Urinary Cotinine Cut-offs for Tobacco Smoke Exposure in Pregnancy and Associations with Child Intelligence Quotient: A Multi-Cohort Analysis

Abstract

Background:

Prenatal exposure to tobacco smoke may impair neurodevelopment in children. However, accurately characterizing this exposure remains challenging.

Methods:

We pursued two objectives in this large population study. First, in 1,708 pregnant women from the Environmental Influences on Child Health Outcomes (ECHO) cohort, we constructed Receiver Operating Characteristic (ROC) curves to determine urinary cotinine cut-offs to classify firsthand (FHS), environmental (ETS), and no exposure, and further distinguished secondhand (SHS) from thirdhand smoke (THS) exposure within ETS. Second, among 1,593 participants in three pregnancy cohorts nested in ECHO, we fit multivariable linear regressions to examine the association between the newly defined smoke exposures and child full-scale intelligence quotient (IQ) at age 4–6 years, and to assess potential effect modification by maternal education or neighborhood deprivation.

Results

Optimal cotinine cut-offs were 17.74 ng/mL and 0.44 ng/mL to discriminate FHS and no exposure, respectively. Among the ETS group, a cut-off of 5.69 ng/mL differentiated SHS from THS. Applying these optimal cut-offs, we estimated a 0.93-point (95%CI: −3.44, 1.59) and a 1.03-point (95%CI: −2.84, 0.79) lower child IQ in the FHS and ETS categories, respectively, compared to no exposure. The inverse association between prenatal ETS and child IQ was mainly driven by SHS. Stronger associations were suggested in subgroups with higher education attainment or those living in less deprived neighborhoods.

Conclusions:

This study provides a novel classification of prenatal tobacco smoke exposures. Although the associations with child IQ were statistically insignificant, the study carries important implications for future research on developmental origins of diseases.

Introduction

The fetal period is widely recognized as a critical window of neurodevelopment.^1,2 This stage is marked by rapid cell differentiation, rapid neurogenesis, and the formation of intricate neural networks.^1,3 Even subtle environmental disturbances in early life, such as tobacco smoke exposure at low levels, may interfere with the normal trajectory of brain development, potentially leading to cognitive impairments that persist across the lifespan.⁴

Despite a significant decline in maternal smoking during pregnancy (i.e., firsthand smoking [FHS]) in the U.S. over the past decades – from 7.2% in 2016 to 3.7% in 2022⁵ – approximately a quarter of expectant women were exposed to daily environmental tobacco smoke (ETS).^6–8 Populations with lower educational attainment^9,10 and those living in socioeconomically deprived neighborhoods^11–13 consistently experience higher active and passive smoking exposure. These disparities are shaped by a confluence of social determinants, including chronic psychological stress, cultural norms, and material hardship.^14,15 Laboratory research supports the cytotoxicity of tobacco smoke exposures on neurocytes.^16,17 Compelling evidence from population studies has also shown that FHS during pregnancy has a detrimental effect on major developmental and health outcomes in children, independent of postnatal smoke exposure.^18,19 However, findings on the harms of maternal ETS on child neurodevelopment remain inconsistent.^20,21 Alongside secondhand smoke (SHS), the term “thirdhand smoke” (THS) has emerged in recent years to describe the smoke residues that linger on surfaces and in dust after smoking and can subsequently reemit into air.²² SHS and THS differ in chemical profiles and routes of uptake.²³ Treating ETS as synonymous with SHS may underestimate the health risks from chronic THS, particularly for young children, who have frequent contact with contaminated surfaces and exhibit hand-to-mouth behaviors.²⁴ Despite growing recognition of the ubiquity of THS, there is a scarcity of reports on the distinct health effects of SHS and THS.

An accurate assessment of tobacco smoke exposure is essential for epidemiological research purposes. Self-reports of active and passive tobacco smoke exposure can be biased, due to poor recall, stigma, and lack of awareness of passive exposures.²⁵ This underscores the need for valid determination of exposure that integrates information from questionnaires and biomarkers. Cotinine, the major metabolite of nicotine, is a well-established marker widely used to measure ETS in nonsmokers. Previous studies, mostly conducted outside the U.S., have recommended cotinine cut-offs for identifying FHS in pregnant women using saliva and urinary samples.^26–30 However, there is a lack of comprehensive data to further differentiate ETS beyond this context.^27,30 No study to date has proposed a specific cut-off to distinguish SHS from THS within the broader category of ETS in pregnant women.

