General principles of screening: A radiological perspective Fergus Coakley MD, Professor and Chair, Diagnostic Radiology, Oregon Health and Science University
General principles of screening: A radiological perspective Fergus Coakley MD, Professor and Chair, Diagnostic Radiology, Oregon Health and Science University Background Interest dates from whole body CT era: January 23, 2005 The Full Body Scan can detect the presence of many common forms of cancer in their early stages giving the head start needed for early intervention. Closed after civil lawsuit for deceptive advertising CA license now surrendered Learning objectives What is screening? Describe scientific basis and rationale for screening Review sources of bias and appropriate endpoints in evaluating screening effectiveness Screening is the systematic application of a test or inquiry, to identify individuals at sufficient risk of a specific disorder to benefit from further investigation or direct preventive action, among persons who have not sought medical attention on account of symptoms of that disorder. J Med Screen. 2008; 15: 50
Conceptual basis What is early? 56 year old smoker with right sided weakness Early diagnosis Better outcome Metastasis may occur at 1-2 mm diameter Cancer--Principles & Practice of Oncology Science 1999; 284: 1994-1998 Goldilocks tumors Goldilocks tumors Aggressiveness Aggressiveness Screen Clinical Screen Clinical diagnosis diagnosis diagnosis diagnosis Incurable Incurable Improved outcome Curable No improved outcome Curable Time Time
Effectiveness of screening Cancer Technique Decrease in mortality Cervix Pap smear 79% Colon Colonoscopy 44% Breast Mammography 21% Lung Chest CT 20% Lung Chest x-ray 0% Obstet Gynecol 1995; 85: 1017-1021 (cervix) Cancer Causes Control 1998; 9: 455-462 (colon) Lancet 2002; 359: 909-919 (breast) J Natl Cancer Inst 2000; 92: 1308-1316 (lung - CXR) N Engl J Med 2011; 365: 395-409 (lung CT) Evaluating outcome Site-specific is the appropriate endpoint for evaluating screening interventions Cannot use 5 year survival: Lead time bias Length time bias Upward stage migration J Med Screen 2007; 14: 178-85 Lead time bias Length time bias SCREEN DIAGNOSIS CLINICAL DIAGNOSIS Subclinical SCREEN Clinical Subclinical Clinical Time Lead time Survival appears better after screen diagnosis Clinical diagnosis: 2/4 indolent (50%) Screen diagnosis: 2/3 indolent (67%) Screen diagnosis always more indolent Slow moving target is easier to hit
Pseudodisease Upward stage migration Microscopic cancer is common Old technology New technology Cancer Autopsy frequency Prostate cancer 40% Renal cell carcinoma 22% Thyroid cancer 6-36% Cancer 1993; 71(3 Suppl): 933-938 AJR 2001; 177: 989-992 Am J Clin Pathol 1988; 90: 72-76 Cancer 1985; 56: 531-538 Subclinical Stage I Stage II Etc. Stage I Stage II Stage III Etc. New stage I appears to have a better prognosis than old stage I, and so on. Will Rogers effect When the Okies left Oklahoma and moved to California, they raised the average intelligence level in both states. SCREENING TREATMENT N Engl J Med 1985;312:1604-1608
Conclusions Thank you Screening for cancer is widely assumed to be highly effective In fact, screening is based on an imperfect concept and requires rigorous cancer specfic mortality based evaluation For now, screening remains of limited but proven benefit for selected cancers coakleyf@ohsu.edu
Karen Y. Oh, MD Chief of Breast Imaging, Associate Professor Department of Radiology Breast cancer screening: Data and Recommendations
Breast cancer screening: Data and Recommendations Importance of screening mammography Breast cancer is the 2 nd leading cause of cancer death in US women NCI estimates lifetime risk of 12% based on SEER data 230,000 estimated new cases in 2013 Karen Y. Oh, MD Chief of Breast Imaging, Associate Professor Department of Radiology USPSTF guidelines. Ann Intern Med 2009. http://seer.cancer.gov/statfacts/html/breast.html HIP trial Health Insurance Plan NY AGE trial looked at screening 40-49 yrs ~61,000 women 40-64 yrs between 1963-1966 25% mortality reduction with annual screening 40-64yrs after 18 years follow-up About 67% of the invited group showed up for initial screening Shapiro S. Periodic screening for breast cancer; the HIP project and its sequelae. 1963-1986. J Natl Cancer Inst Monogr 1997 (22): 27-30 160,921 women 39-41 yrs between 1991-1997 24% mortality reduction with annual screening 40-48 yrs after removing those women who did not participate in first round About 68% of the invited group showed up for initial screening Moss SM et al. Effect of mammographic screening from age 40 years. Lancet 2006; 368:2053-2060.
