Transforming Clinical Trials with Artificial Intelligence

Abstract

Current treatment guidelines are implicitly based on an “average patient” which ignores the complexity of human pathophysiology that manifests with significant inter-individual differences in treatment response. (Mulder et al., 2018) Randomized controlled trials (RCTs) have been utilized for many years 298and are often heralded as the “gold standard” for determining the safety and efficacy of treatments. In a conventional RCT, the average effect of a drug is compared to the effect of placebo or other active treatment, by assigning patients to alternative treatment groups and recording outcomes. To prevent selection bias patients are randomized to different groups, which produces distinct trial arms that are made up of patients who are largely similar except for the intervention they receive. This balances any confounding factors, known and unknown, and means that any difference in effects between groups is likely to be true, unconfounded treatment difference. However, RCTs are by no means faultless and can induce a multitude of problems including but not limited to poor patient recruitment and high attrition rates, expensive trial designs and the ever-increasing costs of conducting trials which tend to be large, international multi-centre endeavours.

Ähnliche Arbeiten