Oral Presentation, NAACCR/IACR Completeness of molecular testing information among metastatic colorectal cancer patients – A comparison of data sources at an academic medical center Amanda Kahl, MPH Mary E. Charlton, PhD, Aaron D. Bossler, MD, PhD, Bradley D. McDowell, PhD, Elizabeth A. Chrischilles, PhD Oral Presentation, NAACCR/IACR June 13, 2019

Background Targeted molecular cancer therapies are an important advancement in cancer treatment Overarching research question: How well have they been disseminated to eligible cancer patients? Targeted molecular cancer therapies are an important advancement in cancer treatment - And have been shown in many cases to extend the lives of patients

Background – the Example of Metastatic Colorectal Cancer (CRC) National Comprehensive Cancer Network (NCCN) 2019 guidelines state that any patient with metastatic CRC should be tested for RAS (KRAS and NRAS) mutations BRAF mutations Microsatellite instability (MSI) Determines if patients are eligible for certain targeted therapies and immunotherapies Depending on the results of these tests determines…

Molecular tests to perform for each therapy   Molecular tests to be performed Therapies KRAS NRAS HRAS BRAF MSI EGFR inhibitors Cetuximab ✔ Panitumumab PD-1 inhibitors Ipilimumab Nivolumab Pembrolizumab VEGF inhibitors Bevacizumab Ramucirumab Ziv-aflibercept Multi-kinase inhibitor Regorafenib BRAF inhibitor Vemurafenib This chart is based on 2019 guidelines.

Major issues in addressing this research question: How well have they been disseminated to eligible cancer patients? Cancer registries Do not capture all of these tests Do not capture all of the names of molecular therapies These tests and therapies are not always performed or given during first course treatment Hospital data sources (EMRs, pathology reports, etc) often only capture the care they had at that particular institution What is the best combination of data sources to answer this? What is the best combination of data sources to answer this? … Well, there is one potential data source -> flip slide

PCORnet CDM PCORI (Patient-Centered Outcomes Research Institute) funded a number of institutions across the US to form the PCORnet CDM (Common Data Model) CDM includes information on diagnosis, procedures, prescribing, dispensing, medication administration, and death data tables Extracted from institutional billing and electronic health record data Biomarker testing pulled via CPT, HCPCS codes Molecular-guided therapies pulled via RxNorm Concept Unique Identifier (RXCUI), National Drug Code (NDC), CPT, HCPCS The National Patient-Centered Clinical Research Network (PCORnet) Current Procedures Terminology (CPT) Healthcare Common Procedure Coding System (HCPCS)

Purpose Discuss the completeness of molecular testing information across different data sources for metastatic CRC patients seen at the University of Iowa Hospitals and Clinics Data Sources Manual review of electronic medical records (EMR) PCORnet CDM: CPT, HCPCS, prescribing, dispensing, etc. tables Statewide Iowa Cancer Registry

Computational extraction Molecular testing Are patients receiving testing? What testing are patients receiving? Compare results across different sources Molecular therapy How well are therapies being disseminated? Does treatment align with test results? Compare results across sources Computational extraction Can molecular pathology data be pulled electronically?

Methods: Study population Patients diagnosed with stage IV CRC or stage I-III CRC and had metastatic ICD-9/10 code, 2013-2016 N=197 Non-appendix cases N=182 Did not have multiple tumors N=174 Metastatic disease occurred before Dec. 31, 2017 N=172 Had evidence of metastatic disease in medical record N=163 Met with UIHC Medical Oncologist N=138 Excluded: cancer of appendix N=15 Excluded: multiple tumors N=8 Excluded: stage I-III metastatic disease occurred after Dec. 31, 2017 N=2 Excluded: stage I-III no evidence of metastatic disease in medical record N=9 Excluded: did not meet with UIHC medical oncologist N=25 Primary med onc at UIHC, N=105 Only received surgery + med onc consult, N=14 Only received med onc consult, N=19 This last criteria was added because there were patients who received only surgery at the University and therefore had no reason to have complete records of molecular testing or targeted therapy; inclusion of these patients in the denominator would have led to an underestimation of rates of use of molecular testing and targeted therapy This left us with 138 patients.

Methods Biomarkers collected KRAS, HRAS, NRAS, BRAF, MSI Molecular-guided therapies Cetuximab, panitumumab, pembrolizumab, nivolumab, ipilimumab Other targeted therapies Bevacizumab, regorafenib, ramucirumab, ziv-aflibercept, vemurafenib

Results

Cohort Majority male, White, married and had private insurance Tumors were mostly of the colon (75%) adenocarcinoma histology (83%) moderately differentiated (54%) diagnosed at stage IV (66%)

Computational extraction Molecular testing Are patients receiving testing? What testing are patients receiving? Compare results across different sources Molecular therapy How well are therapies being disseminated? Does treatment align with test results? Compare results across sources Computational extraction Can molecular pathology data be pulled electronically?

