Inhibitory Control in Siblings Discordant for Exposure to Maternal Smoking during Pregnancy

Abstract

Maternal smoking during pregnancy (SDP) has been linked to poorer offspring executive function across development, but SDP does not occur independent of other familial risk factors. As such, poor and inconsistent control for potential confounds, notably shared familial (i.e., genetic and environmental) confounds, preclude concluding causal effects of SDP on child outcomes. We examined the within-family association between SDP and one component of executive function, inhibitory control, in a sample of families (N=173) specifically selected for sibling pairs discordant for exposure to SDP. Thus, the present study examines if the SDP-inhibitory control association withstands rigorous control for potential child and familial confounds. 79% of the variation in child inhibitory control was attributable to within-family differences and 21% was attributable to differences between families, indicating that the variability in inhibitory control was primarily a function of differences between siblings rather than differences across families. Further, the association between SDP and inhibitory control was fully attenuated when confounds were considered. These findings suggest that co-occurring vulnerabilities act as more salient risk factors for poorer child inhibitory control than SDP and may serve as effective targets of interventions seeking to improve children’s inhibitory control in families with nicotine dependent mothers.

One of the most major, modifiable environmental risks to the developing fetus is prenatal exposure to maternal smoking during pregnancy (SDP). Offspring exposed to SDP have more physical, behavioral, emotional, cognitive, and substance use problems relative to their non-exposed counterparts (Knopik, 2009). As such, SDP is a major public health concern. Due to its relevance to key developmental outcomes (e.g., Snyder, Miyake, & Hankin, 2015) executive function (EF) has emerged as a fundamental neurocognitive outcome potentially influenced by SDP.

Early neurodevelopment sets the stage for later complex cognition. As such, children exposed to SDP may exhibit decreases in later mental development and higher-order capacities, such as EF, resulting from the early insult to fundamental neurodevelopmental processes (Huizink & Mulder, 2006). The foundational and most commonly indexed domains of EF include: (i) inhibitory control; (ii) set-shifting; and (iii) working memory (Best & Miller, 2010; Miyake et al., 2000). Inhibitory control (IC) involves the suppression or delay of a prepotent, salient response for one that is less dominant to achieve a goal and is recruited to, for example, remove the foot from the gas pedal and apply the brake when approaching a yellow light.

To our knowledge, only three studies assessed the association between SDP and IC. In early childhood, exposure to SDP was associated with worse performance on IC tasks requiring children to wait for appealing snacks and toys. However, SDP was not associated with performance on IC tasks requiring children to hold information in mind and inhibit prepotent responses based on that information (Wiebe et al., 2015). In middle childhood, there was no association between SDP and IC assessed by the Stroop Color-Word Interference Test (1935), which requires children to inhibit a prepotent response (i.e., reading words that spell names of colors) for a less dominant response (i.e., naming dissonant ink color that the word is printed in; Cornelius, Ryan, Day, Goldschmidt, & Willford, 2001). Finally, children who were exposed to more severe SDP had poorer IC in a dose-dependent pattern compared to non-exposed children on the Delay Frustration Task (DeFT), which assesses the ability to withhold a response while frustrated (Huijbregts, Warren, de Sonneville, & Swaab-Barneveld, 2008). This study also found no significant group differences in IC on the Sustained Attention Dots Task, requiring children to inhibit a prepotent response (i.e., responding with the non-dominant hand when presented with 3- or 5- dot patterns) for a less-dominant response (i.e., using the dominant hand when presented with 4-dot patterns). Taken together, these findings suggest that SDP—IC associations may be task-specific.

