Discussion 253 DISCUSSION This thesis is the result of several studies in which various aspects of expanded access to unapproved medicine have been investigated. The consequences and limitations of all individual studies, such as methodological flaws or limited research scope, have been addressed in the corresponding papers. These concerns are further elaborated on in the introductory prologues and concluding epilogues of the individual Parts (if applicable), providing context and scope for each of our research questions. This discussion will be dedicated to debate persistent overarching issues, such as the buzz around real-world data, the artificial separation between research and treatment, and the potential benefits and drawbacks of public-private partnerships. Lastly, we suggest promising directions of future research and set realistic expectations for expanded access as it continues to evolve and straddle the line between clinical practice and scientific conduct. Unmasking the hype: misconceptions on ‘real-world data’ During our investigation into the value of expanded access data, the interest in ‘real-world data’ was on a rise.1 A search of PubMed reveals that in 2010, 562 articles were indexed for real-world data, which increased more than twofold in 2014 to 1,243 publications, followed by a substantial upsurge to 2,421 in 2018 and a staggering 9,268 articles in 2022. We gratefully capitalized on this trend, as it is undeniable that the term ‘real-world data’ holds significant appeal. After all, who would prefer to use ‘fake-world data’? The success of popular expressions as ‘precision’ medicine or ‘targeted’ therapies is fueled by the fear for their antonyms,2 the underlying notion that no one wants their drugs to be ‘imprecise’ or ‘off-target’.3 However, epidemiologists and biostatisticians have been scrutinizing non-interventional data for several decades, advancing designs like case-control, case-cohort, propensity score analyses, or target trials, to partly attenuate biases and produce replicable results, but lacking a fancy term to market their ideas.4 The concerns that trial results do not provide information on real-world usage are, in my opinion, in part, based on misunderstanding about trials. One common misconception about clinical trials is that they are based on random sampling, when in fact they are based on randomization.5,6 Trials enroll patients who volunteer, creating a selection bias. Real-world data does not necessarily solve this ‘issue’ as it too does not ensure random sampling. For instance, expanded access programs may disproportionately include wellconnected, affluent, healthy individuals, further clouding the estimate of drug outcomes in ‘realworld’ populations.7 Addressing the issue of selection bias requires more than merely increasing patient numbers. If bias is present, large numbers will simply perpetuate that bias throughout the study design.
RkJQdWJsaXNoZXIy MTk4NDMw