Aortic Elasticity and Arsenic Exposure: A Step Function rather than a Linear Function

While the dose-response relationship for the carcinogenic effects of arsenic exposure indicates nonlinearity with increases only above about 150 μg/L arsenic in drinking water, similar analyses of noncarcinogenic effects of arsenic exposure remain to be conducted. We present here an alternative analysis of data on a measure of aortic elasticity, a risk factor for hypertension, and its relationship to urinary arsenic levels. An occupational health study from Ankara, Turkey by Karakulak et al. compared urinary arsenic levels and a measure of aortic elasticity (specifically, aortic strain) in workers with a linear no-threshold model.  We have examined these data with three alternative models-a fitted step-function, a stratified, and a weighted linear regression model. Discontinuity within the data revealed two subsets of data, one for workers with urinary arsenic levels ≤ 160 μg/L whose mean aortic strain level was 11.3% and one for workers with arsenic levels > 160 μg/L whose mean aortic stain level was 5.33 % (p < 0.0001). Several alternative models were examined that indicated the best model to be the threshold model with a threshold at a urinary arsenic level of 160 μg/L. Observation of a discontinuity in the data revealed their better fit to a threshold model (at a urinary arsenic level of 160 μg/L) than to a linear-no threshold model.  Examinations with alternative models are recommended for studies of arsenic and hypertension and possibly other noncarcinogenic effects.