Studies of SDP and IC are muddied by inconsistent control for potential confounds. Prenatal exposure to SDP is correlated with parental and contextual factors that could act as more proximal risk factors for poorer child outcomes (Micalizzi & Knopik, revise and resubmit). Thus, in addition to environmental risk, mothers who smoke during pregnancy may also be more likely to confer genetic risk for poorer functioning to their offspring. For example, mothers who have IC problems themselves may more commonly smoke during pregnancy and pass on correlated genes and environments to the child that are associated with IC problems, arising in a non-causal association between SDP and IC. Therefore, children of mothers who smoke may present with IC problems either because IC problems are caused by SDP, or because SDP exposure and IC problems are both caused by common familial (genetic and environmental) influences. As such, poor and inconsistent control for potential confounders preclude concluding causal effects of SDP on IC (Knopik, 2009).

Studies that account for specific, measured confounds (e.g., socioeconomic status) typically find the relations between SDP and psychological outcomes attenuated, but still significant. Studies that account for general, unmeasured familial confounds (i.e., genetic and environmental), however, tell a more complex story with potentially causal SDP effects for some birth outcomes (e.g., Knopik et al., 2016b; Kuja-Halkola, D’Onofrio, Larsson, & Lichtenstein, 2014) and behavioral outcomes (Gaysina et al., 2013, Knopik et al., 2016a), but complete familial confounding for other behavioral and cognitive outcomes (e.g., Bidwell et al., in press; D’Onofrio et al., 2008, 2010a, 2010b, 2012; Ellingson et al., 2014; Kuja-Halkola et al., 2014; Lambe, Hultman, Torrång, MacCabe, & Cnattingius, 2006; Meier, Plessen, Verhulst, & Mors, 2017; Skoglund, Chen, D’Onofrio, Lichtenstein, & Larsson, 2014). The reasons for this inconsistent pattern of findings are unknown, but may be due, in part, to differences in familial liability for MSDP (Palmer et al., 2016), sampling, outcome assessment (e.g., medical registry data vs. lab-based assessments) and SDP measurement.

To our knowledge, there are no genetically informed studies of SDP and EF. However, there have been genetically informed sibling-comparison studies of SDP and children’s cognitive abilities and academic achievement. Accounting for specific, measured within-family and family average covariates (e.g., maternal educational attainment) and differential exposure to SDP between children within families (i.e., child-specific exposure) attenuates that associations between SDP and most cognitive abilities (e.g., digit span, math, reading, receptive vocabulary, general cognitive abilities/fluid intelligence, reading recognition was the exception; Ellingson et al., 2014; Kuja-Halkola et al., 2014) and SDP and academic achievement (D’Onofrio et al., 2010b; Kuja-Halkola et al., 2014; Lambe et al., 2006). These results contest the notion of causal effects of SDP on many (but not all) cognitive abilities, and instead, suggest that the link is primarily due to familial effects that influence cognitive abilities in both generations. Although previous studies examined the associations between SDP and cognitive outcomes, EF does not entirely overlap with other neurocognitive constructs (e.g., Arffa, 2007). Consequently, we cannot assume that evidence for familial confounding of the association between SDP and other cognitive abilities, such as intelligence (Kuja-Halkola et al., 2014), will extend to EF. However, evidence for familial transmission of risk for IC problems comes from twin studies that reveal genetic influences on individual differences in IC (e.g., Friedman et al., 2008; Polderman et al., 2009).

Taken together, the existing literature underscores the multifaceted nature of the SDP—IC association and that genetically informed designs are required to disentangle co-occurring vulnerabilities for IC problems from true SDP effects. The sibling comparison design is one of several quasi-experimental designs that can be used to strengthen causal inferences regarding environmental risks by ruling out specific forms of confounding, including confounding by gene-environment correlation (D’Onofrio et al., 2013; 2014; Kendler, 2017). The design allows for the evaluation of siblings within families that are discordant for exposure to SDP (i.e., mother smoked [or smoked more] while pregnant with one sibling and did not smoke [or smoked less] while pregnant with the other sibling), and by nature of the design, are matched for confounding familial and prenatal risk factors). Therefore, the present study sought to examine if the association between SDP and child IC withstands rigorous control for confounds in a sample specifically selected for this purpose.

