Definition of a Positive Test Result in Computed Tomography Screening for Lung Cancer A Cohort Study

Transcription

1 Original Research Annals of Internal Medicine Definition of a Positive Test Result in Computed Tomography Screening for Lung A Cohort Study Claudia I. Henschke, PhD, MD; Rowena Yip, MPH; David F. Yankelevitz, MD; and James P. Smith, MD, for the International Early Lung Action Program Investigators* Background: Low-dose computed tomography screening for lung cancer can reduce mortality among high-risk persons, but falsepositive findings may result in unnecessary evaluations with attendant risks. The effect of alternative thresholds for defining a positive result on the rates of positive results and cancer diagnoses is unknown. Objective: To assess the frequency of positive results and potential delays in diagnosis in the baseline round of screening by using more restrictive thresholds. Design: Prospective cohort study. Setting: Multi-institutional International Early Lung Action Program. Patients: participants with baseline computed tomography performed between 2006 and Measurements: The frequency of solid and part-solid pulmonary nodules and the rate of lung cancer diagnosis by using current (5 mm) and more restrictive thresholds of nodule diameter. Results: The frequency of positive results in the baseline round by using the current definition of positive result (any parenchymal, solid or part-solid, noncalcified nodule 5.0 mm) was 16% (3396/ ). When alternative threshold values of 6.0, 7.0, 8.0 and 9.0 mm were used, the frequencies of positive results were 10.2% (95% CI, 9.8% to 10.6%), 7.1% (CI, 6.7% to 7.4%), 5.1% (CI, 4.8% to 5.4%), and 4.0% (CI, 3.7% to 4.2%), respectively. Use of these alternative definitions would have reduced the work-up by 36%, 56%, 68%, and 75%, respectively. Concomitantly, lung cancer diagnostics would have been delayed by at most 9 months for 0%, 5.0% (CI, 1.1% to 9.0%), 5.9% (CI, 1.7 to 10.1%), and 6.7% (CI, 2.2% to 11.2%) of the cases of cancer, respectively. Limitation: This was a retrospective analysis and thus whether delays in diagnosis would have altered outcomes cannot be determined. Conclusion: These findings suggest that using a threshold of 7 or 8 mm to define positive results in the baseline round of computed tomography screening for lung cancer should be prospectively evaluated to determine whether the benefits of decreasing further work-up outweigh the consequent delay in diagnosis in some patients. Primary Funding Source: The Flight Attendant Medical Research Institute and the American Legacy Foundation. Ann Intern Med. 2013;158: For author affiliations, see end of text. * For a list of members of the International Early Lung Action Program, see the Appendix (available at See also: Print Editorial comment In computed tomography (CT) screening for lung cancer, positive result or the initial low-dose CT in a given round of screening indicates whether further diagnostic work-up is needed before the first scheduled repeated screening. A definition that is too inclusive may cause excessive diagnostic work-up and unnecessary treatment and thus potential harm. If the definition is too restrictive, further diagnostic work-up of patients with lung cancer may be delayed to the first annual repeated round of screening, thus potentially leading to progression of the lung cancer stage before the first annual screening. In ELCAP (the Early Lung Action Project), begun in 1993, we considered identification of a noncalcified nodule (NCN) of any size on the initial, low-dose CT scan in the baseline round to be a positive result, because we had no previous data on which to base a more restrictive definition. Among the first 1000 participants, 23% had a positive result (1). The definition of a positive result in the baseline round was updated on the basis of frequency of cancer in the ELCAP data (2 6). Thus, in the subsequent screenings in New York ELCAP (NY-ELCAP) (7) and International ELCAP (I-ELCAP) (8), a positive result was defined as a nodule with a 5.0-mm or larger diameter and the frequency of a positive result was reduced to 14.4% in NY-ELCAP and 13.3% in I-ELCAP. Although the NLST (National Lung Screening Trial) showed a reduction in mortality with low-dose CT screening (9), the rate of positive results in the baseline round was 28% and concerns have been raised about falsepositive findings leading to unnecessary surgery and the risks of radiation from the follow-up scans (10 12). Because the NLST screenings ended in 2006, we focused on screenings in the most recent I-ELCAP experience from 2006 to 2010 and considered the implications of using higher, thus more restrictive, thresholds for the nodule size in the baseline round of screening. This round typically has the highest frequency of positive results because no previous CT scans are available for comparison and less aggressive cases of lung cancer are typically identified (6 8, 13). Such updating American College of Physicians

