When time itself is the comparator: Methodological problems and a way forward for trials of surgery versus...

Randomised trials that compare a surgical intervention to medical or conservative management occupy a peculiar space in clinical research. The two arms are not symmetric in time: surgery imposes a concentrated, acute risk in exchange for a long-term benefit, whereas conservative management distributes a smaller hazard across the patient's remaining life. In neurosurgery, where the conditions under study are frequently rare and natural histories are measured in decades, this asymmetry exposes weaknesses in conventional trial methodology that have repeatedly led to inconclusive, non-generalisable, or actively misleading results.

Using as illustrations the ARUBA trial of interventional therapy for unruptured brain arteriovenous malformations and the COSS and CMOSS trials of extracranial-intracranial bypass for symptomatic carotid occlusion, we identify the recurrent problems of surgery-versus-conservative trials: contestable equipoise, ambiguous eligibility, heterogeneous interventional arms, mismatched follow-up to the natural history of disease, restricted external validity, and uncritical extrapolation beyond the population actually randomised. We argue that the most consequential of these problems are not statistical but conceptual: investigators repeatedly fail to design trials in which the temporal structure matches the clinical decision the patient is actually making.

We propose a portfolio of design strategies, several borrowed from oncology, critical care, and cardiology, including expertise-based randomisation, cohort-embedded designs, adaptive designs with informative natural-history priors, platform protocols with a shared control arm, risk-stratified enrichment, and pre-specified lifetime microsimulation. Properly combined, these designs may yield evidence that physicians trust and apply, rather than evidence physicians cite while continuing to operate as before.