Prospective Study Testing a Simplified Paclitaxel Premedication Regimen in Patients with Early Breast Cancer

Background

In early trials, hypersensitivity reactions (HSRs) to paclitaxel were common, thus prompting the administration of antihistamines and corticosteroids before every paclitaxel dose. We tested the safety of omitting corticosteroids after cycle 2 during the paclitaxel portion of the dose‐dense (DD) doxorubicin‐cyclophosphamide (AC)–paclitaxel regimen.

Patients, Materials, and Methods

In this prospective, single‐arm study, patients who completed four cycles of DD‐AC for stage I–III breast cancer received paclitaxel 175 mg/m² every 2 weeks for four cycles. Patients received a standard premedication protocol containing dexamethasone, diphenhydramine, and a histamine H2 blocker prior to the first two paclitaxel cycles. Dexamethasone was omitted in cycles three and four if there were no HSRs in previous cycles. We estimated the rate of grade 3–4 HSRs.

Results

Among 127 patients enrolled, 125 received more than one dose of protocol therapy and are included in the analysis. Fourteen (11.2%; 90% confidence interval, 6.9%–20.0%) patients had any‐grade HSRs, for a total of 22 (4.5%; 3.1%–6.4%) HSRs over 486 paclitaxel cycles. Any‐grade HSRs occurred in 1.6% (0.3%–5.0%), 6.5% (3.3%–11.3%), 7.4% (3.9%–12.5%), and 2.6% (0.7%–6.6%) of patients after paclitaxel cycles 1, 2, 3, and 4, respectively. Dexamethasone use was decreased by 92.8% in cycles 3 and 4. Only one patient experienced grade 3 HSR in cycles 3 or 4, for a rate of grade 3/4 HSR 0.4% (0.02%–2.0%) (1/237 paclitaxel infusions). That patient had grade 2 HSR during cycle 2, and the subsequent grade 3 event occurred despite usual dexamethasone premedication. A sensitivity analysis restricted to patients not known to have received dexamethasone in cycles 3 and 4 found that any‐grade HSRs occurred in 2.7% (3/111; 0.7%–6.8%) and 0.9% (1/109; 0.05%–4.3%) of patients in cycle 3 and 4, respectively.

Conclusion

Corticosteroid premedication can be safely omitted in cycles 3 and 4 of dose‐dense paclitaxel if HSRs are not observed during cycles 1 and 2.

Implications for Practice

Because of the potential for hypersensitivity reactions (HSRs) to paclitaxel, corticosteroids are routinely prescribed prior to each dose, on an indefinite basis. This prospective study, including 125 patients treated with 486 paclitaxel cycles, demonstrates that corticosteroids can be safely omitted in future cycles if HSRs did not occur during cycles 1 and 2 of paclitaxel and that this strategy reduces the use of corticosteroids in cycles 3 and 4 by 92.8% relative to current standard of care.

To avoid hypersensitivity reactions, corticosteroids are routinely prescribed before each dose of paclitaxel. This article reports the results of a study that focused on whether corticosteroids could be safely omitted in later cycles of treatment if reactions did not occur during earlier cycles.

Clinical Inflection Point Detection on the Basis of EHR Data to Identify Clinical Trial-Ready Patients With Cancer

PURPOSE

To inform precision oncology, methods are needed to use electronic health records (EHRs) to identify patients with cancer who are experiencing clinical inflection points, consistent with worsening prognosis or a high propensity to change treatment, at specific time points. Such patients might benefit from real-time screening for clinical trials.

METHODS

Using serial unstructured imaging reports for patients with solid tumors or lymphoma participating in a single-institution precision medicine study, we trained a deep neural network natural language processing (NLP) model to dynamically predict patients' prognoses and propensity to start new palliative-intent systemic therapy within 30 days. Model performance was evaluated using Harrell's c-index (for prognosis) and the area under the receiver operating characteristic curve (AUC; for new treatment and new clinical trial enrollment). Associations between model outputs and manual annotations of cancer progression were also evaluated using the AUC.

RESULTS

A deep NLP model was trained and evaluated using 302,688 imaging reports for 16,780 patients. In a held-out test set of 34,770 reports for 1,952 additional patients, the model predicted survival with a c-index of 0.76 and initiation of new treatment with an AUC of 0.77. Model-generated prognostic scores were associated with annotation of cancer progression on the basis of manual EHR review (n = 1,488 reports for 110 patients with lung or colorectal cancer) with an AUC of 0.78, and predictions of new treatment were associated with annotation of cancer progression on the basis of manual EHR review with an AUC of 0.84.

CONCLUSION

Training a deep NLP model to identify clinical inflection points among patients with cancer is feasible. This approach could identify patients who may benefit from real-time targeted clinical trial screening interventions at health system scale.

Landscape Analysis of Oncology Mobile Health Applications

PURPOSE

More than 325,000 mobile health (mhealth) applications (apps) have been developed. We sought to describe the state of oncology-specific apps and to highlight areas of strength and opportunities for future development.

METHODS

We searched for oncology apps in the Apple iOS and Google Play app stores in January 2020. Apps were classified by English language support, date of last update, downloads, intended audience, intended purpose, and developer type.

RESULTS

We identified 794 oncology-specific, English language applications; only 257 (32%) met basic recency standards and were considered evaluable. Of evaluable apps, almost half (47%) were found in the Medical Store Category and the majority were free (88%). The most common intended audience was health care professionals (45%), with 28% being geared toward the general public and 27% being intended for patients. The intended function was education for 36%, clinical decision support for 19.5%, and patient support for 18%. Only 23% of education apps and 40% of clinical decision support apps reported any formal app content review process. Web developers created 61.5% of apps, scientific societies created 10%, and hospitals or health care organizations created just 6%. Of 54 studies that used mobile apps in oncology identified by a recent meta-analysis, only two could be matched to commercially available apps from our study, suggesting a substantial divide between investigation and product dissemination.

