China
Africa
Explore chapters and articles related to this topic
Biostatistics
Published in Arkadiy Pitman, Oleksandr Sverdlov, L. Bruce Pearce, Mathematical and Statistical Skills in the Biopharmaceutical Industry, 2019
Arkadiy Pitman, Oleksandr Sverdlov, L. Bruce Pearce
As a follow-up, let us consider the conduct of the clinical trial. Many modern clinical trials utilize statistical monitoring of data and may have one or more pre-planned interim analysis. The interim analysis (IA) is defined as “any analysis intended to compare treatment arms with respect to efficacy or safety at any time prior to the formal completion of the trial” [61]. The purpose(s) of an IA may be: i) early stopping for overwhelming efficacy; ii) early stopping for futility; iii) stopping for safety reasons; and iv) modifying the study design (e.g. sample size reassessment; dropping an arm; enrichment of a subpopulation, etc.) Regardless of the purpose, the IA must be carefully planned for in the protocol. The IA poses many statistical challenges, such as control of the error rates; the fact that interim data can be immature, highly variable and fluctuating over time; and the fact that dissemination of the IA results can introduce bias and compromise integrity of the ongoing study. Without delving into much details here, one can derive a lot of useful and practical knowledge from studying group sequential design (GSD) methodologies [63] and principles of data monitoring committees (DMCs) [33]. The one-sentence message about IAs is that they may indeed help expedite clinical development and benefit patients; however they are neither free nor a means to satisfy investigators’ curiosity, and biostatisticians should play a pivotal role in both planning and execution of IAs.
Biomarker-Based Phase II and III Clinical Trials in Oncology
Published in Susan Halabi, Stefan Michiels, Textbook of Clinical Trials in Oncology, 2019
Shigeyuki Matsui, Masataka Igeta, Kiichiro Toyoizumi
In addition to its associated ethical benefits, interim analysis can enhance the efficiency of the marker-stratified trial. In particular, the statistical power can be improved by using interim trial data to adaptively narrow the patient population down to a patient subpopulation that can benefit from the treatment. Note that if the enrollment and follow-up are curtailed in M− patients, they may not be affected in M+ patients, unlike in the adaptive-enrichment design where the criteria for patient enrollment can be changed to recruit more M+ patients.
Globally Optimal Adaptive Trial Designs
Published in Mark Chang, John Balser, Jim Roach, Robin Bliss, Innovative Strategies, Statistical Solutions and Simulations for Modern Clinical Trials, 2019
Mark Chang, John Balser, Jim Roach, Robin Bliss
Practically, interim analysis (IA) time can be different from the initially planned due to different availability of Independent Monitoring Committee (DMC) members. Such a change in timing of IA is allowed (will not cause a type-I error inflation) as long as the change does not depend on the trial data and the stopping boundaries are recalculated at the actual information time following the prespecified error-spending function (ESF) – a cumulative error function since stage 1 of GSD. A popular error-spending function is power function family:
COURAGE-ALS: a randomized, double-blind phase 3 study designed to improve participant experience and increase the probability of success
Published in Amyotrophic Lateral Sclerosis and Frontotemporal Degeneration, 2023
Jeremy M. Shefner, Ammar Al-Chalabi, Jinsy A. Andrews, Adriano Chio, Mamede De Carvalho, Bettina M. Cockroft, Philippe Corcia, Philippe Couratier, Merit E. Cudkowicz, Angela Genge, Orla Hardiman, Terry Heiman-Patterson, Robert D. Henderson, Caroline Ingre, Carlayne E. Jackson, Wendy Johnston, Noah Lechtzin, Albert Ludolph, Nicholas J. Maragakis, Timothy M. Miller, Jesus S. Mora Pardina, Susanne Petri, Zachary Simmons, Leonard H. Van Den Berg, Lorne Zinman, Stuart Kupfer, Fady I. Malik, Lisa Meng, Tyrell J. Simkins, Jenny Wei, Andrew A. Wolff, Stacy A. Rudnicki
An independent, unblinded Data Monitoring Committee (DMC) will regularly review the data for safety and will also conduct two planned interim data reviews during this adaptive design study. First, 12 weeks after at least one-third of the planned study population has been randomized, they will assess the effect of reldesemtiv on ALSFRS-R total score change from baseline to week 12. If there appears to be a lack of effect in this first interim analysis, the DMC may recommend stopping the trial due to futility; if futility is not found, the trial will continue. Second, 24 weeks after at least one-third of the planned study population has been randomized, the DMC will assess whether the trial has adequate power to achieve a statistically significant effect on the primary endpoint in the final primary analysis, given the planned enrollment. The DMC may make a recommendation to (1) stop the trial if continuing is futile, (2) to increase the trial by a prespecified fixed number if continuing is promising, or (3) to continue as planned if the interim data do not suggest the first two options. This method is referred to as the CDL adaptive method and was initially proposed by Chen, DeMets, and Lan (17), later extended by Gao, Ware and Mehta (18), and Mehta and Pocock (19).