To address these critical gaps, we conducted a large population study with two objectives. We first determined urinary cotinine cut-offs for classifying tobacco smoke exposure status in pregnant women, using publicly available data from the first cycle of the Environmental Influences on Child Health Outcomes (ECHO) cohort.^31,32 We then examined the associations between the newly defined smoke exposure status during pregnancy and child full-scale intelligence quotient (IQ), among participants in three pregnancy cohorts nested in ECHO with in-depth data collection for maternal and child characteristics. We further examined whether maternal education and neighborhood deprivation modified these associations

Methods

Study population

Launched in 2016 by the National Institutes of Health, the ECHO Program seeks to understand how the environment, from preconception through early childhood, influences child health and development, with the goal of identifying opportunities to mitigate disease risk and optimize health and thriving.^31,32 The first cycle of the ECHO cohort brought together nearly 60,000 mother-child dyads from 69 individual pregnancy or birth cohorts across the U.S. into a large consortium. This design ensures broad representation across racial, geographic, and socioeconomic backgrounds. The ECHO program provides public access to de-identified participant data through periodic releases on the Data and Specimen Hub hosted by the Eunice Kennedy Shriver National Institute of Child Health and Human Development (https://dash.nichd.nih.gov/). To determine urinary cotinine cut-offs for classifying tobacco smoke exposure status in pregnant women, we analyzed data from 1,708 participants in the ECHO cohort included in the second data release. From 11 original cohorts, these participants had both urinary cotinine and hydration status measured in the second trimester and completed one or more questionnaires regarding tobacco smoke exposures during pregnancy.

We further applied the newly derived urinary cotinine cut-offs to a combined sample of three pregnancy cohorts from the ECHO-PATHWAYS consortium³³ nested in the ECHO cohort -- the Conditions Affecting Neurocognitive Development and Learning in Early Childhood (CANDLE) study, The Infant Development and the Environment Study (TIDES), and the Global Alliance to Prevent Prematurity and Stillbirth (GAPPS), aiming to investigate the association between women’s tobacco smoke exposure status during pregnancy and child IQ. CANDLE enrolled 1,503 pregnant women in their second trimester 2006–2011 from either the general community or affiliated medical group clinics in Shelby County, Tennessee.³⁴ TIDES recruited pregnant women in early pregnancy 2010–2012 from academic medical centers in San Francisco, California; Rochester, New York; Minneapolis, Minnesota; and Seattle, Washington, retaining 803 through delivery.³⁵ GAPPS (https://www.gapps.org) was originally a biorepository based in Seattle and Yakima, Washington designed for research on reducing adverse birth outcomes (births in 2017–2020), and 673 of these women who had available prenatal survey data and biospecimens, and their children aged 4–8, were selected for inclusion as the PATHWAYS-GAPPS cohort. Prior to joining ECHO, these three studies had collected a wealth of extant data on maternal and child characteristics. In the second phase of this analysis, we included 1,593 mother-child dyads from these three cohorts who had available second-trimester urinary cotinine, specific gravity measurements, and child IQ assessments at age 4–6 years. Special informed consent procedures for consenting parents and assenting children have been performed at individual cohorts or sites. This analysis was approved by the Human Subjects Protection Program at the San Diego State University.

Self-reported tobacco smoke exposure

We leveraged self-reported data from multiple ECHO-standardized and cohort-specific questionnaires to classify pregnant women into four tobacco exposure subgroups – FHS, SHS, THS, and no exposure. Given that questions ascertaining FHS were more thoroughly completed than those about other smoke exposures, active smokers during pregnancy were overrepresented. Example questions are listed in Appendix Table 1. We classified participants as having FHS if they confirmed any tobacco or nicotine use during the ECHO index pregnancy. Women who reported anyone (excluding themselves) smoking inside their homes or dwellings during pregnancy were categorized as having SHS. Those living with anyone who smoked cigarettes during pregnancy but outside their homes or dwellings were defined as having THS. Participants denying all three exposures were considered no exposure. The subgroups of SHS and THS were also combined as a broader category of ETS.

Urinary cotinine and hydration status measurements

Urinary samples were collected during research visits in the second trimester. In the second release of ECHO data, cotinine measurements were available for participants in 11 out of 69 cohorts, following each cohort’s specific protocols. The majority of these cohorts had a limit of detection (LOD) ranging from 0.002 to 0.1 ng/mL, whereas one cohort that exclusively recruited active smokers during pregnancy used a higher LOD of 5 ng/mL. Among 1,708 ECHO participants with cotinine, two-thirds of the ECHO participants (n=1,194) had urine specific gravity assessed at the time of collection, while the remaining participants (n=514) had available creatinine measurements. In the CANDLE, TIDES, and PATHWAYS-GAPPS studies, urinary samples were processed using hybrid solid phase extraction to obtain cleaned extracts.³⁶ The chromatographic separation of cotinine was accomplished using a Shimadzu LC-30AD HPLC system (Shimadzu; Japan) connected to an Acquity UPLC BEH C18 column (50 × 2.1 mm, 1.7 μm, Waters; Milford, MA, USA). Identification and quantification of cotinine was performed on an ABSCIEX 5500 (Applied Biosystems, Foster City, CA, USA). The cohort-specific LODs ranged from 0.002 to 0.017 ng/mL. Specific gravity was determined via a handheld refractometer at the time of urinary sample collection. For all samples, cotinine concentrations below the limit of detection were assigned a value of $\text{LOD}/\sqrt{2}$.

