Risk-Based Cancer Screening: Insights from the NLST

Transcription

1 N L S T Risk-Based Cancer Screening: Insights from the NLST National Lung Screening Trial National Cancer Institute Christine D. Berg, MD Project Officer, NLST Former, Chief, Early Detection Research Group National Cancer Institute No conflicts of interest to disclose The FDA does not have jurisdiction over use of helical CT or CXR screening for lung cancer as it does for mammography. Takes act of Congress. 2 1

2 Lecture Objectives Identify benefits and risks of low-dose helical CT screening for lung cancer Current lung cancer screening guidelines Identify groups in whom to consider screening Identify groups NOT to screen Radiologic issues for implementation of screening Nodule management Smoking cessation Lung cancer death rates and counts for men and women aged years as observed and for modeled tobacco control scenarios. Moolgavkar S H et al. JNCI J Natl Cancer Inst 2012;jnci.djs136 The Author Published by Oxford University Press. 2

3 Low-Dose Helical CT Allows entire chest to be surveyed in a single breathhold Time: approximately 7-15 seconds Reduces motion artifact Eliminates respiratory misregistration Narrower slice thickness Hourly throughput - 4 patients per hour Radiation dose one fifth of diagnostic CT One arm study Subjects enrolled % positive 5 mm detection rate ELCAP US ,7 2,7 I-ELCAP International ,3 1,3 ALCA Japan ,5 0,9 University Munster, Germany ,8 1,5 Shinshu University, Japan ,1 0,4 Finnish Institute of Occupational Helth, Finland 602 8,0 0,8 SMC, University School of Medicine, Seoul, Korea ,9 0,2 Hitachi Health Care Center, Japan ,8 0,5 PALCAD, Ireland ,7 0,4 University of Milano Italy ,9 1,1 NY-ELCAP, US ,4 1,6 PLuSS, US ,5 1,5 COSMOS Italy ,7 1,1 University of Navarra, Spain ,4 1,3 Univerity of Turin, Italy ,0 1,0 Totale ,6 1,1 media 12,8 1,1 range 5,1-24,7 0,4-1,6 3

4 Survival is not adequate measure of screening benefit Lead-time bias Earlier detection may prolong survival independent of delay in death Cancer CT DX Lead time Sx DX Survival prolongation due to lead time Survival Prolonged survival with mortality from cause other than lung cancer or complication of screening L O Overdiagnosis is an extreme form of length bias 4

5 The National Lung Screening Trial Definitive RCT for effect of lung cancer screening on lung cancer specific mortality NLST design Prospective, randomized trial comparing low-dose helical CT screening to chest x-ray screening for three annual screens with the endpoint of lung cancer specific mortality in 53, 454 high risk participants Eligibility Age 55-74; asymptomatic current or former smoker; 30 pack year smoking history; former smokers: quit within preceding 15 years Parameters 90% power to detect 20% difference in lung cancer mortality; α = 0.05 Median follow-up for outcomes ~ 6.5 years (Maximum: 7.4) Vital status known for 97% LDCT 96% CXR 5

6 Baseline Characteristics by Trial Arm and Comparison with US Census Data for High-risk Population CT CXR US Census* (%) Male 15,769 (59) 15,761 (59) (58) Age at baseline (years) < > 70 11,441 (43) 8,171 (31) 4,756 (18) 2,354 (9) 11,425 (43) 8,197 (31) 4,762 (18) 2,348 (9) (35) (29) (20) (15) Race/Ethnicity Black Hispanic 1,195 (4) 479 (2) 1,181 (4) 456 (2) (6) (2) Smoking status at baseline Former smoker Current smoker 13,857 (52) 12,865 (48) 13,827 (52) 12,905 (48) (43) (57) *Percentages were calculated using data from respondents to the US Census Tobacco Use Supplement of the Continuing Population Survey ( ) who met NLST age and smoking eligibility criteria. Interim analysis: lung cancer mortality Arm Person Years (py) Lung cancer deaths Lung cancer mortality per 100,000 py Reduction in lung cancer mortality (%) Value of test statistic Efficacy boundary LDCT 144, CXR 143, p = Deficit of lung cancer deaths in CT arm exceeds that expected by chance, even allowing for multiple looks at the data. CXR arm compared with matched 30,000 cohort in PLCO, no benefit of CXR seen. 6