Methods

Participants and Procedure

Data for the current study were drawn from the Missouri Mothers and Their Children study (MO-MATCH; Knopik et al., 2015). The goal of MO-MATCH was to investigate prenatal environmental influences on child attention problems, neuropsychological functioning, and substance use initiation (See Table 3 in Knopik et al., 2015 for a summary of core assessments). Families were identified based on report in birth records (obtained from the Missouri Department of Health and Senior Services Bureau of Health Informatics) that mothers changed smoking behavior between any two pregnancies. Over 4000 mothers were initially identified as changing smoking behaviors between two pregnancies during the years 1998–2005 and screening interviews were conducted with 1520 of these mothers. Of those screened, 27% confirmed what was reported on the birth record and were invited to participate in the study. Families were excluded if: 1) mothers failed to understand the elements of informed consent; 2) English was not the primary language spoken in the home; 3) children had a history of head trauma, neurological disorders, uncorrected visual or auditory acuity deficits; or 4) mothers reported using nicotine substitutes in the ‘nonsmoking’ pregnancy. Following consent, formal diagnostic interviews were completed with 173 families (344 pregnancies). Of the participating families, 96 fathers provided data.

We examined possible differences between families where fathers did versus did not participate using Wilcoxon-Mann-Whitney tests (i.e., non-parametric analog to independent samples t test) on study variables (i.e., marital status, maternal age, maternal employment status, maternal education, maternal IC, SDP severity; and IQ and IC for child 1 and child 2 separately). Three differences (of 10 tests) emerged. First, mothers were slightly older in families where fathers participated than in families where fathers did not participate (χ²=25.09, p<.05). Second, families where fathers participated were more likely to have a “married” status (96 and 52% among families with and without participating fathers, respectively), whereas families with fathers who did not participate had a higher proportion of “divorced” (2 and 32% in families with and without participating fathers, respectively), “separated” and “widowed” mothers (χ²=12.89, p<.05). Third, child IQ was higher in families where dads participated for child 2 (χ²=3.89, p=.05; child 2 mean IQ= 90.8 vs. 74.8 in families where fathers did and did not participate, respectively).

Mothers completed diagnostic interviews about their pregnancies and child mental health and behavioral history. Both parents provided information about their own mental health history. Assessments occurred when children were 7–16 years old (Child 1 [older sibling] average age=13.02, standard deviation [SD]=1.93, 54% male; Child 2 [younger sibling] average age=10.22, SD=1.80, 51% male). Parents were primarily of Caucasian ancestry (96%, n=250). See Table 1 and Knopik et al (2015) for additional details on the sample. The Institutional Review Boards of Rhode Island Hospital, Washington University and the State of Missouri Department of Health and Senior Services approved the study, entitled, “Prenatal tobacco exposure: Effects on neuropsychological outcomes and ADHD” (protocol number 211020–20).

Table 1.

Sample characteristics

	Child 1			Child 2
	N	Mean	SD	N	Mean	SD
Study Variables
Smoking during pregnancy severity	173	3.95	2.05	167	2.04	1.77
Inhibitory control	164	9.97	3.15	164	10.44	2.58
Covariates
Maternal age at birth	155	26.55	5.46	160	29.22	5.66
Maternal education (in years) at birth	155	13.31	2.07	160	13.48	1.90
Second-hand smoke exposure by fathers	164	1.82	1.44	157	1.59	1.43
Full Scale IQ	166	100.14	12.64	168	100.04	12.54
	N	Mean	SD
Maternal inhibitory control	168	10.20	2.73
	N	%		N	%
Marital status (percent married) at birth	155	85%		160	83%
Food stamp usage at birth	149	9%		151	13%
Family Demographics (at assessment)	N	Mean	SD
Maternal age	162	39.83	5.62
Paternal age	80	44.04	6.34
Child 1 age	166	13.02	1.93
Child 2 age	166	10.22	1.80
	Maternal			Paternal
	N	%		N	%
Education
Less than high school	7	4%		9	10%
high school	30	18%		19	20%
1–2 years college	50	30%		14	15%
3–4 years college	46	27%		17	18%
More than college	29	17%		21	22%
Not reported	7	4%		14	15%
Mothers’ marital status
Never married	6	4%
Married	130	77%
Separated	5	3%
Divorced	26	15%
Widowed	2	1%

Note. Total Ns vary due to randomly missing data.

