SAN DIEGO – Merck & Co Inc. reported data from two phase III trials of its antibody bezlotoxumab for the prevention of recurrent Clostridium difficile at the International Conference on Antimicrobials and Chemotherapy and the International Congress of Chemotherapy (ICAAC/ICC) 2015 annual meeting Sunday. In the trials, a single infusion of bezlotoxumab reduced recurrent C. difficile infection through 12 weeks, from 25 percent to 15 percent. Based on the data, Merck plans to file for approval of the antibody in the U.S., Canada and the EU.

Bezlotoxumab differs from a classical antibiotic in three ways. Its treatment goal is to prevent recurrence, it is an antibody and it is aimed at a toxin produced by C. difficile rather than C. difficile itself.

The broader issues of how to gain approval for antibiotics that do not conform to the classical model was discussed at a lively Saturday session on "Solving the Riddle of Developing Agents with Novel Approaches to Antimicrobial Therapy" that featured speakers from industry and U.S. and European regulatory authorities.

The field is undergoing a shift that is in some ways similar to targeted agents in oncology, where a classification of cancer based on its anatomical site is being supplemented by an approach that focuses on the genetics of the drug target, whether that target is a bacterium or a cancer cell.

Paratek Pharmaceuticals Inc. has both a broad and a narrow-spectrum antibacterial in phase III trials. Paratek's president, Evan Loh, told BioWorld Today that "theoretically, narrow is great."

In practice, narrow-spectrum agents "certainly have a role, and they really have a role if you know what the bug is."

But therein lies the rub: "When people come into your office [with an infection], you never know what it is."

In the case of bezlotoxumab, development of a narrow-spectrum agent was possible in part because C. difficile infections are easy to identify.

Other infectious agents, such as noroviruses, can cause diarrhea. But "once you are in a hospital with recurrent diarrhea, it's due to C. difficile," Merck's Nicholas Kartsonis, associate vice president, Clinical Research, told BioWorld Today.

This contrasts with the situation in infections like pneumonia, which can be caused by many different bacteria.

Loh said that "it's very hard to find a specific pneumonia infection that's caused by Pseudomonas," or whatever other bacterium a trial might be looking at.

And the situation is exacerbated by the speed by which some kind of treatment has to be initiated. The most extreme case is severe sepsis, where every delay of one hour in therapy raises the mortality rate by 7 percent to 8 percent.

Even in relatively less severe infections, time is much more of the essence than it is for cancer patients. While a lung cancer patient can be tested for the EML-ALK fusion gene mutation and a decision to treat with Xalkori (crizotinib, Pfizer Inc.) can be based on results that are returned three days later, a patient with a severe infection can be dead within those three days.

Those issues slow down enrollment of trials overall. The Medicines Co.'s chief medical officer, Jeffery Loutit, said at the symposium that enrolling a trial for an indication of ventilator-acquired pneumonia that was limited to a particular bacterial species could take five to seven years. "And a lot can happen in those five to seven years," he acknowledged, as bacteria continue to develop resistance and competitors continue to develop their own drugs.

For agents that specifically target resistant strains of bacteria, the problems of enrollment can be such that the EMA does not always require trials to be powered for the ability to statistically compare the experimental agent.

"A randomized trial is always preferred," the EMA's chair of Infectious Disease Working Party, Mair Powell, told the audience at the symposium. "But we completely accept that it may not always be possible to power it for inferential statistics."

Such a trial would still have to randomize patients – "the randomization step is considered by us to be a really important issue," she clarified.

But in a trial of a new agent for a multidrug-resistant bacterial strain, the number of treated patients might be so small that "the focus is not really on these data; the focus is on supporting the evidence of the dose."

On this particular issue, the FDA is stricter, requiring inferential statistics, which necessitate larger trials. The FDA's Deputy Director for Safety Sumathi Nambiar said at the symposium that the agency plans to release draft guidelines for developing antibiotics for unmet medical needs. But those will not change the fact that inferential statistics are required.

Statistical issues highlight another problem that is specific to antibiotic trials in switching from anatomically based indications (for example, urinary tract infection) to specific pathogens. While tumors in different anatomical locations are evaluated by the same criteria – basically, tumor shrinkage – the outcomes used to measure antibiotic success are different for different infections.

Given those differences, Nambiar said, the FDA could be prepared to accept a wider margin when determining noninferiority of an experimental drug.

If a trial does pool across body sites, Nambiar said the trial's sponsors should "make every attempt" to have 50 percent patient with serious form of the infection in question – that is, one that is hospital acquired and/or multidrug resistant.