ASCO RWE Research Image

Real-world evidence at ASCO21: Exploring Flatiron research themes

Last updated: June 5, 2021

Author

  • Neal Meropol, MD

    Neal Meropol, MD

    Vice President, Head of Medical and Scientific Affairs

Flatiron’s presentations at ASCO 2021 demonstrate some of the latest applications of EHR-derived real-world data (RWD) to advance cancer care through research and the development of point-of-care tools. We continue to leverage our partnerships with academia, life sciences companies, and community oncologists to conduct research to advance drug development and patient care.

Aligned with the ASCO 2021 theme of “Equity: Every Patient. Every Day. Everywhere.”, this year’s highlights include an investigation of racial disparities in treatment patterns among breast cancer patients, as well as the development and application of a machine learning tool that can ultimately improve access for patients to clinical trials by improving the efficiency of patient ascertainment, thereby removing a key barrier for practices to take part in studies of rare populations.

An example of the power of RWD in aggregating clinical information in rare settings is our study of treatment patterns and outcomes in patients with Castleman disease. Additionally, to reduce the need for human curation in RWD research, we will be presenting a deep learning algorithm that automates the extraction of dates of key clinical events from unstructured chart notes. We leverage our clinico-genomic database, in partnership with Foundation Medicine, to understand implications of NGS-defined biomarkers in patients with breast cancer and gastroesophageal cancers. And finally, we would be remiss not to touch upon how the ongoing COVID-19 pandemic has demonstrated the value of EHR-derived RWD as a source of rapid insight to describe the impact of the evolving COVID-19 pandemic on cancer treatment.

Flatiron’s research continues to demonstrate the important applications, insights and impact of real-world data derived from electronic health records in discovery, therapeutic development, and cancer care -- with one critical goal in mind - to improve lives by learning from the experience of every patient with cancer.

See all the research that utilized Flatiron RWD, including three oral abstracts, on the ASCO website.

Research Focus

Recognizing racial disparities in the utilization of novel therapies

Racial disparities in second-line treatment patterns and overall survival among real-world patients with hormone receptor positive HER2 negative metastatic breast cancer

Xiaoliang Wang et al.

Xiaoliang Wang et al.

There is a similar incidence of breast cancer among African American and White women in the US, yet African American women have 40% higher breast cancer mortality. Prior studies exploring this racial disparity have focused primarily on first-line treatment, with little data known about racial differences in treatment beyond first-line and the consequential impact these differences may have on breast cancer outcomes. This analysis included patients with metastatic breast cancer initiating second-line treatment and observed associated outcomes for this patient population.

Why this matters

Racial disparities in healthcare delivery and outcomes are a major societal concern. Solutions have to start with gaining a deep grasp of the multitude of factors that contribute to the problem. Studies have to tackle issues of potential pathophysiological differences, gaps in access to care, differences in management and more.

This is why it is important to investigate detailed clinical settings to unravel where the root causes for divergent outcomes may lie. This study offers an example of that approach, in a setting (breast cancer) in dire need for a better understanding of disparity drivers.

View the full poster on the ASCO website

There is a similar incidence of breast cancer among African American and White women in the US, yet African American women have 40% higher breast cancer mortality. Prior studies exploring this racial disparity have focused primarily on first-line treatment, with little data known about racial differences in treatment beyond first-line and the consequential impact these differences may have on breast cancer outcomes. This analysis included patients with metastatic breast cancer initiating second-line treatment and observed associated outcomes for this patient population.

Why this matters

Racial disparities in healthcare delivery and outcomes are a major societal concern. Solutions have to start with gaining a deep grasp of the multitude of factors that contribute to the problem. Studies have to tackle issues of potential pathophysiological differences, gaps in access to care, differences in management and more.

This is why it is important to investigate detailed clinical settings to unravel where the root causes for divergent outcomes may lie. This study offers an example of that approach, in a setting (breast cancer) in dire need for a better understanding of disparity drivers.

