1974-1976
December 1, 2022JUNBO GE
December 1, 2022ConcertAI
ConcertAI is a global pioneer for life science businesses and healthcare professionals in precision oncology, combining AI technology, real-world data...
ConcertAI
ConcertAI is a global pioneer for life science businesses and healthcare professionals in precision oncology, AI technology, real-world data, and evidence solutions. Via alliances, unique real-world data properties, leading AI-based innovations, and the world’s best research and data science talent, their goal is to drive progress in clinical results for cancer patients.
AI Aspects
ConcertAI’s rich suite of independent data products is ideal for satisfying a range of analysis and real-world proof use cases around the enterprise. In collaboration with leading biomedical innovators, healthcare professionals, and medical societies, ConcertAI accelerates insights and results for patients through leading real-world evidence, AI technologies, and scientific expertise.
On-Demand Data for Accelerated Research
Post-approval:
- Comparative effectiveness in real-world populations compared clinical trials.
- Study feasibility analyses.
- Patient population comparisons and analytics.
- Characteristics of the patient journey and treatment sequencing over time.
- Adverse event detection and evaluation.
- Treatment patterns and analysis.
- Biomarker data impact on treatment selection and outcomes.
AI-Enriched Real World Data Product
Clinical Development:
- External control arms
- Clinical trial protocol development and optimization
- Identification of undiagnosed and under-treated patient groups
- Study feasibility analyses
- Patient population comparisons and analytics
- Characteristics of the patient journey and treatment sequencing over time
- Adverse event detection and evaluation
- Treatment patterns and analysis
- Biomarker data impact on treatment selection and outcomes
RWD360
It is a large repository of pan-tumor clinical data with the greatest epidemiological significance for broad research. The program can:
- Compare characteristics: compare demographic characteristics based on different treatments.
- See treatment sequencing: how treatments and drug class exposure are sequenced in a cohort.
- Combine clinical criteria.
Patient360
It is an off-the-shelf, highly abstracted data product that eliminates the need for personalized healing and offers clear insights into evolving disease and treatment trends in the current standard of care. It can:
- Identify study feasibility: Identify base rates of clinical characteristics, disease severity measures, and treatment outcomes treatment to inform the feasibility
- Post-approval RWE: Conduct effectiveness outcome studies in specific populations of interest and compare to clinical trials
- Track safety: See adverse events experienced by real-world populations and compare them to clinical
- Cost of care analysis: Link to closed claims for accurate calculations of healthcare utilization and cost-of-care.
Which Cancers Can be Researched?
Genome360
Provides an in-depth understanding of treatment profiles with data from NGS panels and structured and unstructured fields.
Biomarkers testing results.
Identify patients by gene variant.
Assess treatment and outcomes
Custom Data Services for Specific Research Needs
ConcertAI team of HEOR and clinical oncology experts build unique data assets to support research questions that demand deeper insights.(1)
Concert AI’s Financials Aspects
ConcertAI raised $150 million in aggregate Series B financing. With 90% year-over-year revenue growth, ConcertAI has an expanding list of partners and customers: three of the world’s major biopharmaceutical and oncology organizations have deployed Eureka Health, and 19 out of the top 25 biopharma are customers of ConcertAI technologies and services. The Series B funding will support ConcertAI’s continued innovation in real-world data and technology products and services for regulatory and non-regulatory applications and commercial patient solutions.(2)
Genome360
The Extremely Random Forest model can predict the risk of metastatic recurrence in breast cancer patients and could be beneficial for diagnosing and treating breast cancer patients. Out of 3807 women with breast cancer, 628 had metastatic recurrence within four years post one year of the diagnosis date.
The Extremely Random Forest model for predicting the risk of metastatic recurrence within one year from 1-year post-diagnosis showed satisfactory accuracy and a competent area under the curve. (3)
Application of AI for the Diagnosis of NSCLC
Lung cancer is the leading cause of cancer-related mortality in the United States, with a 5-year relative survival rate of 17.5% from 1995 to 2001 for patients with lung cancer. NSCLC (non-small cell lung cancer) is the most common type of lung cancer, accounting for 84% of all lung cancer diagnoses. A study was performed to explore an artificial intelligence model that can distinguish NSCLC and used retrospective electronic health record (EHR) data from a cohort of lung cancer patients. 56,748 LC patients were selected, of which 85% were labeled to be NSCLC. Test results revealed an AUC-ROC of 0.93 and overall accuracy of 93%. These findings show that AI can be used to diagnose NSCLC and thus, promises to be beneficial for a timely diagnosis and further treatment decisions. (4)
AI Model to Predict Slow Progression for Advanced Non-small Cell Lung Cancer Patients Receiving Second-line Therapies
In this study, a machine learning model was trained using the Concert Health AI system of oncology EMR data. The model predicts patients with progression-free survival (PFS) greater than 180 days from the beginning of second-line therapy. The patients included in the study were those pathologically confirmed with advanced non-small cell lung cancer (aNSCLC). Patients were classified as slow progressors if they had no evidence of progression or death within 180 days and were evaluated for progression for at least 180 days post-index. Results showed that from the 2205 patients included, 1776 were used for model training, and 429 were used for model validation. Of these, 420 were labeled as slow progressors. The features associated with slow progression included the absence of metastatic disease, absence of COPD, previous treatment with an EGFR, and normal BMI. (5)
Development of an Algorithm Using Natural Language Processing to Identify Metastatic Breast Cancer Patients from Clinical Notes
From the Concert Health AI dataset, the information of 20138 patients was used to build and validate a set of algorithms. These algorithms included: 1) Classification of a sentence into three classes: Distal/Local metastasis, Suspicious, and Other 2) Classification of a sentence into two classes: Distal or Local 3) Classification of a patient into two classes: Distal metastasis or Not distal metastasis 4) Multi-label classification for detecting sites of metastasis. At a sentence level, the accuracy for the distal/local vs. suspicious vs. irrelevant model was 0.85 and 0.97 for the distal vs. not distal metastasis model. This study showed that the metastatic status and site of metastasis could be extracted automatically from clinical information using deep learning systems. (6)