Article
Statistics & Probability
Rune Christiansen, Matthias Baumann, Tobias Kuemmerle, Miguel D. Mahecha, Jonas Peters
Summary: This work proposes a causal model for studying the influence of armed conflict on tropical forest loss. Through analysis of geospatial information on conflict events and forest loss in Colombia, the study finds that the overall effect is slightly negative and insignificant at the national level, while positive and negative effects can be observed at the provincial level.
JOURNAL OF THE AMERICAN STATISTICAL ASSOCIATION
(2022)
Article
Mathematical & Computational Biology
Satoshi Hattori, Sho Komukai, Tim Friede
Summary: In randomized clinical trials, incorporating baseline covariates can improve the power in hypothesis testing for treatment effects. The Cox proportional hazards model with baseline covariates as explanatory variables can improve the standard logrank test. We propose a simple strategy for sizing randomized clinical trials utilizing historical data and derive a power formula for the augmented logrank test.
STATISTICS IN MEDICINE
(2022)
Article
Mathematical & Computational Biology
Elisabeth Coart, Perrine Bamps, Emmanuel Quinaux, Genevieve Sturbois, Everardo D. Saad, Tomasz Burzykowski, Marc Buyse
Summary: In randomized trials, the comparability of treatment groups is ensured through random allocation of treatments. However, imbalances in prognostic factors among treatment groups may occur. To achieve balance, some procedures dynamically adapt allocation using covariate information. This tutorial compares the performance of minimization, a popular covariate-adaptive procedure, with two other commonly used procedures.
STATISTICS IN MEDICINE
(2023)
Article
Nutrition & Dietetics
Christian Ritz
Summary: This tutorial provides a detailed introduction to the statistical analysis of parallel-arm RCTs in nutrition, focusing on how trial design and other factors may influence subsequent statistical analysis. It covers all steps of the statistical analysis and includes a practical example.
EUROPEAN JOURNAL OF CLINICAL NUTRITION
(2021)
Article
Health Care Sciences & Services
Yang Li, Wei Ma, Yichen Qin, Feifang Hu
Summary: Concerns have been raised about the validity of statistical inference with covariate-adaptive randomization in clinical trials, especially for continuous responses. This study examines the unadjusted test for treatment effect under generalized linear models and covariate-adaptive randomization, and proposes an adjustment method to achieve a valid size based on asymptotic results. Numerical studies confirm the theoretical findings and demonstrate the effectiveness of the proposed adjustment method.
STATISTICAL METHODS IN MEDICAL RESEARCH
(2021)
Article
Statistics & Probability
Xiudi Li, Sijia Li, Alex Luedtke
Summary: We introduce a framework to utilize external data for assessing the efficiency of covariate-adjusted estimators compared to unadjusted estimators in future randomized trials. The relative efficiencies obtained approximate the required sample size ratio for desired statistical power. We develop semiparametrically efficient estimators for various treatment effect estimands of interest, allowing for flexible statistical learning methods to estimate the nuisance functions. We propose a Wald-type confidence interval and a double bootstrap scheme for statistical inference. Simulation studies demonstrate the performance of the proposed methods, and they are applied to estimate the efficiency gain of covariate adjustment in Covid-19 therapeutic trials.
JOURNAL OF THE ROYAL STATISTICAL SOCIETY SERIES B-STATISTICAL METHODOLOGY
(2023)
Article
Computer Science, Interdisciplinary Applications
Wei Ma, Xiaoqing Ye, Fuyi Tu, Feifang Hu
Summary: carat is an R package that allows for the implementation and evaluation of covariate-adaptive randomization methods. It provides a variety of randomization and testing procedures and includes comprehensive evaluation and comparison tools. carat also allows for power analysis to assist in the planning of covariate-adaptive clinical trials.
