Six Cognivia team members contribute critical elements to a recent paper published in Innovations in Clinical Neuroscience.
Cognivia’s Erica A. Smith, PhD (lead author) and Dominique Demolle, PhD helped author a recent paper published in Innovations in Clinical Neuroscience. The paper, “Using Artificial Intelligence-based Methods to Address the Placebo Response in Clinical Trials,” was a collaborative effort by leaders in artificial intelligence and machine learning, as well as their applications within the clinical trials space, and included prominent researchers working at Janssen Research and Development, Boehringer Ingelheim International GmbH, Johnson & Johnson, and the European Medicines Agency, among others. An additional four Cognivia team members contributed unpublished data in a section on using psychological traits to predict patient placebo responsiveness.
The paper was the culminating work of the International Society for CNS Clinical Trials and Methodology (ISCTM) “Use of AI/ML in Placebo Response” working group. The placebo response working group is one of five subcommittees of the “Utilizing Artificial Intelligence and Machine Learning for CNS Clinical Trials” working group.
“This publication resulted from a tremendous collaboration among more than a dozen scientists with expertise in clinical trial design, biostatistics, AI / ML and regulatory science,” said Erica Smith, PhD, Chief Business Officer at Cognivia. “We were thrilled to be a part of this esteemed group who has set the path for the use of artificial intelligence and machine learning to address the placebo response, a critical challenge to drug development that can prevent the delivery of needed medicines to patients.”
Advances in artificial intelligence and machine learning have unlocked new opportunities to tackle some of the more complex challenges facing the biopharmaceutical industry and its drug development efforts. Placebo response is one such challenge that has complicated the ability to demonstrate efficacy resulting in clinical trial failures based not necessarily on low medication response but on high placebo-response rates, especially in therapeutic indications where subjective, patient-reported outcomes are used as primary endpoints. Additionally, the placebo response can complicate the analysis of drug-treated patients. According to researchers, the placebo response makes up approximately 65 percent of the treatment effect in pain and depression studies.
The complexity of placebo response and its dynamic, often psychological nature mean the current approaches to reduce its influence through manual efforts have struggled to make a significant impact. AI and ML-based approaches, however, allow researchers to employ more sophisticated data analysis techniques at scale and speed to better understand and predict placebo responses in the drug development process.
The authors identified five key areas where various AI and ML approaches can effectively address placebo response:
- Drawing meaningful conclusions from large-scale, multi-trial meta-analyses
- Identifying, quantifying, and modifying trial design factors consistently associated with a high placebo response
- Predicting placebo responders before drug administration to improve targeted recruitment and/or population enrichment
- Stratifying and balancing patient distribution based on placebo responsiveness
- Utilizing predicted placebo response as a covariate during statistical analysis
The authors note some limitations and risks that should be considered, including potential outcomes biases depending on the method selected and “unsupervised associations,” both of which can be mitigated by appropriate clinical trial design, expert interpretation, and the use of independent datasets for evaluating the algorithms. Additionally, simply excluding high placebo responders across the board raises both a number of practical and ethical questions and is not something the authors recommend. However, they do offer some strategies and possible framework elements that would address the risks and provide guide rails for appropriate utilization.