Article
Health Care Sciences & Services
Michele Santacatterina
Summary: In observational studies, covariate balance is crucial for obtaining unbiased estimates of treatment effects. Methods targeting covariate balance have been successfully proposed and widely applied for estimating treatment effects on continuous outcomes. However, in many medical and epidemiological applications, the focus is on estimating treatment effects on time-to-event outcomes.
STATISTICAL METHODS IN MEDICAL RESEARCH
(2023)
Article
Biochemical Research Methods
Mengbo Li, Gordon K. Smyth
Summary: Mass spectrometry proteomics in biomedical research suffers from the problem of missing values in peptides. Many analysis strategies have been proposed to distinguish different types of missing values and estimate detection probabilities. A logit-linear function is used to accurately model the detection probability, showing that missing values are related to peptide intensity. A probability model is developed to infer the distribution of unobserved intensities from observed values.
Article
Health Care Sciences & Services
Luis Antunes, Denisa Mendonca, Maria Jose Bento, Edmund Njeru Njagi, Aurelien Belot, Bernard Rachet
Summary: Missing data is a common issue in epidemiological databases, and multiple imputation has become a popular method for addressing it. By extending a new approach to reduce incompatibility between imputation and substantive models, the study validated the effectiveness of the method through simulation studies.
STATISTICAL METHODS IN MEDICAL RESEARCH
(2021)
Article
Mathematical & Computational Biology
Francisco J. J. Rubio, Hein Putter, Aurelien Belot
Summary: Unobserved individual heterogeneity is a common challenge in population cancer survival studies. We propose an individual excess hazard frailty model to account for individual heterogeneity and investigate its effects in the context of excess hazard models. The methodology is implemented in the R package IFNS and is illustrated through simulation and real-data examples.
STATISTICS IN MEDICINE
(2023)
Article
Public, Environmental & Occupational Health
Chao-Yu Guo, Ying-Chen Yang, Yi-Hau Chen
Summary: This study proposes four machine learning-based imputation strategies for survival data with different missing mechanisms, finding that non-parametric missForest is the only robust method under all missing mechanisms.
FRONTIERS IN PUBLIC HEALTH
(2021)
Article
Environmental Sciences
B. B. Cael, P. Goodwin, C. R. Pearce, D. Stainforth
Summary: All pathways to achieve the Paris Agreement target of limiting global warming require large-scale removal of CO2 from the atmosphere. Different CO2 removal strategies vary in price, storage timescale, and other factors. It is important to assess whether the benefits of deploying these strategies outweigh their costs and to understand how costs depend on socioeconomic assumptions. This study provides a framework to quantitatively evaluate and compare different strategies, guiding future research, development, and policy efforts.
ENVIRONMENTAL RESEARCH LETTERS
(2023)
Article
Mathematical & Computational Biology
Chia-Rui Chang, Yue Song, Fan Li, Rui Wang
Summary: Covariate adjustment is important in analyzing data from randomized clinical trials, but missing data can be a barrier. This study reviews different covariate adjustment methods with incomplete covariate data. The researchers propose a weighting approach that combines inverse probability weighting and overlap weighting to adjust for missing outcomes and covariates, and conduct comprehensive simulation studies to evaluate the performance of the methods.
STATISTICS IN MEDICINE
(2023)
Article
Ecology
Shinichi Nakagawa, Daniel W. A. Noble, Malgorzata Lagisz, Rebecca Spake, Wolfgang Viechtbauer, Alistair M. Senior
Summary: The log response ratio (lnRR) is commonly used in ecology meta-analysis, but missing standard deviations (SDs) pose a challenge in estimating the sampling variance. We propose a new method using weighted average coefficient of variation (CV) from studies reporting SDs to address this issue. Our results show that using the average CV to estimate sampling variances for all observations, regardless of missingness, performs better than the conventional approach using individual study-specific CV with complete data. This approach is broadly applicable and can be implemented in all lnRR meta-analyses.
