Nov. 29, 2022 — Artificial intelligence can help improve diversity, fairness and inclusion in clinical trials and drug development by overcoming some of the traditional human biases in these fields, but we’re not there yet, experts say. The technology can also help doctors conduct data insights to make diagnosis and treatment more precise.
It starts with quality. Artificial intelligence (AI) relies on vast amounts of data to create algorithms, or computer instructions, to develop best practices and predictions. But those instructions are only as good as the data used to create them. Humans are the ones who create data.
“What underpins AI technology is people, and those people have their own biases,” said Naheed Kurji, chair of the board of directors of the Alliance for Artificial Intelligence in Healthcare. “So the algorithm will have its own bias.”
Technology that uses speech to diagnose disease is one example.
“In many cases, companies fail to recognize language differences between cultures,” Kurji said. When technology is based on the speech patterns of a limited population, “then when the model is applied in the real world to different people with different accents, population, the model fails.”
“So it’s not representative.”
Another example is genetic and genomic data.
“Give or take, more than 90 percent of genetic and genomic data comes from people of European descent. It’s not from people from the African continent, Southeast Asia, Asia, or South America,” said Kurji, co-author of Cyclica, a data-driven drug discovery company based in Toronto. Inc.’s President and Chief Executive Officer.
As a result, “a lot of the research done on that level of data is inherently biased,” he said.
for equality
Creating data that takes into account the diversity, equity, and inclusion of people and cultures around the world is not a hopeless challenge. But it will take time, experts say. Once this is achieved, AI should move closer to being free of human and systemic bias.
Raising awareness is crucial.
“The solution to the problem comes from an inherent understanding that bias exists,” Kurji said, and then only includes fair and balanced data that passes the diversity test.
Choose wiser?
Another promising avenue for AI is to simplify the drug development process, narrowing down the pool of potential drug candidates and making clinical trials more cost-effective.
“If there are challenges and limitations in the source data, AI will continue to propagate those limitations,” agrees Sastry Chilukuri, co-CEO of data-driven clinical trials company Medidata and founder and president of Acorn AI. “Source data has to become more representative, and it has to be fairer, so that the AI reflects what’s going on.”
When it comes to human or systemic bias in drug development, “it would be an oversimplification to say that artificial intelligence or machine learning can solve this problem,” says Dr. Angeli Moeller, head of data and integrated insights at Roche in Berlin. “But the responsible use of artificial intelligence and machine learning can help us identify bias and find ways to mitigate any negative effects it may have.”
silent partner?
Experts say that while artificial intelligence is designed to simplify drug development, the technology could also help all doctors do their jobs better. For example, AI will help by widely disseminating knowledge and expertise, sharing best practices of doctors experienced in treating more complex patients. This will help guide those who treat only a small number of such patients each year.
In New York City or Delhi, the volume could run into the hundreds of patients a year, Chilukuri said. “But if you go to the interior of the United States like Nebraska, surgeons don’t see as much volume.”
AI could “help doctors by providing a tool that allows them to deliver the same first-class care to all populations more quickly,” he said.
Improve efficiency
AI can help target treatments by using data to identify patients at highest risk. The technology could also improve bottleneck areas in medicine, such as the time it takes to interpret radiology images, Kurji said.
There is one AI company whose “whole business model is not to replace your radiologist, but to make radiologists better,” he noted. One of the company’s goals is to “prevent death or serious illness from not acting quickly enough for that patient because a radiology scan is missed or piles up.”
Radiologists are so busy they may have 30 seconds or less to interpret each scan, Chilukuri said. AI can flag potential lesions, but it can also compare images to past scans of the same patient. This view provided by artificial intelligence is applicable not only to radiology but also to the data-driven field of medicine.
Advancing Personalized Medicine
AI can also guide an individual’s surgical approach, “because it’s not like humans have small, medium, and large,” Chilukuri said. The technology could help surgeons determine exactly where to operate on individual patients.
Moeller agrees that AI has the potential to advance personalized medicine.
“AI can help with diagnosis and risk prediction, which could mean earlier intervention,” said Moeller, who is also vice chair of the Healthcare Council’s AI Alliance. “For example, if you look at a diabetic patient, what is the likelihood that he or she may have eye problems due to diabetic macular edema?”
The technology can also help to see the big picture.
“Machine learning can look for patterns in a population that might not be in your medical textbook,” Moeller said.
Chilukuri predicts that, in addition to diagnosis and treatment, AI could also aid in recovery by customizing recovery for each patient.
“Not everyone is going to recover in the exact same way. So you have a highly individualized AI plan that can really keep you on track and predict where you’re going.”