2 Computed Tomography Screening for Lung Original Research of the definition of a positive result is needed on a continual basis to keep the unnecessary work-up as low as possible. METHODS We drew from the database of the baseline screenings in I-ELCAP between 2006 and 2010, which were performed according to a common protocol (14). A screening was classified as baseline if no chest CT had been performed in the previous 3 years. Institutional review board approvals were obtained at each of the collaborating institutions, which include sites in large metropolitan hospitals and major academic centers as well as smaller community medical centers. Participants with 1 or more NCNs were identified according to specified criteria in the protocol (14) (Table 1). Each NCN was classified according to consistency (solid, part-solid, or nonsolid) and diameter. Diameter was the average of length and width. The current definition of a positive result of the initial, low-dose CT in the lung parenchyma in the baseline round in I-ELCAP is identification of at least 1 parenchymal, solid or part-solid NCN that is 5.0 or more mm in diameter. Participants with nodules smaller than this threshold or with only nonsolid nodules were classified as having a semipositive result and were scheduled for the first annual repeated screening 12 months later for assessment of nodule growth. A solid endobronchial NCN 5.0 or more mm in diameter also qualified as a positive result. However, such tumors are rare so we focused this article on parenchymal nodules. The diagnostic work-up for a participant with a positive result depends on the diameter of the largest NCN (14). For participants whose largest solid or part-solid NCN is 5 to 14 mm in diameter, the preferred option is to perform low-dose, noncontrast CT in 3 months. If growth at a malignant rate is identified, biopsy ideally by CT-guided fine-needle aspiration is recommended, otherwise the work-up stops with the recommendation that the participant return 12 months after the initial baseline CT for a follow-up CT. Context The threshold to define a positive result on chest computed tomography performed for lung cancer will affect the number of patients who undergo subsequent evaluation. Contribution This retrospective analysis of results from a large cohort study of lung cancer screening found that increasing the threshold for positive results from 5 mm to 7 or 8 mm would have substantially reduced the number of positive results. Delays in the diagnosis of cancer of unclear clinical significance would have occurred in a small number of patients. Caution Whether the estimated delays in cancer diagnosis would have altered patient outcomes could not be assessed. Implication Alternative thresholds for positive results on computed tomography for lung cancer screening should be evaluated prospectively. The Editors If the nodule is solid and greater than 10 mm in diameter or the solid component of a part-solid nodule is greater than 10 mm in diameter, positron emission tomography is another option. If the result of this scan is positive, biopsy is recommended; if negative or indeterminate, follow-up low-dose CT 12 months later is recommended. If the largest NCN is 15 mm or larger in diameter and its appearance is highly suggestive of lung cancer, biopsy preferably nonsurgical is an additional option. For all participants whose work-up was stopped or when the diagnostic test did not lead to a lung cancer diagnosis, a follow-up low-dose CT is recommended 12 months after the initial baseline CT. Although these are our recommendations, the decision on how to proceed is left to each par- Table 1. Characteristics of Participants Characteristic All Participants Participants With >1 NCN Participants With a Positive Result Threshold of 5.0 mm Total, n Sex, n (%) Female 8569 (41) 5190 (43) 1241 (37) Male (59) 6888 (57) 2155 (63) Median age (IQR), y 58 (53 65) 59 (54 66) 61 (55 68) Median pack-years (IQR) 26 (11 40) 27 (13 40) 27 (12 42) Ethnicity, n (%) White (86) (90) 3027 (89) Asian 1653 (8) 569 (5) 158 (5) African American 473 (2) 257 (2) 83 (2) Hispanic 524 (2) 291 (2) 92 (3) Other 204 (1) 125 (1) 36 (1) IQR interquartile range; NCN noncalcified nodule February 2013 Annals of Internal Medicine Volume 158 Number 4 247

