“…2 Such designs routinely enroll drug-naive subjects who leapfrog earlier cohorts into never-before-tried doses, placing new subjects at unnecessary risk of severe and fatal toxicities while consigning earlier-enrolled subjects to possibly subtherapeutic low doses. 7 To treat enrolling persons not as 'subjects' but truly as patients-that is, with therapeutic intent 8 -demands a precautionary coherence 3 under which drug-naive patients enroll at low quantiles of population MTD i , whereas never-before-tried doses are attempted preferentially in those already demonstrating tolerance of an adjacent dose level.This distinction between subjects and patients helps us to gauge what actual importance the studies by Wages et al 1 attain, relative to the supposedly clinical motivations adduced for them. All dose-escalation designs, whether of the Continual reassessment method (CRM) family, interval-based, or algorithmic, evaluate doses as their primary objects of interest, and for this purpose typically exploit subjects for the Bernoulli random variates they yield in dose-limiting toxicity assessments.…”