What to add to nothing? Use and avoidance of continuity corrections in meta-analysis of sparse data

2004 ◽  
Vol 23 (9) ◽  
pp. 1351-1375 ◽  
Author(s):  
Michael J. Sweeting ◽  
Alexander J. Sutton ◽  
Paul C. Lambert
Keyword(s):  
Meta Analysis ◽  
Sparse Data ◽  
Continuity Corrections

scholarly journals Performance of methods for meta-analysis of diagnostic test accuracy with few studies or sparse data

2015 ◽  
Vol 26 (4) ◽  
pp. 1896-1911 ◽  
Author(s):  
Yemisi Takwoingi ◽  
Boliang Guo ◽  
Richard D Riley ◽  
Jonathan J Deeks
Keyword(s):  
Hierarchical Models ◽  
Operating Characteristic ◽  
Meta Analysis ◽  
Sparse Data ◽  
Parameter Estimates ◽  
Test Accuracy ◽  
Summary Receiver Operating Characteristic ◽  
Bivariate Model ◽  
Logistic Regression Models ◽  
Meta Analyses

Hierarchical models such as the bivariate and hierarchical summary receiver operating characteristic (HSROC) models are recommended for meta-analysis of test accuracy studies. These models are challenging to fit when there are few studies and/or sparse data (for example zero cells in contingency tables due to studies reporting 100% sensitivity or specificity); the models may not converge, or give unreliable parameter estimates. Using simulation, we investigated the performance of seven hierarchical models incorporating increasing simplifications in scenarios designed to replicate realistic situations for meta-analysis of test accuracy studies. Performance of the models was assessed in terms of estimability (percentage of meta-analyses that successfully converged and percentage where the between study correlation was estimable), bias, mean square error and coverage of the 95% confidence intervals. Our results indicate that simpler hierarchical models are valid in situations with few studies or sparse data. For synthesis of sensitivity and specificity, univariate random effects logistic regression models are appropriate when a bivariate model cannot be fitted. Alternatively, an HSROC model that assumes a symmetric SROC curve (by excluding the shape parameter) can be used if the HSROC model is the chosen meta-analytic approach. In the absence of heterogeneity, fixed effect equivalent of the models can be applied.

scholarly journals Updating Insights into Rosiglitazone and Cardiovascular Risk through Shared Data: Individual Patient- and Summary-Level Meta-Analyses

2019 ◽  
Author(s):  
Joshua D Wallach ◽  
Kun Wang ◽  
Audrey D Zhang ◽  
Deanna Cheng ◽  
Holly K Grossetta Nardini ◽  
...  
Keyword(s):  
Myocardial Infarction ◽  
Heart Failure ◽  
Systematic Review ◽  
Cardiovascular Risk ◽  
Meta Analysis ◽  
Myocardial Infarctions ◽  
Odds Ratios ◽  
Level Data ◽  
Continuity Corrections ◽  
Meta Analyses

