Long-term cancer risk associated with lung nodules observed on low-dose screening CT scans

Journal Pre-proof Long-term Cancer Risk Associated with Lung Nodules Observed on Low-Dose Screening CT Scans Paul Pinsky (Conceptualization) (Methodology) (Formal analysis) (Investigation) (Resources)Writing–original draft)Writing–review and editing), David S. Gierada (Conceptualization) (Investigation) (Resources)Writing–review and editing)

PII:

S0169-5002(19)30742-1

DOI:

https://doi.org/10.1016/j.lungcan.2019.11.017

Reference:

LUNG 6208

To appear in:

Lung Cancer

Received Date:

30 August 2019

Revised Date:

15 November 2019

Accepted Date:

22 November 2019

Please cite this article as: Pinsky P, Gierada DS, Long-term Cancer Risk Associated with Lung Nodules Observed on Low-Dose Screening CT Scans, Lung Cancer (2019), doi: https://doi.org/10.1016/j.lungcan.2019.11.017

This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. © 2019 Published by Elsevier.

Long-term Cancer Risk Associated with Lung Nodules Observed on Low-Dose Screening CT Scans

Paul Pinsky, Ph.D. 1, David S. Gierada, M.D. 2

of

1. Division of Cancer Prevention, National Cancer Institute, Bethesda, MD. 2. Washington University School of Medicine, St. Louis, MO.

*

ro

Correspondence: Paul F. Pinsky, National Cancer Institute, 9609 Medical Center Dr., Bethesda, MD, 20892; e-mail: [email protected]; phone: 240-276-7014.

-p

The authors have no competing interests to declare.



Jo

ur na



The presence of nodules on low-dose CT screening was associated with increased risk of lung cancer up to 12 years later Lung cancers diagnosed even more than 4 years after nodule detection tended to occur in the same lung lobe as the nodule Long-term lung cancer risk differed based on the size and attenuation of nodules

lP



re

Highlights

Abstract Objective Non-calcified nodules (NCNs) associated with false positive low-dose CT (LDCT) lung cancer screens have been attributed to various causes. Some, however, may represent lung cancer precursors. An association of NCNs with long-term lung cancer risk would provide indirect evidence of some NCNs being cancer precursors. Methods

-p

ro

of

LDCT arm participants in the National Lung Screening Trial (NLST) received LDCT screens at baseline and years 1-2. The relationship between NCNs found on LDCT screens and subsequent lung cancer diagnosis over different time periods was examined at the person and lobe level. For the latter, a lobe had a cancer outcome only if the cancer was located in the lobe. Separate analyses were performed on baseline and post-baseline LDCT findings; for the latter, those with baseline NCNs were excluded and only new (non-pre-existing) NCNs examined. Raw and adjusted rate-ratios (RRs) were computed for presence of NCNs and subsequent lung cancer risk; adjusted RRs controlled for demographic and smoking factors. Results

lP

re

26309 participants received the baseline LDCT screen. Over median 11.3 years follow-up, 1675 lung cancers were diagnosed. Adjusted RRs for time periods 0-4, 4-8 and 8-12 years following the baseline screen were 5.1 (95% CI:4.4-5.9), 1.5 (95% CI:1.3-1.9) and 1.5 (95% CI:1.2-1.8) at the person-level and 14.7 (95% CI:12.6-17.2), 2.6 (95% CI: 2.0-3.4) and 2.2 (95% CI:1.6-2.9) at the lobe-level. 18585 participants were included in the post-baseline analysis. Adjusted RRs for periods 0-4, 4-8 and 8-11 years were 5.6 (95% CI: 4.5-7.0), 1.9 (95% CI: 1.3-2.7) and 1.6 (95% CI: 0.9-2.9) at the person-level and 19.6 (95% CI:14.9-25.3), 2.5 (95% CI:1.3-4.7) and 3.3 (95% CI:1.4-7.6) at the lobe-level. Raw RRs were similar.

ur na

Conclusion

NCNs are associated with excess long-term lung cancer risk, suggesting that some may be lung cancer precursors.

