Artificial Intelligence in Oncology: Current Landscape, Challenges, and Future Directions

Artificial intelligence (AI) in oncology is advancing beyond algorithm development to integration into clinical practice. This review describes the current state of the field, with a specific focus on clinical integration. AI applications are structured according to cancer type and clinical domain, focusing on the four most common cancers and tasks of detection, diagnosis, and treatment. These applications encompass various data modalities, including imaging, genomics, and medical records. We conclude with a summary of existing challenges, evolving solutions, and potential future directions for the field.

SIGNIFICANCE

AI is increasingly being applied to all aspects of oncology, where several applications are maturing beyond research and development to direct clinical integration. This review summarizes the current state of the field through the lens of clinical translation along the clinical care continuum. Emerging areas are also highlighted, along with common challenges, evolving solutions, and potential future directions for the field.

Perspectives of Oncologists on the Ethical Implications of Using Artificial Intelligence for Cancer Care

Importance

Artificial intelligence (AI) tools are rapidly integrating into cancer care. Understanding stakeholder views on ethical issues associated with the implementation of AI in oncology is critical to optimal deployment.

Objective

To evaluate oncologists' views on the ethical domains of the use of AI in clinical care, including familiarity, predictions, explainability (the ability to explain how a result was determined), bias, deference, and responsibilities.

Design, Setting, and Participants

This cross-sectional, population-based survey study was conducted from November 15, 2022, to July 31, 2023, among 204 US-based oncologists identified using the National Plan & Provider Enumeration System.

Main Outcomes and Measures

The primary outcome was response to a question asking whether participants agreed or disagreed that patients need to provide informed consent for AI model use during cancer treatment decisions.

Results

Of 387 surveys, 204 were completed (response rate, 52.7%). Participants represented 37 states, 120 (63.7%) identified as male, 128 (62.7%) as non-Hispanic White, and 60 (29.4%) were from academic practices; 95 (46.6%) had received some education on AI use in health care, and 45.3% (92 of 203) reported familiarity with clinical decision models. Most participants (84.8% [173 of 204]) reported that AI-based clinical decision models needed to be explainable by oncologists to be used in the clinic; 23.0% (47 of 204) stated they also needed to be explainable by patients. Patient consent for AI model use during treatment decisions was supported by 81.4% of participants (166 of 204). When presented with a scenario in which an AI decision model selected a different treatment regimen than the oncologist planned to recommend, the most common response was to present both options and let the patient decide (36.8% [75 of 204]); respondents from academic settings were more likely than those from other settings to let the patient decide (OR, 2.56; 95% CI, 1.19-5.51). Most respondents (90.7% [185 of 204]) reported that AI developers were responsible for the medico-legal problems associated with AI use. Some agreed that this responsibility was shared by physicians (47.1% [96 of 204]) or hospitals (43.1% [88 of 204]). Finally, most respondents (76.5% [156 of 204]) agreed that oncologists should protect patients from biased AI tools, but only 27.9% (57 of 204) were confident in their ability to identify poorly representative AI models.

Conclusions and Relevance

In this cross-sectional survey study, few oncologists reported that patients needed to understand AI models, but most agreed that patients should consent to their use, and many tasked patients with choosing between physician- and AI-recommended treatment regimens. These findings suggest that the implementation of AI in oncology must include rigorous assessments of its effect on care decisions as well as decisional responsibility when problems related to AI use arise.

Biomarker Inference and the Timing of Next-Generation Sequencing in a Multi-Institutional, Cross-Cancer Clinicogenomic Data Set

PURPOSE

Observational clinicogenomic data sets, consisting of tumor next-generation sequencing (NGS) data linked to clinical records, are commonly used for cancer research. However, in real-world practice, oncologists frequently request NGS in search of treatment options for progressive cancer. The extent and impact of this dynamic on analysis of clinicogenomic research data are not well understood.

METHODS

We analyzed clinicogenomic data for patients with non-small cell lung, colorectal, breast, prostate, pancreatic, or urothelial cancers in the American Association for Cancer Research Biopharmaceutical Consortium cohort. Associations between baseline and time-varying clinical characteristics and time from diagnosis to NGS were measured. To explore the impact of informative cohort entry on biomarker inference, statistical interactions between selected biomarkers and time to NGS with respect to overall survival were calculated.

RESULTS

Among 7,182 patients, time from diagnosis to NGS varied significantly by clinical factors, including cancer type, calendar year of sequencing, institution, and age and stage at diagnosis. NGS rates also varied significantly by dynamic clinical status variables; in an adjusted model, compared with patients with stable disease at any given time after diagnosis, patients with progressive disease by imaging or oncologist assessment had higher NGS rates (hazard ratio for NGS, 1.61 [95% CI, 1.45 to 1.78] and 2.32 [95% CI, 2.01 to 2.67], respectively). Statistical interactions between selected biomarkers and time to NGS with respect to survival, potentially indicating biased biomarker inference results, were explored.

CONCLUSION

To evaluate the appropriateness of a data set for a particular research question, it is crucial to measure associations between dynamic cancer status and the timing of NGS, as well as to evaluate interactions involving biomarkers of interest and NGS timing with respect to survival outcomes.

Identifying Oncology Clinical Trial Candidates Using Artificial Intelligence Predictions of Treatment Change: A Pilot Implementation Study

PURPOSE

Precision oncology clinical trials often struggle to accrue, partly because it is difficult to find potentially eligible patients at moments when they need new treatment. We piloted deployment of artificial intelligence tools to identify such patients at a large academic cancer center.

PATIENTS AND METHODS

Neural networks that process radiology reports to identify patients likely to start new systemic therapy were applied prospectively for patients with solid tumors that had undergone next-generation sequencing at our center. Model output was linked to the MatchMiner tool, which matches patients to trials using tumor genomics. Reports listing genomically matched patients, sorted by probability of treatment change, were provided weekly to an oncology nurse navigator (ONN) coordinating recruitment to nine early-phase trials. The ONN contacted treating oncologists when patients likely to change treatment appeared potentially trial-eligible.

RESULTS

Within weekly reports to the ONN, 60,199 patient-trial matches were generated for 2,150 patients on the basis of genomics alone. Of these, 3,168 patient-trial matches (5%) corresponding to 525 patients were flagged for ONN review by our model, representing a 95% reduction in review compared with manual review of all patient-trial matches weekly. After ONN review for potential eligibility, treating oncologists for 74 patients were contacted. Common reasons for not contacting treating oncologists included cases where patients had already decided to continue current treatment (21%); the trial had no slots (14%); or the patient was ineligible on ONN review (12%). Of 74 patients whose oncologists were contacted, 10 (14%) had a consult regarding a trial and five (7%) enrolled.

CONCLUSION

This approach facilitated identification of potential patients for clinical trials in real time, but further work to improve accrual must address the many other barriers to trial enrollment in precision oncology research.

Genomic Characterization and Clinical Outcomes of Patients with Peritoneal Metastases from the AACR GENIE Biopharma Collaborative Colorectal Cancer Registry

Peritoneal metastases (PM) are common in metastatic colorectal cancer (mCRC). We aimed to characterize patients with mCRC and PM from a clinical and molecular perspective using the American Association of Cancer Research Genomics Evidence Neoplasia Information Exchange (GENIE) Biopharma Collaborative (BPC) registry. Patients' tumor samples underwent targeted next-generation sequencing. Clinical characteristics and treatment outcomes were collected retrospectively. Overall survival (OS) from advanced disease and progression-free survival (PFS) from start of cancer-directed drug regimen were estimated and adjusted for the left truncation bias. A total of 1,281 patients were analyzed, 244 (19%) had PM at time of advanced disease. PM were associated with female sex [OR: 1.67; 95% confidence interval (CI): 1.11-2.54; P = 0.014] and higher histologic grade (OR: 1.72; 95% CI: 1.08-2.71; P = 0.022), while rectal primary tumors were less frequent in patients with PM (OR: 0.51; 95% CI: 0.29-0.88; P < 0.001). APC occurred less frequently in patients with PM (N = 151, 64% vs. N = 788, 79%) while MED12 alterations occurred more frequently in patients with PM (N = 20, 10% vs. N = 32, 4%); differences in MED12 were not significant when restricting to oncogenic and likely oncogenic variants according to OncoKB. Patients with PM had worse OS (HR: 1.45; 95% CI: 1.16-1.81) after adjustment for independently significant clinical and genomic predictors. PFS from initiation of first-line treatment did not differ by presence of PM. In conclusion, PM were more frequent in females and right-sided primary tumors. Differences in frequencies of MED12 and APC alterations were identified between patients with and without PM. PM were associated with shorter OS but not with PFS from first-line treatment.

SIGNIFICANCE

Utilizing the GENIE BPC registry, this study found that PM in patients with colorectal cancer occur more frequently in females and right-sided primary tumors and are associated with worse OS. In addition, we found a lower frequency of APC alterations and a higher frequency in MED12 alterations in patients with PM.

Effects of KRAS Genetic Interactions on Outcomes in Cancers of the Lung, Pancreas, and Colorectum

BACKGROUND

KRAS is among the most commonly mutated oncogenes in cancer, and previous studies have shown associations with survival in many cancer contexts. Evidence from both clinical observations and mouse experiments further suggests that these associations are allele- and tissue-specific. These findings motivate using clinical data to understand gene interactions and clinical covariates within different alleles and tissues.

METHODS

We analyze genomic and clinical data from the AACR Project GENIE Biopharma Collaborative for samples from lung, colorectal, and pancreatic cancers. For each of these cancer types, we report epidemiological associations for different KRAS alleles, apply principal component analysis (PCA) to discover groups of genes co-mutated with KRAS, and identify distinct clusters of patient profiles with implications for survival.

RESULTS

KRAS mutations were associated with inferior survival in lung, colon, and pancreas, although the specific mutations implicated varied by disease. Tissue- and allele-specific associations with smoking, sex, age, and race were found. Tissue-specific genetic interactions with KRAS were identified by PCA, which were clustered to produce five, four, and two patient profiles in lung, colon, and pancreas. Membership in these profiles was associated with survival in all three cancer types.

CONCLUSIONS

KRAS mutations have tissue- and allele-specific associations with inferior survival, clinical covariates, and genetic interactions.

IMPACT

Our results provide greater insight into the tissue- and allele-specific associations with KRAS mutations and identify clusters of patients that are associated with survival and clinical attributes from combinations of genetic interactions with KRAS mutations.

Analysis and Visualization of Longitudinal Genomic and Clinical Data from the AACR Project GENIE Biopharma Collaborative in cBioPortal

International cancer registries make real-world genomic and clinical data available, but their joint analysis remains a challenge. AACR Project GENIE, an international cancer registry collecting data from 19 cancer centers, makes data from >130,000 patients publicly available through the cBioPortal for Cancer Genomics (https://genie.cbioportal.org). For 25,000 patients, additional real-world longitudinal clinical data, including treatment and outcome data, are being collected by the AACR Project GENIE Biopharma Collaborative using the PRISSMM data curation model. Several thousand of these cases are now also available in cBioPortal. We have significantly enhanced the functionalities of cBioPortal to support the visualization and analysis of this rich clinico-genomic linked dataset, as well as datasets generated by other centers and consortia. Examples of these enhancements include (i) visualization of the longitudinal clinical and genomic data at the patient level, including timelines for diagnoses, treatments, and outcomes; (ii) the ability to select samples based on treatment status, facilitating a comparison of molecular and clinical attributes between samples before and after a specific treatment; and (iii) survival analysis estimates based on individual treatment regimens received. Together, these features provide cBioPortal users with a toolkit to interactively investigate complex clinico-genomic data to generate hypotheses and make discoveries about the impact of specific genomic variants on prognosis and therapeutic sensitivities in cancer.

SIGNIFICANCE

Enhanced cBioPortal features allow clinicians and researchers to effectively investigate longitudinal clinico-genomic data from patients with cancer, which will improve exploration of data from the AACR Project GENIE Biopharma Collaborative and similar datasets.

Empirical evaluation of language modeling to ascertain cancer outcomes from clinical text reports

BACKGROUND

Longitudinal data on key cancer outcomes for clinical research, such as response to treatment and disease progression, are not captured in standard cancer registry reporting. Manual extraction of such outcomes from unstructured electronic health records is a slow, resource-intensive process. Natural language processing (NLP) methods can accelerate outcome annotation, but they require substantial labeled data. Transfer learning based on language modeling, particularly using the Transformer architecture, has achieved improvements in NLP performance. However, there has been no systematic evaluation of NLP model training strategies on the extraction of cancer outcomes from unstructured text.

RESULTS

We evaluated the performance of nine NLP models at the two tasks of identifying cancer response and cancer progression within imaging reports at a single academic center among patients with non-small cell lung cancer. We trained the classification models under different conditions, including training sample size, classification architecture, and language model pre-training. The training involved a labeled dataset of 14,218 imaging reports for 1112 patients with lung cancer. A subset of models was based on a pre-trained language model, DFCI-ImagingBERT, created by further pre-training a BERT-based model using an unlabeled dataset of 662,579 reports from 27,483 patients with cancer from our center. A classifier based on our DFCI-ImagingBERT, trained on more than 200 patients, achieved the best results in most experiments; however, these results were marginally better than simpler "bag of words" or convolutional neural network models.

CONCLUSION

When developing AI models to extract outcomes from imaging reports for clinical cancer research, if computational resources are plentiful but labeled training data are limited, large language models can be used for zero- or few-shot learning to achieve reasonable performance. When computational resources are more limited but labeled training data are readily available, even simple machine learning architectures can achieve good performance for such tasks.

The GENIE BPC NSCLC Cohort: A Real-World Repository Integrating Standardized Clinical and Genomic Data for 1,846 Patients with Non-Small Cell Lung Cancer

PURPOSE

We describe the clinical and genomic landscape of the non-small cell lung cancer (NSCLC) cohort of the American Association for Cancer Research (AACR) Project Genomics Evidence Neoplasia Information Exchange (GENIE) Biopharma Collaborative (BPC).

EXPERIMENTAL DESIGN

A total of 1,846 patients with NSCLC whose tumors were sequenced from 2014 to 2018 at four institutions participating in AACR GENIE were randomly chosen for curation using the PRISSMM data model. Progression-free survival (PFS) and overall survival (OS) were estimated for patients treated with standard therapies.

RESULTS

In this cohort, 44% of tumors harbored a targetable oncogenic alteration, with EGFR (20%), KRAS G12C (13%), and oncogenic fusions (ALK, RET, and ROS1; 5%) as the most frequent. Median OS (mOS) on first-line platinum-based therapy without immunotherapy was 17.4 months [95% confidence interval (CI), 14.9-19.5 months]. For second-line therapies, mOS was 9.2 months (95% CI, 7.5-11.3 months) for immune checkpoint inhibitors (ICI) and 6.4 months (95% CI, 5.1-8.1 months) for docetaxel ± ramucirumab. In a subset of patients treated with ICI in the second-line or later setting, median RECIST PFS (2.5 months; 95% CI, 2.2-2.8) and median real-world PFS based on imaging reports (2.2 months; 95% CI, 1.7-2.6) were similar. In exploratory analysis of the impact of tumor mutational burden (TMB) on survival on ICI treatment in the second-line or higher setting, TMB z-score harmonized across gene panels was associated with improved OS (univariable HR, 0.85; P = 0.03; n = 247 patients).

CONCLUSIONS

The GENIE BPC cohort provides comprehensive clinicogenomic data for patients with NSCLC, which can improve understanding of real-world patient outcomes.

Machine learning for genetics-based classification and treatment response prediction in cancer of unknown primary

Cancer of unknown primary (CUP) is a type of cancer that cannot be traced back to its primary site and accounts for 3-5% of all cancers. Established targeted therapies are lacking for CUP, leading to generally poor outcomes. We developed OncoNPC, a machine-learning classifier trained on targeted next-generation sequencing (NGS) data from 36,445 tumors across 22 cancer types from three institutions. Oncology NGS-based primary cancer-type classifier (OncoNPC) achieved a weighted F1 score of 0.942 for high confidence predictions ([Formula: see text]) on held-out tumor samples, which made up 65.2% of all the held-out samples. When applied to 971 CUP tumors collected at the Dana-Farber Cancer Institute, OncoNPC predicted primary cancer types with high confidence in 41.2% of the tumors. OncoNPC also identified CUP subgroups with significantly higher polygenic germline risk for the predicted cancer types and with significantly different survival outcomes. Notably, patients with CUP who received first palliative intent treatments concordant with their OncoNPC-predicted cancers had significantly better outcomes (hazard ratio (HR) = 0.348; 95% confidence interval (CI) = 0.210-0.570; P = [Formula: see text]). Furthermore, OncoNPC enabled a 2.2-fold increase in patients with CUP who could have received genomically guided therapies. OncoNPC thus provides evidence of distinct CUP subgroups and offers the potential for clinical decision support for managing patients with CUP.

Practice-Level Variation in Molecular Testing and Use of Targeted Therapy for Patients With Non-Small Cell Lung Cancer and Colorectal Cancer

Importance

All patients with newly diagnosed non-small cell lung cancer (NSCLC) and colorectal cancer (CRC) should receive molecular testing to identify those who can benefit from targeted therapies. However, many patients do not receive recommended testing and targeted therapies.

Objective

To compare rates of molecular testing and targeted therapy use by practice type and across practices.

Design, Setting, and Participants

This cross-sectional study used 100% Medicare fee-for-service data from 2015 through 2019 to identify beneficiaries with new metastatic NSCLC or CRC diagnoses receiving systemic therapy and to assign patients to oncology practices. Hierarchical linear models were used to characterize variation by practice type and across practices. Data analysis was conducted from June 2019 to October 2022.

