Secondhand smoke exposure and higher blood pressure in children and adolescents participating in NHANES

Abstract

We assessed the relationship between acute and intermittent secondhand tobacco smoke (SHS) exposure with child and adolescent blood pressure (BP). We analyzed cross-sectional data from 3579 children and adolescents aged 8–17 participating in the National Health and Nutrition Examination Survey (NHANES) collected between 2007 and 2012, with SHS exposure assessed via serum cotinine (a biomarker for acute exposures) and urine NNAL (4-(methylnitrosamino)-1-(3-pyridyl)-1-butanol, a biomarker for intermittent exposures). BP percentiles and z-scores were calculated according to the 2017 guidelines established by the American Academy of Pediatrics. We used weighted linear regression accounting for the complex sampling weights from NHANES and adjusting for socio-demographic and clinical characteristics. Overall, 56% of the children were non-Hispanic white with a mean age of 12.6 years. There was approximately equal representation of boys and girls. Approximately 15.9% of participants lived in homes where smoking was present. In adjusted models, an interquartile range (IQR) increase in urinary NNAL was associated with 0.099 (95% CI: 0.033, 0.16) higher diastolic blood pressure (DBP) z-score, and with a 0.094 (95% CI: 0.011, 0.18) higher systolic blood pressure (SBP) z-score. For serum cotinine, an IQR increase was associated with 0.097 (95% CI: 0.020, 0.17) higher DBP z-scores, but was not significantly associated with SBP z-scores. The associations of cotinine and NNAL with BP also differed by sex. Our findings provide the first characterization of the relationship between a major tobacco-specific metabolite, NNAL, and BP z-scores in a nationally representative population of US children.

Introduction:

In the United States, up to 24 million nonsmoking children and adolescents are exposed to secondhand tobacco smoke (SHS)¹. While cigarette smoking is on the decline in the US, in 2016 smoking prevalence was still 15.5%². Data from a nationally representative sample suggests that up to 4 in 10 children aged 3 – 11 years, and 1 in 3 adolescents aged 12 to 19 years, are exposed to SHS, as indicated by serum cotinine levels (> 0.05 ng/mL)¹. SHS is a complex mixture of hundreds of gaseous and particulate toxicants including nicotine, tobacco-specific nitrosamines, and metals. Children are especially vulnerable to SHS; not only are they potentially exposed to SHS in their home and social environments, but normal physiological characteristics of childhood lung development make them particularly susceptible to health effects of SHS and other toxicant exposure³. Exposure to SHS is well-documented to cause respiratory outcomes clinically apparent in childhood, including asthma and bronchiolitis.^4,5 Additional cardiovascular consequences to SHS exposure occur as well – these may be occult but could persist into adulthood.

A 2016 Scientific Statement by the American Heart Association³ underscored the significant detrimental cardiovascular effects associated with exposure to SHS in children, which included increased risk of premature atherosclerosis, impaired cardiac autonomic function and changes in heart rate variability. Functionally, SHS in adolescents (verified using serum cotinine) has been linked to decreased aortic elasticity⁶, and exposure to parental smoking during childhood is linked to decreased flow-mediated dilatation even 20 years later in adulthood⁷. Indeed, developmental toxicant exposures are established modifiable factors contributing to altered blood pressure (BP) in childhood and adolescence^8,9. Childhood and adolescence are critically important developmental windows, as elevated BP as early as 7 years of age is associated with altered life-long trajectories for hypertension^10,11.

Few previous studies have examined the effects of SHS on BP in children, and the limited studies have yielded mixed findings^12–14. In this study, we examined the role of SHS and BP among 3579 children and adolescents using a nationally representative sample. Recent (e.g. previous 2–4 days) and intermittent (e.g. previous 1 month or longer) exposures to SHS are measured using serum cotinine and urinary 4-(methylnitrosamino)-1-(3-pyridyl)-butanol (NNAL) biomarkers, respectively. We assess the relationship between cotinine and NNAL with BP.

