Is Bavarian Nordic Dendreon 2.0? Ineffective Cancer Vaccine Masked By Misleading Data by Sahm Adrangi, Kerrisdale Capital
Bavarian Nordic is a $1.3B Danish vaccine-maker whose stock price has recently surged (up 63% YTD) thanks to excitement over its putative prostate cancer treatment, Prostvac-VF, a therapeutic vaccine currently undergoing a Phase III clinical trial. Bavarian Nordic touts its earlier Phase II study of Prostvac as showing the “most pronounced survival to date in prostate cancer,” with an 8.5-month improvement in median overall survival, handily outperforming blockbuster drugs like Zytiga and Xtandi. The announcement in March that Bristol-Myers Squibb was paying $60mm upfront for an exclusive option to license and commercialize the vaccine gave investors great confidence that, despite the uncertainty surrounding any clinical trial, Prostvac is likely to succeed.
This confidence is misplaced. The often cited 8.5-month improvement is an illusion: treatment-arm survival was unexceptional relative to the results of other trials in similar patient populations, while placebo-arm survival was anomalously poor. This strikingly bad placebo performance likely had several causes, but one important one was age: relative to men who received Prostvac, those who received a placebo were much older – indeed, older than any group we have come across in any prostate-cancer clinical trial. Researchers have clearly and consistently found – as common sense would suggest – that elderly men with prostate cancer, compared to their younger counterparts, do in fact live substantially less long. Comparing an unexceptional treatment group to an anomalously bad placebo group is a good way to show a strong benefit where none truly exists.
More recent efforts to demonstrate improved survival in patients receiving both Prostvac and the cancer drug Yervoy only further underscore Prostvac’s inefficacy. In a 30-patient trial with no control group, across a range of Yervoy dose levels, median survival was 31.6 months – compared to the ~30-month survival seen over and over again in the control groups of other late-stage prostate-cancer studies, a negligible “improvement.” Given that Yervoy itself clearly has some standalone anti-tumor activity and has been shown to extend survival by (a non– statistically significant) 1.2 months even in post-chemo prostate-cancer patients, Prostvac’s combination-therapy data look even less impressive. The natural conclusion is that any apparent benefit comes from Yervoy; Prostvac itself accomplishes nothing.
This finding should come as no surprise: the history of therapeutic cancer vaccines is two decades of unmitigated failure. We expect nothing different from Bavarian Nordic.
- Prostvac’s purported 8.5-month survival benefit is an artifact of a bad control. In Prostvac’s Phase II study, which began in late 2003, the Prostvac group lived for 25.1 months; the placebo group lived for 16.6 months. For the relevant subset of patients – men with metastatic, castration-resistant prostate cancer (mCRPC) who are minimally symptomatic or asymptomatic – 25.1 months is an unremarkable outcome. For instance, in the TAX-327 study, initiated back in March 2000, men with minimal symptoms had median survival of 25.6 months (1). More recent studies have yielded even better results: the control groups in trials for tasquinimod (2), abiraterone (3), enzalutamide (4), and orteronel (5), with enrollment start dates ranging from 2007 to 2010, survived 30.4, 30.3, 30.2, and 29.5 months in trials, respectively.
Since there is nothing special about Prostvac’s treatment-group performance, the purported benefit comes entirely from the anomalously bad performance of the placebo group. The original paper conceded that, applying a popular predictive model to a set of baseline prognostic factors, the treatment group had a 2.1-month survival “head start” (6), and an accompanying editorial (7) further noted (emphasis added):
[I]t is of concern that the control group had a median overall survival lower than that predicted by the Halabi et al. model (16.6 months actual compared with 20.4 months predicted). The reasons for this discrepancy are not at all clear, particularly given the eligibility criteria designed to select lower-risk patients.
We hypothesize that an important source for this discrepancy is age: the median age in the Prostvac group was 71.5 years, while in the placebo group it was 79. We have found no group in any other prostate-cancer trial with an age distribution skewed so far to the right. The original paper’s authors dismiss this massive imbalance (favoring the Prostvac group) with the claim that “age is not a significant prognostic factor in prostate cancer,” citing the predictive model published in 2003 by Susan Halabi et al. (8), but this claim is wrong. Halabi’s model was based on data from men with a relatively narrow range of ages, over which small differences may not matter.
By contrast, a study focusing specifically on elderly mCRPC patients (aged 75 years and older) showed that those in relatively good condition experience median survival of 17.5 months (9) – very similar to the outcome for the Prostvac placebo group and approximately 10 months worse than that for younger men with similar disease characteristics (1) (10). Another publication showed that mCRPC patients aged 85 years or older have 5-year survival rates that are less than a third of those for younger men aged 65-74 (9), while a 2006 study by Halabi et al. noted that 60-to-69-year-olds experienced survival similar to that of 70-to-79-year-olds but lived almost twice as long as 80-to-89-year-olds, who constituted roughly half of the Prostvac Phase II placebo group (10).
In short, age does affect overall survival even for men with late-stage prostate cancer, likely explaining, at least in part, the Prostvac placebo group’s unusually bad performance. Regardless of the cause, though, there is no reason to expect such a bad control to recur in the larger Phase III study, set to finish in late 2016 or beyond; thus, there is no reason to expect Prostvac to show any benefit. The treatment will fail.
See full PDF below.