From randomized clinical trials (RCTS) to artificial intelligence (AI) – history and methodological challenges

Abstract

From 1991 to 2020 a considerable number of quality devices (i. e. Equator [2] have been developed and implemented, for the quality of all steps of trials from planning to implementing the knowledge derived from results. This may sound as if it was an easy path but it was not. Quality is a value which is under permanent attack and has to be defended. Bias is a major aggressor, probably the most famous in that camp is the publication bias. Up to 50% of research results are not published, with a heavy bias of what is published and visible [3], accordingly with serious ethical impact [4]. In recent years a new era has been launched, starting 2008 with a provocative article claiming that enough data make the scientific method obsolete [5]. Big Data, artificial intelligence (AI) and personalized medicine (also called precision medicine) have generated a realm of visions and promises where the quality issue seems to have completely disappeared: Unlimited data guarantee any level of needed quality, without particular effort. Can this be expected, or where is the border between realistic expectations and marketing-driven promotion? We are observing a confrontation and a cultural clash between the “old”, methods-driven world and the new “informatics-based” or “data-driven” world which is not receiving the attention it deserves in the current climate of enthusiasm and hype [6]. It is urgent to avoid misleading perspectives and return to strictly quality-driven research agendas and the implementation of these methods [7]. The presentation describes and illustrates these different methods worlds and their tensions and controversies.