Exposures

Oncology practice providing care.

Outcomes

Primary outcomes were rates of molecular testing and targeted therapy use for patients with NSCLC and CRC. Secondary outcomes were rates of multigene testing for NSCLC and CRC.

Results

There were 106 228 Medicare beneficiaries with incident NSCLC (31 521 [29.7%] aged 65-69 years; 50 348 [47.4%] female patients; 2269 [2.1%] Asian, 8282 [7.8%] Black, and 91 215 [85.9%] White patients) and 39 512 beneficiaries with incident CRC (14 045 [35.5%] aged 65-69 years; 17 518 [44.3%] female patients; 896 [2.3%] Asian, 3521 [8.9%] Black, and 32 753 [82.9%] White patients) between 2015 and 2019. Among these beneficiaries, 18 435 (12.9%) were treated at National Cancer Institute (NCI)-designated centers, 8187 (5.6%) were treated at other academic centers, and 94 329 (64.7%) were treated at independent oncology practices. Molecular testing rates increased from 74% to 85% for NSCLC and 45% to 65% for CRC. First-line targeted therapy use decreased from 12% to 8% among patients with NSCLC and was constant at 5% for patients with CRC. For NSCLC, molecular testing rates were similar across practice types while rates of multigene panel use (13.2%) and targeted therapy use (16.6%) were highest at NCI-designated cancer centers. For CRC, molecular testing rates were 3.8 (95% CI: 1.2-6.5), 3.3 (95% CI, 0.4-6.1), and 12.2 (95% CI, 9.1-15.3) percentage points lower at hospital-owned practices, large independent practices, and small independent practices, respectively. Rates of targeted therapy use for CRC were similar across practice types. After adjusting for patient characteristics, there was moderate variation in molecular testing and targeted therapy use across oncology practices.

Conclusions and Relevance

In this cross-sectional study of Medicare beneficiaries, molecular testing rates for NSCLC and CRC increased in recent years but remained lower than recommended levels. Rates of targeted therapy use decreased for NSCLC and remained stable for CRC. Variation across practices suggests that where a patient was treated may have affected access to recommended testing and efficacious treatments.

Elucidating Analytic Bias Due to Informative Cohort Entry in Cancer Clinico-genomic Datasets

BACKGROUND

Oncologists often order genomic testing to inform treatment for worsening cancer. The resulting correlation between genomic testing timing and prognosis, or "informative entry," can bias observational clinico-genomic research. The efficacy of existing approaches to this problem in clinico-genomic cohorts is poorly understood.

METHODS

We simulated clinico-genomic cohorts followed from an index date to death. Subgroups in each cohort who underwent genomic testing before death were "observed." We varied data generation parameters under four scenarios: (i) independent testing and survival times; (ii) correlated testing and survival times for all patients; (iii) correlated testing and survival times for a subset of patients; and (iv) testing and mortality exclusively following progression events. We examined the behavior of conditional Kendall tau (Tc) statistics, Cox entry time coefficients, and biases in overall survival (OS) estimation and biomarker inference across scenarios.

RESULTS

Scenario #1 yielded null Tc and Cox entry time coefficients and unbiased OS inference. Scenario #2 yielded positive Tc, negative Cox entry time coefficients, underestimated OS, and biomarker associations biased toward the null. Scenario #3 yielded negative Tc, positive Cox entry time coefficients, and underestimated OS, but biomarker estimates were less biased. Scenario #4 yielded null Tc and Cox entry time coefficients, underestimated OS, and biased biomarker estimates. Transformation and copula modeling did not provide unbiased results.

CONCLUSIONS

Approaches to informative clinico-genomic cohort entry, including Tc and Cox entry time statistics, are sensitive to heterogeneity in genotyping and survival time distributions.

IMPACT

Novel methods are needed for unbiased inference using observational clinico-genomic data.

Disparities in the Inclusion of Racial and Ethnic Minority Groups and Older Adults in Prostate Cancer Clinical Trials: A Meta-analysis

Importance

Prostate cancer (PCa) is marked by disparities in clinical outcomes by race, ethnicity, and age. Equitable enrollment in clinical trials is fundamental to promoting health equity.

Objective

To evaluate disparities in the inclusion of racial and ethnic minority groups and older adults across PCa clinical trials.

Data Sources

MEDLINE, Embase, and ClinicalTrials.gov were searched to identify primary trial reports from each database's inception through February 2021. Global incidence in age subgroups and US population-based incidence in racial and ethnic subgroups were acquired from the Global Burden of Disease and Surveillance, Epidemiology, and End Results 21 incidence databases respectively.

Study Selection

All phase 2/3 randomized PCa clinical trials were eligible for age disparity analyses. Trials recruiting exclusively from the US were eligible for primary racial and ethnic disparity analyses.

Data Extraction and Synthesis

This study was reported in accordance with Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) reporting guidelines. Data were pooled using a random-effects model.

Main Outcomes and Measures

Enrollment incidence ratios (EIRs), trial proportions (TPs) of participants 65 years or older or members of a racial and ethnic subgroup divided by global incidence in the corresponding age group, or US population-based incidence in the corresponding racial and ethnic subgroup, were calculated. Meta-regression was used to explore associations between trial characteristics and EIRs and trends in EIRs during the past 3 decades.

Results

Of 9552 participants among trials reporting race, 954 (10.8%) were African American/Black, 80 (1.5%) were Asian/Pacific Islander, and 8518 (78.5) were White. Of 65 US trials, 45 (69.2%) reported race and only 9 (13.8%) reported data on all 5 US racial categories. Of 286 global trials, 75 (26.2%) reported the enrollment proportion of older adults. Outcomes by race and age were reported in 2 (3.1%) and 41 (15.0%) trials, respectively. Black (EIR, 0.70; 95% CI, 0.59-0.83) and Hispanic (EIR, 0.70; 95% CI, 0.59-0.83) patients were significantly underrepresented in US trials. There was no disparity in older adult representation (TP, 21 143 [71.1%]; EIR, 1.00; 95% CI, 0.95-1.05). The representation of Black patients was lower in larger trials (meta-regression coefficient, -0.06; 95% CI, -0.10 to -0.02; P = .002).

Conclusions and Relevance

The results of this meta-analysis suggest that Black and Hispanic men are underrepresented in trials compared with their share of PCa incidence. The representation of Black patients has consistently remained low during the past 2 decades.

Utilizing Electronic Health Records (EHR) and Tumor Panel Sequencing to Demystify Prognosis of Cancer of Unknown Primary (CUP) patients

Cancer of unknown primary (CUP) is a type of cancer that cannot be traced back to its original site and accounts for 3-5% of all cancers. It does not have established targeted therapies, leading to poor outcomes. We developed OncoNPC, a machine learning classifier trained on targeted next-generation sequencing data from 34,567 tumors from three institutions. OncoNPC achieved a weighted F1 score of 0.94 for high confidence predictions on known cancer types (65% of held-out samples). When applied to 971 CUP tumors from patients treated at the Dana-Farber Cancer Institute, OncoNPC identified actionable molecular alterations in 23% of the tumors. Furthermore, OncoNPC identified CUP subtypes with significantly higher polygenic germline risk for the predicted cancer type and significantly different survival outcomes, supporting its validity. Importantly, CUP patients who received first palliative intent treatments concordant with their OncoNPC-predicted cancer sites had significantly better outcomes (H.R. 0.348, 95% C.I. 0.210 - 0.570, p-value 2.32 × 10-5). OncoNPC thus provides evidence of distinct CUP subtypes and offers the potential for clinical decision support for managing patients with CUP.

Germline variants associated with toxicity to immune checkpoint blockade

Immune checkpoint inhibitors (ICIs) have yielded remarkable responses but often lead to immune-related adverse events (irAEs). Although germline causes for irAEs have been hypothesized, no individual variant associated with developing irAEs has been identified. We carried out a genome-wide association study of 1,751 patients on ICIs across 12 cancer types. We investigated two irAE phenotypes: (1) high-grade (3-5) and (2) all-grade events. We identified 3 genome-wide significant associations (P < 5 × 10-8) in the discovery cohort associated with all-grade irAEs: rs16906115 near IL7 (combined P = 3.6 × 10-11; hazard ratio (HR) = 2.1); rs75824728 near IL22RA1 (combined P = 3.5 × 10-8; HR = 1.8); and rs113861051 on 4p15 (combined P = 1.2 × 10-8, HR = 2.0); rs16906115 was replicated in 3 independent studies. The association near IL7 colocalized with the gain of a new cryptic exon for IL7, a critical regulator of lymphocyte homeostasis. Patients carrying the IL7 germline variant exhibited significantly increased lymphocyte stability after ICI initiation, which was itself predictive of downstream irAEs and improved survival.

