The Lancet, April 22, 1961
Schroeder,1 2 following Japanese work,3 found substantial and highly significant correlations between average hardness of the finished (drinking) water and mortality from cardiovascular disease in the states and large municipalities of the United States. These associations were negative—i.e., the softer the water the higher the cardiovascular death-rate. A reconnaissance using available official statistics has therefore been made in the county boroughs of England and Wales to see whether there is a case for investigation in this country. A third of the population lives in these towns, and differences in cardiovascular mortality among them are well recognised. Around the 1951 census, for example, in men aged 45-64 there was a range of 168-382 per 100,000 in coronary heart-disease and of 82-182 in cerebrovascular disease.4
Correlations were calculated between (a) death-rates from thirty-eight causes and groups of causes of death in 170 sex-age brackets (i.e., in every instance where data were available and numbers at all adequate), and (b) indices of average hardness of the local water as finished and delivered into supply in the 83 county boroughs. The correlation coefficient is a useful method of exploring a question such as this. The distributions being examined, however, are sometimes very skewed, and problems therefore arise which we are trying to solve in various ways.
Table I gives the results for the leading causes of death, and some others of special interest, in middle age and at 65-74. Attention may be drawn to several features. There are striking correlations (part I, line 1), all negative, between water hardness and total cardiovascular mortality: the softer on average the local water in the county boroughs of England and Wales, the higher on average the cardiovascular death-rate. Among the groupings within cardiovascular disease (there is no choice about these) the correlations are substantial and highly significant for deaths from vascular lesions of the nervous system and for those certified to "myocardial degeneration", &c. In coronary heart-disease the correlation is sizeable only in middle-aged men. Correlations with mortality certified to hypertension and other cardiovascular causes are remarkable.
Correlations with non-cardiovascular causes of death (part II of table I) are small and mostly insignificant at the 1% level. The main exception is bronchitis, and this is interesting in view of the several connections between bronchitis and heart-disease. The correlations for noncardiovascular mortality ("other disease") when bronchitis is excluded fall to -0.12, -0.26, -0.13, and -0.20.
Correlations in diabetes are negligible, and with chronic nephritis, &c., irregular. (There are too few deaths to attempt any study of stone.)* There are no other results of interest.
The main problem in this kind of "ecological" study is to determine whether the correlations found between water hardness and cardiovascular disease are likely to represent cause and effect, or whether they merely reflect other factors with which both the water and the mortality are associated. Thus, so far as the county boroughs are concerned, soft water is found mainly in Lancashire,the West Riding of Yorkshire, and South Wales. There may be host factors in the population which arc responsible for the high mortality from cardiovascular disease in the regions, and water softness could be fortuitous: alternatively other environmental factors in these regions may be responsible. For example, the softness of local water was one of the reasons why the industrial revolution (cotton, wool) started in south-east Lancashire and the West Riding: it may he that industrial and urban conditions increase the cardiovascular death-rate, again regardless of the composition of local water.
Much more work will have to he done on these questions. Meanwhile, it is worth noting that although mortality in general is high in the three main soft-water areas, it is only in cardiovascular disease that important correlations between mortality and the water-supply appear — i.e., there seems to be a specific relationship between water and cardiovascular mortality. Furthermore, when the county boroughs of Lancashire, the West Riding and South Wales are excluded from the calculations, substantial and highly significant correlations between water hardness and cardiovascular death-rates are still found among the remaining 51 towns—a good test in itself of the hypothesis.
Comparison within regions is helpful in this kind of problem. When this is done, the results in general agree with the national figures. However, Birmingham, with very soft water, has low mortality similar to the neighbouring Midland county boroughs with their hard water. This observation may well be crucial, though difficult to interpret in view of the traffic and exchange there is among people, food, and drink across local boundaries within a region.
In brief, questions must be asked whether the association with water is a matter of regional variation. Equally, it must be asked, might regional differences in cardiovascular mortality be due in some measure to differences in their water-supply? Intensive study of this will mean extending the analysis beyond the county boroughs—a daunting prospect. The same problems are arising in the United States.12 It seems unlikely that the same regional factors associated with cardiovascular mortality could be producing indirect correlations with water-supply both in the United States and in England and Wales, and as remarkable if different regional factors happened to do so. In England mortality in general, cardiovascular and other, is higher in the North, but this latitude factor has been found to be irrelevant to the main problem. Correlations between water hardness and local rainfall, temperature, and sunshine averages in the county boroughs will shortly be available.
Some progress has been made with the problem of possible confounding social factors. Correlations were calculated between water hardness in the county boroughs and a wide variety of local environmental indices (table II). Only two of these correlated significantly at the 5% level with water hardness, and none did so at 1%. The associations with air pollution may be relevant; but, when correlations were recalculated to eliminate the effects of air pollution, those between cardiovascular mortality and water hardness were scarcely affected, while the highish correlations already mentioned between bronchitis and water hardness dropped sharply.
So far, then, there is no indication that local water hardness is merely indicating some important social-environmental cause of death from cardiovascular disease. Some interesting associations were, of course, found between some of these social indices and local death-rates, and these will be described elsewhere.
