Artificial intelligence and other forms of advanced data analysis are often viewed as completely unbiased and unbiased. However, when the people who put the processes in place inadvertently introduce biases, the results can be affected.
Rich Christie, Chief Medical Officer of AiCure, recently connected with Outsourcing-Pharma to talk about the problems that AI biases can cause in testing and development.
OSP: Could you please share your perspective on biases around clinical trial data? Specifically, how can AI and other advanced analytical tools (widely seen as objective) be geared towards one group and / or against another?
RC: Bias can arise in AI when the data used in initial training sets is not fully representative of the diversity present in subsequent patient populations. After all, the strength and generalization of AI models reflects the underlying training data sets, which is why consistent assessment of the quality and quantity of the data sets is essential.
For example, the datasets that many computer vision tools have trained on in the past lacked variability and diversity as well as a number of dimensions, usually because many of them were based on programmers. IT volunteers, a group historically limited in diversity. This means that when visual tools are used with patients of diverse backgrounds, the inherent bias can not only impact research operations, but it can also affect patient outcomes and ultimately affect research operations. an impact on the ability to provide meaningful data to study drug safety and efficacy.
The consequences of biased AI can vary depending on how the tool is deployed, whether in clinical development or in the real world. In one scenario, an AI system aimed at identifying optimal investigation sites could sub-optimally skew the sites chosen to participate in recruiting and managing participants in a clinical trial. In another scenario, a study could fail if a trial uses insufficient AI to determine patient engagement and enrollment rates – an incorrect number of premature withdrawals within a previously unidentified subpopulation could result in inadvertently skewed results.
In another example, if the computer vision technologies in a clinical trial do not work uniformly on patients with diverse skin tones, there could be an unintended bias in the study results that leads to misperception of the results. benefits to patients in the general population.
In 2022, we will need to work together to ensure that the technology used by our patients and pharmaceutical companies is built on the foundation it needs to foster equality and reach its potential. By focusing on the diversity of training data sets, formal generalization of results, and real-world confirmation, pharmaceutical companies can develop more powerful algorithms to optimize drug development and patient outcomes.
OSP: How does AI help advance precision medicine?
CR: AI-powered predictive analytics can help drive precision medicine in several ways, including guiding timely, personalized interventions that help patients stay on track and providing sponsors with valuable information about patients. Patient and site behaviors both during trials and even before trials begin to stratify risk.
The more we can directly capture the patient experience and nuance of behavior at the level of individual participants, the better we can match therapy to specific patient needs. AI continues to play an important role in revealing this information. Specifically, digital video and audio biomarkers can capture a patient’s response to treatment particularly well due to their consistency, automation, and ability to span across geographies and patient populations.
Quality of life assessments such as a patient’s ability to button their shirt or sign their name can be conducted to truly understand how patients are responding to treatment and its impact on their quality of life. Instead of relying solely on in-person clinical assessments or self-reported patient data, digital video and audio biomarkers can measure subtle changes in a patient’s behavior that might otherwise go unnoticed.
Additionally, AI can help predict in advance a patient’s ability to adhere to trial protocols based on past behavior, which can help sites focus on patients who may have struggling to stay on track and adapt useful advice to them. Sponsors can also use this data to optimize patient pools by understanding who may be best placed to provide quality data to answer the questions posed by the study.
OSP: In your opinion, what could be the biggest trends to watch in 2022? Don’t hesitate to talk about potential obstacles, as well as areas of opportunity.
RC: Looking to 2022, one of the areas that is expected to grow is the use of open source data platforms to advance AI. Digital video and audio biomarkers are an untapped resource for accurately measuring patient behavior and improving the objectivity of a clinical trial, but currently the proprietary nature of digital biomarker algorithms prevents researchers from exploiting them to their full potential. , which hinders their clinical validation and refinement.
As an industry, we must continue to encourage open source platforms so that the scientific community can access algorithms, jointly contribute to their advancement, and further validate digital biomarkers as a legitimate way to understand disease.
Second, we will continue to see renewed interest in mitigating the impact of bias on AI in healthcare. The role of AI in healthcare is only growing, and AI-powered knowledge will continue to affect every element of drug development, from clinical research to commercialization to healthcare evaluations. Population.
With the increased role of AI comes the possibility for it to perpetuate invisible biases if the industry is not proactive in its management. If AI is impartial and reliable, it has the power to help sponsors cope with real-world variability, better understand how drugs work and their impact on specific patients, and link clinical outcomes to an increasingly diverse real-world data collection environment.
OSP: What does AiCure have in store for the coming year that you might want to preview?
RC: AiCure’s AI platform is evolving to provide the predictive and unbiased information needed to optimize drug development throughout the drug lifecycle. Currently, much of the data used for decision making in clinical trials are lagging indicators of participant behavior, and responses are generally reactive rather than proactive, resulting in wasted resources, longer trial times. long and delayed care.
To address this issue, AiCure strives to enable structured data collection through secure and compliant management of unique PHI data so that users can understand how patient behavior generates value and can access predictive information for allow intervention when it is important. The AiCure platform provides a single place to collect, aggregate, annotate, create and test new assessments and algorithms, providing clients with the ability to develop symptom-specific AI models that enable more accurate measurement of signs of disease. through audio and video data.