Lung Cancer Risk Associated With New Solid Nodules in the National Lung Screening Trial

Cardiopulmonary Imaging Original Research Pinsky et al. Lung Cancer Risk Associated With New Nodules Cardiopulmonary Imaging Original Research Paul F. Pinsky 1 David S. Gierada 2 P. Hrudaya Nath 3 Reginald Munden 4 Pinsky PF, Gierada DS, Nath PH, Munden R Keywords: low-dose CT, lung cancer, new nodules, screening DOI:10.2214/AJR.17.18252 Received March 17, 2017; accepted after revision May 2, 2017. The opinions and assertions contained herein are the private views of the authors are not to be construed as official or as representing the views of the National Cancer Institute or the National Institutes of Health. The National Lung Screening Trial was supported by grants and contracts from the National Cancer Institute. No additional funding was obtained for this analysis. 1 Division of Cancer Prevention, National Cancer Institute, 9609 Medical Center Dr, Bethesda, MD 20892. Address correspondence to P. F. Pinsky (pp4f@nih.gov). 2 Department of Radiology, Washington University School of Medicine, St. Louis, MO. 3 Department of Radiology, University of Alabama at Birmingham School of Medicine, Birmingham, AL. 4 Department of Radiology, Wake Forest School of Medicine, Winston-Salem, NC. This article is available for credit. AJR 2017; 209:1009 1014 0361 803X/17/2095 1009 American Roentgen Ray Society Lung Cancer Risk Associated With New Solid Nodules in the National Lung Screening Trial OBJECTIVE. As low-dose CT (LDCT) lung cancer screening moves into routine clinical practice, evaluation of nodules identified as new becomes critical. We examine the frequency and clinical outcomes of new lung nodules reported at the two postbaseline annual screening examinations (hereafter referred to as postbaseline time 1 [T1] and time 2 [T2]), compared with those detected at baseline in the National Lung Screening Trial. MATERIALS AND METHODS. Radiologists classified nodules detected at T1 and T2 as new or preexisting on the basis of comparison with findings from prior LDCT screening examinations. Subjects were tracked for lung cancer incidence and mortality. We examined the incidence of new nodules and their associated lung cancer risk by nodule size (i.e., mean diameter). RESULTS. A total of 25,002 subjects underwent the baseline LDCT screening examination and either a T1 or T2 LDCT screen. At both T1 and T2, 2.6% of subjects had new solid nodules. Of the new solid nodules, 53.0% were < 6 mm, 29.5% were 6 to < 10 mm, and 17.1% were 10 mm. Lung cancer risk (defined as diagnosis within 2 years of baseline) increased from 1.1% for nodules < 4 mm to 24.0% for those 20 mm. Compared with solid nodules detected at baseline, the cancer risk was higher for new solid nodules that were 4 to < 6 mm (p < 0.001) and 6 to < 8 mm (p < 0.001) but lower for new nodules 20 mm (p = 0.03). Cancers associated with new nodules had significantly poorer survival than did those associated with baseline nodules and were significantly less likely to be adenocarcinoma. CONCLUSION. The incidence of new nodules was 2 3% annually, with the cancer risk increasing by nodule size. New nodules may convey differential lung cancer risks by size, compared with baseline nodules. T he National Lung Screening Trial (NLST) showed that screening for lung cancer with lowdose CT (LDCT) reduces death from lung cancer in high-risk current and former smokers [1]. As screening moves from the initial research phase, when the emphasis was disproportionately on baseline screening, to population-based screening, which is recommended annually for up to 25 years, the preponderance of screens in clinical practice going forward likely will be postbaseline screens. Such screens would be expected to have a prior image available for comparison; in a large study of mammography, more than 90% of postbaseline screens had a comparison image available [2]. Comparison images may not be available for baseline screens, and therefore the recent growth history of nodules detected on baseline screens can be unknown. Nodules found to be new on postbaseline screens after com- parison with an earlier image must have grown in the prior year, and thus they may represent a higher risk for lung