JOURNAL OF STATISTICAL SOFTWARE
(2023)
Article
Medicine, Research & Experimental
Siyun Yang, Fan Li, Laine E. Thomas, Fan Li
Summary: Propensity score weighting methodology is developed to improve precision and power of subgroup analyses in randomized clinical trials. Simulation results show that adjusted estimators have smaller standard errors than unadjusted estimators, and weighting estimators with full-interaction propensity model consistently outperform standard main-effect propensity model.
Review
Cardiac & Cardiovascular Systems
Leah Pirondini, John Gregson, Ruth Owen, Tim Collier, Stuart Pocock
Summary: This article reviews the current practice of covariate adjustment in cardiovascular trials published in major medical journals in 2019. The study finds that contemporary cardiovascular trials do not make best use of covariate adjustment, and that more frequent use could lead to improvements in the efficiency of future trials.
JACC-HEART FAILURE
(2022)
Article
Mathematical & Computational Biology
Hongxiang Qiu, Andrea J. Cook, Jennifer F. Bobb
Summary: Generalized linear mixed models (GLMM) are widely used for analyzing clustered data, but standard statistical tests may have elevated type I error rates when the number of clusters is small to moderate. It remains unknown which tests are appropriate for count outcomes or covariate-adjusted models.
STATISTICS IN MEDICINE
(2023)
Article
Biology
Tong Wang, Wei Ma
Summary: This paper examines the impact of misclassification on covariate-adaptive randomized trials, showing that superior covariate balance can still be achieved compared to complete randomization even with misclassified covariates. Additionally, it is found that the two sample t-test is conservative with reduced Type I error, but this can be corrected using a bootstrap method. Adjusting misclassified covariates in the model used for analysis can maintain nominal Type I error and increase power. These results support the use of covariate-adaptive randomization in clinical trials, even in the presence of covariate misclassification.
Article
Clinical Neurology
Dale J. Langford, Sonia Sharma, Michael P. Mcdermott, Avinash Beeram, Soroush Besherat, Fallon O. France, Remington Mark, Meghan Park, Mahd Nishtar, Dennis C. Turk, Robert H. Dworkin, Jennifer S. Gewandter
Summary: The statistical analysis of baseline factors in chronic pain RCTs is inconsistent, and prespecified adjustments for baseline covariates can improve accuracy and assay sensitivity.
Review
Health Care Sciences & Services
Pascale Nevins, Kendra Davis-Plourde, Jules Antoine Pereira Macedo, Yongdong Ouyang, Mary Ryan, Guangyu Tong, Xueqi Wang, Can Meng, Luis Ortiz-Reyes, Fan Li, Agnes Caille, Monica Taljaard
Summary: This study investigates the methods of randomization and reporting of balance at baseline in stepped-wedge cluster randomized trials (SW-CRTs). The findings suggest that most trials use unrestricted allocation for cluster randomization and there are limitations in reporting balance at baseline. The authors recommend that researchers need more guidance on methods of randomization and assessment of baseline balance.
JOURNAL OF CLINICAL EPIDEMIOLOGY
(2023)
Article
Clinical Neurology
Dale J. Langford, Raissa Lou, Soun Sheen, Dagmar Amtmann, Luana Colloca, Robert R. Edwards, John T. Farrar, Nathaniel P. Katz, Michael P. McDermott, Bryce B. Reeve, Ajay D. Wasan, Dennis C. Turk, Robert H. Dworkin, Jennifer S. Gewandter
Summary: Variability in pain outcomes can hinder the sensitivity of chronic pain clinical trials. Participants' expectations may contribute to this variability and impede the development of new pain treatments. Measurement and management of expectations in clinical trials need to be optimized and standardized. This article provides an overview of research findings on the relationship between baseline expectations and pain outcomes, emphasizing the potential benefit of adjusting for participants' expectations in trial analyses.
Article
Statistics & Probability
Leonard Henckel, Emilija Perkovic, Marloes H. Maathuis
Summary: This article introduces a graphical method for covariate adjustment to estimate the total causal effect. By comparing the asymptotic variances provided by different valid adjustment sets, a simple variance decreasing pruning procedure and a graphical characterization of the optimal asymptotic variance are proposed. These results are applicable to various graphical structures, not only causal linear models.
