CHAPTER 2
1. EVOLUTION OF A GRAPH
Natural Selection is the process which results in the survival and reproduction of only those forms of life which have characteristics appropriate to a particular environment. Such selected creatures thrive and may become common compared to their less-well adapted cousins. In this way the world of living creatures moulds itself to match the environment. This is well known and accepted but it is less well known that facts also change so that they can take their place in a cruel world! Consider how a decades old graph, which first appeared in a scientific journal, has finally ended up now that it is in the public eye.
Figure 2.1 summarises the results of perhaps the most respected of all studies on the effects of drinking on driving ability.
The basic message from Figure 2.1 is clear enough -- with increasing blood alcohol concentration (BAC) there is a rapidly increasing probability of an accident.
Important though this primitive (Borkenstein et al., 1964: 101) specimen is, it is by no means well adapted to our society and has scarcely ever reproduced. Its structural flaw lies in its bottom left corner which implies that a few drinks should improve driving ability. For this reason it is to be found in only a few specialised collections of books and never in public.
Some years later (Birrell, 1974: 50), nurtured by the Victorian Police Surgeon, an improved version was found to have appeared and reproduced itself moderately widely in Sun Books (see Figure 2.2 above). It is obvious that the anatomical disability of its predecessor is much modified for the better, but the crucial anatomy on its right side is virtually unchanged. However, even the flattening in the 0.00 to 0.04 region is a serious obstacle to its reproduction. (This graph appeared unacknowledged but its essentially point-for-point correspondence with Fig. 2.1 leaves little doubt about its origin.)
In 1982, at the Traffic Accident Research Unit appeared the first example of the fully adapted form (Herbert, 1982: 8; reproduced here as Figure 2.3). Notice for the first time the uninterrupted upwards sweep of the curve. Curiously, the 0.00 to 0.02 segment is absent, doubtless because of the unimaginable rigours of the evolutionary process (see Alfred Lord Tennyson, "In Memoriam" for a graphic account of the horrors of "Nature red in tooth and claw"). This segment, as we have seen, is of importance in reproduction, and as a result Figure 2.3 also is rarely encountered. (Herbert, 1982, describes this figure as "based on a graph published by Birrell, 1974" and incorporating data from other studies. The resemblance to Borkenstein's Figure is once again striking, and the absence of low BAC values most curious.)
In Figures 2.4 and 2.5 (below) we see two examples of the most highly evolved form yet found. These were collected at the Office of Road Safety Canberra (Anon, 1984) and The Road Patrol (Anon, 1986). The missing segment has reappeared. The primitive right hand side of the curve, present as we have seen in the most primitive known specimen, continues as a reminder to us that in any evolving organism vital and optimally produced features are always preserved. (Both are described "as adapted by TARU (1982) from J.L. Birrell (1974)".)
It may be considered improper to treat lightly a subject so seriously regarded as drinking and driving. Indeed, the subject is so important that one must ask why Borkenstein's original graph should have been copied virtually point-for-pint, except for low BAC values, in successive transformations. If Borkenstein's data are good enough for high BAC values, why not for low BAC values, particularly since his low BAC results have, as we shall see, aroused such lively interest in the professional literature?
Oddly enough the explanation for the dip in Fig. 2.1 is not that a few drinks improve driving ability. Quite the contrary. However, what appears to be the correct explanation is one that those charged with the duty of regulating and restraining our lives are apparently unwilling to put to the public. To understand it, it is necessary to know a little more about the work which produced Figure 2.1.
The Grand Rapids Dip
R.F. Borkenstein, the inventor of the Breathalyser now widely used around the world to monitor blood alcohol in drivers, was the first scientist to demonstrate this curious phenomenon: although the likelihood of having an accident increases overall with increasing blood alcohol concentration (BAC), between BAC levels of 0.01 and 0.04 there appears to be a reduction in the likelihood of having an accident -- as shown in Figure 2.1. This odd result, known as the "Grand Rapids dip" or "notch" after the town where Borkenstein did his work, has been much discussed and its origin sought. The simplest explanation -- that a few drinks improve your driving -- turns out to be wrong.
