1 / 34

Systematic Review Module 10: Quantitative Synthesis II

Systematic Review Module 10: Quantitative Synthesis II. Thomas Trikalinos, MD, PhD Joseph Lau, MD Tufts EPC. CER Process Overview. Learning objectives of this module. Dealing with between-study heterogeneity Promise and danger of subgroup analyses Meta-regression

nedra
Télécharger la présentation

Systematic Review Module 10: Quantitative Synthesis II

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Systematic Review Module 10: Quantitative Synthesis II Thomas Trikalinos, MD, PhDJoseph Lau, MDTufts EPC

  2. CER Process Overview

  3. Learning objectives of this module • Dealing with between-study heterogeneity • Promise and danger of subgroup analyses • Meta-regression • Control rate meta-regression

  4. Homogeneity From Cochrane Database Syst Rev. 2000;(2):CD000505

  5. Heterogeneity: Patellar resurfacing in total knee arthroplasty for pain J Bone Joint Surg Am. 2005;87(7):1438-45

  6. Heterogeneity Diversity of studies in a meta-analysis Typically abundant Arguably the most important role of meta-analytic methodologies is to quantify, explore, and explain between-study heterogeneity

  7. Heterogeneity Methodological heterogeneity Pertains to specifics of study design and analysis(e.g., type of study, length of follow-up, proportion of dropouts and handling thereof) Clinical heterogeneity Pertains to differences in the populations, intervention and co-interventions, outcomes

  8. Statistical heterogeneity Statistical heterogeneity exists when the results of the individual studies are not “consistent” among themselves Clinical heterogeneity Methodological heterogeneity Biases Chance Statistical heterogeneity

  9. Clinical vs. statistical heterogeneity Clinical and methodological heterogeneity is abundant. Our aim is to explore it, and use these observations to formulate interesting hypotheses. Often, but not always, clinical and methodological heterogeneity will result in a statistically significant test Chance, technical issues or biases may result in statistically significant results in heterogeneity tests

  10. RESPONSE SURFACE modeling individual patient data META-REGRESSION modeling summary data SUBGROUP ANALYSES differentiating effects in subgroups OVERALL ESTIMATE combining summary data

  11. Promises of subgroup analyses

  12. J Am Coll Card 1990

  13. Mortality of thrombolytic therapy for AMI meantime to treatment (0-3 hours)

  14. Mortality of thrombolytic therapy for AMI meantime to treatment (3.1-5 hours)

  15. Mortality of thrombolytic therapy for AMI meantime to treatment (5.1-10 hours)

  16. Mortality of thrombolytic therapy for AMI meantime to treatment (> 10 hours)

  17. Vit E and all cause mortality Ann Intern Med. 2005;142(1):37-46.

  18. Hazards of subgroup analyses

  19. From Fibrinolytic Therapy Trialists’ Collaborative Group: Indications for Fibrinolytic Therapy Lancet 343: 311,1994

  20. ISIS-2. Lancet 1988;ii:349-60.Subgroup analyses

  21. ISIS-2. Lancet 1988;ii:349-60.Subgroup analyses

  22. Beyond subgroup analyses:meta-regression

  23. Subgroup analysis Ann Intern Med. 2005;142(1):37-46.

  24. Univariate meta-regression Ann Intern Med. 2005;142(1):37-46.

  25. Meta-regression: Zidovudine monotherapy vs. placebo ~τ’2 ~τ2

  26. Multivariate meta-regression: Effect of Soy on LDL Dose Baseline LDL

  27. Control Rate Meta-Regression • Single covariate included is event rate in the control group (control rate) • Control rate is surrogate for all baseline differences between the studies, in terms of baseline risk for the event of interest. • Can show that underlying risk of event (severity of illness) may explain differences in the treatment effect across studies

  28. Control rate meta-regression in the streptokinase example Stat Med. 1998;17(17):1923-42.

  29. Two types of covariates in meta-regressions Study level covariates vs. participant level covariates • Study level: presence/absence of blinding, intervention dose (in experimental studies) • Participant level: mean age, proportion of diabetics, mean intake of vitamin D (in observational studies)

  30. Spurious associations in meta-regressions and subgroup analyses Meta-regressions that use participant-level covariates can mislead, as they are susceptible to ecological fallacy Associations of treatment effect and participant-level covariates should be interpreted with caution See the quiz

  31. Summary • Subgroup analyses, meta-regressions and control-rate meta-regressions are tools to explore between-study heterogeneity. Do use them to understand your data. • They are mostly hypothesis forming tools. Especially for meta-regressions on patient-level covariates, ecological fallacy may mislead. • Beware when interpreting their results.

More Related