Measures

SDP

Maternal report of SDP was obtained using a modified version of the Missouri Assessment of Genetics Interview for Children—Parent on Child (Todd et al., 2003). An investigation of these data (Knopik et al., 2016b) compared the predictive ability of maternal report of SDP relative to both birth record report and paternal report of SDP and revealed that retrospective maternal SDP, both any SDP (absent/present) and quantity smoked, was found to be the best assessment of SDP. Thus, to be consistent with prior reports, we focus here on maternal report of SDP. As there is literature suggesting different, and potentially more harmful, effects of SDP later into pregnancy (Dwyer, McQuown, & Leslie 2009; Hebel, Fox & Sexton, 1988), we incorporated timing of exposure into the computed SDP variable. Therefore, the following items were used to create an overall SDP severity score: Any SDP (0=No, 1=Yes) across each pregnancy as a whole, as well as specific to each trimester, and overall quantity smoked during pregnancy was assessed via mothers’ estimate of the number of cigarettes smoked in each trimester. A single SDP severity score was created for each child (Knopik et al., 2016a):

1. did not smoke during pregnancy

2. smoked during first trimester only, 1–10 cigarettes per day

3. smoked during first trimester only, 11–19 cigarettes per day

4. smoked during first trimester only, 20+ cigarettes per day

5. smoked beyond first trimester, 1–10 cigarettes per day (max of all trimesters)

6. smoked beyond first trimester, 11–19 cigarettes per day (max of all trimesters)

7. smoked beyond first trimester, 20+ cigarettes per day (max of all trimesters)

IC

Parent and child IC were assessed with the inhibition condition of the Color-Word Interference Test on the Delis-Kaplin Executive Function System, a reliable and valid assessment (Delis, Kaplan & Kramer, 2001a, 2001b). Based on the Stroop (1935), the inhibition interference measure requires participants to verbally report the ink color of words that spell a dissonant color word. For example, if the word ‘green’ is printed in red ink, the participant must say ‘red’ rather than ‘green.’ The task requires inhibition of an automatic response (i.e., reading) in order to generate a conflicting response (i.e., naming the dissonant ink color; Delis et al., 2001a). Participants completed two baseline conditions, basic naming of color patches and basic reading of words that denote colors printed in black ink, before proceeding to 10 practice items. The task was discontinued if participants required 4 corrections. The testing phase included 50 color words printed in dissonant ink colors. If participants made three consecutive errors in the testing phase, they were prompted to name the ink color once. Raw completion time in seconds (maximum 180 seconds) were normed (mean=10, SD=3) and corrected for the appropriate age group. Higher scaled scores indicate better performance.

Child and Familial Covariates

Maternal, familial and child characteristics that could confound the association between SDP and IC were selected a-priori, on theoretical grounds, and are consistent with prior genetically informed studies of SDP and child outcomes (e.g., D’Onofrio et al., 2012; Knopik et al., 2016a). These comprised: maternal report of her marital status, age, and education at the birth of each child, qualification for food stamps (yes/no) at the time of delivery (collected from birth records), maternal IC, birth order¹, sex, child IQ (Wechsler, 2003), and second-hand smoke exposure during pregnancy (by the father). Missing data for covariates was low, ranging from .5% for sex to 11% for food stamps. See Appendix 1 for intercorrelations among study variables.

APPENDIX 1.

