AB Science's dog drug data raise more questions than answers

AB Science’s plans to turn its canine tumour drug masitinib into a human therapeutic appear to have been given a massive boost thanks to the positive subgroup analysis of a failed phase III study in the notoriously difficult setting of pancreatic cancer.

However, the company has refused to confirm whether the secondary analysis had been pre-specified under the trial’s protocol, let alone revealing patient numbers or the biomarker involved. If it is later revealed that the stunning result is merely the result of data mining investors should brace themselves for a sharp retrenchment in AB’s stock price – something that would be compounded if the European regulator were to request a prospectively designed confirmatory study.

Masitinib, an oral tyrosine kinase inhibitor, is being studied against a range of human cancers. The latest data came from a trial in advanced or metastatic pancreatic cancer in which it was added to Eli Lilly’s Gemzar. News of the positive results sent AB’s shares up 22% to €18.70 yesterday, and the stock has continued climbing, breaching the €20.00 mark today.

Failed primary

The primary endpoint of median overall survival across the 348 patients enrolled was missed, with masitinib plus Gemzar showing 7.7 months versus 7.0 for Gemzar alone.

The excitement was caused by the result in two subgroups that AB says represent particularly aggressive disease. Patients with an undisclosed genetic biomarker survived just 5.0 months on Gemzar, but 11.0 months on Gemzar plus masitinib – reducing risk of death by 71% with a p value of an amazing 0.000038.

And those with cancer pain intensity above a certain (again undisclosed) level also did well, with the active group showing median OS of 8.1 months, versus 5.4 months in the control arm, hitting statistical significance with p=0.010. AB says both the biomarker and high-pain groups represent patients that do particularly poorly on Gemzar.

But it is what has not been disclosed that should give observers pause. We do not know how many times AB analysed the same dataset – something that would affect the statistical significance of the positive result – nor whether the two subgroups had been pre-specified.

The trial’s listing on clinicaltrials.gov mentions nothing about secondary subgroup analyses. If the subgroups had in fact not been pre-specified then it would seem that all the company has done is mined the data until it did hit something where the active and control groups separated – making the result highly questionable in statistical terms.

Moreover, without being told how many patients were actually in the two subgroups we do not know how likely it is that the result is a fluke. AB did say that the mystery biomarker and high degree of cancer pain are manifest in 65% and 45% of the overall pancreatic cancer population respectively.

AB’s CEO, Alain Moussy, said he would not comment on the data beyond that revealed yesterday, but admitted that questions about subgroups, multiple non-predefined analyses and patient numbers were “relevant”. The ASCO-GI meeting in January, at which the full dataset will be presented, will be the “right place to respond”, he told EP Vantage.

Filing basis

The company revealed that it was this same phase III subgroup analysis that had formed the basis of a recent EU filing for pancreatic cancer (Event – AB Science awaiting first test of scattergun approach, September 22, 2011). The European Medicines Agency has just accepted this filing for review.

Masitinib’s human clinical study programme represents an unusual extension for a veterinary drug. The product is already marketed as a treatment for mast cell tumours in dogs, as Masivet in Europe and Kinavet in the US, although sales have been very modest so far.

The drug is also awaiting EU approval for treating gastro-intestinal stromal tumours in patients who have progressed on Novartis’s Gleevec. This filing was based on a 44-patient phase II trial in which masitinib outperformed Pfizer’s Sutent in terms of overall but not progression-free survival (AB Science tastes success in GIST phase II trial, February 2, 2012).

With the latest round of apparently good news AB Sciences’ share price has at last exceeded the €12.65 at which the company floated in 2010, and AB is currently valued at over €650m. At this level, and still without a partner, AB Sciences would probably be best off seizing the moment to raise some much-needed cash.

Certainly a lot more information must be forthcoming before investors know for sure whether such a remarkable market cap is sustainable.

To contact the writer of this story email Jacob Plieth in London at [email protected]