DRIVING UNDER THE INFLUENCE

End-of-year festivities coincide with media campaigns to persuade
people not to drink alcohol and drive, despite the fact that the
number of casualties after drink-related accidents peaks in the
summer. Feelings run high when driving while under the influence is
discussed. The US lobby group, Mothers Against Drunk Driving, heads
its faxes with "Drunk driving is not an accident. It is a violent
crime". Although the "personal freedom" argument--the individual
should be free to drink and drive without interference from the
law--is rarely heard these days, voices against further restrictions
are raised. In the UK, owners of public houses (licensed bars) in
rural communities say that any lowering of the limit would "sound the
death knell" of their businesses, threatening the cohesion of the
local community. The UK government wants to reduce the legal
blood-ethanol concentration for driving from 80 to 50 mg/dL, which
would bring the UK into line with most other EU countries--and would
save an estimated 50 lives a year.

Reports in the press which suggest that the UK government is having
second thoughts are misleading. The Guardian reported on its front
page last week that the government, acting on the advice of police,
had decided not to reduce the limit to 50 mg/dL. The story was wrong.
The UK transport minister is still considering a reduction, and the
Guardian published, the following day, a letter on behalf of chief
constables to say that the police do want a lower limit.

Driving while under the influence does lead to deaths and serious
injuries to drivers, passengers, and other road users. In 1997, in the
USA, just over 16 000 people were killed in crashes involving alcohol,
nearly two-fifths of all traffic deaths. Over a million people were
injured in drink-related accidents. The cost of drink-related
accidents has been estimated at US$45 billion a year, with US$70
billion lost in quality-of-life costs. 90 000 to 129 000 lives have
been saved between 1983 and 1996 due to the fall in alcohol
involvement in fatal accidents in the USA.

In the UK, last year, there were 540 deaths and over 16 000 injuries
(nearly 3000 serious) in road accidents involving illegal alcohol
concentrations, with about half the victims being people other than
the driver. The dramatic fall in the number of crashes involving
illegal alcohol levels since 1986 flattened out in 1993 and remains
flat--hence the UK government's wish to "do something".

Several interventions have been proposed other than reducing the
blood-alcohol limit. These include education or rehabilitation courses
for offenders, increasing police power to breathalyse a suspect,
targeting high-risk or repeat offenders, and even ignition locks with
a breath-alcohol detector.

When experimenting with public policy, the effectiveness of any
intervention must be assessed as carefully as in the limited
population that is exposed to a clinical trial--after all, if the
intervention is of little benefit, its adverse influences (cost,
police-time wasted, loss of the social enjoyment of drinking) become
more important. Drinking alcohol carries risks and benefits (see
Commentary on p 1873 about cardioprotection). Whatever new
drink-driving scheme is tried, and in the UK a reduction in the legal
limit seems the likely next step, how the effect will be audited needs
stating clearly and prospectively, as in a clinical trial protocol.
For instance, the data will be skewed if the police confound the
situation by devoting more resources to breathalysing drivers if the
limit is decreased or by targeting high-risk groups at the same time.
Cars are becoming safer in crashes, which can also confound
year-on-year data. Data analysts already use weightings to allow for
factors such as under-reporting or exposures (eg, miles driven).

More thought needs to be given to prospective designs that can show
effectiveness of a policy change rather than the somewhat blunt death
and injury statistics used now. Designs such as cluster randomisation
(regions are paired in clusters, and the new intervention is tried in
one pair while the other stays as usual) and stepped-wedge techniques
(phased introduction of the intervention by cluster until the whole
population is covered) might prove useful statistically, and, if
carefully explained, might even persuade the public that the
intervention is worthwhile.

Checked-by: derek rea