Systematic Review: Meta-analysis II The nuts and bolts of the statistics Alka M. Kanaya, M.D. Assistant Professor of Medicine, Epi/Biostats April 19, 2007.

Slides:



Advertisements
Similar presentations
Meta-analysis: summarising data for two arm trials and other simple outcome studies Steff Lewis statistician.
Advertisements

EVAL 6970: Meta-Analysis Vote Counting, The Sign Test, Power, Publication Bias, and Outliers Dr. Chris L. S. Coryn Spring 2011.
Systematic Review: Analytical Methods of Meta-analysis Stephen Bent, M.D. Assistant Professor of Medicine, Epidemiology and Biostatistics UCSF.
Significance and effect sizes What is the problem with just using p-levels to determine whether one variable has an effect on another? Don’t EVER just.
1 Case-Control Study Design Two groups are selected, one of people with the disease (cases), and the other of people with the same general characteristics.
Chance, bias and confounding
Design and Analysis of Clinical Study 12. Randomized Clinical Trials Dr. Tuan V. Nguyen Garvan Institute of Medical Research Sydney, Australia.
Critical Appraisal of Systematic Reviews Douglas Newberry.
Statistics By Z S Chaudry. Why do I need to know about statistics ? Tested in AKT To understand Journal articles and research papers.
By Dr. Ahmed Mostafa Assist. Prof. of anesthesia & I.C.U. Evidence-based medicine.
Sample Size Determination
Long-term predictive value of assessment of coronary atherosclerosis by contrast- enhanced coronary computed tomography angiography: meta- analysis and.
Cohort Studies Hanna E. Bloomfield, MD, MPH Professor of Medicine Associate Chief of Staff, Research Minneapolis VA Medical Center.
The Bahrain Branch of the UK Cochrane Centre In Collaboration with Reyada Training & Management Consultancy, Dubai-UAE Cochrane Collaboration and Systematic.
Are the results valid? Was the validity of the included studies appraised?
Department of O UTCOMES R ESEARCH. Daniel I. Sessler, M.D. Michael Cudahy Professor and Chair Department of O UTCOMES R ESEARCH The Cleveland Clinic Clinical.
Systematic Reviews Professor Kate O’Donnell. Reviews Reviews (or overviews) are a drawing together of material to make a case. These may, or may not,
Program Evaluation. Program evaluation Methodological techniques of the social sciences social policy public welfare administration.
Data Analysis in Systematic Reviews-Meta Analysis.
Data Analysis in Systematic Reviews
1 ICEBOH Split-mouth studies and systematic reviews Ian Needleman 1 & Helen Worthington 2 1 Unit of Periodontology UCL Eastman Dental Institute International.
Systematic Reviews.
Study design P.Olliaro Nov04. Study designs: observational vs. experimental studies What happened?  Case-control study What’s happening?  Cross-sectional.
Evidence Based Medicine Meta-analysis and systematic reviews Ross Lawrenson.
Introduction to Systematic Reviews Afshin Ostovar Bushehr University of Medical Sciences Bushehr, /9/20151.
A Systematic Review On The Hazards Of Aspirin Discontinuation Among Patients With Or At Risk For Coronary Artery Disease Giuseppe Biondi Zoccai Hemodynamics.
Systematic Review: Analytical Methods of Meta-analysis Stephen Bent, M.D. Assistant Professor of Medicine, Epidemiology and Biostatistics UCSF.
Simon Thornley Meta-analysis: pooling study results.
A systematic meta-analysis of randomized controlled trials for adjuvant chemotherapy for localized resectable soft-tissue sarcoma Nabeel Pervaiz Nigel.
EBCP. Random vs Systemic error Random error: errors in measurement that lead to measured values being inconsistent when repeated measures are taken. Ie:
Literature searching & critical appraisal Chihaya Koriyama August 15, 2011 (Lecture 2)
