2017-06-14…15 NIAID PK-PD Workshop: A quick summary What did we learn? John H. Rex, MD John.h.rex@gmail.com Please distribute freely 2017-06-15 - NIAID PK-PD - Summary ideas

What PK-PD can / cannot do Early: Derisk dose-selection Mid: Enable programs to happen at all in settings where clinical data are limited Late: Help with labeling questions about dosing and PK in special pops, pediatrics, drug-drug interactions, etc. Late: Provide support to breakpoint determination Cannot (necessarily) Speed the program: PK-PD is often iterative Make overall program smaller: You may well need to spend time exploring / confirming PK in relevant patient settings 2017-06-15 - NIAID PK-PD - Summary ideas

What is robust dose selection? “Robust” should reduce risk. What reduces risk? Standards: Use benchmark compounds as internal controls. We do this implicitly within classes – being explicit about this is helpful Isolates: Consider the use of standard (QC-like) PD isolates Orthogonal data: Test a range of isolates & models. Chose them to explore relevant variation Buffer your P3 study vs. patient-level PK variation Select dose to give maximum exposure within tox limits When all you have is HV data, anticipate greater variance in patients Get PK in suitably diverse patient settings as early as possible Exposure at the site matters: Keep site PK in mind … is plasma a good correlate or do you need to go deeper (e.g., ELF)? 2017-06-15 - NIAID PK-PD - Summary ideas

Standardization might help But let’s avoid the illusion of skill… Simple steps towards reduced variation in PD target measures Step 1: Detailed reporting on methods (ARRIVE animal data guidelines) Step 2: Using known ways to reduce variation (e.g., methods that reduce CFU/ml variation in the inoculum) Step 3: Seek to compare across labs It seems unlikely that we’d do a multi-lab (M23-like) study But, could we begin to build a database of standard (QC?) isolates tested vs. standard benchmark compounds? Is it useful to discuss how we do target attainment math? Current: Worst case ounding limits (MIC90, 95% CI PK limits) For the future: Variance-based (incorporate variance of observed PD targets, PK parameters, MIC ranges) 2017-06-15 - NIAID PK-PD - Summary ideas

2017-06-15 - NIAID PK-PD - Summary ideas Gaps / Transformation We lack validated cIAI & cUTI models Would routine use of real-time PK improve outcomes? Is point-of-care PK possible? How do we develop products without an MIC? It’s not clear that PK-PD provide the orthogonal causality support How do we develop combinations? How do we do preclinical PK-PD? How do we do the clinical testing for a combination regimen? TB as a paradigm for both: a workshop is coming Studies of specific clinical questions Exploring CSF penetration (and its meaning) More use of PBPK (Physiologically-based PK) Do we have the (compartment-level, by-age) data we need? 2017-06-15 - NIAID PK-PD - Summary ideas

Possible outputs from this meeting This workshop was an experiment. Thanks to NIAID for organizing and thanks to all of you for participating. What might we do / see next? Summary of best practice as of today Ex: We might write a summary paper (or two) NB: Discussions at this workshop should not be taken as regulatory guidance! Ideas gaps to address via funder-led work Ex: NIAID might seek to fund work addressing the gaps A further step in the conversation on this topic Ex: We might carry on the debate via further workshop(s) 2017-06-15 - NIAID PK-PD - Summary ideas