China
Africa
Explore chapters and articles related to this topic
Multivariate Meta-Analysis
Published in Christopher H. Schmid, Theo Stijnen, Ian R. White, Handbook of Meta-Analysis, 2020
Dan Jackson, Ian R. White, Richard D. Riley
The methods described above summarize how standard univariate methods for estimating the between-study variance and making inferences about the average effect have been generalized to the multivariate case. However, further results are often presented when communicating the results of univariate meta-analyses and multivariate extensions of some of these have recently been proposed. In particular, descriptive statistics, such as heterogeneity statistics and study weights, have become commonplace in univariate meta-analysis. For example, I2 statistics (Higgins and Thompson, 2002) are often used to quantify the impact of between-study heterogeneity and study weights are often displayed on forest plots in order to show which studies contribute most to the estimated effect. These descriptive statistics have been extended for use in multivariate meta-analysis and we describe these methods next.
Meta-Analysis for Evaluating Diagnostic Accuracy
Published in Ding-Geng (Din) Chen, Karl E. Peace, Applied Meta-Analysis with R and Stata, 2021
In addition, the weighted average approach assumes homogeneity between studies that variations between the reported sensitivity (specificity) values from the selected diagnostic studies are purely due to a random error; however, in reality, due to different clinical settings, tests operators and evaluators, or different patient characteristics, the diagnostic accuracy measures can be truly different between studies. Such a situation is referred to as between-study heterogeneity. The previous chapter in this book discussed measures to test if heterogeneity exists in a MA, and we can apply here to check if heterogeneity exists between DTA studies. A common way of accounting for between-study heterogeneity and correlation between sensitivity and specificity pairs is using the bivariate MA approach, which models the sensitivity (or true-positive rate (TPR)) and specificity (or false-positive rate (FPR)) directly and jointly. The bivariate model in a MA for DTA studies has two levels of modeling. At the first level, it models the variability of TPR (sensitivity) and FPR (1-specificity) within each DTA study, respectively, and assumes normal distributions for each. At the second level, the correlation between TPR and FPR is accounted directly by assuming a bivariate normal distribution, and the between-study heterogeneity was accounted for by using the average sensitivity and specificity Reitsma et al. (2005). Let ξi = logit (F P Ri) and ηi = logit (T P Ri) be the true population parameters for each study. A bivariate normal model for (ξi, ηi) is
Osteoarthritis
Published in John M. Saxton, Exercise and Chronic Disease, 2011
In this same meta-analysis (Fransen and McConnell 2008), the authors describe treatment intensity in the various RCTs as ranging from ‘maximum effort’ muscle strengthening (Schilke et al., 1996; Gür et al., 2002) to low intensity aerobic walking (Bautch et al., 1997; Talbot et al., 2003; Messier et al., 2004). Treatment content varied from mostly aerobic walking programmes (Minor et al., 1989; Kovar et al., 1992; Ettinger et al., 1997; Talbot et al., 2003; Messier et al., 2004) to very complex, comprehensive programmes including manual therapy, upper-limb and/or truncal muscle strengthening and balance coordination (Rogind et al., 1998; van Baar et al., 1998; Peloquin et al., 1999; Deyle et al., 2000; Bennell et al., 2005) in addition to the more usual lower-limb muscle strengthening programmes. Two studies evaluated Tai Chi classes (Song et al., 2003; Fransen et al., 2007). The 32 included studies were categorised according to the main treatment focus: simple quadriceps strengthening, lower limb muscle strengthening (Theraband, cuff weights, isokinetics), strengthening and aerobic component (stationary bicycle or walking), walking programme only or ‘other’ (not specifically focused on lower limb muscle strengthening or increasing aerobic capacity). The simple quadriceps programmes achieved only borderline significance for both pain and physical function and the ‘other’ programmes resulted in an insignificant treatment effect for physical function. However, for both pain and physical function, no significant difference in effect size could be demonstrated between the groups of exercise programmes when testing the Chi2 distribution. Between study heterogeneity was considerable within most of the study categories.
Risk factors for acute kidney injury in COVID-19 patients: an updated systematic review and meta-analysis
Published in Renal Failure, 2023
Jialing Zhang, Qi Pang, Ting Zhou, Jiali Meng, Xingtong Dong, Zhe Wang, Aihua Zhang
Through our meta-analysis, we determined that hypertension, CKD, and diabetes were all independent risk factors for AKI in patients with COVID-19, consistent with the study of Cai X [5]. Considering the significant heterogeneity among studies, we not only applied a random-effect model but also subsequently performed meta-regression and sensitivity analysis. Through our meta-regression analysis, we observed an interaction among country, hypertension, and CKD. By further sensitivity analysis, we identified significant heterogeneity between the study of Chávez-Íñiguez et al. [32] and other studies. Chávez-Íñiguez et al. conducted a prospective, observational cohort study with a relatively young population from Mexico. Clinical heterogeneity is probably high for the country since the country is a heterogeneous group of diverse ethnicities. Disparities in socioeconomic conditions across racial lines were exacerbated during the COVID-19 pandemic [62]. In addition, study heterogeneity may be related to the proportion of subjects with comorbidities, study sample size, COVID severity, and differences among national and regional healthcare systems, which may affect the conclusions of the meta-analysis.
Determinants of return to activity and work after carpal tunnel release: a systematic review and meta-analysis
Published in Expert Review of Medical Devices, 2023
Larry E. Miller, Kevin C. Chung
The mean and 95% confidence interval (CI) were calculated for each outcome in individual studies and the overall pooled estimate. The pooled RTA and RTW estimates were calculated using the DerSimonian and Laird method for random-effects meta-analysis to account for anticipated inter-study heterogeneity and visually depicted with forest plots. For studies where a variance measure (e.g. standard deviation) or the percentage of employed patients were not reported, we imputed these missing data in accordance with Cochrane recommendations [14]. Potential publication bias was assessed by visual examination of funnel plot symmetry, Egger tests [15], and the trim-and-fill method to estimate the number of studies missing from the meta-analysis due to publication bias [16]. The influence of missing data imputation and single-study effects on outcomes were assessed in sensitivity analyses.
Dietary fiber intake, depression, and anxiety: a systematic review and meta-analysis of epidemiologic studies
Published in Nutritional Neuroscience, 2023
Faezeh Saghafian, Maryam Hajishafiee, Parisa Rouhani, Parvane Saneei
Reported ORs, RRs, or HRs (and their 95% CIs) for depression were used to calculate log OR and its standard errors. For two studies, that reported several ORs for different kinds of dietary fiber, we first consolidated them in a preliminary meta-analysis to provide an overall estimate for that study [33,13]. Using a random effects model that takes between-study variation into account, the overall effect size was calculated. Between-study heterogeneity was assessed using Cochran’s Q test and I2. In case of significant between-study heterogeneity, subgroup analysis was used to find possible sources of heterogeneity. Between-subgroup heterogeneity was examined through fixed effects model. Sensitivity analysis was carried out to examine the extent to which inferences might depend on a particular study. Publication bias was assessed by visual inspection of Begg’s funnel plots. In addition, Egger’s regression asymmetry test was used to formally assess funnel plot asymmetry. Statistical analyses were conducted using STATA version 11.2 (STATA Corp., College Station, Texas). P values less than 0.05 were considered statistically significant.