Harry J. de Koning, MD; Rafael Meza, PhD; Sylvia K. Plevritis, PhD; Kevin ten Haaf, MSc; Vidit N. Munshi, MS; Jihyoun Jeon, PhD; Saadet Ayca Erdogan, PhD; Chung Yin Kong, PhD; Summer S. Han, PhD; Joost van Rosmalen, PhD; Sung Eun Choi, SM; Paul F. Pinsky, PhD; Amy Berrington de Gonzalez, PhD; Christine D. Berg, MD; William C. Black, MD; Martin C. Tammemägi, PhD; William D. Hazelton, PhD; Eric J. Feuer, PhD *; Pamela M. McMahon, PhD *
Disclaimer: The contents of this report are solely the responsibility of the authors and do not necessarily represent the official views of the National Cancer Institute or the Agency for Healthcare Research and Quality.
Acknowledgment: The authors thank Melecia Miller, MPH (formerly of Massachusetts General Hospital); Suresh Moolgavkar (Fred Hutchinson Cancer Research Center); and Arry de Bruijn (Erasmus Medical Center).
Grant Support: This report is based on research conducted by the National Cancer Institute's Cancer Intervention and Surveillance Modeling Network through support from an interagency agreement with the Agency for Healthcare Research and Quality, Rockville, Maryland (administrative supplement to U01 CA152956).
Potential Conflicts of Interest: Disclosures can be viewed at www.acponline.org/authors/icmje/ConflictOfInterestForms.do?msNum=M13-2316.
Reproducible Research Statement:Study protocol: Available from Dr. Meza (e-mail, email@example.com). Statistical code: Please go to http://cisnet.cancer.gov/lung/profiles.html. Data set: Please go to https://biometry.nci.nih.gov/cdas.
Requests for Single Reprints: Harry J. de Koning, MD, PhD, Department of Public Health, Erasmus Medical Center, PO Box 2040, 3000 CA Rotterdam, the Netherlands; e-mail, firstname.lastname@example.org.
Current Author Addresses: Drs. de Koning and van Rosmalen and Mr. ten Haaf: Department of Public Health, Erasmus Medical Center, PO Box 2040, 3000 CA, Rotterdam, the Netherlands.
Dr. Meza: Department of Epidemiology, University of Michigan, 1415 Washington Heights SPH-II 5533, Ann Arbor, MI 48109-2029.
Dr. Plevritis: NCI Stanford Center for Cancer Systems Biology, Department of Radiology, Stanford University, 1201 Welch Road, Room P060, MC 5488, Stanford, CA 94305-5488.
Mr. Munshi: Institute for Technology Assessment, Massachusetts General Hospital, 101 Merrimac Street, 3rd Floor, Boston, MA 02114-4724.
Drs. Jeon and Hazelton: Fred Hutchinson Cancer Research Center, 1100 Fairview Avenue North, PO Box 19024, Seattle, WA 98109-1024.
Drs. Erdogan and Han: Department of Radiology, Stanford University, 1201 Welch Road, Room P060, MC 5488, Stanford, CA 94305-5488.
Drs. Kong and McMahon and Ms. Choi: Institute for Technology Assessment, Harvard Medical School, Massachusetts General Hospital, 101 Merrimac Street, 10th Floor, Boston, MA 02114.
Drs. Pinsky, Berrington de Gonzalez, Berg, and Feuer: National Cancer Institute, National Institutes of Health, 6116 Executive Boulevard, Suite 504, Bethesda, MD 20892.
Dr. Black: Department of Radiology, Dartmouth Hitchcock Medical Center, 1 Medical Center Drive, Lebanon, NH 03756.
Dr. Tammemägi: Department of Community Health Sciences, Brock University, Walker Complex, Academic South, Room 306, 500 Glenridge Avenue, St. Catharines, Ontario L2S 3A1, Canada.
Author Contributions: Conception and design: H.J. de Koning, R. Meza, S.K. Plevritis, S.A. Erdogan, C.Y. Kong, W.C. Black, M.C. Tammemägi, E.J. Feuer, P.M. McMahon.
Analysis and interpretation of the data: H.J. de Koning, R. Meza, S.K. Plevritis, K. ten Haaf, V.N. Munshi, J. Jeon, S.A. Erdogan, C.Y. Kong, S.S. Han, J. van Rosmalen, S.E. Choi, P.F. Pinsky, A. Berrington de Gonzalez, C.D. Berg, W.C. Black, M.C. Tammemägi, W.D. Hazelton, E.J. Feuer, P.M. McMahon.