Article
Mathematical & Computational Biology
Chiu-Hsieh Hsu, Yulei He, Chengcheng Hu, Wei Zhou
Summary: Missing covariate problems often occur in biomedical and electrical medical record data studies, making it difficult to evaluate the relationship between a biomarker and a clinical outcome when biomarker data is missing for some subjects. This paper proposes a sensitivity analysis approach using a nonparametric multiple imputation strategy to handle missingness that may not be random (MNAR). The proposed approach is expected to be robust and produces plausible regression coefficient estimates according to simulation results. It is also applied to assess the impact of MNAR on the relationship between post-operative outcomes and incomplete pre-operative Hemoglobin A1c level in patients with advanced atherosclerotic disease who underwent carotid intervention.
STATISTICS IN MEDICINE
(2023)
Article
Mathematical & Computational Biology
Corentin Segalas, Clemence Leyrat, James R. Carpenter, Elizabeth Williamson
Summary: One popular method for addressing confounding in causal inference is propensity score matching, which matches treated patients with untreated patients based on similar propensity scores. Multiple imputation is often used for handling missing data. However, combining propensity score matching and multiple imputation can result in over-coverage of the confidence interval for the treatment effect estimate. In this article, the authors investigate the cause of this over-coverage and propose a correction to remove it.
STATISTICS IN MEDICINE
(2023)
Article
Economics
Zachary Parolin, Megan Curran, Jordan Matsudaira, Jane Waldfogel, Christopher Wimer
Summary: This study introduces a framework to produce monthly estimates of the Supplemental Poverty Measure and official poverty measure based on monthly income, arguing that a shorter accounting period and more timely estimates of poverty can better capture income volatility and inform the public about current economic conditions. The framework is validated using data from the Current Population Survey and reflects trends in poverty rates from 2004 to 2016 and during the COVID-19 pandemic. The study finds that monthly poverty rates generally declined in the 1990s, increased in the 2000s, and declined again after the Great Recession. However, within-year variation in monthly poverty rates has generally increased, particularly among families with children.
JOURNAL OF POLICY ANALYSIS AND MANAGEMENT
(2022)
Article
Mathematics
Shashi Bhushan, Abhay Pratap Pandey
Summary: This article discusses new chain imputation methods using two auxiliary variables under the MCAR approach, testing them for optimality in terms of MSE. The proposed methods are seen as efficient extensions to previous works and show promising results in comparison with conventional chain-type imputation methods.
COMMUNICATIONS IN MATHEMATICS AND STATISTICS
(2023)
Article
Public, Environmental & Occupational Health
Chinchin Wang, Tyrel Stokes, Russell J. Steele, Niels Wedderkopp, Ian Shrier
Summary: Researchers demonstrated that random hot deck imputation can achieve plausible multiple imputation in longitudinal studies, serving as an alternative method when model-based approaches are infeasible.
CLINICAL EPIDEMIOLOGY
(2022)
Article
Psychology, Mathematical
Samantha F. Anderson
Summary: Reporting standardized effects is important in randomized treatment studies. However, estimating the standardized average treatment effect (sATE) accurately becomes challenging when outcome data are missing. In this study, different strategies for handling missing data were compared using a simulation study, and recommendations for improving estimation were provided.
BEHAVIOR RESEARCH METHODS
(2022)
Article
Mathematical & Computational Biology
Francisco J. Rubio, Bernard Rachet, Roch Giorgi, Camille Maringe, Aurelien Belot
Summary: In cancer epidemiology, regression models for excess mortality hazard are commonly used to estimate cancer survival and describe the association between prognosis factors and excess mortality. However, expected mortality rates from general population life tables may not be sufficiently stratified, leading to potential biases in estimation of regression parameters. Proposed parametric corrections in excess hazard regression models show good statistical performance in simulation studies and are successfully applied to real population-based data of lung cancer patients. Suggestions and limitations for practical use are also provided.
