Brittany Avin McKelvey

PURPOSE: Cancer registries are often asked to present cancer data for small geographic areas to inform and facilitate targeted interventions and prevention programs. However, it is challenging to compute and visualize reliable cancer estimates for areas with small case counts and populations to support cancer control planning. METHODS: Leveraging a user-centered design process, we developed a visual analytics platform and interactive graphics to display modeled cancer risk estimates for small areas. Development of our visual analytics platform was informed by cancer registry and public health professionals through focus groups and surveys. The reliable cancer risk estimates for small areas that we displayed on this platform were created using a Bayesian hierarchical model that borrows strength from neighboring areas and over time to produce cancer estimates for small areas. RESULTS: The Cancer Analytics and Maps for Small Areas tool (CAMSA) provided age-adjusted cancer incidence and mortality rates and risk probabilities for eight cancers at the county and ZIP-code tabulation area levels. It allowed the user to identify areas of high cancer incidence, including among subgroups defined by sex and race/ethnicity. Potential end users were enthusiastic about the opportunity to implement CAMSA within their practice, emphasizing the tool's potential for increased collaborative opportunities at local and state levels. Suggestions for improvement included adding map overlays such as additional cancer risk variables and incorporating functionalities such as exportable data tables. CONCLUSION: CAMSA presented cancer rate and risk estimates for small geographic areas where they may have previously been suppressed. Through our user-centered design process, we developed statistical models and data visualizations to support the needs of an array of potential end users.

Supplementary Data DS1 from Breakthrough Therapy Designation Criteria Identify Drugs that Improve Clinical Outcomes for Patients: A Case for More Streamlined Coverage of Promising Therapies

Supplementary Figure S3. Kaplan-Meier Curves for OS and PFS by Baseline ctDNA Detection and Multivariable Associations. Kaplan-Meier plot for (A) OS and (B) PFS and baseline ctDNA detection vs. non-detection, landmarked at 70 days from index; patients with an event during the 70-day landmark were excluded from the analysis. Data presented include the Training and Test datasets.

Supplementary Figure S4. Kaplan-Meier Curves for OS and PFS by Baseline ctDNA Detection and Multivariable Associations. Kaplan-Meier plot for (A) OS and (B) PFS and baseline ctDNA detection vs. non-detection, landmarked at 70 days from index; patients with an event during the 70-day landmark were excluded from the analysis. Data presented include the Training and Test datasets.

Supplementary Figure S1. Kaplan-Meier Curves for OS and PFS by Percent Change in ctDNA Levels and Multivariable Associations. Kaplan-Meier plot for (A) OS and (B) PFS and percent change in ctDNA levels from T0 to T1, landmarked at 70 days from index; patients with an event during the 70-day landmark were excluded from the analysis. Data presented include the Training dataset only.

Supplementary Figure S2. Comparison of Training vs. Test Datasets for Research Objective 1. Training (A,B) vs. test (C,D) datasets for research objective 1. For OS, the multivariable Harrel’s C-Index for Training was 0.749 and for Test was 0.697 while the Uno’s C-Index was 0.736 for Training and 0.645 for Test. For PFS, the multivariable Harrel’s C-Index for Training was 0.653 and for Test was 0.599 while the multivariable Uno’s C-Index was 0.630 for Training and 0.589 for Test.

Supplementary Figure S6. Training vs. Test Datasets for Early Radiographic Response Combined with ctDNA Categories. Training (A,B) vs. Test (C,D) datasets for early radiographic response combined with ctDNA categories. For OS, the multivariable Harrel’s C-Index for Training was 0.753 and for Test was 0.723 while the Uno’s C-Index was 0.744 for Training and 0.660 for Test. For PFS, the multivariable Harrel’s C-Index for Training was 0.649 and for Test was 0.603 while the Uno’s C-Index was 0.646 for Training and 0.595 for Test.

Supplementary Figure S5. Training vs. Test Datasets for Early Radiographic Response and OS/PFS. Training (A,B) vs. Test (C,D) datasets for early radiographic response. For OS focused on early radiographic response, the multivariable Harrel’s C-Index for Training was 0.695 and for Test was 0.688 while the Uno’s C-Index was 0.708 for Training and 0.661 for Test. For PFS focused on early radiographic response, the multivariable Harrel’s C-Index for Training was 0.621 and for Test was 0.541 while the Uno’s C-Index was 0.623 for Training and 0.535 for Test.

Supplementary Table S1. Clinical trials included in the analysis. This table summarizes the clinical trials included in the analysis, detailing the study name, sponsor, trial type, tumor type, treatment, and ctDNA assay used. References to the original trial publications are also provided.

Purpose: Circulating tumor DNA (ctDNA) holds promise as an early endpoint to predict overall survival (OS). The creation and structured interrogation of aggregated datasets inform the hypothesis that ctDNA is reasonably likely to predict treatment benefit. Friends of Cancer Research convened a diverse working group to establish and implement an analysis plan assessing patient-level associations between changes in ctDNA levels with OS and progression-free survival (PFS). Experimental Design: The aggregate dataset included eight clinical trials representing 940 patients with biomarker-positive advanced non–small cell lung cancer treated with tyrosine kinase inhibitors. Detection of baseline and on-treatment ctDNA was assessed for associations with OS and PFS. Additionally, combinations of ctDNA detection and RECIST measurements up to 10 weeks on treatment were considered. Results: Patients with detected ctDNA at baseline that became nondetected on treatment (“clearance”) experienced improved OS compared with patients with persistently detected ctDNA (adjusted HR = 2.12, P < 0.001). This pattern was also seen in the subset of patients with stable disease as measured by RECIST within 10 weeks of treatment initiation (adjusted HR = 4.15, P < 0.001). Results were similar for PFS. Conclusions: In patients with oncogene-driven advanced non–small cell lung cancer treated with tyrosine kinase inhibitors, ctDNA clearance within 10 weeks of treatment initiation was associated with improved OS and PFS. These patient-level results support the growing evidence that demonstrates a change in ctDNA levels during treatment is associated with clinical benefit. Future prospective trials should include predefined thresholds of molecular response to advance the utility of ctDNA as an early endpoint.