Meta-analysis and “statistical aggregation” Dave Thompson Dept. of Biostatistics and Epidemiology College of Public Health, OUHSC Learning to Practice.
Meta-analysis 統合分析 蔡崇弘. EBM ( evidence based medicine) Ask Acquire Appraising Apply Audit.
The Campbell Collaborationwww.campbellcollaboration.org C2 Training: May 9 – 10, 2011 Introduction to meta-analysis.
RevMan for Registrars Paul Glue, Psychological Medicine What is EBM? What is EBM? Different approaches/tools Different approaches/tools Systematic reviews.
META-ANALYSIS: THE ART AND SCIENCE OF COMBINING INFORMATION Ora Paltiel, October 28, 2014.
How to read a paper D. Singh-Ranger. Academic viva 2 papers 1 hour to read both Viva on both papers Summary-what is the paper about.
Understanding Medical Articles and Reports Linda Vincent, MPH UCSF Breast SPORE Advocate September 24,
PH 401: Meta-analysis Eunice Pyon, PharmD (718) , HS 506.
Lecture 9: Analysis of intervention studies Randomized trial - categorical outcome Measures of risk: –incidence rate of an adverse event (death, etc) It.
EBM Conference (Day 2). Funding Bias “He who pays, Calls the Tune” Some Facts (& Myths) Is industry research more likely to be published No Is industry.
Objectives  Identify the key elements of a good randomised controlled study  To clarify the process of meta analysis and developing a systematic review.
CAT 5: How to Read an Article about a Systematic Review Maribeth Chitkara, MD Rachel Boykan, MD.
Sifting through the evidence Sarah Fradsham. Types of Evidence Primary Literature Observational studies Case Report Case Series Case Control Study Cohort.
EBM --- Journal Reading Presenter :呂宥達 Date : 2005/10/27.
1 Lecture 10: Meta-analysis of intervention studies Introduction to meta-analysis Selection of studies Abstraction of information Quality scores Methods.
Systematic Synthesis of the Literature: Introduction to Meta-analysis Linda N. Meurer, MD, MPH Department of Family and Community Medicine.
Younghun Han Department of Epidemiology UT MD Anderson Cancer Center
Making Randomized Clinical Trials Seem Less Random Andrew P.J. Olson, MD Assistant Professor Departments of Medicine and Pediatrics University of Minnesota.
1 Lecture 10: Meta-analysis of intervention studies Introduction to meta-analysis Selection of studies Abstraction of information Quality scores Methods.
Is a meta-analysis right for me? Jaime Peters June 2014.
How to Conduct a Meta-Analysis Arindam Basu MD MPH About the Author Required Browsing.
Meta-analysis of observational studies Nicole Vogelzangs Department of Psychiatry & EMGO + institute.
Brady Et Al., "sequential compression device compliance in postoperative obstetrics and gynecology patients", obstetrics and gynecology, vol. 125, no.
Benefits and Pitfalls of Systematic Reviews and Meta-Analyses
How to read a paper D. Singh-Ranger.
Systematic Review Systematic review
Supplementary Table 1. PRISMA checklist
Randomized Trials: A Brief Overview
Fatimah Al-Ani 1,2,. MD MRCP, Jose Maria Bastida Bermejo3,
Heterogeneity and sources of bias
Random error, Confidence intervals and P-values
Lecture 4: Meta-analysis
Chapter 7 The Hierarchy of Evidence
Narrative Reviews Limitations: Subjectivity inherent:
Comprehensive Meta-Analysis of DES vs
Publication Bias in Systematic Reviews
What is a review? An article which looks at a question or subject and seeks to summarise and bring together evidence on a health topic. Ask What is a review?
Systematic Reviews and Meta-Analysis -Part 2-
Introduction to Systematic Reviews
Presentation transcript:

Systematic Review: Meta-analysis II The nuts and bolts of the statistics Alka M. Kanaya, M.D. Assistant Professor of Medicine, Epi/Biostats April 19, 2007