Jo

Keywords: lung cancer; non-calcified nodule; low-dose CT; screening; long-term risk

Introduction

Non-calcified nodules (NCNs) associated with false positive low-dose CT (LDCT) screens for lung cancer have been attributed to various causes, including infectious and inflammatory processes 1-2. There is also evidence that some NCNs may represent lung cancer precursors. This evidence comes from case series of surgically resected NCNs, where some lesions have been shown to represent atypical

of

adenomatous hyperplasia, a lung adenocarcinoma precursor 3-4. Further indirect support for this

ro

hypothesis comes from a study of participants in the National Lung Screening Trial (NLST) who had NCNs found on LDCT screening 5. Participants with baseline NCNs had significantly higher lung cancer

-p

incidence in the period 5-7 years post-screening controlling for standard lung cancer risk factors. Additionally, the excess risk was location-specific, as the location of the lung cancer was correlated with

re

the location of the baseline NCN 5.

lP

In this paper, we extend the above analysis of NLST LDCT arm participants in several ways. First, we extend the follow-up time for baseline nodules to up to 12 years. Second, we also assess long-term risk

ur na

associated with new (non-pre-existing) nodules reported on post-baseline screens. Finally, we examine lung cancer histology in relation to NCN status and time period.

Jo

Methods

NLST Design

The design of NLST has been reported previously 6,7. Briefly, men and women aged 55-74 years

with at least 30 pack-years of cigarette smoking and who were either current smokers or had quit within the past 15 years were enrolled at 33 medical institutions across the U.S. between 2002 and 2004. Exclusion criteria included previous lung cancer diagnosis, a CT scan in the prior 18 months, unexplained

weight loss in the year before enrollment, or hemoptysis. Participants were randomized into a LDCT or chest radiograph (CXR) arm, with 3 annual protocol screens (T0/T1/T2) for each modality. Participants were actively followed for lung cancer incidence and all-cause mortality until December 31, 2009. During this time, medical records were abstracted for those with a positive screening test or lung cancer diagnosis. Vital status was assessed through periodic questionnaires and by

study and each person provided written consent to participate in the study.

of

linkage with the National Death Index (NDI). Institutional review boards at each center approved the

ro

After the active follow-up period, participants were followed passively through linkages with

-p

state cancer registries and the NDI 8. Among the screening centers, 22 of the 33, comprising 88% of NLST participants, participated in the state registry linkages. Lung cancer incidence follow-up was

re

through the end of 2014 for participating centers; otherwise, it was through the end of 2009. Mortality

ur na

2009 otherwise.

lP

follow-up was through the end of 2015 for centers (32 of 33) with NDI linkage and through the end of

Quantitative Methods

We performed separate analyses for baseline (T0) and post-baseline (T1/T2) nodules; for each,

Jo

assessments were performed at both the person and lung lobe level (Figure 1). For the analysis of baseline nodules, all those receiving the baseline CT screen were included in the person-level analyses and all 5 lobes for these participants were included in the lobe-level analysis. Participants were classified as either having a NCN or not; lobes were similarly classified as to whether an NCN was present in the given lobe. Follow-up time began at the time of the baseline screen.

The post-baseline analysis examined new (non-pre-existing) NCNs reported on post-baseline screens among those completing at least one such screen (and the baseline screen) (Figure 1). To isolate the effect of new NCNs from that of persisting baseline NCNs, participants were excluded from this analysis if they had an NCN at baseline. For the person-level analysis, follow-up began at T1 if the participant had a new nodule at T1, otherwise it started at the last screen (generally T2). All 5 lobes were included in the lobe-level analysis; follow-up began at T1 if there was a new nodule in the lobe at

of

T1, otherwise follow-up began at the last screen. Note this design insures that follow-up does not begin

ro

until the status of the participant (lobe) as to whether a new NCN is present is determined.

For both the baseline and post-baseline analyses, follow-up ended at date of death or the end of

re

cancer diagnosis date for those with lung cancer.

