Simulation-guided selection of a Bayesian adaptive Phase II design for a nine-arm cilostazol-albumin trial in aneurysmal subarachnoid hemorrhage

The Cilostazol Albumin Treatment in Subarachnoid Hemorrhage (CATS) trial evaluates eight active cilostazol-human albumin regimens and a concurrent control in patients with aneurysmal subarachnoid hemorrhage. This case study examines the proposed adaptive regimen-selection design and the tradeoffs associated with allocation, stopping, and reducing the candidate regimens. The binary endpoint was Common Data Elements-defined delayed cerebral ischemia within 14 days. The selected design combines protected control allocation, response-adaptive allocation, interim stopping, and two-dimensional normal dynamic linear model borrowing. Primary operating characteristics were obtained from 1,000 trials per scenario using Fixed and Adaptive Clinical Trial Simulator version 7.0.0. Reproducible independent-arm simulations used 5,000 trials per scenario (50,000 for recalibrated allocation comparisons) to examine purpose-built comparison strategies and sensitivity to missing outcomes, control risk, burn-in, and omitted regimens. We distinguish the probability of any Go decision from power to identify a truly beneficial treatment. The estimated global-null family-wise type I error rate was 0.104 for the selected design and 0.100 for the fixed benchmark. Under albumin-favored and cilostazol-favored gradients, respectively, Go probability was 0.858 and 0.838 versus 0.610 and 0.600 with fixed equal allocation; every active regimen was beneficial, so Go probability also measures power to identify a beneficial regimen. Expected enrollment was 308.3 and 307.9 versus 400.0. In independent-arm simulations without borrowing, the Statistical Analysis Plan allocation rule increased power from 0.697 to 0.777 and from 0.703 to 0.780 relative to fixed allocation under the same Go rule, with both calibrated to a 10% global-null target. Ten percent missing outcomes reduced power from 0.750 to 0.707 in the albumin-favored gradient. Reduced-arm designs could not identify a beneficial monotherapy omitted from the candidate set. For a nine-arm Phase II trial of related regimens with a protected concurrent control, the selected design increased Go probability and reduced expected enrollment relative to fixed equal allocation at a similar global-null family-wise type I error rate. Adaptive allocation added power in independent-arm simulations, whereas removing monotherapy arms could exclude the only beneficial regimen. These comparisons do not establish superiority over standard adaptive design classes or isolate the benefit of model borrowing.

Authors

Institutions

Publication Details

Journal
BMC Medical Research Methodology
Published
2026-11-18
DOI
https://doi.org/10.1186/s12874-026-03080-z
Primary Topic
Statistical Methods in Clinical Trials
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Simulation-guided selection of a Bayesian adaptive Phase II design for a nine-arm cilostazol-albumin trial in aneurysmal subarachnoid hemorrhage

Byron J. Gajewski, José I. Suárez, Jonathan Beall, Renee’ Hebert Martin et al.
BMC Medical Research Methodology
Statistical Methods in Clinical Trials
article

Simulation-guided selection of a Bayesian adaptive Phase II design for a nine-arm cilostazol-albumin trial in aneurysmal subarachnoid hemorrhage

Byron J. Gajewski, José I. Suárez, Jonathan Beall, Renee’ Hebert Martin, Naima Alam, Hassan Raza, Adnan I. Qureshi
article en

Abstract

The Cilostazol Albumin Treatment in Subarachnoid Hemorrhage (CATS) trial evaluates eight active cilostazol-human albumin regimens and a concurrent control in patients with aneurysmal subarachnoid hemorrhage. This case study examines the proposed adaptive regimen-selection design and the tradeoffs associated with allocation, stopping, and reducing the candidate regimens. The binary endpoint was Common Data Elements-defined delayed cerebral ischemia within 14 days. The selected design combines protected control allocation, response-adaptive allocation, interim stopping, and two-dimensional normal dynamic linear model borrowing. Primary operating characteristics were obtained from 1,000 trials per scenario using Fixed and Adaptive Clinical Trial Simulator version 7.0.0. Reproducible independent-arm simulations used 5,000 trials per scenario (50,000 for recalibrated allocation comparisons) to examine purpose-built comparison strategies and sensitivity to missing outcomes, control risk, burn-in, and omitted regimens. We distinguish the probability of any Go decision from power to identify a truly beneficial treatment. The estimated global-null family-wise type I error rate was 0.104 for the selected design and 0.100 for the fixed benchmark. Under albumin-favored and cilostazol-favored gradients, respectively, Go probability was 0.858 and 0.838 versus 0.610 and 0.600 with fixed equal allocation; every active regimen was beneficial, so Go probability also measures power to identify a beneficial regimen. Expected enrollment was 308.3 and 307.9 versus 400.0. In independent-arm simulations without borrowing, the Statistical Analysis Plan allocation rule increased power from 0.697 to 0.777 and from 0.703 to 0.780 relative to fixed allocation under the same Go rule, with both calibrated to a 10% global-null target. Ten percent missing outcomes reduced power from 0.750 to 0.707 in the albumin-favored gradient. Reduced-arm designs could not identify a beneficial monotherapy omitted from the candidate set. For a nine-arm Phase II trial of related regimens with a protected concurrent control, the selected design increased Go probability and reduced expected enrollment relative to fixed equal allocation at a similar global-null family-wise type I error rate. Adaptive allocation added power in independent-arm simulations, whereas removing monotherapy arms could exclude the only beneficial regimen. These comparisons do not establish superiority over standard adaptive design classes or isolate the benefit of model borrowing.

BMC Medical Research Methodology
Johns Hopkins University (US), Medical University of South Carolina (US), Johns Hopkins Medicine (US), University of Kansas Medical Center (US), University of Missouri (US)
Good health and well-being
Statistical Methods in Clinical Trials
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon