Picture the site initiation visit packet for a TMS-based depression trial running across a dozen clinics in three states. The PI at each site has a different background in neuroimaging. The coordinators have different EDC habits. And the EEG capture protocol, however carefully written, still depends on a technician who may or may not have run an ERP test before last Tuesday’s training call. This is the operational problem that every sponsor trying to use objective brain-function data as a trial endpoint eventually hits: the technology can be FDA-cleared and the science can be solid, but endpoint consistency lives or dies at the hands of whoever is running the device that morning.
Radial Therapeutics is now betting that vertical integration solves this problem in a way that site-by-site vendor contracts never have. Following its August 20, 2026 acquisition of TMS Health Partners, the management services organization behind Mindful Health Solutions, Radial manages 27 brain medicine clinics across seven states, making it one of the largest brain medicine clinical networks in the country. Now it has selected Firefly Neuroscience’s Evoke platform, a Class II neurodiagnostic device that received FDA 510(k) clearance in 2017, to feed objective EEG and ERP data into Radial’s AI operating system across that entire network. For clinical operations teams evaluating whether to engage this network for a CNS trial, the integration raises a question that matters more than the press release: what does standardized EEG capture actually require to hold up across 27 sites, and does a single MSO arrangement deliver it?
The Endpoint Consistency Problem Nobody Budgets For
Any coordinator who has worked a neuroimaging study knows that EEG data quality is exquisitely sensitive to things that never make it into the protocol: electrode impedance at setup, ambient electrical interference, patient fatigue before the session, and technician confidence with the software. ERP testing adds another layer, because the stimulus-response paradigm requires consistent administration conditions, the same auditory or visual stimuli, the same inter-stimulus intervals, the same instruction script delivered to the patient the same way. Miss any of those and the waveform morphology shifts. P300 latency, one of the most commonly used ERP markers in depression and cognitive disorder research, can drift by 20 milliseconds or more from confounds that a site coordinator would never flag as a deviation.
A published study of 52 MDD patients using ERP markers to predict antidepressant outcomes found that 61% achieved remission and 77% showed treatment responsiveness over eight weeks, results that are only meaningful if the underlying ERP data is captured consistently enough to be analyzed as a biomarker rather than noise. That level of endpoint integrity demands something most multi-site trials rely on inconsistently: a training and quality oversight mechanism that survives coordinator turnover, site-level distraction, and the ambient chaos of a clinic running five concurrent protocols.
This is where the MSO structure Radial has built changes the operational math, at least in theory. When a single entity manages the non-clinical operations of 27 clinics, staffing models, technology infrastructure, operational SOPs, it can enforce device training requirements and EEG capture protocols in a way that a sponsor CRA visiting four times a year cannot. A network of 27 clinics under unified MSO management means a single Evoke training curriculum, a single quality check against the same device firmware version, and a single escalation path when a site’s data looks anomalous. For sponsors, that is the operational pitch: the monitoring overhead per site drops when the site network is already operating to a common standard before the CRA ever books a flight.
But unified management and standardized execution are not the same thing, and the gap between them is where most multi-site trials actually fail.
What the FDA’s DCT Guidance Demands From Distributed Biomarker Capture
The FDA’s September 2024 final guidance on decentralized clinical trials does not give distributed EEG capture a free pass because the device is 510(k)-cleared. The guidance is explicit that sponsors using technology to collect trial data at locations other than the principal site must establish that the data collection process is equivalent across those locations, equivalent in training, in device qualification, and in documentation. For an EEG/ERP endpoint, that means source documentation for every session: the impedance log, the artifact rejection parameters, the technician identifier, the calibration record. It means a deviation process for sessions that fall outside acceptable signal quality thresholds. And it means a monitoring plan that can detect drift in data quality before it accumulates across enough patients to compromise the endpoint.