-p

incidence follow-up, whichever came first, for participants without a lung cancer diagnosis and at lung

We computed raw lung cancer rate ratios (RRs) comparing participants (lobes) with NCNs to those

lP

without for three follow-up time periods of 0-4, 4-8 and 8-12 years (8-11 years for the post-baseline analysis). RRs were defined as events divided by person-years (PYs) or lobe-years (LYs) of follow-up. For

ur na

the lobe-level analysis, only lung cancers located in the given lobe were counted as lung cancer events; lobes were censored at the time of a lung cancer not located in the given lobe (or of unknown location). All lung cancers were counted as events for the person-level analysis. In addition to raw RRs, we used Poisson regression to compute adjusted RRs. RRs were adjusted for age, sex, current smoking status,

Jo

pack years (<50 vs 50+), self-reported COPD and lobe location (lobe-level analysis). Proc Genmod (SAS version 9.4) was used with a repeated statement to account for correlations among lobes within persons for the lobe-level analysis. Among lobes with NCNs, for the lobe-level analysis we also computed modeled RRs for NCN characteristics, specifically size (average of longest diameter and perpendicular) and attenuation. Since lobes could have multiple NCNs, the factors examined were presence in the lobe

of at least one large NCN (size ≥ 8mm) and at least one solid NCN. The models controlled for age, sex, smoking status, pack-years and lobe location. To further examine the correlation of NCN location and tumor location, we ran the above lobe-level Poisson regression model restricted to subjects with NCNs (for the baseline analysis). Assuming the person-level analysis demonstrated that presence of an NCN was a risk factor generally for lung cancer,

of

this restricted lobe-level analysis would only show a significant association of NCNs with cancer if the eventual tumor location was correlated with NCN location.

ro

Finally, to assess the effect of short-term nodule stability on long-term lung cancer risk, we examined

-p

lung cancer outcomes as a function of baseline nodule follow-up status at the year 2 screen. Specifically, for all lobes with a baseline NCN and a year 2 screen and in which there were no reported

re

new nodules at year 1 or 2, we categorized their status at year 2 as follows: 1) no reported NCNs in lobe, 2) reported pre-existing NCNs in lobe, all stable (no growth), and 3) reported pre-existing NCNs in lobe,

lP

at least one growing. We examined adjusted RRs for these three categories of lobes compared to lobes

Results

ur na

without baseline NCNs for the period 4-12 years.

Jo

Baseline Analysis

A total of 26309 participants received the baseline LDCT screen (Figure 1). For this baseline cohort, 59%

were men, 50% current smokers and 45% age 65 or over. Median (25th/75th) follow-up for cancer incidence was 11.3 (9.3/11.7) years.

Of the 26309 participants, 7090 (26.9%) had a NCN at baseline (Table 1). For the lobe-level analysis, among all 131,545 (i.e., 5*26309) lobes, 9448 (7.2%) had a baseline NCN. Among lobes with NCNs, 75.8% had at least 1 solid NCN, 20.5% had at least one NCN (solid or otherwise) of size ≥8 mm and the mean number of NCNs was 1.17. There were 1676 lung cancers in the baseline cohort, with 1457 having known lobe location (Table 2).

of

In the person-level analysis, raw RRs for presence of an NCN were 5.4 (95% CI: 4.7-6.3), 1.7 (95% CI: 1.42.0) and 1.6 (95% CI: 1.3-1.9) for the 3 periods; modeled RRs were similar. For the lobe-level analysis,

ro

the RRs were higher, with raw RRs of 16.3 (95% CI: 14.0-18.9), 3.0 (95% CI: 2.3-3.8) and 2.3 (95% CI: 1.73.2) for the three periods. Modeled RRs were similar (Table 2). For the lobe-level analysis restricted to

-p

subjects with at least one NCN, modeled RRs were 10.8 (95% CI: 8.5-13.5), 2.3 (95% CI: 1.6-3.1) and 2.0

re

(95% CI: 1.4-2.9) for the periods 0-4, 4-8, and 8-12 years, respectively.