cancer. Recently, the Dutch-Belgian Randomized Lung Cancer Screening Trial (Dutch acronym: NELSON) reported new solid nodules discovered at two postbaseline screens of approximately 8000 subjects [3]. Several other studies have also reported on the presence of and risk associated with new nodules, including the Mayo Clinic study and the International Early Lung Cancer Action Program [4, 5]. These studies reported rates of 5 13% for the detection of new nodules per annual postbaseline screen [3 5]. Overall lung cancer risks for new nodules were generally approximately 5%. When analyzed by new nodule size, as in the NELSON trial, the cancer risk increased substantially with nodule size [3]. The NLST enrolled more than 25,000 subjects in the LDCT arm; subjects underwent three screening rounds (one at baseline and AJR:209, November 2017 1009

Pinsky et al. Baseline Screen T1 Screen T2 Screen 26,309 Subjects 25,002 had T1 screen, T2 screen, or both screens 24,604 Subjects 23,972 Subjects two after baseline) [1]. Although the findings from these two postbaseline screening rounds have been previously published, the analyses did not distinguish between preexisting and new nodules [6]. In this study, we analyze findings from the NLST with respect to new solid nodules. Specifically, we examine the frequency and size distribution of new solid nodules and the cancer risk associated with them, as stratified by nodule size. Among cancers linked to new nodules, we examine the relationship between nodule size and stage, histologic findings, and survival. In addition, we compare new nodules with nodules detected at baseline, with respect to nodule size and cancer risk. Baseline Nodules 5990 Subjects 8811 Nodules New Nodules 644 Subjects 832 Nodules New Nodules 628 Subjects 822 Nodules 238 Linked Cancers 48 Linked Cancers 46 Linked Cancers Fig. 1 Flowchart analysis of nodules detected at baseline low-dose CT screening examination and new nodules detected at two annual postbaseline screening examinations performed at time 1 (T1) and time 2 (T2). Materials and Methods National Lung Screening Trial Design The NLST randomized subjects to undergo either LDCT or chest radiography as a screening examination. Eligibility criteria included a history of smoking ( 30 pack-years) and current smoking status or having quit smoking in the past 15 years [5]. Subjects aged 55 74 years were recruited at 33 U.S. centers during 2002 2004 and underwent either LDCT or chest radiography screens at baseline and annually for two more years after baseline (denoted as time 1 [T1] and time 2 [T2]). The NLST was approved by the institutional review board at each participating screening center, and all subjects provided informed consent. No institutional review board approval was required for this secondary analysis. The study protocol defined a noncalcified nodule with a longest diameter of 4 mm as a positive screening examination result. For each such noncalcified nodule, radiologists used standardized forms to report location (i.e., lung lobe), greatest transverse and perpendicular diameter, margins, and attenuation characteristics. At T1 and T2, they also reported, on the basis of examination of prior images, whether the nodule was preexisting or new. Other abnormalities, including adenopathy or effusion, could also trigger positive screening results. Radiologists also recorded recommendations for diagnostic follow-up, which included LDCT performed at various time intervals (at 3, 3 6, 6, 12, and 24 months), diagnostic CT, CT densitometry, PET, and biopsy. Positive screening results were tracked for resultant diagnostic procedures and lung cancer diagnoses. In addition, subjects were followed by means of annual surveys to ascertain incident cancers and all-cause mortality. All reported cancers were verified with medical records, with stage and histologic findings recorded. Follow-up continued until December 31, 2009. Quantitative Methods The analysis included all subjects in the LDCT arm of the trial who underwent screening LDCT at baseline and at least one subsequent screening LDCT examination. The rate of new solid nodules noted at each postbaseline screen was computed as the proportion of all subjects undergoing the screen that