94
Special
Articles
phlebitis, thrombophlebitis, thrombosis, or embolism in any vein (except the cerebral, coronary, hepatic, and mesenteric veins) or as pulmonary embolism or infarction, are coded by the Registrar General to list nos. 463-466 in the International Classification of Diseases (LC.D.)6; special list nos. (648, 682, 684) are used when such deaths
VENOUS THROMBOHMBOLIC DISEASE AND THE USE OF ORAL CONTRACEPTIVES A Review of
Mortality Statistics in England and Wales
known to have been pregnant or in the the time of death. (These list numbers have been altered in the latest (eighth) revision of the I.C.D.,7 but this does not affect the data considered in this
occur
M. P. VESSEY M.B. Lond. MEMBER
OF
SCIENTIFIC STAFF, MEDICAL RESEARCH COUNCIL’S STATISTICAL RESEARCH UNIT, UNIVERSITY COLLEGE HOSPITAL MEDICAL SCHOOL, LONDON W.C.1
M.B., B.Sc. Edin. STATISTICIAN, GENERAL REGISTER OFFICE, LONDON W.C.2
The mortality statistics for venous thromboembolic disease (International Classification of Diseases list nos. 463-466) in England and Wales for 1953-67 have been reviewed in relation to estimates of the mortality attributable to the use of oral contraceptives. It is concluded that there has been an increase in the mortality from venous thromboembolism in young women in recent years of a magnitude compatible with the existence of a causal relation between the and death from venous use of oral contraceptives thromboembolism.
Summary
INTRODUCTION
RECENT epidemiological investigations 1-3 have indicated that oral contraceptives sometimes cause certain types of thromboembolic disease, particularly venous thrombosis and pulmonary embolism (" venous thromboembolism "). When, however, Swyer4 and a Worldl Health Organisation Scientific Group5 studied the national mortality-rates for the thromboembolic diseases in the United States and the United Kingdom up to the year 1964, they concluded that trends among women of childbearing age had been paralleled by corresponding trends among men, and that vital statistics did not support the suggestion that oral contraceptives cause thromboembolic disease. We therefore thought it would be useful to examine the mortality statistics for venous thromboembolic disease in England and Wales in recent years in relation to estimates which have recently been provided of the mortality attributable to the use of oral
contraceptives.2 Deaths in which the 1.
Royal College
underlying
of General Practitioners.
women
Jl
cause
is certified
as
R. Coll. Gen. Practnrs. 1967,
13, 267. 2. Inman, W. H. W., Vessey, M. P. Br. med. J. 1968, ii, 193. 3. Vessey, M. P., Doll. R. ibid. p. 199. 4. Swyer, G. I. M. ibid. 1966, i, 355. Techn. Rep. Ser. Wld Hlth 5. World Health Organisation. no. 326.
at
article.)
JOSEPHINE A. C. WEATHERALL MEDICAL
in
puerperium
Org. 1966,
The investigation by the Committee on Safety of Drugs showed 2that only about 25% of thedeaths assigned to list nos. 463-466 during 1966 in women, aged 20-44 years, occurred in the absence of wellrecognised predisposing conditions such as recent surgery, heart-disease, or previous thromboembolism. The study also indicated that among women with no known predisposing conditions about 5 deaths from venous thromboembolism in the age-group 20-34 years, and about 9 such deaths in the age-group 35-44 years, were attributable to the use of oral contraceptives. Unfortunately, no reliable estimate is available of the risk of fatal venous thromboembolism from oral contraceptives in women with predisposing conditions, nor of the frequency of use of oral contraceptives by such women. The prevalence of women with predisposing conditions in the population at large, however, is unlikely to exceed 10% and it seems probable that the use of oral contraceptives by such women would be relatively infrequent. We have therefore assumed that most deaths from venous thromboembolism caused by oral contraceptives occur in previously healthy women. RESULTS
accompanying table gives national estimates of the consumption of oral contraceptives for the years 1963-67 provided by Intercontinental Medical Statistics, Ltd. These data represent only oral contraceptives purchased The
by women from chemists, and do not include the relatively small proportion supplied by such agencies as dispensing doctors and family planning clinics.22 Data on the agedistribution of users of oral contraceptives for 1965-67, provided by the same source and by Medical Data, Ltd., showed very little variation between the years. In the absence of specific information for 1963-64, it has been assumed that the age-distribution of users of oral contraceptives in these years was much the same as in the later ones. (Vessey and Doll3 presented some evidence that 6. World Health Organisation. Manual of the International Statistical Classification of Diseases, Injuries and Causes of Death. Geneva, 1957. 7. World Health Organisation. Manual of the International Statistical Classification of Diseases, Injuries and Causes of Death. Geneva, 1967. 8. Inman, W. H. W., Vessey, M. P. Unpublished.