Patients and Methods:

NHANES data collection:

In 2019, we analyzed publicly available data from the National Health and Nutrition Examination Survey (NHANES)⁴. NHANES is a recurring cross-sectional survey of the noninstitutionalized civilian US population, who live in the 50 states and the District of Columbia, with details available elsewhere¹⁵. In our analysis, we combined data collected from 3 cycles of NHANES (2007–2008, 2009–2010, and 2011–2012). These cycles were chosen because they serve as the most recent data; in 2015–2016, NNAL was not collected in NHANES; and in 2013–2014, the technology for detecting NNAL changed compared with the previous cycles. Thus, these three cycles were chosen to maintain consistency. The study sample consisted of children aged 8 years and older, but less than 18 years old who had no missing information on BP and SHS biomarker data. We also excluded those who reported usage of tobacco/nicotine products in the past 5 days (n=151). We further excluded those with any missing data in one of the following variables: age, sex, race/ethnicity, height, waist circumference, the ratio of family income to poverty (PIR), serum lead and cadmium concentrations, and urine albumin-creatinine ratio. In total, 308 participants were excluded due to missing covariates. Family income to poverty ratio was missing in 272 participants, waist circumference was missing in 36 participants and lead and cadmium was missing in 5 participants each. Our final sample size was 3579 children. Written informed consent is obtained from a parent or guardian for participants who are minors. The NHANES procedures and protocol is approved by the CDC/NCHS Research Ethics Review Board¹⁶. This study was exempted from review by the Icahn School of Medicine at Mount Sinai’s Institutional Review Board (#1702145).

Outcomes:

Systolic and diastolic blood pressures (SBP and DBP) were measured by the auscultatory method for children ages 8 years and older using three readings and an optional fourth reading¹⁷. We used the average of all available readings. SBP and DBP percentiles were calculated using a previously described method¹⁸, based on the 2017 Clinical Practice Guidelines from the American Academy of Pediatrics¹⁹. Briefly, the approach uses natural spline quantile regression to estimate sex-specific BP norms based on data from 49,967 normal weighted children using five input variables: sex, age in months, height in cm, and SBP and DBP in mmHg. We calculated BP percentiles for the study sample using the recommended published SAS macro¹⁸. The BP percentiles were transformed into z-scores using inverse-normal methods. This was done because unlike percentiles, z-scores are on an equal interval scale. BP z-scores were used in subsequent regression analyses. We calculated whether each child could be categorized as belonging in the hypertensive range. For participants less than 13 years of age, the hypertensive range is defined as SBP ≥ 120 mmHg or SBP percentile at or above the 90^th percentile, or DBP ≥ 80 mmHg or DBP percentile at or above the 90^th percentile. For participants aged 13 years of age or older, the hypertensive range is defined as SBP ≥ 120 mmHg or DBP ≥ 80 mmHg.

Exposures:

Exposure to SHS and tobacco products was assessed by measuring serum cotinine and urinary total NNAL. Cotinine is a major metabolite of nicotine with a half-life in plasma of 15–20 hours, and is regarded as a surrogate marker for recent (e.g. 2–4 days^20,21) SHS exposure²². NNAL is a major metabolite of the tobacco-specific nitrosamine NNK (4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone) with a half-life of 10–16 days in urine. Because NNK is rapidly metabolized to form NNAL, NNAL provides a biomarker for exposure to second hand smoke over a longer period of time (e.g. a month or longer) than cotinine²⁰.

For laboratory measures, serum and urine specimens were stored in vials at appropriate frozen conditions (at or below −20°C), and were shipped to the National Center for Environmental Health for testing. Serum cotinine was measured using an isotope-dilution high-performance liquid chromatography/atmospheric pressure chemical ionization tandem mass spectrometric (ID HPLC-APCI MS/MS) method²². NNAL was measured using liquid chromatography linked to tandem mass spectrometry using methods previously described elsewhere²³. The limit of detection (LOD) was 0.015 ng/mL for cotinine and 0.0006 ng/mL for NNAL.

