MCDA (Multi-Criteria Decision Analysis)

Recommended for further appraisal for use in benefit-risk assessment

1. Description

MCDA quantifies the overall performance of two or more alternatives. As applied to the benefit-risk balance of a drug and its comparators, performance of the alternatives on the favourable and unfavourable effects are judged for their clinical relevance and all effects are weighted to create a common unit of preference value, or utility. Summing those common units of benefit and risk provides an overall benefit-risk preference value or utility for each alternative, enabling calculation of the difference of the drug against the comparators.

2. Evaluation

2.1 Principle

MCDA is based on decision theory. The framework for MCDA follows PrOACT-URL that ensures transparency. It is important to realise that certain conditions should be met if the simple calculations applied in MCDA are to give valid results. The criteria should be:

• Requisite in number-a complete set showing how the options differ in ways that matter to the decision maker(s), without double-counting.

• Understandable as preferences-an unambiguous definition for each criterion, which indicates a clear direction of preference (more is preferable to less, or less to more).

• Mutually preference independent-so that scores can be assigned on each criterion without having to know the scores on any of the other criteria.

• Accommodating of preferences over time-fully covering the time horizon of the model (possibly a mixture of short-term and long-term criteria).

A key question for applying MCDA to the benefit-risk assessment of drugs is who does the scoring and weighting. Measurable data are usually available, but these must then be translated into preference scores through the use of value functions. These are often linear, direct or inverse, but may be non-linear.

2.2 Features

• MCDA places criteria on the same unit/scale that allows them to be compared directly and to be combined.

• Weighs criteria appropriately.

• Different sources of evidence and time dimension can be incorporated as another criterion in MCDA.

• Compares multiple options simultaneously.

2.3 Visualisation

• Value tree diagram

• Stacked bar graph

• Difference display

• Frontier graph (line area graph)

2.4 Assessability and accessibility

• Knowledge of decision analysis is required.

• Ranking of the final results make interpretation and communication easier.

• Gathering the most relevant experts and stakeholders might be a challenge, but following that building an MCDA model would be straightforward.

• Modelling focusses on incorporating data and clinical judgements, and exploring sensitivity of the results to imprecision in the data, uncertainty and differences of opinions.

• There are specialised software to facilitate the implementation of MCDA, e.g.:

  Hiview 3 (http://www.catalyze.co.uk)

  V.I.S.A (http://www.visadecisions.com/)

  Intelligent Decision System(http://www.e-ids.co.uk)

  Logical Decisions(http://www.logicaldecisions.com)

MCDA was tested in the Efalizumab, Natalizumab, Rimonabant, Rosiglitazone and Telithromycin case studies.

3. References

[1] Mussen F, Salek S, Walker S. Benefit-Risk Appraisal of Medicines. John Wiley & Sons, Ltd.; 2009.
[2] Dodgson J, Spackman M, Pearman A et al. Multi-Criteria Analysis: A Manual. 2000. London, Department of the Environment, Transport and the Regions.
[3] Keeney RL, Raiffa H. Decisions with Multiple Objectives: Preference and Value Tradeoffs. New York: John Wiley; 1976.