For the analysis of nodule follow-up, 118180 (98.6%) lobes met the criteria for analysis (i.e., had

lP

year 2 screen and no new nodules at the year 1 or 2 screen), including 7986 (6.8%) with NCNs at baseline. There were 674 lung cancers during the period 4-12 years post-baseline, including 104 in lobes

ur na

with baseline NCNs. Among lobes with a baseline NCN, those with no NCNs or no growing NCNs at year 2 comprised the majority of the lobes (96%) and lung cancers (81%) and had modestly elevated risk compared to lobes without baseline NCNs; adjusted RRs were 1.7, 95% CI: 1.2-2.4 (no NCNs) and 2.2, 95% CI: 1.6-2.9 (no growing NCNs). In contrast, lobes with growing NCNs comprised a minority of lobes

Jo

(4%) and cancers (19%) but had much higher increased risk, with an adjusted RR of 14.5 (95% CI: 9.123.3). This risk was elevated in both the 4-8 year (RR=18.0, 95% CI: 10.3-31.5) and 8-12 year period (RR=10.3, 95% CI: 4.4-24.0).

Post-baseline Analysis

Of 19219 participants with no baseline NCNs, 18528 (96.4%) received at least one post-baseline screen and were included in the post-baseline analysis (Figure 1). Median (25th/75th) follow-up was 9.3 (8.6/9.7) years. For the person-level analysis, 1175 (6.3%) participants had new NCNs at a post-baseline screen (Table 1). For the lobe-level analysis, new NCNs were reported in 1311 (1.4%) lobes. Of lobes with new NCNs, 62.6% had at least 1 solid NCN and 36.2% had at least 1 NCN of size ≥8 mm (Table 1).

of

There were 801 cancers in the post-baseline cohort, with 666 having known lobe location (Table 3). For the person-level analysis, raw RRs for the 3 periods (0-4, 4-8 and 8-11 years) were 6.3 (95% CI: 5.0-

ro

7.8), 2.2 (95% CI: 1.5-3.1) and 1.8 (95% CI: 1.01-3.2). Modeled RRs were similar (Table 3). As with the baseline analysis, RRs for each period were higher for the lobe-level compared to the person-level

-p

analysis. Raw RRs for the lobe-level analysis were 25.3 (95% CI: 19.7-32.6), 3.3 (95% CI: 1.7-6.2) and 4.4

Nodule Characteristics

lP

re

(95% CI: 1.9-9.9), with modeled RRs similar.

ur na

Table 4 shows model results for nodule characteristics for lobes with at least one NCN. For the baseline analysis, having a large (≥8mm) NCN in a lobe was associated with higher risk as compared to only having small (< 8mm) NCNs, but with the magnitude of excess risk much higher at 0-4 years (RR=15.8, 95% CI: 12.2-20.3) than at later periods - RR=2.3 ( 95% CI: 1.4-3.7) at 4-8 years and RR=4.4

Jo

(95% CI: 2.5-7.7) at 8-12 years (Table 4). In contrast, for NCN attenuation, the RR for having a solid NCN in the lobe as compared to having only non- or semi-solid NCNs was elevated only in the initial period (RR= 1.4; 95% CI: 1.1-1.8). The RR was below one, though not significantly so, in the later periods - (RR= 0.72; 95% CI: 0.5-1.03 at 4-12 years). The RRs for attenuation were significantly different for the initial (0-4 years) versus later (4-12) periods, indicating a significant interaction (p < 0.001) of NCN attenuation with time period. There was a similar significant interaction with time period for NCN size (p < 0.001).

For the post-baseline analysis, the periods 4-8 and 8-11 years were aggregated due to the low number of cancers in each individual period. For having a large NCN, the RR for 0-4 years was significantly elevated with an RR=3.3 (95% CI: 2.1-5.2), though the magnitude of excess risk was much lower than for the corresponding period in the baseline analysis (RR=15.8). The RR was not significantly elevated in the 4-11 year period (RR=1.3; 95% CI: 0.5-3.6). For NCN attenuation, the RR for having a solid nodule was significantly elevated at 0-4 years (RR=3.5, 95% CI: 2.0-6.3) but not at 4-11 years (RR=1.4;

ro

of

95% CI: 0.5-3.9). The interactions with time period were not significant for either size or attenuation.

-p

Histology

Table 5 displays lung cancer histology by NCN status and time period for the baseline and post-

re

baseline analyses. Comparing lobes with baseline NCNs to those without, adenocarcinoma was

lP

significantly more common and small cell and squamous significantly less common for the 0-4 year period; findings were similar in the 4-12 year period, though squamous was no longer significantly different. In contrast, for the post-baseline analysis, there were no significant associations with histology

Discussion

ur na

by NCN status for either time period (Table 5).