Molecular testing received for metastatic CRC patients, all data sources [Go through first graph] The percentage of patients who received the RAS and BRAF tests aren’t all the same. Some patients only received KRAS testing and many patients received multi-panel testing that included KRAS, NRAS, HRAS, and BRAF, but not all. NRAS and BRAF were not required for these anti-EGFR therapies at the start of our study timeframe of 2013. Second graph: After working with different datasets involving CRC I thought it would be of interest to see how many patients receive MSI-DNA testing vs MMR IHC testing. As you can see 45% of patients received MMR testing only, 7% MSI testing only, and 28% received both.

Molecular testing results for metastatic CRC patients, all data sources The results of the molecular testing are fairly consistent with what has been reported elsewhere. 40% of CRC are reported to have a KRAS mutation 2-5% NRAS mutation 5-9% BRAF mutation 3-7% of stage 4 patients are reported to be MSI-H

Comparison of KRAS testing for metastatic CRC captured by EMR review vs Iowa Cancer Registry   Iowa Cancer Registry KRAS Test Results Abnormal Normal Test not done/ unknown TOTAL EMR KRAS Test Results 25 16 41 1 26 22 49 Test not done/unknown 2 45 48 28 27 83 138 65% A 25% difference between the two sources. Of those that were missed by one or the another…[switch slide] 40%

Comparison of KRAS testing for metastatic CRC captured by EMR review vs Iowa Cancer Registry   Iowa Cancer Registry KRAS Test Results Abnormal Normal Test not done/ unknown TOTAL EMR KRAS Test Results 25 16 41 1 26 22 49 Test not done/unknown 2 45 48 28 27 83 138 65% 3 cases were missed by manual review. Received testing elsewhere and it wasn’t mentioned in the clinical notes for those patients. Also, during review found that 1 case was miscoded. One thing to remember is that we only included patients who visited with a medical oncologist. We excluded patients who only came to the University for surgery. If we had included these other patients then there would likely be more patients that the Registry caught compared to manual review. 40%

Comparison of KRAS testing for metastatic CRC captured by EMR review vs Iowa Cancer Registry   Iowa Cancer Registry KRAS Test Results Abnormal Normal Test not done/ unknown TOTAL EMR KRAS Test Results 25 16 41 1 26 22 49 Test not done/unknown 2 45 48 28 27 83 138 65% 38 cases were found during manual review that were missed by the Registry. A few years ago Dr. Mary Charlton and Bobbi Matt at ICR reviewed the KRAS variable at Iowa and found that of those cases that we missed testing for, the majority received their testing >6 months after diagnosis and outside of the abstracting window. So, when were this cohort of patients tested? 40%

Distribution of molecular testing months after diagnosis among metastatic CRC patients 27% 38% 41% 38% Describe axis. Majority of cases received testing within 6 months of diagnosis. (go through numbers) (Go over #s after 6 months) The majority of cases received testing within 6 months of diagnosis, but there were still patients who received testing after this time frame and so the likelihood of the Registry capturing this data goes down. Of those 38 cases that the Registry missed, 63% received KRAS testing >6 months after diagnosis and 55% were stage I-III and 45% stage IV

Comparison of molecular testing for metastatic CRC captured by EMR manual review vs CDM/CPT Data This table is comparing testing found during manual review vs the PCORnet CDM data, which is found via CPT codes. The codes used to extract these tests are provided in the table. (Go over percentages) MSI/MMR discrepancies – only found a CPT code for MSI-DNA testing. There are 4 IHC MMR codes, but no one in this cohort had one of these codes.

Computational extraction Molecular testing Are patients receiving testing? What testing are patients receiving? Compare results across different sources Molecular therapy How well are therapies being disseminated? Does treatment align with test results? Compare results across sources Computational extraction Can molecular pathology data be pulled electronically?

