Easy to bolt-on to existing trial designs
20x fewer errors vs conventional methods
Stop failed trials on average 30% into trial & detect successful ones less than halfway in.
Interim analysis is one of the most powerful tools in clinical trials. Done well, it can stop a futile trial years early, accelerate a successful one, or trigger the operational groundwork for Phase 3 while Phase 2 is still running. But manually accessing unblinded data is currently complex and risks bias, so the majority of industry-sponsored blinded trials conduct no planned interim analysis at all.
Presentient is solving this with its easy-to-integrate, standardised and validated BRAKES platform. BRAKES uses a patented Bayesian-like automated prediction algorithm running near-continuously on unblinded data inside an ultra-secure enclave which is blinded to all humans.
BRAKES continuously checks whether a trial still has a fair chance of success, creating opportunities to save millions per trial and de-risk an overall programme. Or, using BRAKES to continuously estimate the optimal time for a conventional interim analysis, it identifies successful trials to stop or trigger downstream commercial activities. One biotech client acted on an early signal, saving ~30% of their trial budget and redirecting resources to their pipeline.
BRAKES’ futility and efficacy interim analyses are decoupled for higher configurability and better performance for each type:
BRAKES_Futility
Stop signals come less than third of the way in and it would only make the wrong call once per decade (running on 20 trials per year for major pharma company) vs conventional conditional power that might make 2-3 wrong calls each year on the same set of trials. Save $10m+ per trial.
Incorporated into conventional frequentist framework protocols at the design stage with a standardised bolt-on paragraph, no changes to powering or sample size required.
Suitable for designs with endpoints that are continuous, binary & time-to-event as well as single arm designs.
Evaluates key secondary endpoints and incorporates veto logic between them, e.g. if a key secondary is performing well but the primary is highly likely to fail, then BRAKES_Futility stop signal can be suppressed to keep the trial running.
On Presentient’s Phase II empirical dataset with 10,000 trials, BRAKES’ Type II error rate is controlled to 0.5% vs 11.8% for conventional conditional power (at 20% threshold, at midpoint)
Already discussed with the EMA Innovation Task Force
BRAKES_Success
Either stop successful trials the moment the unblinded data shows evidence of overwhelming benefit, or keep trial running but collapse the Phase 2/3 white space. ROI 100x.
BRAKES_Success requires modest sample size adjustments in line with (or lower than) a single conventional interim analysis.
BRAKES_Success is deployed as an automated, pre-specified prediction tool to identify when is the optimal time to run a conventional O’Brien-Fleming interim analysis.
BRAKES_Success has been submitted for review by the FDA via their ISTAND qualification pathway.
On Presentient’s Phase II empirical dataset with 10,000 trials, 43% of successful trials would stop early, collapsing the whitespace between trials.
Scope and limitations
BRAKES provides the most value on blinded trials statistically powered for a primary endpoint (i.e. most Phase 2 & 3 trials, not Phase I). BRAKES needs a minimum sample size of 10 per arm. We are developing BRAKES functionality to run on dose escalation studies and medical device studies.
Stay blinded, decide sooner
Want to speed up your clinical trials with early decision making? Want to run 20+ automated interim analyses while eliminating operational bias?