FDA Workshop on Pathogen-Specific Antibiotic Development - Part 1

For the last two days, July 18-19, I was pleased to be able to attend a fascinating FDA Workshop on the development of antibiotics for use in a subset of patients with unmet needs. I am writing this blog while waiting for my very delayed flight from DC back home.

Unmet needs patients include, mostly, those with infections due to highly resistant pathogens where options for therapy are limited and/or where antibiotics for more usually resistant pathogens are not available in an oral formulation necessitating intravenous therapy. The subset of products specifically considered during this workshop is that targeting specific pathogens – e.g. an antibiotic active only against a single species of bacteria such as Pseudomonas aeruginosa or Acinetobacter baumanii. You can find all the slides used and other information here. I have blogged in the past regarding the FDA in particular and this thorny clinical development problem (1, 2, 3).

My impression in general is that, even if the FDA does not yet know what is the best approach to the development of pathogen-specific antibiotics, they wanted a public discussion of the issue to highlight both the need for these agents to be developed and the consequential need for a feasible pathway for their development.  They also wanted to be able to point to a public discussion of the extreme difficulty of designing and carrying out a standard trial with inferential statistics in this setting.  This public understanding would then make it just a little more comfortable for them to venture into heretofore unexplored territory in antibiotic development.

The meeting started with the usual general statements both from FDA and from EMA, the European regulatory agency. These were helpful and only emphasized the additional tools available in Europe for the approval of such agents compared to the US.  But, at least on day 1 of the meeting, the most important presentations, I thought, were the real life experiences of two companies attempting such trials. There were questions on the rationale for designing and carrying out the trials - I will review this below.

Example 1 – the CARE trial by Achaogen – presented by Ian Friedland. Achaogen is developing a new aminoglycoside antibiotic with activity against resistant pathogens called Plazomicin. Their original plan was to carry out a randomized superiority trial in infections caused by highly resistant pathogens –a noble but ultimately doomed endeavor. The original design called for the study of bloodstream infection and nosocomial pneumonia caused by carbapenem-resistant Gram negative pathogens. The endpoint was all cause mortality at day 28 (because the FDA requires this for non-inferiority nosocomial pneumonia trials?). The plan called for a 1:1 randomization between plazomicin + meropenem or tigecycline vs. colistin + meropenem or tigecycline calling for the enrollment of 360 evaluable patients (assumed 80% evaluability). Note the use of combination therapy – thought to be essential in these desperately ill patients with complicated infections. The trial initially screened 694 patients to enroll 14. Screening failures were caused by the lack of a carbapenem-resistant pathogen or the presence of more than 72 hours of previous antibiotic therapy (excluded by protocol) in these very sick patients in intensive care units (no surprise to most of us here).  The company quickly decided to carry out a standard, non-inferiority trial in urinary tract infection and altered their CARE trial design to be more accommodating to enrollment. Having started the original trial in early 2014, they project an enrollment of 100 patients by the end of this year – still far from their original goal of 360 patients. Enrollment appears to be speeding up slightly after their study amendments – but not enough to make this trial feasible as a stand-alone pivotal study for registration under today’s FDA standards. At this point the company plans to use the data in this smaller number of patients to support an application for approval based on the urinary tract infection trial data.

Example 2 – TANGO trials of meropenem plus a new beta-lactamase inhibitor, vaborbactam, from the Medicines Company presented by Mike Dudley. TANGO I was a standard, non-inferiority trial in urinary tract infection and the top line data has already been presented. They actually showed superiority (just) over their comparator, piperacillin-tazobactam.  Would meropenem have done the same without vaborbactam? One must ask . . . 

The company also is carrying out a TANGO 2 trial of meropnem-vaborbactam vs. best available therapy in urinary tract infection, intra-abdominal infection, nosocomial pneumonia and bacteremia suspected to be caused by carbapenem- resistant Gram negative pathogens. The trial is a comparative superiority trial with a 2:1 randomization. To support the control data for their trial, they carried out a retrospective study to gather data on the efficacy of best available therapy in regions where they would carry out the prospective clinical trial. The results of this retrospective study were surprising in that the patients were very sick and the mortality rates were high with a surprising 18% 28 day all cause mortality rate in patients with urinary tract infection. They also discovered an astounding 69 different best available therapy regimes in the different centers studied and they noted that combination therapy seemed not to improve outcome over monotherapy. They amended their trial protocol based on the findings of their retrospective study. They note that this trial will probably not yield inferential data.

Paul Ambrose raised the following question. If we can’t do inferential studies in these so-called pathogen specific indications, why don’t we just rely on PK/PD data to show that therapy is feasible and likely to be efficacious? While scientifically, one can’t argue with this position, the retort from companies is that physicians want to see actual clinical data in the patient population they will treat.  But does non-inferential data count? Can we not, rather, educate physicians to understand the importance of scientific basis of the PK/PD argument?

Day 2 was even more interesting.  Hold on to your hats!