CONCLUSION

Relatively few oncology-related apps exist in the commercial marketplace, up-to-date apps are uncommon, and there is a notable absence of key oncology stakeholders in app development. Meaningful development opportunities exist.

Natural Language Processing to Ascertain Cancer Outcomes From Medical Oncologist Notes

PURPOSE

Cancer research using electronic health records and genomic data sets requires clinical outcomes data, which may be recorded only in unstructured text by treating oncologists. Natural language processing (NLP) could substantially accelerate extraction of this information.

METHODS

Patients with lung cancer who had tumor sequencing as part of a single-institution precision oncology study from 2013 to 2018 were identified. Medical oncologists' progress notes for these patients were reviewed. For each note, curators recorded whether the assessment/plan indicated any cancer, progression/worsening of disease, and/or response to therapy or improving disease. Next, a recurrent neural network was trained using unlabeled notes to extract the assessment/plan from each note. Finally, convolutional neural networks were trained on labeled assessments/plans to predict the probability that each curated outcome was present. Model performance was evaluated using the area under the receiver operating characteristic curve (AUROC) among a held-out test set of 10% of patients. Associations between curated response or progression end points and overall survival were measured using Cox models among patients receiving palliative-intent systemic therapy.

RESULTS

Medical oncologist notes (n = 7,597) were manually curated for 919 patients. In the 10% test set, NLP models replicated human curation with AUROCs of 0.94 for the any-cancer outcome, 0.86 for the progression outcome, and 0.90 for the response outcome. Progression/worsening events identified using NLP models were associated with shortened survival (hazard ratio [HR] for mortality, 2.49; 95% CI, 2.00 to 3.09); response/improvement events were associated with improved survival (HR, 0.45; 95% CI, 0.30 to 0.67).

CONCLUSION

NLP models based on neural networks can extract meaningful outcomes from oncologist notes at scale. Such models may facilitate identification of clinical and genomic features associated with response to cancer treatment.

Time-Driven Activity-Based Costing in Breast Cancer Care Delivery

BACKGROUND

Accurate measurement of healthcare costs is required to assess and improve the value of oncology care.

OBJECTIVES

We aimed to determine the cost of breast cancer care provision across collaborating health care organizations.

METHODS

We used time-driven activity-based costing (TDABC) to calculate the complete cost of breast cancer care-initial treatment planning, chemotherapy, radiation therapy, surgical resection and reconstruction, and ancillary services (e.g., psychosocial oncology, physical therapy)-across multiple hospital sites. Data were collected between December 2019 and February 2020. TDABC steps involved (1) developing process maps for care delivery pathways; (2) determine capacity cost rates for staff, medical equipment, and hospital space; (3) measure the time required for each process step, both manually through clinic observation and using data from the Real-Time Location System (RTLS); and (4) calculate the total cost of care delivery.

RESULTS

Surgical care costs ranged from $1431 for a lumpectomy to $12,129 for a mastectomy with prepectoral implant reconstruction. Radiation therapy was costed at $1224 for initial simulation and patient education, and $200 for each additional treatment. Base costs for chemotherapy delivery were $382 per visit, with additional costs driven by chemotherapy agent(s) administered. Personnel expenses were the greatest contributor to the cost of surgical care, except in mastectomy with implant reconstruction, where device costs equated to up to 60% of the cost of surgery.

CONCLUSION

The cost of complete breast cancer care depended on (1) treatment protocols; (2) patient choice of reconstruction; and (3) the need for ancillary services (e.g., physical therapy). Understanding the actual costs and cost drivers of breast cancer care delivery may better inform resource utilization to lower the cost and improve the quality of care.

Population-based estimates of survival among elderly patients with brain metastases

BACKGROUND

Prognostic estimates for patients with brain metastases (BM) stem from younger, healthier patients enrolled in clinical trials or databases from academic centers. We characterized population-level prognosis in elderly patients with BM.

METHODS

Using Surveillance, Epidemiology, and End Results (SEER)-Medicare data, we identified 9882 patients ≥65 years old with BM secondary to lung, breast, skin, kidney, esophageal, colorectal, and ovarian primaries between 2014 and 2016. Survival was assessed by primary site and evaluated with Cox regression.

RESULTS

In total, 2765 versus 7117 patients were diagnosed with BM at primary cancer diagnosis (synchronous BM, median survival = 2.9 mo) versus thereafter (metachronous BM, median survival = 3.4 mo), respectively. Median survival for all primary sites was ≤4 months, except ovarian cancer (7.5 mo). Patients with non-small-cell lung cancer (NSCLC) receiving epidermal growth factor receptor (EGFR)- or anaplastic lymphoma kinase (ALK)-based therapy for synchronous BM displayed notably better median survival at 12.5 and 20.1 months, respectively, versus 2.8 months exhibited by other patients with NSCLC; survival estimates in melanoma patients based on receipt of BRAF/MEK therapy versus not were 6.7 and 2.8 months, respectively. On multivariable regression, older age, greater comorbidity, and type of managing hospital were associated with poorer survival; female sex, higher median household income, and use of brain-directed stereotactic radiation, neurosurgical resection, or systemic therapy (versus brain-directed non-stereotactic radiation) were associated with improved survival (all P < 0.05).

