Limitations in Using Multiple Imputation to Harmonize Individual Participant Data for Meta-Analysis

Individual participant data (IPD) meta-analysis is a meta-analysis in which the individual-level data for each study are obtained and used for synthesis. A common challenge in IPD meta-analysis is when variables of interest are measured differently in different studies. The term harmonization has been coined to describe the procedure of placing variables on the same scale in order to permit pooling of data from a large number of studies. Using data from an IPD meta-analysis of 19 adolescent depression trials, we describe a multiple imputation approach for harmonizing 10 depression measures across the 19 trials by treating those depression measures that were not used in a study as missing data. We then apply diagnostics to address the fit of our imputation model. Even after reducing the scale of our application, we were still unable to produce accurate imputations of the missing values. We describe those features of the data that made it difficult to harmonize the depression measures and provide some guidelines for using multiple imputation for harmonization in IPD meta-analysis.     

Individual Participant Data Meta-Analysis of Mechanical Workplace Risk Factors and Low Back Pain

Objectives. We used individual participant data from multiple studies to conduct a comprehensive meta-analysis of mechanical exposures in the workplace and low back pain.Methods. We conducted a systematic literature search and contacted an author of each study to request their individual participant data. Because outcome definitions and exposure measures were not uniform across studies, we conducted 2 substudies: (1) to identify sets of outcome definitions that could be combined in a meta-analysis and (2) to develop methods to translate mechanical exposure onto a common metric. We used generalized estimating equation regression to analyze the data.Results. The odds ratios (ORs) for posture exposures ranged from 1.1 to 2.0. Force exposure ORs ranged from 1.4 to 2.1. The magnitudes of the ORs differed according to the definition of low back pain, and heterogeneity was associated with both study-level and individual-level characteristics.Conclusions. We found small to moderate ORs for the association of mechanical exposures and low back pain, although the relationships were complex. The presence of individual-level OR modifiers in such an area can be best understood by conducting a meta-analysis of individual participant data.In the past 4 decades there have been more than 30 systematic reviews of workplace biomechanical risk factors and low back pain (LBP).1–33 Although these reviews were rigorously done, controversy remains about the role of mechanical workplace exposures34 among other possible factors,35 and important gaps in our knowledge persist.36 To date, authors of comprehensive systematic reviews that include multiple definitions of LBP and mechanical exposures have not attempted to statistically combine data. This is likely attributable to the diversity of study designs, study populations, methods of exposure measurement, and assessments of LBP. Differences in the way mechanical exposure and LBP are reported make it difficult to comprehensively summarize this literature.To overcome these problems we developed methods to categorize “like” definitions of LBP that could be combined in a homogeneous meta-analysis37 and to create combinable mechanical exposure measures.38 We also obtained individual participant data for LBP studies in workers. We used these components to conduct an individual participant data meta-analysis. Use of individual participant data allowed us to go beyond the typical meta-analysis and make full use of all data collected. For example, we were not restricted to definitions of LBP reported in the published literature; instead, we were able to explore all definitions of LBP collected within a study. Individual participant data also enabled us to consistently adjust for potential confounding factors and explore individual-level odds ratio (OR) modifiers.     

缺失数据的多重估算

目的 探讨多重估算方法在缺失数据分析中的应用。方法 利用Bayesian理论与MCMC方法，在NORM软件中模拟得到m个完整数据集。结果 对m个重复测量数据集用SAS软件分析，合并m个分析结果可见，由NORM软件合并数据集的标准差比缺失数据集更稳定。结论 多重估算法既能反映缺失数据的不确定性，又可充分利用资料信息，对模型估计结果更可信。     

居民健康调查资料中的缺失数据的多重估算

目的 解决居民健康调查数据中存在的数据缺失问题。充分利用所采集到的数据。得出更有效的统计推断，方法 运用建立在马尔科夫链蒙特卡罗方法基础上的多重估算技术，对缺失数据进行替换，产生多个完整的数据集，进行联合统计推断。结果 弥补了由于缺失数据所造成的信息损失，改善了统计推断的质量。结论 多重估算技术是解决社会调查资料中数据缺失问题的有效工具。     

Meta-analyses Using Individual Participant Data From Cardiovascular Cohort Studies in Japan: Current Status and Future Directions

