108
Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion Escher 3.2: towards effective, transparent and accountable assessment of benefit-risk using information technology and evidence synthesis Gert van Valkenhoef Department of Epidemiology, University Medical Center Groningen (NL), Faculty of Economics and Business, University of Groningen (NL) Escher 3.x Cluster Meeting, 8 Dec 2010 Utrecht, The Netherlands

Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

  • Upload
    others

  • View
    5

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Escher 3.2: towards effective, transparent andaccountable assessment of benefit-risk

using information technology and evidence synthesis

Gert van Valkenhoef

Department of Epidemiology, University Medical Center Groningen (NL),Faculty of Economics and Business, University of Groningen (NL)

Escher 3.x Cluster Meeting, 8 Dec 2010Utrecht, The Netherlands

Page 2: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Outline

1 Introduction2 Meta-analysis3 Network meta-analysis4 Benefit-risk analysis5 Discussion

After every part, there will bean opportunity to askquestions.

Page 3: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Outline

1 Introduction2 Meta-analysis3 Network meta-analysis4 Benefit-risk analysis5 Discussion

After every part, there will bean opportunity to askquestions.

Page 4: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Escher 3.2 Goals

Develop a drug information system:

Effective knowledge access and management

Answer drug efficacy and safety questions

in an efficient, transparent and accountable waywithin and across compoundsfor a broad audience (including regulators)

Improve consistency in regulatory decision making

Based on systematic review and meta-analysis

Page 5: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Effective knowledge access: problems

Review of existing systems:

Evidence-based decision making time-consuming/error-prone

No comprehensive source of trial information existsTrial information is insufficiently structured

Missed opportunities to introduce more structure

Trial registration, regulatory submission and systematic review

It is unclear how the information should be structured

Prototypes should be developed now, to discover thisRelated manuscripts:1) G. van Valkenhoef, T. Tervonen, B. de Brock, H. Hillege, Deficiencies in the transfer and availability of clinicalevidence in drug development and regulation. Manuscript under review.2) T. Tervonen, E.O. de Brock, P.A. de Graeff and H.L. Hillege (2010). Current status and future perspectives onDrug Information Systems.Proceedings of the 18th European Conference on Information Systems (ECIS2010),June 6-9, 2010, Pretoria, South Africa.

Page 6: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Prototype: global requirements

Interviews with major stakeholders

To develop the overall vision for the prototype

Database of clinical trialsAnswer efficacy/safety questionsStreamline benefit-risk decision makingFor regulatory authoritiesUsing aggregated data

Page 7: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Aggregate data

Regulatory submissions in Europe contain aggregate data

EMA, SmPC, Galvus (EMEA/H/C/000771 -II/0007), updated 2010-04-27.

Page 8: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Aggregate data

Journal articles report aggregate data

Chouinard G, Saxena B, Belanger MC, Ravindran A, Bakish D, Beauclair L, et al. A Canadian multicenter,double-blind study of paroxetine and fluoxetine in major depressive disorder. J Affect Disord. 1999;54:39-48.

Page 9: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Aggregate data

Trials registered with ClinicalTrials.gov have aggregate results

GlaxoSmithKline, “Controlled-release Paroxetine in Major Depressive Disorder (Double-blind, Placebo-controlledStudy)”, ClinicalTrials.gov NCT00866294, updated October 14, 2010.

Page 10: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

ADDIS: Aggregate Data Drug Information System

Assisted evidence synthesis and benefit-risk assessment

Based on a database of clinical trials

Focussed on aggregated dataRelated manuscripts:3) G. van Valkenhoef, T. Tervonen, T. Zwinkels, B. de Brock, H. Hillege, ADDIS: a decision support system forevidence-based medicine. Manuscript under review.