Jo

We have shown here that NCNs identified on LDCT screening were predictive of lung cancer risk up

to ten or more years following the screen. Further, the risk was spatially correlated with nodule location, with the rate-ratio higher for the lobe-level than person-level analysis. The excess risk persisted when controlling for other lung cancer risk factors, suggesting that the NCNs themselves, at least a subset,

may be lung cancer precursors. If this is the case, this sheds light on the natural history of lung cancer and its precursors. With respect to NCN characteristics (e.g., size and attenuation), risk factors differed by time-period post screening, with size and solid attenuation posing greater risk in the short-term as compared to the long-term. Solid NCNs may pose greater risk short-term than long-term because when cancerous they

of

represent more aggressive tumors; thus, they are probably benign if not presenting as cancer within 4 years.

ro

A prior analysis of NLST showed that lung cancer risk as predicted by a commonly used model was

-p

significantly correlated with the false positive rate, or essentially the rate of finding an NCN with no associated lung cancer within one year 9. Since most risk factors for lung cancer would also likely be risk

re

factors for lung cancer precursors, this finding is consistent with the hypothesis that at least some NCNs are such precursors. An alternative hypothesis is that the nodules are markers for some field effect that

(e.g., within a lobe).

lP

increases lung cancer susceptibility, with that field effect generally spatially confined to a small area

ur na

The examination of histology showed, for the baseline analysis, that of cases diagnosed in the 4-12 year period, the tumor was more likely to be adenocarcinoma if there was a NCN in the cancer lobe. In contrast, the tumor was slightly (but not statistically significantly) less likely to be squamous cell if there

Jo

was an NCN in the cancer lobe. In the post-baseline analysis, the low number of cancers in the late period (N=16) limited the informativeness of the data, but similar trends were observed. Adenocarcinoma has the precursor of atypical adenomatous hyperplasia while squamous cell carcinoma has the precursor sequence of bronchial hyperplasias, metaplasias and dysplasias 10. Therefore, it is possible that these lesions have differential properties on CT imaging, with only the adenocarcinoma precursor being generally visible on low-dose CT scans.

A limitation of this analysis is the lack of imaging at the time of long-term diagnosis; therefore, it was not possible to determine whether the nodules in question were actual cancer precursors versus potential markers of a field effect. Clinical experience suggests that a solid nodule stable for more than two years is exceedingly unlikely to be a cancer, and current management is based on this expectation (i.e. it is typically recommended to stop following stable solid nodules after 2 years) 11 . In this study, most lobes with nodules at baseline had only stable, or non-reported, nodules at two years, and these

of

lobes still showed about a two-fold increased risk at 4-12 years compared to lobes without baseline

ro

nodules. A minority (about 20%) of the baseline nodules were growing at year 2, and the increased risk for lobes with such nodules was much greater (about 14-fold). Therefore, it is possible that the

-p

increased long-term risk associated with baseline nodules was due to both a field effect and to the presence of some cancer precursors among the nodules. When restricted to subjects with a baseline

re

NCN, the lobe-level analysis still showed significant RRs for presence of an NCN and long-term cancer

lP

outcome, indicating that tumor location and lobe location were significantly correlated.

ur na

Although the magnitude of excess risk 4-12 years following nodule detection was modest, these findings have potential implications for patient management. For example, for persons with NCNs and no lung cancer diagnosis within a year or two but who phase out of screening eligibility based on age or years since quit, should they continue with screening or some other form of increased surveillance? The

Jo

current recommendation is to stop screening after age 80, or after smoking cessation for more than 15 years, based in part on population models of lung cancer risk and life expectancy 12. Use of a prediction model that estimates the probability of an individual developing or dying from lung cancer based on their unique combination of risk factors, or that estimates the probability that a nodule found at screening will eventually be diagnosed as lung cancer based on individual risk factors, may be an

alternative approach 13-14. Computer-aided methods such as radiomics or deep learning may help further quantify risk based on individual CT findings 15-16. In conclusion, in this study the presence of NCNs on LDCT lung cancer screens were predictive of future lung cancer in the same location (lobe) up to a decade or more post-screening.

of

Long Term Authors

ro

Paul Pinsky: Conceptualization, Methodology, Formal Analysis, Investigation, Resources, Writing – Original Draft, Writing – Review & Editing

-p

David Gierada: Conceptualization, Investigation, Resources, Writing – Review & Editing

re

Declarations of interest: none

lP

Acknowledgments:

This research did not receive any specific grant from funding agencies in the public, commercial or

ur na

not-for-profit sectors. The authors report no conflicts.

