Correlation of cancer death rates with altitude and with the quality of water supply of the 100 largest cities in the United States

Abstract
Data on water treatment and the quality of the finished water supply of the 100 largest USA cities (population over 60 million) were combined with data on [human] cancer deaths (age-adjusted) of the county or counties in which the cities lie, to look for significant correlations with altitude and with quality of water supply. The previously discovered decrease in cancer rates with increasing altitude (for countries around the world) was confirmed for these cities. Correlation coefficients [R], -0.46 for males and -0.32 for females, were very highly significant (P < 0.001). The rates were then altitude-adjusted for the subsequent analysis. Of the general measures of water quality, the highest correlation was with specific conductivity (R = - 0.41, P < 0.001), total dissolved solids (R = - 0.38, P < 0.001) and hardness (R = - 0.29, P < 0.03). The pH was not significantly correlated. Significant correlations were found for Ca2+, Mg2+, Na+, HCO3- and Cl- concentrations, but not for F-, NO3- or K+ concentrations, which were very low. Regression slopes for specific ions were much more uniform in terms of milliequivalents per liter than parts per million. The reduction in cancer appears to be nonspecific for the major sites of cancer, as is the previously reported altitude effect. A possible common explanation of the altitude and water-quality correlations is discussed in terms of the acid-base balance and base reserve of the body, but there are no adequate data for the effect of increased dissolved solids in water on the blood chemistry to support speculation at this time. A regression equation combining the altitude and the specific conductivity of the water of these cities predicts the death rate from cancer to the 95% level of confidence, within .+-. 17.4% for males and .+-. 17.0% for females. Of the total variance in death rates, 36% is associated with these 2 parameters combined.