We also extracted five additional biomarker variables: albumin and creatinine measured in urine, the first urine albumin-creatinine ratio, cadmium and lead measured in blood. Urinary albumin was assessed using a solid-phase fluorescent immunoassay²⁴, and urinary creatinine was assessed using the Roche/Hitachi Modular P Chemistry Analyzer, using methods described elsewhere²⁴. The first urine albumin-creatinine ratio was calculated as urine albumin measured in mg/dL divided by urine creatinine measured in g/dL²⁴. Cadmium and lead are major components of cigarette smoke²⁵ and SHS, and can affect children’s BP^26–28. Whole blood lead and cadmium concentrations are measured using inductively coupled plasma mass spectrometry²⁹. Any lead or cadmium measures under the limit of detection were replaced by the analyte’s corresponding detection limit divided by the square root of two²⁹.

Covariates:

Participant characteristics, such as age, race/ethnicity (non-Hispanic white, non-Hispanic black, Hispanic, other), sex, income and education level were collected using questionnaires. Socioeconomic status was assessed using an index for the ratio of family income to poverty, in which the poverty threshold is defined using the Department of Health and Human Services’ poverty guidelines and are specific to family size, survey year and state³⁰. Body measurements, such as height, weight, and waist circumference were measured during a physical examination. Assessment of SHS was also evaluated from the household smoker questionnaire regarding the smoking status of the household members. Analyses including urinary NNAL also adjusted for urinary creatinine to account for changes in hydration status.

Statistical analyses:

We linked demographic data, laboratory data and physical exam data using a unique survey participant identifier. Our analyses accounted for the NHANES complex survey design by using SAS survey procedures with sampling strata, cluster, weights, and domain parameters, thus the results can be considered representative of the US population³¹. We reported the weighted frequency for categorical variables and weighted means or geometric means for continuous variables. All biomarker variables were natural log transformed due to their skewed distributions. Differences in NNAL and cotinine concentrations by different covariates were assessed using survey linear regression models.

We used weighted linear regressions to study the association between cotinine or NNAL with SBP and DBP z-scores, as well as continuous SBP and DBP, and belonging to the hypertensive range, in four separate models. Our models for BP z-scores adjusted for race/ethnicity (non-Hispanic white, non-Hispanic black, Hispanic, other), family poverty to income ratio, waist circumference, cadmium (ug/L), lead (ug/dL), first albumin creatinine ratio (mg/g), and urinary creatinine (mg/dL). In addition, we conducted sex-stratified analysis. In models using continuous BP in mmHg as the outcome, we additionally adjusted for age, sex and height. In all models with NNAL as the main predictor, urine creatinine was included in the model as a covariate³². In order to aid interpretability of the models, we back-transform the beta coefficients to report the adjusted BP change associated with an interquartile range (IQR) difference in NNAL or cotinine.

Below LOD NNAL and cotinine concentrations (38.8% and 26.5%, respectively) were replaced by the corresponding detection limit divided by the square root of two²⁹, which is a common substitution method and implemented by the CDC in NHANES²². We also conducted a sensitivity analysis using a different method for dealing with concentrations below LOD. Our sensitivity analysis used an approximate Bayesian framework for multiple imputation³³, to impute NNAL and cotinine concentrations below LOD. This was implemented using the R package “Missing Data Imputation and Model Checking”³³.

We used weighted logistic regression to study the association between SHS exposure and being in the hypertensive range. In order to increase interpretability, SHS exposure was categorized into cotinine (NNAL) levels below the detection limit, detectable cotinine (NNAL) levels less than or equal to the survey-weighted 75^th percentile of exposure, and cotinine (NNAL) levels greater than the survey-weighted 75^th percentile of exposure. Weighted logistic regression models were adjusted for age, sex, height, race/ethnicity, family poverty to income ratio, waist circumference, cadmium (ug/L), lead (ug/dL), first albumin creatinine ratio (mg/g), and urinary creatinine (mg/dL, for NNAL model only).