Child IQ assessments

Child IQ at age 4–6 years was assessed using validated examiner-administered batteries, with instruments varying by cohorts. CANDLE used the Stanford-Binet Intelligence Scales, 5^th edition (SB-5);^37,38 TIDES employed the abbreviated five-subtest version of the Wechsler Intelligence Scale for Children, 5^th Edition (WISC-V);^39,40 and PATHWAYS-GAPPS administered the Wechsler Preschool & Primary Scale of Intelligence, 4^th Edition (WPPSI-IV, age 4:0–7:7 version).^41,42 In CANDLE, full-scale IQ was derived from 10 subtests in the SB-5 covering five cognitive domains: knowledge, fluid reasoning, quantitative reasoning, visual-spatial processing, and working memory. For TIDES and PATHWAYS-GAPPS, the metric incorporated five cognitive domains – verbal comprehension, visual-spatial, fluid reasoning, working memory, and processing speed. Despite differences in the underlying domains, full-scale IQ scores are highly correlated across these standardized instruments and provide a consistent and reliable metric of overall cognitive performance.⁴³

Covariates

To estimate associations between tobacco smoke exposure in pregnancy and child IQ, we identified several maternal and child characteristics from a literature review as major confounders or precision variables, as depicted in a conceptual model (Appendix Figure 1). Maternal characteristics included the highest education attained (less than high school vs. high school degree/General Educational Development or equivalent vs. some college and above), race (White, Black, Asian, Native Hawaiian or Pacific islander, Native American or Alaska Native, multiple races, or other races), ethnicity (Hispanic vs. non-Hispanic), region- and inflation-adjusted household income, household size, age at delivery, alcohol consumption during pregnancy (ever vs. never), and IQ measured by the Wechsler Abbreviated Scale of Intelligence (the 1^st Edition⁴⁴ in CANDLE and the 2^nd Edition^45,46 in TIDES and PATHWAYS-GAPPS). Child characteristics were sex (female vs. male), age at IQ assessment, and birth order (firstborn vs. non-firstborn), which were considered precision variables. These three variables were associated with the outcome but not necessarily with the exposure, and their inclusion can improve the precision of effect estimates by reducing residual variance, particularly for continuous outcomes such as child IQ⁴⁷. Children’s secondhand smoke exposure at home (ever vs. never) was obtained from parental report. Neighborhood-level covariates – study site and Neighborhood Deprivation Index (NDI)⁴⁸ at the census tract level – were further included.

Statistical analysis

To determine urinary cotinine cut-offs for classifying tobacco smoke exposure status, we first implemented the method proposed by Boeniger et al. to calculate hydration adjusted cotinine by either specific gravity or creatinine.⁴⁹ The equations are specified as $E_{sg}=E_o\times \frac{{SG}_{\mathit{\text{median}}}-1}{{SG}_o-1}$ and $E_{cr}=E_o\times \frac{{CR}_{\mathit{\text{median}}}-1}{{CR}_o}$, respectively, where $E_o$ is the raw cotinine concentration in urine, ${SG}_o$ is the measured specific gravity, ${CR}_o$ is the measured creatinine, ${SG}_{\mathit{\text{median}}}$ and ${CR}_{\mathit{\text{median}}}$ are the median of specific gravity or creatinine in each individual cohort. We combined $E_{sg}$ and $E_{CR}$ – hydration adjusted cotinine – and treated them equally in the following analyses. We then described the distribution of hydration-adjusted cotinine overall and by self-reported tobacco smoke exposure status.

Using questionnaire-defined exposure subgroups as the reference, we constructed Receiver Operating Characteristic (ROC) curves to estimate optimal urinary cotinine cut-offs for distinguishing FHS-predominant, ETS-predominant, and no exposure-predominant. We made a further attempt to determine the cut-off to distinguish SHS-predominant from THS-predominant among individuals with ETS exposure. Cut-offs were selected by maximizing the weighted Youden index to optimize diagnostic accuracy. This index is defined as $2\times \left[w\times \mathit{\text{sensitivity}}+\left(1-w\right)\times \mathit{\text{specificity}}\right]-1$.⁵⁰ Considering the right-skewed distribution of urinary cotinine and the subsequent potential high false positive rate, we assigned a 70% weight to specificity and a 30% weight to sensitivity when determining the cut-offs.⁵¹ For validation, we employed an expectation-maximization (EM) algorithm for mixtures of univariate Gaussian distributions with a 50:50 random split of data for training and testing in a sensitivity analysis.^52–54 This data-driven method derives optimal mixture components by fitting the best univariate empirical distribution to hydration-adjusted cotinine data without requiring a gold-standard reference. We compared the maximum likelihood estimates for models with 1 to 8 mixture components.

In the combined CANDLE, TIDES, and PATHWAYS-GAPPS sample, self-reported data to define tobacco exposure status is incomplete. We applied the optimal cut-offs identified in prior analyses to second trimester urinary cotinine measurements and further investigated the associations between the newly defined tobacco smoke exposure status and child full-scale IQ. In the primary analysis, we fit multivariable linear regressions with adjustment for maternal education, maternal race, maternal ethnicity, maternal age at delivery, pregnancy alcohol consumption, maternal IQ, an interaction between region- and inflation-adjusted household income and household size, child sex, child age at IQ assessment, birth order, study sites, and NDI. Child secondhand smoke exposure, which may correlate with prenatal smoking exposure and serve as a mediator, was additionally included in an extended model. Multiple imputation by chained equations (MICE) was employed to address missing data in covariates.⁵⁵ Each missing value was imputed 10 times with 100 iterations between each round of imputation using predictive mean matching. To estimate the potential modified associations, we introduced cross product terms of maternal education and tobacco smoke exposure, as well as NDI and tobacco smoke exposure, in separate linear regressions. All analyses were performed in R (version 4.5.0; R Development Core Team).

