Tracking recurrence In population-based cancer surveillance : Where to start? Literally!

1. Tracking recurrence In population-based cancer surveillance: Where to start? Literally! Loria Pollack, MD, MPH Senior Medical Officer | Epidemiologist CDC National Programs of Cancer Registries NAACCR 2019, Vancouver June 13, 2019

2. Tracking Recurrence • Cancer survivors are concerned about recurrence • Knowledge about recurrence is sparse • Population-based cancer registries do not collect recurrence • Hospital-based registries have known limitations1 • Published studies are not generalizable • Response Evaluation Criteria In Solid Tumors (RECIST)2 • Requires intense follow-up (clinical trials) • Not practical for cancer surveillance • 1In H, et al. J Surg Oncol, 2015. 111(6): p. 657-62; • 2 Eisenhauer, EA, et al., Eur J Cancer, 2009. 45(2): p. 228-47. • Background

3. Patient-Centered Outcome Research • Goal: Collect ‘patient-centered’ outcomes in population-based registries • Cancer progression • Cancer recurrence • Data items, definitions & analysis • Findings & Recommendations • Starting points (Loria) • Ending points (Trevor) • CDC National Programs of Cancer Registries

4. Comparative Effectiveness Research (CER) • 10 US states • Breast, Colorectal, CML diagnosed in 2011 • Capture of additional treatment/biomarkers • New data items (Tobacco use, height/weight) Patient-Centered Outcome Research • Patient-Centered Outcome Research (PCOR) • 5 US states - among CER states • Continued follow-up of 2011 incident CER cases (Breast, Colorectal) • CDC National Programs of Cancer Registries

5. Patient-Centered Outcomes Research Project Active follow-up Minimum 32 months Majority followed for 60 months ~18,000 cases 5 states* Patient-Centered Outcomes Disease-free status Recurrence, Progression, Residual Disease Breast and Colorectal Cancers Diagnosed in 2011 Record abstraction at facilities Remote access to EMRs 2011-2015 * Colorado, Idaho, Louisiana, New Hampshire, Rhode Island

6. ‘Active’ Follow-up • Evidence for outcomes: • Patient history • Physical exam • Laboratory tests • Imaging results • Mammogram • CT scan • MRI • Bone scan • Chest imaging • Colonoscopy • Biopsy • Dedicated staff to collect follow-up data • Abstractor training • Educate on definitions at start of project • Ongoing monthly project meetings • Medical record review • Abstract records on-site • Access electronic record remotely • Contact provider for follow-up • No patient contact • NPCR Patient-Centered Outcome Research

7. Study population

8. Case Count by State Study population NPCR Patient-Centered Outcome Research 17,802 Total Cases 11,769 Breast Cancer 6,033 Colorectal Cancer Coverage of 4.4% of U.S. Reflective of entire U.S.

9. Study Population: Demographic • NPCR Patient-Centered Outcome Research

10. Study Population: Socioeconomic factors From Medical Record From Census Tract • NPCR Patient-Centered Outcome Research

11. Tumor characteristics: Stage at diagnosis • NPCR Patient-Centered Outcome Research One primary only: Breast 73.0%; Colorectal 71.0% Two or more primaries: First primary Breast 8%; Colorectal 8% | Second primary Breast 16%; Colorectal 18%

12. OutcomesDefinitions & Findings

13. Cancer Recurrence • NCI Dictionary of Cancer Terms1 • Cancer that has recurred (come back), usually after a period of time during which the cancer could not be detected. The cancer may come back to the same place as the original (primary) tumor or to another place in the body. • American College of Surgeons (Coding Instructions) 2 • Record the date the physician diagnoses the first progression, metastasis, or recurrence of disease after a disease-free period. • 1https://www.cancer.gov/publications/dictionaries/cancer-terms • 2https://www.facs.org/quality-programs/cancer/ncdb/registrymanuals • Working Definitions

14. Outcome variable definitions • NPCR Patient-Centered Outcome Research

15. NPCR Patient-Centered Outcome Research • Outcome variables

16. Outcomes for Breast Cancer Minimum 32 months active follow-up

17. Outcomes for Breast Cancer Minimum 32 months active follow-up

18. Outcomes for Breast Cancer Minimum 32 months active follow-up

19. Outcomes for Breast Cancer Minimum 32 months active follow-up

20. Outcomes for Colorectal Cancer • Disease Free? • Unknown (If Disease Free) • 301 (5.0%) • Documented Disease Free • 4091 (67.8%) • Never Disease Free • 1641 (27.2%) • Recurrence • 530 (13.0%) • Documented Progression • 701 (42.7%) • Unclear if Ever Disease Free • 201 (66.8%) • No Info. other than Diagnosis • 100 (33.2%) • Residual Disease • 117 (7.1%) • DF until end of study • 3237 (79.1%) • Uncertain if Progression • 823 (50.2%) • Uncertain if recurrence • 324 (7.9%) Minimum 32 months active follow-up