NATIONAL ESTIMATES (PROVIDED BY INTERCONTINENTAL MEDICAL STATISTICS, LTD.) FOR THE CONSUMPTION OF ORAL CONTRACEPTIVES IN THE YEARS 1963-67, TOGETHER WITH ESTIMATES OF THE NUMBERS OF DEATHS ATTRIBUTABLE TO THE USE OF ORAL CONTRACEPTIVES IN EACH YEAR
95
Male and female age-specific mortality-rates per million per (shown as unbroken lines).
For further
explanation
of
figures,
annum
venous
thromboembolism (I.C.D. list
nos.
463-466) by
year
see text.
using oral contraceptives in 1964 were both older and of higher parity on the average than those using them in 1966. The results presented here are not materially altered if this was correct.) On this basis, estimates of the numbers of deaths from venous thromboembolism attributable to the use of oral contraceptives in each year from 1963 to 1967 have been calculated by scaling up or down the number of deaths in 1966 in direct proportion to the relative frequency of oral-contraceptive use in the other years (see table). The accompanying figure shows the male and female age-specific mortality-rates per million per annum from venous thromboembolism (list nos. 463-466) for 1953-67. Before 1958, the Sixth Revision of the I.C.D.9 was used for coding death certificates, but changes relating to the venous thromboembolic diseases in the Seventh Revision were minor and did not produce any obvious distortion of the mortality curves shown in the figure. Regression lines have been fitted to each set of agespecific mortality-rates for the ten years 1953-62 during which oral contraceptive use in this country was negligible. The slopes of these lines are, of course, subject to considerable standard errors, especially at the lower ages, at which numbers of deaths are small. However, for each sex and age group a linear regression provides a reasonable description of the rates from 1953-62, and we have the lines to provide a rough basis for predictextrapolated " ing expected " death-rates in 1963-67 independent of any change attributable to the introduction of oral women
contraceptives. For men, aged 20-34 trend in mortality from
years, there was a slight upward 1953 to 1962, and the regression line fitted to the data for these years (shown as the straight, broken line in the male section of the figure) is seen to provide an excellent prediction of the rates observed in 1963-67. For women, aged 20-34 years, the trend, if anything, was very slightly downward during 1953-62, but the observed values for 1963-67 are all seen to lie above the values predicted by the regression line (shown as the straight, broken line in the female section of the figure). These observed values do not, however, show any clear trend with time. Also shown in the figure is the modification to the regression line for women, aged 20-34 years, that is pro9. World Health
for
Organisation. Manual of the International Statistical Classification of Diseases, Injuries and Causes of Death. Geneva, 1948.
duced by adding in the estimated numbers of deaths attributable to the use of oral contraceptives (see table). It is clear that the very slight increase in female mortality from venous thromboembolism observed during 1963-67 is more than sufficient to account for the estimated effect of oral contraceptives on the death-rate. The mortality-rate was fairly stable for both men and women, aged 35-44 years, during 1953-62. In the interval 1963-67 there was an increase in the male deathrate, though no clear trend with time is apparent. Among women, however, there has been a slow but steady increase in the mortality-rate during 1963-67. Even if it is assumed that the factors which led to an increase in the male death-rate at ages 35-44 years during 1963-67 also affected the female death-rate, the excess deaths observed among women are still sufficient to account for the estimated effect of oral contraceptives on the death-rate
(see figure). Death-rates from venous thromboembolism have shown a steady increase in the older age-groups (45-54 years, 55-64 years) in both sexes from 1953 to 1962. In the age-group 45-54 years for both sexes, the regression lines fitted to the rates during 1953-62 provide excellent estimates of the rates actually observed in 1963-67. In the age-group 55-64 years, however, the rate of increase in the female death-rate during 1963-67 has been slower than that predicted, while the reverse is true of the male death-rate. COMMENT
The
interpretation of trends in mortality statistics must be always made with due regard to the many factors which may influence them. Quite apart from variation in the true pattern of a disease in the population, the frequency with which doctors make certain diagnoses and the importance attached to them may change. The latter will influence the way in which diseases are recorded on death certificates, and this in turn will affect the selection of the underlying cause of death when more than one condition is reported. Difficulties of this type were recently discussed in relation to asthma mortality by Speizer et al.I0 We do not, therefore, wish to suggest that the mortality data we have presented necessarily offer any positive evidence to support a causal association between the use of oral contraceptives and death from venous thrombo10.