Intercorrelations among study variables by child.
	1	2	3	4	5	6	7	8	9	10	11	12	13	14	15	16	17
1. IC	1	−.13	−.05	.08	.29a	.05	.12	.15	.14	.18*	−.09	−.09	.01	−.16 c	.22b	.09	−.06
2. SDP (CS)	−.04	1	−.15	−.20b	−.14	−.05	−.17 c	−.02	−.21b	−.07	.12	.19*	.29a	.15	−.06	.03	.01
3. SDP (FA)	−.06	.15	1	.03	−.01	−.13	−.02	−.05	.09	.04	−.08	−.04	−.01	.20b	−.19 c	−.18 c	−.06
4. IQ (CS)	.19c	−.20b	−.03	1	−.01	.04	−.02	<.01	.09	−.01	.17 c	<.01	−.04	−.02	−.04	−.03	.08
5. IQ (FA)	.41a	.14	−.01	.01	1	<.01	.38a	.01	.19 c	.22b	−.1	−.18 c	−.14	−.09	.19 c	.03	.06
6. Mother EDU (CS)	<.01	−.05	.13	.04	.01	1	−.15	.1	.03	.03	.01	−.03	−.04	−.08	.07	.04	−.07
7. Mother EDU (FA)	.01	.17b	−.10	.02	.39a	.16 c	1	.05	.52a	.25b	−.1	−.28a	−.1	−.18 c	.13	.07	.17 c
8. Mother Age (CS)	.12	−.01	.03	−.01	.06	.10	.03	1	.12	.28a	−.13	−.22b	<.01	−.05	.17 c	−.06	.13
9. Mother Age (FA)	−.05	.21b	.09	−.09	.19 c	−.03	.52a	−.09	1	.29a	−.09	−.22b	−.20 c	−.24b	.11	.17 c	.15
10. Mother Marital Status	−.05	.05	.05	−.01	.19 c	−.02	.20 c	−.22b	.26 b	1	−.13	−.37a	−.04	.05	.04	−.04	.11
11. Food Stamps (CS)	−.05	.12	.08	.18 c	.10	.01	.10	−.14	.09	.15	1	.15	.13	.03	−.08	.02	−.05
12. Food Stamps (FA)	.01	−.19 c	−.04	<.01	−.19 c	.03	−.28a	.22b	−.22b	−.31a	−.15	1	<.01	.08	−.15	−.01	.01
13. SHSE (CS)	.01	.29a	.02	−.03	.15	−.03	.11	.01	.22b	.04	.14	−.04	1	−.02	−.03	−.06	−.02
14. SHSE (FA)	−.02	−.15	.20b	.02	−.11	.08	−.16 c	.04	−.24b	.08	−.03	.08	.01	1	−.17c	−.25b	−.1
15. Mother IC	.30a	.06	−.19 c	.04	.18 c	−.06	.13	−.14	.12	<.01	.08	−.16	.03	−.18 c	1	.03	−.04
16. Sex (CS)	.07	.03	.18 c	−.03	−.03	.04	−.07	−.05	−.17 c	.05	.02	.01	−.05	.25b	−.03	1	<.01
17. Sex (FA)	.03	−.01	−.06	−.08	.04	.09	.17 c	−0.05	.16 c	.05	.06	<.01	.01	−.11	−.06	<.01	1

Note. Child 1 correlations are below the diagonal; child 2 correlations are above the diagonal. IC=Inhibitory control; SDP=smoking during pregnancy; CS= child specific; FA= family average; EDU= education; SHSE=second-hand smoke exposure

^ap<.001;

^bp<.01;

^cp<.05

Statistical Analysis

The sibling comparison approach (e.g., D’Onofrio et al., 2008), involved fitting a series of hierarchical linear models (HLM) executed using SAS PROC MIXED in order to assess the within-family and family average associations between SDP and IC while accounting for non-independence of the data. Prior to fitting the HLM models, an unconditional ‘intercept-only’ model was fit to the data to decompose the variance in the IC measure into within-family (i.e., individual child-level) and family level variation. The intraclass correlation (ICC; family level variance/[family level variance+individual child-level variance]) was used to describe the initial proportion of the within-family variance relative to total variance (Raudenbush & Byrk, 2001; Singer & Willett, 2003). The unconditional model provided a baseline against which subsequent models were compared in order to understand how much additional within-family (and, therefore, potentially causal) variance was explained by each conditional model.

