You are here

Article Text

Article menu

Download PDFPDF

Commentary
The legitimacy of modelling the impact of an intervention based on important intermediate outcomes in a trial
  1. Simon Cousens1,
  2. Roy Head2
  1. 1 Centre for Maternal Adolescent Reproductive and Child Health (MARCH), London School of Hygiene and Tropical Medicine, London, UK
  2. 2 Development Media International, London, UK
  1. Correspondence to Roy Head; roy{at}developmentmedia.net

https://doi.org/10.1136/bmjgh-2018-001222

Statistics from Altmetric.com

    Request Permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

    Summary box

    • A cluster randomised controlled trial in Burkina Faso evaluating the impact of a mass media behaviour change campaign showed a highly significant impact on child consultations for malaria (56% increase in 1st year), diarrhoea (73%) and pneumonia (39%), but did not detect an effect on child mortality.

    • The trial was not, and could not be, adequately powered to detect a reduction in child mortality of 10% or less, so a well-established modelling tool was used to estimate the effect of these intermediate outcomes on lives saved.

    • Modelling was considered justified in the context of strong improvement in these important intermediate outcomes and good availability of the drugs needed to treat malaria, diarrhoea and pneumonia.

    • Demand-side interventions are not a substitute for health system strengthening or other supply-side interventions, but treatment seeking is a necessary precondition for any child to receive life-saving treatment.

    Colbourn et al 1 raise three interesting questions. (1) Whether, if a randomised controlled trial (RCT) fails to detect an effect on the primary outcome but shows a significant effect on an intermediate outcome, it is legitimate to model the effect? (2) Whether the modelling was performed correctly? and (3) Whether it is meaningful to focus on demand-side factors in isolation?

    (1) We would argue that it is legitimate to use a model to evaluate an effect that the RCT was not, and could not be, powered to detect, where there is strong evidence of an impact on an important intermediate outcome, provided the limitations of such an exercise are appropriately acknowledged.

    Measuring the effect of media campaigns is challenging. Very few countries have the very low national radio and television penetration and …

    View Full Text

    Linked Articles