Performance of Cancer Recurrence Algorithms After Coding Scheme Switch From International Classification of Diseases 9th Revision to International Classification of Diseases 10th Revision

PURPOSE

We previously developed and validated informatic algorithms that used International Classification of Diseases 9th revision (ICD9)-based diagnostic and procedure codes to detect the presence and timing of cancer recurrence (the RECUR Algorithms). In 2015, ICD10 replaced ICD9 as the worldwide coding standard. To understand the impact of this transition, we evaluated the performance of the RECUR Algorithms after incorporating ICD10 codes.

METHODS

Using publicly available translation tables along with clinician and other expertise, we updated the algorithms to include ICD10 codes as additional input variables. We evaluated the performance of the algorithms using gold standard recurrence measures associated with a contemporary cohort of patients with stage I to III breast, colorectal, and lung (excluding IIIB) cancer and derived performance measures, including the area under the receiver operating curve, average absolute prediction error, and correct classification rate. These values were compared with the performance measures derived from the validation of the original algorithms.

RESULTS

A total of 659 colorectal, 280 lung, and 2,053 breast cancer cases were identified. Area under the receiver operating curve derived from the updated algorithms was 89.0% (95% CI, 82.3% to 95.7%), 88.9% (95% CI, 79.3% to 98.2%), and 80.5% (95% CI, 72.8% to 88.2%) for the colorectal, lung, and breast cancer algorithms, respectively. Average absolute prediction errors for recurrence timing were 2.7 (SE, 11.3%), 2.4 (SE, 10.4%), and 5.6 months (SE, 21.8%), respectively, and timing estimates were within 6 months of actual recurrence for more than 80% of colorectal, more than 90% of lung, and more than 50% of breast cancer cases using the updated algorithm.

CONCLUSION

Performance measures derived from the updated and original algorithms had overlapping confidence intervals, suggesting that the ICD9 to ICD10 transition did not affect the RECUR Algorithm performance.

Management of Male Breast Cancer: ASCO Guideline

PURPOSE

To develop recommendations concerning the management of male breast cancer.

METHODS

ASCO convened an Expert Panel to develop recommendations based on a systematic review and a formal consensus process.

RESULTS

Twenty-six descriptive reports or observational studies met eligibility criteria and formed the evidentiary basis for the recommendations.

RECOMMENDATIONS

Many of the management approaches used for men with breast cancer are like those used for women. Men with hormone receptor-positive breast cancer who are candidates for adjuvant endocrine therapy should be offered tamoxifen for an initial duration of five years; those with a contraindication to tamoxifen may be offered a gonadotropin-releasing hormone agonist/antagonist plus aromatase inhibitor. Men who have completed five years of tamoxifen, have tolerated therapy, and still have a high risk of recurrence may be offered an additional five years of therapy. Men with early-stage disease should not be treated with bone-modifying agents to prevent recurrence, but could still receive these agents to prevent or treat osteoporosis. Men with advanced or metastatic disease should be offered endocrine therapy as first-line therapy, except in cases of visceral crisis or rapidly progressive disease. Targeted systemic therapy may be used to treat advanced or metastatic cancer using the same indications and combinations offered to women. Ipsilateral annual mammogram should be offered to men with a history of breast cancer treated with lumpectomy regardless of genetic predisposition; contralateral annual mammogram may be offered to men with a history of breast cancer and a genetic predisposing mutation. Breast magnetic resonance imaging is not recommended routinely. Genetic counseling and germline genetic testing of cancer predisposition genes should be offered to all men with breast cancer.

Assessment of Temporal Selection Bias in Genomic Testing in a Cohort of Patients With Cancer

This cohort study assesses for temporal selection bias in patients with lung, breast, colorectal, pancreatic, or urothelial cancer from a single institution who had tumor profiling using a next-generation sequencing protocol between 2013 and 2017.

Mobile health applications in oncology

e14115

Background: Over 325,000 mobile health (mhealth) applications (apps) have been developed. There has been a substantial increase in mhealth investment, with over $8.1 billion invested in digital health startups in 2018. While apps have been studied within clinical oncology, we are aware of no comprehensive evaluation of the commercial footprint of oncology-specific apps. We sought to describe the state of oncology-specific apps and highlight notable areas of development. Methods: We conducted a systemic search for oncology apps in the Apple iOS and Google Play app stores in January 2020. Search terms included “cancer,” “oncology,” “radiotherapy,” and “chemotherapy.” All apps were manually reviewed and classified by English language support, date of last update, downloads, intended audience, intended purpose, and developer type. We also compared commercially available apps with those described in a recently conducted meta-analysis of oncology-app studies. We performed descriptive statistics using RStudio V1.2.335. Results: We identified 794 oncology-specific, English-language applications, but only 257 (32%) met basic quality standards and were considered evaluable. The primary reason for exclusion was lack of a recent update. Of included apps, almost half (47%) were found in the “Medical” Store Category and the majority were free (88%). The most common intended audience was healthcare professionals (45%), with 28% being geared towards the general public and 27% being intended for patients. The intended function was education for 37%, clinical decision support (CDS) for 19%, and patient support for 18%. Only 22% of education apps and 40% of CDS apps reported any formal app content review process. Web developers created 61% of apps, scientific societies created 10%, and hospitals/healthcare organizations created just 6% (Table). The most frequently downloaded apps tended to be geared toward educating/supporting the public. Of 54 studies that utilized mobile apps in oncology identified by a recent meta-analysis, only 2 could be matched to commercially available apps from our study, suggesting a substantial divide between investigation and product dissemination. Conclusions: Our analysis of oncology-related apps in the commercial marketplace found few high-quality, up-to-date apps, and a notable absence of key oncology stakeholders in app development. Future studies should explore barriers to developing and disseminating apps designed to advance oncology care delivery. [Table: see text]

