Imputed Datasets Research Articles

Causal relationship between physical exercise and risk of ischemic stroke recurrence based on the potential outcome theory

To explore the causal relationship between physical exercise and ischemic stroke recurrence using a multiple imputation approach based on the potential outcomes framework. We collected data from 636 patients who were diagnosed with first-ever ischemic stroke between July, 2010 and December, 2018 at West China Hospital, Sichuan University.All the patients had motor ability and were followed up at least for 1 year.According to the intensity of physical exercise, the patients were divided into low-level physical exercise (LPE) group (n=244) and moderate-to high-intensity exercise (MHPE) group (n=392).With both the result and the time of recurrence as the potential outcomes, a multiple imputation method was used to impute the missing values based on a predictive matrix.Based on the imputed datasets, a causal inference model was built to estimate the average causal effect of physical exercise on the risk of stroke recurrence. Among the enrolled patients, 148 experienced recurrent stroke with a median recurrence time of 24.0 months and a cumulative recurrence rate of 23.3%during follow-up.As there were no outliers and the marginal distributions of the potential outcomes were basically consistent, both the Strip plots and Kenel density plots indicated that the imputed values could have been plausible measurements if they had not been missing.The average causal effect (ACE) of physical exercise on stroke recurrence was 0.578 (95%CI: 0.186-0.970, P=0.012). There is a causal relationship between physical exercise and the risk of ischemic stroke recurrence, indicating the value of moderate or high-level exercise after stroke in reducing the risk of stroke recurrence.

The development and validation of prognostic models for overall survival in the presence of missing data in the training dataset: a strategy with a detailed example

BackgroundThe United Kingdom Myeloma Research Alliance (UK-MRA) Myeloma Risk Profile is a prognostic model for overall survival. It was trained and tested on clinical trial data, aiming to improve the stratification of transplant ineligible (TNE) patients with newly diagnosed multiple myeloma. Missing data is a common problem which affects the development and validation of prognostic models, where decisions on how to address missingness have implications on the choice of methodology.MethodsModel buildingThe training and test datasets were the TNE pathways from two large randomised multicentre, phase III clinical trials. Potential prognostic factors were identified by expert opinion. Missing data in the training dataset was imputed using multiple imputation by chained equations. Univariate analysis fitted Cox proportional hazards models in each imputed dataset with the estimates combined by Rubin’s rules. Multivariable analysis applied penalised Cox regression models, with a fixed penalty term across the imputed datasets. The estimates from each imputed dataset and bootstrap standard errors were combined by Rubin’s rules to define the prognostic model.Model assessmentCalibration was assessed by visualising the observed and predicted probabilities across the imputed datasets. Discrimination was assessed by combining the prognostic separation D-statistic from each imputed dataset by Rubin’s rules.Model validationThe D-statistic was applied in a bootstrap internal validation process in the training dataset and an external validation process in the test dataset, where acceptable performance was pre-specified.Development of risk groupsRisk groups were defined using the tertiles of the combined prognostic index, obtained by combining the prognostic index from each imputed dataset by Rubin’s rules.ResultsThe training dataset included 1852 patients, 1268 (68.47%) with complete case data. Ten imputed datasets were generated. Five hundred twenty patients were included in the test dataset. The D-statistic for the prognostic model was 0.840 (95% CI 0.716–0.964) in the training dataset and 0.654 (95% CI 0.497–0.811) in the test dataset and the corrected D-Statistic was 0.801.ConclusionThe decision to impute missing covariate data in the training dataset influenced the methods implemented to train and test the model. To extend current literature and aid future researchers, we have presented a detailed example of one approach. Whilst our example is not without limitations, a benefit is that all of the patient information available in the training dataset was utilised to develop the model.Trial registrationBoth trials were registered; Myeloma IX-ISRCTN68454111, registered 21 September 2000. Myeloma XI-ISRCTN49407852, registered 24 June 2009.