7 Probability of survival: Participants with lung cancer Probability of survival: ALL participants 4/22/2013 Kaplan-Meier curves for lung cancer and all-cause mortality CT arm lung cancer CXR arm lung cancer CT arm all-cause CXR arm all-cause Years from randomization Lung cancer case survival Kaplan Meier curve CXR compared with matched group in PLCO and no difference Years from diagnosis CT arm CXR arm 7

8 Cumulative Numbers of Lung Cancers and of Deaths from Lung Cancer. The National Lung Screening Trial Research Team. N Engl J Med 2011;365: Table 5. Results for National Lung Screening Trial Subset. Oken, M. M. et al. JAMA 2011;306: Copyright restrictions may apply. 8

9 Screen positivity rate by screening round & arm Number screened Low-dose helical CT Number positive % Positive Number screened CXR Number positive % Positive Screen 1 26,314 7, ,049 2, Screen 2 24,718 6, ,097 1, Screen 3 24,104 4, ** 23,353 1, ** All screens 75,136 18, ,499 5, * Positive screen: nodule 4 mm or other findings potentially related to lung cancer. ** Abnormality stable for 3 rounds could be called negative by protocol. True and false positive screens Screening Result Screen 1 Low-dose Helical CT Round 2 Round 3 Round 1 CXR Round 2 Round 3 Total Positives 7,193 (100) 6,902 (100) 4,054 (100) 2,387 (100) 1,482 (100) 1,175 (100) Lung cancer No lung cancer 270 (4) 6,923 (96) 168 (2) 6,734 (98) 211 (5) 3,843 (95) 136 (6) 2,251 (94) 65 (4) 1,417 (96) 78 (7) 1,097 (93) Data reflect the final interpretation, including benefit of historical comparison exams. Positive Screens were > 3-fold higher in the LDCT arm 9

10 lung cancers diagnosed in NLST Screen Result and Time Period CT (%) CXR (%) Total T0-T2 Screen [+] lung cancers 649 (61.2%) 279 (29.6%) Total T0-T2 Screen [-] lung cancers 44 (4.2%) 137 (14.6%) Total NO screen lung cancers 367 (34.6%) 525 (55.8%) Total lung cancers in arm 1060 (100.0%) 941 (100%) 892 NO screen cancers include: post-screen time period (N = 802) never screened (N= 35) due for screen (N = 55) NLST Investigators; NEJM 2011 Definitions of Performance Measures Sensitivity: probability of a positive test given that the patient is ill # of true positives / # of true positives + # of false negatives Specificity Probability of a negative test given that the patient is well # of true negatives / # of true negatives + # of false positives Negative Predictive Value # of true negatives / # of true negatives + number of false negatives Positive Predictive Value # of true positives / # of true positives + # of false positives 10