Sites I work with across distributed CNS programs routinely underestimate how much site-level documentation a neurophysiology endpoint generates compared to a standard PRO or clinical scale. A coordinator managing a MADRS assessment produces a completed form and a source note. A coordinator managing an Evoke session produces a session log, a signal quality record, and potentially a deviation report if the impedance values fell outside protocol spec, and those artifacts need to land in the TMF and the EDC in a way that a remote data review can actually interrogate. Across 27 sites, that documentation volume is not trivial, and a unified MSO helps only if it has built the SOP infrastructure to capture it consistently.
Radial’s stated model, backed by $50 million in Series A and seed funding led by General Catalyst as of December 2025, is an AI-native operating system that aggregates clinical data across its network. The Evoke integration is designed to feed into that system rather than generate isolated site-level data files. If the data pipeline from device capture to the central platform is robust, a sponsor’s data management team could theoretically see a real-time signal quality dashboard across all 27 sites rather than waiting for monitoring visit findings. That is a genuine operational advantage, provided the integration has been validated to the standard the FDA’s DCT guidance requires for remote data collection tools.
The Sponsor Questions That Need Answers Before Site Activation
For clinical operations teams evaluating Radial’s network as a trial execution partner, the Evoke integration creates a vendor oversight question that sits upstream of the typical site qualification checklist. Firefly’s Evoke platform earned its 510(k) clearance as an aid to clinical diagnosis, as validated in a pilot study evaluating EEG biomarkers in MDD patients treated with vortioxetine. Using it as a trial endpoint rather than a clinical tool requires the sponsor to specify, in the protocol and the SAP, exactly which Evoke output variables constitute the endpoint, what ranges are acceptable, and how out-of-range sessions are handled. That analytical definition has to survive FDA review, and it has to be operationally enforceable at the site level.
The pre-startup work here is more demanding than sponsors typically scope. Before the SIV, the sponsor’s clinical operations team needs to confirm that every Radial clinic in the trial has completed Evoke device qualification under a documented procedure, that the training records are in the TMF, and that the data transfer pathway from the Evoke platform to the EDC has been tested and validated. A PSV checklist built for a standard Phase II depression trial does not cover those items, they require a device-specific readiness assessment that looks more like a lab qualification audit than a site initiation walk-through.
Sponsors who plan to run risk-based monitoring against this network should build their central monitoring triggers around EEG signal quality metrics, not just data entry patterns. If a site’s average impedance values or artifact rejection rates drift outside the protocol-specified range across consecutive sessions, that is a data integrity signal that warrants a triggered on-site visit, not a query. Building that trigger into the monitoring plan before first-patient-in is the difference between catching a site-level execution problem in month two and discovering it at database lock.
The operational truth here is that Radial’s 27-clinic network and the Evoke integration represent a genuinely new site activation model for CNS trials, one where the sponsor is not assembling a site network from scratch but inheriting an existing operational infrastructure with its own standards, technology stack, and management chain. That is faster to activate and cheaper to monitor than building from individual sites. It also means the sponsor’s oversight responsibility shifts: less time qualifying individual sites, more time qualifying the MSO’s systems. For CNS sponsors accustomed to one model, the transition requires updating the vendor oversight framework before the first contract amendment is signed.
References
- GlobeNewswire, “Firefly Neuroscience’s Evoke™ Platform Selected to Bring Objective Brain-Function Insights to Radial’s AI-Native Operating System for Brain Medicine”
- Evoke Neuroscience, FDA 510(k) clearance and device description for the Evoke EEG/ERP platform
- Fierce Healthcare, “Radial acquires TMS Health Partners, MSO behind network of brain medicine clinics” (August 20, 2026)
- Radial Therapeutics — “$50 Million Series A and Seed Funding, December 2025”
- PubMed, ERP markers predicting antidepressant outcomes in 52 MDD patients (61% remission, 77% treatment responsiveness)
- ResearchGate, “Evaluating an EEG-based tool for assessing acute clinical and cognitive changes in adult outpatients with MDD treated with open-label, flexible-dose vortioxetine: A pilot study”
- Federal Register, FDA Final Guidance: “Conducting Clinical Trials With Decentralized Elements” (September 18, 2024)