Cancer incidence data have been provided by the following state cancer registries: Alabama, Arizona, California, Colorado, District of Columbia, Georgia, Hawaii, Idaho, Indiana, Iowa, Kentucky, Louisiana, Maryland, Massachusetts, Michigan, Minnesota, Missouri, Nevada, North Carolina, Ohio, Pennsylvania,

Jo

Rhode Island, Texas, Utah, Virginia and Wisconsin. All are supported in part by funds from the Centers for Disease Control and Prevention, National Program for Central Registries, local states, or by the National Cancer Institute, Surveillance, Epidemiology, and End Results Program. The results reported here and the conclusions derived are the sole responsibility of the authors.

References

1. Huang C, Hung S, Chang W, Sheu, C. Benign features of infection-related tumor-like lesions of the lung: a retrospective imaging review study. J Medical Imaging Radiation Oncology 2017; 61: 481-488.

of

2. Starnes SL, Reed MF, Meyer CA, et al. can lung cancer screening by computed tomography be effective in areas with endemic histoplasmosis? J Thorac Cardiovasc Surg 2011; 141: 688-693.

ro

3. Kim HY, Shim YM, Lee KS, Han J, Yi CA, and Kim YK. Persistent pulmonary nodular ground-glass opacity at thin-section CT: histopathologic comparisons. Radiology 2007; 245:267-275.

re

-p

4. Ohtsuka T, Watanabe K, Kaji M, Naruke T, and Suemasu K. A clinicopathological study of resected pulmonary nodules with focal pure ground-glass opacity. Eur J Cardiothorac Surg 2006; 30:160-163.

lP

5. Pinsky PF, Nath P, Gierada DS, et al. Short- and long-term lung cancer risk associated with noncalcified nodules observed on low-dose CT. Cancer Prev Res 2014; 7: 1179-1185.

ur na

6. Aberle DR, Berg CD, Black WC, et al. The National Lung Screening Trial: overview and study design. Radiology. Jan 2011;258:243-253. 7. The National Lung Screening Trial Research Team. Reduced lung-cancer mortality with low-dose computed tomographic screening. N Engl J Med. 2011;365:395-409.

Jo

8. National Lung Screening Trial Research Team. Lung cancer incidence and mortality with extended follow-up in the National Lung Screening Trial. J Thoracic Oncol 2019; doi.org/10.1016/j.tho.2019.05.044 9. False-positive screens and lung cancer risk in the National Lung Screening Trial: Implications for shared decision-making. Pinsky PF, Bellinger CR, Miller DP. J Med Screen 2018; 25: 110-112. 10. Kadara H, Scheet P, Wistuba II, Spira AE. Early events in the molecular pathogenesis of lung cancer. Cancer Prev Res 2016; 9: 518-527.

11. MacMahon H, Naidich DP, Goo JM, et al. Guidelines for management of incident pulmonary nodules detected on CT images: From the Fleischner Society 2017. Radiology 2017; 284: 228-243.

12. Moyer VA, US Preventive Services Task Force. Screening for lung cancer: U.S. Preventive Services Task Force recommendation statement. Ann Intern Med 2013; 160: 330-338. 13. McWilliams A, Tammemagi MC, Mayo JR, et al. Probability of cancer in pulmonary nodules detected on first screening CT. New Engl J Med 2013; 369: 910-919. 14. Tammemagi MC, Katki HA, Hocking WG, et al. Selection criteria for lung-cancer screening. New Engl J Med 2013; 368: 728-736. 15. Wilson R, Deveraj A. Radiomics of pulmonary nodules and lung cancer. Trans Lung Cancer Res 2017; 6: 86-91.

