exclude: true --- class: top left hide-count background-image: url(img/cdc-QEU-QgIOJKA-unsplash_darkest.jpg) background-size: cover .move-top20[ .title-text[<span style='font-size:0.75em'>Designing Microbiome Trials</span>] .title-subtext[<span style='font-size:1.2em'>Unique Challenges & Considerations</span>] ] .callout-url-leftcorner[ .title-nametext[ Brendan J. Kelly, MD, MS Infectious Diseases, Microbiology & Epidemiology University of Pennsylvania GMFH World Summit 12 March 2022 ] ] .footnote-right[<span style='color: white'>photo: CDC @unsplash</span>] --- ## Disclosures .pad-left[ - No conflicts of interest. - Research supported by: - NIAID K23 AI121485 - CDC BAA 200-2016-91964 - CDC BAA 200-2016-91937 - CDC BAA 200-2018-02919 - CDC BAA 200-2021-10986 ] --- class: full-screen hide-count <div class="grid-3-1"> <div class="extension-tile gridset-a animated flipInY"><div>How Randomized Trials Go Wrong</div></div> </div> --- layout: false class: full-screen hide-count <div class="grid-3-1"> <div class="extension-tile gridset-a"><div>How Randomized Trials Go Wrong</div></div> <div class="extension-tile gridset-b animated flipInX"><div>RCTs of FMT</div></div> </div> --- layout: false class: full-screen hide-count <div class="grid-3-1"> <div class="extension-tile gridset-a"><div>How Randomized Trials Go Wrong</div></div> <div class="extension-tile gridset-b"><div>RCTs of FMT</div></div> <div class="extension-tile gridset-c animated bounceInDown"><div>Strategies for<br>the LBP Era</div></div> </div> --- layout: false class: full-screen hide-count <div class="grid-3-1"> <div class="extension-tile gridset-a"><div>How Randomized Trials Go Wrong</div></div> <div class="extension-tile gridset-grey"><div>RCTs of FMT</div></div> <div class="extension-tile gridset-grey"><div>Strategies for<br>the LBP Era</div></div> </div> --- ## "Understanding and Misunderstanding RCTs" .pad-left[ - Randomized clinical trials = gold standard for clinical research: - balance known & unknown confounders - powerful tool to measure causal effects - RCTs can still mislead: - selection bias (subjects ≠ population) - "trial interventions are interactions" resulting in heterogenous effects (subject features modify effect of trial intervention) - "subjects aren't even fully representative of themselves" (<u>time-varying features</u> also interact with trial intervention) ] .footnote-left[Deaton & Cartwright _Social Science & Medicine_ 2018; Andrew Gelman] --- ## Time-Varying Gut Microbiota <img src="./img/kelly_bioinf_2016.png" width="110%" style="display: block; margin: auto;" /> .footnote-left[Kelly et al _Bioinformatics_ 2015] --- ## Time-Varying Gut Microbiota .pad-left[ - Dietary changes and antibiotics exert large effects... ... larger effects when followed by colonization from healthcare environment - Gut microbiota at time of intervention ≠ microbiota at time of enrollment: - misclassification of inclusion criteria kills randomized trials - "the medicine doesn't work if the patient isn't sick" - microbiome trials run a **high risk of eligibility misclassification** ] .footnote-left[Kelly et al _ICHE_ 2021] --- ## How Microbiota Trials Go <u>Right</u> .pad-left[ - RCTs can overcome effect modification by changing gut microbiota if: - very precise disease phenotype (less likely time-varying gut microbiota) - precision medicine: tailor treatment to near-real-time measures of gut microbiota - interventions with huge effects (antibiotic conditioning prior to FMT) ] --- layout: false class: full-screen hide-count <div class="grid-3-1"> <div class="extension-tile gridset-grey"><div>How Randomized Trials Go Wrong</div></div> <div class="extension-tile gridset-b"><div>RCTs of FMT</div></div> <div class="extension-tile gridset-grey"><div>Strategies for<br>the LBP Era</div></div> </div> --- ## Fecal Microbiota Transplant (FMT) Trials .pad-left[ - FMT in Infectious Diseases @ the University of Pennsylvania: - FMT dose finding for recurrent _C. difficile_ infection (NCT03973697) - serial FMT for severe _C. difficile_ infection (NCT03970200) - FMT for MDRO colonization (CDC sponsored) - Successful applications of FMT: - recurrent & severe _C. difficile_, inflammatory bowel diseases, potentiation of anti-PD1 immunotherapy, MDRO colonization, food allergy mitigation - heterogeneity of gut microbiota community composition? ] --- ### Microbial Heterogeneity in _C. difficile_ .pad-left[ - **Aim**: microbiome features that discriminate _C. difficile_ colonization / infection - **Population**: 384 consecutive positive _C. difficile_ tests (in- & outpatient)