11 Performance Characteristics at T0 Sensitivity LDCT: 93.8% (270/288) CXR: 73.5% (136/185) Specificity LDCT: 73.4% (19,043/25,954) CXR: 91.3% (23,5477/25,790) Negative Predictive Value LDCT: 99.9% CXR: 99.8% Positive Predictive Value (PPV1) LDCT: 3.8% (267/7,010) CXR: 5.7% (136/2,379) PPV for positive finding that led to biopsy (PPV3) LDCT: 52.9% (265/501) CXR: 70.2% (132/188) Follow-up of Positive Screens by Screening Round and Trial Arm CT CXR Round 1 Round 2 Round 3 Round 1 Round 2 Round 3 Total Positives with documented follow-up Imaging procedures Clinical procedures Other procedures No procedures 7,051 (100) 5,718 (81) 5,086 (72) 219 (3) 680 (10) 6,741 (100) 2,517 (37) 3,190 (47) 130 (2) 2,876 (43) 3,911 (100) 2,009 (51) 2,151 (55) 102 (3) 1,389 (36) 2,348 (100) 2,009 (86) 1,413 (60) 72 (3) 173 (7) 1,456 (100) 968 (67) 723 (50) 36 (2) 378 (26) 1,149 (100) 906 (79) 658 (57) 42 (4) 192 (17) Biopsy procedures Cytology procedures Resection 240 (3) 248 (4) 249 (4) 144 (2) 109 (2) 169 (2) 146 (4) 125 (3) 193 (5) 121 (5) 88 (4) 106 (4) 48 (3) 59 (4) 48 (3) 62 (5) 63 (6) 54 (5) Total Invasive/Lung cancers 737/ / / / /65 179/

12 Histopathological Type by Trial Arm and Stage at Diagnosis* CT CXR I II III IV Total I II III IV Total Adenocarcinoma 217 (58.5) 27 (7.3) 62 (16.7) 65 (17.5) 371 (100) 121 (38.2) 29 (9.1) 56 (17.7) 111 (35.0) 317 (100) Squamous cell carcinoma 122 (51.5) 25 (10.5) 58 (24.5) 32 (13.5) 237 (100) 72 (37.5) 22 (11.5) 51 (26.6) 47 (24.5) 192 (100) Bronchiolo-alveolar carcinoma 88 (81.5) 4 (3.7) 10 (9.3) 6 (5.6) 108 (100) 18 (51.4) 3 (8.6) 8 (22.9) 6 (17.1) 35 (100) Large cell carcinoma 21 (51.2) 3 (7.3) 12 (29.3) 5 (12.2) 41 (100) 14 (33.3) 2 (4.8) 17 (40.5) 9 (21.4) 42 (100) Small cell carcinoma 6 (5.0) 7 (5.8) 42 (34.7) 66 (54.5) 121 (100) 12 (8.6) 8 (5.8) 42 (30.2) 77 (55.4) 139 (100) Other 55 (40.7) 7 (5.2) 32 (23.7) 41 (30.4) 135 (100) 35 (21.2) 9 (5.5) 48 (29.1) 73 (44.2) 165 (100) Total 509 (50.2) 73 (7.2) 216 (21.3) 215 (21.2) 1,013 (100) 272 (30.6) 73 (8.2) 222 (24.9) 323 (36.3) 890 (100) *Table includes diagnosed cancers with data on both stage and histopathological type. screening-related complications CT lung cancer CT NO cancer CXR lung cancer CXR No cancer N % N % N % N % Positive screens , , Major complication < 0.1 Death 60 days after invasive procedure < Overall complications were low. Complications in patients with no cancer diagnosis were minimal. The 20 types of major complication included hemothorax requiring tube placement, myocardial infarction, bronchopulmonary fistula 12

13 Radiation Risk LDCT Comparison to Standard Chest CT NLST Effective Dose vs Standard Chest CT male female std chest ct msv EURO GROUP CTEXPO & ICRP 60 CTEXPO & ICRP 103 STD CHEST CT Acceptable chest CT screening can be accomplished at a small fraction of the dose of a standard chest CT 13

14 radiation dose Whole body effective dose (weighted average dose to each organ) Low dose helical CT: 1.5mGy Mammogram: 0.7mGy CXR: 0.01 mgy Low dose helical CT: estimates of organ specific dose Lung: 4 mgy Breast: 4 mgy Red bone marrow, stomach, liver and pancreas: each ~1 mgy Screening mammogram organ specific dose: Breast : 4mGy Other organs: < 0.1mGy CXR: effective dose ~ 0.1 msv 3 screens Smokers Age Radiation Risks vs Benefits Radiation risk from screens 1-3 lung cancer deaths per 10,000 screened 0.3 breast cancers per 10,000 females screened Radiation risk from follow-up CT scans Low-dose or thin-section chest CT + 25% Diagnostic chest CT + 100% Cumulative mortality reduction NLST 30 lung cancer deaths per 10,000 screened 14

