FDA Panel Sees Role For PK/PD Modeling In Revising Antibiotic “Breakpoints”

In certain circumstances it may be appropriate to change the criteria for determining when a pathogen is susceptible to a drug and recommended doses for systemic antibacterials absent clinical outcomes data, Anti-Infective Drugs Advisory Committee says.

Pharmacokinetic and pharmacodynamic data simulations could assist in revising “breakpoints” and dosing recommendations for previously approved systemic antibacterials, but reliance on such an approach absent clinical outcomes data may hinge on a host of factors, including the source of the data, the drug’s toxicity profile and the type of infection, an FDA panel suggested.

FDA convened the Anti-Infective Drugs Advisory Committee Oct. 17 to discuss the role of PK/PD data in setting susceptibility interpretive criteria, or breakpoints, in labeling. The agency also sought panel input on revising dosing recommendations based on PK/PD, clinical safety and efficacy data.

To spur discussion on issues encountered in updating breakpoints for previously approved drugs, FDA devised two hypothetical scenarios and asked the panel to consider various breakpoint modifications, dosing and labeling changes based, in part, on PK/PD data generated through modeling simulations (Also see "FDA Panel To Discuss “Breakpoint” Changes For Systemic Antibiotics" - Pink Sheet, 15 Oct, 2013.).

The committee gave a mixed verdict on use of simulated PK/PD data to modify breakpoints for systemic antibiotics.

In one scenario, the panel narrowly voted that it would be acceptable to recommend a higher dose than was clinically studied for a particular indication to ensure a uniform breakpoint across indications for a given pathogen. In the other example, the committee overwhelmingly opposed setting different breakpoints for a specific pathogen for each of three approved indications and dosing regimens, believing such an approach would complicate laboratories’ reporting of susceptibility information and create confusion in clinical practice.

In the process of commenting on labeling options for the two hypothetical scenarios, the committee gave a mixed verdict on use of simulated PK/PD data to modify breakpoints for systemic antibiotics. Panelists indicated they were more comfortable relying on such simulations when they were based on data from patients with infections, rather than healthy subjects.

FDA is expected to use the panel’s advice and guidance in deciding whether, in resetting breakpoints, it is appropriate to extrapolate efficacy from PK/PD data when there is other evidence that pathogens have developed resistance to a particular drug and actual clinical outcomes data are lacking.

Simulated PK/PD data “looks like … something that clearly adds value and one that should be easily used by both the FDA and sponsors coming to the FDA to assist them in creating the label and the best intended use of these products,” committee member Alan Magill, Bill and Melinda Gates Foundation, said.

“Having said that, I also take very near and dear to the heart some of the comments also made about the desire to see clinical outcomes. I think clearly we’d like to see clinical outcomes,” he said. “It’s up to the sponsor to come [to FDA] with why this makes scientific sense to use PK/PD modeling in this indication, this dose, and such.”

Challenges In Revising Breakpoints

Breakpoints are the concentrations of a drug at which bacteria are considered to be clinically susceptible, intermediately susceptible or resistant to treatment. They are used in testing patient samples to identify appropriate therapies but are periodically revised to reflect increased bacterial resistance to antibiotics.

Breakpoints set at the time of initial drug approval are based on several sources of information: in vitro microbiology data for distribution of minimum inhibitory concentrations (MICs) or zone diameters for the drug against recent clinical isolates of the target pathogens; data from animal models of infection including PK/PD information; and correlation of the MIC or zone diameter with clinical and microbiological outcomes from clinical trials.

Under the FDA Amendments Act of 2007, the agency is required to identify and periodically update breakpoints for antibacterial drug products and to make those findings publicly available (Also see "Antibiotic Trial Standards To Be Revised By FDA Under Congressional Mandate" - Pink Sheet, 15 Oct, 2007.). FDA laid out its approach to updating breakpoints in a June 2008 draft guidance (Also see "Drug Makers Face Higher Relabeling Burden For Antibacterial Resistance Info" - Pink Sheet, 7 Jul, 2008.). The guidance was finalized a year later.

However, FDA has encountered various scientific challenges related to the types of data that can be used to support breakpoint revisions and the quantitative and qualitative limitations of such data.

Labeling for some antibiotics reflects different approved dosing regimens for different indications; different regimens potentially could lead to different breakpoints for each indication. Various sources of data may support increasing the recommended dose for one use to make it consistent with higher dosing regimens approved for other labeled indications, and “consistency in the dosing regimens could also lead to similar susceptibility test interpretive criteria for a specific bacterium across the approved indications,” FDA stated.