Page 11: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Development of ADDIS: concurrent engineering

Software Development

Methodology Research

Open problems Knowledge, methods Case studies

Feedback, use cases

Open problems

Page 12: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Development of ADDIS: concurrent engineering

Software Development

Methodology Research

Open problems

Knowledge, methods Case studies

Feedback, use cases

Open problems

Page 13: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Development of ADDIS: concurrent engineering

Software Development

Methodology Research

Open problems Knowledge, methods

Case studies

Feedback, use cases

Open problems

Page 14: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Development of ADDIS: concurrent engineering

Software Development

Methodology Research

Open problems Knowledge, methods Case studies

Feedback, use cases

Open problems

Page 15: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Development of ADDIS: concurrent engineering

Software Development

Methodology Research

Open problems Knowledge, methods Case studies

Feedback, use cases

Open problems

Page 16: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Development of ADDIS: concurrent engineering

Software Development

Methodology Research

Open problems Knowledge, methods Case studies

Feedback, use cases

Open problems

Page 17: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Development of ADDIS: agile

ADDIS requirements highly uncertain

Only vague goals can be set

Much is expected to be discovered ‘on the way’

Agile software development

No full up-front specification of requirements

But: short-term plans and periodic re-evaluation

Supported by 2-3 part-time programmers (since Oct 2009)Related manuscripts:4) G. van Valkenhoef, T. Tervonen, B. de Brock, D. Postmus, Product and Release planning practices for ExtremeProgramming. Proceedings of the 11th International Conference on Agile Software Development (XP2010).5) G. van Valkenhoef, T. Tervonen, B. de Brock, D. Postmus, Quantitative release planning in ExtremeProgramming. Manuscript under review.

Page 18: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Development of ADDIS: agile

ADDIS requirements highly uncertain

Only vague goals can be set

Much is expected to be discovered ‘on the way’

Agile software development

No full up-front specification of requirements

But: short-term plans and periodic re-evaluation

Supported by 2-3 part-time programmers (since Oct 2009)Related manuscripts:4) G. van Valkenhoef, T. Tervonen, B. de Brock, D. Postmus, Product and Release planning practices for ExtremeProgramming. Proceedings of the 11th International Conference on Agile Software Development (XP2010).5) G. van Valkenhoef, T. Tervonen, B. de Brock, D. Postmus, Quantitative release planning in ExtremeProgramming. Manuscript under review.

Page 19: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Development of ADDIS: agile

Page 20: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Development of ADDIS: open

Open development (http://drugis.org/)

Nightly builds (daily), development builds (bi-weekly)

Release: ca. every 3 months

Mailing list

Subscribe if you’re interested!

Public issue tracker

Anyone can report bugs and track progressRoadmap: short-term plans

Page 21: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Development of ADDIS: open

Page 22: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Development of ADDIS: open

Page 23: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Development of ADDIS: open source

Open source

Aiming for scientific impact

Ensures others will be able to continue the project

Anyone worried about bugs can review the source code

Allows us to re-use many existing OSS components

Page 24: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Development of ADDIS: open source

Page 25: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Questions?

Page 26: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Introduction

ADDIS global requirements:

Database of clinical trials

Answer efficacy/safety questions

Streamline benefit-risk decision making

For regulatory authorities

Using aggregated data

Intermediate goal: ‘dynamic Cochrane’ (automated meta-analysis)

Store trials in sufficient detail to do meta-analysis

Discover required data-model

Page 27: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Introduction

ADDIS global requirements:

Database of clinical trials

Answer efficacy/safety questions

Streamline benefit-risk decision making

For regulatory authorities

Using aggregated data

Intermediate goal: ‘dynamic Cochrane’ (automated meta-analysis)

Store trials in sufficient detail to do meta-analysis

Discover required data-model

Page 28: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Meta-analysis

Hansen et al. Ann Intern Med 2005;143:415-426

Page 29: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Meta-analysis

Hansen et al. Ann Intern Med 2005;143:415-426

Page 30: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Meta-analysis

Hansen et al. Ann Intern Med 2005;143:415-426

Page 31: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Meta-analysis

Hansen et al. Ann Intern Med 2005;143:415-426

Page 32: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Meta-analysis in ADDIS

Supported since ADDIS v0.4 (December 2009)

Database of trials + characteristics + outcomes

Development of data model

Related manuscripts:3) G. van Valkenhoef, T. Tervonen, T. Zwinkels, B. de Brock, H. Hillege, ADDIS: a decision support system forevidence-based medicine. Manuscript under review.