View the full poster on the ASCO website

View more arrow_viewmore_02

Research Focus

The impact of the COVID-19 pandemic on the practice of cancer care

Time to treatment initiation for patients with advanced cancer documented during the COVID-19 pandemic

Samuel U Takvorian et al.

Samuel U Takvorian et al.

The COVID-19 pandemic disrupted US healthcare delivery, however the impact on diagnosis and timely care for patients with cancer is unknown. Researchers assessed the pandemic’s impact on time from metastatic diagnosis to systemic treatment initiation for patients diagnosed with solid tumors.

Why this matters

Researchers are taking stock of the extent of disruption brought by the COVID-19 pandemic onto healthcare delivery, and real-world data sources are becoming instrumental in this endeavor. In this study, EHR-derived data proves to be a unique source to characterize and understand the interplay between timing management and the mitigating strategies applied by providers in routine care.

While further and longer term research will be required to learn the ultimate effect on patient outcomes, these initial observations may already be invaluable to guide efforts towards improving resilience in healthcare delivery systems.

View the full poster on the ASCO website
Fewer advanced diagnoses in 2020 than 2019, with greater proportion de novo
Figure: Fewer advanced diagnoses in 2020 than 2019, with greater proportion de novo metastatic.

The COVID-19 pandemic disrupted US healthcare delivery, however the impact on diagnosis and timely care for patients with cancer is unknown. Researchers assessed the pandemic’s impact on time from metastatic diagnosis to systemic treatment initiation for patients diagnosed with solid tumors.

Why this matters

Researchers are taking stock of the extent of disruption brought by the COVID-19 pandemic onto healthcare delivery, and real-world data sources are becoming instrumental in this endeavor. In this study, EHR-derived data proves to be a unique source to characterize and understand the interplay between timing management and the mitigating strategies applied by providers in routine care.

While further and longer term research will be required to learn the ultimate effect on patient outcomes, these initial observations may already be invaluable to guide efforts towards improving resilience in healthcare delivery systems.

View the full poster on the ASCO website

View more arrow_viewmore_02

Research Focus

Characterizing treatment patterns and clinical outcomes of rare diseases

Clinical characteristics, treatment patterns, and overall survival of real-world patients with idiopathic multicentric Castleman disease

Aaron B. Cohen et al.

Aaron B. Cohen et al.

Idiopathic multicentric Castleman disease (iMCD) is an incredibly rare disease with both poor outcomes and poor treatment options, with only one FDA-approved therapy available (siltuximab in April 2014). Because little is known about this patient population due in part to the lack of Castleman disease-specific ICD codes until 2017, researchers used RWD to understand their clinical characteristics and treatment patterns, ultimately finding that less than half of iMCD patients diagnosed on or after siltuximab’s approval actually received it.

Why this matters

Clinical research of ultra-rare diseases continues to be a challenge, largely due to the difficulties in identifying those patients and accruing them to studies. This area remains one where RWD-based research holds the greatest promise.

This study represents a quintessential example of that approach. Investigators used a large EHR-derived database to generate a sizable cohort of patients with Castleman disease, and to extract valuable findings and insights on treatment patterns and overall outcomes.

View the full poster on the ASCO website

Idiopathic multicentric Castleman disease (iMCD) is an incredibly rare disease with both poor outcomes and poor treatment options, with only one FDA-approved therapy available (siltuximab in April 2014). Because little is known about this patient population due in part to the lack of Castleman disease-specific ICD codes until 2017, researchers used RWD to understand their clinical characteristics and treatment patterns, ultimately finding that less than half of iMCD patients diagnosed on or after siltuximab’s approval actually received it.

Why this matters

Clinical research of ultra-rare diseases continues to be a challenge, largely due to the difficulties in identifying those patients and accruing them to studies. This area remains one where RWD-based research holds the greatest promise.