Design of eSyM: An ePRO-based symptom management tool fully integrated in the electronic health record (Epic) to foster patient/clinician engagement, sustainability, and clinical impact

e14120

Background: Chemotherapy and surgery can cause distressing symptoms, which can be a burden for health system to address. Programs that directly engage patients, including electronic tracking of patient-reported outcomes (ePROs), can improve symptom control and decrease the need for acute care. Previous ePRO programs have relied on third party vendors with limited EHR integration, constraining their clinical utility and scalability. An integrated solution could offer distinct advantages. Methods: As part of NCI’s Moonshot-funded IMPACT consortium, 6 health systems and Epic built an electronic symptom management program (eSyM) based on the PRO-CTCAE questionnaire that is fully integrated into the EHR. The agile, user-centered design process engaged patients, clinicians, and institutions. The core functional components include: 1) symptom surveys in the postoperative period or between chemotherapy visits, 2) self-management tip sheets, 3) clinician alerts, and 4) dashboards for population management. Critical points of integration with supporting EHR functions and workflow impacts were identified; and major challenges of integration and implementation were described. Results: eSyM, which was implemented at two health systems (Baptist Memorial in Tennessee and Mississippi and West Virginia University Health) in the fall of 2019, required multiple supporting EHR functions: 1) access a secure, HIPPA-compliant patient portal/messaging system (MyChart); 2) record diagnosis, procedure and chemotherapy treatment plan data; 3) identify target populations and track metrics/events; 4) define and execute autonomous logic-based workflow rules; 5) generate reports for clinicians/patients; and 6) documentation. Major challenges included: 1) working within pre-existing EHR system standards and capabilities, which limited the ability to customize interfaces and workflows specifically for the eSyM use case; and 2) adapting to different EHR configurations and polices across multiple health systems. Conclusions: The eSyM build leveraged many existing EHR capabilities and addressed regulatory hurdles; but it required design and workflow compromise. Integration of ePRO-based symptom management programs into the EHR could help overcome barriers, consolidate clinical workflows, and foster scalability/sustainability. Ongoing efforts include launching eSyM at four more sites and evaluating its adoption, usability, and impact on clinical outcomes.

Development and validation of a novel EHR-based tumor progression outcome to support biomarker discovery

e19297

Background: Obtaining clinical outcomes for analysis has historically been a critical barrier to cancer genomics research. EHRs could constitute an important data source to bridge this gap, but EHRs rarely capture structured outcomes such as cancer progression. Novel, robust methods are needed to capture clinically relevant outcomes from EHRs. Methods: Among patients with lung adenocarcinoma whose tumors were sequenced via the Dana Farber Cancer Institute/Brigham and Women’s PROFILE study from 2013-2018, imaging reports following first palliative-intent systemic therapy were annotated using natural language processing (NLP) models trained to capture cancer progression according to the structured “PRISSMM” framework. NLP-based cancer progression and imaging report frequency were jointly modeled using inverse-intensity weighted generalized estimated equations, censored at six months, to explore associations between alterations in lung cancer biomarkers (ALK, EGFR, ROS1, BRAF, KRAS, SMARCA4) and progression. Among patients with KRAS mutations who received immunotherapy, we also analyzed the association between STK11 mutations and progression. The novel outcome generated by the model – imaging report-based progression (iPROG) – corresponded to the difference in the mean log odds of progression per inverse-intensity weighted report associated with a given biomarker; it was reported as adjusted mean probability and in exponentiated form as an odds ratio (OR). Results: Among 690 patients with lung adenocarcinoma, associations between tumor mutations and the iPROG outcome are listed in the Table. Conclusions: A deep NLP model applied to EHR data can capture a novel cancer progression outcome, which is associated with known prognostic markers in lung cancer. Application of this method to large “real world” datasets, with attention to interactions between treatment and genomics, could speed biomarker discovery. [Table: see text]

Patient-reported health literacy and numeracy among new patients seeking consultation at a comprehensive cancer center

7038

Background: Health literacy and numeracy are essential for patients to make informed cancer treatment decisions. Oncologists do not typically evaluate literacy and numeracy and vary in their ability to adapt health discussions to meet patients’ needs. Systematic ascertainment of literacy and numeracy may provide oncologists with useful information to help guide initial oncology consultations. Methods: We deploy an electronic new patient intake questionnaire (NPIQ) that includes health literacy and numeracy, basic demographics and cancer risk screening. Patients are considered to have limited health literacy and/or numeracy if they respond with either “somewhat”, “a little bit” or “not at all” to a single question: “How confident are you filling out medical forms?” or “How confident are you in understanding medical statistics?” respectively. Results: Between January 2018 and August 2019, 8418 (24.6%) of patients presenting for a new patient consultation responded to the NPIQ. Among respondents with non-missing data, limited health literacy was reported by 19.4% respondents with 13.9% reporting “not at all” and 33.1% reporting “not at all” or only “a little bit” of confidence completing medical forms. Limited health numeracy was reported by 33.2% respondents with 9.1% reporting “not at all”. Nearly 20% of respondents reported both limited health literacy and numeracy. Patients reporting lack of confidence completing medical forms or understanding medical statistics were older (20.3%, 30.7% ³ 70 years old), male (20.2%, 30.1%), and non-white (21.3%, 32.1%). Conclusions: A substantial proportion of cancer patients report lack of confidence in their ability to complete medical forms or understand medical statistics, potentially limiting the ability to actively engage in shared decision-making. Prospective identification of these social determinants of health prior to consultations may provide oncologists with information necessary to tailor health discussions and to provide materials that promote understanding and informed decision-making. [Table: see text]

