EMA starts drawing lines around external control arms
Amsterdam is finally writing down what counts as a credible non-randomised comparator, just as a year of methods papers shows how often submissions still skip the basics.
Biostatistics Desk
The evidence-synthesis stack is being rebuilt, all at once
An arXiv preprint hands clustering decisions to an LLM while a cluster of new JCE papers concedes that GRADE, PRISMA-P, and guideline appraisal each need patching. The substrate under every HTA dossier is moving.
CRT inference: three papers, one estimand problem
The cluster-randomized headline number is doing less work than it looks. May–July 2026 brought a closed-form bias proof, a reporting indictment, and a workaround that sidesteps the ICC entirely.
CNS trials reach for Bayesian neurophysiology while blinding quietly leaks
A semaglutide-for-cocaine protocol, a scoping review on active placebos, and an EMA concept paper on smoking cessation all land in the same week — and all point at the same unresolved problem.
The OS Toolkit Arrives Just as ASCO Stress-Tests It
SBR's special section gives oncology biostatisticians a coordinated framework for OS monitoring, safety, and interim looks — and ASCO 2026 supplied the awkward case studies before the ink dried.
The Win Ratio's Methodology Bill Comes Due
A simultaneous wave of papers formalizes the win ratio's pathologies — including a 2×2 reversal in which treatment beats control on every component and still loses overall — while cardiovascular sponsors keep pre-specifying it as primary.
Bayesian Borrowing Methods Converge on a Single Problem: Defending Type I Error
Across four journals in one cycle, independent groups are re-engineering robust mixture, power, and synthetic priors with frequentist error control as the design constraint — not the afterthought.