Maximum Likelihood Multiple Imputation: Faster Imputations and Consistent Standard Errors Without Posterior Draws

Multiple imputation (MI) is a method for repairing and analyzing data with missing values. MI replaces missing values with a sample of random values drawn from an imputation model. The most popular form of MI, which we call posterior draw multiple imputation (PDMI), draws the parameters of the imputation model from a Bayesian posterior distribution. An alternative, which we call maximum likelihood multiple imputation (MLMI), estimates the parameters of the imputation model using maximum likelihood (or equivalent). Compared to PDMI, MLMI is faster and yields slightly more efficient point estimates. A past barrier to using MLMI was the difficulty of estimating the standard errors of MLMI point estimates. We derive, implement and evaluate three consistent standard error formulas: (1) one combines variances within and between the imputed datasets, (2) one uses the score function and (3) one uses the bootstrap with two imputations of each bootstrapped sample. Formula (1) modifies for MLMI a formula that has long been used under PDMI, while formulas (2) and (3) can be used without modification under either PDMI or MLMI. We have implemented MLMI and the standard error estimators in the mlmi and bootImpute packages for R.

Evaluating the impact of multivariate imputation by MICE in feature selection.

Demonstrate the positive impact of multivariate imputation in the feature selection process on datasets with missing values. We compared the effects of the feature selection process using complete datasets, incomplete datasets with missingness rates between 5 and 50%, and imputed datasets by basic techniques and multivariate imputation. The feature selection algorithms used are well-known methods. The results showed that the datasets imputed by multivariate imputation obtained the best results in feature selection compared to datasets imputed by basic techniques or non-imputed incomplete datasets. Considering the results obtained in the evaluation, applying multivariate imputation by MICE reduces bias in the feature selection process.

Elucidating Age and Sex-Dependent Association Between Frontal EEG Asymmetry and Depression: An Application of Multiple Imputation in Functional Regression

Frontal power asymmetry (FA), a measure of brain function derived from electroencephalography, is a potential biomarker for major depressive disorder (MDD). Though FA is functional in nature, it is typically reduced to a scalar value prior to analysis, possibly obscuring its relationship with MDD and leading to a number of studies that have provided contradictory results. To overcome this issue, we sought to fit a functional regression model to characterize the association between FA and MDD status, adjusting for age, sex, cognitive ability, and handedness using data from a large clinical study that included both MDD and healthy control (HC) subjects. Since nearly 40% of the observations are missing data on either FA or cognitive ability, we propose an extension of multiple imputation (MI) by chained equations that allows for the imputation of both scalar and functional data. We also propose an extension of Rubin’s Rules for conducting valid inference in this setting. The proposed methods are evaluated in a simulation and applied to our FA data. For our FA data, a pooled analysis from the imputed datasets yielded similar results to those of the complete case analysis. We found that, among young females, HCs tended to have higher FA over the θ, α, and β frequency bands, but that the difference between HC and MDD subjects diminishes and ultimately reverses with age. For males, HCs tended to have higher FA in the β frequency band, regardless of age. Young male HCs had higher FA in the θ and α bands, but this difference diminishes with increasing age in the α band and ultimately reverses with increasing age in the θ band. Supplementary materials for this article, including a standardized description of the materials available for reproducing the work, are available as an online supplement.

Sickle Cell Disorders and Severe COVID-19 Outcomes: A Cohort Study

Sickle Cell Disorders and Severe COVID-19 Outcomes: A Cohort Study

Predicting dissatisfaction following total hip arthroplasty using a Bayesian model averaging approach: Results from the Australian Arthroplasty Clinical Outcomes Registry National (ACORN).

Total hip arthroplasty (THA) provides excellent pain relief and improved function in patients with painful arthritis. The aim of this study was to identify rates and predictors of dissatisfaction following THA. Data were collected prospectively from the Australian Arthroplasty Clinical Outcomes Registry National (ACORN) database between 2014 and 2016 from 2096 patients who underwent THA. Data included baseline demographics, patient-reported outcome measures (PROMs) and postoperative clinical outcomes. Patients were dichotomized into two groups based on their 6-month response to the satisfaction question answered on a Likert scale. Eighteen predefined variables were analyzed. PROMs included full Oxford Hip Score, EQ-5D, and patient satisfaction. A Bayesian model averaging approach was used to build the best predictive model for dissatisfaction. Multiple logistic regression techniques were applied to quantify the effect size of the best model. At 6 months following THA, 95.4% of patients (n=2000) were satisfied with surgical outcome and 4.6% (n=96) were dissatisfied. The only variable that was significantly associated with dissatisfaction after THA was "complications after discharge." This result was consistent for both the complete and imputed dataset (odds ratio 4.78, 95% confidence interval 2.60-8.80, P < 0.001 and odds ratio 3.8, 95% confidence interval 2.60-5.60, P < 0.001, respectively). Our study confirms the high rates of patient satisfaction following THA, with postoperative complications being the only determinant of dissatisfaction. Optimization of patients prior to surgery, reducing postoperative complications, may further improve satisfaction rates after THA.