Results

Among 1,708 pregnant women participating in ECHO, 67% self-identified as White, 12% as Black, 20% as other race, and 17% as Hispanic/Latino. The mean age at delivery was 30.8 (SD: 5.7) years. Approximately 40% had attained some college education or higher. Based on questionnaire responses, 379 (22%) women were identified as having FHS, followed by 59 (3.5%) with SHS and 60 (3.5%) with THS, while 1,210 reported no exposure. Detection rates of urinary cotinine varied between 18.6% and 100% across 11 cohorts, with an overall rate of 56% in the analytic sample. The geometric means in hydration-adjusted cotinine were 530.67 (geometric standard deviation [GSD]: 34.4), 1.82 (GSD: 24.81), 0.20 (GSD: 21.63), and 0.03 (GSD: 11.03) ng/mL in the subgroups of FHS, SHS, THS, and no exposure, respectively.

Optimal cotinine cut-offs derived from the ROC analysis were 17.74 ng/mL to discriminate FHS-predominant (Figure 1A; the weighted Youden index: 0.88, sensitivity: 0.87, specificity: 0.97) and 0.44 ng/mL to discriminate no exposure-predominant (Figure 1B; the weighted Youden index: 0.73, sensitivity: 0.87, specificity: 0.88). Among the ETS-predominant group, the optimal cotinine cut-off of 5.69 ng/mL (Figure 1C; the weighted Youden index: 0.52, sensitivity: 0.44, specificity: 0.9) was further determined to distinguish SHS-predominant from THS-predominant. The sensitivity analysis of the EM algorithm combined with cross-validation showed a substantial increase in log-likelihood for models with up to three components (log-likelihood: −1886.5) and four components (log-likelihood: −1873.8), indicating significant improvements in fit (Appendix Figure 2). We further plotted the Gaussian distributions of three and four mixture components with hydration adjusted cotinine and the optimal cut-offs (Appendix Figure 3). The optimal cut-off to distinguish SHS- from THS-predominant aligned with the Gaussian distributions in the 4-component analysis; however, the cut-offs for discriminating FHS and no exposure did not match well with the distributions. This disagreement likely arises because cotinine levels do not follow the expected Gaussian distribution in subgroups with a small sample.

Figure 1. Receiver Operating Characteristic (ROC) curves to establish optimal urinary cotinine cut-offs for different tobacco smoke exposure status.

FHS: firsthand smoke exposure; SHS: secondhand smoke exposure; THS: thirdhand smoke exposure; no exp: no tobacco smoke exposure; any exp: any tobacco smoke exposure. Figure 1A: the optimal cotinine cut-off derived from the ROC analysis – 17.7 ng/mL – to discriminate FHS from non-FHS. Figure 1B: the optimal cotinine cut-off derived from the ROC analysis – 0.4 ng/mL – to discriminate no exp from any exp. Figure 1C: the optimal cotinine cut-off derived from the ROC analysis – 5.69 ng/mL – to discriminate SHS from THS within participants with environmental tobacco smoke exposure.

Among 1,593 mother-child dyads in CANDLE, TIDES, and PATHWAYS-GAPPS, 48% of participating women self-identified as White, 42% as Black, and 5% as Hispanic/Latino (Table 1). They had an average age of 28.7 (standard deviation [SD]: 6) years at delivery, and more than half (55%) had attained some college or higher. The median region- and inflation-adjusted household income was $51,167 (interquartile range: 59,736). Compared to the 1,708 ECHO participants included for identifying urinary cotinine cut-offs, these women were younger, more educated on average, and included a higher proportion of Black participants. Children were on average 5 years of age (SD: 1) at IQ assessment, and there was an approximately equal number of boys and girls. Child full-scale IQ was normally distributed, with a mean of 103 (SD:15) (Appendix Figure 4). Across cohorts, urinary cotinine detection rates varied from 31% to 86.7%, and the overall detection rate was 66.4%. Applying the optimal cut-offs newly derived to urinary cotinine levels in this sample, we defined 155 (9.7%) participating women with FHS during pregnancy, 422 (26.5%) with ETS (96 [6%] with SHS and 326 [20.5%] with THS), and 1016 (63.8%) without exposure. Active smokers tended to have less education, lower average IQ scores, and lived in more disadvantaged neighborhoods, compared to other subgroups.

Table 1.