Page 33: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Limits of meta-analysis (1)

Hansen et al. (2005) systematic review:

46 studies comparing n = 10 second-generation AD

On efficacy (HAM-D responders) and adverse events

Page 34: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Limits of meta-analysis (1)

Fluoxetine

Paroxetine

Sertraline

Venlafaxine

6

5

6

3 1

2

Hansen et al. Ann Intern Med 2005;143:415-426

Page 35: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Limits of meta-analysis (1)

Fluoxetine

Paroxetine

Sertraline

Venlafaxine

6

5

6

3 1

2

Hansen et al. Ann Intern Med 2005;143:415-426

Page 36: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Limits of meta-analysis (1)

Fluoxetine

Paroxetine

Sertraline

Venlafaxine

6

5

6

3 1

2

Hansen et al. Ann Intern Med 2005;143:415-426

Page 37: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Limits of meta-analysis (2)

Fluoxetine

Paroxetine

Sertraline

Venlafaxine

6

5

6

3 1

2

Hansen et al. (2005) systematic review:

46 studies comparing n = 10 second-generation AD

Only 3 meta-analyses, all against fluoxetine

Page 38: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Limits of meta-analysis (2)

Fluoxetine

Paroxetine

Sertraline

Venlafaxine

6

5

6

3 1

2

How to compare paroxetine, sertraline and venlafaxine?

Can we compare sertraline/venlafaxine?

Only one direct trialIgnoring the 11 trials sertr-fluox-venlaIs this justified?

When comparing fluox/parox or fluox/sertr?

Can we ignore the 3 parox-sertr trials?

Parox as comparator → same conclusions?

Page 39: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Limits of meta-analysis (2)

Fluoxetine

Paroxetine

Sertraline

Venlafaxine

6

5

6

3 1

2

How to compare paroxetine, sertraline and venlafaxine?

Can we compare sertraline/venlafaxine?

Only one direct trialIgnoring the 11 trials sertr-fluox-venlaIs this justified?

When comparing fluox/parox or fluox/sertr?

Can we ignore the 3 parox-sertr trials?

Parox as comparator → same conclusions?

Page 40: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Limits of meta-analysis (2)

Fluoxetine

Paroxetine

Sertraline

Venlafaxine

6

5

6

3 1

2

How to compare paroxetine, sertraline and venlafaxine?

Can we compare sertraline/venlafaxine?

Only one direct trialIgnoring the 11 trials sertr-fluox-venla

Is this justified?

When comparing fluox/parox or fluox/sertr?

Can we ignore the 3 parox-sertr trials?

Parox as comparator → same conclusions?

Page 41: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Limits of meta-analysis (2)

Fluoxetine

Paroxetine

Sertraline

Venlafaxine

6

5

6

3 1

2

How to compare paroxetine, sertraline and venlafaxine?

Can we compare sertraline/venlafaxine?

Only one direct trialIgnoring the 11 trials sertr-fluox-venlaIs this justified?

When comparing fluox/parox or fluox/sertr?

Can we ignore the 3 parox-sertr trials?

Parox as comparator → same conclusions?

Page 42: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Limits of meta-analysis (2)

Fluoxetine

Paroxetine

Sertraline

Venlafaxine

6

5

6

3 1

2

How to compare paroxetine, sertraline and venlafaxine?

Can we compare sertraline/venlafaxine?

Only one direct trialIgnoring the 11 trials sertr-fluox-venlaIs this justified?

When comparing fluox/parox or fluox/sertr?

Can we ignore the 3 parox-sertr trials?

Parox as comparator → same conclusions?

Page 43: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Limits of meta-analysis (2)

Fluoxetine

Paroxetine

Sertraline

Venlafaxine

6

5

6

3 1

2

How to compare paroxetine, sertraline and venlafaxine?

Can we compare sertraline/venlafaxine?

Only one direct trialIgnoring the 11 trials sertr-fluox-venlaIs this justified?

When comparing fluox/parox or fluox/sertr?

Can we ignore the 3 parox-sertr trials?

Parox as comparator → same conclusions?

Page 44: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Conclusion

Meta-analysis is good if we compare two drugs

It is problematic for more

Selection bias: choice of common comparator?Are results of different comparisons consistent?

We need a way to include all trials/drugs

Page 45: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Questions?