Borkenstein measured many characteristics of drivers (both these who were and those who were not involved in accidents) in an effort to determine what it is that predisposes a driver to be accident-prone. One of these was "self-reported drinking frequency". When Borkenstein's data is graphed (Hurst, 1973) to include this variable an even odder result than the "Grand Rapids dip" appears (Figure 2.6).
At all blood alcohol levels up to 0.09, people who say that they drink daily are far safer drivers than those who say they drink "yearly or less". The reason for this is unknown, but the most likely explanation is probably that "regular drinkers" includes a large number of middle-aged, experienced drivers. (See Hurst (1973), and Walls and Brownlie (1985) for more information.) Intermediate drinkers -- those who drink 3 times a week, weekly, and monthly -- fall between the two extremes. It now becomes possible to explain the "Grand Rapids dip". If all drivers on the roads could be breathalysed at the same moment, it would be expected that the group with a BAC of 0.02 per cent would consist mostly of frequent drinkers: it would be unusual to find many yearly drinkers just after their yearly drink. In effect, the driving population includes a group of safer-than-average drivers who have had a few drinks.
It should be noted (as Fig. 2.6 shows) that all drivers, whether teetotaller or tippler, become worse drivers as they consume more alcohol. (5) But, on average, regular drinkers can exceed the blood alcohol limit of most countries yet remain better than the average driver.
Given these results, it would be more logical to use the Breathalyser to remove the infrequent drinkers from the road. Random breath testing has the opposite effect: it removes some, indeed a significant proportion, of the safest drivers on the road. Anyone who has seen a drunk attempting to start a car, let alone drive it, is well aware of the danger that can lie in the combination of drinking and driving. Such bad drivers should be stopped. It is the purpose of this account, however, not to produce a safe traffic policy but to disclose elements in the drink-driving story that are virtually unknown to the public. The "Grand Rapids dip" and its disturbing explanation are not the only example.
Teetotal Alcoholics
The Nobel Prizewinner, Sir Hans Krebs, some years ago demonstrated a most surprising result: the animal body generates its own alcohol and disposes of it (Krebs and Perkins, 1970). However, this sequence of events does not always happen. Some people combine a propensity to produce excessive amounts of alcohol from their food with a poor disposal system (Kaji et al., 1984). Such people can fall over dead-drunk simply from eating breakfast. Others are merely rather intoxicated. Naturally enough they never suspect the cause of their condition. Such people, who form an unknown, but probably very small, proportion of the Australian population, nevertheless pose a serious problem for anyone interested in the ethics of drink-driving laws.
Cadavers and Alcohol
There is evidence to show that the bodies of those killed in car accidents are likely to contain alcohol and this has been used as an argument to support drink-driving laws. The full story is not so simple. All cadavers generate alcohol, and the greater the delay between death and the taking of a blood sample, the greater the BAC (Nanikawa and Kotoku, 1971). For example, seven days post mortem in a room at 15°C, the average cardiac BAC of 6 rabbits was 0.4% (range 0.11-0.88). (6) This fact is not generally known. It makes interpretation of alcohol levels in dead accident victims most difficult and raises doubts about statistics based on such evidence. As the generation of alcohol in cadavers tends to increase with the passage of time and is greater at higher temperatures, this phenomenon would seem to have particular application to accidents in non-metropolitan areas where relative isolation and high temperatures often go hand-in-hand.
The Road Traffic Authority and Causes of Accidents
According to the Western Australian Road Traffic Authority Research and Statistics Division Report No. 7 (Kirkham, 1978) "intoxicated drivers and riders" were responsible for only 3.4 per cent of all accidents and 8.0 per cent of fatal accidents. 92 per cent of fatal accidents involved sober drivers. This is only one of a number of studies of the causes of accidents. Its interest lies firstly in its status as a publication of the traffic authority and secondly in its obscurity. Its obscurity is hardly surprising: until recently it was not available for public viewing.
2. QUIT!
This "evolution of facts" is by no means unique to the drink-driving story. Public health is littered with similar cases. For example, the Health Department of Western Australia has for some time funded a campaign, "Quit", intended to encourage people to give up smoking. It has done this in spite of the fact that the only controlled trial ever conducted showed that simply giving up smoking is certainly not beneficial. It showed a significant increase in deaths from cancer other than lung cancer in those who gave up smoking, with no change in deaths from lung cancer, other causes or all causes (Rose et al., 1982). But that is another story. Amongst other initiatives "Quit" has published full-page advertisements in local newspapers (for example, Subiaco Post, 28 June 1988: 12). In the text is the statement: "If present trends continue, lung cancer will soon overtake breast cancer as the most common malignant cancer in women." Yet the same Health Department in a statement of May 1987 by W.M. Hatton and M.D. Clarke-Hundley says:
Both incidence and mortality in both sexes have shown some indication of levelling off or at least provide no evidence of a sustained rise between 1982 and 1985.