Tabar Swedish 2-county trial, 2011 Longest running breast screening trial (29+yrs) 30% reduction in breast 40-74 yr old women invited to screening Absolute benefit in mortality increases with longer follow-up times Screening 300 women for 10 years prevents 1 death from breast cancer When no shows were placed in the control group, there was a 63% decrease in the death rate of the screened group Tabar et al, Swedish 2-county trial. Radiology 2011. British Columbia trial service screening results Women 40-79 yrs between Jan 1988 Dec 2003 598,690 women underwent average of 3.7 screenings during study analysis 40% mortality reduction Coldman A, et al. Breast after screening. Int J Cancer 2007. USPTSF 2009 Screening ages 50-69 yrs produced projected 17% (15-23%) reduction in mortality (2009) USPSTF recognizes benefit of screening seems equivalent for ages 40-49 and 50-59 Incidence of breast cancer less in 40-49 yr age group Mammography reduces mortality by 15% in 39-49 yr age group Randomized controlled trial Study group realities cause dilution of benefits Screened with mammography Invited to screen Did not get any mammograms Control group Randomized controlled trials Screening and USPTSF calculations Actually based on invitation off protocol to screen (NNI), screened not actual number of women screened Demissie et al. J Clin Epidemiol 1998; 51:81-91.
NNI = # of women needed to invite to screening to prevent 1 death Many women randomized to invited group do not screen Some will screen then fail to continue NNI assumes all the women in the invited group are assumed to have received screening (not true) NNS = # of women needed to screen to prevent 1 death Counts the women who actually screen only Looks at the mortality benefit in those who get screening mammography Demissie et al. J Clin Epidemiol 1998; 51:81-91. Hendrick, Helvie. AJR 2012; 198:723-728. Demissie et al. J Clin Epidemiol 1998; 51:81-91. Hendrick, Helvie. AJR 2012; 198:723-728. Hendrick and Helvie Used same data set as USPTSF 39.6% mortality reduction with annual screening 40-84yrs 71% more lives saved than USPSTF regimen 64,889 more lives at 65% compliance 84 women needed to be screened annually 40-84 years to save 1 life Hendrick, Helvie. AJR 2012; 198:723-728. Hendrick, Helvie. AJR 2011; 196:W112-W116. Harms of mammography Psychological harms Unnecessary imaging tests Unnecessary biopsies in women without cancer Pain associated with procedure Inconvenience due to false-positive screening results * These factors have been shown to have minimal effect on patients returning to screening Armstrong et al. Screening mammography in women 40-49 years. Ann Intern Med, 2007 Nelson et al. Risk Factors for Breast Cancer for Women Aged 40-49 Years. Ann Int Med, April 2012.
Harms of mammography Overdiagnosis Radiation exposure Minimal risk over lifetime of screening Annual 2 view screening age 40-80 has smaller lifetime attributable risk than a single pelvic, abdomen or chest CT Overdiagnosis: Detection of a cancer that would not become clinically apparent during a woman s lifetime Overtreatment: Unnecessary earlier treatment of breast cancer that would have become apparent, but would not shorten a woman s life Hendrick. Radiation doses and cancer risks from breast imaging studies. Radiology, 2010. The cancers which will kill you cannot be reliably predicted mammographically (or with any imaging) Low grade tumors or DCIS could be slower growing, but we don t know how to predict which ones will eventually become deadly Estimated to be 0.5%-52% Duffy et al. Absolute numbers of lives saved and overdiagnosis. J Med Screen 2010. Jorgensen KJ, Goetzsche PC. Overdiagnosis in screening. BMJ 2009. Overdiagnosis Where has this confusion led us? Not even looking for cancers or diagnosing cancer is not the answer In the future can we find them and test for future deadly potential? Harms of overtreatment could be mitigated Failure Analysis analyzing a "failure" (death from breast cancer) and why it occurred Estimated effect of screening mammography on mortality by review of the screening histories of women who died from breast CA Webb ML, Cady B, Michaelson JS, et al. A Failure Analysis of Invasive Breast Cancer. Cancer 2013.
Where has this confusion led us? What s next? 609 confirmed breast cancer deaths 70.8% of deaths from breast cancer occurred among the women who were unscreened (~20%) 50% of deaths in women under 50 yrs 13% of deaths in women age 70+ Median diameter, grade, LN mets prognostically worse in unscreened women November 14, 2013 USPTSF issued a draft research plan to guide a systematic review of evidence on breast screening Basis for new task force recommendation Will include 3D mammography, MRI, US, CBE Webb ML, Cady B, Michaelson JS, et al. A Failure Analysis of Invasive Breast Cancer. Cancer 2013. http://www.uspreventiveservicestaskforce.org/draftresplan2.htm Conclusion: Conclusion: Providers should discuss benefits and potential harms of screening with women beginning at age 40 Including the reduced chance of dying from breast cancer Mortality benefits are maximized by screening annually beginning at age 40 Meta-analysis of trials underestimate true effectiveness of mammography. Compliance and contamination biases Population-based evaluations of women actually screened usually show reductions in mortality greater than those found by randomized controlled trials.
Conclusion: Regarding statistical modelling "Given enough time, enough attempts, and enough imagination, almost any pattern can be teased out of any data set. Andrew Lo, Charles E. and Susan T. Harris Professor, Professor of Finance, Director of the Laboratory for Financial Engineering at the MIT Sloan School of Management.