Results:

Participant characteristics:

In Table 1, we present distributions of socio-demographics and clinical risk factors of the 3579 children. Overall, 56% of the children were non-Hispanic white with a mean age of 12.6 years. There were approximately equal representation of boys and girls. The geometric means of NNAL and cotinine were 1.62 pg/mL (interquartile range (IQR):0.40–4.28 pg/mL) and 0.06 ng/mL (IQR: 0.01–0.16 ng/mL), respectively. Approximately 13.9% of children live in homes where smoking was present. Table 2 presents the survey-weighted summary of NNAL and cotinine levels across demographic groups. The geometric means of NNAL and cotinine were 16 times and 24 times higher, respectively, among children living in a home where smoking was present versus those living in a nonsmoking home. We observed significant sex-specific differences for cotinine (p=0.015) (higher among boys) but not for NNAL. Both NNAL and cotinine levels differed significantly by race/ethnicity (p<0.0001), and presence of smoking at home (p<0.0001).

Table 1.

Summary of sociodemographic and clinical covariates.

Covariate	Summary measure
	Weighted frequency (%)
Sex
Male	50.3
Female	49.7
Race/Ethnicity
Other	7.6
Hispanic	21.9
non-Hispanic black	14.2
non-Hispanic white	56.3
Smoking at home
No	86.1
Yes	13.9
Hypertensive range
No	87.6
Yes	12.4
	Median (interquartile range)
Age (years)	12.1 (9.7–14.4)
Family income to poverty ratio	2.3 (1.1–4.1)
SBP percentile	46.8 (24.1–69.4)
DBP percentile	27.6 (11.0–54.5)
SBP (mmHg)	105.1 (98.4–111.7)
DBP (mmHg)	57.2 (49.5–65.1)
Waist circumference (cm)	73.4 (65.5–82.7)
Height (cm)	157.1 (145.1–166.5)
Urinary NNAL (pg/mL)	1.02 (0.40–4.28)
Urinary creatinine (mg/dL)	112 (63.9–161.9)
Serum cotinine (ng/mL)	0.03 (0.01–0.16)
Cadmium (ug/L)	0.13 (0.11–0.19)
Lead (ug/dL)	0.66 (0.48–0.95)
First albumin creatinine ratio (mg/g)	8.57 (5.29–16.4)

SBP denotes systolic blood pressure; DBP denotes diastolic blood pressure. Interquartile range denotes the interval covering the 25th to 75th percentile. Weighted frequencies and summary statistics were calculated using SAS survey procedures that took into account the NHANES complex survey design.

Table 2.

Survey-weighted summary of NNAL and cotinine levels across demographic groups.

Covariates	NNAL (pg/mL)		Cotinine (ng/mL)
	Geometric mean	p-value	Geometric mean	p-value
Sex
Boys	1.8	0.37	0.07	0.015*
Girls	1.5		0.05
Race/Ethnicity
Other	1.2	0.0049*	0.05	0.08
Hispanic	1.1	0.0003*	0.03	0.0006*
non-Hispanic black	2.6	0.08	0.12	0.0001*
non-Hispanic white	1.8		0.06
Smoking at home
No	1.1	<.0001*	0.04	<.0001*
Yes	15.9		0.98
Hypertensive range
No	1.6	0.0168*	0.06	0.0018*
Yes	2.1		0.09

^*denotes p-value < 0.05 significance level using t-tests.

Association between NNAL and BP:

We assessed the associations between NNAL and DBP and SBP z-scores, as well as sex-stratified analyses (Figure 1). In adjusted models, NNAL was associated with DBP z-scores: an IQR increase in urinary NNAL was associated with a 0.099 (95% CI: 0.033, 0.16) higher DBP z-score, and with a 0.094 (95% CI: 0.011, 0.18) higher SBP z-score. Among boys, an IQR increase in NNAL was associated with a 0.16 (95% CI: 0.079, 0.23) higher DBP z-score, but as not associated with SBP z-scores. Meanwhile, among girls, an IQR increase in NNAL was associated with a 0.14 (95% CI: 0.018, 0.26) higher SBP z-score, but was not associated with DBP z-scores.

