Last week, Sanofi-Aventis (SNY) announced disappointing results from a phase III trial evaluating iniparib in breast cancer. The drug failed to improve survival and progression-free survival (PFS) in breast cancer patients and although actual data were not published, approval is unlikely even for a subset of patients. Failed phase III trials are quite common in oncology, a field with one of the highest attrition rates in the pharmaceutical industry. Nevertheless, iniparib’s failure is particularly disturbing, as the phase III was supported by compelling results from a randomized controlled phase II trial as well as strong scientific rationale. Importantly, this trial could have broader implications as it raises questions regarding the role of randomized phase II trials as a go/no go decision point for pivotal trials.
The weakest link
Randomized phase II trials have become all the rage, hailed as the cure to the frustratingly low success rate in phase III cancer trials. Traditionally, oncology drugs have been evaluated in three phases. Phase I studies were typically dose escalation studies for evaluating safety and initial signs of efficacy in 20-30 patients. Phase II studies were intended for fine tuning the dosing regimen and obtaining a clear efficacy signal. These single arm trials usually included 40-70 patients, all of whom received the drug. If the phase II data looked promising compared to historic controls, then a phase III trial would be initiated. Phase III trials typically accrue 400-800 patients (depending on the indication and statistical assumptions) and included a control arm on top of the active arm. The control arm can be an approved treatment or placebo.
As the number of approved treatments and the efficacy bar for obtaining regulatory approval got higher, it became harder and harder to generate positive results in phase III studies. It also became apparent that single arm phase II trials often overstate the drug’s efficacy, as drugs performed poorly in phase III trials compared to the phase II results. This was partially attributed to bias in patient selection as patients with favorable prognosis were often picked. In addition, the control arm in phase III trials often achieved better results than past trials, making it harder to beat. Thus, the industry identified single arm phase II trials as the weakest link, as they simply did not deliver in terms of predicting phase III outcome.
In the early 00’s, a new model was gaining momentum: randomized controlled phase II trials. The idea was to make phase II more difficult to cross by adding a control arm and increasing the number of patients to 80-150. Patients were to be randomized to receive either the investigational agent combined with standard of care or standard of care alone. This design was hoped to eliminate the patient selection bias as well as provide a real life reliable control arm. In other words, oncology phase II’s turned into mini phase III trials. Of course, the cost doubled or tripled, but the ability of distinguishing promising drugs from those that are doomed to fail seemed to more than compensate.
Although it is too early to reach a verdict regarding randomized phase II trials in oncology, so far they haven’t panned out as expected in terms of predicting phase III outcomes. The table below includes some high profile drugs with good performance in randomized phase II trials that was not corroborated in phase III. In many cases, the phase II trials enabled small companies to sign lucrative partnership deals such as Antisoma’s $890M deal with Novartis (NVS) or Synta’s (SNTA) $965M deal with GSK (GSK).
Agents with positive phase II that eventually failed phase III
Of note, in most cases, the phase II randomized trials did not exactly mirror phase III studies in several aspects. To begin with, phase II suffer from the inherent problem of being too small to produce a real overall survival signal, which is the primary endpoint in many phase III trials. Instead, they are powered to show a statistically significant difference in response rate and PFS. This was the case in the elesclomol and AS404 studies, which showed an impressive numeric difference in OS which was not statistically significant.
In addition, the phase II trials often did not mirror the phase III’s in terms of patient population and dosing regimen. Pfizer excluded a certain histology subtype from figitumumab’s phase III trial, as opposed to the phase II. Vandetanib’s phase II trial included two doses in combination with chemotherapy, only one of which demonstrated an improvement in PFS. BiPar decided to enroll both treatment naïve and relapsed patients in the phase III whereas the phase II included only relapsed patients.
Another issue with randomized phase II trials is that they typically add the investigational agent on top of an existing regimen. This makes it hard to tell whether the efficacy signal in the combination arm is a result of adding the investigational agent or merely the presence of good responders to the standard of care.
The value of randomized phase II trials is a matter of active debate. For those interested in delving deeper into the different arguments an excellent article titled “Randomized Phase II Trials: Inevitable or Inadvisable?” was published last year in the May issue of Journal of Clinical Oncology and included some interesting comments as well.
Iniparib’s phase III failure demonstrates that the industry has a lot to learn about randomized phase II studies, from trial design to the decision of starting pivotal studies based on their results. Although at the moment success rates are frustratingly low, one cannot deny the fact that randomized phase II probably identified compounds that were unlikely to succeed in phase III even though this can never be proven.
It is clearly still too early to throw in the towel on randomized phase II’s, as hopefully the experience gained so far will enable the design of more reliable studies. One important factor would be incorporating biomarkers for patient selection in order to better define the desired patient population.
In that sense, Endocyte (ECYT), which is expected to complete its IPO this week, is a great example of using a robust biomarker in phase II and designing a pivotal trial accordingly. At last year’s ASCO, Endocyte reported phenomenal data in ovarian cancer in a fairly large randomized phase II study. The company’s innovative approach includes injecting a diagnostic agent to patients prior to treatment. This provides valuable data that enabled Endocyte to accurately characterize patients’ tumors and select those who are very likely to derive benefit from its drug. Consequently, the company is currently conducting a relatively small pivotal trial with improved likelihood of success, at least in a subset of patients.
For investors who are looking to avoid the nerve wrecking dependence on randomized phase II trials data, there are agents which can reach proof of concept and even approval without randomized studies. Typically, these agents are so effective that they can be given as mono-therapy and demonstrate clear cut activity, especially in biomarker-defined populations. In some cases, they might even become approved based on single arm studies. Examples for such agents include Roche/Plexxikon’s PLX4032, Pfizer’s (PFE) crizotinib, Exelixis’ (EXEL) XL184, Seattle Genetics’ (SGEN) SGN-35 (discussed here), Roche/Immunogen’s (IMGN) T-DM1 and Incyte’s (INCY) INCB424 (discussed here).
Scarcity of late stage assets
Many expect Sanofi to in-license another late stage oncology drug to fill in the gap created by iniparib’s failure. Unfortunately, the past couple of years left only few available attractive late stage oncology assets with proof of concept. One unpartnered agent with a fairly high likelihood of reaching the market by 2013 is Exelixis’s XL184. As I previously discussed, this drug could represent a true revolution in the way bone metastases are treated across a variety of indications. Nevertheless, published data is still limited and investors are anxiously awaiting an update later this month from the ASCO GU meeting to see whether the bone effect is real and durable.