David's New Book

Friday, August 27, 2010

Guidance?

Pus oozing from an abscess caused by bacteria—...Image via Wikipedia

The long-promised Draft Guidance on Acute Bacterial Skin and Skin structure Infections (ABSSSI): Developing Drugs for Treatment has arrived (http://www.fda.gov/downloads/Drugs/GuidanceComplianceRegulatoryInformation/Guidances/UCM071185.pdf).  As discussed on several previous blogs, the basic trial design, with its requirement for cellulitis (75 sq. cm) for all infections studied (abscess, wounds and cellulitis) and its 48-72 hour endpoint of halting progression of the lesion and lack of fever seem feasible. One issue is that infections such as diabetic foot infections, commonly encountered by clinicians, are not encompassed within this guidance.  Any drug to be approved for ABSSSI that offers any sort of advantage over existing compounds, or even that is promoted when generics are not, will be used off label for many of the indications not covered in the guidance.  Is that what we want?  When I posed that question to the FDA back in march they responded with an emphatic “No”. 

Another issue, consistently raised by Brad Spellberg, is that the endpoint is clinically irrelevant.  Who cares if spread of the lesion is halted by 48-72 hours or not?  We all want to see a cured patient at the end of therapy or in the days or weeks that follow.    It is true that the FDA endpoint probably correlates with a relevant clinical endpoint, but why not use the relevant endpoint?  The answer is because the FDA is stuck in some political battle with statisticians who vociferously require we use endpoint established 75 years ago when sulfonamides were compared to UV light therapy for erysipelas.

The FDA also has sponsors conducting a number of other “studies” for them so they can gather data around skin infections – nothing wrong with that.  They suggest that sponsors develop an instrument for Patient Reported Outcomes for ABSSSI during Phase II.  Right!

But to me, the most upsetting part of the guidance is in their explanation of their approach to setting the non-inferiority margin.  This is the crucial statistical argument that will make the difference, based on the numbers of patients required for study, between feasibility or not.  In a previous blog, Discounting Antibiotics (http://antibiotics-theperfectstorm.blogspot.com/2010/08/discounting-antibiotics.html), I reiterated points made by the FDA at the August workshop where trial design for ABSSSI was discussed all of which are included in the new guidance.  Briefly –

·      They look at the difference between UV light and antibiotic treatment where UV treatment has an effect.  Therefore, the treatment difference (M1) they calculate is not with placebo but with another effective therapy thus underestimating M1.
·      They then calculate a 95% confidence interval and then take the worst number (lowest difference or poorest treatment effect) that would occur one time in 40 trials and use this (under)estimate of M1.
·      Then, wanting to be extra-conservative and desiring to cover possible differences in patient population and trial design from the 75 year old trials until now, they discount the treatment effect still further by some arbitrary percentage.  They assume arbitrarily that antibiotics today would perform worse than they did 75 years ago. But, in fact, both published data (Corwin, P, L Toop, G McGoech, et al., 2005, Randomized Controlled Trial of Intravenous 1356 Antibiotic Treatment for Cellulitis at Home Compared With Hospital, BMJ, 330(7483):129), data  from Pfizer and from Trius, presented at the FNIH meeting , all show that modern trials duplicate the 75 year old results of Snodgrass quite faithfully.
·      The above gets us from a 30-40% treatment effect down to the FDA’s estimate of a whopping 12%.
·      The non-inferiority margin cannot be greater than the treatment effect. Of course, the FDA has already reduced the treatment effect by using improbable numbers as shown above. So, the FDA then says, arbitrarily, that the margin cannot be more than 50% of the treatment effect.  This gets us to a non-feasible NI margin of 6%! But there is no particular scientific basis for 50%.  It could be 90% or 10%. 

But take heart everyone out there in antibiotics land.  The FDA has already indicated that they will be flexible in their discounting of M1 depending on arguments presented by sponsors.  And, in fact, Trius has already negotiated at 10% non-inferiority margin for their upcoming Phase III trials in ABSSSI. 

So that leaves with the question as to why the FDA constructed their whole non-inferiority section the way they did. The only explanation I can see is that it was done to satisfy some pressure from statisticians within and outside the agency who are all stuck on Snodgrass and the pre-antibiotic era trials.  But the only thing this approach, formalized within FDA guidance, will do is alienate and frighten sponsors and investors away from research and development in antibiotics. 

So, FDA, in our era of emerging superbugs, which is more important, your political standing with statisticians or the pursuit of public health?
Enhanced by Zemanta