CONCLUSIONS

Elderly patients with BM have a poorer prognosis than suggested by prior algorithms. If prognosis is driven by systemic and not intracranial disease, brain-directed therapy with potential for significant toxicity should be utilized cautiously.

Survival in male breast cancer (MaBC) over the past three decades

569

Background: Breast cancer mortality in women has declined significantly over the past several years. In men, it is unclear whether survival has changed over time. The aim of this study was to evaluate changes in breast cancer-specific survival (BCSS) and overall survival (OS) in MaBC over the past three decades. Methods: We evaluated men diagnosed with breast cancer between 1988 and 2017, with known cause of death reported in the Surveillance, Epidemiology, and End Results registry. Patients were categorized into 3 groups by year of diagnosis: 1988-1997, 1998-2007 and 2008-2017. BCSS and OS were estimated by Kaplan-Meier and differences between groups were compared by log-rank test. Cox proportional hazards regression was used to evaluate the independent association of tumor and patient characteristics with BCSS and OS. Results: We included 8,412 men diagnosed between 1988-1997 (N = 1,033), 1998-2007 (N = 2,938) and 2008-2017 (N = 4,441). Median age for the overall population and within each decade of diagnosis was 68 years. Median follow-up was 23.6 years, 14.3 years and 4.5 years in periods 1988-1997, 1998-2007 and 2008-2017, respectively. Overall, BCSS at 5 years was 83.5%, 83.6% and 84.3% in periods 1988-1997, 1998-2007 and 2008-2017, respectively; p = 0.8. There was no significant difference in BCSS between the three periods of diagnosis within each stage of breast cancer (stage I, II, III and IV). Among all patients, OS at 5 years was 64.7%, 67.2% and 69.3% in periods 1988-1997, 1998-2007 and 2008-2017, respectively; p = 0.01. In multivariate Cox models, older age at diagnosis, black race, grade 3 disease, increasing stage, hormone receptor negative status and no surgery were all independently associated with worse BCSS and OS. In these adjusted Cox models, each additional year of diagnosis had no significant association with BCSS (hazard ratio, 1.0; 95% CI, 0.99 – 1.01; p = 0.78), and a significant improvement in OS (hazard ratio, 0.99; 95% CI, 0.98 – 0.99; p = 0.01). Conclusions: Over the past three decades, there has been no significant improvement in BCSS in MaBC. The changes in OS over time suggest increasing life expectancy. Efforts to improve BCSS in MaBC are warranted.

Barriers to web-based symptom management systems (web-SyMS)

6545

Background: Web-SyMS can reduce the burdens of cancer and its treatment. While patients frequently express willingness to use these systems, only a subset actively engages with them. Some patients may lack the tools and confidence needed to benefit from web-SyMS. We sought to characterize these barriers among community-based cancer patients receiving care across six diverse healthcare systems. Methods: We surveyed patients receiving chemotherapy at three healthcare systems (Baptist, TN; Maine Medical, ME; Dana-Farber, MA) and patients recovering from cancer-directed surgery at three healthcare systems (West Virginia University, WV; Dartmouth-Hitchcock, NH; Lifespan, RI). Surveys were conducted as part of a pre-implementation analysis of eSyM – an EHR-embedded web-SyMS that collects, tracks, and manages patient reported outcomes during cancer therapy. Results: Among 563 respondents, access to tech devices (i.e., tablet, computer, or smartphone) was high: 78% reported access to ≥2 devices and only 5% reported access to no devices. However, confidence using tech devices to accomplish online tasks varied: 45% very confident, 38% somewhat confident, 11% little-no confidence. Compared to medical oncology patients, surgery patients were more likely to report being very confident (57% vs. 31%). There were significant differences based on patients’ self-reported tech confidence (Chi-square P<.05 for all values in the table). Conclusions: Low self-reported tech confidence may identify patients who are at high risk for experiencing the burdens of cancer but may be less likely to benefit from web-SyMS. Addressing this barrier is critical to improving outcomes and addressing disparities. Clinical trial information: NCT03850912. [Table: see text]

Factors associated with late risks of breast cancer-specific mortality in the SEER registry

Purpose

Most reports describing the risk of late relapse in breast cancer (BC) have been based on selected patients enrolled into clinical trials. We examined population-based long-term risks of BC-specific mortality (BCSM), the risks of BCSM conditional on having survived 5 years, and factors associated with late BCSM.

Methods

Using SEER, we identified women diagnosed with BC (T1-T2, N0-N2, M0) between 1990 and 2005 with known hormone receptor (HR) status. Kaplan–Meier analyses determined cumulative risks of BCSM. We performed Fine and Gray regression stratified by HR status.

Results

We included 202,080 patients (median follow-up of 14.17 years). Of all BC deaths, the proportion that occurred after 5 years was 65% for HR-positive vs 28% for HR-negative (p < 0.001) BC. In HR-positive BC, the cumulative risks of BCSM during years 5–20 were 9.9%, 21.9%, and 38% for N0, N1, and N2 disease. For HR-negative BC, the risks were 7.9%, 12.2%, and 19.9%, respectively. For T1a/b, N0, HR-positive BC, the risk of BCSM was 6 times lower than the risk of non-BCSM. In N2, HR-positive BC, the risk of BCSM was 43% higher than the risk of non-BCSM. In adjusted Fine and Gray models stratified by HR status, the risks of BCSM conditional on having survived 5 years for both HR-positive and HR-negative depended on T-N status, age, and year of diagnosis. In HR-positive, the risks also depended on race and grade.

Conclusion

The risks of BCSM beyond 5 years, although different, remain important for both HR-positive and HR-negative BC. Strategies to prevent early and late recurrences are warranted.

