David's New Book

Sunday, August 12, 2012

Rebooting Hospital Acquired Pneumonia at FDA.


As I mentioned in my last blog, I was recently invited to participate in another meeting with the FDA at the Brookings Institution.  At the last meeting (summarized in a previous blog) Janet Woodcock made the extraordinary admission that the FDA would have to reboot in order to get antibiotic development out of the cellar in the US. 

In this installment, I would like to explore a feasible way forward for hospital acquired and ventilator associated pneumonia.  The FDA guidance for these indications currently suggests designs where the endpoint is mortality and where patients are precluded from receiving prior antibiotics.  These designs have been discussed several times in this blog and are completely infeasible and in many ways not at all real world (over 80% of ICU patients receive antibiotics for e.g.).

Ventilator-associated pneumonia is a particularly problematic indication because of great controversy around diagnostic accuracy and the fact that the disease incidence is shrinking making these patients difficult to find for enrollment in clinical trials.  Adding to the difficulty are US rules suggesting that nosocomial infections might not be reimbursed by Medicare that, in turn, provide a disincentive to actually make an official diagnosis in the chart of hospital-acquired pneumonia.

Many studies suggest that antibiotics for this indication have a treatment effect of 40-60%. When looking at mortality, but more importantly for our purposes, even when looking at clinical outcome via pharmcometrics these numbers hold up. In spite of this, the FDA has proposed a 10% non-inferiority margin within the microbiologically documented population and the EMA proposes a 12.5% margin. The patient numbers required by these margins, even when looking at clinical outcome as an endpoint, may be difficult to achieve today. But the treatment effect numbers suggest that in a clinical trial setting, we can think about non-inferiority margins of 15-20% that still retain at least 50% of the treatment effect. This range of margin would almost certainly bring these trials within the range of feasible patient numbers.

So here is a proposed design with trial numbers that might be required (note that for a combined HAP/VAP trial at least 30% of patients would be required to have VAP).

Patient population.



Patients would be allowed up to 24 hours of prior antibiotic within 72 hours of enrollment in the trail (as per EMA). Patients who had failed prior therapy (predefined in protocol) would be allowed to enroll regardless of the time course of the prior failing antibiotic.

Endpoint - clinical response at test of cure. 

Analysis population – ITT and modified ITT (per protocol).

Non-inferiority margin – 17%.  For a trial with a 60% treatment success rate, 90% power  and 70% evaluability rate, 500 patients would be required per trial or 1000 for two trials.

Option for a single trial – using support from a previous trial in severely ill (PORT III-IV) trial in community acquired pneumonia, a single trial in HAP/VAP would suffice for approval. For a 17% NI margin as noted above but at 80% power, the trial population would be 750 patients total.

This proposal harmonizes, to a large extent, the FDA and EMA proposed addendum.  I have increased the non-inferiority margin above that suggested by EMA to account for both the very large treatment effect of antibiotics for this indication including that seen for an endpoint of clinical outcome, and to allow for increased feasibility at a time when the disease incidence and therefore patient availability are decreasing.