had one or more new solid nodules reported. To assess the lung cancer risk associated with new (solid) nodules, the lobe location of the nodule and cancer was analyzed. Specifically, any lung cancer diagnosed within 2 years of the screen at which a new (solid) nodule was first identified and found in the same lung lobe as that new nodule was ascribed (i.e., linked) to that nodule. Note that the NLST did not attempt to link cancers back to specific nodules as part of the trial findings. Cancer risk was analyzed by new nodule size category as measured by mean diameter (the mean of the longest diameter and longest perpendicular values in the transverse plane), which is more relevant to current screening practice than the longest diameter criterion used to define a positive screen result in the NLST. For multiple new nodules in the same lobe, any cancer in the lobe was ascribed to the largest new nodule in that lobe. To isolate the effects of a given new nodule on cancer risk, we performed an analysis restricted to new nodules with no other new or preexisting (solid) nodules in the same lobe (referred to as solitary in-lobe nodules ). In addition to nodule size, we also examined nodule margins with respect to cancer risk. Because margins were correlated with nodule size, we stratified nodules into small (< 10 mm) and large ( 10 mm) categories to examine risk in this analysis. For new nodules detected at the T1 screen, we analyzed the findings at the T2 screen to assess their persistence. Specifically, if a nodule reported at the T2 screen was found in the same lobe as the new nodule reported at the T1 screen and was classified as preexisting, the new nodule was denoted as persistent; otherwise, it was deemed not persistent. To compare the cancer risk of new versus baseline (solid) nodules, we compared the proportion of nodules of each type that were linked to lung cancer by size category. Linked cancers were defined similarly for baseline nodules and new nodules (i.e., they were diagnosed within 2 years and in the same lobe). For lung cancers linked to new solid nodules, we examined the relationship of nodule size to cancer characteristics, including histologic findings, stage, and survival. Statistical significance was assessed using a p value for trend over quartiles of nodule size. We also compared the histologic findings and stage and survival information for cancers linked to new nodules versus those linked to baseline nodules. Lung cancer specific survival was assessed using Kaplan-Meier analysis [7]. Results Figure 1 shows a flowchart of the analysis. A total of 25,002 subjects in the LDCT arm of the trial were included in the analysis of new nodules on the basis of having undergone a baseline screen and a T1 screen, a T2 screen, or both postbaseline screens. Of these subjects, 24,604 (98.4%) underwent a T1 screen, 23,972 (95.9%) underwent a T2 screen, and 23,574 (94.3%) underwent both screens. The median age of the subjects was 60 years (25th percentile, 57 years; 75th percentile, 65 years), 59% were men, and 47% were current smokers. 1010 AJR:209, November 2017

Lung Cancer Risk Associated With New Nodules TABLE 1: New Solid Nodules Detected at Two Annual Postbaseline Screening Examinations Performed at Time 1 (T1) and Time 2 (T2) No. of New Solid Nodules Detected at T1 Screen (n = 24,604) Table 1 shows the number of subjects with new solid nodules at the T1 and T2 screens (note that all references to nodules refer to solid nodules only, unless otherwise specified). The proportion of subjects with at least one new nodule was 2.6% at the T1 screen, and it was also 2.6% at the T2 screen; 66 subjects (0.3%) had new nodules at both the T1 and T2 screens. For comparison, at baseline, 21.3% of all subjects had at least one nodule. Of 644 subjects with new nodules at the T1 screen, 217 (33.7%) also had preexisting nodules (located anywhere in the lung, although not necessarily in the same lobe). Of 628 subjects with new nodules at the T2 screen, 217 (34.6%) also had preexisting nodules (located anywhere in the lung). Table 2 shows lung cancer risk by new nodule size. For T1 and T2 screens combined, Detected at T2 Screen (n = 23,972) Detected at T1 and T2 Screens (n = 25,002) a 0 23,960 (97.4) 23,344 (97.4) 23,796 (95.2) 1 524 (2.1) 506 (2.1) 931 (3.7) 2 78 (0.3) 80 (0.3) 177 (0.7) 3 24 (0.1) 24 (0.1) 53 (0.2) 4 18 (0.1) 18 (0.1) 45 (0.2) 1 644 (2.6) 628 (2.6) 1206 (4.8) Total 832 822 1654 Note Data are the number or number (%) of new solid nodules detected. a Nodules detected at T1 and T2 combined; all subjects who underwent screening at either T1 or T2 were included. 