Participant characteristics in the CANDLE, TIDES, and GAPPS cohort in overall sample and by tobacco smoke exposure status defined by urinary cotinine cut-offs

	Overall sample	FHS-predominant	ETS-predominant	No exp-predominant
	n = 1,593	n = 155	n = 422	n = 1,016
Maternal characteristics
Education
Less than high school	131 (8%)	44 (28%)	60 (14%)	27 (3%)
High school/GED	581 (37%)	90 (58%)	245 (58%)	246 (24%)
Some college and above	875 (55%)	21 (14%)	113 (27%)	740 (73%)
Missing	6 (0.4%)	0 (0%)	3 (1%)	3 (0%)
Race
White	764 (48%)	40 (26%)	71 (17%)	653 (64%)
Black	664 (42%)	102 (66%)	307 (73%)	255 (25%)
Asian	33 (2%)	0 (0%)	1 (0.01%)	32 (3%)
Native Hawaiian/Pacific islander	2 (0%)	0 (0%)	0 (0%)	2 (0%)
Native American/Alaska Native	6 (0%)	0 (0%)	2 (0.01%)	4 (0%)
Others	77 (5%)	8 (5%)	32 (8%)	37 (4%)
Multi-race	32 (2%)	3 (2%)	6 (1%)	23 (2%)
Missing	15 (1%)	2 (1%)	3 (1%)	10 (1%)
Ethnicity
Hispanic	78 (5%)	8 (5%)	17 (4%)	53 (5%)
Non-Hispanic	1506 (95%)	146 (94%)	401 (95%)	959 (94%)
Missing	9 (1%)	1 (1%)	4 (1%)	4 (0%)
Age at delivery	28.7 (6)	25.7 (5.4)	25.3 (5.5)	30.6 (5.5)
Missing	13 (1%)	2 (1%)	1 (0.01%)	10 (1%)
Household income	51167 [59736]	17812 [23443]	22901 [30115]	81426 [68540]
Missing	75 (5%)	8 (5%)	27 (6%)	40 (4%)
Household size
≤3	337 (21%)	26 (17%)	101 (24%)	210 (21%)
4	625 (39%)	48 (31%)	129 (31%)	448 (44%)
5	349 (22%)	36 (23%)	91 (22%)	222 (22%)
≥6	256 (16%)	43 (28%)	96 (23%)	117 (12%)
Missing	26 (2%)	2 (1%)	5 (1%)	19 (2%)
Maternal IQ	100.8 (18.3)	86.1 (15)	89.2 (15)	108.3 (15.8)
Missing	86 (5%)	2 (1%)	8 (2%)	76 (7%)
Prenatal alcohol use	134 (8%)	14 (9%)	24 (6%)	96 (9%)
Missing	6 (0.4%)	1 (1%)	1 (0.1%)	4 (0%)
Child characteristics
Age at outcome assessment	5 (1)	4.8 (1)	4.7 (0.9)	5.2 (1)
Missing	18 (1%)	2 (1%)	2 (0.01%)	14 (1%)
Sex
Male	758 (48%)	77 (50%)	203 (48%)	478 (47%)
Female	835 (52%)	78 (50%)	219 (52%)	538 (53%)
Firstborn	657 (41%)	44 (28%)	174 (41%)	439 (43%)
Missing	10 (1%)	0 (0%)	2 (0.01%)	8 (1%)
Neighborhood Deprivation Index	0.16 (0.8)	0.7 (0.8)	0.7 (0.8)	−0.51 (0.7)
Missing	59 (4%)	3 (2%)	13 (3%)	43 (4%)
Cohort: study site
CANDLE: Memphis, TN	963 (60%)	123 (79%)	342 (81%)	498 (49%)
TIDES: Minneapolis	111 (7%)	2 (1%)	7 (2%)	102 (10%)
TIDES: Rochester	108 (7%)	22 (14%)	38 (9%)	48 (5%)
TIDES: San Francisco	100 (6%)	1 (1%)	4 (1%)	95 (9%)
TIDES: Seattle	91 (6%)	3 (2%)	6 (1%)	82 (8%)
GAPPS: Seattle	99 (6%)	3 (2%)	6 (1%)	90 (9%)
GAPPS: Yakima	121 (8%)	1 (1%)	19 (5%)	101 (10%)

FHS: firsthand smoke exposure; ETS: environmental tobacco smoke exposure.; no exp: no tobacco smoke exposure; GED: General Educational Development or equivalent. Shown in the table are mean (SD), counts (percentage), and median [interquartile range]. Shown in the table are mean (± SD), counts (%), and median [interquartile range].

Compared to the no exposure-predominant category, FHS-predominant and ETS-predominant during pregnancy were associated with a statistically insignificant decrease in early childhood IQ scores of 0.93 points (coefficient: −0.93, 95% confidence interval [CI]: −3.44, 1.59) and 1.03 points (coefficient: −1.03, 95%CI: −2.84, 0.79), respectively (Table 2). These conclusions remained unchanged after postnatal secondhand smoke exposure was additionally controlled. The suggestive inverse association between prenatal ETS and child IQ in our samples was mainly driven by SHS: children with prenatal SHS exposure had an average 2.18-point lower IQ (coefficient: −2.18, 95%CI: −5.07, 0.71) than children without major tobacco smoke exposure, while the average IQ in children with prenatal THS exposure was 0.76-point lower (coefficient: −0.76, 95%CI: −2.66, 1.15). Among mothers with some college education or higher, stronger associations of pregnancy FHS and ETS with child IQ were detected (coefficient: −3.57, 95%CI: −9.16, 2.01 for FHS-predominant and coefficient: −2.41, 95%CI: −5.06, 0.25 for ETS-predominant), although the interaction term between tobacco smoke exposure status and maternal education was insignificant ($p_{\mathit{\text{interaction}}}:$ 0.43) (Table 3). The effect modification by NDI was also insignificant ($p_{\mathit{\text{interaction}}}$: 0.10) (Table 4). However, in the least deprived neighborhood, FHS and ETS were associated with a 1.79-point (coefficient: −1.76, 95%CI: −4.73, 1.14) and a 1.59-point lower (coefficient: −1.59, 95%CI: −3.52, 0.35) child IQ, while these associations reversed in the most deprived neighborhood (coefficient: 5.23, 95%CI: −2.08, 12.55 for FHS-predominant and coefficient: 4.14, 95%CI: −1.47, 9.76 for ETS-predominant).

