Even by the low standards of failures in Phase III studies, Targacept’s failure is phenomenal. Phenomenal because TC-5214 had shown such promise in a Phase II depression study – demonstrating a six point improvement in the Hamilton D (or HAM-D) depression score of the combination of TC-5214 and Forest’s (NYSE:FRX) Celexa compared to Celexa alone. A two or three point improvement would have been good enough. In addition, secondary endpoints in that Phase II study were also all positive (for example, in the Montgomery-Asberg Depression Rating Scale, or MADRS, TC-5214 plus Celexa demonstrated a greater than seven-point improvement over Celexa alone). This was enough for Astra Zeneca (NYSE:AZN) to pony up $200 million in up-front fees and in-license TC-5214 for a total of $1.24 billion. The attraction was not just based on the study, but also on a new mode of action – TC-5214 is a modified version of mecamylamine, used last century to treat hypertension but in trials for Tourette’s were found to have anti-depressant properties. In contrast, most anti-depressants work through the serotonin or nor-epinephrine pathway.
Fast forward to Phase III, and the first of four Phase III studies looking at efficacy, Renaissance 3 (R3) was a bust. 624 European patients were recruited to receive standard therapy of SSRI and SNRI. After eight weeks of treatment, 295 patients who did not respond adequately were randomized to receive either a flexible dose of TC-5214 or placebo on background therapy. The dosage of TC-5214 was initially 2 mg/day and could be increased at the discretion of the investigator to 4 mg/day and 8 mg/day based on tolerability and therapeutic response.
So what happened in Phase III? How could a drug that was so promising in Phase II fail in Phase III? Several points need to be considered.
1. STUDY SITE: The Phase II trial was conducted in U.S. and (repeat and) India, not just India alone. The R3 was conducted in Europe. There are several cultural differences between the East and West that people conducting trials should be aware of. Indian patients are much more trusting than patients in the U.S. or Europe. Thus, even when told they could be getting placebo, they are liable to show a higher response rate than those in the West. In a highly subjective disease such as depression, where there are no objective measures, that can have a huge effect. But, as mentioned, the Phase II study was in U.S. AND India, not just India alone. Were there too many Indian patients that skewed the data in the Phase II study? Possible and I’m sure both companies have already evaluated that possibility. Looking at efficacy in just the U.S. patients of the Phase II study and comparing it to the Indian patients should have been done. If there was a huge difference, that would have been a red flag. If there was no difference, the answer lies elsewhere.
2. STUDY DESIGN: Both studies had an unusual “flexible design” rather than the standard parallel study design. The putative advantage of this study design is that it is supposed to give an answer with fewer patients. I have read several articles on this study design, but I’m still skeptical. I have yet to see a Phase III study, irrespective of the indication, that confirms the Phase II data of this design. I hope some reader who is better than me in statistics will comment on this below and educate U.S. all. Also in the design, were the inclusion and exclusion criteria the same in both studies? For example, in Phase II, only citalopram failures or partial responders were included whereas in the R3, failures could have been with any SSRI or SNRI. The populations, strictly speaking, are not the same.
3. MECHANISM OF ACTION: A single negative study, even if in Phase III, may not kill new mechanisms of action, but it will make it much harder to interest big pharma to go after novel targets. It will further decrease an already depressed CNS focus in big pharma, with many companies exiting this therapeutic area. For a small company with CNS focus, the job just got harder. An additional point about MOA – were there genetic differences that can explain the differences? Indians are genetically not that different from Caucasians, especially Germans (bet that surprises a lot of readers), but it should still be considered since this study was conducted in Poland, Baltic states, Finland, Czech Republic, France and Sweden in addition to Germany.
The other Phase III studies have already started. The data should be intriguing.
Disclosure: I have no positions in any stocks mentioned, and no plans to initiate any positions within the next 72 hours.