Okay, listen up, @Bioconlimited @kiranshaw, and others
I've now seen the youtube press conference. Let me take up one issue with Biocon's Chief Medical Officer, who addresses the Q of small sample size (from 1:35 onwards in the video
I've now seen the youtube press conference. Let me take up one issue with Biocon's Chief Medical Officer, who addresses the Q of small sample size (from 1:35 onwards in the video
He says quite clearly, that the small sample size doesn't matter because the trial achieved statistical significance. What does that really mean? rhetorical question. It means that if the null hypothesis of no difference between Itoluzimab and standard care was true, then..
this particular trial 30 patients, 20 Arm A with 0 deaths, 10 Arm B with 2 deaths (I presume those numbers are accurate) provided evidence to reject that Null Hypothesis. But in doing so we may be committing a Type 1 error.
His next statement is even more disturbing. "if we had not found such a dramatic difference , ie. if none of the control arm patients died, then we'd have carried on with the trial". I do not believe he understands the full import of this admission. What he is in effect admitting
Is that the trial design did not really have a pre-defined sample size. It was designed to generate a p value in favour of the Itoluzimab arm. That is simply using science and the altruism of the patients who consented in good faith, to generate a result favourable to the company
I have heard of trial that have a pre-defined stopping rule but not one with a pre-agreed but undeclared continuing rule. Now let me explain what the p value does not do.
A) it does not mean that the trial is generalisable to the population of patients with #Covid19 ARDS and CRS
A) it does not mean that the trial is generalisable to the population of patients with #Covid19 ARDS and CRS
B) it does not in and of itself, get around the problem of a poor trial design. A shaky marksman who hits the bullseye in Round1 could have just been lucky.
C) Beyond a fixation with p-values there is confidence intervals and NNTs. So my question to this panel is
C) Beyond a fixation with p-values there is confidence intervals and NNTs. So my question to this panel is
1. What % age reduction in death rate can we expect if this drug is rolled out for all Covid19 ARDS cases? Are they claiming that the NIL deaths in 20 cases will translate into 0 deaths in 500 cases?
2. What is the 95% C.I. range for this mean %age reduction in death rate
2. What is the 95% C.I. range for this mean %age reduction in death rate
3. What is the NNT, i.e how many #Covid19 ARDS patients need to be treated with #Itoluzimab in order to prevent 1 addition death. Can this figure be as low as 1. as their data seem to suggest? really? that would make it a guaranteed death preventer.
4. One of the key ingredients of clinical trials is reproducibility. What would the response be if someone else did a similar trial and reported a different result? Would they rubbish that trial or accept that trial results are prone to error
5. If a p value is all that matters why is the US approval for a trial of #Itoluzimab in lupus nephritis only for a Phase 1b and even so the plan is to enrol 56 patients.
End of rant
End of rant