Table 2.

Association between pregnancy tobacco smoke exposure status defined by urinary cotinine cut-offs and child IQ in CANDLE, TIDES, and GAPPS participants

	Primary Model	Extended Model
Exposure	coefficient (95% CI)	coefficient (95% CI)
No exp-predominant	Ref	Ref
FHS-predominant	−0.93 (−3.44, 1.59)	−0.99 (−3.64, 1.66)
ETS-predominant	−1.03 (−2.84, 0.79)	−1.07 (−2.95, 0.81)
No exp-predominant	Ref	Ref
FHS-predominant	−1.00 (−3.52, 1.52)	−1.09 (−3.75, 1.57)
SHS-predominant	−2.18 (−5.07, 0.71)	−2.24 (−5.19, 0.71)
THS-predominant	−0.76 (−2.66, 1.15)	−0.81 (−2.77, 1.14)

FHS: firsthand smoke exposure; ETS: environmental tobacco smoke exposure.; no exp: no tobacco smoke exposure; Ref: reference. Using the no exp-predominant as the reference, we fit multivariable linear regressions to estimate the associations of tobacco smoke exposure status with child IQ. The primary model was adjusted for maternal education, maternal race, maternal ethnicity, maternal age at delivery, pregnancy alcohol consumption, maternal IQ, an interaction between region- and inflation-adjusted household income and house size accounting for non-proportional financial needs of a household grow with additional family members, child sex, child age at IQ assessment, birth order, study sites, and Neighborhood Deprivation Index. The extended model was additionally controlled for child secondhand smoke exposure. Missing data were imputed by Multiple Imputation by Chained Equations.

Table 3.

Modified associations between tobacco smoke exposure status in pregnancy defined by urinary cotinine cut-offs and child IQ by maternal education in CANDLE, TIDES, and GAPPS participants

Exposure	coefficient (95% CI)
Among the subgroup of less than high school
No exp-predominant	Ref
FHS-predominant	2.77 (−3.32, 8.87)
EHS-predominant	1.74 (−4.02, 7.51)
Among the subgroup of high school
No exp-predominant	Ref
FHS-predominant	−0.55 (−3.72, 2.63)
EHS-predominant	−0.21 (−2.58, 2.15)
Among the subgroup of some college and above
No exp-predominant	Ref
FHS-predominant	−3.57 (−9.16, 2.01)
EHS-predominant	−2.41 (−5.06, 0.25)

^*Wald Global test $p_{\mathit{\text{interaction}}}=0.43$

FHS: firsthand smoke exposure; ETS: environmental tobacco smoke exposure.; no exp: no tobacco smoke exposure; Ref: reference. Using the no exp-predominant as the reference, we fit a multivariable linear regression with an interaction term between tobacco smoke exposure status and maternal education, controlling for maternal race, maternal ethnicity, maternal age at delivery, pregnancy alcohol consumption, maternal IQ, an interaction between region- and inflation-adjusted household income and house size accounting for non-proportional financial needs of a household grow with additional family members, child sex, child age at IQ assessment, birth order, study sites, and Neighborhood Deprivation Index. Missing data were imputed by Multiple Imputation by Chained Equations.

Table 4.

Modified associations between tobacco smoke exposure status in pregnancy defined by urinary cotinine cut-offs and child IQ by Neighborhood Deprivation Index in CANDLE, TIDES, and GAPPS participants

Exposure	coefficient (95% CI)
Among the subgroup living in the least deprived neighborhood
No exp-predominant	Ref
FHS-predominant	−1.76 (−4.73, 1.14)
EHS-predominant	−1.59 (−3.52, 0.35)
Among the subgroup living in the most deprived neighborhood
No exp-predominant	Ref
FHS-predominant	5.23 (−2.08, 12.55)
EHS-predominant	4.14 (−1.47, 9.76)

^*Wald Global test $p_{\mathit{\text{interaction}}}=0.10$

FHS: firsthand smoke exposure; ETS: environmental tobacco smoke exposure.; no exp: no tobacco smoke exposure; Ref: reference. Using the no exp-predominant as the reference, we fit a multivariable linear regression with an interaction term between tobacco smoke exposure status and Neighborhood Deprivation Index, controlling for maternal race, maternal ethnicity, maternal age at delivery, maternal education, pregnancy alcohol consumption, maternal IQ, an interaction between region- and inflation-adjusted household income and house size accounting for non-proportional financial needs of a household grow with additional family members, child sex, child age at IQ assessment, birth order, and study sites. Missing data were imputed by Multiple Imputation by Chained Equations.