This study represents a quintessential example of that approach. Investigators used a large EHR-derived database to generate a sizable cohort of patients with Castleman disease, and to extract valuable findings and insights on treatment patterns and overall outcomes.

View the full poster on the ASCO website

View more arrow_viewmore_02

Research Focus

Point-of-care EHR solutions to improve clinical trial participation

An automated EHR-based tool to improve patient identification for biomarker-driven trials

Shailendra Lakhanpal et al.

Shailendra Lakhanpal et al.

Next-generation sequencing (NGS) test results contribute to clinical trial eligibility; the lack of structured NGS results hinders the automation of trial matching for biomarker-driven trials. As a result, researchers in this study developed a machine learning tool to infer the presence of NGS results in the EHR, facilitating clinical trial matching.

Why this matters

The growing complexity of oncology clinical research demands ever more detailed clinical protocols and eligibility screening processes. The burdens associated with managing patient accrual at the clinical site level may have the detrimental effect of deterring trial participation.

This work aims to address that issue, presenting an example of a tool that lessens the burden placed on local research personnel and is easy to implement in workflows. In this case, the tool can automatically categorize patients as ‘having received NGS testing yes/no,’ an increasingly relevant criterion, based on information contained in unstructured clinical notes.

View the full poster on the ASCO website

Next-generation sequencing (NGS) test results contribute to clinical trial eligibility; the lack of structured NGS results hinders the automation of trial matching for biomarker-driven trials. As a result, researchers in this study developed a machine learning tool to infer the presence of NGS results in the EHR, facilitating clinical trial matching.

Why this matters

The growing complexity of oncology clinical research demands ever more detailed clinical protocols and eligibility screening processes. The burdens associated with managing patient accrual at the clinical site level may have the detrimental effect of deterring trial participation.

This work aims to address that issue, presenting an example of a tool that lessens the burden placed on local research personnel and is easy to implement in workflows. In this case, the tool can automatically categorize patients as ‘having received NGS testing yes/no,’ an increasingly relevant criterion, based on information contained in unstructured clinical notes.

View the full poster on the ASCO website

View more arrow_viewmore_02

Research Focus

Advancing machine learning-based EHR data extraction for RWD

Extracting non-small cell lung cancer diagnosis and diagnosis dates from electronic health record text using a deep learning algorithm

Alexander S. Rich et al.

Alexander S. Rich et al.

A patient’s cancer diagnosis and the date of that diagnosis are often not accurately recorded in a structured form in the electronic health record. These data, however, are important for selecting real-world research cohorts and conducting downstream analyses. As such, this study looked to develop a deep learning model for identifying patients with NSCLC and their initial and advanced diagnosis date(s).

Why this matters

While real-world data is expected to have major strengths (large sample sizes, representativeness of the population at large), completeness and quality are contingent on the nature of documentation at the originating point.

When EHRs are used as a source, capture of specific diagnoses and dates can lack standardization or precision in structured data fields; yet these elements are critical for outcome analyses, which then necessitate reliable and scalable approaches to the extraction of that information from unstructured sources. This is why it is important to investigate detailed clinical settings to unravel where the root causes for divergent outcomes may lie. This study offers an example of that approach, in a setting (breast cancer) in dire need for a better understanding of disparity drivers.

View the full poster on the ASCO website

A patient’s cancer diagnosis and the date of that diagnosis are often not accurately recorded in a structured form in the electronic health record. These data, however, are important for selecting real-world research cohorts and conducting downstream analyses. As such, this study looked to develop a deep learning model for identifying patients with NSCLC and their initial and advanced diagnosis date(s).

Why this matters

While real-world data is expected to have major strengths (large sample sizes, representativeness of the population at large), completeness and quality are contingent on the nature of documentation at the originating point.

When EHRs are used as a source, capture of specific diagnoses and dates can lack standardization or precision in structured data fields; yet these elements are critical for outcome analyses, which then necessitate reliable and scalable approaches to the extraction of that information from unstructured sources. This is why it is important to investigate detailed clinical settings to unravel where the root causes for divergent outcomes may lie. This study offers an example of that approach, in a setting (breast cancer) in dire need for a better understanding of disparity drivers.