An application of Multiple Imputation under the Two Generalized Parametric Families

Multiple imputation under the multivariate normality assumption has often been regarded as a viable model-based approach in dealing with incomplete continuous data. Considering the fact that real data rarely conform with normality, there has been a growing attention to generalized classes of distributions that cover a broader range of skewness and elongation behavior compared to the normal distribution. In this regard, two recent works have shown that creating imputations under Fleishman’s power polynomials and the generalized lambda distribution may be a promising tool. In this article, essential distributional characteristics of these families are illustrated along with a description of how they can be used to create multiply imputed data sets. Furthermore, an application is presented using a data example from psychiatric research. Multiple imputation under these families that span most of the feasible area in the symmetry-peakedness plane appears to have substantial potential of capturing real missing-data trends that can be encountered in clinical practice.

Predicting microcystin occurrence in freshwater lakes and reservoirs: assessing environmental variables

ABSTRACT Determining the environmental conditions that influence the occurrence and concentration of the cyanobacterial toxin microcystin (MC) is a critical step for predicting cases in which the toxin will adversely affect drinking water sources, recreational waterbodies, and other freshwater ecosystems. Although widely studied, little consensus exists regarding the factors that influence MC on a global scale. The objective of this study was to identify the environmental variables most strongly associated with MC concentrations using observational data from lakes and reservoirs around the world while also addressing the substantial proportions of missing values that a large aggregated dataset often involves. A total of 124 studies containing data from an estimated 2040 lakes and reservoirs in 22 countries was used to construct a global dataset. Variables including <35% of non-missing observations were removed prior to analysis. Missing values for the remaining 12 predictors of MC were imputed using an iterative imputation algorithm based on a random forest approach. Variable selection was performed with generalized additive modeling on the complete case and imputed datasets. Models applied to the imputed data produced lower prediction errors than those fit to the complete dataset. Variables of greatest significance to MC concentration included location (longitude–latitude pairs), total nitrogen, turbidity, and pH. Total phosphorus was not found to be a strong predictor of MC. In addition to assisting water resource managers in protecting their waterbodies against MC, the presented methodologies may provide a useful framework for future water quality modeling while accounting for varying proportions of missing data.

Abstract 31: Genome-wide association study of African esophageal squamous cell carcinoma

