921 Africans without reference to colour or creed. In this respect South Africa is probably second to none on the African continent. Almost anything that takes place these days in South Africa is grist to the political mill, but we would urge the medical profession to insist on the same rigour and objectivity in obtaining facts and evaluating situations as they do in clinical practice. Since the W.H.O. document contains so many inaccuracies and misrepresentations, fairness would dictate that only a full on-the-spot investigation will do. The facts are available and the South African Government has made it clear that any organisation with a legitimate interest may inspect our mental health services, private or public, without let or hindrance. Let us trust they will do so. The supporting documents referred to2-s may not be readily available in other countries. Anyone who wishes to have them should contact the Department of Health, Private Bag X88, Pretoria 0001, South Africa.
Department of Psychiatry, University of Cape Town, Groote Shuur Hospital, Cape Town 7925, South Africa
L. S. GILLIS Chairman, Executive Committee,
Society of Psychiatrists of South Africa (M.A.S.A.)
MORTALITY ASSOCIATED WITH THE PILL little alarm, has attended the publication from the Oral Contraception Study of the Royal College of General Practitioners (Oct. 8, p. 727). One principal finding in that paper was that the ratio of the mortality-rate in ever-users to that in controls, attributable to all diseases of the circulatory system, was approximately 5, and that this result was statistically significant at the level of 0.01. The report notes that this and other associated ratios were approximate and were based on small numbers, but it did not estimate the resulting imprecision, a point noted in your accompanying editorial. A statement of the confidence limits
SIR,-Considerable publicity, and
finite rate ratio. The lower limits bear out the reported levels of statistical significance. The seemingly dramatic ratio of 5, underpinned by the striking level (0.01) of statistical significance, may well range from a value as low as 1.6 to an infinitely high value, when confidence limits are considered. Pill optimists and pill pessimists alike may draw ammunition from that statement, but what real meaning should it convey to the uncommitted? I think that the lesson is simply that where a condition is rare, the increased risk of its occurrence in an individual due to the voluntary adoption of a course of medication is not rationally to be measured by a ratio of rates of occurrence. Rate ratios in such circumstances are highly uncertain and are highly irrelevant, and both of those properties follow from the rarity of the condition. The criterion of relevancy answers the question: "How much worse off (in respect of this particular condition) is the woman who decides to take the pill?" If she has a risk of dying of 5/100 000 each year without the pill, and 26/100 000 with it, her chance of survival declines from 99 995/100 000 to 99 974/100 000, which is an absolute reduction of 21 x 10-5 or a relative reduction of virtually the same extremely small amount. It is regrettable that so few journalists, and surprising that so few epidemiologists, appear to take this rational view of the situation. But hot news will always evaporate cold reason.
Department of Community Medicine, University of Manchester,
Manchester M13 9PT
CONFIDENCE LIMITS OF RATE RATIOS IN R.C.G.P. STUDY
SIR,-Evidence on the relation of diet to coronary heart-disand other vascular disorders, has been assessed indepen-
dently by 18 national committees. Each has advised reduction in the consumption of saturated fat, either by the population as a whole (14/18) or by groups considered to be at high risk (4/18) of which women on oral contraceptives is one. Partial substitution of saturated by polyunsaturated fat is recommended by 16/18 committees and restriction of cholesterol intake by 16/18. A "fat switch" may be particularly important in reducing platelet adhesiveness and aggregation, and thus a tendency to thrombus formation. However, routine measurement of the concentration of blood lipids is not yet practical and there are also difficulties over the interpretation of this concentration. There is no cut-off point between normal and abnormal, and assessment of the significance of any level must take account of other risk factors, including inherited predisposition, smoking habit, raised blood-pressure and increased permeability of the arterial wall. It has been shown that these are not important in people who habitually eat a healthy diet and thus have a low plasma-cholesterol, becoming so only when they migrate and change their dietary habits. Women who choose to continue with oral contraceptives should be given advice about smoking and have their bloodpressures recorded, but should also be given a clear explanation of the recommended diet. This could be done by the family doctor, a practice nurse or the family-planning clinic. This practice may reduce the incidence of vascular complications but the
which can be ascribed to these ratios helps to put the matter into perspective. Using a simulation method-based on the assumptions that the observed (and published) numbers of deaths are equal to the "expected" numbers and that the numbers "observed" in hypothetical replications of the study follow Poisson distributions with those expectations-I have estimated the confidence limits (see table, together with details extracted from the R.C.G.P. paper). The method I used is simplistic, but it provides a reasonable guide to the limits. The "undefined" upper limits arise because the numbers of controls were extremely small and thus in many replications (e.g., in more than 50 out of 10 000 for I.C.D. code 390-458) the "observed" number was zero, yielding an indefinite or in-
should be put to the test and monitored of lipid levels and platelet func-
University of Edinburgh, 21 Buccleuch Place, Edinburgh EH8 9LN
R. W. D. TURNER
SIR,-The report from the R.C.G.P. Oral Contraceptive Study is open to several criticisms. The study began in 1968. The smoking habits of the women were those "at entry". We seem to
believe that 46 000
changed smoking habits for nearly eight years. This is very un-
the absence of updated information about smoking habits must cast some doubt on the value of the results, particularly since an earlier report of this study’ indicated that in October, 1968, pill takers smoked more often and more heavily. Mortality from arterial disease has well-established connections with smoking, but it is the number of cigarettes smoked rather than smoking itself which correlates with both morbidity and mortality due to arterial disease. Pill takers who smoked more heavily in 1968 might have increased their amount of smoking by 1976, and this factor alone would discriminate between "pill takers" and controls.