View the full poster on the ASCO website

View more arrow_viewmore_02

Research Focus

Understanding the implications of NGS biomarkers and their outcomes with novel therapies

Concordance of HER2+ status by IHC/ISH and ERBB2 status by NGS and analysis of real-world outcomes in patients with metastatic breast cancer receiving first-line HER2 therapy

Cheryl D. Cho-Phan et al.

Cheryl D. Cho-Phan et al.

HER2 overexpression or amplification is a predictive biomarker for HER2-targeted therapies. With the advent of next-generation sequencing, researchers were able to examine how patients with HER2+ metastatic breast cancer and ERBB2 amplification responded to 1L HER2 therapy. Among patients in the real-world clinicogenomic dataset, researchers saw discordance between ERBB2 amplification and in situ hybridization/immunohistochemistry (ISH/IHC) HER2 testing methods. Additionally, patients with HER2+ tumors but negative for ERBB2 amplification had worse overall survival after receiving HER2 therapy.

Why this matters

The increasing sophistication of diagnostic tools continues to deepen our understanding of tumor genomics and their clinical implications. The present study investigates nuances behind HER2 positivity, unveiling clinical differences in responsiveness to HER2-targeted therapy between patients with tumors where protein overexpression is concordant with gene amplification vs those with discordant status.

Studies like this show that even for biomarkers with a long research history behind them, such as HER2 in breast cancer, detailed investigation remains critical in order to deliver on the full promise of precision oncology and personalized medicine.

View the full poster on the ASCO website

HER2 overexpression or amplification is a predictive biomarker for HER2-targeted therapies. With the advent of next-generation sequencing, researchers were able to examine how patients with HER2+ metastatic breast cancer and ERBB2 amplification responded to 1L HER2 therapy. Among patients in the real-world clinicogenomic dataset, researchers saw discordance between ERBB2 amplification and in situ hybridization/immunohistochemistry (ISH/IHC) HER2 testing methods. Additionally, patients with HER2+ tumors but negative for ERBB2 amplification had worse overall survival after receiving HER2 therapy.

Why this matters

The increasing sophistication of diagnostic tools continues to deepen our understanding of tumor genomics and their clinical implications. The present study investigates nuances behind HER2 positivity, unveiling clinical differences in responsiveness to HER2-targeted therapy between patients with tumors where protein overexpression is concordant with gene amplification vs those with discordant status.

Studies like this show that even for biomarkers with a long research history behind them, such as HER2 in breast cancer, detailed investigation remains critical in order to deliver on the full promise of precision oncology and personalized medicine.

View the full poster on the ASCO website

View more arrow_viewmore_02

Analysis of real-world data for metastatic breast cancer patients with somatic BRCA or other homologous recombination pathway alterations treated with PARP inhibitors

Felipe Batalini et al.

Felipe Batalini et al.

In order to better understand populations of patients with metastatic breast cancer treated with PARPi, researchers assessed outcomes for patients with germline BRCA1/2 mutation compared to patients with either somatic BRCA or other HR-pathway mutations treated with PARPi. Findings suggest similar benefit from PARPi when tumor comprehensive genomic profiling detects a sBRCA mutation or germline or somatic mutation in other homologous recombination-pathway genes.

Why this matters

With the wide availability of NGS testing for routine practice, our understanding of the genetic underpinnings of cancer is becoming increasingly expansive. One underexplored aspect so far is whether the origin of known oncogenic drivers, somatic or germline, makes a difference in disease course and/or responsiveness to targeted therapy.

This work investigates this question as it relates to BRCA and other genes in the homologous recombination pathway in breast cancer, and their effect on responsiveness to PARPi therapy. In this case, the answer is that the origin of the mutation appears to have no effect on responsiveness to PARPi treatment.