Meta-analysis of individual participant data (IPD meta-analysis) has several advantages over meta-analysis using aggregated published data, including the possibility of using statistical methods such as a fine stratification analysis, interaction analysis between 2 risk factors, and absolute risk estimation. The Evidence for Cardiovascular Prevention from Observational Cohorts in Japan Study (EPOCH-JAPAN), which was initiated in 2005, is a collaborative research project for IPD meta-analysis and includes 13 participating cohort studies in Japan. We generated 2 pooled databases with data on all-cause mortality (n = 199 047) and cardiovascular outcomes (n = 90 528) and applied a stratified Cox model to account for the different baseline hazards between cohorts. The results of our analyses show the age- and sex-specific associations between all-cause and cardiovascular disease mortality and established cardiovascular risk factors (blood pressure, smoking, total cholesterol, proteinuria, and kidney function). During the 9 years of its existence, the results generated by EPOCH-JAPAN have had important implications for clinical medicine and public health policy in Japan. The project is expected to draw upon new analytical methods such as interaction analysis and absolute risk evaluation in the near future. We believe that, over the next decade, this project will continue to provide new insights that can be applied to research on other Asian populations.Key words: meta-analysis, individual participant data, premature death, cardiovascular disease, population attributable fraction     

多重填补在随机干预试验研究中的应用

目的 利用多重填补方法实现对含缺失值的随机干预试验进行分析.方法 结合心理干预试验研究数据,利用SAS程序PROC MI和PROC MIANALYZE实现缺失数据的填补,应用稳健协方差分析评价心理健康干预效果.结果 填补与未填补分析结果一致,心理健康指标在干预组和对照组差别均无统计学意义,但CBO结局与干预有交互作用.结论 干预对学生心理健康起到一定的作用,但差别无统计学意义.     

Sex-Specific Effects of Nutritional Supplements for Infants Born Early or Small: An Individual Participant Data Meta-Analysis (ESSENCE IPD-MA) II: Growth

Neonatal nutritional supplements may improve early growth for infants born small, but effects on long-term growth are unclear and may differ by sex. We assessed the effects of early macronutrient supplements on later growth. We searched databases and clinical trials registers from inception to April 2019. Participant-level data from randomised trials were included if the intention was to increase macronutrient intake to improve growth or development of infants born preterm or small-for-gestational-age. Co-primary outcomes were cognitive impairment and metabolic risk. Supplementation did not alter BMI in childhood (kg/m²: adjusted mean difference (aMD) −0.11[95% CI −0.47, 0.25], p = 0.54; 3 trials, n = 333). Supplementation increased length (cm: aMD 0.37[0.01, 0.72], p = 0.04; 18 trials, n = 2008) and bone mineral content (g: aMD 10.22[0.52, 19.92], p = 0.04; 6 trials, n = 313) in infancy, but not at older ages. There were no differences between supplemented and unsupplemented groups for other outcomes. In subgroup analysis, supplementation increased the height z-score in male toddlers (aMD 0.20[0.02, 0.37], p = 0.03; 10 trials, n = 595) but not in females, and no significant sex interaction was observed (p = 0.21). Macronutrient supplementation for infants born small may not alter BMI in childhood. Supplementation increased growth in infancy, but these effects did not persist in later life. The effects did not differ between boys and girls.     

Using Multiple Imputation to Incorporate Cases with Missing Items in a Mental Health Services Study

When data analysis tools require that every variable be observed on each case, then missing items on a subset of variables force investigators either to leave potentially interesting variables out of analysis models or to include these variables but drop incomplete cases from the analysis. For example, in a study considered here, mental health patients were interviewed at two time points about a variety of topics that reflect successful adaptation to outpatient treatment, such as support from family and friends and avoidance of legal problems, although not all patients were successfully interviewed at the second time point. In a previous analysis of these data, logistic regression models were developed to relate baseline patient characteristics and recent treatment cost history to binary outcomes capturing aspects of adaptation. In these models, years of education was omitted as a covariate because it was incompletely observed at baseline. Here, we carry out analyses that include information from partially observed cases. Specifically, we use a multivariate model to produce multiple plausible imputed values for each missing item, and we combine results from separate logistic regression analyses on the completed data sets using the multiple imputation inference technique. Although the majority of inferences about specific regression coefficients paralleled those from the original study, some differences are noted. We discuss the implications of having flexible analysis tools for incomplete data in health services research and comment on issues related to model choice.     

