Specific Aims

Precision oncology has transformed cancer treatment for patients whose tumors are molecularly profiled and matched to targeted therapies, yet real-world data from Flatiron Health’s longitudinal EHR database reveals that these advances reach patients unevenly. Racial and ethnic minorities, rural populations, and patients at community oncology practices receive precision therapies at significantly lower rates than patients at academic centers, even when controlling for tumor type and stage. Traditional contract research organizations compound this inequity by recruiting from academic centers that already deliver precision oncology, systematically excluding the community practices where underserved patients receive care. What is not known is whether a governed federation of board-certified clinicians, operating as a distributed CRO grounded in real-world disparity evidence, can identify and close precision oncology treatment gaps in communities where traditional CROs do not operate.

Our long-term goal is to establish a governed distributed CRO federation that aligns precision oncology research with real-world evidence on treatment equity. The objective of this application is to use Flatiron’s de-identified EHR data to characterize treatment disparities and determine whether a credentialed clinician federation can function as a distributed CRO for underserved populations. Our central hypothesis is that governed community learning by ABOPM and CAOH clinicians, grounded in Flatiron real-world data and structured by CANONIC’s mathematical governance framework, can match traditional CRO performance on data quality while reaching populations that traditional CROs miss. This hypothesis is based on: (1) MammoChat, with two registered trials (NCT06604078, 199 enrolled; NCT07214883, 20,000 target) and AdventHealth support, demonstrating governed community learning in clinical settings; (2) CaribChat, live across 15 OECS member states through CAOH, demonstrating the same governance across jurisdictions without US structural inequity confounders; (3) SaveLife.AI, demonstrating that clinicians deliver governed intelligence remotely without traditional site infrastructure; and (4) RGN Med’s 75+ active protocols demonstrating physician-sovereign evidence capture at the point of care. The rationale is that the clinician network itself becomes the research infrastructure that traditional CROs cannot replicate.

Aim 1: Characterize precision oncology treatment disparities and distributed CRO site feasibility using Flatiron real-world data. We will analyze Flatiron’s EHR data across cancer types, demographics, and practice settings to map where precision treatments are underutilized and where ABOPM clinicians already practice in those communities. Working hypothesis: Disparity patterns will cluster by practice setting and demographics, and ABOPM clinicians will be co-located with the highest-disparity populations.

Aim 2: Validate governed community learning as a distributed evidence-capture mechanism across US and Caribbean networks. We will deploy OmicsChat (ABOPM) and CaribChat (CAOH) with RGN Med structured protocols, comparing community learning signals against Flatiron treatment pathways. The Caribbean arm provides a natural comparator where treatment gaps exist without US structural inequity confounders. Working hypothesis: Clinician-generated intelligence will independently identify Flatiron-documented treatment gaps, and US-Caribbean comparison will distinguish structural from clinical drivers of disparity.

Aim 3: Pilot the governed distributed CRO on Flatiron-identified disparity cohorts. We will deploy the full stack (OmicsChat/CaribChat, CANONIC governance, RGN Med evidence capture) at ABOPM, Howard, and CAOH sites, measuring clinician enrollment, governed data quality versus traditional CRO benchmarks, and treatment recommendation shifts toward equity. Working hypothesis: The governed federation will demonstrate comparable data quality while enrolling patients from communities where no traditional CRO operates.

Expected outcomes. Aim 1: a quantitative disparity map paired with distributed CRO site feasibility scores. Aim 2: validation of community learning as a disparity detection mechanism, with the first US-Caribbean evidence distinguishing structural from clinical drivers. Aim 3: first evidence that a governed distributed CRO can close precision oncology’s equity gap. These outcomes will demonstrate that the clinicians closest to underserved patients, federated through governed infrastructure and grounded in real-world evidence, are the research mechanism for closing precision oncology’s equity gap.

---

LOI Research Plan (page 2)

Description of Population Studied

Inclusion: Adults (≥18) in Flatiron’s de-identified EHR with confirmed cancer diagnosis and documented treatment decisions for cancer types with FDA-approved precision therapies. Caribbean arm: patients navigating care through CAOH-affiliated providers across 15 OECS states.

Exclusion: Pediatric patients. Cancer types without FDA-approved precision therapies. Records with insufficient demographic or treatment data.

Data Elements Required

Variable names pending Flatiron data dictionary access (NDA executed 2026-03-30, data dictionary requested from Dr. Cleo Ryals at Flatiron).

Demographics (age, sex, race/ethnicity, geography, insurance), clinical (cancer type, stage, histology, biomarkers: HER2, PD-L1, EGFR, ALK, BRCA, MSI, TMB), treatment (targeted therapy receipt, immunotherapy, genomic testing, line of therapy, time to treatment), practice setting (academic vs. community, size, geography), outcomes (OS, PFS, response).

