Pull up the site list for any active GBA1 or LRRK2 Parkinson’s trial and run a quick geography check. The sites cluster in Western Europe and the continental United States. The IRB submissions went to central boards with English-language consent templates. The patient identification algorithms were built on neurology clinic databases that skew heavily toward non-Hispanic white populations. Nobody made a deliberate decision to exclude anyone. It just happened, incrementally, one feasibility questionnaire at a time, and now a landmark study published in Lancet Neurology has put the operational consequences in plain view.
That study, from the Global Parkinson’s Genetics Program (GP2), analyzed 99,783 individuals — 58,559 Parkinson’s disease patients and 41,224 controls — across multiple ancestries. The headline finding for clinical operations is not the genetics. It is what the genetics reveal about who is missing from every trial currently running. Overall, 2.1% of Parkinson’s patients in the cohort carried a causal genetic variant, but that number fractures dramatically by ancestry in ways that should force sponsors to reopen their site feasibility models today, not at the next protocol amendment cycle.
What the Ancestry Data Actually Changes
Consider the rs3115534-G variant in GBA1. Approximately 50% of West African Parkinson’s disease cases carry it, with a dose-dependent risk relationship: one copy confers roughly 1.5 times the risk of developing PD, two copies confers roughly 3.5 times the risk. That is a clinically significant carrier frequency in a population that is almost entirely absent from current genetically-informed trials. The variant is not rare in West African ancestry. The patients are rare in site databases, which is a different problem entirely and a correctable one.
The operational implication is immediate. Sponsors running GBA1-targeted trials built their sample size assumptions on European carrier frequency data. Their site activation decisions followed from those assumptions. A site in Lagos or Accra was never in the feasibility model because the feasibility model was never built to ask whether it should be. That circularity is exactly how research populations calcify over decades, and the GP2 data now makes the cost of that circularity legible in a way that regulators and payers will not ignore for much longer.
Sites I work with across diverse urban networks already feel this pressure arriving from two directions simultaneously. Sponsors are starting to ask, mid-Phase II, whether their enrollment population will support genetically-informed labeling claims across ancestries. At the same time, IRB reviewers at several institutions are flagging consent documents that lack translated materials for the communities a protocol theoretically aims to reach. The friction between those two pressures lands on the CRC’s desk in the form of a protocol that was designed for one population and is now being asked to serve another, without any corresponding change to the startup documents, the subject identification workflow, or the per-patient budget.
That budget gap deserves its own paragraph. Adding ancestrally diverse sites mid-pipeline is not a checkbox exercise. Phase III trial expansions mid-study carry substantial activation costs, driven by site qualification visits, contract and budget renegotiation, IRB submissions at new institutions, staff training, and the startup burn that precedes first-patient-in at any new site. A sponsor who budgets for 40 sites and then decides at Month 18 that they need community health centers in Atlanta, Houston, and South London has not budgeted for those sites. They have budgeted for the sites they already activated. The difference comes out of someone’s contingency reserve, and in many trials, that reserve was already thin.
The Startup Documents Nobody Revised
The BLAAC PD study run by GP2 offers a concrete benchmark for what intentional diversity recruitment actually requires at the operational level. As of August 2025, BLAAC PD had enrolled over 350 Black and African American Parkinson’s patients and over 500 controls across twelve US sites, four years after initiation. Twelve sites, four years, 850-plus participants. That pace reflects the real resource intensity of community-embedded recruitment when you build it correctly from the start. It is not a pace that sponsors can replicate by adding a diversity supplement to an existing protocol amendment at Month 20.
The startup documents are where the operational gap is most visible and most fixable. Consent forms written in English only, at a reading level calibrated to academic medical center patients, will not clear ethics review at community health centers serving West African or South Asian diaspora populations. The translation workflow alone — document preparation, certified translation, back-translation, IRB resubmission, site-level staff training on the revised consent — adds four to eight weeks to site activation timelines at minimum. Any coordinator who has managed a consent amendment at a site with a non-English-speaking patient population knows that the IRB turnaround is not the bottleneck. The bottleneck is the three rounds of back-and-forth between the sponsor’s medical writing team and the translation vendor before a document that reads like a research protocol in English becomes a document that reads like a conversation in Yoruba or Mandarin or Spanish.
ICH E6(R3), finalized in 2023, strengthened expectations around participant-centered design and proportionate approaches to trial conduct. Section 4 of the revised guideline places direct responsibility on sponsors to ensure that the trial design, including recruitment and consent processes, is appropriate for the intended participant population. That language has teeth in a genetically-informed Parkinson’s trial where the intended population now demonstrably includes carriers who are not being reached by current site configurations. An FDA BIMO inspection of a GBA1 trial that enrolled 98% European-ancestry participants when the science shows 50% West African GBA1 carrier prevalence is a finding waiting to happen.
What Changes Monday Morning
For site operations teams, the immediate action is a feasibility reconciliation audit on any active genetic Parkinson’s trial. Pull the current enrollment data by ancestry. Compare it against the GP2 carrier frequency data by population. If the enrolled population does not reflect the carrier frequency distribution, document why, and document it now, before a monitor asks. The TMF entry that explains a deliberate, evidence-based enrollment decision is a different document from the TMF entry that explains the absence of any decision at all.
For sponsors, the concrete ask is earlier. Add an ancestry-stratified carrier prevalence analysis to the site feasibility template at protocol design, not at interim analysis. The GP2 data now provides the frequency inputs needed to model which community health centers, academic sites, and patient advocacy networks in non-European populations can realistically contribute to a genetically-defined enrollment target. That analysis will change the site list. It will change the consent document requirements. It will change the budget. Running it at protocol design costs a few weeks of work. Running it at Month 18 costs a restructured site network, compressed activation timelines, and the kind of enrollment rescue that no CRO bids into their original contract.
Across our network, the trials that will navigate this transition without a crisis are the ones that treat the GP2 findings not as a policy statement about equity, but as a feasibility constraint that changes the arithmetic of genetic enrollment. The sites that can serve West African, South Asian, and Latin American Parkinson’s patients are not generic neurology clinics. They are specific institutions with specific community relationships, and building those relationships takes longer than a site initiation visit.
The GP2 study did not create this problem. It made the problem’s dimensions precise enough to act on. Sponsors who read the Lancet Neurology paper as a scientific publication and not as an operational directive will be having a different, harder conversation with regulators when their genetically-informed label hits populations the trial never enrolled.
References
- Lancet Neurology — “Parkinson’s disease genetics across diverse ancestries: an observational genetic study of causal and risk variants with translational implications”
- PubMed — GBA1 rs3115534-G variant prevalence in West African Parkinson’s disease cases
- NORC — “Practical Approaches to Inclusive Recruitment Practices in Parkinson’s Disease Research” (BLAAC PD enrollment data, August 2025)
- Intuition Labs — Clinical Trial Start-Up Costs: Phase III mid-study network expansion cost benchmarks
- DZNE — GP2 multi-ancestry analysis: 99,783 individuals, GBA1 and LRRK2 variant frequency data across ancestries
