You are misunderstanding several aspects of the analysis.
First, the external control arm used in the DCVax-L analysis was not constructed by randomly pooling “all patients.” It was built from individual patient-level data drawn from multiple contemporary randomized trials, and those patients were filtered to match the Phase 3 inclusion criteria as closely as possible. Age, performance status, MGMT status where available, extent of resection, and other clinical variables were used to align the cohorts. That is exactly how external control arms are constructed in oncology research.
Second, the trial did not require “near total resection.” The protocol required intent for gross or near-total resection, but the enrolled population still included the range of real-world surgical outcomes typical for newly diagnosed GBM. The comparator datasets used in the ECA include patients with comparable surgical status distributions. Cherry-picking the MRD subgroup from one study and claiming it represents the appropriate baseline is not how regulators evaluate overall survival comparisons.
Third, the suggestion that reconstructing survival data from Kaplan–Meier curves is somehow illegitimate is simply incorrect. Reconstructing individual patient data from KM curves is a well-established statistical method used in peer-reviewed oncology research for meta-analysis when raw datasets are unavailable. It has been published and validated repeatedly and is widely used by academic groups.
Most importantly, regulators such as the MHRA and FDA are not relying on social-media interpretations of graphs. They have access to the full clinical trial dataset and statistical analysis plan, and they evaluate the evidence based on the complete data package.
So the premise that the comparison is based on a casual reconstruction of curves or on an inappropriate patient mix is simply not accurate.
Market Data 
Markets 
AI