Page 46: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Introduction

ADDIS global requirements:

Database of clinical trials

Answer efficacy/safety questions

Streamline benefit-risk decision making

For regulatory authorities

Using aggregated data

Intermediate goal: automated network meta-analysis

Meta-analysis of > 2 drugs

No existing software does this

Immediate value to scientific community

Page 47: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Introduction

ADDIS global requirements:

Database of clinical trials

Answer efficacy/safety questions

Streamline benefit-risk decision making

For regulatory authorities

Using aggregated data

Intermediate goal: automated network meta-analysis

Meta-analysis of > 2 drugs

No existing software does this

Immediate value to scientific community

Page 48: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Network meta-analysis

46 studies comparing n = 10 second-generation AD

Paroxetine

Bupropion

(1)

Duloxetine

(1)

Mirtazapine

(2)

Venlafaxine

(2)

Sertraline

(3)

(1)

Escitalopram

(2)

Fluoxetine

(8)

(2)

(1)

(1)

(7)

Fluvoxamine

(2)

(6)

Citalopram

(1)

(3) (1) (2)

(1) (2)

Network meta-analysis: include all evidence in one analysis

Page 49: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Network Meta-Analysis models

Network meta-analysis models are difficult to specify

Automated in ADDISRelated manuscripts:3) G. van Valkenhoef, T. Tervonen, T. Zwinkels, B. de Brock, H. Hillege, ADDIS: a decision support system forevidence-based medicine. Manuscript under review.6) G. van Valkenhoef, T. Tervonen, B. de Brock, H. Hillege, Algorithmic Parameterization of Mixed TreatmentComparisons. Manuscript under review.7) G. van Valkenhoef, B. de Brock, H. Hillege, Automating network meta-analysis. Initiated (conference paper).

Page 50: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Network Meta-Analysis models

Network meta-analysis models are difficult to specify

Automated in ADDIS

Related manuscripts:3) G. van Valkenhoef, T. Tervonen, T. Zwinkels, B. de Brock, H. Hillege, ADDIS: a decision support system forevidence-based medicine. Manuscript under review.6) G. van Valkenhoef, T. Tervonen, B. de Brock, H. Hillege, Algorithmic Parameterization of Mixed TreatmentComparisons. Manuscript under review.7) G. van Valkenhoef, B. de Brock, H. Hillege, Automating network meta-analysis. Initiated (conference paper).

Page 51: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Network Meta-Analysis models

Network meta-analysis models are difficult to specify

Automated in ADDISRelated manuscripts:3) G. van Valkenhoef, T. Tervonen, T. Zwinkels, B. de Brock, H. Hillege, ADDIS: a decision support system forevidence-based medicine. Manuscript under review.6) G. van Valkenhoef, T. Tervonen, B. de Brock, H. Hillege, Algorithmic Parameterization of Mixed TreatmentComparisons. Manuscript under review.7) G. van Valkenhoef, B. de Brock, H. Hillege, Automating network meta-analysis. Initiated (conference paper).

Page 52: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Example: data

Study Fluox Parox VenlaChouinard et al, 1999 67/101 67/102De Wilde et al, 1993 25/41 24/37Fava et al, 1998 31/54 32/55Fava et al, 2002 57/92 64/96Gagiano, 1993 27/45 30/45Schone and Ludwig, 1993 9/52 20/54Alves et al, 1999 30/47 25/40De Nayer et al, 2002 27/73 37/73Dierick et al, 1996 95/161 107/153Rudolph and Feiger, 1999 52/103 57/100Silverstone and Ravindran, 1999 77/121 84/128Tylee et al, 1997 58/170 67/171Ballus et al, 2000 23/43 25/41McPartlin et al, 1998 128/178 137/183

Page 53: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Example: consistency

Fluox

Parox Venla

df ,vdf ,p

dp,v

pair-wise OR network OR

df ,p 1.24 (0.92, 1.67)

1.22 (0.92, 1.61)

df ,v 1.30 (1.03, 1.65)

1.34 (1.08, 1.67)

dp,v 1.20 (0.80, 1.82)

1.11 (0.82, 1.50)

Page 54: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Example: consistency

Fluox

Parox Venla

df ,vdf ,p

dp,v

pair-wise OR network OR

df ,p 1.24 (0.92, 1.67)

1.22 (0.92, 1.61)

df ,v 1.30 (1.03, 1.65)

1.34 (1.08, 1.67)

dp,v 1.20 (0.80, 1.82)

1.11 (0.82, 1.50)

Page 55: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Example: consistency

Fluox

Parox Venla

df ,vdf ,p

dp,v

pair-wise OR network OR

df ,p 1.24 (0.92, 1.67)

