P1‐120: Higher groundwater vanadium levels are associated with better cognitive functioning in a sample of cases with diabetes mellitus: A Project FRONTIER study

association, where hypertension appears harmful at midlife but not in late life. This study aims to reproduce this pattern observed across studies within a single study and investigates whether two potential explanations, selection bias and duration of hypertension, can account for this pattern. Methods: We used data from 1284 participants from the Normative Aging Study who were under age 45 and non-hypertensive at enrollment. These participants were assessed for hypertension every 3-5 years since 1963, and 758 completed a battery of cognitive tests between 1993 and 2005. We modeled the association between having a history of hypertension at different ages from 45 to 65 and mean cognitive test z-score at end of follow-up from a series of point exposure studies. To investigate the impact of selection bias from drop-out we compared analyses using inverse probability weights for censoring to un-weighted analysis. To investigate the impact of duration of hypertension, we stratified theweighted analyses by duration of hypertension. Results: We found within our study sample the same the age-dependent association observed across studies, finding strong adverse associations with hypertension at midlife and little association in late-life. Weighting for censoring did not eliminate the age-dependent pattern, but produced stronger adverse associations throughout. Restricting to participants with longer durations of hypertension, the age-dependent pattern disappeared or reversed direction, such that we observed similar or stronger adverse associationswith hypertension at later ages.Conclusions: Selection bias cannot account for the age-dependent association between hypertension and cognition, but may contribute to protective associations observed in other studies of older populations. The proportion of persons with short duration of hypertension at follow-up accounts, in part, for the overall age-dependent pattern of association.