Supplementary Information

The online version contains supplementary material available at 10.1007/s10549-021-06233-4.

Sociodemographic Factors Associated With Rapid Relapse in Triple-Negative Breast Cancer: A Multi-Institution Study

BACKGROUND

Triple-negative breast cancer (TNBC) accounts for disproportionately poor outcomes in breast cancer, driven by a subset of rapid-relapse TNBC (rrTNBC) with marked chemoresistance, rapid metastatic spread, and poor survival. Our objective was to evaluate clinicopathologic and sociodemographic features associated with rrTNBC.

METHODS

We included patients diagnosed with stage I-III TNBC in 1996 through 2012 who received chemotherapy at 1 of 10 academic cancer centers. rrTNBC was defined as a distant metastatic recurrence event or death ≤24 months after diagnosis. Features associated with rrTNBC were included in a multivariable logistic model upon which backward elimination was performed with a P<.10 criterion, with a final multivariable model applied to training (70%) and independent validation (30%) cohorts.

RESULTS

Among all patients with breast cancer treated at these centers, 3,016 fit the inclusion criteria. Training cohort (n=2,112) bivariable analyses identified disease stage, insurance type, age, body mass index, race, and income as being associated with rrTNBC (P<.10). In the final multivariable model, rrTNBC was significantly associated with higher disease stage (adjusted odds ratio for stage III vs I, 16.0; 95% CI, 9.8-26.2; P<.0001), Medicaid/indigent insurance, lower income (by 2000 US Census tract), and younger age at diagnosis. Model performance was consistent between the training and validation cohorts. In sensitivity analyses, insurance type, low income, and young age were associated with rrTNBC among patients with stage I/II but not stage III disease. When comparing rrTNBC versus late relapse (>24 months), we found that insurance type and young age remained significant.

CONCLUSIONS

Timing of relapse in TNBC is associated with stage of disease and distinct sociodemographic features, including insurance type, income, and age at diagnosis.

Twenty-year risks of breast cancer-specific mortality for stage III breast cancer in the surveillance, epidemiology, and end results registry

PURPOSE

We aimed to report the 20-year risk of breast cancer-specific mortality (BCSM), report the risk of BCSM conditional on having survived 5 years, and identify factors associated with late deaths in stage III breast cancer.

METHODS

Using Surveillance, Epidemiology, and End Results data, we included women with stage III breast cancer diagnosed from 1990 to 2005. We excluded women with unknown hormone receptor (HR) status, women who did not undergo resection of the primary tumor or axillary nodes, or unknown cause of death. We estimated risks of BCSM using cumulative incidence function and used Fine and Gray regression to identify factors associated with late deaths.

RESULTS

Final sample was 36,500 patients with 14 years of median follow-up. For each stage subgroup, the risk of BCSM at 20 years was significantly higher for HR-negative vs HR-positive tumors (IIIA: 49.8% vs 43.2%, P < 0.0001; IIIB: 60.9% vs 47.6%, P < 0.0001; IIIC: 68.7% vs 63.1%, P < 0.0001). Compared with the risks of non-BCSM, the risks of BCSM at 20 years were four times higher in stage IIIC HR-positive disease and seven times higher in stage IIIC HR-negative disease. Risks of BCSM conditional on having survived 5 years depended on tumor size, nodal status, race, and tumor grade for HR-positive disease and depended on tumor size, nodal status, and age for HR-negative disease.

CONCLUSIONS

In stage III breast cancer, the risk of BCSM at 20 years is very high and remains important even beyond the first 5 years in both HR-positive and HR-negative disease. Late BCSM depends on traditional clinicopathologic factors.

Abstract PS6-13: Population-based tool to estimate residual risks of breast cancer specific mortality (BCSM) and non-BCSM

Background: The risk of breast cancer death persists for at least 20 years from diagnosis. Most reports describing this risk have been based on selected patients (pts) enrolled into clinical trials. These studies report absolute risks at fixed timepoints (i.e. 10 or 20 years) and do not consider the dynamic changes in risks over time. The aim of this study was to develop a tool to calculate residual risks of BCSM and non-BCSM based on individual pt and tumor characteristics, at any given time after breast cancer diagnosis. Methods: Using data from the Surveillance, Epidemiology, and End Results (SEER) program, we identified women diagnosed with non-metastatic invasive breast cancer between 1990-2005, with one primary cancer in their lifetime, and known hormone receptor (HR) status. We estimated the effect of baseline clinical and pathologic variables known to be prognostic, including pt age, HR status, tumor size (T), nodal status (N), and tumor grade, on residual cumulative risks of BCSM and non-BCSM over time. The tool generates the residual death risk (RDR), which is a nonparametric estimate of the cumulative risk of BCSM and non-BCSM by year 20 after any given time from initial diagnosis, among patients defined by baseline clinical and pathologic variables using the method of Gray (1988) implemented in the cmprsk package in R. Results: We included 235,015 pts (median follow-up = 14 years). Among all breast cancer deaths, the proportion occurring after 5 years was 60% for HR+ vs 24% for HR- (p&lt;0.001). The table shows risks of BCSM and non-BCSM by HR and N status. The cumulative risk of BCSM in year 5-20 ranged from 4.2% in HR+ T1a N0 to 50.1% in HR+ N3. Using the RDR tool, a 54 year-old pt, diagnosed 7 years prior with a HR+, T1c, N1, grade 2 breast cancer, has a 16.6% residual cumulative risk of BCSM over the following 13 years, and a residual cumulative risk of non-BCSM over the same period of 4.0%. For a 66 year-old pt, diagnosed 10 years prior with a HR+, T1c, N0, grade 1 breast cancer, her residual cumulative risk of BCSM over the following 10 years is 2.9%, and her residual cumulative risk of non-BCSM over the same period is 10.0%. For a 45 year-old pt, diagnosed 8 years prior with a HR-, T2, N1, grade 3 breast cancer, her residual cumulative risks of BCSM and non-BCSM over the following 12 years are 4.4% and 4.9%, respectively. Conclusions: For HR+ breast cancer, risks of BCSM remain high beyond 5 years from diagnosis. For HR- breast cancer, the risk of BCSM is highest within 5 years from diagnosis; however, risks beyond 5 years are still considerable. The RDR tool provides population-based long-term estimates of cumulative risk of BCSM and non-BCSM over time, based on individual pt and tumor characteristics.

BCSMnon-BCSMAll-cause mortality% Event-FreeCumulative risk (%)Cumulative risk (%)Cumulative risk (%)at 5 yat 10 yy 5-20y 0-20y 0-20y 0-20Nodal status by HR status (any T)HR+N097.293.98.610.633.243.8N192.183.920.025.225.550.8N281.566.236.145.621.266.8N367.749.150.163.116.179.2HR-N089.285.47.317.022.739.7N174.268.312.234.317.051.3N257.049.320.053.514.568.0N340.733.526.268.79.378.0Tumor size among N0 onlyHR+T1a99.097.64.25.030.435.4T1b98.997.25.15.934.440.2T1c97.694.38.710.532.943.3HR-T1a97.394.74.97.323.030.3T1b95.191.96.210.527.137.7T1c91.787.97.514.823.037.8

Citation Format: Jose P Leone, Sara M Tolaney, Bernardo A Leone, Rachel A Freedman, Michael J Hassett, Julieta Leone, Carlos T Vallejo, Eric P Winer, Nancy U Lin, Nabihah Tayob. Population-based tool to estimate residual risks of breast cancer specific mortality (BCSM) and non-BCSM [abstract]. In: Proceedings of the 2020 San Antonio Breast Cancer Virtual Symposium; 2020 Dec 8-11; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2021;81(4 Suppl):Abstract nr PS6-13.

Frequency and distribution of gastrointestinal oncology patient-reported symptomatic adverse events (SAEs) at a comprehensive cancer center

463

Background: In clinical trials, the systematic collection of patient (pt) reported outcomes has been shown to improve quality of life & overall survival. To develop predictive care models for symptom management, we explored the frequency & distribution of SAEs reported by pts who reported electronic patient reported outcomes (ePRO) prior to outpatient visits to the Gastrointestinal Cancer Center (GCC) at Dana Farber Cancer Institute (DFCI). Methods: ePRO is a modified NCI Patient Reported Outcomes – Common Terminology Criteria for Adverse Events instrument distributed weekly to GCC pts with a medical/surgical/radiation oncology encounter. Responses are available to the care team in the electronic health record. ePRO consists of presence/frequency/severity/interference 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). Responses are scored 0 to 3 (with 2 and 3 indicating moderate and severe SAEs, respectively). We examined the frequency & distribution of grade 2 and 3 SAEs in ePRO responders by age, gender, race/ethnicity. All pts had gastrointestinal cancer and an outpatient visit for treatment, symptom management, follow-up care. Results: From 9/1/2018 to 8/31/2020, 1912 unique pts responded (response rate 23%). Most respondents were age 50-69 years (58% compared to 15% age <50, 27% age ≥70; range 18-95), male (53%), white (75%). Grade 3 SAE frequencies were pain (12%), fatigue (11%), anxiety/constipation/insomnia/decreased appetite (5%), sadness/numbness and tingling/diarrhea (3%), concentration/shortness of breath (2%), nausea/rash (1%), fever/vomiting (0%). Across pts, fatigue, general pain, insomnia, anxiety were the most common grade 2 and 3 SAEs. Shortness of breath, vomiting, rash, fever were least common (Table). Conclusions: In GCC pts responding to ePRO, the most frequent SAEs were pain, fatigue, insomnia, anxiety. Shortness of breath, nausea, vomiting, diarrhea were less often severe. Pts <50 were more likely to report severe anxiety but there were no other major differences based on age, sex, race/ethnicity. Ongoing efforts will increase pt/provider engagement and develop predictive models & symptom management interventions from ePRO responses. [Table: see text]

Selection of Optimal Adjuvant Chemotherapy and Targeted Therapy for Early Breast Cancer: ASCO Guideline Update

PURPOSE

The aim of this work is to update key recommendations of the ASCO guideline adaptation of the Cancer Care Ontario guideline on the selection of optimal adjuvant chemotherapy regimens for early breast cancer and adjuvant targeted therapy for breast cancer.

METHODS

An Expert Panel conducted a targeted systematic literature review guided by a signals approach to identify new, potentially practice-changing data that might translate into revised guideline recommendations.

RESULTS

The Expert Panel reviewed abstracts from the literature review and identified one article for inclusion that reported results of the phase III, open-label KATHERINE trial. In the KATHERINE trial, patients with stage I to III human epidermal growth factor receptor 2 (HER2)-positive breast cancer with residual invasive disease in the breast or axilla after completing neoadjuvant chemotherapy and HER2-targeted therapy were allocated to adjuvant trastuzumab emtansine (T-DM1; n = 743) or to trastuzumab (n = 743). Invasive disease-free survival was significantly higher in the T-DM1 group than in the trastuzumab arm (hazard ratio, 0.50; 95% CI, 0.39 to 0.64; P < .001), and risk of distant recurrence was lower in patients who received T-DM1 than in patients who received trastuzumab (hazard ratio, 0.60; 95% CI, 0.45 to 0.79). Grade 3 or higher adverse events occurred in 190 patients (25.7%) who received T-DM1 and in 111 patients (15.4%) who received trastuzumab.

