Article
Clinical Neurology
Mengmeng Wang, Zhizhong Zhang, Marios K. Georgakis, Ville Karhunen, Dandan Liu
Summary: This study used a two-sample Mendelian randomization (MR) analysis to evaluate the causal effect of genetically proxied AMP-activated protein kinase (AMPK) activation on functional outcome following ischemic stroke. The results indicated that genetically predicted AMPK activation was significantly associated with lower odds of poor functional outcome, suggesting that metformin-mediated AMPK activation may have a beneficial effect on functional outcome following ischemic stroke.
Article
Immunology
Feiwu Long, Chenghan Xiao, Huijie Cui, Wei Wang, Zongze Jiang, Mingshuang Tang, Wenqiang Zhang, Yunjie Liu, Rong Xiang, Li Zhang, Xunying Zhao, Chao Yang, Peijing Yan, Xueyao Wu, Yutong Wang, Yanqiu Zhou, Ran Lu, Yulin Chen, Jiayuan Li, Xia Jiang, Chuanwen Fan, Ben Zhang
Summary: This study conducted a Mendelian randomization (MR) analysis using genome-wide association studies (GWAS) data and found causal associations between genetically predicted IgG N-glycosylation traits and COVID-19 susceptibility, hospitalization, and severity. The results suggest that modulating IgG N-glycosylation levels may have potential implications for improving COVID-19 outcomes.
FRONTIERS IN IMMUNOLOGY
(2023)
Article
Medicine, Research & Experimental
Leandro Garcia Barrado, Tomasz Burzykowski
Summary: The study found that the use of an imperfect biomarker assay in a Bayesian biomarker-driven outcome-adaptive randomization design can significantly reduce power and increase the type-I error probability. The effects depend on the sensitivity, specificity of the assay, and the distribution of the biomarker in the patient population. With an imperfect biomarker assay, careful consideration is required for implementing a biomarker-driven outcome-adaptive randomization design.
Article
Health Care Sciences & Services
Anders Granholm, Benjamin Skov Kaas-Hansen, Theis Lange, Olav Lilleholt Schjorring, Lars W. Andersen, Anders Perner, Aksel Karl Georg Jensen, Morten Hylander Moller
Summary: This article provides thorough guidance on key methodological considerations for adaptive clinical trials. It discusses the pros and cons of adaptive stopping, adaptive arm dropping, and response-adaptive randomization, and provides guidance on using simulation to compare different adaptive trial designs. The article focuses on Bayesian multi-arm adaptive trials and covers various aspects such as intervention and possible common control, outcome selection, timing of adaptive analyses, decision rules for adaptive stopping and arm dropping, randomization strategies, performance metrics, simulations, and reporting.
JOURNAL OF CLINICAL EPIDEMIOLOGY
(2023)
Article
Medicine, Research & Experimental
Edward L. Korn, Boris Freidlin
Summary: Response-adaptive randomization is less efficient in terms of increased required sample size compared to fixed 1:1 randomization ratio, and it leads to biased treatment effects when there are time trends. Analysis methods that account for potential time trends can eliminate this bias but contribute to additional inefficiency of response-adaptive randomization.
Article
Clinical Neurology
Zhizhong Zhang, Mengmeng Wang, Dipender Gill, Xinfeng Liu
Summary: This study investigated the association of genetically predicted smoking and alcohol consumption on poststroke outcomes. The results suggest a causal association between smoking and worse functional outcome after ischemic stroke, while alcohol consumption was not associated with functional outcome.
Article
Health Care Sciences & Services
Andrea Callegaro, B. S. Harsha Shree, Naveen Karkada
Summary: In this study, the performance of asymptotic and re-randomization tests under covariate-adaptive randomization was compared through simulation. The results showed that re-randomization tests are as powerful as asymptotic tests when the model is correct, and even more powerful when adjusting for covariates. Minimization and permuted blocks were found to provide similar results in the study.
