Sunday, November 6, 2011
The FDA AIDAC on Pneumonia Nov 3-4.
I usually try and limit my blogs to 800 words or less. That’s just not happening today – sorry. I attended the recent FDA Anti-infectives Drug Advisory Committee (fondly known as AIDAC) meetings on community-acquired (CABP) and hospital-acquired or ventilator-associated pneumonia (CABP). The FDA briefing materials, can be found by following these links - (CABP), (HABP/VABP). It is clear from the briefing materials and the meeting itself that the FDA wants to make their trial design requirements feasible. For CABP they are getting there. For oral antibiotics for CABP - who knows? For HABP/VABP we have a long way to go.
The discussion was extensive and took the entire day. There were notable presentations from Tom File, IDSA (John Bartlett) and PhRMA (Barry Eisenstein).
Dr. Sumati Nambiar began the day with a nice review of 4 years of FDA discussions and guidance documents in this area including comments to the docket. Basically, by reviewing the preantibiotic era literature and through work with FNIH, the FDA was able to convince itself that an NI margin based on early clinical outcome (symptoms only) could be justified since the treatment effect comparing sulfonamides to placebo was so dramatic. The FNIH process defined an early endpoint of symptom improvement at study day 4 (72 hours of therapy) as a one point improvement in at least two symptoms of cough, dyspnea, pleuritic chest pain, sputum production AND no worsening of other symptoms. An absence of elevated body temperature and improvement in important measures of physiological stability would be important but not part of the primary endpoint. Likewise, the need for a later assessment to assure long-term response was recognized but only included as an “important” secondary endpoint.
The FDA, based mainly on daptomycin data as published by Pertel et al, and partly based on a potential masking of ceftaroline superiority to ceftriaxone in the FOCUS studies, believes that prior antibiotics, even a single dose of a short acting drug, might confound a non-inferiority trial and drive the results towards the null - hence the importance of the question on prior antibiotic use.
Presentations by John Bartlett, Tom File and PhRMA made the strong argument that trials would be difficult to enroll in general, but impossible to enroll in the US if all prior antibiotic use was precluded. In Forest’s FOCUS 1 and 2 trials only 2% of all patients came from the US. I presented the recent experience of a phase II study of an IV-oral antibiotic for CABP where most US centers refused to participate because of the exclusion of all prior antibiotic use. Even overseas this was a problem. The study was carried out in the US, Canada, Hungary and Poland. Of 860 patients screened, only 32 were enrolled. 220 failed to enroll because of prior antibiotic use. The rest of the screening failures were a mixture of causes with the most common being that the patients were not felt to have pneumonia after evaluation or refused consent.
The discussion of FDA’s question 1 centered around whether the symptom assessment at the early endpoint was sufficiently objective. Many physicians wanted to include signs such as temperature, pulse and respiratory rate as part of the assessment as noted in FDA’s option 2. The usual objection from Tom Fleming was that these are biomarkers. But surprisingly, a few clinicians noted that they thought the pulse and respiratory rates recorded on hospital wards were highly unreliable and that getting temperature measurements more than 2-3 times per day would require an act of God. Conclusion – physiological signs are not necessary but could be used at the sponsor’s option. From my point of view this is good since option 2 is infeasible just based on trial size alone. Option 1 would require 1376 subjects if powered at 80% for the overall program and therefore 90% for each trial. This is quite feasible.
As noted in the FDA’s options, the NI margin for clinical outcome was proposed as 10% but they allowed pooling of the microbiologically documented population for two trials in options 1 and 2 and proposed a margin of 15% for that analysis. Both would have to be met to achieve approval. In the single trial option (FDA’s option 3), they proposed several margins where the primary analysis population is the microbiologically documented one. An 80% powered trial at a 10% NI margin would require 1862 patients – probably infeasible. A 12.5-15% NI margin would require 1192 and 828 respectively at 80% power. These are feasible. The committee suggested a 10% margin for the single trial.
A further discussion of the margins led to questioning of the entire rationale for the early endpoint. This exists because the FDA could find historical data to justify a treatment effect. I noted that one could easily do ths is the context of previous prospective, controlled clinical trials using a pharmacometric approach as noted in previous blogs. The thought is to use pharmacometrics to define an NI margin for the more clinically relevant test of cure endpoint that has been used traditionally and is still used in Europe as an endpoint. The topic was taken up by John Rex and others and became even more important during the discussion of HABP/VABP on Nov. 4. The pharmcometric method has been shown to be robust and applicable across many indications and many antibiotic classes. I remain baffled as to why the FDA continually refuses to look at this option.
