July 23, 2019

Predicting Risk of Mortality in Lung Cancer Patients for Improved Clinical Trials and Patient Access

The costs associated with every cancer patient recruited into a clinical trial are enormously high. A recent study of 59 new therapeutic agents approved by the FDA between 2015 to 2016 found that the average cost of pivotal efficacy trials was $19 million, with those costs going up with increasing patient recruitment and enrollment needs.

Improving clinical trials and finding ways to design them with greater speed and efficiency has become mission critical for life sciences companies.

At ASCO 2019, Concerto HealthAI researchers presented a research abstract poster on an Artificial Intelligence model that can accurately predict risk of mortality in patients at various points. This model can now help guide clinical development teams to more effectively select attributes of patients for different pre- and post-approval studies. In addition to helping determine clinical trial eligibility faster than traditional processes, AI models like this can also be useful for assessing patient risk and treatment options and evaluating cost and quality of care.

Our AI model focused on lung cancer and can predict survival of lung cancer patients 3-12 months from their last clinical visit. This can allow clinical researchers to gain deeper insights into key variables impacting patient survival. Researchers can now determine with a high degree of confidence which patients are expected to survive the next three months and recruit accordingly to ensure a successful clinical trial at a faster pace.

A Gradient Boosting model was built using de-identified structured data from 55,000 lung cancer patients in ASCO’s CancerLinQ Discovery® database. Nearly 4,000 unique variables, including diagnostic and therapeutic codes, biomarkers, surgeries and lab tests, were accessible to the model. Model validation was done on a randomly selected and reserved set of 8,468 patients. These results were significantly better than a baseline Cox-PH model and compared very favorably to other survival models in the literature created using AI and comparable machine learning techniques.

This kind of AI model can impact patient accessibility to new treatments and clinical drug development timelines by identifying the right patients for clinical trials and enabling researchers to design and execute more effective and efficient studies. To learn more about this AI model, access the full ASCO 2019 abstract or read our full press release.