15 Change in cancer risk per 10,000 screened (assuming 20% mortality reduction) Female never smokers 10 Age Age Age Age Lung cancer risk Lung cancer benefit Breast cancer risk Berrington de Gonzalez A, Kim KP, Berg CD. J Med Screen 2008;15: NLST LSS Study of Factors Associated with Smoking Behavior Screening Result Odds Ratios (95% CI; p-values) Normal no abnormalities (Referent group) 1.00 Negative for lung cancer, minor other abnormalities 0.85 ( ; p = 0.037) Negative for lung cancer, significant other abnormalities Odds ratios from multivariable* longitudinal logistic analysis for being a current smoker given the previous screening result ( ; p = 0.004) Positive for lung cancer 0.46 ( ; p < 0.001) Positive for lung cancer, stable, no significant change from previous screen 0.67 ( ; p = 0.003) Conclusion: The probability of subsequent smoking is inversely associated with the abnormality of screening result in a dose-response fashion. Tammemagi, M and Taylor K 15

16 European Randomized Trials European randomized CT screening trials: Post-NLST J.K. Field, R. van Klaveren, J.H. Pedersen, U. Pastorino, E.Paci, N. Becker, M. Infante, Matthijs Oudkerk H.J. de Koning on behalf of the European Randomised Screening Trial Group. NELSON DLCST LUSI DANTE ITALUNG MILD UKLS No. of rounds or 5 1 Screen interval 1, 2 and year 1 year 1 year 1 year Randomisation NA yrs 1 or 2 yrs Screen arm and ,000 Control arm ,000 Age at randomisation (yrs ± SD) Current smokers at randomisation 59 (6) 57 (5) 58 (5) 65 (5) 61 (4) 59 (6) NA 55% 76% 61% 55 % 65% 63% NA Mean pack-years (± SD) 42 (19) 36 (13) 36 (18) 47 (25) 43 (18) 43 NA Females 16% 45% 34% 0 % 35% 32% NA Follow-up since 6 yrs 5 yrs 3 yrs 6 yrs 6 yrs 5 yrs NA randomisation PY of follow-up** 90,655 23,248 2,031 13,541 14,453 5,589 NA EU total ~17,000 screen arm * Last update August 2010; NA: not available or applicable; PY: Person-years; LC: lung cancer; SD: standard deviation; yrs: years;**: cut off date ; Study recruiting. 16

17 What next and for whom? IASLC J Thoracic Oncol 2012:Jan 7(1): Strategic CT Screening Advisory Committee plans: identification of high-risk individuals for lung cancer CT screening programs develop radiological guidelines for use in developing national screening programs develop guidelines for the clinical work-up of "indeterminate nodules" resulting from CT screening programmers guidelines for pathology reporting of nodules from lung cancer CT screening programs; recommendations for surgical and therapeutic interventions of suspicious nodules identified through lung cancer CT screening programs; and integration of smoking cessation practices into future national lung cancer CT screening programs. 17

18 U.S. Population Estimates from NHANES Courtesy of Peter Bach Criteria Total Population Women Men Current Smokers Former Smokers 55-74, 30 pack years, 15 years quit 50-79, 20 pack years, 10 years quit 50, smoking history 21, 10 years of smoking any amount 7,425,000 3,220,000 4,205,000 4,027,000 3,398,000 13,500,000 5,816,000 7,684,000 9,114,,000 4,386,000 46,481,000 20,147,000 26,334,000 14,729,000 31,752,000 77,005,000 33,805,000 43,200,000 39,883,000 37,122,000 Existing Guidelines Organization Type of Statement NLST Like Subjects? Others? ACCP/ ASCO/ ATS endorsed Evidence based guideline Suggest it be offered No ACS Guideline May be considered No ALA Guidance Recommended No NCCN Consensus guidelines Recommended Yes, for some individuals AATS Guideline Recommended Age 50-79; 20 pack years and five year cumulative risk of 5% or more; lung cancer survivors 18