Jo

ur na

lP

re

-p

ro

of

16. Ardila, D, Kiraly AP, Bharawaj S, et al. End-to-end lung cancer screening with threedimenisonal deep learning on low-dose chest commuted tomography. Nature Med 2019; 25: 954-961.

Figure Caption

Jo

ur na

lP

re

-p

ro

of

Figure 1. Flowchart for baseline and post-baseline analysis. N and N[Lobe} show number of participants and number of lung lobes for the person-level and lobe-level analyses, respectively. LDCT yearly screens were at baseline (T0) and years 1 & 2 (T1, T2). Follow-up was 12 years from T0 and 11 years from T1/T2. See text for more details.

Table 1. Lobes and non-calcified nodules (NCNs) at baseline and post-baseline screens

1.66

1.38

131545 9448 (7.2) 7159 (75.8)

92640 1311 (1.4) 821 (62.6) 475 (36.2)

ro

1937 (20.5)

1.17 2

1.19

ur na

lP

re

1. Those with baseline nodules excluded 2. Only new (non pre-existing) NCNs included.

Jo

2

of

Post-Baseline 1 18528 1175 (6.3) 2

-p

Number of Participants Participants with NCNs - # (% of participants) Mean # of NCNs among participants with 1+ NCN Number of Lobes in Analysis Lobes with NCNs - # (% of all lobes) Lobes With 1+ Solid NCN - # (% of all lobes with an NCN) Lobes with 1+ NCN of size ≥ 8mm - # (% of all lobes with an NCN) Mean # of NCNs among lobes with 1+ NCN

Baseline 26309 7090 (26.9)

Table 2. Baseline non-calcified nodules (NCNs) and associated cancers by time period

774 486 416 1676

507 177 144

Lobe-level Analysis

All Lobes (N=131545)

Lobes with NCNs (N=9448)

Lobes without NCNs (N=122097)

Follow-up Period 0-4 Years

704

13.8

386

109.0

318

4-8 Years 8-12 Years All

404 349 1457

8.8 10.3

72 50

23.3 22.5

Rate per 10,000 PY|LY Participants without NCNs (N=19219) 267 35.3 309 45.1 272 52.7

Raw RR (95% CI) 1

5.4 (4.7-6.3) 1.7 (1.4-2.0) 1.6 (1.3-1.9)

-p

re

190.6 74.7 82.0

# Cancers

Modeled RR (95% CI) 2

5.1 (4.4-5.9) 1.5 (1.3-1.8) 1.5 (1.20-1.8)

of

0-4 Years 4-8 Years 8-12 Years All

332 299

6.7

16.3 (14.0-18.9)

7.8 9.5

3.0 (2.3-3.8) 2.3 (1.7-3.2)

14.7 (12.617.2) 2.6 (2.0-3.4) 2.2 (1.6-2.9)

lP

75.8 52.7 60.2

Rate per 10,000 PY|LY Participants with NCNs (N=7090)

ro

# Cancers

Person-level Analysis

# Rate Cancers per 10,000 PY|LY All Participants (N=26309)

ur na

1. Rate ratio of participants/lobes with NCNs compared to participants/lobes without NCNs. 2. Modeled rate ratio (participants/lobes with NCNs compared to participants/lobes without NCNs) controlling for sex, age, current smoking status, pack-years and lobe location. LY- lobe year, PY- person year.

Jo

Note: For lobe-level analysis, 219 cancers with unknown or non-lobe location are excluded; number of cancers include only those cancers located in the given lobe.

Table 3. Post-Baseline non-calcified nodules (NCNs) and associated cancers by time period

380 304 116 801

109 34 13

52.1 51.0 54.8

248 103.8 93.8

# Rate Cancers per 10,000 PY|LY Participants without NCNs (N=17353) 271 39.5 270 47.9 103 52.0

Raw RR (95% CI) 1

Modeled RR (95% CI) 2

6.3 (5.0-7.8) 2.2 (1.5-3.1) 1.8 (1.01-3.2)

5.6 (4.5-7.0) 1.9 (1.3-2.7) 1.6 (0.9-2.9)

of

# Rate Cancers per 10,000 PY|LY Participants with NCNs (N=1175)

ro

Personlevel Analysis 0-4 Years 4-8 Years 8-11 Years All

# Rate Cancers per 10,000 PY|LY All Participants (N=18528)

All Lobes (N=92640)