Genomic analysis of early-stage lung cancer reveals a role for TP53 mutations in distant metastasis

Patients with non-small cell lung cancer (NSCLC) who have distant metastases have a poor prognosis. To determine which genomic factors of the primary tumor are associated with metastasis, we analyzed data from 759 patients originally diagnosed with stage I-III NSCLC as part of the AACR Project GENIE Biopharma Collaborative consortium. We found that TP53 mutations were significantly associated with the development of new distant metastases. TP53 mutations were also more prevalent in patients with a history of smoking, suggesting that these patients may be at increased risk for distant metastasis. Our results suggest that additional investigation of the optimal management of patients with early-stage NSCLC harboring TP53 mutations at diagnosis is warranted in light of their higher likelihood of developing new distant metastases.

Disparities in Representation of Women, Older Adults, and Racial/Ethnic Minorities in Immune Checkpoint Inhibitor Trials

PURPOSE

We aim to describe reporting and representation of minority patient populations in immune checkpoint inhibitor (ICI) clinical trials and assess predictors of enrollment disparity.

METHODS

Trial-level data were acquired from eligible phase II and III trials. Population-based estimates were acquired from the SEER 18 and Global Burden of Disease incidence databases. Trials reporting race, age, and sex were summarized using descriptive statistics. Enrollment-incidence ratio (EIR) was used to assess representation of subgroups. Average annual percentage change (AAPC) in EIR was calculated using Joinpoint Regression Analysis. Trial-level characteristics associated with EIR were assessed using multivariable linear regression.

RESULTS

A total of 107 trials with 48,095 patients were identified. Participation of Black, White, Asian, Native American, Pacific Islander, and Hispanic participants was reported in 65 (61%), 77 (72%), 68 (64%), 40 (37%,) and 24 trials (22%), respectively. Subgroup analyses of clinical outcomes by race, age, and sex were reported in 17 (22%), 62 (78%), and 57 (57%) trials, respectively. Women (trial proportion [TP]: 32%; EIR: 0.90 [95% confidence interval [CI]: 0.84-0.96]), patients aged ≥65 years (TP: 42%; EIR: 0.78 [95% CI: 0.72-0.84]), Black participants (TP: 1.9%; EIR: 0.17 [95% CI: 0.13-0.22]) and Hispanics (TP: 5.9%; EIR: 0.67 [95% CI: 0.53-0.82]) were underrepresented. Representation of Black patients decreased significantly from 2009 to 2020 (AAPC: -23.13). Black participants were significantly underrepresented in phase III trials (P < .001).

CONCLUSION

The reporting of participation by racial or ethnic subgroup categories is inadequate. Women, older adults, as well as Black and Hispanic participants are significantly underrepresented in ICI clinical trials.

Variation of use of targeted therapies and molecular diagnostic testing by practice type for non-small cell lung cancer and colorectal cancer

6551

Background: Targeted therapies are important first-line treatments for many patients with non-small cell lung cancer (NSCLC) and colorectal cancer (CRC). All patients with newly-diagnosed metastatic NSCLC and CRC should undergo molecular diagnostic testing to guide treatment selection. Methods: We used 100% Medicare fee-for-service data from 2015 through 2019 to identify beneficiaries with incident metastatic NSCLC or CRC receiving systemic therapy and to assign beneficiaries to oncology practices. We then assessed for use of molecular diagnostic testing and targeted therapies in these cohorts. We used linear mixed effects models to assess patient and practice characteristics associated with molecular diagnostic testing and targeted therapy use. Results: Rates of molecular diagnostic testing increased between 2015 and 2019 for NSCLC and CRC. In 2019, rates of molecular diagnostic testing were 85% and 65% for NSCLC and CRC, respectively. Rates of targeted therapy use did not increase over time for NSCLC or CRC, and were 8% and 5%, respectively, in 2019. Compared to National Cancer Institute (NCI)-designated cancer centers, rates of molecular diagnostic testing for CRC were 3.7 percentage points lower at practices associated with non-academic hospitals and 10.6 percentage points lower at small independent practices. Rates of targeted therapy use for NSCLC were 4.8, 5.9 and 5.5 percentage points lower at academic medical centers, large independent practices and small independent practices, respectively, compared to NCI centers. Conclusions: Rates of molecular diagnostic testing for NSCLC and CRC increased in recent years, but testing rates remain below recommended levels, and targeted therapy use remains low. Substantial variation in testing and targeted therapy use by practice type suggest that the practice where a patient is treated may impact access to recommended testing and efficacious treatments. [Table: see text]

Deep Learning for Cancer Symptoms Monitoring on the Basis of Electronic Health Record Unstructured Clinical Notes

PURPOSE

Symptoms are vital outcomes for cancer clinical trials, observational research, and population-level surveillance. Patient-reported outcomes (PROs) are valuable for monitoring symptoms, yet there are many challenges to collecting PROs at scale. We sought to develop, test, and externally validate a deep learning model to extract symptoms from unstructured clinical notes in the electronic health record.

METHODS

We randomly selected 1,225 outpatient progress notes from among patients treated at the Dana-Farber Cancer Institute between January 2016 and December 2019 and used 1,125 notes as our training/validation data set and 100 notes as our test data set. We evaluated the performance of 10 deep learning models for detecting 80 symptoms included in the National Cancer Institute's Patient-Reported Outcomes version of the Common Terminology Criteria for Adverse Events (PRO-CTCAE) framework. Model performance as compared with manual chart abstraction was assessed using standard metrics, and the highest performer was externally validated on a sample of 100 physician notes from a different clinical context.

RESULTS

In our training and test data sets, 75 of the 80 candidate symptoms were identified. The ELECTRA-small model had the highest performance for symptom identification at the token level (ie, at the individual symptom level), with an F1 of 0.87 and a processing time of 3.95 seconds per note. For the 10 most common symptoms in the test data set, the F1 score ranged from 0.98 for anxious to 0.86 for fatigue. For external validation of the same symptoms, the note-level performance ranged from F1 = 0.97 for diarrhea and dizziness to F1 = 0.73 for swelling.

CONCLUSION

Training a deep learning model to identify a wide range of electronic health record-documented symptoms relevant to cancer care is feasible. This approach could be used at the health system scale to complement to electronic PROs.

Rates of Guideline-Concordant Surgery and Adjuvant Chemotherapy Among Patients With Early-Stage Lung Cancer in the US ALCHEMIST Study (Alliance A151216)

Importance

Standard treatment for resectable non-small cell lung cancer (NSCLC) includes anatomic resection with adequate lymph node dissection and adjuvant chemotherapy for appropriate patients. Historically, many patients with early-stage NSCLC have not received such treatment, which may affect the interpretation of the results of adjuvant therapy trials.

Objective

To ascertain patterns of guideline-concordant treatment among patients enrolled in a US-wide screening protocol for adjuvant treatment trials for resected NSCLC.

Design, Setting, and Participants

This retrospective cohort study included 2833 patients with stage IB to IIIA NSCLC (per American Joint Committee on Cancer 7th edition criteria) who enrolled in the Adjuvant Lung Cancer Enrichment Marker Identification and Sequencing Trial (ALCHEMIST) screening study (Alliance for Clinical Trials in Oncology A151216) from August 18, 2014, to April 1, 2019, and who did not enroll in a therapeutic adjuvant clinical trial; patients had tumors of at least 4 cm and/or with positive lymph nodes. Statistical analysis was conducted from June 1, 2020, through October 1, 2021.

Exposures

Care patterns were ascertained overall and by sociodemographic and clinical factors, including age, sex, race and ethnicity, educational level, marital status, geography, histologic characteristics, stage, genomic variant status, smoking history, and comorbidities.

Main Outcomes and Measures

Five outcomes are reported: whether patients (1) had anatomic surgical resection, (2) had adequate lymph node dissection (≥1 N1 nodal station plus ≥3 N2 nodal stations), (3) received any adjuvant chemotherapy, (4) received any cisplatin-based adjuvant chemotherapy, and (5) received at least 4 cycles of adjuvant chemotherapy.

Results

Of the 2833 patients (1505 women [53%]; mean [SD] age, 66.5 [9.2] years) included in this analysis, 2697 (95%) had anatomic surgical resection, 1513 (53%) had adequate lymph node dissection, 1617 (57%) received any adjuvant chemotherapy, 1237 (44%) received at least 4 cycles of adjuvant platinum-based chemotherapy, and 965 (34%) received any cisplatin-based adjuvant chemotherapy. Rates were similar across race and ethnicity.

Conclusions and Relevance

This cohort study found that among participants in a screening protocol for adjuvant clinical trials for resected early-stage NSCLC, just 53% underwent adequate lymph node dissection, and 57% received adjuvant chemotherapy, despite indications for such treatment. These results may affect the interpretation of adjuvant trials. Efforts are needed to optimize the use of proven therapies for early-stage NSCLC.

Trial Registration

ClinicalTrials.gov Identifier: NCT02194738.

