PathAI is partnering with GSK on a randomised Phase IIb trial in non-alcoholic steatohepatitis (NASH). The Boston based artificial intelligence (AI)-powered pathology company will be responsible for generating, digitising, and analysing liver biopsy slides for central pathologist evaluation.
In addition, its AI-based Measurement of NASH Histology (AIM-NASH) tool will be used for histological evaluation. The AIM-NASH tool has been trained to detect and quantify histological features of NASH.
The collaboration between PathAI and GSK is part of a multi-year partnership which is focusing on drug development in NASH and oncology.
The Phase IIb study (NCT05583344) will measure liver fibrosis and inflammation improvements with GSK4532990 compared to the placebo. The trial aims to enrol 246 participants with pre-cirrhotic NASH and metrics generated by the AIM-NASH tool will be included as exploratory endpoints in this trial.
Earlier this month, PathAI announced the launch of AISight, a digital pathology platform, and AIM-PD-L1 NSCLC RUO algorithm, which evaluated the percentage of PD-L1 positive tumour and immune cells in non-small cell lung cancer (NSCLC). 13 anatomic pathology laboratories across the US will participate in the early access programme to gather real-world evidence about the use of digital tools in pathology to advance precision medicine.
AI continues to shake the conservative nature of the pharmaceutical industry as its various applications are being tested in different stages of clinical trial conduct. For example, the combination of predictive analytics and AI has the potential to bypass animal testing and forecast simulated interventions.
The technology is also being used to speed up patient identification and recruitment, and ensure patient adherence and retention in a trial. Additionally, the implementation of AI can improve the data analysis that is generated from digital biomarkers.
More novel AI applications, such as digital twins, are also entering the clinical trial space. Digital twins allow researchers to digitally replicate trial participants and observe them in real-time while simulating and predicting a different clinical outcome for the same person.
However, regulators are not keeping apace with the evolution of AI. Currently, AI-powered tools are regulated as medical devices, but they require a different level of oversight as most of these technologies generate human unreadable code.
Moreover, there has been concern over bias in AI as algorithms are trained on already existing datasets, which are often biased. This creates a risk for AI-generated outcomes being incorrect, and potentially dangerous for patients.