RECOMMENDATIONS

Patients with HER2-positive breast cancer with pathologic invasive residual disease at surgery after standard preoperative chemotherapy and HER2-targeted therapy should be offered 14 cycles of adjuvant T-DM1, unless there is disease recurrence or unmanageable toxicity. Clinicians may offer any of the available and approved formulations of trastuzumab, including trastuzumab, trastuzumab and hyaluronidase-oysk, and available biosimilars.Additional information can be found at www.asco.org/breast-cancer-guidelines.

Stakeholder feedback at four ePRO-naïve healthcare institutions about the need, effectiveness, and barriers to usage of a fully EHR-integrated ePRO tool

165

Background: Collecting patient-reported outcomes (PROs) is a proven method to enhance doctor-patient communication and care. With the influx of technology and usage of telehealth services, electronic PROs (ePROS) have become the mainstay for ascertaining how a patient is doing at home. Collection of ePROs is particularly valuable for providers caring for rural and vulnerable populations with limited access to high quality care. A fully EHR-integrated ePRO collection system could help bridge the gap. Methods: To inform the design, function, and deployment of a new EHR-integrated ePRO symptom management system, focus groups with stakeholders were conducted at four institutions caring for largely rural-based populations (Baptist Cancer Center, West Virginia University Cancer Institute, Dartmouth-Hitchcock Medical Center, Maine Medical Center). Sessions were conducted 2 to 3 months prior to the launch of a new ePRO platform and included oncologists, surgeons, practice nurses, tech analysts, operations staff, and institutional leadership. Each group included a 30-minute overview of the new tool followed by a 30-minute discussion with qualitative open-ended questions and clicker-enabled multiple-choice questions. Developed questions utilized the CFIR and RE-AIM implementation frameworks. Results: In total, 134 stakeholders participated from the four institutions. RNs made up nearly half of respondents (47%). 97% of participants felt a new ePRO system would complement existing healthcare initiatives and 64% felt it would be extremely effective/very effective in improving symptom management. Each group was asked to rate the barriers to patient usage of an ePRO system in the home-care setting. Computer literacy (51%) and access to an internet-enabled device (48%) ranked as the highest barriers. Other barriers perceived to be of less significance included privacy, distrust, and limited English-language proficiency. Consequently, two-thirds of respondents felt patients would only be somewhat likely/not so likely to use an ePRO system; one-third felt patients would be likely/extremely likely to utilize the system. Conclusions: From the perspective of stakeholders at four engaged institutions, an integrated ePRO tool is considered a widely acceptable symptom management solution, but uncertainty remains around patient acceptance and uptake. Future research will include post-implementation discussions with stakeholders and evaluation of patient utilization and clinical outcomes.

Development of self-management tip sheets for medical oncology and surgical patients electronically reporting symptoms in the home-care recovery setting

299

Background: Patients receiving cancer treatments, including chemotherapy and surgery, often face immense morbidities. Poor symptom control frequently leads to decreased quality of life and an increased need for acute care services. For patients undergoing chemo, adverse side effects can deter them from receiving life-saving therapies. Similarly, poorly managed postoperative symptoms can delay recovery and timely receipt of adjuvant therapies. Empowering patients to proactively monitor, electronically report, and effectively treat symptoms in the home-care setting is critical to improving clinical outcomes. Methods: Through the NCI’s Moonshot-funded IMPACT consortium, 6 health systems developed a library of 70 open source symptom management tip sheets for medical oncology and surgical patients. The study team went through an iterative process with medical oncologists, surgeons, practice nurses, health educators, and patient advocates. Careful attention was paid to minimize the usage of regional dialects or idioms to ensure scalability and acceptability. The tip sheets achieved passing scores on two validated healthy literacy and readability tools. Results: Tip sheets were accessible to patients participating in the novel eSyM (electronic symptom management) program, a fully EHR-integrated ePRO model.eSyM and the incorporated tips were deployed at four health systems between fall 2019 and spring 2020 (Baptist Cancer Center, West Virginia University, Dartmouth-Hitchcock Medical Center, and Maine Medical Center). Patients enrolled in eSyM had access to the tip sheet library through their patient portal and could view them at any time. In addition, after completing an ePRO questionnaire, patients were given dedicated links to the tips for symptoms they reported. Each developed tip sheet included 4 sections: 1) things you can do on your own, 2) with over-the-counter medications, 3) with the help of your care team, 4) when to call your care team for help. This simplified structure allowed patients to determine how to manage symptoms on their own and when to seek out assistance. Conclusions: Presenting self-management tip sheets in response to patient-reported symptoms through a fully integrated patient portal platform is a novel approach to symptom management. Future efforts will include deploying the library and platform at two additional health institutions and evaluating the adoption, acceptability, and utilization of the tip sheets and their impact on clinical care outcomes.

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

164

Background: Chemotherapy and surgery can cause distressing symptoms, which can be a burden for health systems 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 overcame 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.