Artificial intelligence-aided clinical annotation of a large multi-cancer genomic dataset

To accelerate cancer research that correlates biomarkers with clinical endpoints, methods are needed to ascertain outcomes from electronic health records at scale. Here, we train deep natural language processing (NLP) models to extract outcomes for participants with any of 7 solid tumors in a precision oncology study. Outcomes are extracted from 305,151 imaging reports for 13,130 patients and 233,517 oncologist notes for 13,511 patients, including patients with 6 additional cancer types. NLP models recapitulate outcome annotation from these documents, including the presence of cancer, progression/worsening, response/improvement, and metastases, with excellent discrimination (AUROC > 0.90). Models generalize to cancers excluded from training and yield outcomes correlated with survival. Among patients receiving checkpoint inhibitors, we confirm that high tumor mutation burden is associated with superior progression-free survival ascertained using NLP. Here, we show that deep NLP can accelerate annotation of molecular cancer datasets with clinically meaningful endpoints to facilitate discovery.

To accelerate cancer research that correlates biomarkers with clinical endpoints, methods are needed to ascertain outcomes from electronic health records at scale. Here, the authors train natural language processing to extract outcomes for participants in a precision oncology study.

Implications of Selection Bias Due to Delayed Study Entry in Clinical Genomic Studies

Importance

Real-world data sets that combine clinical and genomic data may be subject to left truncation (when potential study participants are not included because they have already passed the milestone of interest at the time of study recruitment). The lapse between diagnosis and molecular testing can present analytic challenges and threaten the validity and interpretation of survival analyses.

Observations

Effects of ignoring left truncation when estimating overall survival are illustrated using data from the American Association for Cancer Research (AACR) Project Genomics Evidence Neoplasia Information Exchange Biopharma Collaborative (GENIE BPC), and a straightforward risk-set adjustment approach is described. Ignoring left truncation results in overestimation of overall survival: unadjusted median survival estimates from diagnosis among patients with stage IV non-small cell lung cancer or stage IV colorectal cancer were overestimated by more than 1 year.

Conclusions and Relevance

Clinicogenomic data are a valuable resource for evaluation of real-world cancer outcomes and should be analyzed using appropriate methods to maximize their potential. Analysts must become adept at application of appropriate statistical methods to ensure valid, meaningful, and generalizable research findings.

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.

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.

Correlation Between Surrogate End Points and Overall Survival in a Multi-institutional Clinicogenomic Cohort of Patients With Non-Small Cell Lung or Colorectal Cancer

IMPORTANCE

Contemporary observational cancer research requires associating genomic biomarkers with reproducible end points; overall survival (OS) is a key end point, but interpretation can be challenging when multiple lines of therapy and prolonged survival are common. Progression-free survival (PFS), time to treatment discontinuation (TTD), and time to next treatment (TTNT) are alternative end points, but their utility as surrogates for OS in real-world clinicogenomic data sets has not been well characterized.

OBJECTIVE

To measure correlations between candidate surrogate end points and OS in a multi-institutional clinicogenomic data set.

DESIGN, SETTING, AND PARTICIPANTS

A retrospective cohort study was conducted of patients with non-small cell lung cancer (NSCLC) or colorectal cancer (CRC) whose tumors were genotyped at 4 academic centers from January 1, 2014, to December 31, 2017, and who initiated systemic therapy for advanced disease. Patients were followed up through August 31, 2020 (NSCLC), and October 31, 2020 (CRC). Statistical analyses were conducted on January 5, 2021.

EXPOSURES

Candidate surrogate end points included TTD; TTNT; PFS based on imaging reports only; PFS based on medical oncologist ascertainment only; PFS based on either imaging or medical oncologist ascertainment, whichever came first; and PFS defined by a requirement that both imaging and medical oncologist ascertainment have indicated progression.

MAIN OUTCOMES AND MEASURES

The primary outcome was the correlation between candidate surrogate end points and OS.

RESULTS

There were 1161 patients with NSCLC (648 women [55.8%]; mean [SD] age, 63 [11] years) and 1150 with CRC (647 men [56.3%]; mean [SD] age, 54 [12] years) identified for analysis. Progression-free survival based on both imaging and medical oncologist documentation was most correlated with OS (NSCLC: ρ = 0.76; 95% CI, 0.73-0.79; CRC: ρ = 0.73; 95% CI, 0.69-0.75). Time to treatment discontinuation was least associated with OS (NSCLC: ρ = 0.45; 95% CI, 0.40-0.50; CRC: ρ = 0.13; 95% CI, 0.06-0.19). Time to next treatment was modestly associated with OS (NSCLC: ρ = 0.60; 0.55-0.64; CRC: ρ = 0.39; 95% CI, 0.32-0.46).

CONCLUSIONS AND RELEVANCE

This cohort study suggests that PFS based on both a radiologist and a treating oncologist determining that a progression event has occurred was the surrogate end point most highly correlated with OS for analysis of observational clinicogenomic data.

Abstract 2620: Ignoring left truncation in overall survival within real-world genomic-phenomic data leads to inflated survival estimates

Studies linking genomic and phenomic data are subject to selection biases, including delayed entry or immortal time bias. Delayed entry can be problematic for time-to-event analyses, but utilization of appropriate statistical methods to account for delayed entry are underutilized. Delayed entry commonly occurs when genomic sequencing results are obtained after the start time for survival estimation.

To evaluate the impact of left truncation on overall survival (OS) estimates, we explored outcomes in patients with de novo stage IV non-small cell lung cancer (NSCLC) and colorectal cancer (CRC) from the AACR GENIE Biopharma Collaborative, who had genomic sequencing within a specified timeframe. We analyzed OS from diagnosis and from start of the most common first-line regimen, carboplatin/pemetrexed for NSCLC (N = 212 patients) and FOLFOX for CRC (N = 369 patients). We compared median OS using standard Kaplan-Meier methods to median OS using left truncation methods to account for delayed entry. All NSCLC and CRC patients underwent genomic sequencing after their diagnosis date. Among NSCLC patients on carboplatin/pemetrexed, 41% and among CRC patients on FOLFOX, 14% had sequencing determined after starting first-line regimen. The survfit function in R package survival was used, and the absolute differences and percent differences in median OS estimates were calculated.

Failure to account for delayed entry leads to an overestimation of OS, regardless of cohort and start date. Adjusting survival outcomes using left truncation methods reduces the influence of some aspects of selection bias and results in better estimates of time to event outcomes. Analyses from these cohorts can provide meaningful insights about survival outcomes outside the clinical trial setting and may support trial design and reliable selection of control arms. As such, it is imperative that analytic methods to account for the inflated survival estimates are incorporated.

EstimateCRC Stage IV (N = 658)NSCLC Stage IV (N = 722)Unadjusted Median (IQR) Overall Survival from Diagnosis (Years)3.2 (2.9, 3.4)2.3 (2.0, 2.5)Median (IQR) Overall Survival from Diagnosis in Years, Adjusting for Delayed Entry2.1 (1.9, 2.4)1.3 (1.1, 1.6)Difference in Medians (Years)1.11.0% Difference in Medians34%44%EstimateCRC Stage IV (N = 369)NSCLC Stage IV (N = 212)Unadjusted Median (IQR) Overall Survival from Most Common First-Line Regimen (Years)2.9 (2.6, 3.4)1.3 (1.0, 1.6)Median (IQR) Overall Survival from Most Common First-Line Regimen in Years, Adjusting for Delayed Entry2.1 (1.8, 2.5)0.9 (0.7, 1.2)Difference in Medians (Years)0.80.4% Difference in Medians28%31%

Citation Format: Samantha Brown, Jessica A. Lavery, Eva M. Lepisto, Caroline McCarthy, Hira Rizvi, Celeste Yu, Kenneth L. Kehl, Shawn M. Sweeney, Julia E. Rudolph, Nikolaus Schultz, Ritika Kundra, Brooke Mastrogiacomo, Phillipe Bedard, Jeremy L. Warner, Gregory J. Riely, Deborah Schrag, Katherine S. Panageas, The AACR Project GENIE Consortium. Ignoring left truncation in overall survival within real-world genomic-phenomic data leads to inflated survival estimates [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2021; 2021 Apr 10-15 and May 17-21. Philadelphia (PA): AACR; Cancer Res 2021;81(13_Suppl):Abstract nr 2620.

Abstract 2619: Defining real-world recurrence in the AACR Project GENIE Biopharma Collaborative Data

Obtaining information regarding cancer recurrence from a retrospective, EHR-based dataset poses several challenges primarily due to the lack of structured data. Patients are at risk for cancer recurrence beginning at a time point at which they are characterized as having no evidence of disease. The absence of cancer may be indicated on a radiology report or a medical oncologist assessment, requiring manual review and interpretation of potentially ambiguous free text. Further, the recurrence event itself can be defined based on several distinct data sources including pathology, imaging, clinician assessments, or tumor markers. The likelihood of ascertaining recurrence is dependent on the frequency and type of surveillance performed and varies based on tumor type and based on clinicians' thresholds for pursuing workup of borderline or suspicious findings; if follow up assessments are infrequent, there are fewer opportunities to detect recurrence. Given these challenges, there is currently no standardized approach to evaluating cancer recurrence in EHR data, impeding analyses of rare molecular tumor subtypes in multi-institutional linked clinico-genomic databases.