STATISTICAL METHODS IN MEDICAL RESEARCH
(2021)
Article
Mathematical & Computational Biology
Wei Ma, Fuyi Tu, Hanzhong Liu
Summary: This article investigates several intuitive and commonly used regression models in randomized clinical trials, demonstrating that all these models can robustly estimate the treatment effect, even with arbitrary misspecification. The article also proposes consistent non-parametric variance estimators and compares them to existing model-based variance estimators. Recommendations are made for the effective use of regression under different scenarios.
STATISTICS IN MEDICINE
(2022)
Article
Economics
Federico A. Bugni, Mengsi Gao
Summary: This paper investigates inference in a randomized controlled trial (RCT) with covariate-adaptive randomization (CAR) and imperfect compliance of a binary treatment. It focuses on studying the local average treatment effect (LATE) and proposes an estimator based on instrumental variable (IV) linear regression. The paper demonstrates the asymptotic normality of the proposed estimator and characterizes its asymptotic variance in terms of the problem's parameters. It also provides consistent estimators of standard errors and asymptotically exact hypothesis tests, and explores strategies to minimize the asymptotic variance in a hypothetical RCT.
JOURNAL OF ECONOMETRICS
(2023)
Article
Multidisciplinary Sciences
Jiwoo Lee, Sakari Jukarainen, Antti Karvanen, Padraig Dixon, Neil M. Davies, George Davey Smith, Pradeep Natarajan, Andrea Ganna
Summary: Understanding the causal impact of clinical risk factors on healthcare costs is crucial for evaluating healthcare interventions. Using a genetically-informed design, this study found that waist circumference, body mass index, and blood pressure have significant causal impact on healthcare costs. Increased waist circumference is a major contributor to annual total healthcare costs.
NATURE COMMUNICATIONS
(2023)
Article
Optics
Farzin Salek, Masahito Hayashi, Andreas Winter
Summary: Adaptiveness is a key principle in information processing, and this study investigates its usefulness in asymptotic binary hypothesis testing for quantum channels. The results show that adaptive and nonadaptive strategies have the same error exponents for classical-quantum channels, and adaptive strategies do not outperform nonadaptive strategies when restricted to classical feed-forward and product state channel inputs.
Article
Clinical Neurology
Huan Cai, Hao Zhang, Jialin Liang, Zhonghua Liu, Guozhi Huang
Summary: Using the Mendelian randomization (MR) framework, this study found evidence of a possible causal effect of frailty on poor functional outcome after ischemic stroke. The findings highlight the potential of targeting frailty as an intervention to improve recovery after ischemic stroke.
INTERNATIONAL JOURNAL OF STROKE
(2023)
Article
Health Care Sciences & Services
Kim May Lee, J. Jack Lee
Summary: Bayesian adaptive randomization is a method that adjusts random assignment based on data trends, but there is a risk of patients being assigned to inferior treatment arms. Researchers have proposed an adaptive clip method and utility approach to address these issues and aid in selecting a randomization procedure.
STATISTICAL METHODS IN MEDICAL RESEARCH
(2021)
Editorial Material
Oncology
Jack Bowden
Summary: MR-PHeWAS is a powerful new design for discovering causal mechanisms between a disease and its candidate risk factors, with great potential in cancer research when utilizing powerful and principled statistical approaches.
BRITISH JOURNAL OF CANCER
(2021)
Article
Medicine, General & Internal
Gonzalo Villanueva-Martin, Marialbert Acosta-Herrera, Martin Kerick, Elena Lopez-Isac, Carmen P. Simeon, Jose L. Callejas, Shervin Assassi, Lorenzo Beretta, Yannick Allanore, Susanna M. Proudman, Mandana Nikpour, Carmen Fonseca, Christopher P. Denton, Timothy R. D. J. Radstake, Maureen D. Mayes, Xia Jiang, Javier Martin, Lara Bossini-Castillo
Summary: Obesity is a known risk factor for immune-mediated diseases, but its role in systemic sclerosis (SSc) is still unclear. This study used genetic correlation analyses and Mendelian randomization methods to investigate the relationship between body fat distribution parameters and SSc. The results showed no genetic correlation or causal relationship between these parameters and SSc, except for a negative causal association between WHRadjBMI and SSc.
