J Chron Dis 12;586-591 (Dec) 1960
From the Department of Physiology, Dartmouth Medical
School and the Brattleboro Retreat
(Received for publication April 9, 1960)
Death rates vary widely from state to state. Significant negative correlations between weighted average hardness of water supplies and death rates from cardiovascular diseases in the United States have been reported.1 In States supplied with harder-than-average water, average annual age-adjusted death rates from all causes and from cardiovascular diseases for both sexes, for males, for females, and for white males aged 45 to 64 years, were in general lower than national averages. These relationships applied to death rates from coronary heart disease as well as to rates from all other cardiovascular diseases. Rates for Coronary heart disease in older white males correlated significantly with hardness and with a number of constituents of drinking water in the 163 largest municipalities of the country. Correlations were absent for deaths from all other causes.1
In order to define these interesting relations further, the present report examines correlations by state between weighted average hardness of "finished"* water (an index of exposure per person) and average annual death rates from a number of chronic and degenerative diseases for the years 1949 to 1951.2 Correlation coefficients (r values) were calculated by the usual method.3 Values for water were measured in 1950 to 1951.4 Diseases listed among the ten leading causes of death and subdivisions thereof with relatively high death rates in the decade 55 to 64 years were chosen in order to avoid, as much as possible, rates possibly erroneous because of small numbers of cases. This consideration applied in general to death rates less than 10 per 100,000 persons.
States which have harder water than the national average are all those in the East North Central (5 states), West North Central (7 states), and Mountain (8 states) areas, plus Florida, Kentucky, Oklahoma, Texas, and California.5 Therefore, 25 states have water harder and 23 states and the District of Columbia have water softer than the national average of 97 parts per million.
It so happens that many "soft-water" states have large nonwhite populations. The correlation coefficient of hardness of water and per cent of the non white population per state was -0.341, which is significant at the 2 per cent level of confidence. Therefore, death rates of white persons were analyzed in order to avoid this influence. In all cases, however, correlation coefficients were calculated as well for all races and, in most cases, for all ages; these data are reported when they appear of interest to the problem.
In Table I are shown the correlation coefficients and their levels of significance for water hardness by state and death rates from the ten leading causes of death for all ages, races, and both sexes. Significant values appear mainly in the cardiovascular group of diseases, motor vehicle accidents, and congenital malformations. In Table II, confined to white persons aged 55 to 64 years, major categories are subdivided into their constituent parts. In white persons the highest negative correlations occurred with those disorders commonly believed to be associated with atherosclerosis and with arterial hypertension, (hypertension with heart disease, vascular lesions affecting the central nervous system, general arteriosclerosis, and arteriosclerotic heart disease, including coronary disease); correlations were not significant in the case of rheumatic heart disease, nonrheumatic chronic endocarditis and other myocardial degeneration, chronic nephritis, or "other diseases of heart" (Table II).
In the case of malignant neoplasms, one probably significant negative correlation (p = 0.02) appeared for neoplasms of the respiratory tract in males aged 55 to 64 years. In this decade there was some relationship (p < 0.05) between water hardness and death rates from all neoplastic diseases and from those of unspecified sites; for other carcinomas, lymphomas, and leukemias, there were no correlations of significance.
Only in the decade 55 to 64 did a possibly significant negative correlation (p < 0.05) appear for cirrhosis of the liver. In the case of motor vehicle accident deaths a high positive correlation was found; for other accidental deaths it was also positive but not as significant (Table I). The correlation for congenital malformations, at all ages, was also positive (p < 0.01). Correlation coefficients were small in the cases of other major causes of death, e.g., "certain diseases of early infancy," influenza and pneumonia and total infant deaths per 1,000 live births (Tables I and II).
There seems to be clearly established a relationship between weighted average hardness of "finished" water and average annual death rates by states in 1949 to 1951 for diseases associated with arterial hypertension and disorders secondary to atherosclerosis. The high correlation coefficients, -0.5 or more for two of these disorders are significant at probability levels of about one in one thousand. Therefore, if one grants the validity of the data, this relationship can be accepted. Lack of significant correlations between water hardness and three major categories of heart diseases — chronic rheumatic heart disease, "nonspecific chronic endocarditis and other myocardial degeneration" (which undoubtedly includes an unknown number of cases of arteriosclerotic heart disease), and "other diseases of heart" — points further to a true relation between water hardness and atherosclerotic and or hypertensive disorders.
Sauer and Enterline6 recently examined a number of factors which might affect data on death rates and concluded that variations from state to state were real and were unaccountable by the quality or quantity of physicians. They also believe that population and total death rates are fairly accurate. The present study is based upon their conclusions and upon chemical and analytical data on specimens of municipal water collected during the period when the death rates were calculated (1949-1951). Errors in annual variations were thus largely avoided.
Among the noncardiovascular diseases, a possibly significant correlation with malignant neoplasms appeared only in respect to male respiratory tract neoplasia Analyses of less common neoplasms included in the total were not made, for the death rates were based on too few cases for valid analyses. A suggestive relationship appeared for cirrhosis of the liver but for no other common diseases.
Of the ten leading causes of death, correlations were not statistically significant for diabetes mellitus, infant deaths, diseases of early infancy, or pneumonia and influenza, either for total population or older white persons, with the use of the rates available for 1949 to 1951 or for 1957. Curiously enough, a significant positive correlation did appear for congenital malformations in the total population and in white persons of all ages, which could not be explored further because of the low rates in adults. Rates for other diseases were too low in many states for accurate computation.