Editorial Material
Pediatrics
David Tappin, Edwin A. Mitchell, James Carpenter, Fern Hauck, Lynsay Allan
ARCHIVES OF DISEASE IN CHILDHOOD
(2023)
Article
Mathematical & Computational Biology
Matthew J. Smith, Mohammad A. Mansournia, Camille Maringe, Paul N. Zivich, Stephen R. Cole, Clemence Leyrat, Aurelien Belot, Bernard Rachet, Miguel A. Luque-Fernandez
Summary: The main purpose of medical studies is to estimate treatment effects, but sometimes randomization is not possible; observational studies are used in such cases. Challenges in observational studies include confounding, which is typically controlled by adjusting measured confounders; recent advances in causal inference have focused on addressing confounding. However, a lack of computational tutorials has caused some confusion for researchers using these methods.
STATISTICS IN MEDICINE
(2022)
Article
Oncology
S. F. Lee, B. A. Vellayappan, L. C. Wong, C. L. Chiang, S. K. Chan, E. Y-F Wan, I. C-K Wong, P. C. Lambert, B. Rachet, A. K. Ng, M. A. Luque-Fernandez
Summary: In an Asian population-based cohort, it was found that incident cardiovascular diseases (CVDs) can occur soon after treatment for diffuse large B-cell lymphoma (DLBCL) and continue to occur throughout survivorship. Age, hypertension, diabetes, and baseline use of aspirin were associated with an increased risk of incident CVD. In a subgroup of higher-risk patients, the time in the CVD state was relatively short, with other causes of death surpassing DLBCL-related death after about 5 years.
Article
Mathematical & Computational Biology
Brennan C. Kahan, Tim P. Morris, Beatriz Goulao, James Carpenter
Summary: Factorial trials offer an efficient method to evaluate multiple interventions, but careful handling of additional treatments is necessary to avoid misinterpretation of results.
STATISTICS IN MEDICINE
(2022)
Article
Mathematical & Computational Biology
Corentin Segalas, Clemence Leyrat, James R. Carpenter, Elizabeth Williamson
Summary: One popular method for addressing confounding in causal inference is propensity score matching, which matches treated patients with untreated patients based on similar propensity scores. Multiple imputation is often used for handling missing data. However, combining propensity score matching and multiple imputation can result in over-coverage of the confidence interval for the treatment effect estimate. In this article, the authors investigate the cause of this over-coverage and propose a correction to remove it.
STATISTICS IN MEDICINE
(2023)
Article
Oncology
Kueshivi Midodji Atsou, Bernard Rachet, Edouard Cornet, Marie-Lorraine Chretien, Cedric Rossi, Laurent Remontet, Laurent Roche, Roch Giorgi, Sophie Gauthier, Stephanie Girard, Johann Bockle, Stephane Kroudia Wasse, Helene Rachou, Laila Bouzid, Jean-Marc Poncet, Sebastien Orazio, Alain Monnereau, Xavier Troussard, Morgane Mounier, Marc Maynadie
Summary: This study examines the care pathways of acute myeloblastic leukemia patients in France and found that different pathways could lead to unequal access to treatment, resulting in excess mortality. Referral from a general practitioner to a specialized hematology unit is the most common care pathway. Age, medical condition, diagnostic evaluations, and the type of hospital also influence whether patients can receive treatment at a specialized hematology unit.
Article
Oncology
Laure Tron, Laurent Remontet, Mathieu Fauvernier, Bernard Rachet, Aurelien Belot, Ludivine Launay, Ophelie Merville, Florence Molinie, Olivier Dejardin, Guy Launoy
Summary: The study found that the social gradient in cancer net survival could be partially explained by socially-determined co-morbidities, and using simulated deprivation-specific life tables reduced the social gradient. However, the lack of deprivation-specific life tables in the original data overestimated the social gradient in cancer net survival, leading to potential inaccuracies in conclusions. This highlights the importance of creating proper deprivation-specific life tables for accurate analysis of social inequalities in cancer net survival.
Article
Clinical Neurology
Tom Foltynie, Sonia Gandhi, Cristina Gonzalez-Robles, Marie-Louise Zeissler, Georgia Mills, Roger Barker, James Carpenter, Anette Schrag, Anthony Schapira, Oliver Bandmann, Stephen Mullin, Joy Duffen, Kevin McFarthing, Jeremy Chataway, Mahesh Parmar, Camille Carroll
Summary: Multi-arm, multi-stage platform designs have improved the efficiency of clinical trials in the field of oncology. Foltynie et al. discuss the challenges and considerations of using this approach to assess potential disease-modifying treatments in progressive neurological conditions such as Parkinson's disease.