Lobes with NCNs (N=1311)

Lobes without NCNs (N=91329)

320

8.7

82

166.9

238

4-8 Years 8-11 Years All

252 94 666

8.7 9.4

10 6

27.7 39.4

-p

Lobe-level Analysis Follow-up Period 0-4 Years

re

6.6 8.5 8.9

19.6 (14.925.3) 2.5 (1.3-4.7) 3.3 (1.4-7.6)

lP

242 88

25.3 (19.732.6) 3.3 (1.7-6.2) 4.4 (1.9-9.9)

ur na

1. Rate ratio of participants/lobes with NCNs compared to participants/lobes without NCNs. 2. Modeled rate ratio (participants/lobes with NCNs compared to participants/lobes without NCNs) controlling for sex, age, current smoking status, pack-years and lobe location. LY- lobe year, PY- person year.

Jo

Note: 135 participants had lung cancer with unknown or non-lobe location; these cancers are excluded from the lobe-level analysis.

Table 4. Rate ratios for nodule characteristics at baseline and post-baseline by time period

Period 4-8 Years

8-12 Years

4-12 Years

Modeled RR 1

Modeled RR 1

Modeled RR

Ref 15.8 (12.220.3) Ref

Ref 2.3 (1.4-3.7)

Ref Ref 4.4 (2.5-7.7) 3.0 (2.1-4.3)

Ref

Ref

1.4 (1.1-1.8) 0-4 Years

0.79 (0.51.3) 4-8 Years

0.64 (0.41.1) 8-11 Years

Ref 3.3 (2.1-5.2) Ref

* * *

3.5 (2.0-6.3)

ur na

Post-Baseline All NCNs Size < 8 mm 1+ NCN Size ≥ 8 mm All NCNs non- or semisolid 1+ Solid NCN

*

of

ro

All NCNs non- or semisolid 1+ Solid NCN

P-value Interaction of RR (0-4 years vs 4-12 years) 2

1

<0.001

Ref

0.71 (0.501.03) 4-11 years

-p

Baseline All NCNs Size < 8 mm 1+ NCN Size ≥ 8 mm

lP

Modeled RR 1

re

0-4 Years

<0.001

* * *

Ref 1.3 (0.5-3.6) Ref

0.11

*

1.4 (0.5-3.9)

0.11

Note: Only lobes with 1+ NCN included.

* Too few cases to reliably estimate RR.

Jo

1. Modeled rate ratio controlled for sex, age, current smoking status, pack-years and lobe location. 2. Test for difference in RR for 0-4 year versus 4-12 (or 4-11) year period.

Table 5. Lung Cancer Histology by NCN presence and time period. Time Period NCN in cancer lobe

P-value

Baseline Adenocarcinoma Squamous Other NSCLC Small Cell Other/Unknown All

127 (39.9) 87 (27.4) 52 (16.4) 46 (14.5) 6 (1.9) 318

251 (65.0) 59 (15.3) 53 (13.7) 16 (4.2) 7 (1.8) 386

<0.001 <0.001 0.33 <0.001 0.94

233 (36.9) 194 (30.7) 78 (12.4) 107 (17.0) 19 (3.0) 631

62 (50.8) 31 (25.4) 15 (12.3) 8 (6.6) 6 (4.9) 122

0.004 0.23 0.98 0.004 0.28

Post-baseline Adenocarcinoma Squamous Other NSCLC Small Cell Other/Unknown All

90 (37.8) 65 (27.3) 36 (0.15) 41 (17.2) 6 (2.5) 238

28 (34.2) 27 (32.9) 14 (0.17) 11 (13.4) 2 (2.4) 82

0.55 0.33 0.68 0.42 0.97

131 (39.7) 107 (32.4) 35 (10.6) 48 (14.6) 9 (2.7) 330

8 (50.0) 3 (18.8) 3 (18.8) 1 (6.3) 1 (6.3) 16

0.41 0.25 0.31 0.35 0.41

ur na Jo

of

4-12 years No NCN in cancer lobe

ro

P-value

-p

NCN in cancer lobe

re

0-4 years No NCN in cancer lobe

lP

Histology