Figure 1.

Adjusted associations between SHS biomarkers, as measured by serum cotinine and urinary 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanol (NNAL), and blood pressure among children in NHANES 2007–2012. DBP=diastolic blood pressure in mmHg, zDBP =diastolic blood pressure z-score, zDBP_M = diastolic blood pressure z-score among males, SBP=systolic blood pressure in mmHg, zSBP =systolic blood pressure z-score, zSBP_F = systolic blood pressure percentile among females.

In adjusted models, NNAL was significantly positively associated with DBP (in mmHg): an IQR increase in NNAL was associated with a 1.14 (95% CI: 0.39–1.90) mmHg higher DBP, but was not associated with SBP (in mmHg). Furthermore, for adjusted models with DBP (in mmHg), we found a significant interaction between NNAL and child sex (p = 0.02), but this interaction was not significant for SBP (in mmHg). Sensitivity analysis using a multiple imputation framework to impute NNAL concentrations below the LOD showed similar results.

Among children in the top quartile of NNAL exposures, 16.8% were in the hypertensive range, while among children with undetectable NNAL, 10.3% were in the hypertensive range. The odds of being in the hypertensive range was 1.966 (95% CI: 1.31, 2.951) times greater among children in the top quartile of NNAL exposures compared to those with undetectable NNAL.

Association between cotinine and BP:

Similarly, results from our adjusted models (Figure 1) showed that cotinine was associated with DBP z-scores. An IQR increase was associated with a 0.097 (95% CI: 0.020, 0.17) higher DBP z-score, but was not associated with SBP z-scores. Furthermore, among boys, an IQR increase in cotinine was associated with a 0.16 (95% CI: 0.090, 0.23) increase in DBP z-scores, but was not significant among girls. Furthermore, an IQR increase in cotinine was associated with a 1.19 (95% CI: 0.35–2.03) mmHg higher DBP, but was not associated with SBP (in mmHg). We did not find a significant interaction between cotinine and child sex, for models of SBP (in mmHg) or DBP (in mmHg). Sensitivity analysis using multiple imputation to impute cotinine concentrations below LOD showed similar associations.

Among children in the top quartile of cotinine exposures, 16.4% of children were in the hypertensive range, while among children with undetectable cotinine, 9.1% were in the hypertensive range. The odds of being in the hypertensive range was 1.979 (95% CI: 1.435, 2.73) times greater among children with the top quartile of cotinine exposures, compared to those with undetectable cotinine. We did not find a significant interaction between child sex and cotinine (or NNAL) and risk of being in the hypertensive range in adjusted models.

Because education is another indicator for socio-economic status, we conducted secondary analyses to include education (<high school, high school, >high school) as another covariate. This reduced sample sizes from n = 3579 to n = 2287. The sensitivity analyses confirmed associations between cotinine and higher SBP z-scores, but we did not find associations between NNAL and SBP or DBP z-scores

Discussion:

Our findings demonstrate higher DBP in association with SHS exposure (assessed via urinary NNAL and serum cotinine concentrations) during childhood, when adjusting for relevant socio-demographic and clinical characteristics in a nationally representative sample. In our analysis, we found significant associations between SHS and DBP and SBP. Notably, SHS (NNAL or cotinine) was associated with DBP z-scores among boys, but not among girls. In addition, NNAL was associated with SBP z-scores both overall and among girls, though cotinine was not associated with SBP. Together these results suggest a sex-specific relationship that could be explored in future analysis. Importantly, this study assessed childhood SHS using serum cotinine and urinary NNAL biomarkers, which is an important distinction from several other studies that have reported links between parental smoking and cardiovascular function. In prior studies, parental smoking was assessed using questionnaire data, which may be vulnerable to responder bias and lack of accounting for other sources of smoke exposure.