53.0% of new nodules were size < 6 mm, 29.5% were 6 to < 10 mm, and 17.1% were 10 mm. Ninety-four cancers were linked to new nodules in 92 distinct individuals. The cancer rate increased monotonically from 1.1% for nodules < 4 mm to 24.0% for nodules 20 mm. Note that because the NLST did not record the location or preexistence status of nodules < 4 mm in longest diameter, the mean diameter category of < 4 mm in the present study is restricted to nodules 4 mm in longest diameter. For the analysis restricted to solitary in-lobe new nodules (i.e., no other new or preexisting nodules were found in the same lobe), the results were similar (with a cancer risk of 1.7% noted for nodules < 4 mm to 20.0% for nodules of 20 mm). Note that 1183 of 1654 new nodules (70.2%) were solitary in-lobe nodules. With respect to new nodule margins, 52.8% were smooth, 19.6% were spiculated, 26.5% were poorly defined, and 1.0% were not classified. New nodules with spiculated margins were more likely to be 10 mm (33.2%) than were nodules with smooth (10.4%) or poorly defined (18.5%) margins. For nodules < 10 mm, the cancer rate was significantly higher for nodules with spiculated margins (7.8%) than for nodules with smooth margins (3.2%) (p = 0.005). In contrast, for nodules 10 mm, the cancer rate was not significantly different for those with spiculated margins (17.6%) than for those with smooth margins (23.1%) (p = 0.34). New nodules with poorly defined margins had lower cancer rates both for nodules < 10 mm (1.1%) and those 10 mm (8.6%), compared with nodules with smooth or spiculated margins. Of 832 new nodules noted at the T1 screen, 717 were found in subjects who underwent a T2 screen. Of these 717 nodules, 211 (29.4%) were deemed persistent, including 11 that were linked to subsequent cancers. The other 115 nodules were found in subjects who did not undergo the T2 screen and included 37 nodules linked to cancers. Table 3 shows the relationship between nodule size and cancer characteristics for cancers linked to new nodules. The proportion with adenocarcinoma significantly decreased in association with the nodule size quartile, from 47% with adenocarcinoma for the lowest quartile (nodule size, < 7 mm) to 12% for the highest quartile (nodule size, 18 mm). A borderline significant asso- TABLE 2: Cancer Risk According to the Size of New Solid Nodules Detected at Two Annual Postbaseline Screening Examinations Performed at Time 1 (T1) and Time 2 (T2) Mean Nodule Diameter (mm) Nodules Detected at T1 Screen Nodules Detected at T2 Screen Nodules Detected At T1 and T2 Screens Solitary In-Lobe Nodules a Detected at T1 and T2 Screens All Linked to Cancer All Linked to Cancer All Linked to Cancer All Linked to Cancer < 4 b 82 (9.9) 0 (0) 98 (11.9) 2 (2.0) 180 (10.9) 2 (1.1) 120 (10.1) 2 (1.7) 4 to < 6 342 (41.1) 9 (2.6) 355 (43.2) 7 (2.0) 697 (42.1) 16 (2.3) 483 (40.8) 16 (3.3) 6 to < 8 160 (19.2) 11 (6.9) 158 (19.2) 6 (3.8) 318 (19.2) 17 (5.4) 234 (19.8) 13 (5.6) 8 to < 10 87 (10.4) 6 (6.9) 84 (10.2) 5 (6.0) 171 (10.3) 11 (6.4) 124 (10.4) 8 (6.5) 10 to < 15 96 (11.5) 11 (11.5) 64 (7.8) 11 (17.2) 160 (9.7) 22 (13.8) 123 (10.4) 13 (10.6) 15 to < 20 39 (4.7) 7 (17.9) 33 (4.0) 7 (21.1) 72 (4.4) 14 (19.4) 55 (4.6) 11 (20.0) 20 21 (2.5) 4 (19.0) 29 (3.5) 8 (27.6) 50 (3.0) 12 (24.0) 40 (3.4) 8 (20.0) Unknown 5 (0.6) 0 (0) 1 (0.1) 0 (0) 6 (0.4) 0 (0) 4 (0.3) 0 (0) All 832 (100) 48 (5.8) 822 (100) 46 (5.4) 1654 (100) 94 (5.7) 1183 (100) 71 (6.0) Note Data are number (%) of nodules. Two subjects had two nodules that each were linked to a lung cancer (different diagnosis dates and histologic findings were noted in both instances), so only 92 distinct subjects had lung cancer linked to new nodules. a Excludes nodules with other new or preexisting solid nodule in the same lobe. b Longest diameter had to be 4 mm. AJR:209, November 2017 1011