1.22 (0.92, 1.61)

df ,v 1.30 (1.03, 1.65)

1.34 (1.08, 1.67)

dp,v 1.20 (0.80, 1.82)

1.11 (0.82, 1.50)

Page 56: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Example: consistency

Fluox

Parox Venla

df ,v = df ,p + dp,vdf ,p

dp,vassume consistency: direct andindirect estimates lead to thesame conclusions.

pair-wise OR network OR

df ,p 1.24 (0.92, 1.67)

1.22 (0.92, 1.61)

df ,v 1.30 (1.03, 1.65)

1.34 (1.08, 1.67)

dp,v 1.20 (0.80, 1.82)

1.11 (0.82, 1.50)

Page 57: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Example: consistency

Fluox

Parox Venla

df ,v = df ,p + dp,vdf ,p

dp,vassume consistency: direct andindirect estimates lead to thesame conclusions.

pair-wise OR network OR

df ,p 1.24 (0.92, 1.67) 1.22 (0.92, 1.61)df ,v 1.30 (1.03, 1.65) 1.34 (1.08, 1.67)dp,v 1.20 (0.80, 1.82) 1.11 (0.82, 1.50)

Page 58: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Consistency

We assume consistency: direct and indirect estimates lead to thesame conclusions.

Estimate all relative effects simultaneously

Including all studies

Leading to consistent conclusions

Also estimate missing comparisons

Page 59: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Extended example

Fluox

Parox Venla

Sertr

Escit

Complexity → consistency greaterconcern

Pair-wise against Fluox, Escitexcluded

Yet, evidence suggests Escit>Fluox

How do the other drugs compare?

Escit 0.59 (0.37, 0.94) 0.69 (0.41, 1.15) 0.74 (0.44, 1.25) 0.81 (0.53, 1.24)Fluox 1.18 (0.91, 1.52) 1.27 (0.99, 1.63) 1.38 (1.10, 1.72)

Parox 1.08 (0.77, 1.51) 1.17 (0.86, 1.59)Sertr 1.09 (0.80, 1.48)

Venla

Page 60: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Extended example

Fluox

Parox Venla

Sertr

Escit

Complexity → consistency greaterconcern

Pair-wise against Fluox, Escitexcluded

Yet, evidence suggests Escit>Fluox

How do the other drugs compare?

Escit 0.59 (0.37, 0.94) 0.69 (0.41, 1.15) 0.74 (0.44, 1.25) 0.81 (0.53, 1.24)Fluox 1.18 (0.91, 1.52) 1.27 (0.99, 1.63) 1.38 (1.10, 1.72)

Parox 1.08 (0.77, 1.51) 1.17 (0.86, 1.59)Sertr 1.09 (0.80, 1.48)

Venla

Page 61: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Extended example

Fluox

Parox Venla

Sertr

Escit

Complexity → consistency greaterconcern

Pair-wise against Fluox, Escitexcluded

Yet, evidence suggests Escit>Fluox

How do the other drugs compare?

Escit 0.59 (0.37, 0.94) 0.69 (0.41, 1.15) 0.74 (0.44, 1.25) 0.81 (0.53, 1.24)Fluox 1.18 (0.91, 1.52) 1.27 (0.99, 1.63) 1.38 (1.10, 1.72)

Parox 1.08 (0.77, 1.51) 1.17 (0.86, 1.59)Sertr 1.09 (0.80, 1.48)

Venla

Page 62: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Extended example

Fluox

Parox Venla

Sertr

Escit

Complexity → consistency greaterconcern

Pair-wise against Fluox, Escitexcluded

Yet, evidence suggests Escit>Fluox

How do the other drugs compare?

Escit 0.59 (0.37, 0.94) 0.69 (0.41, 1.15) 0.74 (0.44, 1.25) 0.81 (0.53, 1.24)Fluox 1.18 (0.91, 1.52) 1.27 (0.99, 1.63) 1.38 (1.10, 1.72)

Parox 1.08 (0.77, 1.51) 1.17 (0.86, 1.59)Sertr 1.09 (0.80, 1.48)

Venla

Page 63: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Extended example

Fluox

Parox Venla

Sertr

Escit

Complexity → consistency greaterconcern

Pair-wise against Fluox, Escitexcluded

Yet, evidence suggests Escit>Fluox

How do the other drugs compare?