Speizer,
F. E.
Doll, R., Heaf,
P. Br.
med. J. 1968, i, 335.
96
embolism. We do, however, conclude that there has been a slight increase in mortality in young women from venous thromboembolism in recent years of a magnitude entirely compatible with the existence of a causal association. Requests for reprints should be addressed to M. P. V., M.R.C. Statistical Research Unit, University College Hospital Medical School, 115 Gower Street, London W.C.I.
TABLE II-PRESCRIBING OF AMPHETAMINES BY PRACTICES
USE OF AMPHETAMINES IN GENERAL PRACTICE HELEN HOOD B.Sc. Belf. RESEARCH ASSISTANT TABLE III-GEOGRAPHICAL VARIATION IN AMPHETAMINE PRESCRIBING
O. L. WADE Cantab., F.R.C.P.
M.D.
PROFESSOR OF THERAPEUTICS AND PHARMACOLOGY
DEPARTMENT OF THERAPEUTICS AND THE
QUEEN’S
PHARMACOLOGY,
UNIVERSITY OF BELFAST
THE effectiveness of amphetamines and related drugs in the treatment of psychiatric illness and obesity has been questioned.12 Because some patients become dependent on these drugs, and because amphetamine abuse has become a feature of certain teenage cults and is not uncommon in middle-aged women, examination of the prescribing of amphetamines and related drugs is timely. This paper analyses the prescribing of these drugs by general practitioners in Northern Ireland. THE SURVEY
We examined all
prescriptions for amphetamines issued by general practitioners in the 3 months from April to June, 1966, excluding the small proportion issued to private patients and by doctors who do their own dispensing. In order to compare the use of preparations of different sizes and of different composition, a unit of amphetamine was defined as the equivalent of 5 mg. amphetamine. Phenmetrazine (’ Preludin ’) and methylamphetamine hydrochloride Methedrine ’) were included in this survey. RESULTS
In 3 months 11,400 prescriptions were written for nearly a million units of amphetamine. Nine different preparations were prescribed,Durophet ’ being the most common (table i). In the analysis of prescribing habits the data from 32 practices each with less than 500 patients were excluded. These practices accounted for 26,486 units of amphetamine (2-5% of the total). In the 444 practices examined there were 56 (12%) in which no ampheta1. 2.
Macgregor, A. G. Prescrib. J. 1963, 3, Kiloh, L. G. ibid. 1964, 4, 57.
25.
mine was prescribed, 293 (67%) in which 1-999 units per 1000 patients on the list were prescribed, and 89 (20%) in which 1000-4999 units per 1000 patients were prescribed (table 11). 9% of the total amphetamine units were prescribed by 1.3% of the practices. Examination of the geographical variation in amphetamine prescribing (table ill) showed that it was highest in Belfast, with a gradient of high use in the eastern counties to low use in the west. The prescribing of amphetamines has declined slightly since the time of this survey (see figure), but a recent reexamination of the prescribing of these drugs does not suggest that the pattern of prescribing has changed. CONCLUSIONS
The survey suggests a considerable divergence of opinion among doctors on the value of these drugs. In 12% of practices they were not prescribed at all, yet l-3°o of the practices accounted for 9% of all the prescriptions. The geographical differences in prescribing are unex-
TABLE I-PRESCRIBING OF AMPHETAMINES