配比设计中缺失数据的hot-deck估算

目的探讨hot-deck估算在含有缺失数据的配比设计资料分析中的应用.方法在Stata7.0软件上利用Adrian Mander和David Clayton开发的hot-deck程序进行数据模拟.结果hot-deck估算可以使缺失数据所包含的信息得到一定程度的弥补,当含有缺失值的变量是分类或等级变量时,效果更为显著.结论hot-deck估算是解决配比设计资料中离散型数据缺失问题的一个有效工具.     

Sex-Specific Effects of Nutritional Supplements for Infants Born Early or Small: An Individual Participant Data Meta-Analysis (ESSENCE IPD-MA) I—Cognitive Function and Metabolic Risk

Neonatal nutritional supplements are widely used to improve growth and development but may increase risk of later metabolic disease, and effects may differ by sex. We assessed effects of supplements on later development and metabolism. We searched databases and clinical trials registers up to April 2019. Participant-level data from randomised trials were included if the intention was to increase macronutrient intake to improve growth or development of infants born preterm or small-for-gestational-age. Co-primary outcomes were cognitive impairment and metabolic risk. Supplementation did not alter cognitive impairment in toddlers (13 trials, n = 1410; adjusted relative risk (aRR) 0.88 [95% CI 0.68, 1.13]; p = 0.31) or older ages, nor alter metabolic risk beyond 3 years (5 trials, n = 438; aRR 0.94 [0.76, 1.17]; p = 0.59). However, supplementation reduced motor impairment in toddlers (13 trials, n = 1406; aRR 0.76 [0.60, 0.97]; p = 0.03), and improved motor scores overall (13 trials, n = 1406; adjusted mean difference 1.57 [0.14, 2.99]; p = 0.03) and in girls not boys (p = 0.03 for interaction). Supplementation lowered triglyceride concentrations but did not affect other metabolic outcomes (high-density and low-density lipoproteins, cholesterol, fasting glucose, blood pressure, body mass index). Macronutrient supplementation for infants born small may not alter later cognitive function or metabolic risk, but may improve early motor function, especially for girls.     

时间序列中随机型缺失数据的填补及预测效果比较

目的本文旨在通过填补时间序列资料中的随机型缺失数据并拟合ARIMA模型,比较三种填补方法的填补和预测效果。方法利用SAS产生平稳、有周期性的时间序列并构造不同比例的随机型缺失,分别采用周期性填补法、均值填补法和三次样条函数插值法进行缺失数据的填补,并对填补后序列拟合ARIMA模型进行序列预测。采用配对t检验对三种填补方法的填补误差和序列预测误差进行比较。结果三种填补方法的填补值与真值的差异均无统计学意义(P>0.05);随着缺失比例的增大,周期性填补法的填补误差和序列预测误差均小于三次样条函数插值法和均值填补法。结论周期性填补法对于含有确切周期信息的时间序列缺失数据,填补效果较优。     

Explicating the Conditions Under Which Multilevel Multiple Imputation Mitigates Bias Resulting from Random Coefficient-Dependent Missing Longitudinal Data

Random coefficient-dependent (RCD) missingness is a non-ignorable mechanism through which missing data can arise in longitudinal designs. RCD, for which we cannot test, is a problematic form of missingness that occurs if subject-specific random effects correlate with propensity for missingness or dropout. Particularly when covariate missingness is a problem, investigators typically handle missing longitudinal data by using single-level multiple imputation procedures implemented with long-format data, which ignores within-person dependency entirely, or implemented with wide-format (i.e., multivariate) data, which ignores some aspects of within-person dependency. When either of these standard approaches to handling missing longitudinal data is used, RCD missingness leads to parameter bias and incorrect inference. We explain why multilevel multiple imputation (MMI) should alleviate bias induced by a RCD missing data mechanism under conditions that contribute to stronger determinacy of random coefficients. We evaluate our hypothesis with a simulation study. Three design factors are considered: intraclass correlation (ICC; ranging from .25 to .75), number of waves (ranging from 4 to 8), and percent of missing data (ranging from 20 to 50%). We find that MMI greatly outperforms the single-level wide-format (multivariate) method for imputation under a RCD mechanism. For the MMI analyses, bias was most alleviated when the ICC is high, there were more waves of data, and when there was less missing data. Practical recommendations for handling longitudinal missing data are suggested.     