Abstract Esophageal squamous cell carcinoma (ESCC) has a high incidence in sub-Saharan Africa and a poor prognosis, but little is known about the genetic contribution to ESCC susceptibility in this region. We carried out a GWAS in the South African Black (SAB) population with 1,686 ESCC cases and 3,217 controls as part of the Evolving Risk Factors for Cancer in African populations (ERICA-SA) study after ethical approval and informed consent. ESCC case samples from ERICA-SA and controls from the H3A AWI-Gen study were genotyped on the 2.3 million H3 Africa SNP Illumina array and additional SNPs imputed using the Wellcome Sanger Institute imputation service. Measures for correction of population structure included Eigen decomposition for Principal Component (PC) analysis. The final imputed dataset of 14.4M SNPs for 1,686 OSCC cases and 3,217 population controls was used in linear-mixed model analysis to test for association using GEMMA. A meta-analysis of SNP sets common to both our study and a GWAS of 2,013 Chinese ESCC cases and 2,701 controls was done using METAL. The GWAS of the SAB population identified a total of 62 SNPs with suggestive association with ESCC (P &amp;lt; 5x10-6), from 29 independent potential risk loci. A strong signal at genome-wide significance is located upstream of the FAM120A gene on chromosome 9 (rs1237966, p=4.58x10-8) and is supported by multiple SNPs at this locus. FAM120A encodes an RNA binding protein which is a critical component of oxidative stress-induced signalling. A strong association was also observed on chromosome 2, led by rs142741123 (p=5.49x10-8) in a region spanning the MYO1B and STAT4 genes with SNPs that are absent in non-African populations, suggesting that this is an African-specific risk locus for ESCC. There was limited support for the association of ESCC risk loci reported in other populations in our study. We did a meta-analysis of the SAB GWAS with a published GWAS of Chinese ESSC cases and controls with 4.85 million SNPs shared between the two studies. A total of 12 SNPs were associated with ESCC at genome-wide significance. These included 3 loci: chromosome 9, led by rs12379660 (p-valuemeta = 9.36x10-10), chromosome 10 led by rs7099485 (p-valuemeta = 1.48x10-8) and chromosome 22 led by rs1033667 (p-valuemeta = 1.47x10-9). The chromosome 9 locus at FAM120A is the novel locus identified in our African ESCC GWAS, the chromosome 10 locus at PLCE1 shows association only in the Chinese GWAS, and the chromosome 22 locus at CHEK2 was previously identified in the Chinese ESCC GWAS, and is supported further by the African ESCC GWAS. This study is, to our knowledge, the first cancer GWAS conducted in solely in a resident African population, and has detected two novel risk loci for ESCC at or near genome-wide significance, one of which is likely specific to African populations. It also demonstrates the power of trans-ethnic meta-analysis to identify common or distinct risk loci in populations of diverse ancestry. Citation Format: Wenlong C. Chen, Jean-Tristan Brandenburg, Ananyo Choudhury, Mahtaab Hayat, Dhriti Sengupta, Chantal Babb de Villiers, Lucien Ferndale, Cassandra Soo, Sang H. Lee, Charles Curtis, Robert Newton, Tim Waterboer, Frederic Sitas, Michele Ramsay, Christian C. Abnet, Colleen Aldous, Mohamed I. Parker, Elvira Singh, Deborah Bradshaw, Cathryn M. Lewis, Christopher G. Mathew. Genome-wide association study of African esophageal squamous cell carcinoma [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2021; 2021 Apr 10-15 and May 17-21. Philadelphia (PA): AACR; Cancer Res 2021;81(13_Suppl):Abstract nr 31.

The Leeds Risk Profile Predicts Outcome for Older Myeloma Patients: A Prognostic Score Using Frailty Biomarkers Derived and Tested Using 2371 Clinical Trial Patients

