. 2022 Jun 22;130(6):067007. doi: 10.1289/EHP10393

Table 7.

Potential effect modification of the key characteristics on the relation between PFOS exposure and all-cause mortality in the NHANES follow-up study from 1999–2014 to 31 December 2015 ( $lowercase italic n equals 11,747$ ).

Characteristics	Deaths/observations ( $n / N$ )		Adjusted HR (95% CI) $greater than or equal to 17.1$ vs. $less than 7.9 nanograms per milliliter$ (tertile 3 vs. tertile 1)	$p$ for interaction^a
Characteristics	$less than 7.9 nanograms per milliliter$ (tertile 1)	$greater than or equal to 17.1 nanograms per milliliter$ (tertile 3)		$p$ for interaction^a
Age (y)				0.064
$less than 50$	47/2,606	45/1,702	1.03 (0.46, 2.23)	—
$greater than or equal to 50$	152/1,280	635/2,242	1.35 (1.02, 1.78)	—
Sex				$less than 0.001$
Male	102/1,382	418/2,381	1.23 (0.85, 1.78)	—
Female	97/2,504	262/1,563	1.98 (1.37, 2.86)	—
Race/ethnicity				0.556
Non-Hispanic White	112/1,527	419/1,953	1.72 (1.23, 2.40)	—
Non-Hispanic Black	32/674	129/992	2.33 (1.38, 3.91)	—
Mexican American	35/821	94/610	2.16 (1.07, 4.36)	—
Other	20/864	38/389	1.59 (0.69, 3.65)	—
Smoking status				0.127
Never smoke	83/2,259	278/2,012	1.99 (1.34, 2.95)	—
Past or current smoker	116/1,627	402/1,932	1.67 (1.18, 2.36)	—
Hypertension				0.299
No	135/3,356	387/2,997	1.57 (1.11, 2.23)	—
Yes	64/530	293/947	1.59 (1.08, 2.36)	—
Diabetes				0.011
No	139/3,477	530/3,488	1.76 (1.30, 2.39)	—
Yes	60/409	150/456	1.13 (0.67, 1.88)	—
Obesity				0.981
No	128/2,446	461/2,620	1.86 (1.33, 2.60)	—
Yes	71/1,440	219/1,324	1.78 (1.15, 2.73)	—
Renal function [ $Ccr (mL / min)$ ]				0.291
Normal ( $greater than or equal to 70$ )	99/3,472	335/3,157	1.55 (0.89, 2.67)	—
Impaired ( $less than 70$ )	100/414	345/787	0.72 (0.34, 1.56)	—

Note: HRs were estimated using Cox proportional hazards models and were weight adjusted using NHANES-specified sampling weights. Other adjusted variables include PFAS excluding PFOS (categorized into three groups based on k-means algorithm), sex, age, race/ethnicity, education, smoking status, physical activity, hypertension, healthy eating index, Ccr, serum total cholesterol, and serum cotinine. Missing data on covariates were processed using multiple imputation algorithm. Samples with test values below the LOQ were substituted with the value of the LOQ divided by the square root of 2. —, not applicable; Ccr, creatinine clearance rate; CI, confidence interval; HR, hazard ratio; LOQ, limit of quantification; NHANES, National Health and Nutrition Examination Survey; PFAS, per- and polyfluoroalkyl substances; PFOS, perfluorooctane sulfonic acid.

^{^a}

$p$ for interaction was calculated by log-likelihood ratio test.