Replies to post #100564 on Avid Bioservices Inc (CDMO)

[entdoc]

k_t,wow.what an effort. mightily appreciated. thanks eom

[horselover45]

Never ceases to amaze me the info that you-FTM-CJ-RR-Bungler and Entdoc provide for us amateurs.

Def "you all" make it much easier to hold and do a little buying through these testing times.

[iwfal]

There is also a question about which patient in the count represents the median patient for the overall survival plot when there was an even number of patients in the control arm. The 1 mg/kg Bavi arm included 40 patients, raising question whether the 20th patient breached the MOS or the 21st. ... As of the July update, only 19 patients in the 1 mg/kg Bavi arm had died, requiring at least one or two more patient deaths to be documented to reach an MOS endpoint similar to that reached in the control arm.

Suggest this is a little confused. Either:

a) the median is announced once the KM plot goes through 50% (which is what >99% of companies do - including this one. See the graph). And that has already occured for this trial and could theoretically happen with as few as 2 or 3 events out of an enrolled population of 100 (puzzle for the reader - why is this unlikely in real life?)

OR

b) the median is only announced when it can no longer move because the first 50% have died. And that could take anywhere from 0 more events to 10+ more events.

Either way hitting 20 events total doesn't, per se, trigger 'a median' per either of the above definitions.

[EYEBUYSTOX]

God, I wish we could add a "clinical trial site location" column to that spreadsheet. Based on King's remarks that "they could not have conducted a more balanced trial" with regard to locations, it sure would speak volumes to have a location next to those censored patients. Nice work, KT. Fingers crossed.

[jq1234]

Your numbers are off. At 3, 6, 9 etc time intervals, you can't have patients died plus censored plus patients at risk > total number of patients. For example, placebo arm at 3 months, you have 32 survivors, 6 died, 3 censored, adding up to 41 while there are only 38 patients in this arm. It looks like only 5 died, 1 censored at that point.

I agree with iwfal. NO ONE defines MOS for survival analysis the way you and many other on this board do!

[Protector]

KT, I never exactly understood why there is so much fuss about those censored patients. Now, with a simple look on the tables you provided, I see why some posters have a problem with it.

If I understand it correctly then those censored patients don't count as "death" even if they die after being censored.

So if many patients become censored, and if those patients would be the ones in bad shape (near death), then censoring them keeps them from counting for the MOS event and they also don't count to establish the 100% patient count. As I understand from the table MOS is reach if 50% of the patients died, NOT including the censored ones.

Now it becomes clear why RRDog's comment about the censoring not being done by Peregrine is so important.

PPHM is not in charge of censoring in a double blind gold standard clinical trial. That is done by the IDMC and clinicians.

Except for the fact that the trials wouldn't be double blinded if Peregrine needs access to the data for censoring it would also open the door to "cherry picking" :) when censoring.

It again all emphasizes FTM's post remarks about 1-Year survivors.

Censored or not censored, if you are alive then you are alive and if you are a stage 4 NSCLC patients that is not because they give you placebo, fresh water or pure air!

In other words, the 1-YS data will allow us to give a better interpretation on the possible weight, impact and importance of the amount of censored patients on the quality of the MOS. If the 1-YS data is good then those censored patients, whatever reason they where censored for, are less important then when we get poor 1-YS data.

[iwfal]

One more comment on your conversion of the curves into tables - you've made several substantive errors wrt censors. For example in the first three months of placebo you have 3 censors when you can tell from the at risk numbers at the bottom of the curve there can be only 1 ( there are 5 steps on the curve and the at risk numbers on the bottom show a change in patients at risk from 38 to 32 so there was only 1 censored).

Censor ticks are notoriously difficult to count in the best of circumstances and much more so in a graphic like this one that is an enhanced copy of a poor quality graphic.