Intended for healthcare professionals


Effect of seeing tobacco use in films on trying smoking among adolescents: cross sectional study

BMJ 2001; 323 doi: (Published 15 December 2001) Cite this as: BMJ 2001;323:1394

Smokier films or simply more films?

Sargent et al demonstrate that in a large series of adolescents,
those who have ever tried smoking have a higher level of exposure to
watching smoking in films than those who have not. The smoking outcome
studied is not restricted to instances of starting smoking for the first
time. An unusual feature of this context is that most likely practically
everyone in the series has some exposure to films containing smoking, so
what is shown here should be regarded as dose-response, not a comparison
of the exposed and unexposed.

Both the authors and responders (Richard Hutchinson and Mike Males)
discuss confounding as a major issue in interpreting an observational
study such as this. Among many confounders considered by the authors, age
plays a crucial role: as an individual ages, both the total number of
films ever watched can only increase, and also the only possible smoking
transition is from "never" to "ever". In recognition of this, figure 3
shows how the relationship is demonstrable within age bands also.

However, in this instance, there is a related issue, which is not
confounding as such, hence more covert. The exposure measure has been
developed as a proxy for total exposure to viewing smoking on films and is
probably valid for this purpose. But it is a composite measure.
Adolescents in the study varied widely in how many of the 50 designated
films they had seen (figure 2), as well as in the kinds of films they
chose. The authors state that the latter factor is unlikely to explain
the findings. Hence, might it not be the case that the study is largely
picking up some sort of association of smoking with total amount of
viewing? Both factors may be inversely related to physical activity,
which does not appear to have been considered. Surely it would be helpful
to split up the exposure measure and to examine how smoking and non-
smoking adolescents differed both in the "smokiness" of the average film
they watched and also how many of the films they had seen, the latter as a
proxy for overall film-watching activity (if that is not an oxymoron).

My purpose in pointing this out is not in any way to seek to disparage the
hypothesis that the portrayal of smoking in films may be harmful, merely
to reiterate the principle that our prior belief that a hypothesis is
correct does not obviate the need to consider carefully alternative
explanations of what is observed. Indeed, many other practices
deleterious to health and wellbeing feature strongly in films, and given
the choice, adolescents tend to select such films. The issue of
appropriate film content extends far beyond the depiction of smoking.

Competing interests: No competing interests

20 December 2001
Robert G Newcombe
Senior Lecturer in Medical Statistics
University of Wales College of Medicine, Heath Park, Cardiff CF14 4XN