Decision Analysis of Colorectal Cancer Screening Tests by Age to Begin, Age to End, and Screening Intervals: Report to the United States Preventive Services Task Force from the Cancer Intervention and Surveillance Modeling Network (CISNET) Agency for Healthcare Quality and Research September 8, 2008 MISCAN Memorial Sloan-Kettering Cancer Center - Ann Zauber Erasmus MC - Marjolein van Ballegooijen, Iris Lansdorp-Vogelaar, Janneke Wilschut SimCRC University of Minnesota – Karen Kuntz Massachusetts General Hospital – Amy Knudsen

What CMS reimbursement for a new CRC test? 2003 and 2007 Stool DNA test? $ to be determined $4.54 $22.22 National Coverage Determination (NCD) on stool DNA (PreGen- Plus test, version 1.1 every 5 years for average risk population)

Questions addressed by CISNET for USPSTF 2007  USPSTF addresses updates for 2002 colorectal cancer screening recommendations  Evidence based literature review  Task Force requested a decision analysis for recommended CRC screening tests for  age to begin  age to end  rescreening interval  Should the current recommendations be changed?  Microsimulation models (MISCAN and SimCRC) of CISNET consortium used for the decision analysis to inform health policy

Adenoma to Carcinoma Pathway Normal Epithelium Small Adenoma Colorectal Cancer Advanced Adenoma

Microsimulation Modeling of Adenoma Carcinoma Sequence with Potential Interventions adenoma 6-9 mm adenoma >=10 mm ADENOMA Preclinical screen-detectable adenoma phase No lesion adenoma <=5 mm preclinical stage I preclinical stage II preclinical stage III preclinical stage IV Preclinical CANCER screen-detectable cancer phase clinical stage I clinical stage II clinical stage III clinical stage IV Clinical CANCER phase death colorectal cancer Datasources: Adenoma Autopsy studies Colonoscopy studies Preclinical Cancer Dwell time Clinical Cancer SEER Incidence Death US Mortality Screening

Colorectal Cancer Screening Strategies Current Age and Interval Recommendations* Age Begin 50 Screening Tests Hemoccult II Hemoccult SENSA FIT Flex Sig Flex Sig + SENSA Colonoscopy Rescreening Intervals 1 – FOBT 5 – Flex Sig 10 - Colonoscopy Age End None Surveillance No stop age * MultiSociety and ACS

Colorectal Cancer Screening Strategies Cohort of 40 year olds in 2005 Age Begin Screening Tests Hemoccult II Hemoccult SENSA FIT Flex Sig* Flex Sig* + SENSA Colonoscopy (No Screening) * With biopsy Rescreening Intervals 1,2,3 – FOBT 5,10,20- Endos Age End Surveillance** No stop age Adherence 100% ** 3 year for advanced adenomas, 5-10 (5) for non-advanced adenomas 145 Test Strategies

>10mm6-9mm<5mm CRC SensitivitySpecificity Percent Hemoccult II Hemoccult SENSA FIT Sigmoidoscopy Sig + Hemoccult SENSA Colonoscopy Adenoma Sensitivity by Size Sensitivity and Specificity of Tests from Literature Review Sigmoidoscopy sensitivity for lesions within range

Screening Test Costs $ per Test USPSTF requested NOT to use costs

Outcome Measures  Most effective = Greatest life years gained relative to no screening  Weigh effectiveness against resources required and exposure to risks: Colonoscopy as resource and risk indicator  Endoscopy resources  Perforation risk  Life years gained (LYG) vs Total colonoscopies in lifetime (per 1000 persons in population).

Effectiveness-Risk Analysis  Determine efficient strategies for each test  Plot life years gained versus colonoscopies required  If strategy requires more colonoscopies but has fewer life years gained (LYG) (ie less effective) then eliminate  Of the remaining strategies, rank by increasing effectiveness (LYG) Derive relative to each other:  Incremental number of colonoscopies = ΔCol  Incremental LYG = ΔLYG  Incremental number colonoscopies to gain a life yr = ΔCol/ ΔLYG  Efficiency Ratio of measure of the additional number of colonoscopies required to gain one year of benefit when considering a more effective strategy relative to the next less effective strategy  Efficiency frontier – all strategies NOT dominated (eliminated)  Near the efficiency frontier – those strategies that are with 98% of the LYG on the frontier

Colonoscopy-MISCAN

Colonoscopy-SimCRC

Efficient Colonoscopy Strategies Strategy*# Col (per 1000) # LYG (per 1000)  Col (per 1000)  LYG (per 1000)  Col/  LYG MISCAN 1 COL, 60-75, 202, COL, 50-75, 203, , COL, 50-75, 104, COL, 50-85, 104, COL, 50-75, 55, , COL, 50-85, 56, SimCRC 1 COL, 60-75, 201, COL, 50-75, 202, , COL, 50-75, 103, COL, 50-85, 104, Dominated 5COL, 50-75, 55, , COL, 50-85, 56, * Test, begin age – end age, interval  Col = incremental number of colonoscopies compared with the next best strategy  LYG = incremental number of life years gained compared with the next best strategy

Age to End Screening  No prior recommendations on stop age for CRC screening  Age 75 and 85 considered  Comorbidity and life expectancy rather than chronological age important  Example for colonoscopy: If start screening at age 50 and stop at age 75  Negative colonoscopy at age 50, 60, and 70  3 negative exams before stopping  Those with adenomas or colorectal cancer detected at screening colonoscopy would be in a surveillance program with no stopping age

Hemoccult II-MISCAN

Hemoccult SENSA-MISCAN

FIT-MISCAN

Flexible Sigmoidoscopy-MISCAN ,0001,5002,0002,500 Colonoscopies per 1,000 persons Life-yeasr gained per 1,000 persons flexible sigmoidoscopy strategiesFrontierstart age 40Frontier , , , ,5

Combination-MISCAN

Comparisons Among Tests without comparator of costs  To compare among tests, it is important to consider that tests other than colonoscopy are required (ie, FOBT, Flex Sig)  To pick an efficient strategy for each test we would expect to find an ordering to the efficiency ratios as follows: COL > SENSA > [FIT, HII] > [FSig, FSig+SENSA]  Eg, SENSA should require fewer colonoscopies to gain a benefit of 1 year compared with COL because of the added number of FOBTs needed in addition to the colonoscopies to achieve that benefit.

Approach to Choosing Efficient Strategies  Assume that a single start and end age would be recommended for screening  Select strategies from all tests (including combination of tests) that: 1.are efficient (or near efficient) within the test 2.have efficiency ratios with expected ordering (to account for the burden of other testing) 3.have comparable effectiveness (LYG)  Example: start age = 50; stop age = 75; anchored with 10-year colonoscopy (as a starting strategy)

Efficient (near efficient) strategies for start age 50 and stop age 75-(Table 9 bolded strategies) Strategy* # Col (per 1000) # LYG (per 1000)  Col/  LYG # FOBT# Fsig MISCAN COL, 50-75, 104, SENSA ®, 50-75, 13, ,5410 FIT, 50-75, 12, ,7720 Hem II ®, 50-75, 11, ,2320 Fsig, 50-75, 51, ,139 FsigSENSA ®, 50-75, 5,32, ,145 SimCRC COL, 50-75, 103, SENSA ®, 50-75,12, ,5730 FIT, 50-75,12, ,8300 Hem II ®, 50-75, 11, ,2390 Fsig, 50-75, ,483 FsigSENSA ®, 50-75, 5,31, ,679

Sensitivity Analysis  Comparative modeling (2 models) give similar results  Similar rankings of strategies even if assume better or worse estimates on sensitivity and specificity  Adherence varied from 100%, 80%, 50%

MISCAN Adherence Plot

CONCLUSIONS  Current recommended guidelines* are on or close to the efficiency frontier  Beginning at age 50 balances life years gained and number of colonoscopies required and associated risk of perforation  To increase efficiency of current guidelines*, stop screening at age 75  should depend on life expectancy of person rather than strict chronological age *MultiSociety and ACS

CONCLUSIONS (Continued 1)  Annual SENSA or FIT have similar LYG as colonoscopy every 10 years but with lower colonoscopy requirements – PROVIDED high compliance for all tests  FlexSig every 5 years with annual FOBT with Sensitive FOBT not recommended (high efficiency ratio)  Original strategy for Flex Sig+ FOBT was for Hemoccult II with lower sensitivity  Combination of Flex Sig and Hemoccult SENSA ® could have one mid-interval FOBT between the 5 year repeat Flex Sig screening rather than annual FOBT  FlexSig every 5 years and Hemoccult II not as good in terms of effectiveness

CONCLUSIONS for Adherence  Adherence conclusions  Life years gained and colonoscopies decreased with decreased adherence BUT  The overall conclusions did not change substantially as adherence varied from 50% to 100%.  Hemoccult II and flexible sigmoidoscopy every 5 years remained the least two attractive alternatives re life years gained  Colonoscopy every 10 years improved a bit relative to the other strategies when adherence was 80% but lost its health benefit advantage when adherence as 50%

