Meta‐analysis of continuous outcomes combining individual patient data and aggregate data

Riley, Richard D; Lambert, Paul C.; Staessen, Jan A.; Wang, Ji‐Guang; Gueyffier, François; Thijs, Lutgarde; Boutitie, Florent

doi:10.1002/sim.3165

Cited by 237 publications

(350 citation statements)

References 43 publications

Supporting

Mentioning

340

Contrasting

Unclassified

Order By: Relevance

“…Our findings agree with previous work and simulations for continuous and binary outcomes 9, 23: it is crucial to separate within‐trial and across‐trial interactions, to avoid ecological bias caused by unexplained trial‐level confounding 8, 54, 55. Otherwise, clinical conclusions about interactions may be driven by ecological, trial‐level information rather than solely within‐trial information at the individual‐level.…”

Section: Discussionsupporting

confidence: 90%

“…In situations where IPD are limited and most information comes from across trials (for example, when IPD are not available for all studies 9, 20, 23 or the variation in particular covariate values within trials is small or even zero), again, this does not provide credence for making recommendations based on the trial‐level information because of the aforementioned issues. At best, meta‐regression analyses using the trial‐level should only be viewed as exploratory when the aim is to identify individual‐level associations and should not inform clinical recommendations.…”

Section: Discussionmentioning

confidence: 99%

“…In other words, the within‐ and the across‐trial interaction estimates are now uncorrelated with each other and thus disentangled 9, 26. In contrast, models (1) and (2) provide some weighted average of β W and β A , which will increase power but at the expense of potential ecological bias.…”

Section: Estimation Of Treatment‐covariate Interactions In a One‐stagmentioning

confidence: 99%

“…Further, it is also prone to study‐level confounding and ‘ecological bias’ 8, which means the observed across‐study relationships do not properly reflect the individual‐level relationships within trials. For example, meta‐regression may identify that studies containing a larger proportion of male participants have a larger overall treatment effect; however, this may be due to such studies also having a higher dose of the treatment, and therefore, improved effect is due to the dose and not being male rather than female 9.…”

Section: Introductionmentioning

confidence: 99%

“…In contrast to the two‐stage approach, one‐stage meta‐analysis models allow a more exact likelihood to be specified and automatically account for the correlations amongst parameters 17. However, when investigating treatment‐covariate interactions, it has been shown that the one‐stage approach does not automatically avoid ecological bias when estimating treatment‐covariate interactions; that is, estimation of interaction terms in a one‐stage meta‐analysis might merge (amalgamate) both within‐trial and across‐trial information 9, 20, 21. To avoid potential ecological bias, one needs to separate out within‐trial and across‐trial interaction effects in the model specification 9, 12, 14, 22, 23, which is also recognized in areas outside the meta‐analysis field that contains clustering 24, 25, 26, 27.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

One-stage individual participant data meta-analysis models: estimation of treatment-covariate interactions must avoid ecological bias by separating out within-trial and across-trial information

et al. 2016

Self Cite

View full text Add to dashboard Cite

Stratified medicine utilizes individual‐level covariates that are associated with a differential treatment effect, also known as treatment‐covariate interactions. When multiple trials are available, meta‐analysis is used to help detect true treatment‐covariate interactions by combining their data. Meta‐regression of trial‐level information is prone to low power and ecological bias, and therefore, individual participant data (IPD) meta‐analyses are preferable to examine interactions utilizing individual‐level information. However, one‐stage IPD models are often wrongly specified, such that interactions are based on amalgamating within‐ and across‐trial information. We compare, through simulations and an applied example, fixed‐effect and random‐effects models for a one‐stage IPD meta‐analysis of time‐to‐event data where the goal is to estimate a treatment‐covariate interaction. We show that it is crucial to centre patient‐level covariates by their mean value in each trial, in order to separate out within‐trial and across‐trial information. Otherwise, bias and coverage of interaction estimates may be adversely affected, leading to potentially erroneous conclusions driven by ecological bias. We revisit an IPD meta‐analysis of five epilepsy trials and examine age as a treatment effect modifier. The interaction is −0.011 (95% CI: −0.019 to −0.003; p = 0.004), and thus highly significant, when amalgamating within‐trial and across‐trial information. However, when separating within‐trial from across‐trial information, the interaction is −0.007 (95% CI: −0.019 to 0.005; p = 0.22), and thus its magnitude and statistical significance are greatly reduced. We recommend that meta‐analysts should only use within‐trial information to examine individual predictors of treatment effect and that one‐stage IPD models should separate within‐trial from across‐trial information to avoid ecological bias. © 2016 The Authors. Statistics in Medicine published by John Wiley & Sons Ltd.

show abstract

Section: Discussionsupporting

confidence: 90%

Section: Discussionmentioning

confidence: 99%

Section: Estimation Of Treatment‐covariate Interactions In a One‐stagmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

One-stage individual participant data meta-analysis models: estimation of treatment-covariate interactions must avoid ecological bias by separating out within-trial and across-trial information

et al. 2016

Self Cite

View full text Add to dashboard Cite

show abstract

Preferred Reporting Items for a Systematic Review and Meta-analysis of Individual Participant Data