Efficacy of neoadjuvant chemotherapy (NAC) in male breast cancer (MaBC) compared with female breast cancer (FBC): A National Cancer Database (NCDB) study

587

Background: NAC is frequently used in the treatment of FBC. The efficacy of NAC in MaBC is unclear. Few studies have compared outcomes for MaBC and FBC after similar treatment. The aim of this study was to compare proportions of pathologic complete response (pCR) between MaBC and FBC according to tumor subtype (TS). Secondary aims were clinical response and overall survival (OS). Methods: We evaluated men and women with breast cancer treated with NAC between 2010 and 2016 with known hormone receptor (HR) status and human epidermal growth factor receptor 2 (HER2) status at NCDB centers. The proportion with pCR (ypT0/Tis ypN0) was compared for MaBC and FBC for each TS by Fisher’s exact test. Logistic regression evaluated odds of pCR. OS was estimated by Kaplan-Meier and compared by log-rank test. Results: Of 7,721 MaBC and 859,096 FBC patients, 385 MaBC (5%) and 68,065 FBC (7.9%) underwent NAC and were included in this study. Median age for MaBC was 58 years (y) (range 23-88) and for FBC was 53 y (range 18-90). Within each TS, there were no significant differences in the distribution of tumor grade between MaBC and FBC. Clinical stage in MaBC and FBC were: Stage I: 8% v 11%, Stage II: 54% v 59%, Stage III: 38% v 30%; respectively. Median time from initiation of NAC to surgery was 143 days in MaBC and 148 days in FBC. Compared with FBC, MaBC had a lower proportion of complete clinical response (18% v 31%) and a higher proportion of no clinical response (14% v 7%); p < 0.001. Proportions and odds of pCR were numerically lower in MaBC compared with FBC for each TS and statistically significant for HR+/HER2- and HR+/HER2+ (table). pCR was associated with OS in both MaBC and FBC. Specifically, in MaBC who achieved pCR v not, 5 y OS was 90% v 64.7%; p = 0.02. In FBC who achieved pCR v not, 5 y OS was 91.9% v 75.3%; p < 0.01. Among pts receiving NAC, MaBC had worse OS at 5 y than FBC (67.1% v 79.0%; p = 0.02). Conclusions: Men receiving NAC achieved lower proportions of pCR than women and had significantly worse OS. However, pCR is prognostic in both MaBC and FBC. Limitations include small sample sizes for HR-/HER2+ and triple-negative TS and lack of detailed regimen information. Nevertheless, our results suggest that, compared with FBC, MaBC may be intrinsically more resistant to NAC. [Table: see text]

Self-reported overall wellbeing (OWb), physical function (PFn), and PRO-CTCAE symptom scores in post-operative and chemotherapy patients

2064

Background: A standardized, validated tool for capturing symptoms from cancer patients, PRO-CTCAE, has been used to reduce symptom burden, decrease acute care needs, and preserve quality of life. The association between specific PRO-CTCAE symptom scores and single item measures of OWb and PFn were characterized to understand symptom constellations. Methods: A novel Epic-based symptom management program (eSyM) was deployed for GI, GYN, and thoracic cancer patients starting chemotherapy (Memphis Baptist) or having surgery (WVU Medicine). Patients received automated prompts to complete surveys via the patient portal (MyChart) on a fixed schedule, approximately twice/week. Each survey included one OWb item, one PFn item, and at least 6 PRO-CTCAE items (pain, nausea, vomiting, fatigue, anxiety, insomnia). The OWb and PFn items, which were created de novo, included 5 ordinal response options with corresponding pictograms (emojis from very happy to very sad for OWb; a figure walking to one prone in bed for PFn). Composite scores were generated: 0 for no symptoms, 1-2 for mild/moderate symptoms, and 3 for severe symptoms. We describe OWb and PFn and analyze associations between these items and PRO-CTCAE symptom scores. Results: Between 9/10/19-1/22/20, we collected 908 eSyM responses from 166 chemotherapy patients at Baptist (Age, M = 65), and 480 eSyM responses from 97 postoperative patients at WVU (Age, M = 57). The OWb and PFn scores demonstrated moderate correlation with PRO-CTCAE symptom scores (Baptist r = 0.63; WVU r = 0.75), and moderate correlation with mean symptom scores among surgery patients at WVU (r = 0.74); but lower correlation among chemotherapy patients at Baptist (r = 0.53-0.55). Scores improved over time following surgery, but not after initiation of chemotherapy. Among the 730 eSyM responses with none/mild values for both OWb and PFn (52.9% of all responses), only 4.5% reported any severe symptom; among 651 responses with impairment of OWb and/or PFn, 45.2% reported at least one severe symptom. Conclusions: Integration of eSyM into the Epic EHR enabled tracking of OWb, PFn, and PRO-CTCAE items. When asked alongside PRO-CTCAE symptom items, two single item OWb and PFn measures provided distinct information and correlated with symptom burden. These results demonstrate the feasibility of integrating ePRO collection into routine post-operative and medical oncology care and that PRO-CTCAE items provide information that is distinct from that obtained from global metrics of well-being. Clinical trial information: NCT03850912.

