Article Text

Download PDFPDF

The preference effect in an unblinded health promotion intervention trial: how important is it?
  1. C. Hayes1,
  2. C. Collins2,
  3. M. Geary3,
  4. C. C. Kelleher4
  1. 1
    Department of Public Health, Dr Steevens’ Hospital, Dublin 8, Ireland
  2. 2
    Irish College of General Practitioners, Dublin, Ireland
  3. 3
    Rotunda Hospital, Dublin, Ireland
  4. 4
    School of Public Health and Population Science, University College Dublin, Dublin, Ireland

    Statistics from

    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.


    Maternal smoking is one of the few potentially preventable factors associated with miscarriage, complications of pregnancy, low birth weight, perinatal death and poorer long term growth, development and health of the child. Health promotion initiatives pose a particular challenge especially for disadvantaged women. The preference effect, whereby people support an intervention because it conforms with their own understanding of a behaviour or disease under study, is one such determinant and is especially relevant to unblinded trials.


    To determine whether assignment to intervention or control group and smoking status affected participation in an unblinded intervention study, one of the aims of which was to identify the key determinants of sustained effective brief intervention for smoking cessation in pregnancy and post partum.


    The sample comprised a cohort of 1000 disadvantaged pregnant smokers who were followed up at five time points—twice during pregnancy at the hospital, once directly after birth and at two subsequent public health nurse visits (3–4 and 7–9 months post partum).


    The methodology comprised a quasi-experimental historical cohort design as blinding was not feasible with a non-intervention group (500) recruited first. The intervention group was recruited following a washout period. Interviewer administered questionnaires were completed at each time point and reported quitting validated by urinary cotinine measurement once in pregnancy (visit 2) and once after (visit 4).


    The overall proportion of responder and non-responder women who reported being current smokers at the first ante-natal visit was similar, however differences emerged within the groups with a higher proportion of the intervention group who were current smokers (56.8%), being non-responders at visit 5, compared to the reference group (41.8%) (p<0.05). Spontaneous quitters at the booking visit were also less likely to stay with the programme (61.4% vs. 38.6%, p<0.5), with no difference in effect for intervention and reference groups (p = 0.8). Fall off in response was an important element of the programme with 46.8% of the intervention group taking part at visit 5 compared to 53.2% in the reference group (p = 0.05).


    These findings indicate that smoking status and intervention arm each affected participation, which illustrates the importance of estimating such a preference effect in assessing an effective health promotion programme.