C. difficile
Category
Subject
Count
Proportion
GDH+ Toxin EIA- NAAT-
94
24.5%
GDH+ Toxin EIA- NAAT+
213
55.5%
GDH+ Toxin EIA+
77
20.1%
- **Sampling**: stool 16S rRNA gene sequencing & 16S rRNA gene qPCR - **Comparison**: toxin EIA+ (infection) versus NAAT+ only (colonization) - **Outcome**: EIA+, with fecal lactoferrin as sensitivity analysis ] .footnote-left[Tkatch et al _ASM World Microbiome Forum_ 2021] --- background-image: url(figs/p_toxineia_clostridioides_boxplot_tp.png) background-size: 65% .footnote-left[Tkatch et al _ASM World Microbiome Forum_ 2021] --- background-image: url(figs/p_cdi_cat_fecal_lacto_boxplot_tp.png) background-size: 65% .footnote-left[Tkatch et al _ASM World Microbiome Forum_ 2021] --- ## Phenotype Definitions & FMT Outcomes .pad-left[ - The success of FMT depends on homogeneity of gut microbiota phenotype: - moderate heterogeneity of _C. difficile_ infection - large heterogeneity across IBD - Pre-FMT antibiotic conditioning as a method to enforce homogeneity: - which antibiotics are used? - effects of other medications active on gut microbiota (e.g., PPIs)? - FMT as part of a bundled intervention? ] --- layout: false class: full-screen hide-count <div class="grid-3-1"> <div class="extension-tile gridset-grey"><div>How Randomized Trials Go Wrong</div></div> <div class="extension-tile gridset-grey"><div>RCTs of FMT</div></div> <div class="extension-tile gridset-c"><div>Strategies for<br>the LBP Era</div></div> </div> --- ## Into the LBP Era .pad-left[ - Live biotherapeutic products (LBPs) replacing FMT: - Ferring, Seres, Vedanta, Finch... - ensure greater homogeneity of intervention - larger effects? - Are bundled interventions still necessary? - role for pre-LBP antibiotic conditioning? - restrictions on medications that reshape gut microbial communities? - role for precision medicine & pre-treatment microbiota profiles? ] --- layout: false class: full-screen hide-count <div class="grid-3-1"> <div class="extension-tile gridset-a animated flipInY"><div>How Randomized Trials Go Wrong</div></div> <div class="extension-tile gridset-b animated flipInX"><div>RCTs of FMT</div></div> <div class="extension-tile gridset-c animated flipInY"><div>Strategies for<br>the LBP Era</div></div> </div> --- ## Topics for Discussion .pad-left[ - Every RCT intervention is an interaction: - gut microbiota community composition varies over time - account for changes in diet & medications, environment - RCTs can overcome effect modification if: - very precise disease phenotype (less likely time-varying gut microbiota) - precision medicine: tailor treatment to near-real-time measures of gut microbiota - interventions with huge effects (antibiotic conditioning prior to FMT) - How to translate lessons from FMT trials to the LBP era? ] --- exclude: true ## Acknowledgements .pull-left[ - __ARES Group @ Penn__ Sean Loughrey, Laura Cowden, Laurel Glaser, Kyle Rodino, Magda Wernovsky, Emily Reesey, Erik Clarke, Michael David, Matt Ziegler, Lauren Dutcher, Ebbing Lautenbach, Jim Harrigan, Hatem Abdallah - __Bushman Laboratory @ Penn__ Arwa Abbas, Aoife Roche, Andrew Marques, Aubrey Bailey, Jacob Leiby, Rick Bushman - __PennCHOP Microbiome Program__ Lisa Mattei, Casey Hofstaedter, Huanjia Zhang, Kyle Bittinger ] .pull-right[ - __Collman Laboratory @ Penn__ Ize Imai, Aurea Simon Soro, John McGinniss, Ron Collman - __Division of ID @ Penn__ Ian Frank, Pablo Tebas, Robert Gross, Emily Blumberg - __Rutgers University & Penn DBEI__ Jason Roy, Arman Oganisian - __CDC Prevention Epicenters__ Clifford McDonald, Alison Laufer Halpin - __Funding__ <u>CDC</u>: BAAs 200-2016-91964, 200-2018-02919, 200-2021-10986, 200-2021-10986 & <u>NIAID</u>: K23 AI121485 ] .center[ ### brendank@pennmedicine.upenn.edu ] --- class: middle center hide-count background-image: url(img/cdc-QEU-QgIOJKA-unsplash_darkest.jpg) background-size: cover .title-subtext[Questions?] .callout-url-bottom[ .fade-in[<span style="font-size:0.5em; color:white">slides ↓ </span>] <span style="link-color:white">[bjklab.org](http://www.bjklab.org)</span> ]