FDA and Pathogen Specific Antibiotics

As I promised in my blog on FDA, Roche and Polyphor, I met by telephone with the FDA today (Ed Cox, Sumathi Nambiar).   I began by reminding them of my experience at IDSA at the origin of the last blog.  Specifically, many people both from academia and from industry were complaining that there was no existing pathway for approval of a pathogen specific drug at FDA currently.  Various folks said that the FDA was insisting on some “inferential” (read statistically significant) data to support an approval for a new antibiotic.  But we all know that for rare pathogens like Pseudomonas aeruginosa and Acinetobacter baumanni, there are not enough patient numbers to fulfill this requirement.

For example, lets say that you have a new antibiotic active only against Acinetobacter and no other bacteria. To show that your drug provides an advantage of currently available therapy, you target patients likely to have infections caused by resistant Acinetobacter.  Such infections might be pneumonia, urinary tract or skin infections. As such, you will probably have to treat at least 300 patients to end up with about 30 resistant Acineotbacter infections in your trial.  Such a trial will take at least 18-24 months (the longest reasonable time to spend on a enrolling a trial). And, lets say that standard therapy of patients with pneumonia, urinary tract infection and skin infections provides a clinical response in about 50% of such patients. You hope that your drug will bring the response rate up to 70-80% for a 20-30% improvement over standard therapy. We didn’t even get into the idea of stratifying patients by site of infection – but I’m not sure it would be necessary. To power the study to achieve statistical significance would require studying 100-250 patients with the desired infections – clearly not feasible in a two-year period – to say nothing of the expense involved.

It is clear to me, and to FDA, that these trials will be externally controlled.  Historical controls like retrospective but contemporaneous chart reviews or prospective observational studies where inclusion and exclusion criteria are similar to those used to enroll patients in the treatment trial would be acceptable.  Another approach would be to use pharmacometric data (I think) if robust data were available. The FDA agreed that external controls are OK.  This is not the problem for them.

In general, the FDA understands that they do not have a clear pathway at the moment for such drugs. They have always preferred to have enough data such that statistical significance at P less than 0.05 in a two-sided test was possible. They understand that for these pathogens, this will not be possible in any reasonable time frame.  What they need is a way to approve such drugs given their current regulatory constraints.

I mentioned orphan drugs where I thought that there would be examples where non-significant data would be used for approvals.  Ed corrected me.  He had just studied many such approvals and found that the P values were highly significant for most of these drugs.  The diseases tend to be chronic.  Enrollment seems not to be a problem.  And the effect size is very large making the trial size requirement fairly small.

I suggested that for a pathogen specific antibiotic they accept studies not powered at the 0.05 level, but rather that they accept more statistical trends (P of less than 0.1 for example). This could only occur in the presence of a very substantial PK/PD package including a strong dose-justification rationale. In the case of such an approval, the label would specify use is approved only for those patients with few or no alternatives. There would be a post-approval commitment for the sponsor to continue gathering data to try and provide a more substantial inferential argument.  In the case that this was not achieved, that is, with greater numbers of patients the new therapy could not be shown to be statistically superior to standard of care, both the FDA and the sponsor would have to rethink. This solution is fallible since standard of care might change in the interim making continuing study difficult or even impossible. Both the FDA and the sponsor would have to be willing to accept this risk.

Ed reminded me that I had proposed something very similar back in 2005-6 during a project I carried out with the Manhattan Institute.  I had totally forgotten about this proposal, but it was very similar to what I have proposed above.

Ed asked if I thought that rapid diagnostics would provide a way forward here.  I said – not tomorrow.  For this to be helpful, the diagnostic would have to be “point of care” and CLIA-waived.  I don’t think we are anywhere near something like that for Gram-negative infections today.

I am sure that there are other ways to approach this problem as well.  Ed told me that the FDA has been working hard on this internally, and in fact had spent virtually the entire day today on the subject.  He also told me that he has presented this as a problem to the President’s Advisory Council on Combating Antibiotic Resistance and asked for their help here.

I hope for all our sakes that we find a way forward – and soon.  Pathogen specific drugs are coming and they will be good for the microbiome , good for patients and good for us as a society. Apparently the ball is in the court of the President’s Council for now . . . .