Follow Us

Select Language
Mobile

AI Eşleştirme Sistemi

INTRODUCTION AND PURPOSE

  • Clinical research teams manually screen patients for cancer trials, a process that can take ~25 minutes per single trial using highly specialized resources, restricting trial options available to patient, and negatively impacting enrollment rates.

  • Precision Cancer Consortium, including seven of the top 20 pharmaceutical companies, collaborated with Massive Bio to explore an innovative and efficient multi-trial matching method.

  • This initiative focused on shifting towards a multi-trial prescreening approach, incorporating NGS testing results, and using artificial intelligence to enhance accuracy and efficiency of trial matching, particularly for targeted therapies.

METHODS

  • AI-Enabled System: Massive Bio developed a state-of-the-art AI system that utilizes computer vision and natural language processing to extract 180 structured clinical parameters from medical records. This system also features an AI recommendation algorithm for matching patients to digitized inclusion/exclusion (I/E) criteria from over 14,000 actively recruiting interventional cancer trials.

  • Data Analysis and Algorithm Development: The study analyzed a sample of 5,707 cancer patients from the Massive Bio Synergy AI real-world database, focusing on tumor types relevant to 23 selected trials. A decision-tree algorithm was developed for retrospective matching. Results were compared to the theoretical matching rate based on specific criteria including tumor type, biomarker prevalence, disease extent at diagnosis, and prior treatment history.

RESULTS

  • Records of 5,707 unique patient were selected, of which only 1,557 (27%) had NGS testing results available.

  • 690 unique patients were matched to at least one trial. The total number of matching pairs (patient-trial) was 1,254 resulting in a 1.82-fold increase as a result of multi-trial matching.

  • In a simulated scenario, where 100% of patients were assumed to have NGS testing results, a total number of 2,635 matching patient-trial pairs were found. This implied a potential 2.1-fold increase from NGS testing.

  • In a multi-trial case study, specific for lung cancer trials targeting wild-type or mutated EGFR gene, multi-trial matching to 11 trials resulted in 12.4-fold increase in unique patients compared to a single trial.

  • Compared to the theoretical matching rate, Massive Bio’s matching rate was significantly higher 50% of trials, and equivalent in 30%.

  • The total savings in effort was estimated to be 19,500 hours (99.9%).

Multi-trial matching and impact of genomic testing in multi-trial matching

Sponsor Trial Tumor Types Patients Analyzed Had NGS Results Exact Matches Partial Matches No-Matches: Fails NGS Partial Matches: Missing only NGS Partial Matches: Other than NGS Potential Exact Matches Assuming 100% NGS
AZ NCT03997123 Breast 781 190 16 3 0 (0%) 3 (100%) 16 (2.05%)
AZ NCT04305496 Breast 781 190 54 0 (0%) 8 (100%) 8 0 54 (6.91%
Bayer NCT03188965 Lymphoma, Prostate, Breast, Colon, Rectal, Ovarian, Endometrial, Cervical 2637 678 86 406 (72%) 9 (2% 564 305 182 (6.92%)
Bayer NCT02576431 Solid Tumors 4939 1487 0 852 541 759 (89%) 0 (0%) 0 (0%)
Bayer NCT04147819 Solid Tumors 4939 1487 29 9 0 0 (0%) 9 (100%) 29 (0.59%)
Bayer NCT05099172 Lung 841 260 47 194 175 139 (72%) 5 (3%) 79 (9.37%)
GSK NCT02064387 Multiple Myeloma 223 9 91 23 0 0 (0%) 23 (100%) 91 (40.81%)
GSK NCT02655016 Ovarian 163 77 9 27 18 27 (100%) 0 (0%) 18 (11.04%)
GSK NCT02715284 Solid Tumors 4939 1487 587 1647 340 1291 (78%) 80 (5%) 1442 (29.20%)
GSK NCT05723562 Rectal 81 45 0 11 3 5 (45%) 1 (9%) 1 (1.53%)
JNJ NCT05601973 Lung 841 293 9 3 10 1 (33%) 2 (67%) 9 (1.13%)
JNJ NCT04538664 Lung 841 293 6 213 206 125 (59%) 2 (1%) 11 (1.27%)
JNJ NCT03390504 Urothelial 49 9 2 12 6 11 (92%) 0 (0%) 5 (9.69%)
JNJ NCT04487080 Lung 841 293 11 81 107 69 (85%) 0 (0%) 17 (2.07%)
JNJ NCT04988295 Lung 841 293 24 11 12 1 (9%) 8 (73%) 25 (2.94%)
Lilly NCT05307705 Solid Tumors 4939 1487 73 1733 700 1253 (72%) 24 (1%) 226 (4.57%)
Lilly NCT05614739 Solid Tumors 4939 1487 29 1671 882 1280 (77%) 15 (1%) 90 (1.82%)
Lilly NCT04956640 Solid Tumors 4939 1487 65 1557 1009 1180 (76%) 13 (1%) 150 (3.03%)
Lilly NCT04194944 Lung 841 293 0 181 220 153 (85%) 0 (0%) 0 (0%)
Novartis NCT05132075 Lung 841 293 17 100 107 90 (90%) 1 (1%) 30 (3.56%)
Novartis NCT05445843 Lung 841 293 79 117 83 71 (61%) 17 (15%) 117 (13.92%)
Novartis NCT05791097 Lung 841 293 6 19 8 9 (47%) 3 (16%) 11 (1.28%)
Roche NCT02091141 Solid Tumors 4939 1487 14 1395 822 950 (68%) 2 (0%) 32 (0.65%)
Unique: 5,707 1,557 690 2,216(100%) 1,123 1.729(78%) 172(8%) estimated 1,290
Total: 46,857 14,211 1,254 10,431(100%) 5,554 7,820(75%) 225(2%) 2,635
2.1-fold increase in patient-trial matches assuming 100% NGS testing

Multi Trial Matching in biomarker driven Lung Cancer trials

CONCLUSION

The combination of Next-Generation Sequencing (NGS) with Artificial Intelligence (AI) in a multi-trial matching approach resulted in a nearly 2-fold increase potential patient eligibility for trials for all tumor types and 12-fold increase in a single focused tumor profile, along with a dramatic reduction in manual effort.

A prospective nationwide analysis is being considered to further explore these findings.

QR Code

#ASCO24 Abstract e13501