RESPONSE TO FILLMORE & CHIKRITZHS (2010)
Article first published online: 10 MAR 2010
© 2010 The Authors. Journal compilation © 2010 Society for the Study of Addiction
Volume 105, Issue 4, pages 647–648, April 2010
How to Cite
BRITTON, A. and SHIPLEY, M. (2010), RESPONSE TO FILLMORE & CHIKRITZHS (2010). Addiction, 105: 647–648. doi: 10.1111/j.1360-0443.2010.02938.x
- Issue published online: 10 MAR 2010
- Article first published online: 10 MAR 2010
Our first response to Fillmore and Chikritzhs  is to re-emphasise that the aim in our paper  was to explore whether the relationship between alcohol consumption and coronary heart disease (CHD) and mortality is different if exposure to alcohol is measured in a variety of ways. We compared using a baseline measure only, an average intake during follow-up, and a measure of change in drinking during follow-up. We did not set out to look at the reasons for changes in consumption. These upstream determinants are likely to include financial aspects, life events as well as health problems.
Indeed, we share their belief that people may reduce alcohol consumption concomitant with detection of health problems. We therefore excluded participants with prevalent CHD at baseline in recognition that they may have already cut back drinking alcohol. We did take account of onset of illness/medication as that is partly what defines an incident CHD event. Furthermore, in post-hoc analyses, detailed in the paper, we discovered that those who declined their intake during the follow-up period were at higher risk of death. This supports the ‘sick-quitter’ hypothesis.
Contrary to the commentators' belief that alcohol consumption in this cohort is declining over time, our data show an upward trend in alcohol consumption across the phases of data collection. Using Table 1 in our paper, we see that overall, 31% of participants move to a higher alcohol category at the next phase in contrast to 24% who move to a lower alcohol category. This upward trend is also seen for all three alcohol variability groups (low, medium and high) in our study. For example, the proportions of participants in the high alcohol variability group defined in our paper as having a downward trend, no trend or increasing trend are 27%, 28% and 46% respectively. Hence, the group of participants who we show to be at increased risk of death represents approximately one quarter of those with high alcohol variability.
The commentators quite rightly recognise, as we do in the paper, that cohort studies are at risk of selective attrition. We used mortality as one of our key endpoints as 99.9% of the baseline cohort is flagged for fatal events using their unique NHS number. Therefore loss-to-follow-up for fatal events is minimal. The Whitehall II cohort is in the process of being linked with the England and Wales Hospital Episodes Statistics database so that in future we will be able to detect non-fatal events among participants even if they no longer respond to screening phases.
In order to explore whether the chronic health consequences of alcohol consumption are affected by varying intake over the life course requires large, longitudinal datasets with repeated exposure and outcome measurements on the same individuals. Such studies are rare. We wholly agree that collaborative research is needed in which data from cohort studies with multiple measurement points are combined to ensure sufficient statistical power and we welcome such endeavours.