Pinsky et al. TABLE 3: Mean Nodule Diameter and Cancer Characteristics of Lung Cancers Linked to New Solid Nodules Quartile of Mean Nodule Diameter Among Lung Cancer Cases Histologic Finding ciation with cancer stage was found, with the proportion of stage I cancers decreasing with nodule size (p = 0.046). No association was seen between nodule size and death from lung cancer (p = 0.71). Of 26,309 subjects who underwent the baseline screen, 5590 (21.3%) had at least one nodule (Fig. 1). Table 4 compares new nodules with baseline nodules on the basis of size and cancer risk. For large nodules (mean diameter, 20 mm), the cancer risk was significantly greater for nodules detected at baseline compared with new nodules (p = 0.03). For smaller nodules (those with a mean diameter of < 4, 4 to < 6 mm, or 6 to < 8 mm), however, the risk was significantly higher for new nodules compared with baseline nodules (p = 0.01 for < 4 mm and p < 0.001 for 4 to < 6 mm and 6 to < 8 mm). First (< 7 mm) (n = 19) Second (7 10.9 mm) (n = 23) Third (11 17.9 mm) (n = 24) A total of 237 subjects had cancers linked to baseline nodules. For six of these subjects, the cancer was also linked to a new nodule. The analysis comparing these groups excluded these six subjects, leaving 86 and 231 subjects with cancers linked to new and baseline nodules, respectively. Cancers linked to new nodules were significantly less likely to be adenocarcinoma (25.6% vs 61.0%, p < 0.001) and were more likely to be small cell lung cancer (17.4% vs 3.9%, p < 0.001) than were those linked to baseline nodules. The cancer stage distribution was similar between the two groups; 57.0% of cancers linked to new nodules were in stage I versus 61.0% of cancers linked to baseline nodules. Fig. 2 compares lung cancer specific survival in the two groups. Cancers linked to nodules detected at baseline Fourth ( 18 mm) (n = 26) All a (n = 92) Adenocarcinoma c 9 (47) 8 (35) 5 (21) 3 (12) 25 (27) 0.006 e Other non small cell lung cancers d 8 (42) 14 (61) 14 (58) 15 (58) 51 (55) 0.40 e Small cell lung cancer 2 (11) 1 (4) 5 (21) 7 (27) 15 (16) 0.06 e Stage I cancer f 14 (74) 14 (61) 16 (70) 10 (40) 54 (60) 0.046 Lung cancer death 8 (42) 11 (48) 10 (42) 13 (50) 42 (46) 0.71 Note Except where otherwise indicated, data are number (%) of lung cancer cases. a Ninety-two subjects had a cancer linked to a new nodule. For the two subjects with two lung cancers that were each linked to a new nodule, the cancer with the earliest diagnosis was used in the analysis. b For trend. c Includes bronchioloalveolar adenocarcinoma. d Includes squamous cell carcinoma, large cell cancer, other non small cell lung cancer, and non small cell lung cancer not otherwise specified. e For comparison of this histologic finding with all other histologic findings. f One subject in the third and one subject in the fourth quartile had unknown stages; percentages exclude these subjects. TABLE 4: Size and Cancer Risk of New Nodules Detected at Two Annual Postbaseline Screening Examinations at Time 1 (T1) and Time 2 (T2) Versus Those of Solid Nodules Detected at Baseline Mean Diameter of Largest New Nodules (mm) No. (%) of New Nodules Detected at T1 and T2 Screens No. (%) of Nodules Detected at Baseline Screen All With Cancer All With Cancer < 4 b 180 (10.9) 2 (1.1) 1029 (11.7) 1 (0.1) 0.01 4 to < 6 697 (42.1) 16 (2.3) 4693 (53.3) 17 (0.4) < 0.001 6 to < 8 318 (19.2) 17 (5.4) 1728 (19.6) 23 (1.3) < 0.001 8 to < 10 171 (10.3) 11 (6.4) 548 (6.2) 32 (5.8) 0.78 10 to < 15 160 (9.7) 22 (13.8) 477 (5.4) 58 (12.2) 0.60 15 to < 20 72 (4.4) 14 (19.4) 160 (1.8) 39 (24.4) 0.41 20 50 (3.0) 12 (24.0) 166 (1.8) 68 (41.0) 0.03 Unknown 6 (0.4) 0 (0) 10 (0.1) 0 (0.0) All 1654 (100) 94 (5.7) 8811 (100) 238 (2.7) < 0.001 Note One subject had two baseline nodules that were each linked to a cancer, so 237 subjects had a cancer linked to a baseline nodule. Ninety-two subjects had a cancer linked to a new nodule. Dash ( ) denotes not applicable. a For comparison of cancer risk for nodules detected at baseline versus new nodules detected at T1 or T2. b Longest diameter had to be 4 mm. had significantly improved survival compared with those linked to new nodules, with 5-year survival rates of 66.6% versus 51.1% (p = 0.001). When small cell lung cancer was excluded, the results were similar (5-year survival, 68.8% vs 55.7%, p = 0.01). Discussion In this analysis of the NLST, we found that 2.6% of subjects had a new solid nodule reported at each postbaseline screening examination. The lung cancer risk increased markedly with new nodule size, with rates of 1.1% noted for nodules with a mean size of < 4 mm versus 24.0% for those with a mean size of 20 mm. The results of this analysis are not directly comparable to those of the NELSON trial because the NELSON trial classified new p a p b 1012 AJR:209, November 2017

