The FDA reboot of antibiotic development.

In May 2012, a number of us listened spellbound as Janet Woodcock of the FDA announced at a meeting at the Brookings Institution that the agency would “reboot” their entire approach to antibiotic development (J. Woodcock, presented at an Expert Workshop on Facilitating Antibacterial Drug Development, 9 May 2012). She recognized that there is a public health crisis of antibiotic resistance that continues to grow worse and that development of new antibacterial agents to deal with the threat is inadequate. Finally, she acknowledged that the approach taken by the Office of Antimicrobials at FDA to change clinical trial designs for antibacterial agents over the past decade had contributed to this crisis. Dr. Woodcock went on to emphasize the need for a new pathway for development focusing on patients with unmet medical needs—that is those with infections due to pathogens that are pan-drug resistant (PDR) or that are extremely drug resistant (XDR) (1). Further discussion showed that the agency also understands that development for traditional indications such as pneumonia also needs “rebooting.” Their thoughts on the reboot follow the recommendations of a working group from the pharmaceutical industry (2). How and why did we get here? The first warning of the exodus of industry from antibacterial development was published as far back as 2002 (3). The change in industry interest in this area was due to a convergence of unattractive economics around antibacterial development (4–6) combined with a fundamental rethink regarding statistical principles of non-inferiority trial designs that began in the late 1990s (7–9). Although this rethink was not specific to antibacterial agents, it disproportionately affected antibacterial development (8). A dangerous inflection point occurred in 2006, after the public spectacle surrounding telithromycin, which was discovered to cause very rare but life-threatening hepatotoxicity only after the drug was approved (8–10). Ironically, this statistical rethink focusing entirely on proving efficacy (and doubting that antibiotics were effective) had been triggered by a safety problem with telithromycin (9). In the aftermath of telithromycin, FDA rules governing trial conduct became increasingly stringent to the point of making antibacterial trials infeasible, nonsensical, or both. For example, changes to trial design meant to increase the scientific purity of the trials (such as excluding any patients who receive even 1 dose of prestudy antibiotic from enrollment in a clinical trial) made it virtually impossible to enroll patients into the trials in the United States. Worries that antibiotics were no more effective than placebo were finally put to rest by the FDA’s own analysis of preantibiotic era data showing treatment effects consistently higher than 20% and frequently over 50% depending on the infection (11, 12). In spite of these large treatment effects, some of which would justify noninferiority margins greater than 20%, the FDA always “discounted” the treatment effect by a sufficient amount such that the resulting noninferiority margins were always 10% (lower margins higher sample sizes, more patients, and greater costs) (13). A new endpoint for skin infections was developed in which patients whose infections have not improved at all after 3 days of therapy are declared treatment successes merely because they stop getting worse (14, 15). This endpoint was convenient for statisticians because they could comfortably calculate a treatment effect of oral sulfonamide antibiotics versus placebo equivalent (UV lamp therapy) based on unverifiable data from two studies published in 1937 (in which the background therapy for skin infections consisted of a liquid diet and a mandatory hot-liquid paraffin soap-and-water enema) (16, 17). The endpoint used in these 80-year-old studies was cessation of spread of the skin lesion. Although this did provide a feasible way forward for companies desiring to develop antibiotics for skin infection, this endpoint, in our view, is invalid and has little clinical relevance (B. Spellberg, presented at an Expert Workshop on Facilitating Antibacterial Drug Development at the Brookings Institution, 9 May 2012 [http://www.brookings.edu/~/media/events/2012/5/09%20antib acterial%20drug%20development/panel%201%20brad%20spell berg%20presentation]). Clearly, the FDA process of determining how antibacterial trials should be conducted has badly lost its way. (For a general overview on this opinion, see the Spellberg presentation cited above.) There have been three results. (i) Many companies do not invest in the trials. (ii) Those that do invest in trials enroll patients in countries where it is possible to withhold therapy while the patients are enrolled in trials, with resulting ethical concerns. (iii) The results become less meaningful and relevant to patients in the United States, because U.S. patients are not enrolled. Pharmaceutical companies have voted with their feet. Twenty years ago, more than 20 large companies had active discovery and development programs for antibacterial agents; in 2013, only four have active discovery programs (18). Our approval rate for new antibiotics has fallen to dismally low levels (Fig. 1). The combination of the above has logically contributed to the hesitancy of companies to invest in the area and has led to a lack of both antibacterial drugs in the pipeline and to an absence of ongoing trials in indications where trial designs required by FDA are infeasible such as pneumonia. The FDA now recognizes that not