Clinical trial enrollment intelligence
Every patient who should be in a trial. Found.
GhostTrials de-risks the protocol before it’s finalized, identifies eligible patients across hospital systems with privacy-preserving federated matching, and orchestrates the enrollment funnel end-to-end — without raw patient data ever leaving the originating institution.
The problem
Enrollment fails before the first patient is screened.
Over-tight eligibility criteria are written into the protocol, then high-value candidates are lost in fragmented data silos, filtered out by slow manual screening, and never contacted because no one knew they existed. The cost compounds every single day a trial runs behind.
~75%
of protocols need at least one substantial amendment — a median $141K–$535K each.
86%
of clinical trials fail to meet their enrollment timelines.
1 in 20
eligible patients is ever asked to join a trial that fits them.
How it works
From protocol design to enrolled patient — in one system.
Parse the protocol
AI reads the trial protocol — from a PDF or an NCT id — and extracts structured eligibility criteria (biomarkers, labs, prior therapy, staging) with ontology codes and confidence scores.
De-risk the design
Before you finalize, a federated dry-run shows which criteria strangle enrollment and the exact edit that fixes it — grounded in real de-identified patient counts, not guesswork.
Match — privately
A federated query runs inside each hospital. Only pseudonyms and differentially-private counts come back. PHI never leaves the institution.
Resolve the ambiguous
Every uncertain record enters a human-in-the-loop queue with the exact blocking criterion, source evidence, and protocol citation. The engine never guesses.
Orchestrate enrollment
Identification, outreach, eConsent, scheduling, and regulatory submission — tracked, auditable, and re-routed automatically when a site nears capacity.
New — Protocol de-risking
Fix the criteria that strangle enrollment — before you lock the protocol.
Point GhostTrials at a draft protocol and get a Criterion Sensitivity & De-risking Report: a ranked view of how many additional patients each criterion would unlock if relaxed, an explicit what-if editor to test the trade-offs, and a per-criterion data-completeness score that flags criteria the source records can’t even support.
What-if, on real counts
Relax a threshold or drop a criterion and re-run a fresh federated pass. Every “+N patients” is a differentially-private count from an actual hospital network — the provenance edge claims-based tools lack.
A report you can act on
Baseline-vs-de-risked cohort projection, baseline covariate distribution, and small-cell-safe counts — exported as a shareable PDF for the protocol team.
Inside the platform
Every stage of the funnel, measured and auditable.
Federated cohort
De-identified candidate pool per site with pseudonymous IDs and per-criterion pass/fail — clinical detail stays withheld at the source.
Feasibility & relaxation
Per-criterion sensitivity on the live cohort: which criteria gate the most patients, and the near-miss gains from relaxing each.
Diversity & DAP analytics
DP-noised demographic breakdown against sponsor-set representation goals, with a draftable Diversity Action Plan.
Enrollment forecast
Transparent, rule-based projection to target with confidence bands and per-site expected enrollable counts — no black-box ML.
Dropout risk
A transparent weighted score per enrollment with the contributing factors shown, recomputed on every stage transition.
HITL review queue
First-class human review with SLA timers, blocking-criterion context, and source evidence for every ambiguous record.
Sites & capacity
Live capacity per site with automatic re-routing to sites with headroom when one nears its enrollment ceiling.
eConsent & outreach
Generate consent documents and candidate outreach in-funnel; signature status and delivery are tracked end-to-end.
Regulatory drafting
Assemble IRB/HREC submission packages from real trial and site data, editable before you send.
Conflicts & audit
Append-only conflict resolutions with protocol citations, and a full org-scoped audit trail of every mutation.
Why it’s different
Built privacy-first, for regulated data.
Raw patient data never moves
Bloom-filter pseudonyms and Laplace-noised counts are computed at each hospital’s connector. Names, dates of birth, and MRNs stay put — by design.
HITL is a feature, not a fallback
Human review is a first-class compliance mechanism. SLA breaches auto-escalate; nothing ambiguous is silently decided.
Every decision is auditable
Conflict resolutions are append-only — stored with the protocol citation, model used, and reasoning. Multi-tenant by construction: one organisation never sees another’s data.
For sponsors & CROs
De-risk protocols pre-finalization, then see your whole portfolio: extracted criteria, the federated cohort, feasibility, diversity, forecast, dropout risk, and regulatory status — in real time.
For site coordinators
Work the queue that matters first. Review candidates, resolve ambiguous records with source evidence, run eConsent, and advance enrollment — without leaving the chart.
The patients are already there. Find them.
GhostTrials turns enrollment from a leaky funnel into a measured, auditable, privacy-safe pipeline — so trials finish on time and patients reach the therapies meant for them.
Sign in to your portal