Table 3.

Average predicted values of systolic blood pressure at percentiles of blood lead $\left(\overline{p}\left(y|x_q\right)\right)$, marginal difference in predicted systolic blood pressure equivalent to 25 percentile change in the total population $\left(\overline{p}\left(y\mid x_{q1}\right)-\overline{p}\left(y\mid x_{q0}\right)\right.$, where exposure contrast from 50^th to 25^th percentile=0.47 ug/dL and 75^th to 50^thpercentile=0.78 ug/dL), and their model fit by analytic approach in NHANES 1999–2016 (n=23,113)^a

	Regression models
	Linear	Robust	Log-transformed	Quadratic
Crude
AIC	−296581	−296539	−297174	−296814
$\overline{p}\left(y\mid x_{25}\right)$	115.69 (115.41, 115.96)	115.83 (115.56, 116.09)	114.8 (114.55, 115.05)	115.19 (114.86, 115.44)
$\overline{p}\left(y\mid x_{50}\right)$	116.20 (115.98, 116.42)	116.47 (116.26, 116.69)	116.71 (116.49, 116.93)	116.09 (115.87, 116.31)
$\overline{p}\left(y\mid x_{75}\right)$	117.01 (116.77, 117.28)	117.51 (117.27, 117.75)	118.69 (118.41, 118.97)	117.51 (117.26, 117.89)
$\overline{p}\left(y\mid x_{50}\right)-\overline{p}\left(y\mid x_{25}\right)$	0.51 (0.41, 0.63)	0.64 (0.53, 0.76)	1.9 (1.8, 2.1)	0.90 (0.79, 1.13)
$\overline{p}\left(y\mid x_{75}\right)-\overline{p}\left(y\mid x_{50}\right)$	0.82 (0.65, 1.0)	1.0 (0.86, 1.2)	2.0 (1.8, 2.1)	1.4 (1.3, 1.8)
Adjusted for age, gender, race/ethnicity, education, NHANES cycle, waist circumference
AIC	−301635	−301614	−301648	−301635
$\overline{p}\left(y\mid x_{25}\right)$	118.22 (117.96, 118.48)	118.48 (118.24, 118.73)	117.99 (117.69, 118.29)	118.15 (117.84, 118.43)
$\overline{p}\left(y\mid x_{50}\right)$	118.28 (118.04, 118.51)	118.54 (118.32, 118.77)	118.29 (118.07, 118.52)	118.25 (118.01, 118.48)
$\overline{p}\left(y\mid x_{75}\right)$	118.37 (118.14, 118.60)	118.63 (118.42, 118.86)	118.60 (118.34, 118.88)	118.41 (118.18, 118.66)
$\overline{p}\left(y\mid x_{50}\right)-\overline{p}\left(y\mid x_{25}\right)$	0.057 (−0.0061, 0.12)	0.058 (0.0052, 0.11)	0.31 (0.13, 0.48)	0.10 (0.011, 0.22)
$\overline{p}\left(y\mid x_{75}\right)-\overline{p}\left(y\mid x_{50}\right)$	0.092 (−0.0097, 0.20)	0.093 (0.0083, 0.18)	0.31 (0.13, 0.49)	0.16 (0.017, 0.34)

Abbreviations: linear, general linear model (GLM) with untransformed X; robust, robust linear model with untransformed X; log-transformed, GLM with log-transformed X; quadratic, GLM with quadratic term for X; AIC, Akaike information criterion.

^aAdjusted for age, gender, race/ethnicity, education, NHANES cycle, waist circumference and 95% confidence intervals bootstrapped.