Lung Cancer Risk Associated With New Nodules (solid) nodules according to volume, whereas nodules were classified by mean diameter in the present analysis. Nonetheless, the overall lung cancer rate (5.7%) was similar to that found in the NELSON trial (4.2%); note that each study categorized outcome as a lung cancer diagnosis within 2 years of detection of a new nodule [3]. In addition, in each study the lung cancer risk sharply increased with size, regardless of how it was measured. An approximate direct comparison of risks determined on the basis of nodule size category in the two studies can be produced, under the assumption that nodules are spherical (i.e., volume = [π / 6]D 3 0.52D 3 where D denotes nodule diameter). Therefore, the highest volume category in the NELSON trial ( 500 mm 3 ) would correspond to a mean diameter of 10 mm; the associated cancer risks were 25% in the NEL- SON trial and 17% in the NLST. Likewise, the NELSON trial volume category of < 200 mm 3 corresponds to a mean diameter of approximately < 8 mm, with associated cancer risks of 2.0% in the NELSON trial and 2.9% in the present analysis. In the NELSON trial, the new nodule rate per screen (5.9%) was approximately twice the rate of 2.6% noted here [3]. This difference in part could be due to differences in defining a nodule, where the NELSON trial used a volume threshold of 15 mm 3 for defining a new nodule and the NLST used a longest diameter threshold of 4 mm. An earlier Mayo Clinic study of approximately 1500 subjects found a 13% rate of new nodules on the first incidence scan (8.3% after excluding nodules < 4 mm) [5]. The corresponding lung cancer rate was 1.5% (3/191). The International Early Lung Cancer Action Program study, which included approximately 27,500 postbaseline screens, reported that 5.3% of such screens had new nodules (including those < 4 mm); of these, 5.1% had a subsequent lung cancer diagnosis [4]. Of interest, when new nodules were compared with baseline nodules, small size (4 6 and 6 8 mm) was associated with a higher cancer risk for new nodules, but the reverse was true for large size ( 20 mm), with the risk higher for baseline nodules. For new nodules to reach a large size, growth must be rapid, which might be more likely with an infectious or inflammatory process than with malignancy. This may partially explain why the lung cancer risk was lower for new nodules than for baseline nodules among large ( 20 mm) nodules. Smaller (< 6 mm) nodules Fig. 2 Lung cancer specific survival rate for subjects with cancers linked to new (solid) nodules (solid line) versus those with cancers linked to baseline (solid) nodules (dashed line). 90 80 70 60 50 40 30 20 10 Lung Cancer Specific Survival Rate (%)100 0 0 1 2 3 4 5 6 7 No. of Years From Diagnosis are common at baseline and likely have been stable for years; they thus have a low risk of malignancy. In contrast, small new nodules must have grown recently but at a slower rate than large new nodules and at a rate more consistent with malignancy. Among new nodules linked to subsequent lung cancer, larger nodules were less likely to have histologic findings of adenocarcinoma. A similar finding was reported in the NELSON trial for new nodules [3]. Because, for new nodules, size is highly correlated with recent growth rates, this is consistent with current evidence that adenocarcinomas tend to be slower growing than other lung cancer histologic findings [9]. However, no significant association was observed between new nodule size and lung cancer death, and only a borderline significant relationship with stage was noted. With a presumed lead time resulting from early