Goals 1.Understand statistical issues for MA  summary estimate and variance models methods  heterogeneity  publication bias 2.Carry on an intelligent conversation with your statistician 3. Know if published MA used appropriate methods

Meta-analysis: the Steps 1.Formulate a question, eligibility criteria 2.Perform a systematic literature search 3.Abstract the data 4.Perform a statistical analysis 5.Calculate the summary effect size 6.Calculate the summary effect size for subgroups 7.Check for heterogeneity 8.Check for publication bias

Clinical Case  5 y.o. girl c/o ear pain and is found to have an acute otitis media.  Should she get antibiotics? Research Questions: 1.Are antibiotics effective for pain relief in children with acute OM? 2.Do antibiotics reduce complications of OM (mastoiditis, hearing problems)?

Systematic Review part  Inclusion criteria: –RCT of antibiotic vs. placebo –Children –Without tympanostomy tubes –With OM (regardless of setting of recruitment) –Patient-relevant outcomes  8 trials with total of 2,287 kids Glasziou, Cochrane Review, 2005

Goal #1: “Best” estimate  Combine findings from several studies to get the "best" estimate  Calculate weighted mean effect estimate, or a summary effect estimate 700Total RRNStudy Do antibiotics reduce pain? 3 RCTs: summary effect estimate= Σ (N i x effect estimate i ) = 640 =0.914 Σ(N i ) 700

Goal #1: Calculate weighted mean effect estimate Summary = Σ (weight i x effect estimate i ) = 30.3 = 0.99 effect estimate Σ(weight i ) 30.5 StudyNRRVar RRWeight Total700

Goal #2: Determine if the summary effect is significant  Calculate variance of summary effect estimate, or the 95% CI around the summary estimate Variance of summary estimate = 1 Σ(weights i ) Variance of summary estimate = _1_ = % CI = √0.03 = Summary OR and 95% CI = 0.99 ( )

Type of Model? Variance RR s = 1/  w i Weight i = 1 + D variance RR i Weight i = 1 variance RR i Variance of individual studies + variance of differences between studies Weights: variance of individual studies Existing studies are a random sample Existing studies are the entire population Goal: estimate the “true” effect Goal: weighted average of risk from existing studies Random EffectsFixed Effects

Model: Random Effects Model: Summary RR b Summary RR a b a

Random VS. Fixed Effects Models Practical Implications of the Choice  Summary estimates: usually similar  Variance: RE model produces large variance of the summary estimate  Confidence intervals: RE model produces wider confidence intervals  Statistical significance: less likely with RE model BOTTOM LINE:  If the individual study findings are similar, the model makes little difference in estimate or statistical significance.  If the individual study findings are heterogeneous, the model can affect the statistical significance.

Which method? ModelMethodEffect estimate Data Fixed Effects: Mantel- Haenszel Ratio (OR, RR)Crude (2x2) General Variance-based Ratio (OR, RR) Difference (risk, rate) Adjusted ratio & CI Crude (2x2) Random Effects: DerSimonian & Laird Ratio, difference Crude (2x2) General Variance-based Ratio (OR, RR) Difference (risk, rate) Adjusted ratio & CI Crude (2x2)

Which method to use? Types of Studies in Meta-analysis Method to Use Randomized trials: Any method (usually Mantel-Haenszel or DerSimonian & Laird Observational studies: General Variance Based

Mantel-Haenszel Method (Fixed Effects Model) DiseasedNot diseased Treated (exposed) a i c i Not treated (unexposed) b i d i OR i = a i / c i = a i x d i lnOR mh = Σ (w i x lnOR i ) b i / d i b i x c i Σw i variance lnOR i = variance OR mh = 1 a i b i c i d i Σ w i weight i = (w i ) = 1 variance lnOR i 95% CI = e lnORmh  (1.96 x √variance lnOR mh )