In the briefing document and at the meeting, FDA explained how breakpoints are currently set at the time of approval and potential options for revising them before putting the specific scenarios before the committee

When an antibacterial agent is first approved, clinical efficacy data are generally available from randomized trials at the doses studied, although safety data at higher doses may be available if the drug is approved for another indication at a higher dose.

“In the setting of increasing MICs for a drug, this raises the question whether the labeling should be revised to recommend a higher dose for all indications in the absence of any or limited clinical efficacy data for that particular indication at the higher dose,” FDA said. When considering a higher dose, it generally would be expected that efficacy will be preserved, although safety and the associated benefit/risk of the higher dose must be considered.

“A second option would be to base the susceptibility test interpretive criteria on the lowest approved dose, which would result in lowering the susceptibility test interpretive criteria such that a greater proportion of pathogens may be categorized as resistant to the particular antibacterial drug,” the briefing document states. “An undesired consequence of such a change would be the earlier loss of a therapeutic option and increased use of alternative antibacterial drugs.”

A third option would be to have multiple breakpoints for a pathogen, each specific to an indication and supported by an approved dosing regimen in labeling.

Third-Party Presentations

The committee heard presentations from several professional and industry groups that advocated use of PK/PD modeling simulations to support revising breakpoints and dosing regimens.

Cubist Pharmaceuticals Inc. Director of Global Medical Affairs Strategy John Mohr, who spoke on behalf of the Pharmaceutical Research and Manufacturers of America, said establishment and revision of breakpoints should be based on the totality of evidence, including the combination of PK/PD, clinical and microbiological data. PK/PD methodologies have evolved over the last 25 years, allowing for improved understanding of the link between drug exposure and drug effect, he said.

AstraZeneca PLC VP and Head of Infection John Rex spoke on behalf of the Clinical and Laboratory Standards Institute, which is involved in setting and revising breakpoints. He said that within limits of approved dosages and exposures, recommended dosages for specific uses can be adjusted based upon PK/PD considerations and modeling.

“As long as higher doses have been studied for safety, it’s entirely rational to encourage their use for specific situations, even if not specifically studied for that indication at that dosage,” Rex said. Furthermore, using a higher dose where justified by new insight is good stewardship, he said, because using suboptimal doses will drive both resistance and clinical failure.

The committee also heard from Bill Brasso, a staff scientist at BD Diagnostic Systems and president of the Susceptibility Testing Manufacturers Association, which represents companies that provide antimicrobial susceptibility testing (AST) results on cultured microbiology isolates. Brasso discussed operational hurdles to adoption of indication- and dosing-specific breakpoints.

While some testing systems can manage multiple breakpoints for a single drug, others cannot. Furthermore, labs rarely receive information regarding the indication that is being treated; a specimen is usually labeled only with a source, such as urine, blood or wound. “Using three separate breakpoints would require the lab and the AST system that you have in the lab to provide separate SIR [susceptible, intermediate or resistant] interpretations for each possible indication in case the patient happens to have one,” Brasso said.

During the open public hearing, two representatives from the National Physicians Alliance spoke to oppose reliance on PK/PD modeling in setting or revising breakpoints absent confirmatory clinical data.

“Experimental data obtained in vitro or in animals is invaluable, and PK/PD modeling is useful to generate hypotheses, but these must be validated in people, said Lisa Plymate, a Seattle practitioner. She asked the committee to consider whether there is “clinical evidence that changing sensitivity breakpoints will help patients in meaningful ways, or might changes simply lead clinicians to select newer, more expensive antibiotics that might in fact be less effective, or more toxic. My patients are not test tubes, they are not immune-deficient leukopenic mice, nor can their responses to medications always be predicted by [PK/PD] modeling. Our reference standard has to be what happens in people.”

Revising The Recommended Dose

FDA presented the committee with the two hypothetical scenarios to frame discussion of the challenges faced in revising breakpoints. FDA based the scenarios on data that were masked or modified to de-identify the drugs at issue. Notably, both drugs were deemed to have relatively benign safety profiles.

In the first scenario, a drug initially approved in 1998 carries three indications with distinct dosing regimens – acute bacterial skin and skin structure infections (ABSSSI) (dosed 1 g every 8 hours), hospital-acquired bacterial pneumonia (HABP) (2 g every eight hours) and osteomyelitis (4 g every eight hours). Pseudomonas aeruginosa is listed as a pathogen in the ABSSSI and HABP indications. The current breakpoint of 8 mcg/mL was based upon the HABP dosing regimen.

