Picture the enrollment feasibility call for a GBA1-targeted Parkinson’s trial. The CTM pulls up the site list: twelve sites in the US, four in the UK, two in Germany. The clinical operations lead flags that GBA1 carrier prevalence in the target population is “well-characterized.” The biostatistician has modeled a screen failure rate of 35%. Nobody on the call has asked which population the variant prevalence data came from — or whether those numbers actually hold for the sites they are activating.
A study published this month in Lancet Neurology makes that oversight harder to defend. The large-scale, multi-ancestry genetic analysis of Parkinson’s disease documents population-specific variant architecture that diverges meaningfully from what European-ancestry GWAS data has trained the field to expect. The operational consequences of ignoring that divergence are not abstract. They land in screen failure budgets, enrollment timelines, and IRB amendment cycles — and they land hard.
The Prevalence Numbers Your Feasibility Model Probably Got Wrong
The practical starting point is variant frequency. Sponsors designing GBA1 and LRRK2 trials have been doing so with prevalence estimates derived overwhelmingly from European cohorts. According to a review published in Frontiers in Genetics, more than 75% of participants in Parkinson’s disease GWAS studies are of European ancestry, with fewer than 4% representing African, Latin American, or Asian populations outside of Chinese-ancestry cohorts. When a feasibility model draws carrier prevalence from that data, it is essentially extrapolating from one population to all populations — and the Lancet Neurology study shows that extrapolation is not reliable.
Consider what this means at the site level. A site in Lagos or Nairobi activated to enroll GBA1 carriers will encounter a different variant landscape than a site in London or Boston. Research published in Brain documents that the GBA1 risk allele rs3115534-G carries a frequency of 0.23 in African-ancestry Parkinson’s cases compared to 0.17 in controls — a variant largely absent from European-ancestry analysis panels. If the genetic screening vendor’s panel was built to detect the European GBA1 variant profile, that site’s coordinator is running screens that will systematically miss carriers who actually qualify. Screen failure rates at non-European sites will come in above projection, the sponsor will flag the site as underperforming, and the root cause will never surface in a monitoring visit finding because nobody interrogated the panel design at SIV.
That is not a hypothetical failure mode. It is the predictable consequence of activating sites in diverse geographies without first reconciling the genetic testing vendor’s panel against the population-specific variant architecture the Lancet Neurology data now quantifies. The question every CTM should be walking into SIV with is specific: does this vendor’s GBA1 detection panel include the variants prevalent in this site’s patient population, or only the ones prevalent in the cohort that built the original assay?
The LRRK2 picture is similarly fractured by ancestry. Variant prevalence and pathogenicity differ across populations in ways that a single-ancestry enrollment strategy cannot capture — and the Phase 2 ACTIVATE trial (BIOHAVEN’s BIA 28-6156, targeting GCase enzyme function in GBA1-Parkinson’s patients, with sites tracked in the 2026 emerging trials watch) is running into exactly this enrollment calculus. Trials built on genetically stratified entry criteria carry an operational obligation that standard therapeutic trials do not: the entry criterion itself must be verified as analytically valid for the population being enrolled. When it is not, you do not get protocol deviations. You get systematic exclusion dressed up as a clean screen.
Where the IRB and Vendor Chains Break Down
Activating a genetically stratified trial across diverse geographies creates a cascade of operational dependencies that standard startup checklists were not built to handle. The IRB or IEC submission package for a site in an African or Latin American country must address the local regulatory framework for genetic sample collection, storage, and data transfer — and those frameworks vary substantially from the FDA or EMA expectations baked into most sponsors’ template packages. An amendment to add a non-European site mid-study does not just require a protocol deviation process review. It requires renegotiating the informed consent to address population-specific genetic privacy concerns, revalidating the central genetic testing vendor against local sample logistics, and sometimes re-engaging the ethics committee on ancestry-specific research risks. Sites I work with in early feasibility have flagged amendment cycles of 60 to 90 additional days when cross-border genetic data transfer requirements were not scoped into the original IEC submission.
The Global Parkinson’s Genetics Program (GP2), launched in 2020 through the ASAP initiative and The Michael J. Fox Foundation, has been building exactly the kind of site network and genomic infrastructure that genetically stratified commercial trials need — aiming to gather data from over 250,000 individuals across underrepresented populations. Sponsors who are not mapping their site activation plans against GP2’s established infrastructure are spending startup time rebuilding capacity that already exists in the field. A site that has been through GP2’s genetic data collection protocols has coordinators who understand the consent complexity, labs familiar with the sample handling requirements, and PIs who have navigated the local ethics landscape. That operational readiness does not appear on a standard site qualification scorecard, but it should.
There is also a vendor qualification gap that rarely surfaces until the first shipment fails. Central genetic testing labs that serve European and US sites have optimized their cold chain, kit design, and turnaround time for those shipping corridors. Activating a site in sub-Saharan Africa or Southeast Asia against the same central lab SOP — without reconciling kit stability data against local temperature ranges and customs clearance timelines — produces sample integrity failures that look like site performance problems on the enrollment dashboard. ICH E6(R3) Section 5.13 requires sponsors to document vendor qualification against the specific conditions of use. “Conditions of use” in a multi-ancestry trial includes the shipping conditions that samples actually travel through, not the conditions the vendor validated in a Frankfurt warehouse.
What Changes Monday Morning
For site operations teams, the immediate ask is to pressure-test the genetic screening panel before SIV, not after the first screen failure report arrives. Get the vendor’s assay documentation and map it against the published variant frequency data for each site’s target population. If the panel was built on European-ancestry cohort data and the site is enrolling from a predominantly African or Asian community, that gap needs to surface in the site qualification visit report — not six months into enrollment when the screen failure rate is twice the budget assumption.
For sponsors, the structural fix runs earlier in the process. Feasibility questionnaires sent to candidate sites for genetically stratified Parkinson’s trials need a section on local genetics infrastructure: has the site collected genetic samples under GCP before, do they have experience with the specific vendor panel being used, have their coordinators completed genetics-specific consent training? None of that is standard in a site feasibility questionnaire today. It needs to be, because the Lancet Neurology data has just formalized what GP2’s field experience has been demonstrating since 2020 — the genetic architecture of Parkinson’s disease is population-specific, and a trial enrollment strategy that does not account for that is not a diversity problem. It is a data integrity problem.
The next trial that misses its enrollment timeline because non-European sites are underperforming will almost certainly trace that failure to a screen failure rate nobody modeled correctly. The science to model it correctly now exists. The operational systems to act on it are the next gap to close.
References
- Lancet Neurology — “Parkinson’s disease genetics across diverse ancestries: an observational genetic study of causal and risk variants with translational implications”
- Frontiers in Genetics — “Enrollment disparities in Parkinson’s disease GWAS: >75% European ancestry, <4% African/Latin American/Asian representation”
- Brain — “GBA1 risk allele rs3115534-G frequency in African-ancestry Parkinson’s cases (0.23 in cases vs 0.17 in controls)”
- NeurologyLive — “Previewing Parkinson’s Disease Pipeline: Emerging Trials to Watch in 2026 (ACTIVATE trial, BIA 28-6156)”
- PubMed — “Global Parkinson’s Genetics Program (GP2): launched 2020 via ASAP/Michael J. Fox Foundation, targeting 250,000+ individuals across underrepresented populations”