Resource utilization rates among English versus limited English proficient patients (pts) by patient-report of low health literacy (LHL)

107

Background: About 30 million people in the US report Limited-English Proficiency (LEP). LEP cancer pts are less likely to understand their medical condition(s) and are at increased risk of LHL, emergency department (ED) visits or hospitalizations. We examined the relationship between LEP, LHL, and ED visits/hospitalization in oncology. Methods: Dana-Farber Cancer Institute’s New Pt Intake Questionnaire (NPIQ) documents clinical and social determinants of health, including LHL. Pts reported LHL if they responded “a little bit”, “somewhat” or “not at all” to 1 of 2 questions: 1) “How confident are you in filling out medical forms?” and 2) “How confident are you in understanding medical statistics?”. Pts reported LEP if they noted a primary language other than English at registration. ED visits/hospitalizations were determined from Partners Healthcare System records. Statistically significant relationships between LEP, LHL and ED visits/hospitalizations and pt demographics (age, sex, race/ethnicity, zip code) and clinical (disease center, treatment intent) characteristics were determined with χ2 tests. Results: From 5/30/15 – 4/30/20, 21570 of 98200 eligible pts responded to NPIQ (response rate 22.0%). LHL differed by age (p-value < 0.001), gender (p-value < 0.001) and race/ethnicity (p-value = 0.007). Among LEP pts reporting LHL, financial distress (p-value = 0.004), emotional distress (PROMIS score; p-value = 0.014), and prior cancer (p-value = 0.006) were more prevalent; however, there was no significant statistical increase in ED visits (p-value = 0.237) or hospitalizations (p-value = 0.965) compared to LEP not reporting LHL. Conclusions: The results indicate that sociodemographic and other pt characteristics contribute to ED and hospital utilization in LEP cancer pts. Future studies will employ prospective data to examine the covariates’ predictive ability for resource utilization with LHL among LEP pts. [Table: see text]

Empirical power comparison of statistical tests in contemporary phase III randomized controlled trials with time-to-event outcomes in oncology

BACKGROUND

More than 95% of recent cancer randomized controlled trials used the log-rank test to detect a treatment difference making it the predominant tool for comparing two survival functions. As with other tests, the log-rank test has both advantages and disadvantages. One advantage is that it offers the highest power against proportional hazards differences, which may be a major reason why alternative methods have rarely been employed in practice. The performance of statistical tests has traditionally been investigated both theoretically and numerically for several patterns of difference between two survival functions. However, to the best of our knowledge, there has been no attempt to compare the performance of various statistical tests using empirical data from past oncology randomized controlled trials. So, it is unknown whether the log-rank test offers a meaningful power advantage over alternative testing methods in contemporary cancer randomized controlled trials. Focusing on recently reported phase III cancer randomized controlled trials, we assessed whether the log-rank test gave meaningfully greater power when compared with five alternative testing methods: generalized Wilcoxon, test based on maximum of test statistics from multiple weighted log-rank tests, difference in t-year event rate, and difference in restricted mean survival time with fixed and adaptive τ.

METHODS

Using manuscripts from cancer randomized controlled trials recently published in high-tier clinical journals, we reconstructed patient-level data for overall survival (69 trials) and progression-free survival (54 trials). For each trial endpoint, we estimated the empirical power of each test. Empirical power was measured as the proportion of trials for which a test would have identified a significant result (p value < .05).

RESULTS

For overall survival, t-year event rate offered the lowest (30.4%) empirical power and restricted mean survival time with fixed τ offered the highest (43.5%). The empirical power of the other types of tests was almost identical (36.2%-37.7%). For progression-free survival, the tests we investigated offered numerically equivalent empirical power (55.6%-61.1%). No single test consistently outperformed any other test.

CONCLUSION

The empirical power assessment with the past cancer randomized controlled trials provided new insights on the performance of statistical tests. Although the log-rank test has been used in almost all trials, our study suggests that the log-rank test is not the only option from an empirical power perspective. Near universal use of the log-rank test is not supported by a meaningful difference in empirical power. Clinical trial investigators could consider alternative methods, beyond the log-rank test, for their primary analysis when designing a cancer randomized controlled trial. Factors other than power (e.g. interpretability of the estimated treatment effect) should garner greater consideration when selecting statistical tests for cancer randomized controlled trials.

Association between the 21-gene recurrence score and isolated locoregional recurrence in stage I-II, hormone receptor-positive breast cancer

Background

Although the 21-gene recurrence score (RS) assay is widely used to predict distant recurrence risk and benefit from adjuvant chemotherapy among women with hormone receptor-positive (HR+) breast cancer, the relationship between the RS and isolated locoregional recurrence (iLRR) remains poorly understood. Therefore, we examined the association between the RS and risk of iLRR for women with stage I-II, HR+ breast cancer.

Methods

We identified 1758 women captured in the national prospective Breast Cancer-Collaborative Outcomes Research Database who were diagnosed with stage I-II, HR+ breast cancer from 2006 to 2012, treated with mastectomy or breast-conserving surgery, and received RS testing. Women who received neoadjuvant therapy were excluded. The association between the RS and risk of iLRR was examined using competing risks regression.

Results

Overall, 19% of the cohort (n = 329) had a RS ≥25. At median follow-up of 29 months, only 22 iLRR events were observed. Having a RS ≥25 was not associated with a significantly higher risk of iLRR compared to a RS < 25 (hazard ratio 1.14, 95% confidence interval 0.39–3.36, P = 0.81). When limited to women who received adjuvant endocrine therapy without chemotherapy (n = 1199; 68% of the cohort), having a RS ≥25 (n = 74) was significantly associated with a higher risk of iLRR compared to a RS < 25 (hazard ratio 3.66, 95% confidence interval 1.07–12.5, P = 0.04). In this group, increasing RS was associated with greater risk of iLRR (compared to RS < 18, hazard ratio of 1.66, 3.59, and 7.06, respectively, for RS 18–24, 25–30, and ≥ 31; P_trend = 0.02).

