NorthWest Biotherapeutics Inc (NWBO)

[dennisdave]

Nope its you failing to understand

You’re right about the prognosis hierarchy. Nobody disputes that the extent of resection (EoR) is a dominant covariate in GBM: GTR (near-total) > sub-/near-GTR > partial/biopsy for survival. If the ECA contains more partials/biopsies, its raw median will look worse. Point taken.

But several of your leaps don’t follow unless we accept unproven assumptions.
1) “No benefit in near-total resection” based on what statistic?
Saying “no benefit” because medians look similar is weak. Median OS is a blunt instrument in datasets with long-tail survivors. Two fixes are standard:
* Hazard ratio (adjusted) in the near-total stratum (with balance checks on age, KPS, MGMT, steroid use, time-to-treatment).
* Restricted Mean Survival Time (RMST) to capture area under the curve (long-tail effects) rather than a single percentile.
If the adjusted HR ˜ 1 and RMST difference ˜ 0 in near-total, fine then “no benefit” holds. But you haven’t shown that; you’ve asserted it from medians.

2) Collapsing “substantial remaining disease” into one bucket (2 + 3) corrupts the comparison
Group 2 (“close to near-total”) and Group 3 (partials/biopsies) are not interchangeable. Residual enhancing volume is continuous, prognosis is monotonic, and biopsy-only is an outlier.

Lumping 2 + 3 together, then claiming the apparent effect is “just because ECA has more 3,” is post-hoc pooling that can create or erase signals. What’s required: analyze 2 and 3 separately, and crucially exclude biopsy-only ECA patients in a sensitivity run to see if the effect persists in true residual disease (partial resections) rather than pure biopsies.

3) “Apples to apples” needs time-zero alignment and EoR harmonization
Two typical biases can dwarf treatment effect:
* Immortal time bias: If DCVax-L time-zero is “vaccination-ready” but ECA time-zero is “diagnosis” or “surgery,” you’ve gifted DCVax time patients didn’t actually “survive” under risk.
* EoR definition drift: “Near-total” must be defined by the same residual volume threshold and imaging rules on both sides (preferably re-read or algorithmically harmonized).
* Unless time-zero and EoR thresholds are aligned, your “apples to apples” claim is not apples to apples.

4) Balance the other killers: MGMT, age, KPS, steroids, location, center effects
EoR is big, but not alone. A credible ECA analysis shows standardized mean differences (SMDs) for covariates < 0.1 after matching/weighting, and reports stabilized weights diagnostics. Without that, you can’t attribute any difference to treatment rather than case-mix.

5) “Burden of proof is on NWBO” you say. Agreed; here’s the bar that would settle it
If NWBO can (and to a regulator likely did) deliver:
* Pre-specified ECA plan (timestamped SAP), not retrofitted;
* Aligned time-zero and harmonized EoR definitions;
* Adjusted HRs + RMST within each EoR stratum (near-total, sub-near-total, partial),
Sensitivity analyses that (a) exclude biopsy-only ECA, (b) restrict to contemporaneous controls, (c) demonstrate covariate balance (age, KPS, MGMT, steroid dose/use, time-to-TMZ/RT, center effects), and (d) include a tipping-point analysis for unmeasured confounding,

Then your “it’s all Group-3 contamination” argument loses oxygen. If those analyses still show neutral results in near-total and benefit only in residual disease, then we’d be looking at a biologically plausible interaction (vaccine benefit greater with antigen load) rather than a mere artifact. But we need the adjusted stats to say that not pooled medians.

6) What your current argument does and does not prove
Valid: Raw medians can be skewed if ECA has more partial/biopsy patients.
Not proven: That all observed benefit in the “residual-disease” DCVax cohort is explained by that skew. You haven’t shown an EoR-matched, time-aligned, covariate-balanced null effect.

If, after EoR-matched, time-aligned, covariate-balanced analysis with RMST and sensitivity exclusions of biopsy-only, the near-total stratum still shows no improvement, call it no benefit there fair. But you can’t dismiss the residual-disease signal as “just ECA biopsies” without running the very controls that regulators require. Until those controls are presented (and they may already be in confidential submissions), your are just guessing around. (as most of the time)