A Method of Moments Estimator for Random Effect Multivariate Meta-Analysis

ArticleinBiometrics 68(4) · May 2012with30 Reads
DOI: 10.1111/j.1541-0420.2012.01761.x · Source: PubMed
Abstract
Meta-analysis is a powerful approach to combine evidence from multiple studies to make inference about one or more parameters of interest, such as regression coefficients. The validity of the fixed effect model meta-analysis depends on the underlying assumption that all studies in the meta-analysis share the same effect size. In the presence of heterogeneity, the fixed effect model incorrectly ignores the between-study variance and may yield false positive results. The random effect model takes into account both within-study and between-study variances. It is more conservative than the fixed effect model and should be favored in the presence of heterogeneity. In this paper, we develop a noniterative method of moments estimator for the between-study covariance matrix in the random effect model multivariate meta-analysis. To our knowledge, it is the first such method of moments estimator in the matrix form. We show that our estimator is a multivariate extension of DerSimonian and Laird's univariate method of moments estimator, and it is invariant to linear transformations. In the simulation study, our method performs well when compared to existing random effect model multivariate meta-analysis approaches. We also apply our method in the analysis of a real data example.
    • "Odds ratios (ORs) and hazards ratios (HRs) with 95% confidence intervals (95%CIs) were estimated using fixed-or random-effects models to evaluate the association between MALAT1 expression and the clinical pathological features and prognosis of cancers. AZ-test was employed to assess the significance of the pooled effect size (Chen et al., 2012) and forest plots were generated to display between-group comparisons of HRs and ORs with 95%CIs. The Q (Jackson et al., 2012) and I 2 tests were applied to assess between-study heterogeneity (Peters et al., 2006). "
    [Show abstract] [Hide abstract] ABSTRACT: Metastasis-associated lung adenocarcinoma transcript 1 (MALAT1) has been identified as a potential cancer biomarker, yet the mechanism by which it influences the development of cancer remains unknown. In this study, we aimed to correlate MALAT1 expression with pathological features and prognosis in cancer patients. Several databases were searched using combinations of keywords relating to MALAT1 and cancer. After selection of relevant cohort studies according to strict criteria, a meta-analysis was conducted. Twelve studies were analyzed, involving 958 cancer patients. Elevated MALAT1 expression was associated with poor prognosis and larger tumors [prognosis: hazard ratio = 3.11, 95% confidence interval (CI) = 1.98-4.23, P = 0.000; tumor size: odds ratio (OR) = 0.40, 95%CI = 0.21-0.74, P = 0.003]. However, no connection with histological grade, T-stage, lymph node (LN) metastasis, or distant metastasis was established (all P > 0.05). A correlation between increased expression and poor prognosis was observed in the large and small sample-size subgroups (all P< 0.05), as was a relationship with large tumor size (OR = 0.30, 95%CI = 0.13-0.71, P = 0.006). Expression was correlated with T-stage and distant metastasis in the small sample-size subgroup (all P < 0.05), but no association was detected regarding histological grade, LN metastasis in either subgroup (all P > 0.05). Our findings demonstrate that elevated MALAT1 expression correlates with large tumor size, advanced tumor stage, and poor prognosis, and might therefore be utilized to evaluate clinical pathological features and prognostic out come for cancer patients.
    Full-text · Article · Dec 2015
    • "A Z test (Chen et al., 2012) was carried out to assess the significance of the overall effect size and forest plots were used to display the values of OR at 95%CIs between patient and control groups. Heterogeneity was assessed by the Cochran's Q-statistic (Chen et al., 2012), with P < 0.05 indicating the existence of heterogeneity. For quantification of the degree of heterogeneity, the I-squared (I 2 ) statistic (Peters et al., 2006) was calculated ranging from 0 to 100% (0%, no heterogeneity; 100%, maximal heterogeneity). "
    [Show abstract] [Hide abstract] ABSTRACT: This meta-analysis investigated the correlation between the PPARγ2 Pro12Ala polymorphism and cardiovascular disease (CVD). Electronic database and manual searches were conducted to retrieve studies published relevant to the PPARγ2 Pro12Ala polymorphism and CVD. Rigorous inclusion and exclusion criteria were employed for selection of high-quality patients-control studies. Statistical data analyses on allelic, dominant, homozygous, heterozygous, and recessive inheritance models were performed using the R 3.1.0 and Stata 12.0 software. We enrolled 12 case-control studies consisting of 10,189 patients with CVD [1070 with myocardial infarction (MI), 7849 with coronary artery disease (CAD), and 1270 with acute coronary syndromes (ACS)] and 17,899 controls. The results of meta-analyses revealed that the PPARγ2 Pro12Ala (rs1801282) polymorphism was correlated with a higher risk of CVD under both allelic and dominant models, while no statistical significance was found under homozygous, heterozygous, or recessive models. Subgroup analysis based on disease showed that the PPARγ2 Pro12Ala (rs1801282) polymorphism was correlated with a higher risk of MI under both allelic and dominant models, while no statistical significance was found for association with CAD or ACS under allele or dominant models. Furthermore, under homozygous, heterozygous, and recessive models, the PPARγ2 Pro12Ala (rs1801282) polymorphism had no statistically significant association with MI, CAD, or ACS. The results of this meta-analysis suggest that the PPARγ2 Pro12Ala (rs1801282) polymorphism might be correlated with a higher risk of CVD, particularly MI, and could serve as an important early indicator for CVD.
    Article · Dec 2015
    • "Correlations between APOA5 and APOC3 SNPs and risk of CHD were estimated by the calculation of odds ratios (ORs) and 95% confidence intervals (95%CIs). A Z-test was employed to detect the significance of overall effect size (Chen et al., 2012), and forest plots were generated to display OR values and 95%CIs between case and control groups. Heterogeneity among studies was evaluated using Cochran's Q-statistic (P < 0.05 was considered to signify evident heterogeneity) and the I 2 test, which measures the percentage of total variation across studies, ranging from 0 to 100% (Peters et al., 2006; Jackson et al., 2012). "
    [Show abstract] [Hide abstract] ABSTRACT: The aim of this study was to investigate correlations between apolipoprotein A-V (APOA5) -1131T>C and apolipoprotein C-III (APOC3) -455T>C polymorphisms and coronary heart disease (CHD). PubMed, Ovid, Cochrane Library, Embase, China National Knowledge Infrastructure, and Wanfang databases were searched using combinations of keywords relating to these polymorphisms and CHD. Studies retrieved from database searches were screened using our stringent inclusion and exclusion criteria, and Comprehensive Meta-Analysis Version 2.0 software was used for statistical analyses. In total, 115 studies were initially retrieved and after further selection, 11 were included in the meta-analysis. These 11 articles comprised 4840 patients with CHD in the case group and 4913 healthy participants in the control group. Meta-analysis revealed that APOA5 -1131T>C and APOC3 -455T>C polymorphisms increased CHD risk. In addition, subgroup analysis by ethnicity showed that while the -1131T>C polymorphism elevated the risk of CHD in the Caucasian population under both allelic and dominant models, this increased risk was observed only under a dominant model in the Asian population. The results of our meta-analysis point to a strong link between both APOA5 -1131T>C and APOC3 -455T> C polymorphisms and an increased risk of CHD. Thus, these polymorphisms constitute important predictive indicators of CHD susceptibility.
    Article · Dec 2015
Show more