detection, the follow-up time was not that long (5 6 years), and continued follow-up for death may yet elucidate survival differences by nodule size. When we compared characteristics of lung cancers associated with new nodules versus baseline nodules, the latter were significantly more likely to be adenocarcinoma and to have significantly improved survival than the former. This is consistent with the idea of cancerous nodules detected at baseline being, on average, slower growing than new cancerous nodules. For solid new nodules, the Lung CT Screening Reporting and Data System (Lung-RADS) assigns categories 3, 4A, and 4B for mean nodule sizes of 4 to < 6 mm, 6 to < 8 mm, and 8 mm, respectively [9]. The cancer risks observed in the present study (per nodule) for new nodules of these sizes were 2.3%, 5.4%, and 13.0%, which shows that the increasing intensity of recommended diagnostic follow-up for these categories does track with the increasing risk for new solid nodules. For the 4B category (nodules 8 mm), the risk was actually similar to that for new nodules 8 to < 10 mm (6.4%) as for those 6 to < 8 mm (category 4A), with risk increasing only for new nodules of 10 mm. This suggests that 10 mm, rather than 8 mm, may be a more appropriate cutoff between Lung-RADS categories 4A and 4B for new nodules, although more evidence is needed. A limitation of this analysis is that cancers could not be definitively linked to specific nodules. Therefore, linkage had to be assumed on the basis of lobe location. In addition, nodule volume was not measured, making direct comparison with findings from European studies difficult. Results also may have been affected by the 2-year follow-up limit for cancer diagnosis; if any new nodules were monitored for longer than 2 years or were lost to follow-up before cancer was diagnosed, the cancer rate would be underestimated in this study. Also, for those cancers linked to new nodules reported at the last NLST screen, it is possible that the cancer could have arisen from an entirely new nodule arising after this last screen. However, because the annual rate of new nodule formation was only 2.6%, this likely was a rare occurrence. Conclusion The present study showed an incidence of new solid nodules of 2.6% annually. The cancer risk associated with new nodules increased with nodule size and differed from that of baseline nodules of the same size. References 1. The National Lung Screening Trial Research Team. Reduced lung-cancer mortality with low- AJR:209, November 2017 1013

Pinsky et al. dose computed tomographic screening. N Engl J Med 2011; 365:395 409 2. Hubbard RA, Kerlikowske K, Flowers CI, et al. Cumulative probability of false-positive recall or biopsy recommendation after 10 years of screening mammography. Ann Intern Med 2011; 155:481 492 3. Walter JE, Heuvelmans MA, de Jong PA, et al. Occurrence and lung cancer probability of new solid nodules at incidence screening with lowdose CT: analysis of data from the randomized, controlled NELSON trial. Lancet Oncol 2016; 17:907 916 4. The International Early Lung Cancer Action Program Investigators. Survival of patients with stage I lung cancer detected on CT screening. N Engl J Med 2006; 355:1763 1771 5. Swensen SJ, Jett JR, Sloan JA, et al. Screening for lung cancer with low-dose spiral computed tomography. Am J Respir Crit Care Med 2002; 165:508 513 6. Aberle DR, DeMello S, Berg CD, et al. Results of the two incidence screenings in the National Lung Screening Trial. N Engl J Med 2013; 369:920 931 7. Kaplan EL, Meier P. Nonparametric estimation from incomplete observations. J Am Stat Assoc 1958; 53:457 481 8. West H, Harpole D, Travis W. Histologic considerations for individualized systemic therapy approaches for the management of non-small cell lung cancer. Chest 2009; 136:1112 1118 9. American College of Radiology (ACR). Lung CT Screening Reporting and Data System (Lung-RADS ). ACR website. www.acr.org/ Quality-Safety/Resources/LungRADS. Accessed August 18, 2016 FOR YOUR INFORMATION This article is available for CME and Self-Assessment (SA-CME) credit that satisfies Part II requirements for maintenance of certification (MOC). To access the examination for this article, follow the prompts. 1014 AJR:209, November 2017