. Similarly, if nodules were either new or grew in size in comparison to prior scans, malignancy was much more likely than if stability was noted. However, two thirds of new or growing nodules were still benign. As a result, although nodule size and growth are important when evaluating surveillance CT scans, selecting scans with these two characteristics alone for further investigation will not enhance the positive predictive value of the surgeon's impression. In addition, nodule multiplicity did not correlate with malignancy in this study. However, an important observation from this analysis is that although the vast majority of surveillance scans with pulmonary nodules revealed multiple, subcentimeter nodules, only 1 of these patients had malignancy. This observation implies that the common finding of multiple, tiny nodules in a patient with previously resected NSCLC may not need further investigation. Finally, the role of PET in the postoperative setting may eventually help to discriminate benign from malignant lesions; however, that remains to be investigated.
Pleural fluid on the operated side is a common finding immediately after pulmonary resection for lung cancer. However, it is our contention (unpublished observation) that the majority of these pleural effusions dissipate within the first 6 months postoperatively. In this regard, the only pleural effusions that were malignant in the present study were those that developed after the first postoperative year, and were not present on prior postoperative surveillance scans. Further, when a pleural effusion was present on surveillance CT performed in the first year postoperatively, none proved to be malignant. As shown in Table 4, the positive predictive value of the surgeon's impression of a surveillance scan revealing pleural fluid was only 44%; however, if all effusions present in the first postoperative year were not regarded as suspicious, this value would rise to 100%.
It is difficult to draw conclusions from the patients with adenopathy on their surveillance scans owing to the small number of scans with this finding. It seems intuitive, however, that patients with N2 disease at the time of their initial lung cancer resection would have a higher likelihood of recurrence in this area. In this regard, of the 4 patients in this study with recurrence in the mediastinum, 2 had mediastinal involvement at the time of their initial resection; however, no patient with benign adenopathy on surveillance CT had an initial stage of IIIA. Although not statistically significant, this trend may prove to be real in a large, prospective evaluation of surveillance CT. In this regard, our group is currently conducting a prospective trial evaluating the role of PET/CT for the follow-up of patients with completely resected NSCLC.
In summary, although the majority of surveillance CT scans performed after complete resection for NSCLC are read as abnormal by the radiologist, the vast majority of abnormalities do not prove to be malignant. Intrathoracic, recurrent cancer was rarely missed by the surgeon utilizing surveillance CT, but workup of irrelevant lesions is not uncommon. Finally, characteristics of abnormal surveillance CT findings exist that correlate with the presence of malignancy.
|
Discussion |
|---|
|
TopAbstractIntroductionPatients and MethodsResultsCommentDiscussionReferences |
|---|
Now, granted that this is a little unfair, because the radiologist's job is not to decide necessarily what is relevant but in fact to just identify what is abnormal, and also they very often have inadequate clinical history. In a moment I am going to ask Dr Korst to tell us what parameters the surgeon used to identify which scans he or she termed suspicious, but before I do that, I think it bears mentioning that perhaps what this study shows is the importance of being clinically relevant, the importance of continuity of care, and the importance of clinical judgment. And perhaps in this day of outsourcing so much of what we really can probably do better ourselves, maybe this outsourcing is doing our patients a great injustice.
Think for a moment of what would have happened to those 92 patients if their postoperative follow-up had been relegated to just the internist and the radiologist? I propose that the picture would have been a lot different and a lot more costly.
So I have two related questions for Dr Korst, and they are, what specifically were the parameters that the thoracic surgeon used to determine which of the abnormal CT scans were okay, and also, are you prepared to say, based on your study, that in terms of reading postoperative CT scans, when the thoracic surgeon reads the CT scan, the radiologist is superfluous? Thank you.
DR KORST: I would like to thank Dr Feins for those insightful comments. They are obviously very important points. With regard to how we decided which scans were suspicious, the ability to compare a scan to a prior surveillance scan was of utmost importance. In addition, nodule size and character—whether it was speculated or ground glass—was also important. But it is also very important to note that the clinical history is so important, and the surgeon knows the clinical history and the radiologist does not, so that plays a major role.
I don't believe the radiologist is superfluous. I can tell you what I do in the clinic. Our patients come in in the morning and they have their scan, and then I look at the scan when I see the patient in the afternoon. I look at both the scan myself and then I read the radiology report to see if I missed anything on that scan. I agree that it is not the radiologist's job to tell us what is lung cancer and what is not lung cancer, but I think it is the radiologist's job to go over that scan with a finetoothed comb, and we can compare our readings to their readings and determine if we missed anything that perhaps would need to be further investigated.
|
References |
|---|
|
TopAbstractIntroductionPatients and MethodsResultsCommentDiscussionReferences |
|---|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| ANN THORAC SURG | ASIAN CARDIOVASC THORAC ANN | EUR J CARDIOTHORAC SURG |
| J THORAC CARDIOVASC SURG | ICVTS | ALL CTSNet JOURNALS |
