Vascular’s dressed-up data call for caution

It is ironic that on the day the biotech sector took a battering over concerns that the bubble was about to burst, the strongest-performing stock – Vascular Biogenics – was driven by some decidedly strange-looking phase I/II data.

Vascular’s 30% rise yesterday crowned a rollercoaster ride in which the Israeli group suffered a clinical hold, had to take a 57% price cut to get its Nasdaq IPO away, and saw another R&D project fail in phase II. Its hopes now rest on the anti-angiogenic gene therapy VB-111, though based on yesterday’s data investors would do well to proceed cautiously.

On the face of it Vascular’s phase I/II results with VB-111 in relapsed glioblastoma multiforme look like a success, the company touting a statistically significant median overall survival benefit for patients who progressed and were given VB-111 plus Avastin, versus those who went on to get Avastin alone.

However, closer reading reveals a stark difference between these interim data and the interim data Vascular had presented at last November’s Society for Neuro-Oncology meeting. Back then it said median OS had been reached and had failed to hit significance, with p=0.22.

Moving the goalposts

The answer might lie in a decision to change the statistical analysis method. Vascular said yesterday’s analysis had used the so-called Wilcoxon test to yield a p value of 0.05, while November’s data had used the more stringent log-rank.

Applying Wilcoxon to the earlier analysis would have given a still non-significant but much lower p=0.09, it said. The company did not respond to a request to clarify when the decision to change the analysis method had been taken.

In any case, the trial had an unusual design. It had started as a single-arm study in which patients got VB-111, and those who progressed switched to Avastin; part-way through the design was changed to allow progressors to change to a combo of VB-111 and Avastin, on a sequential, non-randomised, open-label basis.

While this could have introduced bias, Vascular said the baseline patient characteristics were “very much comparable”. Of 46 progressors, 22 continued with Avastin, 23 with the combo and one remained on VB-111 monotherapy; it is not clear how the last patient was treated in the statistical analysis.

However, another imbalance is that patients in the combo arm, which performed better, got on average three doses of VB-111, versus one in those who switched to only Avastin. While the study was designed to measure safety and tolerability – which Vascular said were good – the group’s highlighting of a survival benefit thus looks premature.

Less is more?

Back in November Vascular had reported a non-statistical median OS benefit of 269 days for the combo arm; yesterday it said that had been a relatively immature result, and said the median OS benefit was now 179 days, which was statistically significant.

A survival benefit falling yet becoming statistically significant is unusual, but can be explained by a change in the shape of the survival curves, though the group did not reveal hazard ratios.

This is not the first time that VB-111 has hit the headlines, having been placed on US clinical hold last July pending submission of additional potency data. That was lifted in February – on the same day that another project, VB-201, failed a phase II trial in psoriasis and ulcerative colitis.

One thing in the company’s favour is the intractable nature of brain cancer and the lack of treatments; indeed, Avastin itself is underwhelming here. This sets the bar low, and under a US special protocol assessment Vascular aims to run a 200-patient phase III study with an open-label, two-arm design looking to increase OS.

As for yesterday’s trial, this is to be presented in full at a scientific meeting, with Vascular considering submitting it for a late-breaker at Asco. Before investors speculate on the outcome of phase III they should subject the full data to a rigorous analysis.

To contact the writer of this story email Jacob Plieth in London at [email protected] or follow @JacobPlieth on Twitter