JOURNAL OF THE ROYAL STATISTICAL SOCIETY SERIES B-STATISTICAL METHODOLOGY
(2022)
Article
Biology
Wen Wei Loh, Beatrijs Moerkerke, Tom Loeys, Stijn Vansteelandt
Summary: This article addresses the problem of decomposing the effects along multiple causal pathways from a treatment to an outcome with multiple possible mediators. It proposes a novel estimation strategy using nonparametric estimates of the mediator distributions to avoid the need for modeling the joint mediator distribution, providing more flexibility and overcoming limitations of current estimation approaches.
Article
Statistics & Probability
Stijn Vansteelandt, Oliver Dukes
Summary: Inference for parameters in generalised linear models is usually based on specified assumptions, but often there is excess uncertainty due to the data-adaptive model selection process. We propose novel nonparametric definitions that capture the association and interaction between variables even when the models are misspecified.
JOURNAL OF THE ROYAL STATISTICAL SOCIETY SERIES B-STATISTICAL METHODOLOGY
(2022)
Editorial Material
Statistics & Probability
Stijn Vansteelandt, Oliver Dukes
JOURNAL OF THE ROYAL STATISTICAL SOCIETY SERIES B-STATISTICAL METHODOLOGY
(2022)
Article
Public, Environmental & Occupational Health
Vanessa Gorasso, Isabelle Moyersoen, Johan Van der Heyden, Karin De Ridder, Stefanie Vandevijvere, Stijn Vansteelandt, Delphine De Smedt, Brecht Devleesschauwer
Summary: The study in Belgium estimated the annual health care and lost productivity costs associated with excess weight among the adult population. It found that 48.6% of adults in Belgium were affected by overweight or obesity, with significant medical costs and higher prevalence of chronic conditions compared to normal weight individuals.
Article
Statistics & Probability
Stijn Vansteelandt, Oliver Dukes, Kelly Van Lancker, Torben Martinussen
Summary: In this study, we propose a nonparametric estimand that accurately captures the conditional association of interest even in the case of model misspecification. This assumption-lean inference approach based on the influence function under the nonparametric model allows the use of data-adaptive procedures.
JOURNAL OF THE AMERICAN STATISTICAL ASSOCIATION
(2022)
Article
Mathematical & Computational Biology
Pawel Morzywolek, Johan Steen, Wim Van Biesen, Johan Decruyenaere, Stijn Vansteelandt
Summary: The optimal timing for initiating renal replacement therapy in patients with acute kidney injury remains challenging. This study explores different timing strategies based on serum potassium, pH, and fluid balance, using routinely collected observational data. Statistical techniques are employed to evaluate the impact of dynamic treatment regimes, considering ICU discharge as a competing event. Two approaches - nonparametric and semiparametric - are discussed, along with a cross-validation technique to assess out-of-sample performance.
STATISTICS IN MEDICINE
(2022)
Article
Biology
Linbo Wang, Eric Tchetgen Tchetgen, Torben Martinussen, Stijn Vansteelandt
Summary: This paper provides a response to the comments received on our paper titled "Instrumental variable estimation of the causal hazard ratio."
Article
Biology
Linbo Wang, Eric Tchetgen Tchetgen, Torben Martinussen, Stijn Vansteelandt
Summary: Cox‘s proportional hazards model is widely used in statistical analysis to evaluate the relationship between exposure and a censored failure time outcome. However, when confounding factors are not fully observed, the estimated hazard ratio can be biased due to unmeasured confounding. In this paper, we propose a novel approach using a binary instrumental variable to identify and estimate the causal hazard ratio, while accounting for unmeasured confounding. Our approach provides a consistent estimator for the causal hazard ratio, and we derive its asymptotic distribution and an estimator for its asymptotic variance. We demonstrate the effectiveness of our approach through simulation studies and a real data application.