Survival associated with mutations in SWI/SNF chromatin remodeling complex genes

3643

Background: The SWI/SNF (SWitch/Sucrose NonFermentable) chromatin remodeling complex (CRC) - a combinatorial assembly of products from multiple genes - alters histone/DNA interactions and thereby impacts transcription, DNA replication/repair, and cell division. Studies suggest that over 20% of human cancers contain mutations in at least one SWI/SNF gene, implying that it is the most highly mutated CRC in human cancer. To address existing knowledge gaps, we sought to evaluate the association between SWI/SNF mutations and overall survival (OS). Methods: We identified adult cancer patients who consented to have OncoPanel testing (Dana-Farber/Brigham & Women’s Hospital’s next generation sequencing platform) from June 2013-August 2019. These data were merged with institutional electronic health records and National Death Index vital status. We determined mutation frequency and co-occurrence for the nine SWI/SNF genes included in OncoPanel (ARID1A, ARID1B, ARID2, BCL11B, PBRM1, SMARCA4, SMARCB1, SMARCE1, and SS18). We assessed the association between mutation and OS (from time of OncoPanel testing) for cancers with at least 500 analyzed and 20 mutated cases, controlling for age and TP53 status. Exploratory analyses were conducted using cBioPortal and SAS (no multiple comparison adjustment). Results: Among 25,434 samples from 24,648 patients, a mutation in at least one evaluated SWI/SNF gene was identified in 26% of cases (ARID1A 10.5%, ARID1B 7.2%, SMARCA4 5.5%, PBRM1 4.9%, ARID2 4.8%, BCL11B 3.5%, SMARCE1 1.1%, SMARCB1 1.0%, and SS18 0.7%). The most frequently mutated cancers included small bowel (52%), endometrial (49%), ampullary (48%) and bladder (45%). Co-occurrence was common (30 of 36 potential gene-pairs), with the largest associations (odds ratio; all P < .05) seen for SMARCB1:BCL11B (4.19), ARID1B:BCL11B (3.87), ARID2:BCL11B (3.85), and SMARCA4:BCL11B (3.78). Associations between having a mutation and OS were seen for the following cancers/genes (odds ratio; all P < .05): ARID1A (colorectal 0.72, pancreatic 1.46), ARID1B (melanoma 0.32), SMARCA4 (esophagogastric 1.48, non-small cell lung 1.89, ovarian 0.43), SMARCB1 (non-small cell lung 2.04), and SS18 (soft tissue sarcoma 2.06). Conclusions: Mutations in SWI/SNF genes are widespread, with mutation rates varying by cancer type. Co-occurrence was common, especially with BCL11B. Associations with OS were both favorable and unfavorable, with variability seen by gene and cancer type. Future research should explore the mechanisms by which mutations in SWI/SNF genes influence treatment response/OS.

Distribution and frequency of patient-reported symptomatic adverse events at a comprehensive cancer center

e19117

Background: Systematic review of electronic patient reported outcomes (ePRO) has been shown to improve quality of life and overall survival in clinical trial. We previously demonstrated feasibility of ePRO across Dana-Farber Cancer Institute (DFCI). We sought to examine the distribution and frequency of first symptomatic adverse events (SAEs) among ePRO responders in ambulatory oncology practice. Methods: The ePRO tool uses the validated NCI developed Patient Reported Outcomes – Common Terminology Criteria for Adverse Events (PRO-CTCAE) instrument to assess attributes of 15 core SAEs (fatigue, insomnia, general pain, decreased appetite, nausea, vomiting, constipation, diarrhea, shortness of breath, numbness and tingling, rash, concentration, fever, anxiety, sadness) selected by clinician stakeholders and deployed via any internet-enabled device once every 7 days. Responses are viewable in the EHR, scored 0 to 3 using an algorithm, with scores of 3 highlighted to indicate severe grade SAEs. Results: We examined the distribution and frequency of the first 5183 unique ePRO reports for unselected patients seen in the medical, radiation and surgical oncology outpatient clinics of four pilot multidisciplinary clinics (Breast, Genitourinary, Gastrointestinal and Head and Neck) between September 2018-December 2019. Twenty one percent of eligible patients responded to ePRO (5183 of 26,084). Most respondents were female (59%), Caucasian (89%), and age 50-69 years (56% compared to 16% age <50 years, 28% age ≥70; range 19-98 years). The frequency of grade 3 SAEs was pain (10%), fatigue (6%), insomnia (4%), constipation (3%), numbness and tingling/concentration/anxiety/decreased appetite (2%), diarrhea/shortness of breath/sadness (1%), and rash/fever/nausea/vomiting (none) (Table). Conclusions: We observed a consistent distribution of SAEs across cancer types, age and sex. The most frequently reported SAEs are those clinicians struggle to treat with medications - pain, fatigue, insomnia and anxiety. Research to develop effective strategies to address this constellation of SAEs should be prioritized. [Table: see text]

Neoadjuvant treatment strategies for HER2-positive breast cancer: cost-effectiveness and quality of life outcomes

PURPOSE

Achieving a pathologic complete response (pCR) with neoadjuvant therapy for HER2-positive breast cancer is associated with less recurrence and improved clinical outcomes compared to having residual cancer at surgery. However, recent data have demonstrated favorable outcomes for patients with residual HER2-positive cancer who received adjuvant trastuzumab emtansine (TDM-1). Therefore, we sought to determine the optimal chemotherapy/anti-HER2 treatment strategy.

METHODS

We created a decision-analytic model for patients with stage II-III HER2-positive cancer that incorporated utilities based on toxicity and recurrence. We separately modeled hormone receptor-negative (HR-) and positive (HR+) disease and calculated quality-adjusted life years (QALYs) and costs through 5 years. Simulated patients received one of the following neoadjuvant treatments: three 'intensive' regimens (TCHP: docetaxel, carboplatin, trastuzumab, pertuzumab; THP + AC: taxol, trastuzumab, pertuzumab then doxorubicin and cyclophosphamide; THP: taxol, trastuzumab, pertuzumab) and two 'de-escalated' regimens (TH: taxol, trastuzumab; TDM-1) followed by adjuvant treatment based on pathologic response.