Discussion

In this large multi-cohort study, we determined urinary cotinine cut-offs for classifying tobacco smoke exposure status in pregnant women, including an exploration of the boundary between SHS and THS. We further applied these optimal cut-offs to a pooled analysis of three cohorts nested in ECHO and found suggestive but nonsignificant inverse associations between tobacco smoke exposures in pregnancy and child IQ. We further hypothesized that these associations would be stronger in subgroups with lower maternal education or those residing in more deprived neighborhoods; however, our results did not support this hypothesis.

We are one of the very few U.S. studies that infer tobacco smoke exposure status in pregnant women using urinary cotinine cut-offs. We know of only one other study based in the U.S., which examined the pre-defined urinary creatinine-adjusted cotinine cut-off values – 100, 200, and 500 ng/mL – to determine active smoking in 998 pregnant women recruited from East Boston between 1988 and 1992.²⁹ An optimal cut-off of 200 ng/mL was reported from an ROC plot, which is much higher than the cut-off of 17.74 ng/mL in our analysis. Similar studies in Europe and Asia have also documented higher urinary cotinine cut-offs to discriminate active smoking than our results. For example, data from 595 pregnant women who sought prenatal care in a public hospital in Porto, Portugal have indicated a cut-off of 74.1 ng/mL for urinary cotinine.³⁰ A study based on 69 women from the Polish Mother and Child Cohort Study has established a cut-off of 42.3 ng/mL.²⁸ Data from 2263 pregnant women from the Infancia y Medio Ambiente, Environment and Childhood project – a Spanish multicenter birth cohort -- has suggested cut-offs of 50 ng/mL for daily smokers and 25 ng/mL for occasional smokers, given by the Youden index.²⁶ A more recent study in the Japan Environment and Children’s Study (n=89,895) has identified the cut-off as 21.5 ng/mL.²⁷ Over the past two decades, we have observed a decline in urinary cotinine cut-off levels for active smokers,⁵⁶ likely driven by stronger tobacco control policies and regulations, greater access to routine counseling during prenatal visits and cessation support programs, as well as increased public health awareness and broader cultural shifts. Subsequently, women may reduce their tobacco use or quit smoking entirely during pregnancy. Switching from cigarettes to newer products such as nicotine-free e-cigarettes or vapes among some pregnant women may also contribute to this observation.

The optimal cut-off of 0.44 ng/mL to discriminate the no exposure-predominant group from any exposure in our analysis falls between the results from the previous two studies: the Japan Environment and Children’s Study reported a 0.17 ng/mL cut-off,²⁷ while the value was higher in the study in Portugal (1.6 ng/mL).³⁰ For exploratory purposes, we identified a cut-off value of 5.69 ng/mL to distinguish between SHS and THS. To our knowledge, no previous studies have conducted a similar analysis in pregnant women. A study of 466 adolescents from low-income families in San Francisco, CA used 0.05 ng/mL to classify light SHS or THS vs. SHS or light/intermittent active smokers.⁵⁷ Another recent study of 1,033 children recruited from clinical settings in Cincinnati, OH has identified the threshold of 0.13 ng/mL to discriminate THS from SHS.⁵⁸ Pharmacokinetic data has shown that total nicotine clearance increases by 60% during pregnancy, and cotinine clearance increases by 140% compared to the postpartum or non-pregnant state, which may explain the higher cut-off detected in our sample.⁵⁹ Overall, different cut-off values across studies can be attributed to variations in methods, such as the difference in study population (e.g., with different smoking prevalence or clearance of nicotine), the trimester(s) of urinary sample collection, the questions designed to ascertain women’s smoke exposure, and the statistical criteria (e.g., the Youden index or maximum accuracy) to determine the optimal cut-offs.

Within the CANDLE, TIDES, and PATHWAYS-GAPPS cohort, we found suggestive inverse associations between tobacco smoke exposures in pregnancy and child IQ. It is possible that the results are imprecise due to relatively small samples in some exposure categories, such as FHS and SHS. Several mechanisms by which maternal smoke exposure can influence children’s IQ have been proposed. Substances found in cigarettes can impair uteroplacental circulation and cause fetal hypoxia.^60,61 They can also interfere with amino acid transport, protein synthesis, and enzyme activity.⁶² Nicotine, a key component, stimulates the cholinergic system and disrupts critical neurodevelopmental processes such as synaptogenesis and apoptosis, thereby altering brain and cognitive development.⁶³ A meta-analysis with 25 studies reported a pooled estimate of −1.3 (95% CI −1.74, −0.86), showing that children who were exposed to maternal active smoking in pregnancy presented lower IQ scores compared to the unexposed.^18,19 This magnitude of association aligns with our results. Additionally, a substantial number of studies have investigated the association between prenatal ETS and different cognitive parameters in children, with many reporting an inverse though modest association.^20,21 However, we are not aware of any existing studies based on large cohorts illustrating the association between THS exposure during pregnancy and child neurodevelopment. In this analysis, we have made the first attempt to quantify the separate health effects of SHS and THS and detected a potential stronger association with SHS. We have incorporated some major improvements to address residual confounding, recognizing that few existing studies have been able to control for maternal IQ, and socioeconomic status at both individual and neighborhood levels is rarely accounted for. We have also adjusted for postnatal secondhand smoke exposure to tease apart its impact on the association between tobacco smoke exposure in pregnancy and child IQ, which was not accomplished in many previous studies.²⁰