Conclusions

The RS was significantly associated with risk of iLRR in patients who did not receive adjuvant chemotherapy. The utility of the RS in identifying patients who have a low risk of iLRR should be further studied.

Abstract 2062: Delta prognosis: A novel clinical outcome based on automated analysis of unstructured cancer EHR data

Introduction: Efforts such as AACR Project GENIE pool genomic data across institutions to facilitate understanding of the clinical impact of genomic biomarkers in cancer. However, correlating genomic data with treatment outcomes is challenging, since outside clinical trials, outcomes such as improving or worsening cancer may be recorded only in electronic health record (EHR) free text. Automated methods could accelerate discovery by reliably characterizing treatment outcomes from unstructured EHR data.

Methods: We analyzed unstructured imaging reports for patients with solid tumors (any stage) who participated in an institutional next generation sequencing study from 2013-2018 and received palliative-intent systemic therapy. A recurrent neural network was trained to predict overall survival (OS) following each report, using text from the report and the sequence of prior reports for the patient. Model performance was measured using the concordance (c)-index in a 10% held-out validation set. Next, the ‘delta prognosis, (ΔP)' was measured for each report. ‘Delta prognosis' (ΔP) captures the change in the number of months a patient is predicted to be alive out of the 12 months following each report, compared to the exponentially weighted moving average of prior prognostic predictions for that patient. Finally, a second neural network was trained to predict ΔP using each report. Ten-fold cross validation was used to annotate each report so ΔP measures for each patient would not be based on that patient's actual survival time. Using joint modeling of ΔP, biomarker status, and report frequency, associations were measured between ΔP and previously described prognostic biomarkers for cancer types in which prolonged survival with advanced disease is common, including BRAF mutations in colorectal cancer and BRCA1/2 mutations in ovarian cancer.

Results: Model training was performed using 78,371 imaging reports for 4,512 patients with 273 cancer histologies. In a validation set of 9,306 reports for 595 patients, the c-index for survival prediction was 0.76. Within the first 6 months of palliative-intent therapy, among 597 patients with colorectal cancer, BRAF mutations were associated with worse mean ΔP (-0.33 months per report; 95% CI, -0.55 to -0.11 months; p = 0.003); among 395 patients with ovarian cancer, BRCA1/2 mutations were associated with better mean ΔP (+0.39 months per report; 95% CI, 0.17-0.61 months; p &lt; 0.001). The latter association could not have been evaluated statistically using an overall survival endpoint, since no patients with BRCA1/2 mutations died within 6 months.

Conclusion: Neural networks trained to identify shifts in prognosis using EHR text may enable ascertainment of improving and worsening cancer with no manual labeling, even if overall survival data are unavailable or immature at inference time. This technique could be relevant to any analysis of cancer outcomes using EHR data.

Citation Format: Kenneth L. Kehl, Haitham A. Elmarakeby, Michael J. Hassett, Eliezer M. Van Allen, Deb Schrag. Delta prognosis: A novel clinical outcome based on automated analysis of unstructured cancer EHR data [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 2062.

Abstract 2063: Deep natural language processing for automated ascertainment of cancer outcomes from clinician progress notes

Introduction: Clinical research using genomic datasets, such as AACR Project GENIE, requires outcomes such as cancer progression and response to contextualize molecular information. We are developing the “PRISSMM” (Pathology, Radiology/Imaging, Signs/Symptoms, Medical oncologist assessment, and tumor Markers) framework for clinical curation of genomic data. Natural language processing (NLP) models based on this framework could accelerate curation of reproducible endpoints. However, the application of NLP at scale to extract outcomes from oncologist notes, which mix historical and current information, has been limited to date.

Methods: Medical oncologists' progress notes were reviewed for patients with lung cancer whose tumors were sequenced through an institutional precision medicine study from 2013-2018. For each note, curators recorded whether the assessment/plan indicated the presence of (a) any cancer, (b) progression/worsening of disease, and/or (c) response to therapy/improvement of disease. Next, a recurrent neural network was trained to extract the assessment/plan from each note. Finally, convolutional neural networks were trained on the assessments/plans to predict the probability that each curated outcome was present. Model performance was evaluated among a held-out 10% test subset of patients using the area under the receiver-operating characteristic curve (AUC) and area under the precision-recall curve (AUPRC). Associations between curated response or progression endpoints (generated using 10-fold cross-validation) and overall survival were measured using Cox models, treating the endpoints as time-varying covariates, among patients receiving palliative-intent systemic therapy.

Results: Results among 7,597 curated notes for 919 patients are indicated in the Table.

EndpointAUC of NLP models for identifying endpoint in the test setProportion of manually curated notes with endpointAUPRC of NLP models for identifying endpoint in the test setHR (95% CI) for mortality associated with endpoint, as manually curated, among patients receiving palliative- intent treatmentHR (95% CI) for mortality associated with endpoint, as predicted using NLP models using F1-optimal threshold probabilitiesAny evidence of lung cancer0.940.770.97N/AN/AProgression0.860.200.652.93 (2.33-3.67)2.49 (2.00-3.09)Response to treatment0.900.120.570.70 (0.47-1.03)0.45 (0.30-0.67)

Conclusion: Neural network NLP models can extract meaningful outcomes from oncologist notes for clinical curation of electronic health records at scale.

Citation Format: Kenneth L. Kehl, Wenxin Xu, Haitham A. Elmarakeby, Michael J. Hassett, Jackson Nyman, Bruce E. Johnson, Eliezer M. Van Allen, Deb Schrag. Deep natural language processing for automated ascertainment of cancer outcomes from clinician progress notes [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 2063.