Escit 0.59 (0.37, 0.94) 0.69 (0.41, 1.15) 0.74 (0.44, 1.25) 0.81 (0.53, 1.24)Fluox 1.18 (0.91, 1.52) 1.27 (0.99, 1.63) 1.38 (1.10, 1.72)

Parox 1.08 (0.77, 1.51) 1.17 (0.86, 1.59)Sertr 1.09 (0.80, 1.48)

Venla

Page 64: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Extended example

Fluox

Parox Venla

Sertr

Escit

Complexity → consistency greaterconcern

Pair-wise against Fluox, Escitexcluded

Yet, evidence suggests Escit>Fluox

How do the other drugs compare?

Escit 0.59 (0.37, 0.94) 0.69 (0.41, 1.15) 0.74 (0.44, 1.25) 0.81 (0.53, 1.24)Fluox 1.18 (0.91, 1.52) 1.27 (0.99, 1.63) 1.38 (1.10, 1.72)

Parox 1.08 (0.77, 1.51) 1.17 (0.86, 1.59)Sertr 1.09 (0.80, 1.48)

Venla

Page 65: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Extended example

Fluox

Parox Venla

Sertr

Escit

Complexity → consistency greaterconcern

Pair-wise against Fluox, Escitexcluded

Yet, evidence suggests Escit>Fluox

How do the other drugs compare?

Escit 0.59 (0.37, 0.94) 0.69 (0.41, 1.15) 0.74 (0.44, 1.25) 0.81 (0.53, 1.24)Fluox 1.18 (0.91, 1.52) 1.27 (0.99, 1.63) 1.38 (1.10, 1.72)

Parox 1.08 (0.77, 1.51) 1.17 (0.86, 1.59)Sertr 1.09 (0.80, 1.48)

Venla

Page 66: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Extended example: rank probability

Venla > Fluox, Escit > Fluox significant

Not much can be said about the rest

Which should we consider the best?

Which is the worst?

From a Bayesian perspective,

these questions are perfectly reasonable!

We can estimate the probability of Fluoxetine being

‘best’ (rank 1)‘second best’ (rank 2). . .‘worst’ (rank 5)

Page 67: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Extended example: rank probability

Venla > Fluox, Escit > Fluox significant

Not much can be said about the rest

Which should we consider the best?

Which is the worst?

From a Bayesian perspective,

these questions are perfectly reasonable!

We can estimate the probability of Fluoxetine being

‘best’ (rank 1)‘second best’ (rank 2). . .‘worst’ (rank 5)

Page 68: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Extended example: rank probability

Venla > Fluox, Escit > Fluox significant

Not much can be said about the rest

Which should we consider the best?

Which is the worst?

From a Bayesian perspective,

these questions are perfectly reasonable!

We can estimate the probability of Fluoxetine being

‘best’ (rank 1)‘second best’ (rank 2). . .‘worst’ (rank 5)

Page 69: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Extended example: rank probability

Venla > Fluox, Escit > Fluox significant

Not much can be said about the rest

Which should we consider the best?

Which is the worst?

From a Bayesian perspective,

these questions are perfectly reasonable!

We can estimate the probability of Fluoxetine being

‘best’ (rank 1)‘second best’ (rank 2). . .‘worst’ (rank 5)

Page 70: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Extended example: rank probability

Venla > Fluox, Escit > Fluox significant

Not much can be said about the rest

Which should we consider the best?

Which is the worst?

From a Bayesian perspective,

these questions are perfectly reasonable!

We can estimate the probability of Fluoxetine being

‘best’ (rank 1)‘second best’ (rank 2). . .‘worst’ (rank 5)

Page 71: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Extended example: rank probability

Page 72: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Obstacles

Before the conclusions under consistency can be accepted:

Possible inconsistency should be evaluated:

First, by assessing the studies for exchangeabilitySecond, by statistical means (inconsistency/node-split model)

Assess convergence & run-length of the MCMC simulation

Reasonable priors have to be specified

All of these are research topics

And have (preliminary) implementations in ADDISRelated manuscripts:6) G. van Valkenhoef, T. Tervonen, B. de Brock, H. Hillege, Algorithmic Parameterization of Mixed TreatmentComparisons. Manuscript under review.7) G. van Valkenhoef, B. de Brock, H. Hillege, Automating network meta-analysis. Initiated (conference paper).