The latest report is based on 206 689 women-years of observation. In October, 1968, there were 23 611 pill takers and 22 766 controls. What has happened to the missing womenyears of observation? How were they distributed between takers and controls? What were the smoking habits of these women, and how many have died and for what reasons? One cause of error which has been recognised but not fully explained is the quality of certification of death. If the same general practitioner certifies the death and reports it to the R.C.G.P. study one would expect the reporting to be similar. Doubt should, however, exist about the accuracy of reporting bearing in mind the effect which a known association with the pill has on- the precision of identification of cause of death. What proportion of causes of-death among both pill takers and controls is supported by post-mortem evidence? Arterial disease is associated with objective variables such as weight, blood-pressure, and serum-cholesterol. The R.C.G.P. study has omitted to include such measurements: they would have been invaluable in interpreting the findings.
The Oxford F.P.A. study introduces one factor into the evaluation which the R.C.G.P. ignores-namely, the standardised mortality-rate per 100 000 women years for England and Wales. Taking the two studies together, the following table can be constructed. ’
STANDARDISED MORTALITY-RATES PER
This table suggests that, for circulatory disorders, it is the controls rather than the pill takers who deviate most from England and Wales as a whole. It seems that for both studies, the sizes of the samples are inadequate and prolonging observation on diminishing samples is unlikely to correct this fault.
Attribution of statistical significance to numbers such as 2 and 3 (in the R.C.G.P. study) should not be taken seriously. The R.C.G.P. study may have identified that women who smoke more heavily and become older have a higher risk of death due to arterial disease. Any direct connection with oral contraceptives is very far from substantiated. Department of General Practice, University of Manchester, Darbishire House Health Centre, Manchester M13 OFW 1.
of General Practitioners. Oral London: Pitman. 1974.
G. LLOYD Contraceptives and Health.
SIR,-Subscribers to The Lancet who have kept their backnumbers will have little need for alarm over the reports showing increased mortality risks of cardiovascular diseases among
oral-contraceptive users. In 1973 you published a report showing that half of the pill takers who were depressed had an absolute deficiency of vitamin B6 (pyridoxine) and all of these women responded clinically to its administration.’ Young women on the pill were found to excrete only half as much vitamin C as did controls, suggesting that they metabolised vitamin C more quickly.2 Ischaemic heart-disease has been reported to be strongly negatively correlated with vitamin C intake,3and in the 1940s an experiment on ten healthy young volunteers taking a diet devoid of vitamin C had to be abandoned when two of them suddenly became ill with chest pains and heart symptoms after physical exercise.4 And there was the finding that vitamin C protects against deep-vein thrombosis in surgical patients’ and the controversy over the amount required to protect elderly patients from thrombotic episodes. 6.7 In 197$Wynn* concluded that the "combined pill" raised the serum concentration of vitamin A (retinol), but lowered those of vitamins B2 (riboflavine), B6 (pyridoxine), B12 (cyanocobalamin), and C (ascorbic acid) and folic acid. Other vitamins may also be found to be deficient in "pill" takers: lack of vitamin B, (thiamine) causes degeneration of heart muscle, and vitamin E (alphatocopherol) strengthens the walls of blood-vessels and reduces the number of heart-attacks in coronary patients due to clots forming.9 Perhaps vitamin supplements are the answer to the problem of cardiovascular disease in women who take the pill. Regional Cancer Registry, Queen Elizabeth Hospital, Birmingham 15
SIR,-In view of the current interest in vascular accidents in women who are taking oral contraceptives, may I record an observation made while I was pathologist to the Christie Hospital in Manchester (1944-60). Among the large number of tumour specimens which passed through my department a very small group separated out. They were rare (I did not see one a year). They came from the mouth, lips, gums, or nose of a woman who was younger than most patients who had tumours at those sites; and the structure was a capillary angioma. By chance I found that one of these women was pregnant, and after that I suggested to the clinicians that such patients might be pregnant and this was confirmed. I did not pay special attention to the group but the cases remained in my mind as occurring during the luteal phase of the ovarian cycle. Oral contraceptives mimic the luteal phase of ovarian activity which in pregnancy is associated with the great vascular activity in the placenta and perhaps elsewhere in some individuals. It may be that the angiomas of the liver, subarachnoid haemorrhages, and angiomas of the foregut region in my collection are manifestations of an individual peculiarity in some women who take a contraceptive pill. The gravity of small angiomas would depend entirely upon their site and they might be more numerous than we think. Perhaps the site of an ectopic gestation is determined by a small vascular anomaly in a fallopian tube. 28
Douglas Crescent, Edinburgh 12
Adams, P. W., Rose, D. P., Folkard, J., Wynn, V., Seed, M., Strong, R. Lancet, 1973, i, 897. 2. Harris, A. B., Hartley, J., Moor, A. ibid. 1973, 201. 3. Knox, E. G. ibid. 1973, i, 1465. 4. Peters, R. A., Krebs, H. A., Coward, K. H., Mapson, L. W., Parsons, L G., Platts, B. S., Spence, J. C., O’Brien, J. R. P. ibid. 1948, i, 853. 5. Spittle, C. ibid. 1973, ii, 99. 6. Andrews, C. T., Wilson, T. S. ibid. 1973, ii, 39. 7. Loh, H. S., Wilson, C. W. M. ibid. 1973, 317. 8. Wynn, V. ibid. 1975, i, 561. 1.
9. Wilson, M. G. ibid. 1951,