3 Original Research Computed Tomography Screening for Lung Table 2. Frequency of a Positive Result and Lung Diagnosed Within 12 Months of Baseline Enrollment, by Size of Largest NCN* Size of Largest NCN, mm Frequency, n (%) Cases of Diagnosed, n (%) 5.0 to (75.3) 8 (0.3) 9.0 to (16.3) 26 (4.7) (8.4) 85 (29.8) Total 3396 (100.0) 119 (3.5) NCN noncalcified nodule. * Solid or part-solid. ticipant and the referring physician. The actual work-up is documented in the management system. We have gained experience with the 5.0-mm threshold, which was based on the low frequency of cancer in NCNs smaller than 5.0 mm in diameter. We then recognized that the frequency of NCNs in the 5- to 9-mm range was high, whereas the frequency of diagnosed cases of lung cancer in the first year was low (1, 7). We thus believed that higher alternative thresholds of 6.0, 7.0, 8.0, or 9.0 mm should be considered. Therefore, we obtained from the database all instances of parenchymal, solid and partsolid NCNs identified on the initial baseline low-dose CT that were 5.0 mm or larger and calculated the frequency of positive results and resulting diagnosed cases of lung cancer within 12 months of enrollment by the size of the largest NCN (Table 2). For each alternative threshold, we calculated the frequency with which the result was positive because of a solid NCN, a part-solid NCN, or a combination (Table 3). Some participants may have a positive result because they have at least 2 nodules of different consistency and diameter, each qualifying as a positive result; these participants were classified as combination. Of note, a positive result may be due to a combination of solid and part-solid NCNs in a particular participant at a given threshold. However, when the threshold is increased, the smaller NCN may no longer meet the size threshold; thus, the participant is classified according to the larger nodule, be it solid or part-solid, as solid only or part-solid only. Because of the increased size threshold, the number classified as combination would thus decrease, whereas the number of solid-only or part-solid only NCNs would increase. For example, a participant may have a part-solid NCN that is 5.0 mm in diameter and a solid NCN that is 8.0 mm in diameter. This case would be classified under combination by using the 5.0-mm threshold. If the 6.0-mm or higher threshold was used, the 5-mm part-solid NCN would no longer qualify as a positive result; thus, the case would no longer be classified as combination but as solid only. Therefore, the combination cases would be reduced by 1 and the solid-only cases would be increased by 1. We determined the frequency of positive results by using each of the alternative thresholds and the 95% CIs. The frequencies and 95% CIs with which the diagnosis of lung cancer would have been delayed until the first annual repeated screening were also determined for each alternative. All statistical analyses were performed using SAS 9.2 software (SAS Institute, Cary, North Carolina). Role of the Funding Source This work was supported by The Flight Attendant Medical Research Institute and the American Legacy Foundation. The funding sources had no role in the design, conduct, or reporting of this study or in the decision to submit the manuscript for publication. RESULTS The frequency of identifying at least 1 NCN of any size or consistency on the initial low-dose CT scan in the participants undergoing baseline screenings between 2006 and 2010 was 57% ( of ). Table 1 shows the characteristics of the study population. The frequency of positive results by using our current definition (any noncalcified, solid or part-solid NCN 5.0 mm) was 16.1% (3396 of ) (95% CI, 15.6% to 16.6%). In 2558 (75.3%) of the 3396 participants with positive results by using the current threshold, the largest NCN Table 3. Number of Participants With a Positive Result and Lung Within 1 Year, by Threshold Definition and Nodule Consistency* NCN Consistency Threshold 5.0 mm 6.0 mm 7.0 mm 8.0 mm 9.0 mm Positive Result Positive Result Positive Result Positive Result Positive Result Solid only Part-solid only Combination Total, n Total (95% CI), % 16.1 ( ) 3.5 ( ) 10.2 ( ) 5.5 ( ) 7.1 ( ) 7.5 ( ) 5.1 ( ) 10.4 ( ) 4.0 ( ) 13.2 ( ) * within 1yistabulated according to the nodules that met the threshold criterion for a positive result. Thus, a positive result may be due to a combination of solid and part-solid nodules in an individual under 1 threshold but be due only to solid or part-solid nodules under a higher threshold February 2013 Annals of Internal Medicine Volume 158 Number 4

4 Computed Tomography Screening for Lung Original Research was 5.0 mm or more but smaller than 9.0 mm in diameter, and only 8 (0.3%) of these participants were diagnosed with lung cancer within 12 months of baseline enrollment (Table 2). In 553 (16.3%) of the 3396 participants, the largest NCN was 9.0 mm or more but smaller than 15.0 mm in diameter and 26 (4.7%) of those participants were diagnosed with lung cancer within 12 months of enrollment. In 285 (8.4%) of the 3396 participants, the largest NCN was 15.0 mm or more and 85 (29.8%) of those participants were diagnosed with lung cancer within 12 months of enrollment. Thus, the frequency of a positive result was highest among participants who had NCNs between 5.0 mm and 9.0 mm in diameter. However, the frequency of diagnosing lung cancer within 12 months