RESULTS

Among 'intensive' neoadjuvant strategies, treatment with THP was more effective and less costly than TCHP or THP + AC. When 'de-escalated' strategies were included, TH became the most cost-effective. For HR-negative cancer, TH had 0.003 fewer quality-adjusted life years (QALYs) than THP but was less costly by $55,831, resulting in an incremental cost-effectiveness ratio of over $18M/QALY for THP, well above any threshold. For HR-positive cancer, neoadjuvant TH dominated the THP strategy.

CONCLUSION

An adaptive-treatment strategy beginning with neoadjuvant THP or TH followed by tailoring post-operative therapy reduces treatment costs, and spares toxicity compared to more intensive chemotherapy regimens for women with HER2-positive breast cancer.

Patterns of care for older patients with stage IV non-small cell lung cancer in the immunotherapy era

BACKGROUND

Historically, older patients with advanced lung cancer have often received no systemic treatment. Immunotherapy has improved outcomes in clinical trials, but its dissemination and implementation at the population level is not well-understood.

METHODS

A retrospective cohort study of patients with stage IV non-small cell lung cancer (NSCLC) diagnosed age 66 or older from 2012 to 2015 was conducted using SEER-Medicare. Treatment patterns within one year of diagnosis were ascertained. Outcomes included delivery of (a) any systemic therapy; (b) any second-line infusional therapy, following first-line infusional therapy; and (c) any second-line immunotherapy, following first-line infusional therapy. Trends in care patterns associated with second-line immunotherapy approvals in 2015 were assessed using generalized additive models. Sociodemographic and clinical predictors of treatment were explored using logistic regression.

RESULTS

Among 10 303 patients, 5173 (50.2%) received first-line systemic therapy, with little change between the years 2012 (47.5%) and 2015 (50.3%). Among 3943 patients completing first-line infusional therapy, the proportion starting second-line infusional treatment remained stable from 2012 (30.5%) through 2014 (32.9%), before increasing in 2015 (42.4%) concurrent with second-line immunotherapy approvals. Factors associated with decreased utilization of any therapy included age, black race, Medicaid eligibility, residence in a high-poverty area, nonadenocarcinoma histology, and comorbidity; factors associated with increased utilization of any therapy included Asian race and Hispanic ethnicity. Among patients who received first-line infusional therapy, factors associated with decreased utilization of second-line infusional therapy included age, Medicaid eligibility, nonadenocarcinoma histology, and comorbidity; Asian race was associated with increased utilization of second-line infusional therapy.

CONCLUSION

United States Food and Drug Administration (FDA) approvals of immunotherapy for the second-line treatment of advanced NSCLC in 2015 were associated with increased rates of any second-line treatment, but disparities based on social determinants of health persisted.

Abstract P5-06-13: Factors associated with twenty-year (y) risks of breast cancer-specific mortality (BCSM) for locally-advanced breast cancer (LABC) in the surveillance, epidemiology, and end results (SEER) registry

Background: Patients (pts) with LABC have a high risk of early recurrence and death. The risk of late BCSM in pts with LABC is unclear. Recent studies have shown that the risk of BCSM persists beyond 5 y (Pan, NEJM 2017); however, pts with LABC as well as those with hormone receptor (HR)-negative disease have not been studied. The aims of this study were to report on population-based long-term risks of BCSM, and the risks of BCSM conditional on having survived 5 y, among pts with LABC. In addition, we aimed to identify factors associated with late deaths from breast cancer.

Methods: We evaluated women with LABC (T3 or T4 [any N] or N3 [any T], M0 disease), diagnosed between 1990 and 2005, reported to SEER. HR status was known for all pts, but HER2 was unavailable. Pts with another primary tumor either before or after breast cancer were excluded. The dependent outcome is based on the cause-of-death recode variable from SEER (which can include disease or treatment-related deaths). Patients with non-cancer cause of deaths were censored. We used Kaplan-Meier analyses to determine the effect of baseline variables on cumulative risks of BCSM, estimated the annual rate of events per 100 person-years, and performed unadjusted Fine and Gray regressions for T3/T4 pts and for N3 pts stratified by HR status.

Results: We included 24,082 pts (median follow-up = 7.16 y). Of all breast cancer deaths, the proportion that occurred after 5 y was 46% for HR+ vs 13% for HR- (p&lt;0.001), the proportion after 10 y was 13% vs 2% respectively (p&lt;0.001). The table shows risks of BCSM by HR, N and T status, and annual event rates. The cumulative risk of BCSM in y 5-20 ranged from 10.2% in HR- T3/T4,N0 to 59.4% in HR+ T3/T4,N3. Conditional on having survived 5 y, unadjusted hazards of BCSM among T3/T4 pts were the following: T3/T4 HR+ N3 vs N0 (Hazard ratio [HzR] 3.9; 95% confidence interval [CI], 3.4-4.5); T3/T4 HR- N3 vs N0 (HzR 3.6; 95% CI, 2.7-4.7). Unadjusted hazards of BCSM among N3 pts were the following: N3 HR+ T4 vs T1 (HzR 1.5; 95% CI, 1.3-1.8); N3 HR- T4 vs T1 (HzR 1.8; 95% CI, 1.3-2.7).

Conclusions: Among pts with LABC, event rates within 5 y are high, in both HR+ and HR- pts. Beyond 5 y, BCSM still depends on traditional clinicopathologic factors and more late events occur in HR+ disease than in HR- disease. The observed late events lead to an extremely high cumulative risk of BCSM by y 20 in both HR+ and HR- LABC.