Stewart

Clarke

Rovers

et al. 2015

JAMA

Self Cite

1,633

1,197

View full text Add to dashboard Cite

IMPORTANCE Systematic reviews and meta-analyses of individual participant data (IPD) aim to collect, check, and reanalyze individual-level data from all studies addressing a particular research question and are therefore considered a gold standard approach to evidence synthesis. They are likely to be used with increasing frequency as current initiatives to share clinical trial data gain momentum and may be particularly important in reviewing controversial therapeutic areas. OBJECTIVE To develop PRISMA-IPD as a stand-alone extension to the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) Statement, tailored to the specific requirements of reporting systematic reviews and meta-analyses of IPD. Although developed primarily for reviews of randomized trials, many items will apply in other contexts, including reviews of diagnosis and prognosis. DESIGN Development of PRISMA-IPD followed the EQUATOR Network framework guidance and used the existing standard PRISMA Statement as a starting point to draft additional relevant material. A web-based survey informed discussion at an international workshop that included researchers, clinicians, methodologists experienced in conducting systematic reviews and meta-analyses of IPD, and journal editors. The statement was drafted and iterative refinements were made by the project, advisory, and development groups. The PRISMA-IPD Development Group reached agreement on the PRISMA-IPD checklist and flow diagram by consensus. FINDINGS Compared with standard PRISMA, the PRISMA-IPD checklist includes 3 new items that address (1) methods of checking the integrity of the IPD (such as pattern of randomization, data consistency, baseline imbalance, and missing data), (2) reporting any important issues that emerge, and (3) exploring variation (such as whether certain types of individual benefit more from the intervention than others). A further additional item was created by reorganization of standard PRISMA items relating to interpreting results. Wording was modified in 23 items to reflect the IPD approach. CONCLUSIONS AND RELEVANCE PRISMA-IPD provides guidelines for reporting systematic reviews and meta-analyses of IPD.

show abstract

Postoperative Glaucoma Following Infantile Cataract Surgery

et al. 2014

View full text Add to dashboard Cite

IMPORTANCE Infantile cataract surgery bears a significant risk for postoperative glaucoma, and no consensus exists on factors that may reduce this risk.OBJECTIVE To assess the effect of primary intraocular lens implantation and timing of surgery on the incidence of postoperative glaucoma.DATA SOURCES We searched multiple databases to July 14, 2013, to identify studies with eligible patients, including PubMed, MEDLINE, EMBASE, ISI Web of Science, Scopus, Central, Google Scholar, Intute, and Tripdata. We also searched abstracts of ophthalmology society meetings.STUDY SELECTION We included studies reporting on postoperative glaucoma in infants undergoing cataract surgery with regular follow-up for at least 1 year. Infants with concurrent ocular anomalies were excluded. DATA EXTRACTION AND SYNTHESISAuthors of eligible studies were invited to contribute individual patient data on infants who met the inclusion criteria. We also performed an aggregate data meta-analysis of published studies that did not contribute to the individual patient data. Data were pooled using a random-effects model. MAIN OUTCOMES AND MEASURESTime to glaucoma with the effect of primary implantation, additional postoperative intraocular procedures, and age at surgery. RESULTSSeven centers contributed individual patient data on 470 infants with a median age at surgery of 3.0 months and median follow-up of 6.0 years. Eighty patients (17.0%) developed glaucoma at a median follow-up of 4.3 years. Only 2 of these patients had a pseudophakic eye. The risk for postoperative glaucoma appeared to be lower after primary implantation (hazard ratio [HR], 0.10 [95% CI, 0.01-0.70]; P = .02; I 2 = 34%), higher after surgery at 4 weeks or younger (HR, 2.10 [95% CI, 1.14-3.84]; P = .02; I 2 = 0%), and higher after additional procedures (HR, 2.52 [95% CI, 1.11-5.72]; P = .03; I 2 = 32%). In multivariable analysis, additional procedures independently increased the risk for glaucoma (HR, 2.25 [95% CI,; P = .01), and primary implantation independently reduced it (HR, 0.10 [95% CI, 0.01-0.76]; P = .03). Results were similar in the aggregate data meta-analysis that included data from 10 published articles. CONCLUSIONS AND RELEVANCEAlthough confounding factors such as size of the eye and surgeon experience are not accounted for in this meta-analysis, the risk for postoperative glaucoma after infantile cataract surgery appears to be influenced by the timing of surgery, primary implantation, and additional intraocular surgery.

show abstract

Meta‐analysis of continuous outcomes combining individual patient data and aggregate data

Cited by 237 publications

References 43 publications

One-stage individual participant data meta-analysis models: estimation of treatment-covariate interactions must avoid ecological bias by separating out within-trial and across-trial information

One-stage individual participant data meta-analysis models: estimation of treatment-covariate interactions must avoid ecological bias by separating out within-trial and across-trial information

Preferred Reporting Items for a Systematic Review and Meta-analysis of Individual Participant Data

Postoperative Glaucoma Following Infantile Cataract Surgery

Contact Info

Product

Resources

About