Work is proceeding on this. Table III gives the results for the 53 county boroughs with the fullest data. These are mainly larger towns, and they seem to be representative (line 1 of table III corresponds to line 1 of table I). Table III probably gives the best information on local water in the present communication. Correlations with cardiovascular mortality are high for calcium content, contrasting remarkably with magnesium. These are the main cations present in British drinking water; no figures for others (e.g., sodium) are available.†
Table IV illustrates the kind of variation of mortality with water hardness that has been found.
Clearly there is a case for further inquiry, and several lines have already been suggested. It will be possible to extend the mortality analysis to social class, if not to occupation. Historical study of local death-rates from cardiovascular disease is proving very laborious. There is no suggestion that increase in water softening in recent years may be importantly involved in the recent increase of coronary heart-disease (in any event the coronary correlations are less impressive than those with cerebrovascular disease, and there is no suggestion of any rise in the incidence of the latter). Postmortem material may give a clue to which of the main elements of cardiovascular disease in middle age-mural atheroma, intravascular thrombosis, hypertension16 is specially involved. It must be emphasised that all data presented refer to mortality, and some attempt will have to be made to study incidence and morbidity. Ad-hoc examination of samples of drinking water will need to be undertaken. Several modes of action of water vis-a-vis cardiovascular disease suggest themselves: hard water may be protective because of its mineral content (cf., fluorine and dental caries) or, conceivably, soft water could carry harmful trace elements into supply. Both of these, and other possibilities, will have to be investigated.
The softer the water-supply in the county boroughs of England and Wales the higher the local death-rate from cardiovascular disease tends to be. What this means is not at present clear, and further investigation is indicated.
We are greatly obliged to the water authorities of the County boroughs and their chief officers, and to Dr. E. Windle Taylor of the Metropolitan Water Board, for their generous Cooperation, We are also most grateful for access to I.B.M. computers, including , services by the I.B.M. Applied Science Division and its representatives. The General Register Office, the Ministries of Health and of Housing and Local Government, Prof. C. J. O. R. Morris, and Prof. F. L. Warren of the London Hospital Medical College, and the computing and clerical staff of the unit have given us much help.
J. N. Morris
M.A. Glasg., F.R.C.P., D.P.H.
M. D. Crawford
J. A. Heady
Social Medicine Research Unit of the Medical Research Council and Department of Social Medicine, London Hospital Medical College, London, E.2
*Causes of death and groups of causes have been analysed as presented in the tables of the Registrar General. Those of deaths certified to diabetes and hypertension are difficult to interpret. In the case of bronchitis it is worth remembering that there is considerable interchange of this diagnosis with "myocardial degeneration" on death certificates. A more general problem is the doubtful validity of specific certified causes of death after 65 or 70 years of age. (See ref 9 Opposite.)
†It is interesting that the non-cardiovascular mortality and the calcium content of local water also show correlations, but these are largely due to an association of calcium with latitude. The correlations between calcium and cardiovascular mortality are little affected by the latitude factor; they remain substantial and highly significant when latitude is eliminated by partial correlation.
1. Schroeder, H. A. J. Amer. med. Ass. 1960, 172, 1902.
2. Schroeder, H. A. J. chron. Dis. 1960, 12, 586
3. Kobayashi, J. Ber. Ohara Inst. 1957, 11, 12.
4. Social Medicine Research Unit and Registrar-General, Quarterly Return no. 432, appendix D, 21, 1956. H. M. Stationery Office.
5. Registrar General, Statistical Review of England and Wales, 1948-58; part I (tables, medical). H. M. Stationery Office.
6. Registrar General, Decennial Supplement, Engalnd and Wales, 1951: area mortality, 1950-53. H. M. Stationery Office.7. Registrar General, Census, 1951, England and Wales; population tables. H. M. Stationery Office; and unpublished data.
8. Registrar General, Census, 1951, England and Wales; occupation tables. H. M. Stationery Office.
9. Morris, J. N. Uses of Epidemiology; Edinburgh, 1957.
10. Ministry of Housing and Local Government, 1951; rates and rateable values in England and Wales. H.M. Stationery Office.
11. Ministry of Health, Annual Report, 1955, part I; the National Health Service, appendix xviii, table G. H.M. Stationery Office; and unpublished data.
12. United States Public Health Service, personal communications, 1961.
13. Daly, C. Brit. J. prev. soc. Med. 1959, 13, 14.
14. Wilkins, L. T. Appl. Statist. 1952, 1, 27.
15. Morris, J. N., Titmuss, R. M. Med. Offr, 1944, 72, 69, 77, 85; and unpublished data.
16. Morris, J. N., Crawford, M. D. Lancet, 1961, i, 47.
Ministry of Health, Report on Water Softening, England and Wales, 1949, H.M. Stationery Office.
Hollingsworth, D. F. Chem. & Ind. (Rev.) 1955, p. 1510.
This page was first uploaded to The Magnesium Web Site on September 17, 2002