By using both cotinine and NNAL biomarkers, we were able to assess how both acute and intermittent exposure to SHS affects BP, which might partially explain the discrepancies in the SBP-raising effect observed between NNAL and cotinine. NNAL is a urinary metabolite specific for tobacco smoke, and is a major metabolite of the tobacco-specific nitrosamine called 4-(methylnitrosamino)-1-(3)pyridyl-1-butanone (NNK). Unlike cotinine, which has a half-life of 16–24 hours, and is sensitive to tobacco exposure over the past 2–4 days, NNAL has a much longer half-life of 10–16 days and is sensitive to tobacco exposure over the past month or longer. NNAL is considered to be effective for detecting intermittent SHS exposure, as NHANES is a cross-sectional study and children may not be consistently exposed to SHS that would yield detection by cotinine which has a very short detection window. While previous studies^12,13,34 have mainly focused on using cotinine as a biomarker of SHS exposure, recent work has found that urine NNAL detects a higher prevalence of SHS exposure than using cotinine, perhaps due to NNAL being able to detect intermittent exposure²⁰.

A series of recent studies^12,13 found a relationship between hair nicotine and higher BP in urban young children; in the same cohort, hair nicotine was also associated with other markers of cardiovascular diseases, such as soluble I-CAM. A study among the German preschool population reported a significant association between parental smoking and childhood BP¹⁴. Meanwhile, a study found associations between exposure to parental smoking in German adolescents and increased exercise SBP, but not resting BP³⁵. Our findings are consistent with several other studies which have found an association between SHS and DBP^12,14. Increases in DBP are also clinically relevant, as elevated DBP may serve as an early antecedent to prehypertension, Stage I and II hypertension.

This study had limitations. Because NHANES is a cross-sectional study we are unable to draw conclusions regarding temporality and causality of exposure and outcome. Additionally, as cotinine and NNAL have limited half-lives, we had some limitations in assessing the effects of timing and prolonged exposure to SHS on BP. Another limitation is that we lack information on the pubertal status of each participant in the cohort. SHS is associated with age of pubertal onset in girls³⁶, and puberty is irrevocably linked to increasing blood pressure related to somatic growth during that period. Future work is needed to assess whether the effects of SHS on BP differs by pubertal status. There is also potential for unmeasured confounding. To explore this, we included education as an additional variable in the models for sensitivity analysis, as education can be another indicator for socio-economic status. While we confirmed some of our main findings, we note there were differences, potentially because inclusion of the education covariate significantly reduced our sample size. Further, while we excluded participants who reported primary use of tobacco/nicotine products, as we wanted to focus on SHS, there is a possibility of response bias. Furthermore, it is possible that some participants are primary users of e-cigarettes. NHANES did not collect self-reported e-cigarette use during these survey years; this information was only collected starting with the 2013–2014 cycle. In addition to being biomarkers of SHS, NNAL and cotinine are also biomarkers for primary e-cigarette use³⁷. The NHANES survey years in our analysis (2007–2012), overlaps with when e-cigarettes first came on the market in 2007 and the time period that they were gaining popularity. However, the use of NHANES during these survey years is still valuable, as it provides a large sample size of children and adolescents, and the ability to assess both acute and intermittent SHS exposure on BP.

Our recommendations for future studies are in line with those identified by the American Heart Association³ which are to determine: 1) which effects of SHS are reversible, 2) which effects are persistent, and 3) if there is tipping point, after which cardiovascular deficits remain even after a child stops being exposed. Furthermore, it will be important to assess the impact of exposure timing – and whether there are critical time windows of vulnerability to SHS exposure, and if these sensitive time windows are in utero, in early childhood, late childhood or adolescence. Importantly, future longitudinal studies are needed to determine if SHS-associated elevated BP is an intermediary outcome for long-term cardiovascular dysfunction.

