Machine Learning Tools Help Predict Clinical Trial Outcomes – Xtelligent Healthcare Media
– Machine learning technologies can help predict outcomes of clinical trials, leading to faster drug approval times, lower costs, and more funding to develop new treatments, according to a study conducted by researchers at MIT and published in Harvard Data Science Review.
Randomized clinical trials are a high-risk venture for a wide range of stakeholders, from regulators and biopharma leaders to patients and their families.
“Everyone is affected by the risk of a drug failing in its clinical trial process,” said Andrew Lo, the study’s senior author and director of the MIT Laboratory for Financial Engineering. “With more accurate measures of the risk of drug and device development, we hope to encourage greater investment at this unique inflection point in biomedicine.”
The rise of medical breakthroughs like immune-therapies, gene therapies, and gene-editing techniques have also contributed to the growing complexity of biomedical innovation.
“These breakthroughs generate novel therapies for investigation, each of which requires many years of translational research and clinical testing, costing hundreds of millions to billions of dollars and yet often face a high likelihood of failure,” researchers said.
READ MORE: System Obtains More Training Data for Machine Learning Models
“In fact, drug development productivity—the ratio of the number of new drugs approved to R&D spending each year—has declined steadily over the past 50 years despite scientific and technical progress.”
To increase clinical trial efficiency, researchers developed machine learning algorithms using the largest set of data to date. The algorithms analyzed over 140 features, including trial status, accrual rates, duration, and sponsor track record, to predict clinical trial outcomes.
In addition to machine learning techniques, the team used statistical methods to account for missing data. These methods made it possible to estimate missing values along with other algorithm features, resulting in more accurate predictions.
“It’s the difference between looking back at historical wins and losses to predict the outcome of a horse race versus handicapping the likely winner based on multiple factors like the horse’s pedigree, track record, temperament, the training regimen, the condition of the track, the jockey’s skill, and so on,” said Lo, who also serves as principal investigator at the MIT Computer Science and Artificial Intelligence Laboratory.
Researchers found that their method achieved predictive measures of 0.78 for forecasting transitions from phase 2 to approval, and 0.81 for predicting transitions from phase 3 to approval. Additionally, using five-year rolling windows from 2004 to 2014, the team also sees an increase in the predictive power of the machine learning algorithms.
READ MORE: FDA: Real-World Data, Machine Learning Critical for Clinical Trials
These results have implications for stakeholders across the biomedical ecosystem. More accurate predictions can reduce the uncertainty of drug development and increase the amount of capital that investors are willing to lend to clinical trials.
The team noted that these algorithms could be especially helpful in advancing new treatments from phase 2 to regulatory approval, or from phase 3 to regulatory approval.
“Investors and drug developers may use such predictions to evaluate the risks of different investigational drugs at different clinical stages, providing them with much-needed transparency,” the group said.
“Greater risk transparency is one source of improved financial efficiency because it facilitates more accurate matching of investor risk preferences with the risks of biomedical investment opportunities.”
This study further shows the potential for machine learning and other advanced technologies to improve the clinical trial process. The FDA has consistently emphasized the importance of modernizing clinical trials, with leaders stressing the need to take advantage of big data, artificial intelligence, and other innovative tools.
READ MORE: Using Big Data, Machine Learning to Reduce Chronic Disease Spending
“Digital technologies are one of the most promising tools we have for making health care more efficient and more patient-focused. This isn’t an indictment of the randomized controlled trial. Far from it,” Scott Gottlieb, former FDA Commissioner, said in a speech to the Bipartisan Policy Center in January 2019.
“It’s a recognition that new approaches and technologies can help expand the sources of evidence that we can use to make more reliable treatment decisions. And it’s a recognition that this evidence base can continue to build and improve throughout the therapeutic life of an FDA approved drug or medical device.”
MIT researchers expect that their results will help advance the use of innovative technologies in the clinical trial process.
“These results are promising and raise the possibility of even more powerful drug development prediction models with access to better quality data,” researchers said.
“Ultimately, such predictive analytics can be used to make more informed data-driven decisions in the risk assessment and portfolio management of investigational drugs at all clinical stages.”