TrialCast

TrialCast runs 1,000 to 100,000 iterations (default 5,000 in ~8 seconds) to sample from probability distributions for each uncertain variable—enrollment rates, screen failure, site activation timing, and more. The output is a full probability distribution showing P10 (conservative), P50 (most likely), and P90 (pessimistic) outcomes. Gelman-Rubin convergence detection ensures statistical validity before presenting results.

Three scientifically-validated models: NHPP (Non-Homogeneous Poisson Process) for time-varying enrollment rates with site ramp-up periods—ideal when sites activate at different times. Gompertz curves for S-curve patterns with natural enrollment saturation as sites approach capacity. Beta distribution for asymmetric patterns with flexible control over enrollment timing. The system recommends the best model based on your trial characteristics.

TrialCast connects to the AACT database containing 50,000+ historical trials. Enter your trial parameters, and the system identifies 3-5 similar trials based on phase, indication, site count, and enrollment targets. You get similarity scores, cohort-level statistics, and P10/P50/P90 percentiles for enrollment duration—replacing assumptions with evidence from real completed trials.

Absolutely. Unlike black-box AI, TrialCast uses peer-reviewed statistical methods that auditors understand: Monte Carlo simulation, Poisson processes, and standard probability distributions. Every projection includes explicit uncertainty quantification (confidence intervals), full audit trails showing all inputs and assumptions, and version-controlled budget histories. We provide methodology documentation suitable for regulatory and financial audits.

TrialCast uses multi-tenant architecture with complete data isolation. Every database query includes organization UUID validation. Row-level security ensures users only access their organization's data. The platform is architected for SOC 2 Type II compliance, with security, availability, and confidentiality controls built in from the ground up. All trial data, budgets, and scenarios remain completely private to your organization.

Yes—this is a core use case. When enrollment slips, amendments happen, or assumptions change, create a reforecast scenario in minutes. The baseline remains locked, so you can see exact deltas: "Adding 5 sites increases cost by $1.2M but reduces timeline by 6 weeks." Model intervention strategies (more sites, longer enrollment, protocol simplification) and present options to stakeholders with quantified trade-offs.

TrialCast integrates enrollment forecasting with cost modeling. Patient costs are distributed based on when patients are active (per the enrollment model). CRO milestones are placed at specific dates. Time-based costs spread evenly. Site activation follows the enrollment-driven schedule. Result: 9 cost categories stacked across your entire trial lifecycle with month-by-month precision.

Milestone costs (CRO and Labs) are tied to specific deliverables—they hit your budget when the milestone is achieved. Time-based costs are monthly fees regardless of enrollment progress. TrialCast shows both in the monthly distribution so you can see exactly when each type of cost hits.

TrialCast provides 5 pre-configured scenario templates: Conservative (+20% timeline buffer), Optimistic (-20% timeline), Cost Optimization (-20% sites), Fast Enrollment (+50% sites), and Balanced (AI-optimized). Each template adjusts key parameters and instantly shows the impact on cost, timeline, and risk vs. your baseline scenario.

Traditional planning activates all sites at once, leading to idle site fees. TrialCast's enrollment-driven approach activates sites proportionally to patient enrollment demand—sites come online when they're needed. This optimizes your $20K+ per-site activation budget and prevents wasted spend on inactive sites.