Vascular Biogenics Ltd (VBLT)

[wcopeland]

Comparing 17 vs. 4 is not statistically rigorous. VBLT and their collaborators could have done a better job there. However, you (and me and others) can do our own diligence and see how significant 400-800 days survival is for PrOC patients.

Censoring is the accepted way to handle an incomplete observation.
https://en.wikipedia.org/wiki/Survival_analysis#Censoring

For example, imagine that we know a PrOC patient received VB111 6-months prior to a data release (ie. June 2016 at ASCO) and is still alive at the time of data release. Since that patients has not died, we cannot tell you how long she lived following treatment (... at least not until we invent time machines). Any time that we do not know the start and end of an event, we cannot confidently say how long it occured.

One extreme solution would be to ignore any patient where VBLT does not have both a start and an end. In this extreme, they really limit their sample size since only 7/17 had died by the data cut. Also they would ignore a lot of useful inferences if they did that. For example, they knew that someone was still living after about 840 days. If they ignored that knowledge because the patient had yet to die, they would severely underestimate the power of the therapy for many patients.

On another extreme, VBLT could have arbitrarily stopped observations and only used data collected by June 2016. In this extreme, you discard the fact that (in reality) this lady who has survived for 6-months by June 2016 actually goes on to live until September 2019. Do you see how irrationally conservative that is?

Time-to-event analysis was developed as a statistically rigorous method to not end up on either extreme. We make use of all complete data: all 17 VB111-exposed patients as opposed to only the 7 with observed deaths. However, since we cannot predict the future, we acknowledge that the patients lived for AT LEAST 6-months before data became incomplete and we were forced to censor.

That AF guy is an idiot. Heavily left-censored data is likely the result of not enough time progressing since many patients initiated therapy. Outside of not dying, the other likely reason for data censoring is that the patients stopped accepting follow ups from VBLT representatives/doctors. Any other reasons for censoring would be questionable in my book.

I (and most here) absolutely recognize that a positive outcome of Ph3 GLOBE is uncertain and risky. However, something as basic as survival analysis IS NOT uncertain. A strong understanding of it and other aspects of the trial should help you avoid being led astray by trolls like AF or Staccani.

Another helpful link:
http://www.theanalysisfactor.com/censoring-in-time-to-event-analysis/