Antibiotic Drug Development

A recent analysis of drug development success rates from Phase I to approval found that the average infectious disease drug is 10 times more likely to succeed compared to the average oncology drug (a complementary analysis here–  different magnitude, same directionality).  Why?

In our opinion, highly accurate dose selection makes the difference.  In a recent JAMA analysis of drugs that failed to receive FDA approval, uncertainty around dose selection was the most frequent driver of the outcome.For the antibacterials field, a highly refined and accurate dose selection process is a huge asset that is not well appreciated as a way to significantly predict and de-risk a Phase III outcome.In our view, antibiotics drug developers and investors alike benefit greatly from a thorough analysis of dose selection ahead of investing time and capital in a clinical trial. We outline the basics and some key questions investors and trial sponsors alike should ask of the process with some help from the experts.Bugs don’t lie: the antibacterial field’s unique tools to de-risk clinical developmentMuch like driving a car benefits from a fuel gauge you can trust, clinical development benefits from a marker that accurately predicts how levels of drug over time impact activity against disease. Antibiotics offer one of the strongest markers around - we can observe in a dish and in predictive animal models changes in bacterial burden that directly drive our patients’ clinical outcomes. This powerful marker supports a process for predicting efficacious doses in Phase III with high fidelity based on human pharmacokinetics (PK) and pharmacodynamics (PD) as well as good preclinical models.

‍

Predictive PK/PD based Dose Selection in Four Steps (and key questions to consider along the way)

Understand in vitro drug concentrations needed to kill the broader population of target pathogens

• The key area to make sure of here is that the sample set chosen matches the target pathogens that patients in a trial will likely encounter
• Fortunately, there are great services like IHMA and JMI that can curate hundreds or even thousands of pathogens from clinical centers based on our specifications (region, resistance profile, and species of bug)

Model how dose drives bacterial killing over time

• This is where high fidelity in vivo and in vitro models help immensely. Experienced folks like our colleagues at the ICPD, Hope Lab, Drusano Lab, and others can help choose the right in vitro and in vivo assays based on the unique properties of drug
• The key question to ask is what’s are the exposure-response dynamics for the particular antibiotic. This is a critical driver of the ultimate dose and schedule of drug chosen for a trial. As a rule of thumb, antibiotics fall into two camps – the first are time-dependent, or kill bacteria best when drug exposures stay above a minimum threshold for the longest period of time. The second are concentration-dependent or kill best when maximizing the concentrations of drug over the dosing interval.

• Use Phase I healthy volunteer data to measure drug concentrations over time in serum or other effect sites (like the lung) and variability among patients

Using the above data, model a dose and schedule of administration for the study drug

• Mathematical models refined over decades with benefit of clinical data yield a calculated probability of target drug level attainment for a given dose and schedule in a trial – it pays to ask about this probability, as the higher the number the better the risk profile of the trial (see below)

‍

Predictive dose selection gets results:

‍The process works! A picture from our colleague Paul Ambrose and his team speaks louder than words (Paul and his team go into the nuances of dose selection in an excellent white paper here). The analysis looks at dose selection analysis versus approval over several decades to measure the predictive power of PK/PD analysis. Bottom line is – the more likely a drug’s dose is predicted to meet or exceed the level needed to kill bugs based on the four step dose selection process, the more likely it ultimately gets approved.

‍

Some Caveats:

Fidelity of dose selection based on preclinical and Phase I data depends exquisitely on an understanding of a drugs behavior at the site of infection. Much of PK/PD modeling is based on serum levels; if a drug has a hard time getting from serum to the site of infection (drugs with bioavailability challenges for example), one will have to account for this downside beyond what the modeling predicts. Conversely, drugs present at very high levels at the tissue site relative to their serum levels may have efficacy upside beyond the predicted required dose. Also, although most antibiotics are either time or concentration dependent, some antibiotics will be best dosed in ways that optimize both the maximal concentration and the time above a minimum threshold of concentration.

‍

So What?

Along with pathways to accelerated approval, a large and growing unmet need, the antibacterials field has the advantage of highly predictive early studies that de-risk late stage trials. We at Spero intend to leverage this huge advantage we have in the field and show our work to the external world ahead of Phase III about dose selection. In my observation, we as a field are very transparent about how well our drugs kill bugs, but our record is more mixed about support of dose selection. Not all drugs are created equal in their ability to get to the site of infection consistently. We’d hope that the investment, pharmaceutical, and academic community demand this of us in general to ensure that the right drugs with the right PK properties at the right dose make it to Phase III and succeed. A win for our drugs is a win for the field.

‍

‍

‍About the Authors:

Ankit Mahadevia, M.D.

Chief Executive Officer of Spero Therapeutics

David Melnick, M.D.

Chief Medical Officer of Spero Therapeutics

Paul Ambrose, PharmD

President of the Institute for Clinical Pharmacodynamics