The most contentious discussion of the day occurred over whether prior antibiotics would be allowed or not. The basic tension was between whether one wanted to have any US patients enrolled in a trial or not. John Bartlett even raised the ethics of conducting a trial overseas where, if we know that delaying antibiotics in seriously ill CABP patients increases mortality, than those patients should not be subjected to that risk either by the time consuming enrollment procedures. Thus, he argues that they too should be allowed a single dose of a short acting antibiotic and that it might not be ethical to disallow that for overseas patients. I found this a powerful argument – and apparently so did at least some committee members. About half the committee agreed that a single dose of a short acting antibiotic would be acceptable. They requested that the FDA provide a list of acceptable antibiotics. They did not address the use of concomitant therapy for Legionella like the clarithromycin that was allowed in one of the ceftaroline trials since that is required by guidance in North America.
The FDA proposals for feasible trials are shown below. Again – see their briefing materials and the previous blog on this topic!
There was a clear consensus among all present that this indication is the one with the highest unmet medical need and the one with the least feasible regulatory path forward (except those where placebo trials are required like otitis, sinusitis and bronchitis in my view). The key to understanding the issues here is a full comprehension of the implications of the 28-day all cause mortality endpoint that the FDA requires. I noted that in their questions, they do not ask the committee if they agree with this endpoint. Nevertheless, this became a central area of discussion during the day.
A discussion of the statistical considerations in trial design was presented by Dr. Komo of the FDA. He reviewed a large number of trials in HABP and VABP. The microbiological evaluability rates for VABP ranged from 72 to 84% for trials of broad-spectrum agents and 56-63% for those studying Gram positive pathogens. For HABP, the range was from 50-63% for broad-spectrum agents and from 45-53% for Gram positive pathogens.
Dr. Komo followed this with a discussion of the determination of M1 and the NI margin. He reviewed the FDA discounting methods but both he and later Tom Fleming noted that if mortality is the endpoint and the control mortality rate is set at 20%, anything larger than a 10% margin might lead to potentially unacceptable increases in mortality for the treatment arm in any case. Therefore, using a mortality endpoint, no increase in margin would be permissible anyway purely on clinical grounds.
Dr. Komo then reviewed the Risk Difference, Relative Risk and Odds Ratio methods of measuring risk of mortality based on NI trial data. Since I am NOT a statistician, I only have an elementary understanding of these methods. But, what I clearly understand based on the presentation and the FDA briefing materials is that with the OR method, as the mortality rate goes down – which seems likely given current trends – the sample size required increases asymptotically!
There were excellent presentations by Don Craven (IDSA), Tom File and Robert Fromtling (of Merck for PhRMA). Of these, I will focus on the latter. The PhRMA proposal was for a composite endpoint where cure at TOC and alive at 28 days would be required for success. The suggested analysis would use an odds ratio of 1.714. If the FDA insists on a mortality endpoint, PhRMA prefers the approach the FDA took for CABP where there would be a margin of 10% for the ITT population and a margin of 15% for the microbiologically evaluable population. The study would be powered based on a risk difference approach (where the trial size actually decreases with decreasing mortality). PhRMA also pointed out that in the current assumptions by the FDA, drugs which focus on Pseudomonas would be excluded from the non-inferiority approach since the microbiologic evaluability would only be around 10%. They even questioned whether S. aureus infections would be feasible to study even under the proposed PhRMA standards. A final point from PhRMA related to the key issue of setting breakpoint – never discussed again while I was present. They pointed out that if breakpoints were set at the MIC50, this would dissuade virtually all companies from proceeding with anti-infective discovery and development.
During my open public hearing presentation I once again raised the idea of using pharmacometrics to define a treatment effect for a clinical endpoint. This is a much more critical issue for nosocomial pneumonia than for CABP since the FDA was able to finally get comfortable with a clinical endpoint there. To make trials feasible, we have to get away from mortality as the sole endpoint. John Rex raised this topic during the discussion. He hammered the point that this discussion was critical to whether or not antibiotics would eve be available to Americans or not and to whether companies would want to begin, restart or continue work in the antibiotic field. While I believe that many on the AIDAC were deaf to the urgency of his appeal, they did finally get the message of using pharmacometrics. This was discussed by enough committee members that I think even the FDA might have gotten the message. In my view, this is the only way forward here.
For the committee response – most agreed that a single trial would be OK and a significant number agreed that the single trial could be of both HABP and VABP patients. Again – if mortality is the endpoint and the mortality of the control must be 20% to maintain trial size at a reasonable level - this is going to be extremely hard to achieve, plus the microbiologically evaluable rate will decrease. Nevertheless, this represents an extremely important option for companies. Unfortunately, most on the committee agreed with the use of the odds ratio approach which really punishes sponsors for going under 20% mortality at a time when both the incidence of this disease and its mortality are falling.
Finally, there was a discussion of the use of prior antibiotics. The consensus seemed to be that there was little or no evidence that such treatment confounded trial interpretation and therefore that 24 hours or possibly even longer use might be tolerated. There was no real resolution of the issue of concomitant therapy such as aminoglycoside use that was common in previous trails and is frequently required by clinical guidelines. This will likely remain a thorny issue – but one which, might, once again be susceptible to interrogation using PK/PD methods.