The Early Prevention of Obesity in CHildren (EPOCH) Collaboration - an Individual Patient Data Prospective Meta-Analysis

Background

Efforts to prevent the development of overweight and obesity have increasingly focused early in the life course as we recognise that both metabolic and behavioural patterns are often established within the first few years of life. Randomised controlled trials (RCTs) of interventions are even more powerful when, with forethought, they are synthesised into an individual patient data (IPD) prospective meta-analysis (PMA). An IPD PMA is a unique research design where several trials are identified for inclusion in an analysis before any of the individual trial results become known and the data are provided for each randomised patient. This methodology minimises the publication and selection bias often associated with a retrospective meta-analysis by allowing hypotheses, analysis methods and selection criteria to be specified a priori.

Methods/Design

The Early Prevention of Obesity in CHildren (EPOCH) Collaboration was formed in 2009. The main objective of the EPOCH Collaboration is to determine if early intervention for childhood obesity impacts on body mass index (BMI) z scores at age 18-24 months. Additional research questions will focus on whether early intervention has an impact on children's dietary quality, TV viewing time, duration of breastfeeding and parenting styles. This protocol includes the hypotheses, inclusion criteria and outcome measures to be used in the IPD PMA. The sample size of the combined dataset at final outcome assessment (approximately 1800 infants) will allow greater precision when exploring differences in the effect of early intervention with respect to pre-specified participant- and intervention-level characteristics.

Discussion

Finalisation of the data collection procedures and analysis plans will be complete by the end of 2010. Data collection and analysis will occur during 2011-2012 and results should be available by 2013.

Trial registration number

ACTRN12610000789066

     

Using Item Response Theory to Model Multiple Phenotypes and Their Joint Heritability in Family Data

Many important complex diseases are composed of a series of phenotypes, which makes the disease diagnosis and its genetic dissection difficult. The standard procedures to determine heritability in such complex diseases are either applied for single phenotype analyses or to compare findings across phenotypes or multidimensional reduction procedures, such as principal components analysis using all phenotypes. However each method has its own problems and the challenges are even more complex for extended family data and categorical phenotypes. In this paper, we propose a methodology to determine a scale for complex outcomes involving multiple categorical phenotypes in extended pedigrees using item response theory (IRT) models that take all categorical phenotypes into account, allowing informative comparison among individuals. An advantage of the IRT framework is that a straightforward joint heritability parameter can be estimated for categorical phenotypes. Furthermore, our methodology allows many possible extensions such as the inclusion of covariates and multiple variance components. We use Markov Chain Monte Carlo algorithm for the parameter estimation and validate our method through simulated data. As an application we consider the metabolic syndrome as the multiple phenotype disease using data from the Baependi Heart Study consisting of 1,696 individuals in 95 families. We adjust IRT models without covariates and include age and age squared as covariates. The results showed that adjusting for covariates yields a higher joint heritability () than without co variates () indicating that the covariates absorbed some of the error variance.     

An Empirical Comparison of Meta‐analysis and Mega‐analysis of Individual Participant Data for Identifying Gene‐Environment Interactions

For analysis of the main effects of SNPs, meta‐analysis of summary results from individual studies has been shown to provide comparable results as “mega‐analysis” that jointly analyzes the pooled participant data from the available studies. This fact revolutionized the genetic analysis of complex traits through large GWAS consortia. Investigations of gene‐environment (G×E) interactions are on the rise since they can potentially explain a part of the missing heritability and identify individuals at high risk for disease. However, for analysis of gene‐environment interactions, it is not known whether these methods yield comparable results. In this empirical study, we report that the results from both methods were largely consistent for all four tests; the standard 1 degree of freedom (df) test of main effect only, the 1 df test of the main effect (in the presence of interaction effect), the 1 df test of the interaction effect, and the joint 2 df test of main and interaction effects. They provided similar effect size and standard error estimates, leading to comparable P‐values. The genomic inflation factors and the number of SNPs with various thresholds were also comparable between the two approaches. Mega‐analysis is not always feasible especially in very large and diverse consortia since pooling of raw data may be limited by the terms of the informed consent. Our study illustrates that meta‐analysis can be an effective approach also for identifying interactions. To our knowledge, this is the first report investigating meta‐versus mega‐analyses for interactions.