View the full poster on the ASCO website
Comparison of the rwPFS (top) and rwOS (bottom) from start of PARPi in patients with gBRCA vs patients with non-gBRCA HR-pathway gene mutation.
Figure: Comparison of the rwPFS (top) and rwOS (bottom) from start of PARPi in patients with gBRCA vs patients with non-gBRCA HR-pathway gene mutation.

In order to better understand populations of patients with metastatic breast cancer treated with PARPi, researchers assessed outcomes for patients with germline BRCA1/2 mutation compared to patients with either somatic BRCA or other HR-pathway mutations treated with PARPi. Findings suggest similar benefit from PARPi when tumor comprehensive genomic profiling detects a sBRCA mutation or germline or somatic mutation in other homologous recombination-pathway genes.

Why this matters

With the wide availability of NGS testing for routine practice, our understanding of the genetic underpinnings of cancer is becoming increasingly expansive. One underexplored aspect so far is whether the origin of known oncogenic drivers, somatic or germline, makes a difference in disease course and/or responsiveness to targeted therapy.

This work investigates this question as it relates to BRCA and other genes in the homologous recombination pathway in breast cancer, and their effect on responsiveness to PARPi therapy. In this case, the answer is that the origin of the mutation appears to have no effect on responsiveness to PARPi treatment.

View the full poster on the ASCO website

View more arrow_viewmore_02

ERBB2 copy number as a quantitative biomarker for real-world outcomes to anti-HER2 therapy in advanced gastroesophageal adenocarcinoma

Samuel J Klempner et al.

Samuel J Klempner et al.

HER2 ERBB2 amplification may be a predictor of anti-HER2 therapy outcome compared to in situ hybridization/immunohistochemistry (ISH/IHC), but further research was needed. This study investigated whether an increased ERBB2 copy number was associated with a better outcome to trastuzumab in patients with advanced gastroesophageal adenocarcinoma. Results showed that ERBB2 CN was in fact predictive of real-world progression free survival for patients treated with trastuzumab.

Why this matters

The availability of diagnostic tools of increasing sensitivity is driving progress towards better personalized care. This study provides an example of that paradigm. HER2-targeted therapy benefits patients with gastroesophageal adenocarcinoma with HER2 overexpression.

The standard diagnostic approach is to measure protein overexpression, but this study indicates that genetic amplification (measured as gene copy number) may be a more refined approach to identify responders. This research demonstrates that true personalized medicine rests on two pillars: the characterization of biomarkers themselves and the optimization of the diagnostic tools for practical testing.

View the full poster on the ASCO website
Higher ERBB2 CN is associated with longer real-world progression free survival (rwPFS) in advanced GEA treated with ant-HER2 therapy in the first line setting. 
Figure: Higher ERBB2 CN is associated with longer rwPFS in advanced GEA treated with ant-HER2 therapy in the first line setting. 

HER2 ERBB2 amplification may be a predictor of anti-HER2 therapy outcome compared to in situ hybridization/immunohistochemistry (ISH/IHC), but further research was needed. This study investigated whether an increased ERBB2 copy number was associated with a better outcome to trastuzumab in patients with advanced gastroesophageal adenocarcinoma. Results showed that ERBB2 CN was in fact predictive of real-world progression free survival for patients treated with trastuzumab.

Why this matters

The availability of diagnostic tools of increasing sensitivity is driving progress towards better personalized care. This study provides an example of that paradigm. HER2-targeted therapy benefits patients with gastroesophageal adenocarcinoma with HER2 overexpression.

The standard diagnostic approach is to measure protein overexpression, but this study indicates that genetic amplification (measured as gene copy number) may be a more refined approach to identify responders. This research demonstrates that true personalized medicine rests on two pillars: the characterization of biomarkers themselves and the optimization of the diagnostic tools for practical testing.

View the full poster on the ASCO website

View more arrow_viewmore_02