BCSMAll-cause mortality% Event-FreeAnnual rate (%)Cumulative risk (%)Cumulative risk (%)at 5 yat 10 yy 5-&lt;10y 10-&lt;15y 15-20y 5-20y 0-20y 0-20HR status among N3HR+N366.345.57.85.03.554.970.179.2HR-N338.931.04.70.91.027.371.878.0Nodal status among T3/T4HR+N090.281.82.01.40.617.125.256.6N179.866.53.72.41.431.245.164.0N271.251.86.54.13.148.163.176.2N356.936.99.06.03.459.476.983.8HR-N075.571.31.20.30.710.232.251.0N156.449.32.80.80.418.554.167.1N238.831.54.41.21.428.172.180.8N328.820.86.80.91.034.681.285.1Tumor size among N3 onlyHR+T175.956.76.04.32.647.660.271.0T268.446.18.14.94.156.370.179.7T361.740.98.55.83.758.974.781.0T449.730.910.26.32.660.280.288.3HR-T151.141.54.30.81.527.062.769.4T241.934.83.81.00.722.967.775.9T333.325.75.30.70.626.275.479.0T424.616.28.81.43.060.590.393.9

Citation Format: Jose P Leone, Bernardo A Leone, Michael J Hassett, Julieta Leone, Rachel A Freedman, Sara M Tolaney, Eric P Winer, Carlos T Vallejo, Nancy U Lin. Factors associated with twenty-year (y) risks of breast cancer-specific mortality (BCSM) for locally-advanced breast cancer (LABC) in the surveillance, epidemiology, and end results (SEER) registry [abstract]. In: Proceedings of the 2019 San Antonio Breast Cancer Symposium; 2019 Dec 10-14; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2020;80(4 Suppl):Abstract nr P5-06-13.

Determining the Time of Cancer Recurrence Using Claims or Electronic Medical Record Data

PURPOSE

Data from claims and electronic medical records (EMRs) are frequently used to identify clinical events (eg, cancer diagnosis, stroke). However, accurately determining the time of clinical events can be challenging, and the methods used to generate time estimates are underdeveloped. We sought to develop an approach to determine the time of a clinical event-cancer recurrence-using high-dimensional longitudinal structured data.

METHODS

Manual chart abstraction provided information regarding the actual time of cancer recurrence. These data were linked to claims from Medicare or structured EMR data from the Cancer Research Network, which were used to determine time of recurrence for patients with lung or colorectal cancer. We analyzed the longitudinal profile of codes that could help determine the time of recurrence, adjusted for systematic differences between code dates and recurrence dates, and integrated time estimates from different codes to empirically derive an optimal algorithm.

RESULTS

We identified twelve code groups that could help determine the time of recurrence. Using claims data for patients with lung cancer, the optimal algorithm consisted of three code groups and provided an average prediction error of 4.8 months. Using EMR data or applying this approach to patients with colorectal cancer yielded similar results.

CONCLUSION

Time estimates were improved by selecting codes not necessarily the same as those used to identify recurrence, combining time estimates from multiple code groups, and adjusting for systematic bias between code dates and recurrence dates. Improving the accuracy of time estimates for clinical events can facilitate research, quality measurement, and process improvement.

Assessment of Deep Natural Language Processing in Ascertaining Oncologic Outcomes From Radiology Reports

Importance

A rapid learning health care system for oncology will require scalable methods for extracting clinical end points from electronic health records (EHRs). Outside of clinical trials, end points such as cancer progression and response are not routinely encoded into structured data.

Objective

To determine whether deep natural language processing can extract relevant cancer outcomes from radiologic reports, a ubiquitous but unstructured EHR data source.

Design, Setting, and Participants

A retrospective cohort study evaluated 1112 patients who underwent tumor genotyping for a diagnosis of lung cancer and participated in the Dana-Farber Cancer Institute PROFILE study from June 26, 2013, to July 2, 2018.

Exposures

Patients were divided into curation and reserve sets. Human abstractors applied a structured framework to radiologic reports for the curation set to ascertain the presence of cancer and changes in cancer status over time (ie, worsening/progressing vs improving/responding). Deep learning models were then trained to capture these outcomes from report text and subsequently evaluated in a 10% held-out test subset of curation patients. Cox proportional hazards regression models compared human and machine curations of disease-free survival, progression-free survival, and time to improvement/response in the curation set, and measured associations between report classification and overall survival in the curation and reserve sets.

Main Outcomes and Measures

The primary outcome was area under the receiver operating characteristic curve (AUC) for deep learning models; secondary outcomes were time to improvement/response, disease-free survival, progression-free survival, and overall survival.

Results

A total of 2406 patients were included (mean [SD] age, 66.5 [10.8] years; 1428 female [59.7%]; 2170 [90.2%] white). Radiologic reports (n = 14 230) were manually reviewed for 1112 patients in the curation set. In the test subset (n = 109), deep learning models identified the presence of cancer, improvement/response, and worsening/progression with accurate discrimination (AUC >0.90). Machine and human curation yielded similar measurements of disease-free survival (hazard ratio [HR] for machine vs human curation, 1.18; 95% CI, 0.71-1.95); progression-free survival (HR, 1.11; 95% CI, 0.71-1.71); and time to improvement/response (HR, 1.03; 95% CI, 0.65-1.64). Among 15 000 additional reports for 1294 reserve set patients, algorithm-detected cancer worsening/progression was associated with decreased overall survival (HR for mortality, 4.04; 95% CI, 2.78-5.85), and improvement/response was associated with increased overall survival (HR, 0.41; 95% CI, 0.22-0.77).

Conclusions and Relevance

Deep natural language processing appears to speed curation of relevant cancer outcomes and facilitate rapid learning from EHR data.