For this analysis, we leveraged the AACR Project GENIE Biopharma Collaborative data based on the PRISSMM curation model to develop an algorithm for identifying recurrence among patients diagnosed with stage I-III non-small cell lung cancer or with stage I-III colorectal cancer. This algorithm involves using curated pathology report data to identify a definitive surgery as the time at which patients have completed curative intent treatment. Subsequent imaging reports, pathology reports, medical oncologist assessments and tumor marker data are then evaluated in order to characterize the timing of specific recurrence events.

We will present the real-world recurrence algorithm, its underlying rationale and discuss applications of recurrence endpoints. Beyond enabling estimates of recurrence-free survival, identifying cancer recurrence will allow for estimation of progression-free survival among stage I-III patients in addition to estimation of PFS among de novo stage IV patients. Estimating PFS in a large cohort of patients with linked phenomic and genomic data has historically been a limitation of these types of datasets. Overcoming this limitation will allow for precision medicine advances in oncology by facilitating data pooling across institutions and enabling examination of rare molecular subtypes in relation to clinically meaningful endpoints.

Citation Format: Jessica A. Lavery, Samantha Brown, Eva Lepisto, Michele L. Lenoue-Newton, Caroline McCarthy, Hira Rizvi, Celeste Yu, Kenneth L. Kehl, Shawn M. Sweeney, Julia E. Rudolph, Nikolaus Schultz, Brooke Mastrogiacomo, Ritika Kundra, Jeremy Warner, Philippe Bedard, Gregory J. Riely, Katherine S. Panageas, Deborah Schrag, AACR Project GENIE Consortium. Defining real-world recurrence in the AACR Project GENIE Biopharma Collaborative Data [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2021; 2021 Apr 10-15 and May 17-21. Philadelphia (PA): AACR; Cancer Res 2021;81(13_Suppl):Abstract nr 2619.

Automated identification of immune related adverse events in oncology patients using machine learning

1551

Background: Immune related adverse events (irAEs) are a major cause of morbidity among cancer patients treated with immune checkpoint inhibitors (ICIs). irAEs are difficult to identify systematically, which represents a major barrier to the conduct and reproducibility of irAE research. Automated approaches would facilitate cohort identification and understanding of risk factors for irAEs following ICI therapy. Methods: Patients treated with one or more ICIs at a single tertiary cancer center were identified. Patients who received ICIs outside the clinical trial context were used as a development cohort. For each date containing clinical documentation, proxy outcomes expected to correlate with grade 2+ irAEs including irAE related diagnosis codes, key laboratory values, prescriptions for topical and systemic steroids, and irAE keywords were extracted. Intermediate machine learning models were trained to predict the presence of each proxy outcome using structured and unstructured patient data. We used clinical trial irAEs extracted from adverse event tables found in the electronic health record as the “gold standard” outcome for a final training and evaluation cohort. A logistic regression model was used to combine predictions from each intermediate model and generate an overall probability score for each irAE type on a given encounter date. Ten-fold cross-validation was used to evaluate the final machine learning model on a held-out sample of clinical trial patients. Encounter level models were evaluated for predicting the onset of a given irAE on a given date, and patient level models for predicting irAE onset within 6 months of ICI initiation. Results: We identified 3,765 patients treated with ICIs off-trial and 1106 patients treated on ICI clinical trials. Among trial patients, overall incidence of any grade 2+ irAE was 21%. The combined irAE models were able to predict prospective gold standard irAE labels with accurate discrimination at both the encounter and patient level (Table). Conclusions: Machine learning models can identify irAEs among cancer patients in an automated manner, which may facilitate research to mitigate toxicities and optimize clinical outcomes. Validation of these methods in an external institutional cohort is underway.[Table: see text]

Pan-cancer evaluation of homologous repair deficiency somatic mutations and response to first-line anti-neoplastic therapy

10535

Background: Homologous recombination is a major mechanism of defective DNA repair, but it remains uncertain whether homologous repair deficient (HRD) tumors have favorable prognosis or are more/less likely to respond to treatment than tumors lacking such mutations. Objective: To determine whether lung (NSCLC) and colorectal (CRC) HRD+ tumors have better survival or response to chemotherapy than HRD- tumors. Methods: Patients with de novo stage IV NSCLC or CRC who had next generation sequencing (NGS) between 2015-2018 from one of four cancer centers were identified. Records were curated using the PRISSMM framework to ascertain treatment, overall survival (OS) and progression free survival based on imaging (PFS-I) and oncologists’ notes (PFS-M). Each NSCLC or CRC tumor was categorized as HRD+ if NGS revealed an oncogenic/likely oncogenic mutation in: ATM, BAP1, BARD1, BLM, BRCA1, BRCA2, BRIP1, CHEK2, FAM175A, FANCA, FANCC, NBN, PALB2, RAD50, RAD51, RAD51C, RTEL1, or MRE11A based on the OncoKB database. The tumor was categorized as HRD- if no oncogenic mutation in any of these genes was evident and HRD indeterminate (HRD?) if no mutation was identified but the panel did not include all genes. OS, PFS-I and PFS-M from start of first line therapy were reported by HRD status. The percentage with a good response to first line therapy (≥2x the median) and exceptional response (≥3x the median) was estimated for each endpoint. Results: For NSCLC 4% were HRD+, 59% HRD- and 37% HRD?. For CRC there were 5% HRD+, 60% HRD- and 35% HRD?. There were no significant differences for any survival endpoint between patients who were HRD+ vs HRD- in univariable analyses. The proportion of good and exceptional responders to first line systemic chemotherapy also did not vary by HRD status, though patients with HRD+ CRC were potentially more likely to be exceptional responders. Similarly, no differences between HRD+ and HRD- tumors were apparent for the subgroup receiving platinum containing therapy. Conclusions: NSCLC and CRC patients with somatic mutations in HRD oncogenic genes did not differ from patients lacking such a mutation with respect to OS or PFS. CRC patients with HRD+ tumors may be more likely to be exceptional responders, but sample sizes are limited. By May, the analysis will include breast and pancreatic cancer cases.[Table: see text]

Clinical and genomic predictors of brain metastases (BM) in non-small cell lung cancer (NSCLC): An AACR Project GENIE analysis

2032

Background: 30-50% of patients with non-early NSCLC will eventually develop BM, with a median survival of less than one year from BM diagnosis. There are no widely accepted clinical risk models for development of BM in patients without them at baseline. We predicted the binary risk of BM using clinical and genetic factors from a large multi-institutional cohort. Methods: Stage II-IV NSCLC patients from the AACR Project GENIE Biopharma Consortium dataset were eligible. This consisted of 4 academic institutions who curated clinical data of patients who had somatic next-generation tumor sequencing (NGS) between 2015-2017. We excluded patients who had BM at baseline, died within 30 days of NSCLC diagnosis, or did not undergo brain imaging. Covariates included demographics, anticancer therapies (received up to 90 days prior to BM development and within 5 years from NSCLC diagnosis), and NGS data; radiotherapy (RT) data were not available. NGS features included mutations and copy number alterations. These features were restricted to those classified as oncogenic by OncoKB. Univariate feature selection with Fisher’s test (p<.1) was performed on medication and genetic features. We compared 5 different machine learning models for prediction: random forest (RF), support vector machine (SVM), lasso regression, ridge regression, and an ensemble classifier. We split our data into training and test sets. 10-fold cross-validation was done on the training set for parameter tuning. The area under the receiver-operating curve (AUC) is reported on the test set. Results: 956 patients were included, 192 (20%) in the test set. Univariate features associated with BM were treatment with etoposide, Asian race, presence of bone metastases at NSCLC diagnosis, mutations in TP53 and EGFR, amplifications of ERBB2 and EGFR, and deletions of RB1, CDKN2A and CDKN2B. Univariate features inversely associated with BM were older age, treatment with nivolumab, vinorelbine, alectinib, pembrolizumab, atezolizumab, and gemcitabine, as well as mutations in NOTCH1 and KRAS. Ridge regression had the best AUC, 0.73 (Table). Conclusions: We achieved reasonable prediction performance using commonly obtained clinical and genomic information in non-early NSCLC. The biologic role of the associated alterations deserves further scrutiny; this study replicates similar findings for EGFR and KRAS in a much smaller cohort. Certain subsets of NSCLC patients may benefit from increased surveillance for BM and transition to drug therapies known to effectively cross the blood-brain barrier, e.g., nivolumab and alectinib. Inclusion of additional covariates, e.g., brain RT, may further improve model performance.[Table: see text]

Artificial intelligence in oncology: Path to implementation

In recent years, the field of artificial intelligence (AI) in oncology has grown exponentially. AI solutions have been developed to tackle a variety of cancer‐related challenges. Medical institutions, hospital systems, and technology companies are developing AI tools aimed at supporting clinical decision making, increasing access to cancer care, and improving clinical efficiency while delivering safe, high‐value oncology care. AI in oncology has demonstrated accurate technical performance in image analysis, predictive analytics, and precision oncology delivery. Yet, adoption of AI tools is not widespread, and the impact of AI on patient outcomes remains uncertain. Major barriers for AI implementation in oncology include biased and heterogeneous data, data management and collection burdens, a lack of standardized research reporting, insufficient clinical validation, workflow and user‐design challenges, outdated regulatory and legal frameworks, and dynamic knowledge and data. Concrete actions that major stakeholders can take to overcome barriers to AI implementation in oncology include training and educating the oncology workforce in AI; standardizing data, model validation methods, and legal and safety regulations; funding and conducting future research; and developing, studying, and deploying AI tools through multidisciplinary collaboration.

