Aortic Strain -Threshold Analysis of Non-Cancer Outcome from Arsenic Exposure

Abstract

Background: While arsenic has been shown in many studies to behave as a threshold carcinogen, analyses of non-cancer outcomes (e.g., cardiovascular) have generally been presented only using linear models. Alternative models have received little consideration. We choose here to examine aortic strain data, a known risk factor for hypertension and cardiovascular disease, on arsenic-exposed workers with alternative analytic modeling.Method: A previously published study (Karakulak et al., 2016) of aortic strain measurements in arsenic-exposed workers had demonstrated that aortic strain was markedly reduced among workers with high levels of arsenic exposure (urinary arsenic in ug/L). The original analysis was limited to a linear model without model validation. We have analyzed the data using alternative models.Results: Examination of the linear analysis and its residuals demonstrated significant heteroscedascity with unequal variances, particularly at urinary levels of > 160 ug arsenic per liter (ug/L). The test of equal variances from samples with >160 and samples <= 160 demonstrated inequality (p < 0.0001). The better alternative model was a step-function with a mean aortic strain of 11.3% for those with urinary arsenic <= 160 ug/L and with a mean aortic strain of 5.3% for those with urinary arsenic > 160 ug/L. A further stratified analysis with bins of 40 ug/L showed no effect on aortic strain for the urinary arsenic range of <= 160 ug/L and, separately, no effect for the urinary arsenic range of > 160 ug/L.Conclusion: Aortic elasticity parameters show an adverse effect from arsenic exposure only at urinary exposure levels of > 160 ug/L and not below, indicating a step-function rather than a linear function. Restriction of analyses to a linear no-threshold model may have blinded the finding of a more useful demonstration of a dose-response pattern that may offer different toxicological interpretations.