Developing a cancer-specific trigger tool to identify adverse events using administrative data

238

Background: “Trigger tools” identify complications of care and potential patient safety hazards. However, attempts to create triggers that flag treatment-related complications in oncology have been largely unsuccessful. To address this problem, the authors built a set of claims-based oncology-specific triggers based on a promising pilot study conducted at Memorial Sloan-Kettering Cancer Center. Methods: We selected subjects from the OptumLabs data warehouse, a repository of > 160 million de-identified patients drawn from commercial claims. The cohort included patients with breast, colorectal, lung, and prostate cancer undergoing an initial course of cancer-directed therapy from 2008-14. Using ICD and CPT codes, we defined 16 oncology-specific triggers drawn from the pilot study, all with PPVs ≥50%. Triggers included events such as neutropenic fever, abnormal serum potassium or bicarbonate, and initiation of therapeutic anticoagulation. To distinguish treatment-related complications from other comorbidities, we required a logical and temporal relationship between a treatment and the associated trigger. We tabulated the prevalence of cancer triggers by cancer type and metastatic status during a one-year follow up period and created multivariate logistic regression models to examine the association of triggered cases with one-year mortality. Results: The cohort comprised 369,354 unique subjects including 29% with metastatic disease. The prevalence of triggered events was greatest among non-metastatic patients with lung (33%) and colorectal (21%) cancers, and among those with metastatic disease. The most common triggers included abnormal chemistry tests, blood transfusions, hypoxemia, and chest CT following radiation therapy. The mortality rate was substantially higher among patients with at least one trigger compared to patients with none. Experiencing at least one cancer-specific trigger increased the one-year risk of death by 1.69 (95% CI 1.28-2.24). Conclusions: Oncology-specific triggers provide researchers a promising method for studying patient safety in cancer care.

Spending for Advanced Cancer Diagnoses: Comparing Recurrent Versus De Novo Stage IV Disease

PURPOSE

Spending for patients with advanced cancer is substantial. Past efforts to characterize this spending usually have not included patients with recurrence (who may differ from those with de novo stage IV disease) or described which services drive spending.

METHODS

Using SEER-Medicare data from 2008 to 2013, we identified patients with breast, colorectal, and lung cancer with either de novo stage IV or recurrent advanced cancer. Mean spending/patient/month (2012 US dollars) was estimated from 12 months before to 11 months after diagnosis for all services and by the type of service. We describe the absolute difference in mean monthly spending for de novo versus recurrent patients, and we estimate differences after controlling for type of advanced cancer, year of diagnosis, age, sex, comorbidity, and other factors.

RESULTS

We identified 54,982 patients with advanced cancer. Before diagnosis, mean monthly spending was higher for recurrent patients (absolute difference: breast, $1,412; colorectal, $3,002; lung, $2,805; all P < .001), whereas after the diagnosis, it was higher for de novo patients (absolute difference: breast, $2,443; colorectal, $4,844; lung, $2,356; all P < .001). Spending differences were driven by inpatient, physician, and hospice services. Across the 2-year period around the advanced cancer diagnosis, adjusted mean monthly spending was higher for de novo versus recurrent patients (spending ratio: breast, 2.39 [95% CI, 2.05 to 2.77]; colorectal, 2.64 [95% CI, 2.31 to 3.01]; lung, 1.46 [95% CI, 1.30 to 1.65]).

CONCLUSION

Spending for de novo cancer was greater than spending for recurrent advanced cancer. Understanding the patterns and drivers of spending is necessary to design alternative payment models and to improve value.

Avoiding peg-filgrastim (Peg-F) prophylaxis during the paclitaxel (T) portion of the dose-dense (DD) doxorubicin-cyclophosphamide (AC)-T regimen: A prospective study

517

Background: Use of growth factors (GF) adds considerable expense and some toxicity to adjuvant breast cancer chemotherapy. We tested the feasibility and safety of omitting routine GF use during the T portion of DD AC-T. Methods: This is a prospective, single-arm study in which patients (pts) who completed 4 cycles of DD-AC proceeded to DD-T 175 mg/m2 every two weeks (wks) without routine GF (NCT02698891). Key inclusion: age≤ 65, ECOG PS≤1, absolute neutrophil count (ANC) ≥1500/mm3, and no febrile neutropenia (FN) during DD-AC. Criteria to treat for T included ANC ≥1000/mm3. Peg-F was given only if pts had FN in a prior cycle, or at investigator discretion if infection or treatment delay > 1 wk. Once Peg-F was given, pts received it in all future cycles. The primary endpoint was the rate of T completion ≤ 7 wks from cycle 1 day 1 (C1D1) to C4D1. Secondary endpoints included total use of Peg-F, rates of hematologic toxicity and FN, reasons for dose modification or hold. If ≥85% of pts completed T on time, the regimen would be considered feasible. If the true on-time completion rate is 75%, the chance the regimen would be declared infeasible is 91%, and if it is 85% the chance that the regimen is falsely declared infeasible is 10% (power = 0.899). ≥100/125 pts had to complete T on time for the regimen to be deemed successful. Results: Among 127 pts enrolled, 125 received ≥1 dose of protocol therapy and are included in the analysis. Median age at registration was 46 (range 21-65). Median C1D1 ANC was 7500/mm3 (range 1500-20500). 112 (90%) (95% CI 83-94%) pts completed DD-T ≤ 7 wks, and 3 (2%) completed within > 7 wks (2 due to neutropenia); 10 (8%) did not complete all cycles of T. Omission of Peg-F was not causally related to non-completion of T in any pts. The most common reasons for dose reduction or delays were non-hematologic. One pt had FN but was able to complete T on time. Eight (6.4%) pts received Peg-F during the trial. Conclusions: Omission of routine GF use during DD-T according to a pre-specified algorithm appears safe, feasible, and was associated with a 95.7% reduction in use of Peg-F, relative to the current standard of care. Additional analyses including cost implications are ongoing. Clinical trial information: NCT02698891.