among these participants was very low (0.3%) (Table 2). Table 3 gives the frequency of positive results for each nodule diameter threshold of 5.0, 6.0, 7.0, 8.0, and 9.0 mm. For the threshold of 6.0 mm, the result would be positive in 10.2% (CI, 9.8% to 10.6%; 2159 of ) of the participants and would cause the work-up in the baseline year to be reduced by 36.4% (CI, 34.8% to 38.0%; 3396 to 2159 of 3396). Similarly, if thresholds of 7.0, 8.0, and 9.0 mm are used, the corresponding frequencies of a positive result would be 7.1% (CI, 6.7% to 7.4%; 1498 of ), 5.1% (CI, 4.8% to 5.4%; 1077 of ), and 4.0% (CI, 3.7% to 4.2%; 838 of ). This would cause the work-up in the baseline round of screening to be reduced by 55.9% (CI, 54.2% to 57.6%), 68.3% (CI, 66.7% to 69.9%), and 75.3% (CI, 73.9% to76.8%), respectively. Table 3 also provides the frequency of the nodule consistency for each threshold. For a 5.0-mm threshold, the result was positive on the basis of a solid NCN in 2677 (12.7%), a part-solid NCN in 555 (2.6%), or a combination of these in 164 (0.8%) of the participants. For the alternative thresholds, the respective frequencies are also shown in Table 3. Of the 3396 participants with a positive result by using the current threshold of 5.0 mm, the diagnostic work-up led to the diagnosis of lung cancer in 119 (3.5%) within a year of the initial baseline CT: 74 (2.8%) in 2677 participants with solid NCNs, 29 (5.2%) in 555 participants with part-solid NCNs, and 16 (9.8%) in 164 participants with a combination of these (Table 3). Had the higher threshold of 6.0 mm instead of 5.0 mm been used, 2159 rather than 3396 participants would have been recommended for diagnostic work-up (Table 3 and Figure). Through use of the 6.0-mm threshold, the number of cases of lung cancer diagnosed within 1 year of the initial CT scan was 119, the same number as with the 5.0-mm threshold. This was because all 119 cases of cancer diagnosed within 1 year by using the 5.0-mm threshold were in nodules that were 6.0 mm or larger: 77 (4.7%) in 1630 participants with solid NCNs, 34 (7.7%) in 442 participants with part-solid NCNs, and 8 (9.2%) in 87 participants with a combination of these. The 16 cases of Figure. Frequency of a positive result and cases of lung cancer diagnosed within 12 mo of baseline enrollment. Frequency, n Positive result Cases of cancer Minimum Nodule Diameter to Qualify for a Positive Result, mm cancer classified as combination (solid and part-solid NCNs) by using the 5-mm threshold were reduced to 8 under the 6-mm threshold, and the other 8 cases were reclassified under solid only (n 3) and part-solid only (n 5). The use of nodule diameter thresholds of , or 9.0 mm instead of 5.0 mm would have decreased the number of participants diagnosed with lung cancer within 12 months of the baseline CT by 6, 7, and 8 participants, respectively (Table 3). Thus, the number of cases of cancer diagnosed in that first year would have been reduced from 119 to 113, 112, and 111, respectively (Table 3), a reduction of 5.0% (CI, 1.1% to 9.0%), 5.9% (CI, 1.7% to 10.1%), and 6.7% (CI, 2.2% to 11.2%), respectively. Among the actual 8 cases of diagnosed lung cancer, nodule growth was observed before the invasive diagnostic procedures were performed. All were in clinical stage I before resection and subsequently found to be in pathologic stage I. None of the surgical biopsies or resections had complications, even though the nodule size was small. Changing the threshold to 7.0 mm would have resulted in a delay in the diagnostic work-up in 6 of the 8 cases (Table 4). In 5 of those 6 cases, the lung cancer was diagnosed in the NCN, which classified the result as positive, whereas in 1 case with a 6-mm solid nodule, the cancer was actually diagnosed in a larger, 12-mm nonsolid NCN. Five of the 6 cases were adenocarcinoma, mixed subtype, of which 1 was minimally invasive according to the latest adenocarcinoma classification (15); the other case was squamous cell carcinoma. In 3 of these 6 cases, surgery (lobectomy or segmentectomy) was performed 12 or more months after the baseline CT, and all were still in pathologic stage I. The other 3 had surgery 3 to 5 months after the baseline CT. Increasing the threshold from 7.0 to 8.0 mm would have resulted in an 8-month delay in the diagnostic work-up of 1 additional case of adenocarcinoma, mixed subtype. Further increasing the threshold from 8.0 to February 2013 Annals of Internal Medicine Volume 158 Number 4 249

