In Britain it is a criminal offence to drive if the driver's measured ethanol concentration exceeds 35 micrograms per 100 ml in breath or 80 mg per 100 ml in blood. Those found guilty are disqualified from driving unless they can give a special reason why this should not happen.1 Most commonly, a defendant claims a friend has <> a drink by adding spirits, so that he or she has unwittingly consumed an amount of ethanol sufficient to raise the blood ethanol concentration above the limit. 2,3 Doctors are sometimes asked in court whether a defendant would have known that a drink has been laced. We conducted an experiment to discover whether volunteers were able to tell that vodka had been added to a drink. Subjects, methods, and results We recruited healthy volunteers who gave written informed consent. The west Birmingham research ethics committee approved the study. Two venues were chosen, one at the summer ball of the medical students' society (142 subjects) and the other at an ³awayday² held by Birmingham University's department of public health and epidemiology (47 subjects). Different volumes of vodka (0 ml, 12.5 ml, 25 ml, 50 ml, and 75 ml) were made up to 200 ml with orange juice, giving ethanol concentrations of 0 g, 2.0 g, 4.0 g, 7.9 g, and 11.9 g per 100 ml (% weight/volume). Different volumes of vodka (0 ml, 20 ml, 35 ml, 50 ml, and 70 ml) were made up to 200 ml with lager, giving ethanol concentrations of 2.7 g, 5.9 g, 8.2 g, 10.6 g, and 13.7 g per 100 ml (% weight/volume). We guessed that at least eight volunteers in each group were required to give acceptable statistical power. A table of random numbers was used to allocate one of the five concentrations of drink to sequentially numbered beakers. Volunteers took a beaker,tasted the mixture (or drank it entirely), and answered a supervised questionnaire recording their sex and age, the number of units they usually drank, smoking habit, beaker number, and their response, ³yes² or ³no,² to the question ³is this drink pure?² ... A total of 142 subjects attended the medical school ball, of whom 108 (76%) took part in the study, together with six of the researchers. At the awayday, 39 of 47 (83%) subjects contributed, together with two domestic staff. ... Comment ... 1 Ley NJ. Drink driving law and practice. London: Sweet and Maxwell, 1993. 2 Jones AW. Top ten defence challenges among drinking drivers in Sweden. Med Sci Law 1991;31:229-38. 3 Denny RC. The use of breath and blood alcohol values in evaluating ³hip flask² defences. N LawJ1986;26 September:923-4. 4 Armitage P, Berry G. Statistical methods in medical research. Oxford: Blackwell Scientific Publications, 1971. 5 Marlatt GA, Demming B, Reid JB. Loss of control drinking in alcoholics: an experimental analogue. J Abnorm Psychol 1973;81:233-41. ============================================================ To anyone using these data... PLEASE RESPECT THE AUTHORS" WISHES RE USE OF THESE DATA. DO NOT RE_DISTRIBUTE J Hanley ============================================================ Date: Thu, 13 Jan 2000 15:52:35 +0000 (GMT) From: Tim Marshall Subject: Re: (Fwd) drinking To: James Hanley Reply-to: T.Marshall@bham.ac.uk Organization: The University of Birmingham Priority: normal A description of the data. Col. 1: Number in study. Col. 2: Sex (M = 1, F = 0). Col. 3: ?regular drinker, >= 14 units/wk. (Y = 1, N = 0). Col. 4: Age (years). Col. 5: ?regular smoker (Y = 1, N = 0). Col. 6: was the drink "pure" (orange juice or lager)? (Y = 1, N = 0). Col. 7: which drink? (Lager = 1, orange juice = 0). Col. 8: a coded concentration: 0,1,2,3,4, according to the order of concentration in the paper. 0 = no added alcohol to either drink, but 4 is a different concentration for lager than for orange juice. Within drink, the numbers reflect an ordinal scale. Col. 9: actual relative concentration, specific to each drink (e.g the %-ages for OJ have been re-scaled to 0,1,2,4,6, and for lager to 0,4,7,10,14, to reflect the relative magnitudes of the added alcohol. But the scales are specific to each drink - you'd have to multiply the OJ scale by some factor to express it on the same scale as the lager. Or just convert all back to % of drink volume (as explained in the paper). Col. 10: which event? (Medsoc ball = 0, Public Health Awayday = 1). Col. 11 (spiked): the complement of Col 6 (to get the prob. that the drink has been spiked). Col. 12: a filter for when I looked only at the OJ data. Col 13: the index used for the paper, added alcohol expressed as % weight by volume (whatever that means). I'd be interested to know what your students make of it.... Best wishes, Tim Marshall ===================== Date: Mon, 10 Jan 2000 09:32:44 +0000 (GMT) From: Tim Marshall Subject: (Fwd) drinking To: JimH@epid.lan.mcgill.ca Reply-to: T.Marshall@bham.ac.uk Organization: The University of Birmingham Priority: normal I have no problem with letting you have a copy of the data, but I'm just waiting to hear from Nigel Langford, the registrar, that he is OK with this. There are several ways in which the alcohol concentration can be expressed, from "added alcohol, % weight by volume" to a sequential indicator variable set, or a simple ordinal "score" or the % concentration of added alcohol within the 200ml aliquots we used. All give much the same result. I think the first of these was used in the paper because of its relevance to English law on drinking and driving. I don't mind students having access to the data, but would prefer ******************************************************************* they couldn't copy it and take it away with them. ************************************************* As Robin Ferner said, please acknowledge ***************************************** Best wishes Tim Marshall ======================================================================= From: "r.e.ferner" Subject: Re: data in your bmj article To: "James A. Hanley" Cc: t.marshall@bham.ac.uk, nigel langford MIME-version: 1.0 Priority: normal I'm perfectly happy, but will pass this on to our statistician and our co-author. If and when you do use the data, please acknowledge us. Merry Christmas Robin Ferner R.E.FERNER CITY HOSPITAL BIRMINGHAM B18 7QH ENGLAND TEL: ++ 44 121 554 3801 or ++ 44 121 507 4499 FAX: ++ 44 121 523 6125 ========================================================================== > Date: Wed, 22 Dec 1999 22:15:22 -0500 > To: r.e.ferner@bham.ac.uk > From: "James A. Hanley" > Subject: data in your bmj article > Cc: JimH@Epid.Lan.McGill.CA > Dear Dr Ferner > > What a delightful study in the X-mas edition !! > > I am a biostatistician. I teach logistic regression each year in our summer > school in epidemiology and biostatistics. Up to now I have been using > mortality in a toxicology study of cadmium in rats to show the ... > behind ... > > It struck me that the data in your article would be much more interesting > for the students to practice on! > > For regular linear regression I do use a small dataset (also supplied by > another UK author) on drinking in relation to smooth pursuit velocity. The > issue is again around what happens at the legal limit , but that study > had only 12 subjects -- who were told to drink until they felt like > stopping. > > If you were to share your raw data (including covariates) with me, I would > use them only for teaching purposes. I would keep them to myself if you > did not want students having access to them, although of course I would > prefer that they could. > > If you are interested, you can look through my web page for the course > > http://www.epi.mcgill.ca/~web2/hanley/c678/ > the smooth pursuit data are in the second line under "datasets" > > (the cadmium dataset is in the material for course c626 rather than c622 > (same access) > > my web page is http://www.epi.mcgill.ca/~web2/hanley/ > > If you are able to share the data, I would be happy to receive them in > whatever form is easiest for you to prepare! > > Thank you in advance! > > James A Hanley PhD > Department of Epidemiology and Biostatistics > McGill University