Association Between First-Line Immune Checkpoint Inhibition and Survival for Medicare-Insured Patients With Advanced Non-Small Cell Lung Cancer

IMPORTANCE

Immunotherapy is now a cornerstone of treatment for advanced non-small cell lung cancer (NSCLC), but its uptake and effectiveness among older patients outside clinical trials remain poorly understood.

OBJECTIVE

To understand treatment patterns and evaluate the overall survival associated with checkpoint inhibitor immunotherapy, cytotoxic chemotherapy, and combined chemoimmunotherapy for older patients who have advanced NSCLC and Medicare coverage.

DESIGN, SETTING, AND PARTICIPANTS

This retrospective cohort study included Medicare-insured patients in the US aged 66 to 89 years who initiated first palliative-intent systemic therapy for lung cancer between January 1, 2016, and December 31, 2018. Survival follow-up continued through March 31, 2020. A total of 19 529 patients who had advanced lung cancer and were insured by a Medicare fee-for-service plan were included in the analysis.

EXPOSURES

Regimens included pembrolizumab monotherapy (n = 3079), combined platinum-based drug (ie, cisplatin or carboplatin [hereinafter, platinum]) and pemetrexed disodium (n = 5159), combined platinum and a taxane (ie, paclitaxel, nab-paclitaxel, or docetaxel) (n = 9866), and combined platinum, pemetrexed, and pembrolizumab (n = 1425), as ascertained using Medicare claims from the Centers for Medicare & Medicaid Services.

MAIN OUTCOMES AND MEASURES

The primary outcome was overall survival, which was measured using the restricted mean survival time (RMST) with propensity score adjustment for clinical and sociodemographic characteristics. Median survival was also reported for comparison with outcomes from registrational trials.

RESULTS

A total of 19 529 patients (54% male, 46% female; median age, 73.8 [interquartile range, 69.9-78.4] years) were identified for analysis. The uptake of pembrolizumab-containing regimens in the Medicare population was rapid, increasing from 0.7% of first-line treatments in the second quarter of 2016 to 42.4% in the third quarter of 2018. Patients who were older (≥70 years, 2484 [81%]), were female (1577 [51%]), and/or had higher Risk Stratification Index scores (highest quintile, 922 [30%]) were more likely to receive single-agent pembrolizumab than chemotherapy. After propensity score adjustment, pembrolizumab was associated with survival similar to platinum/pemetrexed (RMST difference, -0.2 [95% CI, -0.5 to 0.2] months) or platinum/taxane (RMST difference, -0.7 [95% CI, -1.0 to -0.4] months). Patients receiving platinum/pemetrexed/pembrolizumab chemoimmunotherapy also had adjusted survival similar to those receiving platinum/pemetrexed chemotherapy (RMST difference, 0.5 [95% CI, 0.1-0.9] months). The unadjusted median survival was 11.4 (95% CI, 10.5-12.3) months among patients receiving single-agent pembrolizumab, approximately 15 months shorter than observed among pembrolizumab-treated participants in the KEYNOTE-024 trial. The unadjusted median survival was 12.9 (95% CI, 11.8-14.0) months among patients receiving platinum/pemetrexed/pembrolizumab chemoimmunotherapy, approximately 10 months shorter than observed among platinum/pemetrexed/pembrolizumab-treated participants in the KEYNOTE-189 trial.

CONCLUSIONS AND RELEVANCE

Immunotherapy has been incorporated rapidly into treatment for patients with advanced NSCLC. However, survival estimates in the Medicare population are much shorter than those reported in registrational trials. These results provide contemporary estimates of survival for older patients with advanced NSCLC treated in routine practice, facilitating patient-centered decision-making.

Plasma cfDNA Genotyping in Hospitalized Patients With Suspected Metastatic NSCLC

PURPOSE

Next-generation sequencing (NGS) is an important component of first-line treatment selection for metastatic non–small-cell lung cancer (NSCLC) and is typically ordered by medical oncologists in the outpatient setting after the pathologic diagnosis has been established. Time to treatment initiation is an important clinical challenge, especially for patients with rapidly progressive disease.

METHODS

Plasma cell-free DNA (cfDNA) NGS was performed on 20 patients with suspected metastatic NSCLC hospitalized at an academic cancer center, before pathologic diagnosis. Clinicopathologic and treatment data were analyzed. Time from pathologic diagnosis to genotyping result was compared with standard care groups who underwent plasma or tumor NGS in routine clinical care.

RESULTS

The median time from pathologic diagnosis to the plasma cfDNA NGS result was 3 days in the study cohort, versus 18 days and 35.5 days in the standard care plasma and tumor NGS groups, respectively. 68.4% of evaluable patients had metastatic NSCLC, and 21.1% had another advanced solid tumor. Forty-five percent of plasma cfDNA results demonstrated actionable or informative genomic variants, and 20% of patients received standard or investigational first-line targeted therapy as a direct result of the plasma cfDNA NGS.

CONCLUSION

Plasma cfDNA NGS before pathologic diagnosis in hospitalized patients with suspected metastatic NSCLC results in substantially shorter time to genotyping result compared with standard outpatient workflows. This provides important initial evidence for the use of plasma-based genotyping earlier in the diagnostic journey, especially for patients with clinically aggressive disease. Additional studies and innovative approaches toward regulatory and reimbursement considerations are needed.

Recommendations for patient similarity classes: results of the AMIA 2019 workshop on defining patient similarity

Defining patient-to-patient similarity is essential for the development of precision medicine in clinical care and research. Conceptually, the identification of similar patient cohorts appears straightforward; however, universally accepted definitions remain elusive. Simultaneously, an explosion of vendors and published algorithms have emerged and all provide varied levels of functionality in identifying patient similarity categories. To provide clarity and a common framework for patient similarity, a workshop at the American Medical Informatics Association 2019 Annual Meeting was convened. This workshop included invited discussants from academics, the biotechnology industry, the FDA, and private practice oncology groups. Drawing from a broad range of backgrounds, workshop participants were able to coalesce around 4 major patient similarity classes: (1) feature, (2) outcome, (3) exposure, and (4) mixed-class. This perspective expands into these 4 subtypes more critically and offers the medical informatics community a means of communicating their work on this important topic.

Turnaround Time of Plasma Next-Generation Sequencing in Thoracic Oncology Patients: A Quality Improvement Analysis

PURPOSE

Genomic analysis of plasma cell-free DNA has become a widespread tool for advanced non-small-cell lung cancer care. Whereas accuracy has been reported on widely, its usefulness is also tied tightly to its turnaround time (TAT), which is not well studied.

METHODS

We studied the TAT of commercial plasma next-generation sequencing (NGS; Guardant360) for 533 results from 461 patients at our center between August 2016 and October 2019. The study received institutional review board approval as a quality improvement study; therefore, the results of the test and clinical setting were not analyzed.

RESULTS

TAT from blood draw to result (median of 9 days) was slightly longer than the TAT from laboratory receipt to result, a median of 7 days. Testing volume at our center increased three-fold over the time of the study. During this period, clinical TAT decreased from an initial median of 12 days to a median of 8 days in 2018, but more recently the median increased slightly to 9 days. During the most recent 12 months, 231 (95%) of 247 cases resulted within 14 days from blood draw, with delayed results usually because of billing, whereas 44 cases (18%) resulted within 7 days of blood draw. Studying 92 cases drawn in the most recent 3-month period, the median time of result receipt was 4:01 pm Eastern Time/1:01 pm Pacific Time; 39 results (43%) were returned after 5:00 pm Eastern Time.

CONCLUSION

In a large single-institution experience, we find that plasma NGS results can routinely be expected within 2 weeks, but uncommonly result within 1 week, supporting the need for new strategies to incorporate plasma NGS into the initial genotyping of advanced non-small-cell lung cancer.

Patient and proxy reports regarding the experience of treatment decision-making in cancer care

OBJECTIVE

Shared decision-making, including the elicitation of patient preferences regarding treatment decisions, is considered part of high-quality cancer care. However, patients may not be able to self-report due to illness, and therefore proxy reports may be used. We sought to determine the difference between proxy and patient reports about patient decisions and preferences among patients who received or were scheduled for chemotherapy using data from a large, population-based survey of patients with incident lung or colorectal cancer.