Introduction:Multiple myeloma (MM) is molecularly heterogeneous but so too are the patients it affects. This heterogeneity becomes more significant in older, less fit patients in whom factors affecting their ability to withstand treatment may be more important than disease biology. Treatment tolerability does not deteriorate linearly with age and is better defined as ‘frailty’. Patient-reported activities of daily living and comorbidities have been combined with age to derive the IMWG Frailty Score (Palumbo A et al, Blood 2015), which can predict outcome. This requires completion of 3 patient questionnaires and may be open to bias.Several biochemical markers of frailty have been associated with adverse outcomes including inflammatory markers such as CRP, renal dysfunction and the ratio of lymphocytes to total white cells (L:W). Using biochemical surrogates to measure patient frailty is objective and avoids the need for potentially burdensome questionnaires. Here we describe a novel prognostic score (Leeds Risk Profile; LRP) that combines readily available laboratory measures of patient frailty with age and WHO performance status (PS) to predict outcomes in transplant non-eligible (TNE) MM patients.Methods:1852 TNE patients recruited to the NCRI Myeloma XI (MyXI) trial were used as a training set and 520 TNE patients recruited to the MRC Myeloma IX (MyIX) trial used as a validation set. MyXI compared IMiD-containing triplet combinations: cyclophosphamide and dexamethasone with thalidomide (CTD) or lenalidomide (CRD). Patients with a suboptimal response (<VGPR) were eligible for further induction with a PI-based triplet (cyclophosphamide, bortezomib and dexamethasone). There was a maintenance randomisation between lenalidomide and observation. MyIX compared induction with melphalan and prednisolone to CTD and maintenance thalidomide to observation.Baseline variables considered included PS, LDH, CRP, international staging score (ISS), age and L:W. The primary objective was overall survival (OS). Missing data in the training set (MyXI) for the variables was imputed via multiple imputation by chained equations. The model was built using a Cox model with LASSO penalty within each imputed dataset. The final model was derived by combining the coefficients and standard errors by Rubin's rules. The model was internally validated by bootstrap sampling and externally validated in the test set using the prognostic separation D-statistic.Results:Univariate Cox-regression analysis demonstrated a significant effect of PS, CRP, ISS, Age and L:W (all p<0.001) but not LDH (p=0.333) on OS within the training set.The final multivariable model (LRP) included PS, Age, ISS and CRP. The discrimination and calibration of the model were good, with calibration better closer to trial entry. The model was successfully internally validated by bootstrap re-sampling according to pre-defined criteria. The combined prognostic index was divided into tertiles to define 3 groups: Fit, Intermediate-fitness, Frail. All variables increased in severity across the three groups. The groups predicted outcomes as shown in Figure 1a (median OS: Fit 60 months [95%CI 55-NR], Intermediate-fitness 44 [39-48] and Frail 25 [22-30] (Log-rank p<0.0001)).External validation was successfully carried out using patients from MyIX. The test data was split into three separate risk groups using the same cut-offs derived and implemented within the training data. Significant differences in OS were again predicted (Figure 1b) (median OS: Fit 43 months [35-NR], Intermediate-fitness 34 [27-38] and Frail 20 [14-25] (Log-rank p<0.0001)).In exploratory analyses the risk groups derived for overall survival were applied to progression-free survival, early mortality and treatment delivery modifications and predicted outcome as expected. Predictive ability was maintained in those patients who were exposed to both IMiD and PI therapy.Conclusion:We have defined a novel score predictive of outcome in TNE MM patients (LRP). This score uses only factors already collected at baseline for MM patients using objective laboratory bloods tests combined with age and a simple WHO PS assessment. As such it would be easy and quick to use in the clinic and apply within clinical trials. The LRP will be prospectively validated, including a comparison to the IMWG Frailty Score, within the UK-MRA Myeloma XIV trial (FITNESS). [Display omitted] DisclosuresCook:Janssen: Consultancy, Honoraria, Research Funding, Speakers Bureau; Sanofi: Consultancy, Honoraria, Speakers Bureau; Bristol-Myers Squibb: Consultancy, Honoraria; Glycomimetcs: Consultancy, Honoraria; Takeda: Consultancy, Honoraria, Research Funding, Speakers Bureau; Amgen: Consultancy, Honoraria, Research Funding, Speakers Bureau; Celgene: Consultancy, Honoraria, Research Funding, Speakers Bureau. Pawlyn:Takeda: Honoraria, Other: Travel support; Janssen: Other: Travel support; Celgene: Honoraria, Other: Travel support. Gregory:Janssen: Honoraria; Celgene: Consultancy, Honoraria. Davies:Takeda: Consultancy, Honoraria; Janssen: Consultancy, Honoraria; Celgene: Consultancy, Honoraria. Morgan:Takeda: Consultancy, Honoraria; Bristol Myers: Consultancy, Honoraria; Celgene: Consultancy, Honoraria, Research Funding. Jackson:Chugai: Honoraria; Amgen: Honoraria; Takeda: Honoraria; Celgene: Honoraria; J&J: Honoraria.

Validation of the Medical Research Council (MRC) and the Leukemia Research Foundation (LRF) Score for Hypomethylating Agents (HMAs) Use in Patients with Relapsed and Refractory Acute Myeloid Leukemia (RR-AML)

Validation of the Medical Research Council (MRC) and the Leukemia Research Foundation (LRF) Score for Hypomethylating Agents (HMAs) Use in Patients with Relapsed and Refractory Acute Myeloid Leukemia (RR-AML)

Early onset of overweight among children from low-income families: The role of exclusive breastfeeding and maternal intake of ultra-processed food.