19 Estimates of Screening Avertable Deaths Screen NLST population (55 74; current or former smokers of 30 pack years, smoked within 15 years) % of cancers: 26.7% With 80% compliance 5% of lung cancer deaths averted (8,150) Pinsky and Berg, JMS 2012:19( ) Separate estimate by ACS researchers Screen 8.6 million; Full compliance 12,250 deaths averted; 70% compliance 8575 deaths averted Ma J et al Cancer on-line February 25,

20 Risk Prediction Models to Determine who to screen Several models exist Tammemagi et al JNCI 2011 includes Age, Education, Family history of lung cancer BMI, Self-reported COPD, CXR in last three years Current smoker; pack years smoked, smoking duration Former smokers: years since quit Bach, Spitz, LLP, COSMOS NCCN: Back of the envelope risk Incremental Value Of Pulmonary Function In Lung Cancer Risk Prediction ROC AUC full 0.77 ROC AUC nested 0.72 (Cancer Prev Res; 4(4); ) 20

21 Original Article Selection Criteria for Lung-Cancer Screening Martin C. Tammemägi, Ph.D., Hormuzd A. Katki, Ph.D., William G. Hocking, M.D., Timothy R. Church, Ph.D., Neil Caporaso, M.D., Paul A. Kvale, M.D., Anil K. Chaturvedi, Ph.D., Gerard A. Silvestri, M.D., Tom L. Riley, B.Sc., John Commins, B.Sc., and Christine D. Berg, M.D. N Engl J Med Volume 368(8): February 21,

22 22

23 PLCO Model Versus 30 Pack-Years Pan-Canadian Lung Cancer Screening Study >2% Lung Risk Index N=2537 Cancer N=76 >2% Lung Risk Index and >30 Pack Years N=2306 Cancer N=66 Compares to 30 pack-yrs as threshold, using PLCO Model, 9% <30 pack-yrs but 13% more lung cancers are identified 3% lung cancers in < 2 years of follow-up 3.3% lung cancers age 55 to 74 yrs CISNET Modeling to inform USPSTF Full individual level NLST data to provide to four lung cancer screening modeling teams to validate and improve current models Focus on extrapolating trial results to answer additional questions considered to be of value for medical practitioners, policymakers etc. including: Ages to initiate and end regular screening (encompasses total number of screens) Frequency of screens Impact of longer follow-up and additional screens on mortality reduction Smoking history that determines eligibility for screening CISNET to then work with USPSTF 23

24 NLST Data, Image and Specimen Access Website Access to entire research community Data transfer agreement required Name, , summary of research proposal on website Return of research papers requested for tracking and posting Biospecimens OREPOSITORY.aspx Summary Opportunity to significantly improve outcome of patients with lung cancer Selection of high-risk group to screen critical Research to minimize burden of falsepositives also critical International collaboration to better define the screening process and most costeffective diagnostic and treatment pathways Screening is NOT a substitute for effective tobacco control policies 24

25 Acknowledgements Full list of all collaborators FNEJMoa &viewType=Popup&viewClass=Suppl Special thanks to Martin Tammemägi NLST Executive Committee Denise R. Aberle, MD Christine D. Berg, MD William C. Black, MD Timothy R. Church, PhD, MS Richard M. Fagerstrom, PhD Barbara Galen, MSN, CRNP, CNMT Ilana F. Gareen, PhD Constantine Gatsonis, PhD Jonathan Goldin, MD, PhD Barnett S. Kramer, MD, MPH David Lynch, MD Irene Mahon, RN, MPH Pamela M. Marcus, MS, PhD Dorothy Sullivan Carl J. Zylak, MD National Cancer Institute: DCP, EDRG, Lung Screening Study DCTD, Cancer Imaging Program, Bethesda, MD American College of Radiology Imaging Network, Philadelphia, PA 25