Page 73: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Questions?

Page 74: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Break!

Page 75: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Introduction

ADDIS global requirements:

Database of clinical trials

Answer efficacy/safety questions

Streamline benefit-risk decision making

For regulatory authorities

Using aggregated data

Intermediate goal: quantitative benefit-risk model

Based on clinical trials or meta-analysis

Making trade-offs explicit

Making uncertainty explicit

Page 76: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Introduction

ADDIS global requirements:

Database of clinical trials

Answer efficacy/safety questions

Streamline benefit-risk decision making

For regulatory authorities

Using aggregated data

Intermediate goal: quantitative benefit-risk model

Based on clinical trials or meta-analysis

Making trade-offs explicit

Making uncertainty explicit

Page 77: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

A simple stochastic model

The ‘Lynd & O’Brien’ model:

Based on cost-effectiveness analysis techniques

Compares 2 alternatives

On 2 criteria (benefit vs. risk)

Sample (∆B,∆R) values from a joint distribution

Plot them on a plane

Count how many points are below the threshold µLynd, LD and O’Brien, BJ (2004), “Advances in risk-benefit evaluation using probabilistic simulation methods: anapplication to the prophylaxis of deep vein thrombosis.” Journal of Clinical Epidemiology 57(8):795–803.

Page 78: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Lynd & O’Brien example: set up (ADDIS)

Page 79: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Lynd & O’Brien example: set up (ADDIS)

Page 80: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Lynd & O’Brien example: set up (ADDIS)

Page 81: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Lynd & O’Brien example: data

0.0 0.2 0.4 0.6 0.8 1.0

02

46

8

probability

dens

ity

0.0 0.2 0.4 0.6 0.8 1.0

02

46

8

probability

dens

ity

Fluoxetine

0.0 0.2 0.4 0.6 0.8 1.0

02

46

8

probability

dens

ity

0.0 0.2 0.4 0.6 0.8 1.0

02

46

8

probability

dens

ity

Sertraline

−0.1 0.0 0.1 0.2 0.3

01

23

45

6

−0.2 −0.1 0.0 0.1 0.2

01

23

45

Difference

HAM-D

Dropouts

57/92 70/96

24/92 26/96

Page 82: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Lynd & O’Brien example: results (ADDIS)

Page 83: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Benefit-risk plane

+Benefit A+Benefit B

+R

isk

A+

Ris

kB

µB better

A better

p = aa+b

count b

count a

Page 84: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Benefit-risk plane

+Benefit A+Benefit B

+R

isk

A+

Ris

kB

µ

B better

A better

p = aa+b

count b

count a

Page 85: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Benefit-risk plane

+Benefit A+Benefit B

+R

isk

A+

Ris

kB

µ

B better

A better

p = aa+b

count b

count a

Page 86: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Benefit-risk plane

+Benefit A+Benefit B

+R

isk

A+

Ris

kB

Trade-off

Trade-off

µ

B better

A better

p = aa+b

count b

count a

Page 87: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Benefit-risk plane

+Benefit A+Benefit B

+R

isk

A+

Ris

kB

µThe acceptability threshold.

We are willing to ‘pay’ µ

units risk to get 1 unit of

benefit.

B better

A better

p = aa+b

count b

count a

Page 88: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Benefit-risk plane

+Benefit A+Benefit B

+R

isk

A+

Ris

kB

µThe acceptability threshold.

We are willing to ‘pay’ µ

units risk to get 1 unit of

benefit.

B better

A better

p = aa+b

count b

count a

Page 89: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Benefit-risk plane

+Benefit A+Benefit B

+R

isk

A+

Ris

kB

µB better

A better

p = aa+b

count b

count a

Page 90: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Example: acceptability curve (ADDIS)

Page 91: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

SMAA BR analysis

The Lynd & O’Brien model is limited to 2x2 problems.

Stochastic Multi-criteria Acceptability Analysis (SMAA)allows m × n problems:

m alternativesevaluated on n criteriaperformance of alternative i on criterion j : Ci,j ∼ f (ci,j )

Related manuscripts:3) G. van Valkenhoef, T. Tervonen, T. Zwinkels, B. de Brock, H. Hillege, ADDIS: a decision support system forevidence-based medicine. Manuscript under review.8) T. Tervonen, G. van Valkenhoef, E. Buskens, H. Hillege, D. Postmus, A stochastic multi-criteria model forevidence-based decision making in drug benefit-risk analysis. Manuscript under review.9) G. van Valkenhoef, T. Tervonen, J. Zhao, B. de Brock, H. Hillege, D. Postmus, Multi-criteria benefit-riskassessment using network meta-analysis. Partial manuscript.