Randomized Trials of Antibiotic Rx for acute OM to prevent TM perforation Study 1 PerforationNo Perforation Antibiotic 1114 Placebo 3116 Study 2 PerforationNo Perforation Antibiotic 7 65 Placebo Calculate OR i for each study: OR 1 =1 x 116 = 0.34lnOR 1 = x 114 OR 2 = _______ = 0.58lnOR 2 = -0.54

Randomized Trials of Antibiotic Rx for acute OM to prevent TM perforation 2. Calculate variance lnOR i for each study: Var ln OR 1 = = Var ln OR 2 = ______________ = Calculate w i for each study: w 1 = 1 = w 2 = ________ = 3.85

Study 1 PerforationNo Perforation Antibiotic 1114 Placebo 3116 Study 2 PerforationNo Perforation Antibiotic 7 65 Placebo Calculate the w i x ln OR i for each study: w 1 x lnOR 1 = 0.74 x = w 2 x lnOR 2 = 3.85 x = Randomized Trials of Antibiotic Rx for acute OM to prevent TM perforation

5. Calculate the sum of the w i w 1 + w 2 = = Calculate lnOR mh = Σ (w i x lnOR i ) = = Σ w i 4.59 = OR mh = Calculate variance OR mh = 1 = 1 = 0.22 Σ w i Calculate 95% CI = e lnORmh + (1.96 x √ variance lnORmh) = e (1.96 x √ 0.22) = Summary OR = 0.53 (95% CI 0.21 – 1.34) Randomized Trials of Antibiotic Rx for acute OM to prevent TM perforation

Dersimonian and Laird Method (Random Effects Model) Similar formula to Mantel-Haenszel: ln OR dl = Σ (w i x ln OR i ) w i = 1 Σw i variance i + D  Where D gets larger as the OR (or effect estimate) of the individual studies vary from the summary estimate

General Variance-Based Method (Fixed or Random Effects) Confidence Intervals: ln OR s = Σ (w i x ln OR i )w i = 1 ______ Σw i variance lnOR i (+D) Variance lnOR i = ln OR i / OR l 2 or ln OR u / OR i Where OR i = OR on i th study OR l = lower bound of 95 % CI for i th study OR u = upper bound of 95 % CI for i th study  Should always be used for MA of Observational studies  Uses adjusted effect estimates  Preserves adjustment for confounding

Choice of Model and Method in Meta-Analysis What type of studies are you summarizing? Randomized TrialsObservational Studies  Either Model  Any Method  Either Model  Confidence Interval Method

Heterogeneity  Are you comparing apples and oranges?  Clinical heterogeneity: are studies asking same question?  Statistical heterogeneity: is the variation likely to have occurred by chance?  Measures how different each individual OR/RR is from the summary OR/RR.  Studies whose OR/RRs are very different from the summary OR/RRs contribute greatly to the heterogeneity, especially if they are weighted heavily.

Problem of Heterogeneity  Study findings are different and should not be combined StudyOR StudyOR

Statistical tests of Homogeneity  Is the variation in the individual study findings likely due to chance? H o : Effect estimate in each study is the same (or homogeneous) H a : Effect estimate in each study is not the same (or heterogeneous) Q = Σ(w i x (ln OR mh – ln OR i ) 2 ) df = (N studies -1) p < 0.05 or 0.10 = reject null, i.e., studies are heterogeneous  8 trials of antibiotics vs. P o for OM, pain outcome:  Q for homogeneity: p=0.91

Subgroup & Sensitivity Analysis  Subgroup Analysis – MA of a subgroup of eligible studies  age  ethnicity  risk factors  treatment  Sensitivity Analysis – add or delete questionable studies  eligibility  treatment

Subgroup Analysis OR95% CIN Ever user Of estrogen: All eligible studies Cohort studies Case-Control studies 2.3* 1.7* 2.4* Dose of estrogen: 0.3 mg mg  1.25 mg Duration of use: < 1 year 1-5 years 5-10 years  10 years * Regimen:Cyclic Daily 3.0* 2.9* * p for homogeneity < 0.05