Metastatic CRC patients receiving molecular-guided vs targeted therapy at UIHC, 2013-2017 Overall, 51% of patients received a molecular-guided or other-targeted therapy. With the majority receiving an other-targeted therapy. Breaking it down into specific therapies – yellow is that the patient received treatment at the University and black means the patients received tx at another facility, but clinic notes mentioned the tx was given elsewhere. Of molecular-guided therapies Panitumumab was the most common with 9% of patients receiving it followed by Pembrolizumab and Cetuximab. Cetuximab and Panitumumab are EGFR-inhibitors and patients must not have mutations in KRAS or NRAS genes Pembrolizumab is a PD-1 inhibitor used in patients who have MSI-H/MMR deficient tumors. Now other-targeted therapies were the most common with Bevacizumab being the most common therapy overall. Bevacizumab is a VEGF-inhibitor and requires no biomarker testing to be eligible and is an acceptable alternative treatment compared to EGFR-inhibitors Ramucirumab is another VEGF-inhibitor Regorafenib is a multi-kinase inhibitor ---Extra info--- Several clinical trials have found survival benefits with these therapies and some have found that tumor location can play a role in their effectiveness NCCN guidelines state that All patients with RAS WT tumors can be considered for EGFR-inhibitors in subsequent lines of therapy if not given previously or patients with left-sided colon tumors can be offered cetuximab or panitumumab in first-line tx of metastatic disease NCCN guidelines recommend PD-1 inhibitors as subsequent-line tx options for patients with metastatic MMR deficient CRC

Molecular-guided therapies and testing for metastatic CRC patients   Cetuximab (N=3) Panitumumab (N=13) Pembrolizumab (N=6) Molecular tests and results % KRAS Normal 100 Mutated - Not done NRAS 33 77 67 23 BRAF 62 38 MSI/MMR Stable/Normal High/Abnormal Of those patients who received molecular-guided therapies, were the appropriate tests done and have appropriate results? All cases who received an EGFR-inhibitor had normal KRAS results. Not every case received NRAS or BRAF testing, but if they did were normal. All cases who received Pembrolizumab were MSI-high

Comparison of molecular-guided therapies for metastatic CRC patients captured by EMR review vs CDM Data 4 out of 22 patients who received molecular-guided therapy were not found in the CDM Given outside of UIHC and/or the study window

Computational extraction Molecular testing Are patients receiving testing? What testing are patients receiving? Compare results across different sources Molecular therapy How well are therapies being disseminated? Does treatment align with test results? Compare results across sources Computational extraction Can molecular pathology data be pulled electronically?

Ability to identify computable pathology and genomic data Extracted molecular testing data from EMR vs manual chart review   Test Type KRAS NRAS HRAS BRAF MSI/MMR Tested at UIHC according to manual chart review 85 63 51 61 103 Test performed (y/n) identified via electronic extraction of EMR (100%) 60 (98%) Test result identified via 65 (76%) 42 (67%) 29 (57%) 38 (62%) Electronic extraction of EMR Molecular testing data was computationally extracted from the UIOWA EMR and compared to manual chart review data for patients who received testing at UIOWA. In this table we have test type along the top and the first row represents the number of cases found to have testing during chart review. A binary (yes/no) result (second row of the table) for KRAS, NRAS, HRAS, MSI/MMR, and all but one BRAF test could be computationally extracted from the EMR. Computationally extracting test results was not as successful (last row of the table), we were only able to extract the actual KRAS results for 65 (76%) of patients tested, NRAS results for 42 (67%) of patients tested, HRAS results for 29 (57%) of patients tested, and BRAF results for 38 (60%) of patients tested. Computationally finding MSI/MMR results was successful and results for all 103 patients tested were found. Issues related to identifying results due to multiple mutation (panel) testing Difficult to identify which mutation was present

Conclusions – Molecular testing EMRs contained 98% of molecular testing and therapy information Between EMR and ICR all molecular testing was found EMR vs Registry Data EMR vs CDM Data Cases the Registry missed >60% of missed KRAS tests were performed >6 months after diagnosis 2% of cases were missed via manual review Proportion of KRAS and BRAF testing found in CDM was similar to manual review MMR testing was not found in the CDM because there is no specific CPT for the immunohistochemistry test 2% of cases were missed via manual review And if we changed our inclusion criteria there would probably be more

Conclusions Targeted therapies: Only 16% of patients received a molecular-guided therapy CDM data missed 4 cases vs EMR review Computational extraction Able to pull if a test was done, but results are more difficult Able to pull if test done in >98% of patients Could only extract results for ~2/3 of patients

Thank you amanda-kahl@uiowa.edu Questions? Thank you amanda-kahl@uiowa.edu

Distribution of molecular testing months after diagnosis among metastatic CRC patients by stage 27% 38% 41% 38% The grey represent stages I-III who received testing. The majority of cases tested between 0-6 months are stage 4 cases and stage I-III cases make up the majority of cases tested after 6 months, but there are still stage 4 cases being tested during this time frame. Maybe for these cases their doctors were holding off on tx to see if traditional chemo worked before trying targeted therapies.