Factors associated with twenty-year (y) risks of breast cancer-specific mortality (BCSM) in the Surveillance, Epidemiology, and End Results (SEER) Registry

540

Background: Most reports describing the risk of late relapse in breast cancer have been based on selected patients (pts) enrolled into clinical trials. The aims of this study were to report on population-based long-term risks of BCSM, and the risks of BCSM conditional on having survived 5 y. Additionally, we aimed to identify factors associated with late deaths from breast cancer. Methods: Using SEER data, we identified women with breast cancer (T1-T2, N0-N2, M0) between 1990-2005, with one primary cancer in their lifetime, and known hormone receptor (HR) status. We used Kaplan-Meier analyses to determine the effect of baseline variables on cumulative risks of BCSM, we estimated annual rate of events per 100 person-years, and performed Cox regression stratified by HR status. Results: We included 202,080 pts (median follow-up = 12.25 y). Of all breast cancer deaths, the proportion after 5 y was 65% for HR+ vs 28% for HR- (p < 0.001). The table shows risks of BCSM by HR and N status, and annual event rates. The cumulative risk of BCSM in y 5-20 ranged from 7.9% in HR-N0 to 38% in HR+N2. Among HR+ pts, adjusted risks of BCSM conditional on having survived 5 y were higher for T2 vs T1a (Hazard ratio [HzR] 3.3, p < 0.001), N2 vs N0 (HzR 3.5, p < 0.001), age at diagnosis (dx) > 64 y vs < 50 y (HzR 1.4, p < 0.001), black race vs white (HzR 1.3, p < 0.001) and grade III vs I (HzR 2.3, p < 0.001). For HR- pts, adjusted risks of BCSM conditional on having survived 5 y were higher for T2 vs T1a (HzR 2.0, p < 0.001), N2 vs N0 (HzR 2.8, p < 0.001) and age at dx > 64 y vs < 50 y (HzR 1.6, p < 0.001). Conclusions: For HR+ breast cancer, risks of BCSM remain high beyond 5 y from dx and depend on T-N status, age, race and grade. In HR- breast cancer, the risk of BCSM is highest within 5 y from dx; however, risks beyond 5 y are still considerable and depend on T-N status and age. Our results underscore the need for better adjuvant therapies in both HR+ and HR- breast cancer. [Table: see text]

Development, Validation, and Dissemination of a Breast Cancer Recurrence Detection and Timing Informatics Algorithm

Background

This study developed, validated, and disseminated a generalizable informatics algorithm for detecting breast cancer recurrence and timing using a gold standard measure of recurrence coupled with data derived from a readily available common data model that pools health insurance claims and electronic health records data.

Methods

The algorithm has two parts: to detect the presence of recurrence and to estimate the timing of recurrence. The primary data source was the Cancer Research Network Virtual Data Warehouse (VDW). Sixteen potential indicators of recurrence were considered for model development. The final recurrence detection and timing models were determined, respectively, by maximizing the area under the ROC curve (AUROC) and minimizing average absolute error. Detection and timing algorithms were validated using VDW data in comparison with a gold standard recurrence capture from a third site in which recurrences were validated through chart review. Performance of this algorithm, stratified by stage at diagnosis, was compared with other published algorithms. All statistical tests were two-sided.

Results

Detection model AUROCs were 0.939 (95% confidence interval [CI] = 0.917 to 0.955) in the training data set (n = 3370) and 0.956 (95% CI = 0.944 to 0.971) and 0.900 (95% CI = 0.872 to 0.928), respectively, in the two validation data sets (n = 3370 and 3961, respectively). Timing models yielded average absolute prediction errors of 12.6% (95% CI = 10.5% to 14.5%) in the training data and 11.7% (95% CI = 9.9% to 13.5%) and 10.8% (95% CI = 9.6% to 12.2%) in the validation data sets, respectively, and were statistically significantly lower by 12.6% (95% CI = 8.8% to 16.5%, P < .001) than those estimated using previously reported timing algorithms. Similar covariates were included in both detection and timing algorithms but differed substantially from previous studies.

Conclusions

Valid and reliable detection of recurrence using data derived from electronic medical records and insurance claims is feasible. These tools will enable extensive, novel research on quality, effectiveness, and outcomes for breast cancer patients and those who develop recurrence.

Medical Care Costs for Recurrent versus De Novo Stage IV Cancer by Age at Diagnosis

OBJECTIVE

To address the knowledge gap regarding medical care costs for advanced cancer patients, we compared costs for recurrent versus de novo stage IV breast, colorectal, and lung cancer patients.

DATA SOURCES/STUDY SETTING

Virtual Data Warehouse (VDW) information from three Kaiser Permanente regions: Colorado, Northwest, and Washington.

STUDY DESIGN

We identified patients aged ≥21 with de novo or recurrent breast (nde novo  = 352; nrecurrent  = 765), colorectal (nde novo  = 1,072; nrecurrent  = 542), and lung (nde novo  = 4,041; nrecurrent  = 340) cancers diagnosed 2000-2012. We estimated average total monthly and annual costs in the 12 months preceding, month of, and 12 months following the index de novo/recurrence date, stratified by age at diagnosis (<65, ≥65). Generalized linear repeated-measures models controlled for demographics and comorbidity.

PRINCIPAL FINDINGS

In the pre-index period, monthly costs were higher for recurrent than for de novo breast (<65: +$2,431; ≥65: +$1,360), colorectal (<65: +$3,219; ≥65: +$2,247), and lung cancer (<65: +$3,086; ≥65: +$2,260) patients. Conversely, during the index and post-index periods, costs were higher for de novo patients. Average total annual pre-index costs were five- to ninefold higher for recurrent versus de novo patients <65.

CONCLUSIONS

Cost differences by type of advanced cancer and by age suggest heterogeneous patterns of care that merit further investigation.