Literature on environmental and social determinants of health indicates that populations with lower educational attainment^9,10 and those living in deprived neighborhoods^11–13 are more likely to be exposed to tobacco smoke and experience elevated risk of adverse health outcomes. Many studies have documented that the introduction of maternal education as a confounder attenuates the association between tobacco smoke exposure in pregnancy and child cognitive outcomes.^64–70 However, the potential effect modification of this association by maternal education and neighborhood deprivation has rarely been explored. A study in 226 inner-city mother-child dyads in New York State has concluded that the negative impact of prenatal exposure to ETS on 2-year cognitive development is exacerbated under conditions of material hardship.⁷¹ However, although there was no strong statistical evidence, we detected stronger inverse associations in subgroups with higher maternal education or those living in less deprived neighborhoods. Low education attainment and neighborhood deprivation are linked to a range of economic and psychosocial disadvantages, such as limited access to healthcare, lack of nutritious food, inadequate social support, intensified tobacco marketing, poor parent-child interaction, and other environmental contaminants, any of which may contribute to aberrant child development. While it is possible that maternal tobacco smoke exposure is equally toxic across subgroups, failure to fully characterize the aforementioned economic and psychosocial challenges could mask the relationship between maternal smoke exposure and child IQ among disadvantaged families.

This study has several notable strengths and provides unique contributions to the current literature. It encompassed a relatively large sample from multiple U.S. pregnancy cohorts in the ECHO cohort with diverse sociodemographic characteristics. It provides the very first evidence to inform detailed definitions for tobacco smoke exposure status in U.S. pregnant women, combining questionnaire information with urinary cotinine measurements. We also included child IQ assessed by trained examiners using validated testing batteries and further estimated the associations between detailed maternal smoke exposures and child IQ, with thorough control for major confounders. It is also important to acknowledge that our study has limitations.

One is the single cotinine measurement from spot urine in mid-pregnancy, which only captures exposure to tobacco smoke in the past 2–4 days. The absence of exposure data from other trimesters prevents us from tracking changes in maternal smoking behavior over time. These measurements were also performed following each cohort’s specific protocols with different LODs. Future studies are warranted to include biomarkers that reflect relatively long-term exposure, such as 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanol or nicotine in hair or nails, measured repeatedly throughout pregnancy using a standardized protocol. In addition, the analytic samples for our two aims differed in sociodemographic characteristics and tobacco smoke exposures, and neither was representative of the U.S. population. We also applied two data driven approaches -- the ROC curve and an EM algorithm for Gaussian mixtures – to determine cotinine cut-offs. Although the resulting thresholds can be applied more broadly to pregnant women given the validation focus of the analysis, they may not be generalizable to other populations, such as adolescents or older adults. The inferential conclusions regarding the association between maternal smoke exposures during pregnancy and child IQ are likely limited to comparable pregnant populations rather than the general U.S. population. Moreover, because self-reported data on other specific smoking locations near home (e.g., balconies, gardens, or yards) and enforcement of complete home smoking bans was unavailable, classification of SHS and THS is likely subject to misclassification.⁷² The differentiation between SHS and THS may be further limited by small subgroup sizes and the lack of specificity in urinary cotinine measurements. Incorporating THS specific markers from participants from nonsmoking household in future studies, such as nicotine residue measured from hand wipe samples,^73,74 furniture surfaces, or indoor air, will help address this research gap.⁷⁵

Conclusion

Using data from the multiple cohorts in ECHO, this study estimated urinary cotinine cut-offs to classify detailed tobacco smoke exposure status in U.S. pregnant women, and it further estimates the associations between tobacco smoke exposure in pregnancy and child IQ. Overall, our study contributes to the evolving science regarding developmental origins of diseases. The identified cotinine cut-offs may have important implications for future epidemiological studies focused on tobacco smoke exposure in pregnancy. Although findings are statistically imprecise, they suggest potential adverse effects of maternal tobacco smoke exposure across varying intensities and sources on child neurodevelopment, which can be used to inform tobacco control policies, regulatory actions, and targeted intervention aimed at reducing smoke-related harms in pregnancy.

Highlights:

1. Classified tobacco smoke exposure (first-, second-, and thirdhand) in U.S. pregnant women.

2. Found insignificant but inverse associations between prenatal smoke exposure and child IQ.

3. Suggested stronger associations in those with higher education or from less deprived areas.

4. Offers key methods for epidemiological studies and insights on developmental origins of diseases.

Availability of data

The ECHO program provides public access to de-identified participant data through periodic releases on the Data and Specimen Hub hosted by the Eunice Kennedy Shriver National Institute of Child Health and Human Development (https://dash.nichd.nih.gov/). The computing code in R can be obtained from the corresponding author via email request.