JOURNAL OF CLINICAL MEDICINE
(2022)
Article
Statistics & Probability
Huaqing Jin, Guosheng Yin, Binhang Yuan, Fei Jiang
Summary: Motivated by wind turbine anomaly detection, a Bayesian hierarchical model (BHM) is proposed for mean-change detection in multivariate sequences. BHM combines a random order distribution induced from the Poisson-Dirichlet process and nonlocal priors, exhibiting satisfactory performance for mean-shift detection. The method yields the smallest detection error and outperforms other competitive methods for wind turbine anomaly detection in terms of the F1 score.
Article
Statistics & Probability
Jiaqi Gu, Yiwei Fan, Guosheng Yin
Summary: The Kaplan-Meier estimator, a nonparametric estimate for event survival functions, is restructured as an M-estimator through maximizing an M-function based on concordance, providing a new interpretation and establishing the limiting distribution. Empirical results demonstrate that the proposed M-estimator is equivalent to the KM estimator and can derive confidence intervals and bands.
AMERICAN STATISTICIAN
(2022)
Article
Statistics & Probability
Hengtao Zhang, Guosheng Yin
Summary: Rerandomization is gaining attention in literature for randomized experiments due to its ability to achieve balanced allocation using covariate information, but it may raise ethical concerns. A response-adaptive rerandomization scheme is proposed for two-arm comparative clinical trials, showing promising statistical and ethical properties. Extensive simulation studies demonstrate the practicality and superiority of this approach.
JOURNAL OF THE ROYAL STATISTICAL SOCIETY SERIES C-APPLIED STATISTICS
(2021)
Article
Mathematical & Computational Biology
Huaqing Jin, Guosheng Yin
Summary: In clinical trials, incorporating historical data in the analysis is crucial for gaining more information, improving efficiency, and providing a comprehensive evaluation of treatment. The new informative prior UIP can adaptively borrow information from multiple historical datasets, and is intuitive, easy to implement, and requires only summary statistics commonly reported in the literature.
STATISTICS IN MEDICINE
(2021)
Review
Public, Environmental & Occupational Health
Chenyang Zhang, Huaqing Jin, Yi Feng Wen, Guosheng Yin
Summary: This network meta-analysis evaluated the efficacy of various treatments for COVID-19 compared to standard care, showing superiority in terms of mortality, mechanical ventilation, hospital discharge, and viral clearance. Tocilizumab was found to be particularly effective in preventing severe outcomes and increasing discharge rates. Additionally, the study highlighted the clinical efficacy of antineoplastic agents, immunostimulants, and immunosuppressants in reducing mortality, ventilation risk, and increasing discharge rates, providing valuable information for potential COVID-19 treatments.
FRONTIERS IN PUBLIC HEALTH
(2021)
Article
Health Care Sciences & Services
Huaqing Jin, Guosheng Yin
Summary: Recent developments in oncology treatment have focused on targeted therapy and immunotherapy, with a shift towards more tolerable treatments that prioritize efficacy. The calibration-free odds (CFO) design presented in this study offers a model-free approach for determining optimal biological dose (OBD) in cancer trials. Extensive simulation studies show that CFO strikes a good balance between efficiency and safety for identifying maximum tolerated dose (MTD) in phase I trials and performs comparably or slightly better for OBD identification in phase I/II trials compared to other methods.
STATISTICAL METHODS IN MEDICAL RESEARCH
(2022)
Article
Statistics & Probability
Yiwei Fan, Jiaqi Gu, Guosheng Yin
Summary: Ordinal classification is an important area in machine learning, aiming to accurately predict the relative order of instances. We propose a novel concordance-based approach that incorporates variable selection and penalized optimization for sparsity considerations.
SCANDINAVIAN JOURNAL OF STATISTICS
(2023)
Article
Statistics & Probability
Wen Su, Guosheng Yin, Jing Zhang, Xingqiu Zhao
Summary: Big data presents both theoretical and computational challenges as well as tremendous opportunities in various fields. In healthcare research, a novel approach called divide-and-conquer (DAC) is developed to handle massive and high-dimensional data. The proposed DAC method performs well and is applied to a large dataset from the Chinese Longitudinal Healthy Longevity Survey.