Drafting of the article: H.J. de Koning, R. Meza, S.K. Plevritis, J. Jeon, C.Y. Kong, S.S. Han, M.C. Tammemägi, E.J. Feuer, P.M. McMahon.
Critical revision of the article for important intellectual content: H.J. de Koning, R. Meza, S.K. Plevritis, K. ten Haaf, J. Jeon, S.A. Erdogan, C.Y. Kong, J. van Rosmalen, P.F. Pinsky, A. Berrington de Gonzalez, C.D. Berg, W.C. Black, M.C. Tammemägi, W.D. Hazelton, E.J. Feuer, P.M. McMahon.
Final approval of the article: H.J. de Koning, R. Meza, S.K. Plevritis, K. ten Haaf, V.N. Munshi, J. Jeon, C.Y. Kong, J. van Rosmalen, A. Berrington de Gonzalez, C.D. Berg, W.C. Black, M.C. Tammemägi, W.D. Hazelton, E.J. Feuer, P.M. McMahon.
Provision of study materials or patients: S.K. Plevritis, C.D. Berg, M.C. Tammemägi.
Statistical expertise: S.K. Plevritis, S.S. Han.
Obtaining of funding: S.K. Plevritis.
Administrative, technical, or logistic support: S.K. Plevritis.
Collection and assembly of data: S.K. Plevritis.
The optimum screening policy for lung cancer is unknown.
To identify efficient computed tomography (CT) screening scenarios in which relatively more lung cancer deaths are averted for fewer CT screening examinations.
Comparative modeling study using 5 independent models.
The National Lung Screening Trial; the Prostate, Lung, Colorectal, and Ovarian Cancer Screening trial; the Surveillance, Epidemiology, and End Results program; and the U.S. Smoking History Generator.
U.S. cohort born in 1950.
Cohort followed from ages 45 to 90 years.
576 scenarios with varying eligibility criteria (age, pack-years of smoking, years since quitting) and screening intervals.
Benefits included lung cancer deaths averted or life-years gained. Harms included CT examinations, false-positive results (including those obtained from biopsy/surgery), overdiagnosed cases, and radiation-related deaths.
The most advantageous strategy was annual screening from ages 55 through 80 years for ever-smokers with a smoking history of at least 30 pack-years and ex-smokers with less than 15 years since quitting. It would lead to 50% (model ranges, 45% to 54%) of cases of cancer being detected at an early stage (stage I/II), 575 screening examinations per lung cancer death averted, a 14% (range, 8.2% to 23.5%) reduction in lung cancer mortality, 497 lung cancer deaths averted, and 5250 life-years gained per the 100 000-member cohort. Harms would include 67 550 false-positive test results, 910 biopsies or surgeries for benign lesions, and 190 overdiagnosed cases of cancer (3.7% of all cases of lung cancer [model ranges, 1.4% to 8.3%]).
The number of cancer deaths averted for the scenario varied across models between 177 and 862; the number of overdiagnosed cases of cancer varied between 72 and 426.
Scenarios assumed 100% screening adherence. Data derived from trials with short duration were extrapolated to lifetime follow-up.
Annual CT screening for lung cancer has a favorable benefit–harm ratio for individuals aged 55 through 80 years with 30 or more pack-years' exposure to smoking.
National Cancer Institute.
Diagram of how earlier detection (followed by treatment) may have an effect on reducing serious consequences of the disease and/or increasing life expectancy.
All models account for the individual's age-specific smoking-related risk for lung cancer, the date and stage of lung cancer diagnosis, the corresponding lung cancer mortality, and the individual's life expectancy in the presence and absence of screening. By replicating trial detection, models estimate key parameters of the screening-detectable period and/or sensitivity and can subsequently estimate cancer detected in the screening scenarios. In essence, when a model incorporates the exact demographic characteristics of participants and the design of a trial, it should be able to reproduce cumulative incidence of lung cancer (by stage, histologic features, sex, age, type of detection, and round) and lung cancer mortality in both groups as closely as possible. The best fit is often defined as the lowest deviance between observed and model-expected numbers.
Percentage and 95% CI of lung cancer mortality in chest radiography group compared with computed tomography group in the NLST, by follow-up duration and comparison with 5 model group results.
As stated in the Methods section, close calibration to difference in lung cancer mortality between groups of the NLST at 6 years' follow-up was prioritized, but not the slope before year 6. This was done on purpose because mortality differences in the first years of trials are subject to chance and small numbers. E = Erasmus Medical Center; F = Fred Hutchinson Cancer Research Center; M = Massachusetts General Hospital; NLST = National Lung Screening Trial; S = Stanford University; U = University of Michigan.
Appendix Table. Key Similarities and Differences Between the Models in Estimating Effects on Life Expectancy With an Effective Lung Cancer Screening Test
Table 1. Benefits of 26 Selected Efficient Screening Programs and a Screening Program Most Similar to Eligibility Criteria for the National Lung Screening Trial*
Twenty-seven screening scenarios in order of increasing number of CT examinations needed, with the relative increase in screening examinations and lung cancer deaths averted (compared with the prior scenario), and the average number of lung cancer deaths averted in each scenario, for a 100 000-person 1950 cohort followed from ages 45 to 90 years.
Number of CT scans is given in Table 1. The bars show the absolute number of lung cancer deaths averted in 27 screening scenarios and the percentage increases in both screening examinations and deaths averted when inclusion criteria are relaxed. The first 2 triennial scenarios show the effect of stopping through age 80 or 85 years: about 6% more screenings when stopping through age 85 years, leading to 11% more lung cancer deaths averted (compared with stopping through age 80 years). Extending the maximum (quit) time from 10 years to 15 years leads to a 6% increase in deaths averted (at the expense of 15% additional screening examinations), and extending it to 25 years yields an additional 12% in lung cancer deaths averted (at the expense of 20% additional screening examinations). Decreasing the minimum patient-year eligibility criteria from 30 to 20 and to 10 patient-years in annual scenarios shows relatively large increases in additional CT scans needed compared with additional lung cancer deaths averted. The scenario's rank score among 576 possible scenarios (that is, the average distance to the efficient frontier for the 5 models) is shown in parentheses. A = annual; B = biennial; CT = computed tomography; PY = minimum pack-years; T = triennial; YSQ = maximum years since quitting.
Table 2. Harms of 26 Selected Efficient Screening Programs and a Screening Program Most Similar to Eligibility Criteria for the National Lung Screening Trial*
Estimated lung cancer mortality reduction (as percentage of total lung cancer mortality in cohort) and life-years gained (averages of 5 models) from annual CT screening, for programs with minimum eligibility age of 55 years and maximum of 80 years at different smoking eligibility cutoffs and NLST scenario (A-55-75-30-15).
The average number of CT screening examinations (5 models) is shown on x-axis. The graph plots the average number of CT screening examinations against the percentage reduction of lung cancer mortality (top) or life-years gained (bottom) for each screening scenario (versus no screening) that was estimated for 100 000 individuals of the 1950 cohort followed from ages 45 to 90 years. Programs are labeled as follows: frequency–start age–stop age–minimum pack-years–maximum years since quitting smoking. The reductions in lung cancer mortality differ from the point estimate of the reduction reported at the 6.5-year follow-up in the NLST because only eligible persons are screened in this cohort analysis (dilution) and lifetime reduction in lung cancer mortality is modeled. The top panel shows the efficiency frontier for all models combined. When the slope in the efficiency frontier plot levels off, the additional reductions in mortality per unit increase in use of CT screening examinations are small relative to the previous strategies. A = annual; CT = computed tomography; NLST = National Lung Screening Trial.
Table 3. Number of Individuals Having Benefits and Harms of Annual CT Screening From Ages 55 Through 80 Years
Estimated percentage of reduction in lung cancer mortality and overdiagnosed cases (of screening-detected cases) for the highlighted scenarios in Tables 1 and 2 (average number of CT screening examinations is shown on the x-axis) for all individual models and the average of the 5 models.
Presentation of a 100 000-person 1950 cohort followed from ages 45 to 90 years. CT = computed tomography.
Absolute number of lung cancer deaths averted for the scenarios in Table 1, for all model groups separately and the average of 5 models.
Presentation of 100 000 individuals of the 1950 cohort followed from ages 45 to 90 years. The x-axis shows the number of CT screening examinations. Ranking of strategies is similar across models. There is no direct comparison with observed data for this specific 1950 cohort. CT = computed tomography.
de Koning HJ, Meza R, Plevritis SK, et al. Benefits and Harms of Computed Tomography Lung Cancer Screening Strategies: A Comparative Modeling Study for the U.S. Preventive Services Task Force. Ann Intern Med. 2014;160:311–320. doi: https://doi.org/10.7326/M13-2316
Download citation file:
Published: Ann Intern Med. 2014;160(5):311-320.
Cancer Screening/Prevention, Hematology/Oncology, High Value Care, Lung Cancer, Prevention/Screening.
Results provided by:
Copyright © 2020 American College of Physicians. All Rights Reserved.
Print ISSN: 0003-4819 | Online ISSN: 1539-3704
Conditions of Use