A new study conducted by researchers at Imperial College London and the University of Melbourne and published in the February 2019 edition of Nature Communications revealed a potential clinical application for artificial intelligence (AI) in the treatment of ovarian cancer.
Using four “descriptors” from pre-operative CT images, the software named TexLab 2.0 generates a Radiomic Prognostic Vector (RPV), which indicates disease severity and predicts survival with greater accuracy than the current methods, which are based primarily on tumor stage (FIGO staging: I–IV) and tumor location (epithelial, stromal, or germ cell). The method has been validated using two independent, multi-center cohorts. A potentially exciting application of this technology would be to determine whether the algorithm can correctly predict response to certain therapies, and thus bring personalized therapy into the ovarian cancer space.
Ovarian cancer has a number of high unmet medical needs. Despite recent advances to the treatment paradigm, with the addition of Avastin between 2011–2014 in the seven major markets (7MM: US, France, Germany, Italy, Spain, UK, and Japan) and poly ADP ribose polymerase (PARP) inhibitors, which were first approved in 2015 in the US and Europe, outcomes remain poor, with five-year survival rates between 30–50% globally.
First-line treatment typically consists of platinum-based chemotherapy, with subsequent treatment choices based primarily on tumor stage and the degree of response to the first line of therapy—deemed platinum-sensitive, platinum-resistant, or platinum-refractory. However, ovarian cancer represents a heterogeneous group of diseases with varying genetic profiles and pathologies, and the platinum-sensitivity paradigm does not accurately predict treatment response. Over the next 10 years, GlobalData expects that immune checkpoint inhibitors, including cytotoxic T-lymphocyte–associated antigen 4 (CTLA-4) and programmed cell death protein 1 (PD-1)/programmed death ligand 1 (PD-L1) inhibitors, will also enter the ovarian cancer space. Thus, the need for individualized therapy to direct treatment to the proper patients will become even more essential in coming years.
In addition to improved outcomes for patients, individualized therapy for ovarian cancer could excite payers, who aim to use healthcare resources efficiently and reduce overall costs. One promise for individualized treatment, which directs the right therapy to the right patient at the right time, is to reduce the instances of treatment failure, which may be viewed as wasted healthcare expenditure. Improved targeting will benefit patients through improved outcomes and reduce exposure to toxic therapies that are unlikely to show much benefit for a particular patient.
As much promise as this AI technology shows in ovarian cancer, the study authors also suggested that the RPV methodology could be applied to other cancer types. This would be a welcome discovery for other tumors with high unmet medical needs that lack the benefits of predictive biomarkers, such as head and neck cancers, cervical cancer, or pancreatic cancer. As a new approach to healthcare, AI has tremendous potential that has not yet been fully exploited. Practical applications of such technologies, especially in the area of precision medicine, are likely to benefit multiple stakeholders, including physicians, payers, and patients.