Explore chapters and articles related to this topic
The Meta-Analysis of Genetic Studies
Published in Christopher H. Schmid, Theo Stijnen, Ian R. White, Handbook of Meta-Analysis, 2020
Cosetta Minelli, John Thompson
In a standard GWAS meta-analysis, each study provides summary data on the measured or imputed genetic variants in the form of an estimate and its standard error (e.g., log odds ratios for a binary outcome and regression coefficients for a continuous outcome). Once these summary data have been thoroughly checked and aligned, the meta-analysis proper can begin. The most commonly used approach is a common-effect meta-analysis (see Chapter 4) based on inverse-variance weighting applied separately to each variant (Gogele et al., 2012), assuming no between-study heterogeneity in the genetic effects.
Stratified Analysis
Published in Peter Cummings, Analysis of Incidence Rates, 2019
Inverse-variance weighting is easy to describe, so I will use it to illustrate what happens when a variance-minimizing method is used for comparing cancer mortality in Alaska with Florida. This method has also been called direct pooling, precision weighting, the Woolf method, and weighted least squares (Greenland and Rothman 2008c p271). The weights are the inverse of the variance of each rate difference or ratio. The variances are equal to the square of the standard error (SE) of the mean for row 5 (rate differences) or row 10 (rate ratios) in the last column of Table 4.4. Here are Stata commands that produce the pooled (inverse-variance) results for the rate ratio and rate difference, using population in units of 100,000 person-years and using the Alaska population as the standard. The standardized estimates are also reported: . ir count alaska pop2, by(agecat) istandard pool Age, 11 categori | IRR [95% Conf. Interval] Weight ------------+------------------------------------ 1 | 9.63816 .1633664 185.14 .10828 (exact) 1-4 | 1.821176 .2075938 7.408422 .43168 (exact) 5-14 | .7496819 .0183554 4.521574 1.01703 (exact) 15-24 | 1.310156 .4151356 3.174781 1.0656 (exact) 25-34 | 1.019463 .459307 1.975306 1.03125 (exact) 35-44 | .7361211 .4529502 1.133598 .92974 (exact) 45-54 | .7749275 .6306186 .9429465 1.11026 (exact) 55-64 | .7604977 .658696 .8738111 .85909 (exact) 65-74 | 1.061044 .927879 1.208177 .3535 (exact) 75-84 | 1.363523 1.187064 1.559078 .14877 (exact) 85+ | 1.094918 .8746952 1.354099 .04711 (exact) ------------+------------------------------------ Crude | .5643365 .5273743 .6032341 (exact) Pooled (direct) | 1.005856 .9409589 1.075229 I. Standardized | .9776556 .9144901 1.045184 ------------------------------------------------- Test of homogeneity (direct) chi2(10) = 49.92 Pr>chi2 = 0.0000 . ir count alaska pop2, by(agecat) istandard pool ird Age, 11 categori | IRD [95% Conf. Interval] Weight ------------+------------------------------------ 1 | 8.277113 -9.872422 26.42665 .10828 1-4 | 2.089068 -4.419313 8.597449 .43168 5-14 | -.328308 -2.313692 1.657076 1.01703 15-24 | 1.110793 -3.069535 5.29112 1.0656 25-34 | .1666198 -5.659688 5.992928 1.03125 35-44 | -8.096794 -18.00497 1.811381 .92974 45-54 | -26.94473 -45.32296 -8.566496 1.11026 55-64 | -75.14965 -108.5949 -41.70443 .85909 65-74 | 37.92093 -47.52346 123.3653 .3535 75-84 | 390.6702 195.1872 586.1532 .14877 85+ | 158.253 -229.4737 545.9798 .04711 ------------+------------------------------------ Crude | -96.08723 -104.5623 -87.61215 Pooled (direct) | -.4022102 -2.021802 1.217382 I. Standardized | -2.844692 -11.16289 5.473506 ------------------------------------------------- Test of homogeneity (direct) chi2(10) = 48.29 Pr>chi2 = 0.0000
Effect of anxiolytic drug silexan on sleep – a narrative review
Published in The World Journal of Biological Psychiatry, 2022
Erich Seifritz, Siegfried Kasper, Hans-Jürgen Möller, Hans-Peter Volz, Walter E. Müller, Anne Eckert, Martin Hatzinger
The analysis was based on the original data of the included trials which were obtained from the sponsor. In a first step, the absolute score difference between baseline and the end of treatment visit was calculated within each trial, and these differences were compared between silexan and placebo using analysis of covariance models with treatment as a factor and the baseline value of the outcome variable as a covariate. For consistency with the original trial protocols and the resulting publications missing data were replaced by carrying the last valid observation forward. Meta-analysis was then performed in a second step using a random effects model based on the DerSimonian and Laird (1986) method. Inverse variance weighting was used for combining the results of the single trials.
Treatment of difficult-to-treat depression – clinical guideline for selected interventions
Published in Nordic Journal of Psychiatry, 2022
Stine Bjerrum Moeller, Krzysztof Gbyl, Carsten Hjorthøj, Maike Andreasen, Stephen F. Austin, Poul Erik Buchholtz, Line Fønss, Simon Hjerrild, Lise Hogervorst, Martin Balslev Jørgensen, Nicolai Ladegaard, Klaus Martiny, Jonas Meile, Aake Packness, Karen Rodriguez Sigaard, Krista Straarup, Sune P. V Straszek, Claus Havregaard Soerensen, Birgitte Welcher, Poul Videbech
Data extraction was conducted independently by two reviewers with support from the method- and review specialist. We transferred the extracted data to Review Manager software (version 5.3) for analyses and meta-analyses using pooled data for each of the six review questions. For outcomes where meta-analysis was deemed inappropriate (e.g. if heterogeneity was higher than 70%), results were synthesized narratively. We conducted meta‐analyses using the standardized mean difference (SMD) for continuous outcomes and the relative risk (RR) for dichotomous outcomes. Effects were calculated as random effects and with inverse variance weighting. In the case of continuous outcomes, endpoint scores were preferred over ‘change from baseline’ if both were available. In the results, we included measures of uncertainty such as 95% confidence intervals (CI) and estimates of I2. The latter measures the percentage of total variation (between-study and within-study), which is due to heterogeneity (between-study variation).
Preoperative MRI characteristics predict chronic subdural haematoma postoperative recurrence: a meta-analysis
Published in British Journal of Neurosurgery, 2021
Brandon A. Sherrod, Cordell Baker, Nicholas Gamboa, Scott McNally, Ramesh Grandhi
Univariate analysis was performed to calculate unadjusted odds ratios (ORs), 95% confidence intervals (CIs), and p values for preoperative MRI T1 hypo- or isointensity predicting increased risk of SDH recurrence postoperatively. If studies reported univariate analysis results, we used these values in our own analysis. Otherwise, these values were calculated using raw data reported in each study. Multivariate analysis results were obtained from studies where reported including adjusted ORs, 95% CIs, and p values for preoperative MRI T1 hypo- or isointensity predicting increased risk of SDH recurrence postoperatively. Forest plots were generated for both univariate and multivariate analyses, and a combined OR was calculated for recurrence risk across studies. Studies were weighted using inverse variance weighting. Heterogeneity statistics were calculated using Cochrane’s Q statistic and by calculating the resultant I2 values. Statistical analyses were performed using SPSS software version 24.0.