Peregrine took none of these precautions. Seventy-nine patients were enrolled in the two bavituximab arms but only 40 patients were accounted for in the survival analysis. That means 39 bavituximab patients, or 49%, were censored and excluded from the analysis.
[Censored patients show up as "tick marks" on the bavituximab Kaplan-Meier survival curves seen in the slide below.]
To put Peregrine's 49% patient censoring rate in perspective, six phase III lung cancer studies published recently in the medical literature had an average censoring rate of 19%. On average, each of these studies enrolled 710 patients.
Peregrine's bavituximab study enrolled just 117 patients (including the 38 patients in the control arm) but had more than double the rate of censoring. That speaks to very poor trial conduct.
The amount of patient censoring is important but so is the time at which censoring occurs. In well-conducted clinical trials, more censoring occurs
median overall survival is reached. This typically reflects patients who remain alive in the study when the analysis was conducted. If or when these late-censored patients die, the median overall survival estimate of the trial will not change appreciably. Remember, median is the value in the middle, not the average.
In the Peregrine study, however, much more patient censoring occurred
median overall survival was reached. [You can count the tick marks on the survival curves.] As a result, the median overall survival estimate is skewed higher artificially. If any of these early-censored patients are found to have died, the median overall survival of pooled bavituximab will not be 12 months but could be significantly lower.
The problem with early censoring can be seen most clearly in the study's 3 mg bavituximab arm, where the median overall survival was reported to be 13 months. However, this analysis was based on 15 deaths, while 18 patients were censored before the median survival was reached. When more patients are missing than confirmed dead in a survival analysis, the estimate is entirely unreliable.
Early censoring can be justified if a trial is relatively immature, meaning patients haven't been followed long enough to analyze.
That's not the case with the Peregrine trial. The bavituximab study was unblinded in December 2011. When the survival data were presented on Sept. 7, the company had 9 months of follow up at a minimum (and this assumes that all patients enrolled on the last day of the trial, which obviously didn't happen.)
Peregrine hasn't disclosed the reasons why patients were censored and the company did not return a phone call seeking comment. But based on previous statements, the bavituximab study was mature enough to avoid early censoring of patients for too-short follow up.
That leaves just one possibility: All the patients censored early in the study before nine months -- and there are lots of them -- were lost to follow up.
--Written by Adam Feuerstein in Boston.
>To contact the writer of this article, click here:
>To follow the writer on Twitter, go to
>To submit a news tip, send an email to:
and become a fan on