Subgroup Analysis Antibiotics vs. Placebo for acute OM Outcome: abnormal tympanometry Study0 – 1 mo f/u RR (95% CI) 2 – 3 mo f/u RR (95% CI) Appelman, (0.25 – 1.26)NA Burke, (0.62 – 1.84)0.58 (0.31 – 1.08) Mygind, (0.49 – 1.94)1.09 (0.52 – 2.31) Summary estimate : 0.91 (0.62 – 1.32)0.75 (0.47 – 1.21) p-for-heterogeneity:

Sensitivity Analyses Performed to test the robustness of the findings To fairly assess and acknowledge the limitations Address publication bias (funnel plots, number needed to change result, etc..)

Sensitivity Analysis Aspirin + Heparin vs. Aspirin alone for Unstable Angina MI or Death Study (reference)AspirinAspirin + heparinRR (95% CI) Theroux et al, /121 (3%)2/122 (2%)0.50 ( ) RISC Group, /189 (4%)3/210 (1%)0.39 ( ) Cohen et al, /32 (3%)0/37 (0%)0.29 ( ) Cohen et al, /109 (8%)4/105 (4%)0.46 ( ) Holdright et al, /131 (31%)42/154 (27%)0.89 ( ) Gurfinkel et al, /73 (10%)4/70 (6%)0.60 ( ) Total/Summary:68/655 (10%)55/698 (8%)0.67 (0.44-1/02)  Remove Holdright: RR s = 0.45 (95% CI ) ; p-for-hetero=0.71  Add data from two additional trials of LMWH: RR s = 0.56 (95% CI ); p for heterogeneity: 0.52  Fixed effects model, Mantel-Haenszel method = same findings

Publication Bias  Published studies may not be representative of all studies ever conducted.  Selective publication of studies based on strength & direction of results & language.  AKA “positive outcome bias”

Minimizing Publication Bias Search bibliographies of published papers Consult with experts Search for unpublished data Clinical Trial Registries (NIH, VA) Institutional Review Boards Pharmaceutical companies Hand searches Consider studies not published in English Stern, BMJ, 2001

Statistical Approaches to Publication Bias  Correlation between study sample size (or weight or variance) and effect estimate  Funnel plot  Other fancy statistical methods: estimate number of unpublished studies that must exist to invalidate the results of the meta- analysis. “File drawer” “Fail-safe N” eliminate the studies that may have been published due to bias

Association of Estrogen use and Endometrial Cancer Correlation of sample size and RR: rho = 0.68; p = 0.08 FUNNEL GRAPH Relative Risk of Endometrial Cancer

RCTs of Heparin plus ASA vs. ASA Correlation of sample size and RR: rho = 0.25; p= 0.64 FUNNEL GRAPH Relative Risk for MI or Death Favors Heparin+ASA Favors ASA Sample Size

Presentation of the Results Tables:  Study Characteristics  population  sample size  definition of intervention  definition of the outcome  important design features (validity of the data) - randomization - blinding - follow-up - compliance  Study Findings  main and secondary outcomes  outcomes by subgroup  sensitivity analysis findings

Table 1. Characteristics of 6 randomized trials of aspirin + heparin vs. aspirin alone to prevent MI and death in patients admitted with unstable angina Study (ref.)BlindingAspirin DoseGoal PTTDuration of Heparin Theroux, Participants & Investigators 325 mg twice per day x normal6 days RISC Group, None75 mg dailyNot stated5 days Cohen, None 80/325 mg daily* 2 x normal3-4 days Cohen, Participants162.5 mg daily2 x normal3-4 days Holdright, Participants150 mg daily1.5-2 x normal2 days Gurfinkel, Participants & Investigators 200 mg daily2 x normal5-7 days

Bent, AMJ, 1999 No Caption Found Antibiotics vs. Placebo in Acute Bronchitis

Take Home Messages  You can do a meta-analysis  Good start on becoming an expert in your field  Your work should be reproducible  Your conclusions should be obvious  Include a statistician on you team