ITCH Trial: Protocol for a Randomized, Double Blind Placebo-controlled Trial

Model scores of the 29,894 trials active now.

Model score 16% for any early termination, 8% for termination due to slow enrollment. Past trials like this one (phase, sponsor type and disease area, started 2008–2016) ended terminated 15% of the time (107 trials).

Finished trial that started in 2015 or later: the model never saw it during training.

Cumulative chance of ending terminated (red) or completed (green) by each year from its start; it actually ended on 2024-06-04. The rest of the way to 100% is still running.

From the competing-risks survival model (termination vs completion, still-running trials censored), fit to trials that started 2008–2014 and recalibrated to termination rates over 2022–2024, since termination has become more common. Checked on past years, it still ran a little low when the rise continued.

• Wording of the registration (title, summary, criteria) (raises risk) · Wording Phrases that raised it: intravenousstrokehourshemorrhageplacebo
  Phrases that lowered it: acidtrialpatientseffectthe aim
• Accepts healthy volunteers: No (raises risk) · Who can join
• Sponsor's past termination rate: no earlier finished trials (lowers risk) · Sponsor's track record
• Minimum age: 20 (lowers risk) · Who can join
• Maximum age: 100 (lowers risk) · Who can join
• Drug's FAERS reports marked serious: 95.3% (raises risk) · Drug's FDA history
• Responsible party: principal investigator (lowers risk) · Design
• Blinding: triple (raises risk) · Design
• Design: parallel (raises risk) · Design
• Has a data monitoring committee: Yes (raises risk) · Design
• Sponsor type: academic or other (raises risk) · Sponsor's track record
• Phase: 2/3 (raises risk) · Design
• Disease area, neurology: Yes (raises risk) · Disease area
• Drug's FAERS reports with a death: 19.0% (lowers risk) · Drug's FDA history
• Trials the sponsor started in the prior 2 years: 0 (raises risk) · Sponsor's track record
• Primary purpose: treatment (raises risk) · Design
• Disease area, cancer: No (lowers risk) · Disease area
• Number of arms: 2 (raises risk) · Design
• Sponsor's past trials: 0 (lowers risk) · Sponsor's track record
• Disease area, cardiovascular: Yes (raises risk) · Disease area
• Collaborators: 0 (raises risk) · Design

Every input that moved this score, largest first: SHAP contributions in log-odds, red raising risk and green lowering it. They describe the model, not causes of termination. Phrases are traced from the text features back to this registration's words, approximately.

• Ultra-early Tranexamic Acid After Subarachnoid Hemorrhage. NCT02684812 · started 2012 · scored riskier than 83% of active trials completed
• STOP-MSU: Stopping Haemorrhage With Tranexamic Acid for Hyperacute Onset Presentation Including Mobile Stroke Units NCT03385928 · started 2018 · scored riskier than 98% of active trials completed
• Tranexamic Acid for Preventing Progressive Intracranial Haemorrhage in Traumatic Brain Injury NCT00755209 · started 2008 · scored riskier than 16% of active trials completed
• Clot Lysis: Evaluating Accelerated Resolution of Intraventricular Hemorrhage Phase III NCT00784134 · started 2009 · scored riskier than 96% of active trials completed
• PREvention of VENous Thromboembolism In Hemorrhagic Stroke Patients NCT01573169 · started 2012 · scored riskier than 83% of active trials terminated

Closest by meaning of the registration text (PubMedBERT embeddings), among finished trials. How they ended is context, not an input to this trial's score.