CANADIAN JOURNAL OF STATISTICS-REVUE CANADIENNE DE STATISTIQUE
(2023)
Article
Health Care Sciences & Services
Huaqing Jin, Wenbin Du, Guosheng Yin
Summary: In the development of new cancer treatments, determining the maximum tolerated dose in a phase I trial is crucial. Different approaches, such as model-based and algorithm-based designs, have their own strengths and weaknesses. The proposed approximate Bayesian computation design offers a more efficient and model-independent method for dose assignment in phase I trials.
STATISTICAL METHODS IN MEDICAL RESEARCH
(2022)
Article
Health Care Sciences & Services
Hengtao Zhang, Guosheng Yin
Summary: Randomized controlled trials (RCTs) are considered the gold standard for clinical research on treatment effects. Recent interest has focused on integrating real-world evidence from observational studies to enhance and complement RCT results. The unit information prior (UIP) is a newly proposed technique that effectively incorporates information from historical datasets. We extend this approach to synthesize non-randomized evidence into current RCTs, improving statistical efficiency in estimating treatment effects for different outcome variables.
STATISTICAL METHODS IN MEDICAL RESEARCH
(2023)
Article
Statistics & Probability
Hengtao Zhang, Guosheng Yin, Donald B. Rubin
Summary: The Mahalanobis distance of covariate means is commonly used to achieve balance in rerandomization strategies. However, this criterion might not work well for high-dimensional cases as it treats all orthogonalized covariates equally. To address this issue, we propose using PCA to identify relevant subspaces for calculating Mahalanobis distance. PCA not only reduces dimensionality but also simplifies computation by focusing on the top orthogonal components. Our PCA rerandomization approach improves covariate balance and enhances the estimation of average treatment effects, as demonstrated in numerical studies with simulated and real data.
CANADIAN JOURNAL OF STATISTICS-REVUE CANADIENNE DE STATISTIQUE
(2023)
Article
Pharmacology & Pharmacy
Huaqing Jin, Guosheng Yin
Summary: Compared with existing phase I designs, the recently proposed calibration-free odds (CFO) design is robust, model-free, and easy to use. However, the original CFO design cannot handle late-onset toxicities commonly encountered in phase I oncology trials. To address this issue, we extend the CFO design to its time-to-event (TITE) version, which maintains the calibration-free and model-free properties. The TITE-CFO design performs robustly and efficiently compared to interval-based and model-based designs.
PHARMACEUTICAL STATISTICS
(2023)
Article
Health Care Sciences & Services
Jiaqi Gu, Yiwei Fan, Guosheng Yin
Summary: The restricted mean survival time (RMST) is widely used to summarize survival distribution due to its robustness and interpretation. In comparative studies, the RMST-based test is used as an alternative to the log-rank test. We developed an RMST-based omnibus Wald test to detect survival differences between two groups throughout the study follow-up period, considering multiple quantile-based time points.
STATISTICAL METHODS IN MEDICAL RESEARCH
(2023)
Article
Health Care Sciences & Services
Hengtao Zhang, Wen Su, Guosheng Yin
Summary: This article proposes a method called quasi-rerandomization, which improves covariate balance by rerandomizing observational covariates, thus approximating randomized experiments. Through extensive numerical studies, the method demonstrates competitive performance in terms of improving covariate balance and precision of treatment effect estimation.
BMC MEDICAL RESEARCH METHODOLOGY
(2023)
Article
Statistics & Probability
Xiaodong Yan, Guosheng Yin, Xingqiu Zhao
Summary: The proposed method offers a new approach to handling treatment heterogeneity with unknown grouping information, allowing for the identification of subgroup structures and estimation of subgroup-specific treatment effects simultaneously. By combining the Buckley-James iterative procedure and the alternating direction method of multipliers, the method demonstrates good performance in simulation studies and real-data applications.