5 Original Research Computed Tomography Screening for Lung Table 4. Diagnosed Cases of Lung That Would Have Been Delayed If the Stated Threshold Had Been Used to Define a Positive Result* Threshold and Cell Type Consistency Months to Follow-up CT Scans From Baseline Months to Biopsy Months to Surgery 7.0 mm Squamous Solid 2 and Adenocarcinoma, mixed Solid 2 and Adenocarcinoma, mixed Solid Adenocarcinoma, mixed Solid 3 and Adenocarcinoma, mixed Solid 3 and Adenocarcinoma, mixed Nonsolid mm Adenocarcinoma, mixed Solid mm Small cell Solid Diameter at Surgery, mm CT computed tomography. * Timing of the CT scans, biopsies, and surgery are from the initial CT in the baseline round. Video-assisted thoracic surgery resection of a larger nonsolid nodule (12 mm in diameter) that grew. According to the latest classification (15), it was a minimally invasive adenocarcinoma (90% lepidic pattern). Positron emission tomography/ct positive scan (standardized uptake value, 3.8) was done at the same time as the CT scan. mm would have resulted in a9month delay in the surgical treatment of a patient with a peripheral small cell lung carcinoma who had resection 6 months after baseline enrollment and received chemotherapy and radiotherapy 9 months after surgery. More than 1 additional CT was performed in 4 of the 6 cases of cancer below the 8.0-mm threshold before biopsy with further delay until surgical resection, whereas for the larger nodules, biopsy and surgery followed shortly after the follow-up CT scan. Seven of the 8 patients whose diagnoses would have been delayed using the alternative thresholds were alive at the close of follow-up, 1 at 18 months and 6 at 50 or more months after baseline enrollment. The patient with small cell carcinoma died of lung cancer 24 months after enrollment. DISCUSSION Concerns about excessive false-positive findings reported in all CT screening programs for lung cancer are ongoing. The frequency of identifying an NCN of any size on the initial baseline CT scan has almost tripled in the past decade, increasing from 23% in the 1990s (1) to the 57% observed in this article. This increase is largely due to the technological advances in CT scanners and the use of large computer monitors to display the CT images. We caution against deeming such noncancerous nodules as false-positive because these are true lung nodules and not erroneous readings of normal structures (16). Previous screening experience (4) has shown that the overwhelming number of NCNs smaller than 5.0 mm are either benign or, if malignant, too small to identify growth at a malignant rate or to perform other diagnostic tests and biopsy with our current diagnostic tools until at least 1 year later, when an increase in nodule size can be used to proceed to further diagnostic tests. Hence, the definition of positive result should focus on a size threshold for which the frequency of cancer is sufficiently high and growth on follow-up CT scans can be realistically assessed. Moreover, the definition needs to be continually reappraised to minimize potential harms of further unnecessary diagnostic work-up and costs while maximizing the frequency of early diagnosis of lung cancer. As large national programs screen millions of high-risk participants, the definition of a positive result will have enormous cost implications. For the screenings performed in I-ELCAP, using the current nodule diameter threshold of 5.0 mm, the frequency of a positive result in the baseline round was 16%. Increasing the threshold to 6.0, 7.0, 8.0, and 9.0 mm would have lowered the frequency to 10%, 7%, 5%, and 4%, respectively, thus decreasing further work-up by 36%, 56%, 68%, and 75%, respectively. For example, increasing the threshold from 5.0 mm to 8.0 mm would have decreased the number of participants who would have required follow-up CT in 3 months from 3396 to 1077, a decrease of 68.3% (CI, 66.7% to 69.9%). However, the use of the 8.0-mm threshold would have resulted in a delay in 7 cases or 5.4% (CI, 1.7% to 10.1%) of 119 patients who had been diagnosed within 12 months of the baseline screening. Clearly, any screening program needs to balance lowering the frequency of positive results with the resulting delay in the diagnosis of lung cancer, which in some cases may lead to stage progression and, thus, a decrease in curability. The NLST, using a threshold lower than I-ELCAP, defined any nodule 4.0 mm or more as a positive result, and 28% of the participants had such a result in the baseline round (9). That definition was set before the start of the trial in 2002 and was not changed during its course. In February 2013 Annals of Internal Medicine Volume 158 Number 4

