Toxic metal mixtures in private well water and increased risk for preterm birth in North Carolina

Table 3 Summary of results from quantile-based g-computation modeling

Model	Interpretation	Crude OR (95% CI)	Adjusted* OR (95% CI)	Adjusted weights		Adjusted coefficients
A) Standard quantile-based g-computation	Increasing all metals by one quartile (ppb)	0.97 (0.96,0.98)	1.00 (0.99,1.02)	Cadmium Lead Chromium Zinc Arsenic Copper Manganese	0.52 0.33 0.15 -0.37 -0.36 -0.20 -0.07	Cadmium Lead Chromium Zinc Arsenic Copper Manganese	0.015 0.009 0.004 -0.009 -0.009 -0.005 -0.002
B) Positive direction partial effects quantile-based g-computation	Increasing all metals that were in the positive direction in the training set by one quartile (ppb)	1.00 (0.99,1.01)	1.02 (1.01,1.03)	Cadmium Lead Chromium	0.65 0.20 0.15	Cadmium Lead Chromium	0.012 0.004 0.003
C) Negative direction partial effects quantile-based g-computation	Increasing all metals that were in the negative direction in the training set by one quartile (ppb)	0.96 (0.95,0.97)	0.99 (0.97,1.00)	Zinc Arsenic Copper Manganese	-0.51 -0.43 -0.05 1.00	Zinc Arsenic Copper Manganese	-0.007 -0.006 -0.001 0.001

Model A includes all metals in the exposure matrix. Model B and C contain metals that were associated in the positive direction and the negative direction, respectively in the training data set in the quantile-based g computation partial effect modelling. The weights, that sum to 1 or -1 for each direction, represent the proportion of each metal’s contribution to the partial effect in the negative (weight < 0) or positive (weight > 0) direction. The adjusted coefficient for each metal represents the independent effect size for that metal. Models were adjusted for smoking, age, race/ethnicity, education, season of conception, tract-level poverty, tract-level nitrates and nitrites

ISSN: 1476-069X