There is an increasing prevalence of overweight during early childhood in the most impoverished areas in Brazil, although there is a lack of evidence regarding its onset. To investigate the incidence and risk factors associated with overweight among low-income children during their first year of life. We analysed data from a prospective birth cohort study conducted in Brazil, which followed-up children at birth, 3rd, 6th and 12th months (n=196). The overweight incidence (zBMI/A > 2SD) was analysed using the Kaplan-Meier survival estimator and its associated factors by the Coxs regression model. Missing data were addressed with multiple imputations and results on the final adjusted model were calculated by pooling the estimates generated for each imputed dataset. The overweight incidence was 17 events/100 children-year, and most cases occurred before the 6th month. The final model showed that children exclusively breastfed up to 30 days (vs. >30 days; HR 2.68; 95%CI 1.11-6.49) and whose mothers consumed ultra-processed foods more than 4 times/day (vs. ≤4 times/day; HR 3.02; 95%CI 1.28-7.13) presented a higher risk of developing overweight. The overweight incidence was high in this impoverished population. Shorter exclusive breastfeeding duration and an unhealthy household food environment provided a short-term risk increase.

Imputing missing patient-level data and propensity score matching in cost-effectiveness analysis in Crohn's disease

ABSTRACT Objectives The effect of imputing missing data followed by propensity score analysis on the incremental cost-effectiveness ratio (ICER) in a cost-effectiveness analysis is unknown. The objective was to compare alternative approaches in grouping data following imputation and prior to calculating propensity scores for use in economic evaluation. Methods Patient-level data from an observational study of 573 children with Crohn’s disease were used in a microsimulation model to determine the incremental cost of early anti-tumor necrosis factor-α treatment compared to standard care per remission week gained. Multiple imputation of a missing covariate followed by propensity score matching to create comparator groups was approached in two ways. The Within approach calculated propensity scores on each imputed dataset separately, while the Across method averaged propensity scores to create one matched population resulting in multiple sets of health state transition probabilities. Results The incremental cost per remission week gained ranged from CAD$2,236 to CAD$12,464 (mean CAD$4,266) with Within datasets and was CAD$4,679 per remission week gained with the Across dataset. Conclusion Imputation of missing patient-level data and propensity score analysis increases methodological uncertainty in cost-effectiveness analysis. The present study indicated that the Across approach may be less cumbersome, and slightly reduce bias and variance.

Effectiveness of a Yoga-Based Lifestyle Protocol (YLP) in Preventing Diabetes in a High-Risk Indian Cohort: A Multicenter Cluster-Randomized Controlled Trial (NMB-Trial).

IntroductionThough several lines of evidence support the utility of yoga-based interventions in diabetes prevention, most of these studies have been limited by methodological issues, primarily sample size inadequacy. Hence, we tested the effectiveness of yoga-based lifestyle intervention against diabetes risk reduction in multicentre, large community settings of India, through a single-blind cluster-randomized controlled trial, Niyantrita Madhumeha Bharat Abhiyan (NMB). Research Design and MethodsNMB-trial is a multicentre cluster-randomized trial conducted in 80 clusters [composed of rural units (villages) and urban units (Census Enumeration Blocks)] randomly assigned in a 1:1 ratio to intervention and control groups. Participants were individuals (age, 20–70 years) with prediabetes (blood HbA1c values in the range of 5.7–6.4%) and IDRS ≥ 60. The intervention included the practice of yoga-based lifestyle modification protocol (YLP) for 9 consecutive days, followed by daily home and weekly supervised practices for 3 months. The control cluster received standard of care advice for diabetes prevention. Statistical analyses were performed on an intention-to-treat basis, using available and imputed datasets. The primary outcome was the conversion from prediabetes to diabetes after the YLP intervention of 3 months (diagnosed based upon HbA1c cutoff >6.5%). Secondary outcome included regression to normoglycemia with HbA1c <5.7%.ResultsA total of 3380 (75.96%) participants were followed up at 3 months. At 3 months post-intervention, overall, diabetes developed in 726 (21.44%) participants. YLP was found to be significantly effective in halting progression to diabetes as compared to standard of care; adjusted RRR was 63.81(95% CI = 56.55–69.85). The YLP also accelerated regression to normoglycemia [adjusted Odds Ratio (adjOR) = 1.20 (95% CI, 1.02–1.43)]. Importantly, younger participants (≤40 years) were found to regress to normoglycemia more effectively than the older participants Pinteraction<0.001. ConclusionBased on the significant risk reduction derived from the large sample size, and the carefully designed randomized yoga-based intervention on high-risk populations, the study is a preliminary but strong proof-of-concept for yoga as a potential lifestyle-based treatment to curb the epidemic of diabetes. The observed findings also indicate a potential of YLP for diabetes prevention in low/moderate risk profile individuals that needs large-scale validation.Trial RegistrationClinical Trial Registration Number: CTRI/2018/03/012804.