Efficacy and safety of blinatumomab in Chinese adults with Ph-negative relapsed/refractory B-cell precursor acute lymphoblastic leukemia: A multicenter open-label single-arm China registrational study
Published in Hematology, 2022
Hongsheng Zhou, Qingsong Yin, Jie Jin, Ting Liu, Zhen Cai, Bin Jiang, Dengju Li, Zimin Sun, Yan Li, Yanjuan He, Liping Ma, Sujun Gao, Jianda Hu, Aili He, Xin Du, Daihong Liu, Xiaohong Zhang, Xiaoyan Ke, Junling Zhuang, Yue Han, Xiaoqin Wang, Yuqi Chen, Paul Gordon, Dong Yu, Gerhard Zugmaier, Jianxiang Wang
The overall sample size of 120 patients, calculated using the exact method for a single proportion, was estimated to ensure 90% power to detect a significant difference in terms of CR/CRh rate between historical control with 30% CR/CRh rate and blinatumomab, assuming a 45% CR/CRh rate in the alternative hypothesis, at the 2.5% one-sided significance level. A Data Review Team oversaw a planned interim efficacy after 90 patients had a chance to complete 2 cycles of treatment and safety follow-up. The Data Review Team could recommend that the interim analysis become the primary analysis if the observed CR/CRh rate exceeded 42.2%, which was derived from an O’Brien-Fleming alpha spending function [22]. Regardless, of the interim analysis results, the trial would continue to enroll to the planned 120 patients.
Randomised controlled trial of interventions for bothersome tinnitus: DesyncraTM versus cognitive behavioural therapy
Published in International Journal of Audiology, 2022
Sarah M. Theodoroff, Garnett P. McMillan, Caroline J. Schmidt, Serena M. Dann, Christian Hauptmann, Marie-Christine Goodworth, Ruth Q. Leibowitz, Chan Random, James A. Henry
We took a Bayesian approach to estimating the mean TQ score by treatment arm and by hearing aid strata over time. Bayesian analysis is distinguished from classical approaches by identifying a probability distribution for the relevant parameter, which, in the current study, is the difference in population mean TQ score. This probability distribution is conditioned on available data, the fitted model, and investigator expertise, and reflects our varying degree of belief over different values of each unknown quantity. Bayesian approaches have the distinct advantage of being more interpretable than classical “p-values” and “confidence intervals,” and easily accommodate multiple comparisons (Gelman et al. 2014) and interim analysis as data are accumulated during the course of the study. Details of the general approach have been published (McMillan and Cannon 2019; Spiegelhalter, Abrams, and Myles 2004). An application of the Bayesian approach in a tinnitus clinical trial is provided in Theodoroff et al. (2017). Technical details of the model development, fitting, and computation are shown in the statistical appendix.