21. Outcomes for Colorectal Cancer • Disease Free? • Unknown (If Disease Free) • 301 (5.0%) • Documented Disease Free • 4091 (67.8%) • Never Disease Free • 1641 (27.2%) • Recurrence • 530 (13.0%) • Documented Progression • 701 (42.7%) • Unclear if Ever Disease Free • 201 (66.8%) • No Info. other than Diagnosis • 100 (33.2%) • Residual Disease • 117 (7.1%) • DF until end of study • 3237 (79.1%) • Uncertain if Progression • 823 (50.2%) • Uncertain if recurrence • 324 (7.9%) Minimum 32 months active follow-up

22. Summary: Outcomes • Determining outcome was dependent on determining DF status • Completeness (97% Breast , 95% Colorectal) • Among DF patients, 97% breast and 92% colorectal were classified as either remaining DF or having a recurrence • Among Never DF patients, it was difficult to distinguish between residual disease and progression; nearly half were classified as ‘unknown’ • IMPACT: We limited outcome analysis to recurrence among DF • NPCR Patient-Centered Outcome Research

23. Time-to-event:Starting point?

24. NPCR Patient-Centered Outcome Research DD/MM/YYYY* • Diagnosis • Surgery • First course therapy • Start • Completion • Patient disease free status • First (earliest) • Last (most recent date patient visited received care and had no evidence of disease) • Initiation of subsequent therapy • Recurrence • Progression • Active follow-up (last date abstractor looked for follow-up medical information for the patient) • Last contact or death • * If day was missing, it was imputed as the midpoint of the month; if month or year was missing, dates were considered missing data. • Date variables

25. NPCR Patient-Centered Outcome Research • Time-to-event

26. NPCR Patient-Centered Outcome Research • Time-to-event definitions

27. NPCR Patient-Centered Outcome Research • Days from diagnosis to surgery Female Breast Cancer (n=9,940) Colorectal Cancer(n=3,894)

28. NPCR Patient-Centered Outcome Research • Days from surgery to documented disease free status Female Breast Cancer (n=9,941) Colorectal Cancer(n=3,895) A B C D E A B C D E

29. Variation in Capture of Disease-Free Status • Misinterpretation of abstraction guidance • Differences in determination of DF status • Timing of follow-up surveillance of patient • Access/ease of medical record review • Documentation or interpretation of DF • Stage at diagnosis (localized vs/ regional or distant) • Potential Reasons

30. NPCR Patient-Centered Outcome Research • Time-to-event definitions • Disease-Free date was not a reliable start point • Inconsistent • Large variation among states • Surgery was chosen as start point for time-to-outcome analysis • Defined event • Limitations: • Excludes patients who did not undergo surgery • Patient may not be disease-free at the surgery (adjuvant therapy)

31. Summary • Large, comprehensive efforts conducted by population-based cancer registries • Detailed data on intermediate cancer outcomes by extensive medical record review • Recurrence: 6.1% of female breast cancer and 13.0% of colorectal cancer • Recurrence incidence varied by stage, as expected • Capture of Disease-Free (DF) Status was a pivotal… • By definition, cancer recurrence is the return of cancer after a period of remission • Abstractors were able to determine DFS for over 95% patients • …but not as a starting point for time-to-outcome analysis • Concerns on quality and consistency • Needed better definition and QC during study • Surgery date was most reliable • NPCR Patient-Centered Outcome Research

32. Other Lessons Learned • Capture of the following dates were important to outcome analysis(T Thompson • Recurrence; Death • Active Follow-up • Last DF date • Multiple sources were used to determine recurrence • Physician-offices were only source for 5% (colorectal) and 6% (breast) of recurrences • Need to further understand ‘evidence’ used to determine DF and outcomes • Outcomes in “Never DF” patients were challenging to determine • 43.7% of breast and 50.2% of colorectal were unknown if progression or residual disease • Our analyses did not “lump” progression with recurrence • For future studies: Focus on those most at risk for recurrence (Stages II-III)? • NPCR Patient-Centered Outcome Research

33. Future analyses • Multivariate analysis of recurrence outcome • Patient information • Smoking • BMI • Social Determinants • Treatment received (?) • Comparison of our data to novel approaches • Claims-based algorithms • Linkage • Predictive models • NPCR Patient-Centered Outcome Research

34. Funding: This work was supported by the Patient Centered Outcome Research Trust Fund through CDC Cooperative Agreements of the National Program of Cancer Registries: in conjunction with the participating states and a CDC Patient Centered Outcome Research contract to ICF Co-investigators/Acknowledgements