Limitations  Analyses for whole population – not specific by sex or race  Potential of more proximal disease in older women and blacks  Age of onset may vary by sex and race  Inadequate data on adenoma prevalence age 40  Chronological age rather than life expectancy  Life expectancy of men: 10.5 at age 75 and 5.9 at 85  Life expectancy of women: 12.5 at age 75 and 7.0 at 85  Simulation models rely on assumptions of natural history of disease  Comparing two models provides sensitivity analysis of natural history assumptions

‘Best’ Test is the One Which Gets Done- SJ Winawer re Adherence

Thank You Mary Barton, William Lawrence of AHRQ Diana Pettit, Michael LeFevre, George Isham, and Steve Teutsch of USPSTF Acknowledgements

Screening and Treatment Costs by Screening Strategy Hemoccult II Hemoccult SENSA FIT No Screening Sigmoidoscopy Sig + Hemoccult SENSA Colonoscopy Per 1000 screened

Hemoccult II Screening TestPolyp Resection and PathologyFollow-up of Positive Test SurveillanceComplications Hemoccult SENSA FIT Sigmoidoscopy Sig + Hemoccult SENSA Colonoscopy Components of Screening Costs (per 1000 screened) (CMS analysis age 65+)

Model Calibrations  Process of matching model output with data  Useful when data aren’t available to estimate certain model parameters but are available on model outcomes  Compare model output with empirical data 1.Prevalence and number of adenomas (autopsy studies) 2.Location and size of lesions (colonoscopy studies) 3.Incidence, location, and stage of diagnosed CRC (SEER)

SENSA ®, 50-75,1 Specificity of 92.5% (base case) vs 87% (ER) Colonoscopy 50-75,10 given as comparator Strategy MISCAN # Col (per 1000) # LYG (per 1000) Colonoscopy, 50-75, 104, SENSA ®, 50-75,1 (basecase)3, SENSA ®, 50-75,1 (SA at 87% specificity) 3,832 (+14%)233 (+1%) Strategy SimCRC # Col (per 1000) # LYG (per 1000) Colonoscopy, 50-75, 103, SENSA ®, 50-75,1 (basecase)2, SENSA ®, 50-75,1 (SA at 87% specificity) 3,104 (+17%)263 (+1.5%)

Efficient Strategies for start age of 50 and stop age of 75 (Table 9 Page 31) Strategy*# Col (per 1000) # LYG (per 1000)  Col (per 1000)  LYG (per 1000)  Col/  LYG MISCAN COL, 50-75, 104, SENSA ®, 50-75, 13, FIT, 50-75, 12, Hem II ®, 50-75, 11, Fsig, 50-75, 51, FsigSENSA ®, 50-75, 5,32, SimCRC COL, 50-75, 103, SENSA ®, 50-75,12, FIT, 50-75,12, Hem II ®, 50-75, 11, Fsig, 50-75, FsigSENSA ®, 50-75, 5,31, * Test, begin age – end age, interval  Col = incremental number of colonoscopies compared with the next best strategy  LYG = incremental number of life years gained compared with the next best strategy

Efficient Strategies for start age of 50 and stop age of 75 Rank order of strategies Strategy*# Col (per 1000) # LYG (per 1000)  Col/  LYG MISCAN COL, 50-75, 104,136 (1)230 (1)29.6 SENSA ®, 50-75, 13,350 (2)230 (1)30.9 FIT, 50-75, 12,949 (3)227 (4)25.9 Hem II ®, 50-75, 11,982 (5)194 (6)14.3 Fsig, 50-75, 51,911 (6)203 (5)9.7 FsigSENSA ®, 50-75, 5,32,870 (4)230 (1)16.3 SimCRC COL, 50-75, 103,756 (1)271 (1)34.7 SENSA ®, 50-75,12,654 (2)259 (2)22.9 FIT, 50-75,12,295 (3)256 (4)19.7 Hem II ®, 50-75, 11,456 (5)218 (5)9.6 Fsig, 50-75, (6)199 (6)8.4 FsigSENSA ®, 50-75, 5,31,655 (4)257 (3)7.0 * Test, begin age – end age, interval  Col = incremental number of colonoscopies compared with the next best strategy  LYG = incremental number of life years gained compared with the next best strategy

Comparisons  First compare strategies within a screening test  Efficient frontier derived for each screening test or combination test   Col/  LYG – ‘Efficiency Ratio’  A measure of the additional number of colonoscopies required to gain one year of benefit when considering a more effective strategy relative to the next less effective strategy  Colonoscopy resources across tests are comparable but burden of all testing is not