And
Suggestions by some commentators that lung cancer deaths in women will overtake breast cancer deaths in the next few years look increasingly unlikely. Age standardised mortality rates for breast cancer have fluctuated between 19 and 21 per 100,000 for the last 50 years while female lung cancer death rates have fallen for the last 2 years.
These are contradictory statements, each with the imprimatur of the Health Department, one intended for the professionals, the other for the public. Hatton (1987) in a separate publication from the Health Department actually calculated the number of expected breast and lung cancers in women for the years 1988, 1991, 1996 and 2001. For every year he predicted more than 3 times as many new cases of breast cancer. For example, he predicted that in 2001, these would be 698 new cases of breast cancer and 193 lung cancer.
A related example comes from a recent survey of cancer trends in Australia (Armstrong, 1988). Again a statement is made about the trend in deaths from breast and lung cancer in women:
Mortality from breast cancer has remained remarkably stable and, if mortality from lung cancer in women continues to rise, it could ultimately overtake breast cancer -- the mortality from which is comparatively stable -- as the commonest cause of death from cancer in them, which it has done already in some North American women.
However, the reference (Stolley, 1983) quoted as authority for this statement states "An estimated 34,000 American women will die this year of lung cancer and 37,200 of breast cancer; in 1978 the comparable figures were 24,000 and 34,000 respectively". Anti-smoking as Stolley's article might be, nowhere does he claim that lung cancer has overtaken breast cancer as the commonest cause of death from cancer in "some North American women" (whatever that might mean) or indeed in any group of women.
The well-known advice of Dr Routh to a young scholar "You will find it a very good practice always to verify your references, sir!" is nowhere so needed as in public health and epidemiology.
3. MORE GRAPHS
A final example comes from two editions of the report on smoking and health produced by the Royal College of Physicians of London. In the edition of 1971 (Smoking and Health Now, 1971), Figure 4.1 showed the change in lung cancer death rates in 45-64 year old men over the period of this century, supposedly due to an increase in tobacco consumption (Figure 2.7 upper, on the following page). However, it has two features so undesirable it may as well have been produced by a saboteur from the tobacco industry. (7) First, it shows no increase in deaths from bronchitis. Indeed there is a decline, which is most unbecoming for a graph intended to demonstrate the deleterious effects of an increase in smoking during the period 1916-45. Second, there is concomitant with the increase in lung cancer deaths a monotonic decrease in tuberculosis deaths consistent from 1916. If the relentless increase in lung cancer this century has been caused by a concomitant increase in smoking, is it not odd that two other respiratory disorders should have declined? Although not inimical to the smoking-lung cancer idea, the secular changes for TB and bronchitis by no means harmonise well with it. (8) These features apparently were offensive to eye and mind because when Figure 4.1 next made an appearance in the 1977 edition (Smoking or Health, 1977) they had both vanished (Figure 2.7, lower).
In no area of science is it more important to scrutinise the published evidence as it is in the area of epidemiology and public health.
ENDNOTES
5. This does not eliminate the possibility that particular individuals may be improved by a drink or two. Alcohol is known to reduce hand tremor. Indeed, at the Olympic Games in 1968, two pistol shooters were allegedly disqualified for sedating themselves with alcohol (Landauer, 1981).
6. Comparable human measures are difficult to obtain because of the various objections to keeping corpses in warm places.
7. Alternatively, the Royal College compilers may have merely reflected a view they had put forward in 1628 when they had recommended the public should smoke imported Virginian tobacco, the home-grown product "falling short of the perfection of other tobaccos that are brought from more southern parts where it hath his natural maturity, vigour and efficiency" [quoted in Keynes, 1978).
8. The decline in TB is not simply explained. See Berkson (1960) and Taylor (1979).
No comments:
Post a Comment