6 Computed Tomography Screening for Lung Original Research comparison, during the same time, the frequency of positive results in the baseline round was 14.4% in NY-ELCAP and 13.3% in I-ELCAP (7, 8). Predictably, a more restrictive definition of a positive result leads to its lower frequency but has the unintended consequence that some cancer diagnoses are delayed. Our results suggest that changing the threshold for a parenchymal, solid or part-solid NCN to 8.0 mm may be reasonable in the baseline round of CT screening (Figure). However, the tradeoff for the increased efficiency would have been a delay in the diagnostic workup of at most 9 months in 6% of the patients diagnosed with stage I lung cancer within 1 year of the baseline screening. Would a delay of 9 months in the workup of lung cancer have allowed for stage progression and, thus, a decrease in the cancer cure rate? No direct evidence answers this critical question, but inferential information is available. We have previously shown that, in the baseline round of screening, 82% (142 of 174) of the cases of cancer were adenocarcinomas. These are generally less aggressive than squamous cell carcinomas, which were found in 9% (15 of 174) of the cases in the baseline round, and small cell carcinomas, which were found in 9% (16 of 174) (13). A review of 266 resected adenocarcinomas that were of mixed subtype and 30 mm or smaller in length in our screening program showed that 247 (93%) were in stage I (17), and these are the least aggressive adenocarcinomas. We know from the I-ELCAP experience that the average volume-doubling time of the lung cancer diagnosed in annual repeated rounds of screening is 121 days (18). Using this volume-doubling time, a typical malignant nodule identified in annual repeated rounds of screening that is 7 mm in diameter would have a diameter of 14 mm 1 year later. Because cases of cancer diagnosed in the baseline round typically have longer volume-doubling times (that is, slower growth rates) than those diagnosed in annual repeated rounds of screening, 7.0-mm malignant nodules identified at baseline would typically reach a diameter less than 14 mm after 1 year. These cases of cancer would typically still be in stage I because 91% of the screendiagnosed malignant nodules smaller than 15 mm in the baseline rounds in ELCAP and 92% in NY-ELCAP were in stage I (1, 7). In addition, the curability of these cases of stage I lung cancer is high (8), particularly for small stage IA cancer. A final source of inferential information comes from a review of the clinical findings in the 7 cases of lung cancer that were actually diagnosed using the current 5-mm threshold, whose diagnosis would have been delayed if the threshold were increased to 8 mm. All 7 had resection and pathologic stage I disease; in 3 of the cases (1 squamous cell cancer and 2 adenocarcinomas), surgery was performed 12 or more months after baseline enrollment. In the 7 cases, the diameter of the cancer in the pathology specimen ranged from 7.5 mm to 15.0 mm; all 7 participants were alive at last follow-up. Six (85%) of 7 patients were diagnosed with mixed-subtype adenocarcinoma, which is the least aggressive of the lung adenocarcinomas. Thus, it would seem unlikely that stage progression would have occurred in these 7 patients by delaying the work-up. It is, of course, possible to postulate that stage progression probably would have occurred if 1 or more of these had been a small cell carcinoma. But, as mentioned, cases of small cell cancer are typically found in only 9% of the cases of cancer in the baseline round, although they occur in 22% of the cases in annual repeated rounds of screening (13). One should appreciate that this article focuses on the first, baseline round of screening because this round generates the greatest frequency of diagnostic work-up. In the subsequent annual rounds of CT screening, the frequency of a positive result is lower; in I-ELCAP it was 5.3% (8), because only new or growing nodules, determined by comparison with the previous CT scan, require further workup. In addition, cases of lung cancer identified in annual repeated rounds are typically smaller and have a more rapid growth rate and a different cell-type distribution than those identified in the baseline round (13, 18). Therefore, the definition of a positive result for the baseline round should differ from that used for subsequent annual rounds of screening (14). A limitation of all CT screening programs is that the definition of a positive result is based solely on nodule size or volume, whereas other nodule characteristics may help distinguish a small cancer from a nonmalignant nodule. As further experience in CT screening for lung cancer accrues, some distinguishing features of cases of small-size cancer will probably emerge, possibly from imagingprocessing techniques, which will improve definition and reduce the frequency of a positive result. Our study has limitations. The implications of new thresholds for the definition of a positive result are based by necessity on retrospective review of the available data. A fuller assessment of the potential harms of delaying the diagnostic process was limited because of the small number of cancer diagnoses that were delayed. The implications of a new definition ideally should be assessed prospectively once this change is implemented, and such prospective assessment should become an integral part of every screening program. The key point of this article is that the definition of positive result needs to be continually prospectively evaluated and updated in light of emerging evidence from ongoing screening programs to reduce unnecessary surgery for nonmalignant pulmonary nodules and reduce potential harms of the diagnostic work-up, while maximizing the diagnosis and treatment of curable cases of lung cancer. From Mount Sinai Medical Center and Mount Sinai School of Medicine, New York, New York, and Weill Cornell Medical College, New York, New York. Grant Support: In part by the Flight Attendant Medical Research Institute, the American Legacy Foundation, Department of Energy (DE February 2013 Annals of Internal Medicine Volume 158 Number 4 251