How to Apply Multiple Imputation in Propensity Score Matching with Partially Observed Confounders: A Simulation Study and Practical Recommendations

Propensity score matching (PSM) has been widely used to mitigate confounding in observational studies, although complications arise when the covariates used to estimate the PS are only partially observed. Multiple imputation (MI) is a potential solution for handling missing covariates in the estimation of the PS. However, it is not clear how to best apply MI strategies in the context of PSM. We conducted a simulation study to compare the performances of popular non-MI missing data methods and various MI-based strategies under different missing data mechanisms. We found that commonly applied missing data methods resulted in biased and inefficient estimates, and we observed large variation in performance across MI-based strategies. Based on our findings, we recommend 1) estimating the PS after applying MI to impute missing confounders; 2) conducting PSM within each imputed dataset followed by averaging the treatment effects to arrive at one summarized finding; 3) a bootstrapped-based variance to account for uncertainty of PS estimation, matching, and imputation; and 4) inclusion of key auxiliary variables in the imputation model.

Evaluation of model fit in structural equation models with ordinal missing data: a comparison of the D 2 and MI2S methods

ABSTRACT Social science research often utilizes measurement instruments that generate ordinal data (e.g., Likert scales). Many empirical studies also face the challenge of missing data, which can be addressed by performing multiple imputation followed by analyses of the imputed datasets. However, when missing data exist on ordinal variables, there has been limited research on how to evaluate model fit of structural equation models for ordinal variables. Recent studies suggest that two multiple-imputation-based approaches show great promise: The D 2 procedure, and the Multiple Imputation Two-step (MI2S) approach, though the two have not been systematically compared. This study extends previous research by comparing the D 2 with the MI2S fit statistics in a wider range of conditions than previous studies. Our findings revealed a number of factors that can influence the performance of these test statistics.

Development and Validation of a Large Synthetic Cohort for the Study of Cardiovascular Health Across the Life Span.

We developed and validated a synthetic cohort approach to examine numbers of cardiovascular risk factors (CRFs) and adverse clinical events, including incident cardiovascular disease and all-cause mortality, across the life span from ages 20 years to 90 years. The current analysis included 40,875 participants from 7 large, population-based longitudinal epidemiologic studies (1948-2016). On the basis of a joint multilevel imputation model, we multiply imputed each participant's life-span numbers of CRFs and events using available records. To validate the imputed values, we partially removed the observed data and then compared the imputed and observed values. The complete life-span synthetic data set reflected the original observed data trends well. In our validation sample, the distributions of imputed CRFs and events were close to the observed distributions but with less variability. Bland-Altman plots indicated that there was a slightly negative trend in general, and the agreement bias was relatively small for the continuous CRFs. The hypothetical linear regression model suggested that the relationships between the CRFs and events were preserved in the imputed data set. This approach generated valid estimates of CRFs and events across the life span for African-American and White participants. The synthetic cohort may be sufficiently accurate to be useful in assessing the origins and timing of accumulating cardiovascular risk that can inform efforts to avoid cardiovascular disease development.