The curious positive correlation between water hardness and motor vehicle accident deaths may be a geographical accident. The hard water areas are in general those midwestern and mountain states with large areas and low densities of population, where long-distance driving is usual. In many, motor vehicle laws are less strict than in the more crowded seaboard states; perhaps this factor accounts for the correlation. In order to explore this factor, a correlation coefficient was determined on total cardiovascular deaths in older white males by state and density of population per square mile: r = + 0.279 (p = 0.05). The coefficient for hardness of water and density of population was -0.292 ( < 0.05). These coefficients are much lower than that for water and cardiovascular death rates (-0.56).
It was impossible to compare death rates of white and nonwhite persons, since the cases were few in many states and in 17 there were few nonwhite persons, making computations invalid.2 In order to include as many disorders as possible, death rates for persons in the decade below mandatory retirement age were chosen when the values were considered sufficiently accurate from a statistical standpoint.2
To ascertain the effects of the inclusion of deaths in nonwhite persons and errors therefrom on these correlation coefficients, calculations were made for all categories of diseases listed. Comparisons of white and total populations showed relatively minor changes in r values for most of the cardiovascular diseases; values became statistically significant only in the case of chronic nephritis when nonwhites were included. Those for several but not all malignant neoplasms became significant, as did those for diabetes mellitus, pneumonia in persons aged 55-64, and general arteriosclerosis, reflecting the higher incidences of these disorders in nonwhite persons from "soft-water" states. Therefore, unless both white as well as total populations are analyzed, errors due to this factor can appear if death rates differ in the two racial groups.
Yerushalmy and Palmer7 proposed certain procedures for evaluating etiological and accessory factors in chronic diseases based on modifications of Koch's postulates. No inference as to etiological factors is intended in this report, and the influences associated with hardness of water are obviously accessory environmental factors at the most. However, some of the postulates of these authors are partly satisfied by this study, and some are not. In discussing this matter, it must he emphasized that correlation coefficients shown in the tables, while statistically significant, are usually small. In Tables I and II there are only four values each greater than 0.45. Yerushalmy and Palmer list the following points:
"1. The suspected characteristic must he found more frequently
in persons with the disease in question than in persons without.
the disease, or
"2. Persons possessing the characteristic must develop the disease more frequently than do persons not possessing the characteristic."7
In so far as death rates in white persons are concerned, an influence associated with hardness of water is found at ages 55-64 in the cases of hypertensive heart disease and coronary heart disease, and at all ages in motor vehicle accidents and congenital malformations. This is only an influence, for there were no "persons without the disease" in any state or city, although there were many perhaps without "the characteristic."7
The third point of these authors, concerned with testing the association for validity by investigating the relationship of the characteristic with other diseases, has been investigated and little or no relationship can be found as far as this influence is concerned. Four causes of death stand out: smaller coefficients of correlation, while still considered statistically significant, may or may not be of actual significance coefficient of 0.45 is high in any study with such a large number of possible variables as this one has.
This report serves to emphasize a relationship of certain chronic disorders with a variable environmental factor to which all persons are exposed. There may be a cause-and-effect relation, i.e., some factor in, or not in, different waters influences these death rates, or each of the two parameters may be related to a third. In any event, experimental work along these lines may he rewarding.
Further inquiry into the relationship between hardness of
finished potable water by state and death rates from the ten
leading causes of death and subdivisions thereof in white persons
55 to 64 years of age disclosed:
1. A. Negative correlations with hypertensive diseases and with common conditions secondary to atherosclerosis (p < 0.01).
B. No significant correlations with other diseases of the heart.
2. A. A possibly significant negative correlation (p = 0.02) with neoplasms of the male respiratory system.
B. No significant correlations with other common neoplasms.
3. A. A suggestive negative correlation (p < 0.05) with cirrhosis of the liver.
B. No significant correlations with diabetes mellitus, influenza and pneumonia, infantile disorders, ulcer of stomach or duodenum, ''senility,'' or intestinal obstruction.
4. Positive correlations with motor vehicle accident deaths (p < 0.01) and congenital malformations (p < 0.01) in all age groups.
In hardness or softness of potable water may lie a clue to an influence affecting death rates from hypertension and atherosclerosis, and possibly a few other chronic disorders.
1. Schroeder, H. A. : Relation Between Mortality From Cardiovascular Disease and Treated Water Supplies, J.A.M.A.. 172:1902, 1960.
2. National Office of Vital Statistics. Death Rates for Selected Causes by Age, Color, and Sex, United States and Each State, 1949 to 1951. Vital Statistics vol. 49, Special Rep. Nos. 16-27, 30-43, 48, 52, 58-60, 1959.
3. Herdan, G. : Statistics of Therapeutic Trials, Amsterdam, 1955, Elsevier Publishing Company.
4. Lohr, E.W., and Love, S. K.: Industrial Utility of Public Water Supplies in the United States, 1952: Parts 1 and 2, Geological Survey, Water-Supply Papers 1299 and 1300, Washington, D. C., 1954, U. S. Government Printing Office.
5. Statistical Abstract of the United States, ed. 80, U. S. Department of Commerce, Washington, 1959.
6. Sauer, H. I., and Enterline, P. E. : Are Geographic Variations in Death Rates for Cardiovascular Diseases Real? J. Chron.. Dis. 10:513. 1959.
7. Yerushalmy, J., and Palmer, C. E. : Methodology of Investigations of Etiologic Factors in Chronic Diseases, J. Chron.. Dis. 10:27, 1959.
This work was supported by a grant from Ciba Pharmaceutical Products, Summit, N. J., and by a grant from the U. S. Public health Service, Grant H-5076.
*"Finished" water is water which has been treated in municipal treatment plants by procedures such as chlorination, filtration, softening, adjustment of pH, control of algae, removal of iron and manganese, coagulation, and aeration. As such, it is the water delivered to the tap. "Raw" water has received no treatment.
This page was first uploaded to The Magnesium Web Site on September 14, 2002