7 Original Research Computed Tomography Screening for Lung FG02-96SF21260), Israel Association, The Rogers Family Fund, Yad-Hanadiv Foundation, Jacob and Malka Goldfarb Charitable Foundation, Auen/Berger Foundation, Princess Margaret Foundation, Berger Foundation, Mills Peninsula Hospital Foundation, Columbia University Medical Center, Mount Sinai Medical Center, Weill Medical College of Cornell University, Cornell University, New York Presbyterian Hospital, Swedish Hospital, Christiana Care Helen F. Graham Center, Holy Cross Hospital, Eisenhower Hospital, Jackson Memorial Hospital Health System, and Evanston Northwestern Healthcare. Potential Conflicts of Interest: Disclosures can be viewed at M Reproducible Research Statement: Study protocol: Available at Statistical code and data set: Not available. Requests for Single Reprints: Claudia I. Henschke, PhD, MD, Department of Radiology, Mount Sinai School of Medicine, 1 Gustave L. Levy Place, New York, NY 10029; , Claudia.Henschke@mountsinai.org. Current author addresses and author contributions are available at References 1. Henschke CI, McCauley DI, Yankelevitz DF, Naidich DP, McGuinness G, Miettinen OS, et al. Early Lung Action Project: overall design and findings from baseline screening. Lancet. 1999;354: [PMID: ] 2. Henschke CI, Yankelevitz DF, Smith JP, Miettinen OS; ELCAP Group. Screening for lung cancer: the early lung cancer action approach. Lung. 2002;35: [PMID: ] 3. Henschke CI, Yankelevitz DF, Mirtcheva R, McGuinness G, McCauley D, Miettinen OS; ELCAP Group. CT screening for lung cancer: frequency and significance of part-solid and nonsolid nodules. AJR Am J Roentgenol. 2002;178: [PMID: ] 4. Henschke CI, Yankelevitz DF, Naidich DP, McCauley DI, McGuinness G, Libby DM, et al. CT screening for lung cancer: suspiciousness of nodules according to size on baseline scans. Radiology. 2004;231: [PMID: ] 5. Libby DM, Wu N, Lee IJ, Farooqi A, Smith JP, Pasmantier MW, et al. CT screening for lung cancer: the value of short-term CT follow-up. Chest. 2006; 129: [PMID: ] 6. International Early Lung Action Program. International Early Lung Action Program Conferences and Consensus Statements. Accessed at on 15 June New York Early Lung Action Project Investigators. CT Screening for lung cancer: diagnoses resulting from the New York Early Lung Action Project. Radiology. 2007;243: [PMID: ] 8. Henschke CI, Yankelevitz DF, Libby DM, Pasmantier MW, Smith JP, Miettinen OS; International Early Lung Action Program Investigators. Survival of patients with stage I lung cancer detected on CT screening. N Engl J Med. 2006;355: [PMID: ] 9. Aberle DR, Adams AM, Berg CD, Black WC, Clapp JD, Fagerstrom RM, et al; National Lung Screening Trial Research Team. Reduced lung-cancer mortality with low-dose computed tomographic screening. N Engl J Med. 2011; 365: [PMID: ] 10. Harris G. CT scans cut lung cancer deaths, study finds. The New York Times. 4 November 2010: CT scans for lung cancer [Editorial]. The New York Times. 10 November 2010: A Parker-Pope T. The downside of a cancer study extolling CT scans. The New York Times. 16 November 2010: D Carter D, Vazquez M, Flieder DB, Brambilla E, Gazdar A, Noguchi M, et al; ELCAP, NY-ELCAP. Comparison of pathologic findings of baseline and annual repeat cancers diagnosed on CT screening. Lung. 2007;56: [PMID: ] 14. Henschke CI. International Early Lung Action Program: Enrollment and Screening Protocol. Accessed at on 15 June Travis WD, Brambilla E, Noguchi M, Nicholson AG, Geisinger KR, Yatabe Y, et al. International Association for the Study of Lung /American Thoracic Society/European Respiratory Society international multidisciplinary classification of lung adenocarcinoma. J Thorac Oncol. 2011;6: [PMID: ] 16. Miettinen OS. Epidemiologic Research: Terms and Concepts. 1st ed. Dortdrect, the Netherlands: Springer Publishing; Vazquez M, Carter D, Brambilla E, Gazdar A, Noguchi M, Travis WD, et al; International Early Lung Action Program Investigators. Solitary and multiple resected adenocarcinomas after CT screening for lung cancer: histopathologic features and their prognostic implications. Lung. 2009;64: [PMID: ] 18. Henschke CI, Yankelevitz DF, Yip R, Reeves AP, Farooqi A, Xu D, et al; Writing Committee for the I-ELCAP Investigators. Lung cancers diagnosed at annual CT screening: volume doubling times. Radiology. 2012;263: [PMID: ] February 2013 Annals of Internal Medicine Volume 158 Number 4