Study Design and Statistical Analysis Plan

Aim 1: Multivariable logistic regression modeling precision therapy receipt as a function of demographics, tumor characteristics, and practice setting. Geographic clustering analysis for disparity hotspots. ABOPM registry cross-referenced against Flatiron geography for site feasibility.

Aim 2: Prospective OmicsChat/CaribChat deployment with RGN Med structured protocols. Kappa statistics comparing community-identified gaps against Flatiron disparity patterns. US-Caribbean stratified analysis testing convergence (clinical driver) vs. divergence (structural driver).

Aim 3: Pre-post design at 10-20 ABOPM/Howard/CAOH sites. Primary: governed data quality score benchmarked against published CRO metrics. Secondary: treatment recommendation equity shift. CANONIC MAGIC 255 governance score as continuous compliance metric.

Expected Significance and Impact (1-3 sentences)

This study will produce the first evidence that a governed distributed CRO can identify and close precision oncology treatment gaps where traditional CROs do not operate. If validated, the model scales to every cancer type and jurisdiction where credentialed clinicians practice, transforming a board from a credentialing authority into the research infrastructure that reaches the patients traditional trials exclude.

---

LOI Cover

Category	Information
Project Title	Governed Distributed CRO for Precision Oncology Treatment Equity: A Flatiron Real-World Data Study
PI	Dexter Hadley, MD/PhD — ABOPM, Director of AI — dexter@canonic.org
Co-I	Anil Bajnath, MD — ABOPM, Founder — ABajnathMD@gmail.com
Co-I	Marisa Nimrod, MD, MPH — CAOH, CEO — marisa.nimrod@gmail.com
Co-I	Robin Williams, MD, FACS — Howard University, Breast Surgery — Robin.Williams@howard.edu
Co-I	Alexander Evans, MD — Howard University, AI in Healthcare Consortium — alex.evans@howard.edu
Co-I	M. Amoy Fraser, PhD, CCRP, PMP — UCF COM, Director Clinical Research — amoy.fraser@ucf.edu
Co-I	Junaid Kalia, MD — SaveLife.AI — junaidkalia@neurocare.ai
Partner	RGN Med (Michael Tierney, CEO) — RWE infrastructure, 75+ protocols
Consultant	Neville Calleja, MD, PhD — University of Malta, WHO EHII Chair
Disease dataset(s)	aNSCLC, mBC, mCRC, advanced prostate (pan-cancer if available)

---

References

#	Citation
1	MammoChat. Governed clinical AI for breast cancer. AdventHealth (51 hospitals). mammochat.ai
2	CaribChat. Caribbean cancer intelligence. 15 OECS member states endorsed. caribchat.ai
3	ClinicalTrials.gov. NCT06604078: AI-Assisted Breast Cancer Clinical Decision Support. 199 enrolled.
4	ClinicalTrials.gov. NCT07214883: MammoChat v2. Recruiting 20,000 patients.
5	RGN Med. Real-world evidence infrastructure. 75+ active protocols. rgnmed.com
6	SaveLife.AI. Distributed clinical second-opinion infrastructure.
7	CAOH 2026 Annual Scientific Conference. July 17-19, Hilton Trinidad. caohcaribbean.org
8	ACS. Real-World Data Impact Award RFA. cancer.org
9	Flatiron Health. De-identified EHR database. 1000+ care sites.
10	CANONIC Foundation. Governed AI compliance. 6 patent families, 90 claims. canonic.org
11	OECS endorsement of CaribChat. 15 member states. March 2026.

---

MAGIC 255 Governance Closure

Dimension	Reviewer Concern	How Closed
D1: Declaration (1)	Are the rules explicit?	Every aim and protocol declared in governed contracts before execution.
D2: Evidence (2)	Is this evidence-based?	Every claim traces to Flatiron RWD, ClinVar, NCCN, or CAOH proceedings. GOLD/SILVER/BRONZE tiered.
D3: Credential (4)	Are clinicians qualified?	ABOPM board certification is the gate. CAOH provides Caribbean clinical authority.
D4: Community (8)	Will clinicians engage?	OmicsChat and CaribChat are live with demonstrated engagement.
D5: Practice (16)	Does this work clinically?	RGN Med operates 75+ protocols at the point of care. Two registered trials prove viability.
D6: Structure (32)	Is multi-site infrastructure sound?	CANONIC: eight binary dimensions, constructive proof of compliance across all sites.
D7: Learning (64)	Will findings compound?	Every session becomes permanent intelligence. Distributed CRO produces accumulating knowledge.
D8: Language (128)	Can the system self-govern?	Architectural constraints, not policy manuals. Append-only ledger, cryptographic integrity.

1 + 2 + 4 + 8 + 16 + 32 + 64 + 128 = 255. Full compliance.

---

SPECIFIC-AIMS | ACS-FLATIRON | GRANTS