Table 3.

Associations [HRs (95% CIs)] between serum zinc levels and incident ischemic stroke stratified by pre-specified factors, the REGARDS Trace Element Study.^a,b

	Serum zinc levels [mean (SD)]	No. of cases	No. of participants in sub-cohort	Quartiles of serum zinc levels (μg/dL)				P for trend
				Quartile 1 (<105.98)	Quartile 2 (105.98-122.78)	Quartile 3 (122.78-144.59)	Quartile 4 (≥144.59)
Sex
Female	123.57 (26.37)	318	1,283	1 (Ref.)	0.81 (0.59-1.11)	0.75 (0.54-1.03)	0.58 (0.41-0.84)	<0.01
Male	124.32 (26.72)	342	1,063	1 (Ref.)	1.36 (0.97-1.90)	1.26 (0.90-1.75)	1.08 (0.78-1.51)	0.92
P for interaction	--	--	--	0.03
Race
African American	121.81 (26.90)	270	935	1 (Ref.)	1.08 (0.75-1.55)	1.13 (0.79-1.61)	0.74 (0.50-1.09)	0.04
Caucasian	125.31 (26.20)	390	1,411	1 (Ref.)	0.91 (0.68-1.22)	0.83 (0.62-1.12)	0.78 (0.57-1.06)	0.33
P for interaction	--	--	--	0.35
Region
Stroke Buckle	124.08 (27.02)	143	499	1 (Ref.)	0.85 (0.50-1.46)	0.85 (0.48-1.48)	0.74 (0.43-1.28)	0.41
The rest of Stroke Belt	122.98 (26.65)	231	803	1 (Ref.)	0.80 (0.55-1.17)	0.83 (0.57-1.21)	0.60 (0.39-0.91)	0.015
Non-Stroke-Belt region	124.56 (26.19)	286	1,044	1 (Ref.)	1.19 (0.84-1.70)	1.01 (0.71-1.43)	1.00 (0.69-1.44)	0.97
P for interaction	--	--	--	0.16

Abbreviations: CI, confidence interval; HR, hazard ratio; No., number; Ref., referent; REGARDS, REasons for Geographic and Racial Differences in Stroke; SD, standard deviation.

^aAll models were constructed using Barlow-weighted Cox proportional hazards regression analysis with adjustment for covariates in model 3 Table 2. P for trend and P for interaction were examined by using the continuous variable of zinc concentrations with values over 95^th percentile excluded.

^bQuartiles of serum zinc levels were calculated based on the random sub-cohort (n=2,346).