ABSTRACTObjectiveTo conduct a systematic review and meta-analysis of the effects of rosiglitazone therapy on cardiovascular risk and mortality using multiple data sources and varying analytical approaches.DesignSystematic review and meta-analysis of randomized controlled trials.Data sourcesGlaxoSmithKline’s (GSK) Clinical Study Data Request (CSDR) and Study Register platforms, MEDLINE, PubMed, Embase, Web of Science, Cochrane Central Registry of Controlled Trials, Scopus, and ClinicalTrials.gov from inception to January 2019.Study selection criteriaRandomized, controlled, phase II-IV clinical trials comparing rosiglitazone with any control for at least 24 weeks in adults.Data extraction and synthesisFor analyses of trials for which individual patient-level data (IPD) were available, we examined a composite of the following events as our primary outcome: acute myocardial infarction, heart failure, cardiovascular-related deaths, and non-cardiovascular-related deaths. As secondary analyses, these four events were examined independently. When also including trials for which IPD were not available, we examined myocardial infarction and cardiovascular-related deaths, ascertained from summary-level data. Multiple meta-analyses were conducted, accounting for trials with zero events in one or all arms with two different continuity corrections (i.e., 0.5 constant and treatment arm comparator continuity correction), to calculate odds ratios and risk ratios with 95% confidence intervals.ResultsThere were 33 eligible trials for which IPD were available (21156 participants) through GSK’s CSDR. We also identified 103 additional trials for which IPD were not available from which we ascertained myocardial infarctions (23683 patients) and 103 trials for cardiovascular-related deaths (22772 patients). Among trials for which IPD were available, we identified a greater number of myocardial infarctions and fewer cardiovascular-related deaths reported in the IPD as compared to the summary-level data. When limited to trials for which IPD were available and accounting for trials with zero-events in only one arm using a constant continuity correction of 0.5, patients treated with rosiglitazone had a 39% increased risk of a composite event compared with controls (Mantel-Haenszel odds ratio 1.39, 95% CI 1.15 to 1.68). When examined separately, the odds ratios for myocardial infarction, heart failure, cardiovascular-related death, and non-cardiovascular-related death were 1.25 (0.99 to 1.60), 1.60 (1.20 to 2.14), 1.18 (0.64 to 2.17), and 1.13 (0.58 to 2.20), respectively. When all trials for which IPD were and were not available were combined for myocardial infarction and cardiovascular-related deaths, the odds ratios were attenuated (1.13 (0.92 to 1.38) and 1.10 (0.73 to 1.65), respectively). Effect estimates and 95% confidence intervals were broadly consistent when analyses were repeated including trials with zero events across all arms using constant continuity corrections of 0.5 or treatment arm continuity corrections.ConclusionsResults of this comprehensive meta-analysis aggregating a multitude of trials and analyzed using a variety of statistical techniques suggest that rosiglitazone is consistently associated with an increased cardiovascular risk, likely driven by heart failure events, whose interpretation is complicated by varying magnitudes of myocardial infarction risk that were attenuated through aggregation of summary-level data in addition to IPD.Systematic review registrationhttps://osf.io/4yvp2/What is already known on this topic-Since 2007, there have been multiple meta-analyses, using various analytic approaches, that have reported conflicting findings related to rosiglitazone’s cardiovascular risk.-Previous meta-analyses have relied primarily on summary-level data, and did not have access to individual patient-level data (IPD) from clinical trials.-Currently, there is little consensus on which method should be used to account for sparse adverse event data in meta-analyses.What this study adds-Among trials for which IPD were available, rosiglitazone use was consistently associated with an increased cardiovascular risk, likely driven by heart failure events.-Interpretation of rosiglitazone’s cardiovascular risk is complicated by varying magnitudes of myocardial infarction risk that were attenuated through aggregation of summary-level data in addition to IPD.-Among trials for which IPD were available, we identified a greater number of myocardial infarctions and fewer cardiovascular deaths reported in the IPD as compared to the summary-level data, which suggests that IPD may be necessary to accurately classify all adverse events when performing meta-analyses focused on safety.

Meta-Analysis and Sparse-Data Bias

2020 ◽  
Author(s):  
David B Richardson ◽  
Stephen R Cole ◽  
Rachael K Ross ◽  
Charles Poole ◽  
Haitao Chu ◽  
...  
Keyword(s):  
Logistic Regression ◽  
Meta Analysis ◽  
Sparse Data ◽  
Maximum Likelihood Estimates ◽  
Finite Sample ◽  
Logistic Regression Models ◽  
Confidence Interval Coverage ◽  
The Individual ◽  
Meta Analyses ◽  
Sparse Data Bias

Abstract Meta-analyses are undertaken to combine information from a set of studies, often in settings where some of the individual study-specific estimates are based on relatively small study samples. Finite sample bias may occur when maximum likelihood estimates of associations are obtained by fitting logistic regression models to sparse data sets. Here we show that combining information from small studies by undertaking a meta-analytical summary of logistic regression estimates can propagate such sparse-data bias. In simulations, we illustrate 2 challenges encountered in meta-analyses of logistic regression results in settings of sparse data: 1) bias in the summary meta-analytical result and 2) confidence interval coverage that can worsen rather than improve, in terms of being less than nominal, as the number of studies in the meta-analysis increases.

An Exact Test for Meta-analysis with Binary Endpoints

2007 ◽  
Vol 46 (06) ◽  
pp. 662-668 ◽  
Author(s):  
C. Gromann ◽  
O. Kuss
Keyword(s):  
Random Effects ◽  
Fixed Effects ◽  
Meta Analysis ◽  
Artery Bypass ◽  
Sparse Data ◽  
Random Effects Model ◽  
Exact Test ◽  
Standard Methods ◽  
Meta Analyses ◽  
Empirical Size

Summary Objectives : We reintroduce an exact Mantel-Haenszel (MH) procedure for meta-analysis with binary endpoints which is expected to workespeciallywell i sparse data, e.g., in meta-analyses of safety or adverse events. Methods : The performance of the exact MH procedure in terms of empirical size and power is compared to the asymptotic MH and to the two standard procedures (fixed effects and random effects model) in a simulation study. We illustrate the methods with a metaanalysis of postoperative stroke occurrence after offpump or on-pump surgery in coronary artery bypass grafting. Results : We find that in almost all situations the asymptotic MH procedure outperforms its competitors; especially the standard methods yield poor results in terms of power and size. Conclusions : There is no need to use the reintroduced exact MH procedure; the asymptotic MH procedure will be sufficient in most practical situations. The standard methods (fixed effects and random effects model) should not be used in the sparse data situation.

Sign in / Sign up

Export Citation Format

Share Document