Article
Medicine, Research & Experimental
Sunita Rehal, Suzie Cro, Patrick P. J. Phillips, Katherine Fielding, James R. Carpenter
Summary: This article discusses statistical methods for handling intercurrent events and missing values in non-inferiority studies. Using a tuberculosis clinical trial as a case study, the authors propose primary and additional estimands suitable for non-inferiority studies. Multiple imputation methods are used for estimation, and the results show that these methods provide a more accurate interpretation of the estimand.
Article
Health Care Sciences & Services
Laura Viviani, Ian R. White, Elizabeth J. Williamson, James Carpenter, Jan van der Meulen, David A. Cromwell
Summary: This study evaluated the performance of the DetectDeviatingCells (DDC) algorithm in detecting data anomalies at the observation and variable level in continuous variables. The DDC algorithm showed promising results in improving error detection processes for observational data, particularly in detecting complex error patterns.
JOURNAL OF CLINICAL EPIDEMIOLOGY
(2023)
Article
Health Care Sciences & Services
Elinor Curnow, James R. Capenter, Jon E. Heron, Rosie P. Cornish, Stefan Rach, Vanessa Didelez, Malte Langeheine, Kate Tilling
Summary: Epidemiological studies often have missing data, which are commonly handled by multiple imputation (MI). This study examines the bias caused by the default option of using simple linear covariate functions in the imputation model and provides practical guidance for researchers. The results show that mis-specification of the relationship between outcome and exposure, or between exposure and confounder, can cause bias in MI estimates, and the method of predictive mean matching can mitigate model mis-specification.
JOURNAL OF CLINICAL EPIDEMIOLOGY
(2023)
Letter
Mathematical & Computational Biology
Tim P. Morris, Ian R. White, Suzie Cro, Jonathan W. Bartlett, James R. Carpenter, Tra My Pham
Summary: For simulation studies evaluating methods of handling missing data, generating partially observed data by fixing complete data and simulating missingness indicators repeatedly is rarely appropriate.
BIOMETRICAL JOURNAL
(2023)
Article
Health Care Sciences & Services
Mia S. Tackney, Elizabeth Williamson, Derek G. Cook, Elizabeth Limb, Tess Harris, James Carpenter
Summary: Clinical trials often use accelerometers to measure step count at a granular level, but missing data is common due to non-compliance. Existing approaches to handling missing data in the literature lead to loss of information on the time of day when data are missing. We propose a finer epoch-level approach and two methods for handling missingness using multiple imputation, showing that the non-parametric approach yields the least biased estimates of treatment effect with small standard errors.
STATISTICAL METHODS IN MEDICAL RESEARCH
(2023)
Article
Mathematical & Computational Biology
Rebecca M. Turner, Tim Band, Tim P. Morris, David J. Fisher, Julian P. T. Higgins, James R. Carpenter, Ian R. White
Summary: In this study, new local and global tests for inconsistency in network meta-analysis were proposed and applied to three example networks. The models were designed to handle treatments symmetrically and locate inconsistency in loops rather than in nodes or treatment comparisons. The global model showed potential for increased power compared to existing approaches.
STATISTICS IN MEDICINE
(2023)
Article
Health Care Sciences & Services
Michael R. Elliott, Orlagh Carroll, Richard Grieve, James Carpenter
Summary: Randomized controlled trials are considered the gold standard for assessing causal effects, but may still be biased in estimating population-level causal effects. Recent research suggests that incorporating information from probability samples can improve population causal inference in randomized controlled trials. This paper reviews recent work on transporting causal effect estimates from trials to populations, and proposes estimators using inverse probability weighting or prediction methods. The proposed methods do not require specific functional form or interaction, and can accommodate unequal probability of selection in benchmark or population samples.
STATISTICAL METHODS IN MEDICAL RESEARCH
(2023)