8 Annals of Internal Medicine Current Author Addresses: Drs. Henschke and Yankelevitz: Department of Radiology, Mount Sinai School of Medicine, 1 Gustave L. Levy Place, New York, NY Ms. Yip: Early Lung and Cardiac Action Program, Department of Radiology, Mount Sinai Medical Center, 1 Gustave L. Levy Place, Box 1234, New York, NY Dr. Smith: Weill Cornell Medical College, 1300 York Avenue, New York, NY Author Contributions: Conception and design: C.I. Henschke, J.P. Smith, D. Yankelevitz. Analysis and interpretation of the data: C.I. Henschke, R. Yip, D.F. Yankelevitz, J.P. Smith. Drafting of the article: C.I. Henschke, J.P. Smith, D.F. Yankelevitz. Critical revision of the article for important intellectual content: C.I. Henschke, R. Yip, J.P. Smith, D.F. Yankelevitz. Final approval of the article: C.I. Henschke, R. Yip, J.P. Smith, D. F. Yankelevitz. Provision of study materials or patients: C.I. Henschke. Statistical expertise: C.I. Henschke, R. Yip. Obtaining of funding: C.I. Henschke, D.F. Yankelevitz. Administrative, technical, or logistic support: C.I. Henschke. Collection and assembly of data: C.I. Henschke, R. Yip. APPENDIX: INTERNATIONAL EARLY LUNG CANCER ACTION PROGRAM SITES AND INVESTIGATORS Mount Sinai School of Medicine, New York, New York: Claudia I. Henschke (principal investigator), David F. Yankelevitz; Rowena Yip, Ali Farooqi, Dongming Xu. Weill Cornell Medical College, New York, New York: Dorothy I. McCauley, Mildred Chen, Daniel M. Libby, James P. Smith, Mark Pasmantier. Cornell University, Ithaca, New York: A.P. Reeves, A. Biancardi. Center for the Biology of Natural Systems, City University of New York at Queens College, Queens, New York: Steven Markowitz, Albert Miller. Fundacion Instituto Valenciano de Oncologia, Valencia, Spain: Jose Cervera Deval. Princess Margaret Hospital, Toronto, Ontario, Canada: Heidi Roberts, Demetris Patsios. Clinica Universitaria de Navarra, Pamplona, Spain: Javier Zulueta, Luis Montuenga, Maria D. Lozano, Gorka Bastarrika. Christiana Care, Helen F. Graham Center, Newark, Delaware: Thomas Bauer. National Institute Regina Elena, Rome, Italy: Salvatore Giunta, Marcello Crecco, Patrizia Pugliese. Swedish Hospital, Seattle, Washington: Ralph Aye. St. Agnes Center, Baltimore, Maryland: Enser Cole. Columbia University Medical Center, New York, New York: John H.M. Austin, Belinda M. D Souza, Gregory D.N. Pearson. Hadassah Medical Organization, Jerusalem, Israel: Dorith Shaham. Holy Cross Hospital Institute, Silver Spring, Maryland: Cheryl Aylesworth. South Nassau Communities Hospital, Long Island, New York: Shahriyour Andaz. Nebraska Methodist Hospital, Omaha, Nebraska: Patrick Meyers. Dorothy E. Schneider Center, Mills-Peninsula Health Services, San Mateo, California: Barry Sheppard. Eisenhower Lucy Curci Center, Rancho Mirage, California: Davood Vafai. ProHealth Care Regional Center, Waukesha & Oconomowoc Memorial Hospitals, Oconomowoc, Wisconsin: M. Kristin Thorsen, Richard Hansen. Wellstar Health System, Marietta, Georgia: William Mayfield. Sylvester Comprehensive Center and Jackson Memorial Hospital, University of Miami, Miami, Florida: Richard Thurer, Tammy Baxter. The 5th Affiliated Hospital of Sun Yat-Sen University, Zhuhai, China: Xueguo Liu. Shin Kong Wu Ho-Su Memorial Hospital, Taipei, Taiwan: Diana Yuwung Yeh. St. Joseph Health Center, St. Charles, Missouri: Dan Luedke. Staten Island University Hospital, Staten Island, New York: Mary Salvatore, Joseph Lowry. Hospital of the Chinese Academy of Sciences, Beijing, China: Ning Wu. Central Maine Medical Center, Lewistown, Maine: Carmine Frumiento. The Valley Hospital Center, Paramus, New Jersey: Robert Korst. Mercy Medical Center, Rockville Center, New York: Gary Herzog. Alta Bates Summit Medical Center, Berkeley, California: Gary Cecchi. Morristown/Atlantic Medical Center, Morristown, New Jersey: Mark Widmann. John Muir Institute, Concord California: Michaela Straznicka. Aurora St. Luke s Medical Center, Milwaukee, Wisconsin: David Olsen. City of Hope National Medical Center, Duarte, California: Fred Grannis, Arnold Rotter. St. Joseph s Hospital, Atlanta, Georgia: Paul Scheinberg. Evanston Northwestern Healthcare Medical Group, Evanston, Illinois: Daniel Ray. Greenwich Hospital, Greenwich, Connecticut: David Mullen. Glen Falls Hospital, Glens Falls, New York: Louis DeCunzo. Baylor University Medical Center, Dallas, Texas: Ed Cheung. Sequoia Hospital, Redwood City, California: Melissa Lim. Rush University Medical Center, Chicago, Illinois: Mark Yoder. Beth Israel Hospital Center, New York, New York: Cliff Connery. Comprehensive Center, Bend Memorial Hospital, Bend, Oregon: Albert Koch February 2013 Annals of Internal Medicine Volume 158 Number 4 W-123