In the year 1959, Arthur Samuel coined the term ‘machine learning‘ and defined it as the “field of study that gives computers the ability to learn without being explicitly programmed.” MIT researchers are using machine learning to predict clinical trials outcomes using statistical techniques. This can enable them to control the drug and device approval process better. A study in the Harvard Data Science Review aims to furnish data that can help all stakeholders manage their resources efficiently, leading to a low failure rate, fast drug approval times, low capital and more funding opportunities for the development of newer therapies.
Machine learning to predict clinical trials
Researchers at the MIT Laboratory for Financial Engineering, have developed the study that leverages the largest set of data to predict the success or failure of clinical trials and uses more than 140 features, including ‘trial outcome, trial status, trial accrual rates, duration, prior approval for another indication, and sponsor track record — to forecast clinical trial outcomes.’ To account for missing data, the researchers used machine-learning techniques with statistical methods that lead to more accurate forecasting. According to one of the researchers, ‘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.’
AI and Machine Learning: Accuracy in Predictions
The market for clinical research is poised to grow at a CAGR of about 12 percent, and has led to regulatory changes have brought clinical research in India back on track. There was a significant rise in the number of clinical approvals from 84 in 2016 to 178 in 2017. The average time for clinical trial application approval used to take 6-7 months in 2016 and is now decreased to 4 months. The issue with successful clinical trials however, is that they occur at a ratio of 1 in 10, cost about $2 to $3 billion and take 10-12 years to be approved. The costs for the trials are high, they have a failure rate of 90 percent and have a lengthy timeline. This is where AI and machine learning can help, and increase the success rate of clinical trials.
According to a report called ‘Data Age 2025’, there will be a 61 percent compound annual rate of growth in data, that may lead to a DataSphere of 175ZB by the year 2025. It is also predicted that medical data will grow 300 percent between 2017 and 2020. The clinical trial stage is the most expensive and crucial stage of the process and needs high levels of accuracy. Out of 10,000 chemical compounds tested in preclinical, 250 will show promise for animal testing and 10 will qualify for the clinical stage. This points to a real need for AI techniques that can be applied on clinical data, lower the failure rate of the trials and shorten approval timelines for drugs. AI can identify genetic markers in populations which can be used to develop drugs for individuals. AI is effective is because it can note patterns in data and go through vast amounts of it. Machine learning can help determine if a patient is suffering from a particular disease by honing in on subtle, data-driven differences, that a human cannot find, thereby increasing its accuracy levels. One study found that machine learning could improve the prediction of long-term outcomes in ischemic stroke patients. Studies such as the above prove that machine learning techniques are highly accurate and can be used to predict clinical trial outcomes.