METHODS

Of 3573 patients who received or were scheduled for chemotherapy, 3108 self-reported and 465 had proxies reporting on their behalf about preferred and actual decision roles regarding this treatment. Preferred and actual decision roles were assessed using the Control Preferences Scale, and categorized as shared, patient-controlled, or doctor-controlled. Multivariable logistic regression models were used to assess the association between patient and proxy responses and whether preferences were met. The models adjusted for sociodemographic and clinical variables and patient/proxy-reported health status.

RESULTS

Sixty-three percent of all respondents reported actual roles in decisions that matched their preferred roles (role attainment). Proxies and patients were similarly likely to report role attainment (65% vs 63%). In adjusted analyses, proxies were more likely report role attainment (OR = 1.27, 95%CI = 1.02-1.59), but this difference was smaller if health variables were excluded from the model (OR = 1.14, 95%CI = 0.92-1.41).

CONCLUSION

Most patients' preferences for treatment participation were met. Surveys from proxies appear to yield small differences on the reports of attainment of preferred treatment decision-making roles in cancer care vs surveys from patients.

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.

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.

Concurrent osimertinib plus gefitinib for first-line treatment of EGFR-mutated non-small cell lung cancer (NSCLC)

9507

Background: First-line treatment with an EGFR tyrosine kinase inhibitor (TKI) is standard of care for patients (pts) with EGFR-mutated NSCLC. The EGFR TKI osimertinib is active against the acquired gefitinib-resistant mutation EGFR T790M, as is gefitinib against the osimertinib-resistant EGFR C797S. Preclinical evidence suggests dual EGFR inhibition with gefitinib + osimertinib may delay emergence of acquired resistance. Methods: This ongoing phase I/II study enrolled pts with stage IV EGFR-mutated (L858R or del19) NSCLC, without prior therapy for metastatic disease. Treatment in dose escalation (n = 6): concurrent osimertinib 40 mg or 80 mg + gefitinib 250 mg daily. In dose expansion (n = 21): osimertinib + gefitinib at the maximum tolerated dose (MTD). Prior to protocol amendment 6 pts received alternating monthly cycles of TKI monotherapy and were excluded from this analysis. The primary endpoints in the dose escalation and expansion phases were, respectively, identification of the MTD and feasibility, defined as receipt of combination therapy for ≥ 6 four-week cycles. Secondary endpoints included overall response rate (ORR), survival outcomes, plasma EGFR mutation clearance (cell free DNA by droplet digital PCR (ddPCR)), and mechanisms of acquired resistance. Results: From May 2017 to July 2019 27 pts were enrolled and evaluable for the primary endpoints. The MTD was osimertinib 80 mg plus gefitinib 250 mg orally daily. In feasibility analysis, 81.5% completed ≥6 cycles combination therapy (1 pt discontinued for progression, 4 for toxicity). The ORR was 85.2% (95% CI 67.5%-94.1%). Best response: 85.2% partial response, 14.8% stable disease. The most common treatment-related adverse effects (TRAEs) (% any grade, % grade 3) were rash (96.3%, 3.7%), diarrhea (85.2%, 11.1%) and dry skin (70.4%, 0%). Plasma ddPCR (n = 25 pts) detected the driver EGFR mutation at baseline in 68% of pts. In these pts, plasma EGFR cleared to undetectable at 2 weeks treatment in 82.4%. At 14.8 months median follow up the median progression free survival was not yet reached. Conclusions: Combination therapy with osimertinib and gefitinib is tolerable for first-line treatment of EGFR-mutated NSCLC and resulted in rapid plasma clearance of the EGFR mutation. The observed ORR is consistent with previously reported first-line response rates to osimertinib. Analysis of survival outcomes and acquired resistance mechanisms are pending data maturity and will facilitate understanding of the role of first-line dual EGFR TKI therapy for this pt population. Clinical trial information: NCT03122717 .

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]

Progression-free survival estimates in non-small cell lung cancer when RECIST is unavailable: Project GENIE’s integration of genomic, therapeutic and phenomic data

9622

Background: Molecular tumor profiling has become an integral component of oncology practice but linked genomic-phenomic data remain scarce. Recurrence, treatment response and progression are not structured consistently in medical records and this deficit has been a roadblock to discovery of biomarkers that are associated with favorable outcomes. Methods: The Genomics Evidence Neoplasia Information Exchange (GENIE) consortium is an AACR sponsored project to link and share genomic and phenomic data to promote discovery in precision medicine. 3 cancer centers that routinely perform somatic tumor profiling for advanced cancers agreed to curate anti-neoplastic treatment exposures and outcomes including recurrence, progression, response and survival using a standard method. 6 cancer types (lung, colorectal, breast, prostate, pancreas and bladder) were selected and a REDCAP database captures anti-neoplastic treatments, and specific elements from pathology, radiology and oncology reports. Curators abstract data using data fields that rely on the PRISSMM standard. “Real world” progression free survival (PFS) was identified based on curation of: 1) the text of radiologists’ reports for CT, PET/CT, PET and MRI scans (PFSI) and 2) medical oncologists’ notes (PFSM). PFSI and PFSM were estimated from the start of 1st line anti-neoplastic systemic therapy until progression or death for all patients with molecularly characterized non-small cell lung cancer (NSCLC). Results: Genomic sequencing was performed between 2015 and 2017 for 748 patients with NSCLC treated at three major cancer centers. Median age at diagnosis was 66 years (interquartile range 58, 73) and 43% were male. As shown in the table, when RECIST assessments are unavailable, estimates of PFS vary based on whether they are derived from radiologists’ or oncologists’ interpretations. Conclusions: Radiologists’ reports and oncologists’ reports provide different PFS estimates. Cohort studies should specify the method used to define “real world” endpoints. Project GENIE will have 1800 NSCLC patients with curated endpoints by the ASCO meeting. [Table: see text]

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.

Safety of Immune Checkpoint Inhibitors in Patients With Pre-Existing Inflammatory Bowel Disease and Microscopic Colitis

PURPOSE

Enterocolitis is among the leading adverse events associated with immune checkpoint inhibitors (ICIs). There are limited retrospective data regarding the safety of ICIs in patients with inflammatory bowel disease (IBD; ulcerative colitis, Crohn's disease) because they have been generally excluded from clinical trials testing ICIs. Furthermore, there are no outcome data available in patients with microscopic colitis, a leading cause of chronic diarrhea. We aimed to study the safety of ICIs in patients with cancer with pre-existing IBD or microscopic colitis.

METHODS

We retrospectively reviewed the records of patients with cancer treated at our institution who received at least 1 dose of either a programmed cell death-1 (PD-1)/ PD-1 ligand inhibitor, cytotoxic T-lymphocyte-associated antigen 4 inhibitor, or both between 2011 and 2018. We identified patients with pre-existing IBD or microscopic colitis.

RESULTS

Of 548 patients with solid tumor treated with an ICI, we identified 25 with pre-existing colitis (21 IBD; 4 microscopic colitis). An enterocolitis flare occurred in 7 patients (28%): 3 of 4 patients (75%) with microscopic colitis and 4 of 21 (19%) with IBD. All were treated with systemic corticosteroids, 2 required an anti-tumor necrosis factor agent, and one required an anti-integrin agent and colectomy for treatment of refractory colitis. ICI therapy was discontinued in all patients who experienced an enterocolitis flare.

CONCLUSION

In our cohort, exacerbation of enterocolitis occurred in a notable percentage of patients with IBD and a majority of patients with microscopic colitis, leading to discontinuation of ICIs. Although these data suggest that patients with cancer with pre-existing IBD/microscopic colitis may be treated with ICIs, additional studies are needed to validate our results.

Race, Poverty, and Initial Implementation of Precision Medicine for Lung Cancer

Data are limited regarding whether the availability of biomarker-directed therapy for lung cancer exacerbates racial and socioeconomic disparities. Patients diagnosed with stage IV lung adenocarcinoma from 2008 to 2013 were identified using Surveillance, Epidemiology, and End Results Program-Medicare. The primary outcome was a Medicare claim for molecular testing within 60 days of diagnosis, analyzed using multivariable logistic regression; the secondary outcome was overall survival, analyzed using multivariable Cox proportional hazards models. All statistical tests were two-sided. Of 5556 patients, 1437 (25.9%) had molecular testing. Testing rates were 14.1% among black, 26.2% among white, and 32.8% among patients of Asian/other descent (adjusted P < .001); 20.6% among patients with Medicaid eligibility vs 28.4% among those without (adjusted P = .01); and 19.9% among patients in the highest census tract-level poverty rate quintile vs 30.7% among patients in the lowest quintile (for all quintiles, adjusted P = .18). Median survival from 60 days was 8.2 months among patients with molecular testing within 60 days of diagnosis and 6.1 months among those without (hazard ratio = 0.92, 95% confidence interval = 0.86 to 0.99; adjusted P = .02). Equitable precision medicine requires concerted implementation efforts.

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.