Causality assessment of adverse drug reactions: comparison of the results obtained from published decisional algorithms and from the evaluations of an expert panel

Authors

  • Ana Filipa Macedo,

    Corresponding author
    1. Núcleo de Farmacovigilância do Centro, Faculdade de Medicina, Faculdade de Farmácia, Universidade de Coimbra, Administração Regional de Saúde do Centro, Portugal
    • Faculdade de Ciências da Saúde. Universidade da Beira Interior, 6201-001 COVILHÃ. Portugal.
    Search for more papers by this author
  • Francisco Batel Marques,

    1. Núcleo de Farmacovigilância do Centro, Faculdade de Medicina, Faculdade de Farmácia, Universidade de Coimbra, Administração Regional de Saúde do Centro, Portugal
    Search for more papers by this author
  • Carlos Fontes Ribeiro,

    1. Núcleo de Farmacovigilância do Centro, Faculdade de Medicina, Faculdade de Farmácia, Universidade de Coimbra, Administração Regional de Saúde do Centro, Portugal
    Search for more papers by this author
  • Frederico Teixeira

    1. Núcleo de Farmacovigilância do Centro, Faculdade de Medicina, Faculdade de Farmácia, Universidade de Coimbra, Administração Regional de Saúde do Centro, Portugal
    Search for more papers by this author

  • Ana Filipa Macedo is PhD student supported by a research grant from “Fundação para a Ciência e a Tecnologia—programa Praxys BD/758/2000—Portugal”.

Abstract

Purpose

To compare the results of causality assessments of reported adverse drug reactions (ADR's) obtained from decisional algorithms with those obtained from an expert panel using the WHO global introspection method (GI) and to further evaluate the influence of confounding variables on algorithms ability in assessing causality.

Method

Two hundred sequentially reported ADR's were included in this study. An independent researcher used algorithms, while an expert panel assessed the same reports using the GI, both aimed at evaluating causality. Reports were divided into three groups according to the presence, absence or lack of information on confounding variables.

Results

For the total sample, observed agreements between decisional algorithms compared with GI varied from 21% to 56%, average of 47%. When confounding variables were taken into account, agreements varied between 41% and 69%, average of 58%; 8% and 65%, average of 46% and 15% and 53%, average of 42% accordingly to the absence, lack of information or presence of confounding variables, respectively. The extend of reproducibility beyond chance was low for the total sample (average Kappa = 0.26) and within the groups considered.

Conclusion

The overall observed agreement between algorithm and GI was moderate although poorly different from chance, confounding variables being a shortcoming of algorithms ability in assessing causality. Copyright © 2005 John Wiley & Sons, Ltd.

Ancillary