26 NLST Lead Radiologists ACRIN Gerald Abbott MD, Denise Aberle MD, Judith Amorosa MD, Richard Barr MD, William Black MD, Phillip Boiselle MD, Caroline Chiles MD, Robert Clark MD, Lynn Coppage MD, Robert Falk MD, Elliot Fishman MD, Jonathan Goldin MD PhD, Eric Goodman MD, Eric Hart MD, Elizabeth Johnson MD, Phillip Judy PhD, Ella Kazerooni MD, Robert Mattrey MD, Barbara McComb MD, Geoffrey McLennan MD, Reginald Munden MD, James Ravenel MD, Michael Sullivan MD, Stephen Swensen MD, Drew Torigian MD, Kay Vydareny MD, John Worrell MD LSS Peter Balkin MD, Matthew T. Freedman MD MBA, Kavita Garg MD, David S.Gierada MD, Subbarao Inampudi MD, Howard Mann MB BCh, William Manor DO, Hrudaya Nath MBBS DMR MD, David L. Spizarny MD, Diane C. Strollo MD, John Waltz MD NLST Physicists CT Physics Committee Dianna Cody, PhD MD Anderson Cancer Center Mike McNitt-Gray, Phd Christopher Cagnon, PhD Philip Judy, PhD Fred Larke, PhD Randell Kruger, PhD Mike Flynn, PhD Xizeng Wu, PhD UCLA UCLA Brigham and Women s Hospital University of Colorado Marshfield Clinic Henry Ford Hospital CXR Physics Committee J. Anthony Seibert, PhD UC Davis Chris Cagnon, PhD Philip Judy, PhD Randell Kruger, PhD Mike Flynn, PhD University of Alabama UCLA Brigham and Women s Hospital Marshfield Clinic Henry Ford Hospital 26

27 NLST Committees Endpoint Verification Team Anthony B. Miller, MB, Chair, Martin J. Edelman, MD, William K. Evans, MD, Robert S. Fontana, MD, Mitchell Machtay, MD Oversight Committee Robert C. Young, MD, Chair, David Alberts, MD, David DeMets, PhD, Peter Greenwald, MD, Dr PH, Paula Jacobs, MD, Theresa C. McLoud, MD, David P. Naidich, MD, James Tatum, MD Data and Safety Monitoring Board Edward A. Sausville, MD, PhD, Chair, Wylie Burke, MD, PhD, Gene Colice, MD, Brenda Edwards, PhD, Scott Emerson, MD, PhD, John Fletcher*,MD, Sylvan Green*,MD, Russell Harris, MD, MPH, Jeffrey S. Klein, MD, Edward L. Korn, PhD, Robert Mayer, MD, Joe V. Selby, MD, MPH, David W. Sturges, JD, Bruce W. Turnbull, PhD, Thomas J. Watson, MD * Deceased Additional Partners National Cancer Institute: Cancer Imaging Program, DCTD Early Detection Research Group, DCP ACRIN Westat IMS CARE Communications Our many, many site investigators and research staff Colleagues NLST ACRIN Tissue Bank & Biomarker Oversight Committee NLST ACRIN Research Evaluation Panel ACRIN Specimen Biorepository at University of Colorado UCLA Tissue Microarray Laboratory American Cancer Society: in-kind assistance with recruitment Advocates & members of lung cancer community who supported NLST 27

28 With appreciation 53,454 trial participants without whom these studies would not have been possible I told you smoking is bad for you 28