Our paper shows a step towards assessing the impact of tobacco smoke exposure on BP in children. Currently, cigarette smoking is now on the decline in the US, at 15.5% in 2016 as compared with 20.9% in 2005². However, globally, the prevalence of cigarette smoking is still high - in 2015, over 1.1 billion people smoked tobacco³⁸. Furthermore, tobacco smoking prevalence is increasing in some regions, such as the World Health Organization Mediterranean Region and the African Region³⁸. The important work of studying smoke exposure on cardiovascular outcomes will also need to continue in light of new technologies and policies (e.g. legalization of marijuana). Furthermore, the recent drastic increased use of e-cigarettes among adolescents (estimated use by over 3 million high school students in 2018³⁹), is an important area for future research and a growing public health concern.

Conclusion:

Our findings provide strong evidence for the relationship between tobacco smoke exposure, as measured by cotinine or NNAL, and increased DBP in a nationally representative population of US children. Adolescents, parents and caregivers alike should be aware that exposure to tobacco smoke may have a negative impact on children’s cardiovascular health.

Table 3.

Associations between SHS biomarkers, as indicated by quartiles of serum cotinine and urinary 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanol (NNAL), and hypertensive status among children in NHANES 2007–2012.

SHS exposure level	% in hypertensive range	OR	95% CI
Overall
Cotinine below detection limit	9.1	ref
Detectable cotinine ≤ 75th percentile	12.2	1.345	(0.948, 1.907)
Cotinine > 75th percentile	16.4	1.979	(1.435, 2.73)
Males
Cotinine below detection limit	12.1	ref
Detectable cotinine ≤ 75th percentile	15.9	1.297	(0.86, 1.955)
Cotinine > 75th percentile	20.6	1.972	(1.238, 3.143)
Females
Cotinine below detection limit	6.4	ref
Detectable cotinine ≤ 75th percentile	8.4	1.286	(0.641, 2.58)
Cotinine > 75th percentile	11.7	1.746	(0.907, 3.36)
Overall
NNAL below detection limit	10.3	ref
Detectable NNAL ≤ 75th percentile	11.6	1.124	(0.812, 1.556)
NNAL > 75th percentile	16.8	1.966	(1.31, 2.951)
Males
NNAL below detection limit	13.5	ref
Detectable NNAL ≤ 75th percentile	15.4	1.111	(0.747, 1.653)
NNAL > 75th percentile	20.8	1.993	(1.185, 3.352)
Females
NNAL below detection limit	7.2	ref
Detectable NNAL ≤ 75th percentile	7.8	1.159	(0.666, 2.017)
NNAL > 75th percentile	12.1	1.839	(1.014, 3.335)

For participants less than 13 years of age (n=1980, weighted proportion= 48.7%), the hypertensive range is defined as SBP ≥ 120 mmHg or SBP percentile at or above the 90^th percentile, or DBP ≥ 80 mmHg or DBP percentile at or above the 90^th percentile. For participants aged 13 years of age or older (n=1599, weighted proportion=51.3%), the hypertensive range is defined as SBP ≥ 120 mmHg or DBP ≥ 80 mmHg. Odds ratios and associated confidence intervals are calculated using weighted logistic regression models adjusted for age, sex, height, race/ethnicity, family poverty to income ratio, waist circumference, cadmium (ug/L), lead (ug/dL), first albumin creatinine ratio (mg/g), and urinary creatinine (mg/dL, for NNAL model only). Detection limits for cotinine and NNAL were 0.015 ng/mL and 0.6 pg/mL, respectively. Exposure concentrations for the 75^th percentiles of cotinine and NNAL were 0.16 ng/mL and 4.28 pg/mL, respectively.

Highlights.

• Secondhand tobacco smoke exposure associated with higher blood pressure in children

• Acute exposure associated with higher diastolic blood pressure

• Intermittent exposure associated with higher systolic, diastolic blood pressure

• Sex-specific differences in associations between cotinine/NNAL and blood pressure