Page 92: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

SMAA BR analysis

SMAA models for benefit-risk:

Can be based on a single trial (8)

Or (network) meta-analysis (9)

And is implemented in ADDIS (3)Related manuscripts:3) G. van Valkenhoef, T. Tervonen, T. Zwinkels, B. de Brock, H. Hillege, ADDIS: a decision support system forevidence-based medicine. Manuscript under review.8) T. Tervonen, G. van Valkenhoef, E. Buskens, H. Hillege, D. Postmus, A stochastic multi-criteria model forevidence-based decision making in drug benefit-risk analysis. Manuscript under review.9) G. van Valkenhoef, T. Tervonen, J. Zhao, B. de Brock, H. Hillege, D. Postmus, Multi-criteria benefit-riskassessment using network meta-analysis. Partial manuscript.

Page 93: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

SMAA example (ADDIS)

SMAA modelbased on networkmeta-analysis.

Page 94: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

SMAA example (ADDIS)

Measurements (input distributions).

Page 95: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

SMAA example (ADDIS)

Model without preference information.

Page 96: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

SMAA example (ADDIS)

Model without preference information.

Page 97: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

SMAA example (ADDIS)

Preferences for severe depression.

Page 98: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

SMAA example (ADDIS)

Severe depression results.

Page 99: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

SMAA example (ADDIS)

Preferences for mild depression.

Page 100: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

SMAA example (ADDIS)

Mild depression results.

Page 101: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Relevance: EMA BR methodology project

Approach/method Relevance to regulators UsefulnessProbabilistic simulation Can illuminate the risk/benefit trade-off when uncertainty is a major

feature of a regulatory decision.Medium

Bayesian statistics Can integrate evidence and its uncertainty, both pre- and post-approval, with multiple criteria in decision models.

High

MCDA Multi-criteria decision analysis extends decision theory to accommo-date multiple, conflicting objectives. Provides common units of valuefor both benefits and risks.

High

Table: MTC/SMAA integrates 2 of 3 quantitative approaches rated’High’ on usefulness, and 1 rated ’Medium’.

EMA (2010). Benefit-risk methodology project work package 2 report: Applicability of current tools and processesfor regulatory benefit-risk assessment. EMA/549682/2010.

Page 102: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Questions?

Page 103: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Escher 3.2 progress

ADDIS software:

database of trialsautomated (network) meta-analysisstochastic benefit-risk models

Research:

Survey of exisiting information systemsAutomating network meta-analysisDevelopment of benefit-risk method

Publications:

Presented at a number of conferencesSeveral papers under review (5)Journal and conference paper in preparationCase study being initiated

Page 104: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Escher 3.2 progress

ADDIS software:

database of trialsautomated (network) meta-analysisstochastic benefit-risk models

Research:

Survey of exisiting information systemsAutomating network meta-analysisDevelopment of benefit-risk method

Publications:

Presented at a number of conferencesSeveral papers under review (5)Journal and conference paper in preparationCase study being initiated

Page 105: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Escher 3.2 progress

ADDIS software:

database of trialsautomated (network) meta-analysisstochastic benefit-risk models

Research:

Survey of exisiting information systemsAutomating network meta-analysisDevelopment of benefit-risk method

Publications:

Presented at a number of conferencesSeveral papers under review (5)Journal and conference paper in preparationCase study being initiated

Page 106: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Future plans

Case studies

(Network) meta-regression

Handle covariates (dose, baseline severity, . . . )Refine the data model

Extend benefit-risk model

Hierarchical model/value treeQualitative attributes

More links with data sources, data sharing

A collaborative web portal?

Page 107: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Questions?

Page 108: Escher 3.2: towards e ective, transparent and …drugis.org/files/valkenhoef-pres-cluster2010.pdfEscher 3.2: towards e ective, transparent and accountable assessment of bene t-risk

Introduction Meta-analysis Network meta-analysis Benefit-risk analysis Discussion

Thank you!