A Phase 3 Trial in Participants With Obesity Due to Bardet-Biedl Syndrome or Alström Syndrome: Efficacy and Safety of the Melanocortin 4 Receptor Agonist Setmelanotide

Introduction: This randomized Phase 3 trial evaluated the effect of setmelanotide, a melanocortin 4 receptor agonist, on weight loss, hunger reduction, and safety outcomes in individuals (aged ≥6 years) with obesity and a genetically confirmed diagnosis of Bardet-Biedl syndrome (BBS) or Alström syndrome (AS), conditions believed to disrupt hypothalamic leptin-melanocortin signaling. Methods: For inclusion, obesity was defined as body mass index ≥30 kg/m2 (in those aged ≥16 years) or weight >97th percentile (in those aged 6–15 years). Individuals were randomized and received setmelanotide or placebo for 14 weeks, followed by open-label setmelanotide so that all participants received at least 1 year of drug. Body weight, height, hunger scores, and treatment-emergent adverse events (AEs) were assessed. The primary endpoint was the proportion of participants (≥12 years) who achieved ≥10% reduction in body weight from baseline after 52 weeks of treatment. For statistical analysis, the primary endpoint had binomial proportions calculated for each of the 100 multiple imputed data sets, which were combined using Rubin’s Rule to compare against the null hypothesis with 95% confidence intervals (CIs) and P values. Efficacy analyses (including change in body weight, body mass index Z score, and hunger) were conducted in participants ≥12 years old at baseline. Safety analyses were conducted in all participants. Results: A total of 38 individuals with BBS (n=32) or AS (n=6) were enrolled. Five participants <12 years and 2 participants ≥12 years who discontinued before receiving active therapy were not included in the primary analysis. The prespecified significance cut points for the primary and key secondary endpoints were met. After ~52 weeks of setmelanotide, 34.5% (95% CI, 17.5%-51.6%; P=0.0024) of participants achieved ≥10% reduction in body weight from baseline. All observed responders had BBS. Mean ± SD percent change in body weight from baseline was −6.2% ± 8.6% (P<0.0001). In participants with BBS aged ≤17 years (n=14), mean ± SD percent change in body mass index Z score from baseline was −24.5% ± 22.3%. Mean ± SD percent Job #11307-1 1/27/2021Haws BBS AS Phase 3 ENDO 2021 EncorePage 2 change in maximal daily hunger score (based on participant responses to scoring their “most” hunger during the day) from baseline was −30.8% ± 25.0% (P<0.0001); 60.2% (95% CI, 35.3%-85.1%; P<0.0001) of participants achieved ≥25% reduction in weekly average daily hunger score from baseline. Common AEs included skin hyperpigmentation (57.9%), injection site erythema (44.7%), and nausea (34.2%). There was 1 serious treatment-related AE of anaphylactic reaction that occurred in a participant receiving placebo. Conclusions: In this Phase 3 trial in patients with BBS and AS, setmelanotide was associated with significant body weight and hunger reduction, with responses being greater in individuals with BBS.

Balancing Inferential Integrity and Disclosure Risk Via Model Targeted Masking and Multiple Imputation

There is a growing expectation that data collected by government-funded studies should be openly available to ensure research reproducibility, which also increases concerns about data privacy. A strategy to protect individuals’ identity is to release multiply imputed (MI) synthetic datasets with masked sensitivity values. However, information loss or incorrectly specified imputation models can weaken or invalidate the inferences obtained from the MI-datasets. We propose a new masking framework with a data-augmentation (DA) component and a tuning mechanism that balances protecting identity disclosure against preserving data utility. Applying it to a restricted-use Canadian Scleroderma Research Group (CSRG) dataset, we found that this DA-MI strategy achieved a 0% identity disclosure risk and preserved all inferential conclusions. It yielded 95% confidence intervals (CIs) that had overlaps of 98.5% (95.5%) on average with the CIs constructed using the full, unmasked CSRG dataset in a work-disability (interstitial lung disease) study. The CI-overlaps were lower for several other methods considered, ranging from 73.9% to 91.9% on average with the lowest value being 28.1%; such low CI-overlaps further led to some incorrect inferential conclusions. These findings indicate that the DA-MI masking framework facilitates sharing of useful research data while protecting participants’ identities. Supplementary materials for this article, including a standardized description of the materials available for reproducing the work, are available as an online supplement.