A change in the drug’s breakpoint for P. aeruginosa is being considered based on a shift in MIC distribution, an increase in the proportion of isolates expressing resistance at MICs <8 mcg/mL and supportive PK/PD data. With regard to the latter, a population PK model was developing using data from about 100 patients treated with the drug, and the PK model simulated a larger population of 8,000 subjects to determine target attainment, which is the frequency with which the desired drug exposure is attained.

The committee voted 10-8 that it would be acceptable to set the P. aeruginosa breakpoint for the drug at 4 mcg/mL based on the 2 g q8h dosing regimen and to recommend this higher dosing regimen for ABSSSI despite the lack of clinical efficacy and safety data in this indication. However, the vote’s narrow margin suggested no clear mandate for this approach, with some panelists in the minority favoring separate breakpoints for ABSSSI and HABP based upon the currently approved dosing regimens.

Committee member Luis Ostrosky, University of Texas Medical School, favored doubling the dose for ABSSSI and rejected concerns that such an approach amounts to “dose blasting.”

“This is a very well thought out dose increase,” he said. “It’s not even the highest dose. It’s a studied dose and safe, and it recognizes a shift in reality that we’re seeing.”

Committee member Michael Neely, University of Southern California, also voted in the majority. “There really is more of a risk for this particular drug, since we were given a relatively benign safety profile, of under-dosing, especially up front,” he said. “I think when you’ve got critically ill patients you really want to get not only the drug but the dose right, and for a drug that is relatively non-toxic I think we want to err on the higher side.”

However, temporary voting member Arjun Srinivasan, Centers for Disease Control and Prevention, disagreed with this approach. “I want to see flexibility for clinicians,” he said. “I don’t want to see situations where we box clinicians into one single dose for all indications. The drug was demonstrated to be effective in the treatment of skin and soft tissue infections at a dose of 1g q8 and now we’re contemplating changing the dose, doubling the dose of the drug because of slightly elevated resistance among a pathogen which we all agree almost never causes skin and soft tissue infections. So my question and my concern is how many patients are now going to be exposed to a doubled dose of the drug?”

Setting Dose-Specific Breakpoints

The second hypothetical drug was approved in the 1990s with different dosing regimens for three indications – ABSSI (0.5 g q12h), complicated intra-abdominal infections (CIAI) (1 g q8h) and HABP (2 g q8h) – with Escherichia coli listed as the pathogen for all three indications. The original breakpoint of 8 mcg/mL was set based upon the highest approved dose. However, there are no clinical efficacy or safety data available from randomized controlled trials of ABSSI or CIAI at that dose.

A population PK model was developed using healthy adult data from six Phase I studies, and the PK model simulated a larger population of 6,000 subjects to determine target attainment. This analysis, coupled with limited clinical data on drug failures and recent microbiologic surveillance data, support different breakpoints for each approved dosing regimen.

However, the committee voted 14-4 that it would not be acceptable to have different breakpoints for E. coli for each indication in labeling.

Temporary voting member Matthew Goetz, Veterans Affairs Greater Los Angeles Healthcare Systems, said he could not accept multiple breakpoints for multiple indications. “The principal driving reason was the complexity with getting that information to clinicians and the reliability that the laboratory would know exactly where the site came from,” he said.

Labeling drugs with breakpoints for a given pathogen specific to site of infection and dose “would be a nightmare,” panel member Ostrosky said.

Ostrosky said the idea of having breakpoints for a given pathogen that are specific to site of infection and dose “is very noble, and it would potentially be very useful to people, but it will have the result of paralyzing with over-information.” Such an approach potentially could be workable in five to 10 years as technology evolves, he said, “but today in October 2013, I think this would be a nightmare.”

Several panelists voting in the majority also said they had concerns about the reliability of the PK simulation results because they were based on data from healthy subjects, not patients with infections.

“I wasn’t sure I trusted the PK/PD simulations in this setting because I don’t know exactly what the target attainment rate is in these specific tissues,” Neely said. “If we had that sort of information and we could derive a breakpoint from that information, I would be much more interested in having multiple breakpoints.”

In keeping with his vote on the first hypothetical scenario, Srinivasan was among the minority of committee members who favored a multiple breakpoint approach on the second example. “This is the direction I would eventually like to see us head,” he said. “We may not be there today in terms of being able to implement something like this, but I really believe that we need to promote the use of more clinical information, not less, among clinicians who are treating infections and we need to increase the number of dosing options that people have to allow us to use more different drugs to treat different infections.”