Article
Statistics & Probability
Tyler J. VanderWeele, Stijn Vansteelandt
Summary: Factor analysis is commonly used to evaluate the relationship between a set of indicators and a latent construct. This paper introduces a statistical test to examine the structural interpretation of a latent factor model and applies it to the association between life satisfaction and mortality rates.
JOURNAL OF THE ROYAL STATISTICAL SOCIETY SERIES B-STATISTICAL METHODOLOGY
(2022)
Article
Biology
Kelly Van Lancker, Oliver Dukes, Stijn Vansteelandt
Summary: The problem of variable selection for confounding adjustment is a key challenge in evaluating exposure effects in observational studies. Current routine procedures do not guarantee adequate performance in delivering exposure effect estimators and confidence intervals. This study proposes a novel procedure using penalized Cox regression to overcome the problem of confounding variables for survival data. The proposed methods yield valid inferences even with high-dimensional covariates, as shown by simulation results.
Article
Health Care Sciences & Services
Ingrid Pelgrims, Brecht Devleesschauwer, Stefanie Vandevijvere, Eva M. De Clercq, Stijn Vansteelandt, Vanessa Gorasso, Johan van der Heyden
Summary: This study aimed to assess the agreement between self-reported and measured height, weight, hypertension and hypercholesterolemia and to identify an adequate approach for valid measurement error correction. The study confirmed the underestimation of risk factor prevalence based on self-reported data and found that random-forest multiple imputation is the method of choice for correcting this bias.
BMC MEDICAL RESEARCH METHODOLOGY
(2023)
Article
Mathematical & Computational Biology
Ruth H. Keogh, Jon Michael Gran, Shaun R. Seaman, Gwyneth Davies, Stijn Vansteelandt
Summary: This article discusses how longitudinal observational data can be used to study the causal effects of time-varying treatments on time-to-event outcomes. It compares two methods: marginal structural models (MSM) and sequential trials approach, and discusses their assumptions and differences in data usage. Simulation study shows that the sequential trials approach is more efficient in estimating the marginal risk difference in most follow-up times, but this efficiency may be reversed at later time points and relies on modelling assumptions. The methods are applied to estimate the effect of dornase alfa on survival in longitudinal observational data from the UK Cystic Fibrosis Registry.
STATISTICS IN MEDICINE
(2023)
Article
Public, Environmental & Occupational Health
Vanessa Gorasso, Johan van der Heyden, Robby De Pauw, Ingrid Pelgrims, Eva M. De Clercq, Karin De Ridder, Stefanie Vandevijvere, Stijn Vansteelandt, Bert Vaes, Delphine De Smedt, Brecht Devleesschauwer
Summary: This study analyzes the health and economic impact of musculoskeletal disorders in Belgium, including low back pain, neck pain, osteoarthritis, and rheumatoid arthritis. The results show that these disorders have a significant burden on both health and economy in Belgium.
POPULATION HEALTH METRICS
(2023)
Article
Mathematical & Computational Biology
Erin E. Gabriel, Michael C. Sachs, Torben Martinussen, Ingeborg Waernbaum, Els Goetghebeur, Stijn Vansteelandt, Arvid Sjolander
Summary: There is a concerning trend in the applied literature to believe that the combination of a propensity score and an adjusted outcome model automatically results in a doubly robust estimator and to misuse more complex established doubly robust estimators. This article introduces a simple alternative method and aims to help readers understand why it is doubly robust. It also provides examples and code for implementation.
STATISTICS IN MEDICINE
(2023)
Article
Critical Care Medicine
Pawel Morzywolek, Johan Steen, Stijn Vansteelandt, Johan Decruyenaere, Sigrid Sterckx, Wim Van Biesen
Summary: This study aimed to compare the impact of different dynamic treatment regimes (DTRs) on the initiation of renal replacement therapy (RRT) in patients with acute kidney injury (AKI). Through causal analysis of the collected data, it was found that delaying RRT until specific thresholds of potassium, pH, and urinary output can reduce the 30-day ICU mortality.