Standard Models

We then fit a series of models for IC: 1) standard model without covariates compared children whose mothers smoked (or smoked more) during pregnancy to those whose mothers who did not smoke (or smoked less) on IC; 2) standard model with covariates, which is the same as model 1, but statistically accounts for potential confounds. Thus, the standard model with covariates examines SDP—IC associations in the entire sample and is representative of how SDP and associated confounds are typically modeled in non-sibling samples, without capitalizing on the sibling comparison aspect of the data. While the hierarchical nature of this model adjusts for the dependency due to siblings nested within family, it does not distinguish between whether SDP is operating at a within-family level (i.e., contributing to differences in IC in one sibling versus another, within families) or family level (i.e., contributing to differences in overall, average levels of siblings’ IC across families). These questions are addressed by the sibling comparison models.

Sibling Comparison Models

Two sibling comparison models were conducted. The 3) sibling comparison without covariates model assessed both within-family and family-average associations between SDP and IC, allowing for a direct test of unique SDP exposure effects on child IC; and 4) sibling comparison with covariates model included potential covariates. Therefore, the sibling comparison with covariates model specifically examined within-family associations of SDP and IC, allowing for a direct test of unique SDP exposure effects on IC while controlling for genetic and environmental variables that siblings share, as well as the family-average analog to traditional research (and the Standard models; Ellingson et al., 2014; Knopik et al., 2016a, 2016b; Lahey & D’Onofrio, 2010).

Two variables were used to capture SDP severity in the sibling comparison model with covariates. Family average SDP severity for each family was the average score for SDP severity across both siblings. Child-specific SDP severity relative to family average for each child was the resulting value when the family average SDP was subtracted from each child-specific SDP severity score (i.e., the SDP severity scores used in the Standard models). Thus, if a mother smoked the same amount for both pregnancies, regardless of total severity, then both siblings would have a child-specific SDP severity relative to family average of zero. In families where mothers changed her smoking behavior from one pregnancy to another, the sibling for whom mothers smoked, or smoked more, would have a positive score, whereas the sibling for whom mothers did not smoke, or smoked less, would have a negative score of the same magnitude (Knopik et al., 2016b).

The effect of the child-specific SDP severity relative to family average on IC assessed the potentially causal within-family effect of SDP on IC (comparing across siblings within a family, a test of any unique effect of SDP on child specific outcomes beyond familial and genetic factors that siblings share). In the sibling comparison models, the child-specific relative to family average SDP severity score was entered as a level 1 predictor, and the family average SDP severity score was entered as a level 2 predictor. In sum, we fit two standard models (i.e., one without covariates and one with covariates) and two sibling comparison models (i.e., one without covariates and one with covariates).

Results

Table 2 presents parameter and variance estimates for all models. Decompositions of the variance in the unconditional model revealed that 79% of the variation in IC was attributable to within-family differences and 21% was attributable to family level differences, suggesting that the variability in IC was primarily a function of differences between siblings rather than differences across families. Child IQ and maternal IC were significant predictors of child IC in each model that included covariates². In the standard model without covariates, higher severity of SDP was associated with worse IC, suggesting a main effect of SDP when child and familial confounds are not controlled. This effect was no longer significant, however, after accounting for these confounds (i.e., standard model with covariates). In the sibling-comparison without covariates model, only the within-family association (i.e., child-specific SDP) was significant, suggesting an association between SDP and IC. However, this relationship did not hold after controlling on child and familial confounds (i.e., sibling-comparison with covariates model).

Table 2.

Parameter estimates (95% confidence intervals) and variances

	Model
	Unconditional	Standard without covariates	Standard with covariates	Sibling-comparison without covariates	Sibling-comparison with covariates
	b (CI)	b (CI)	b (CI)	b (CI)	b (CI)
Intercept	10.20b (9.86, 10.55)	10.72 (10.24, 11.19)b	−.06 (−3.65, 3.53)	10.67 (9.59, 11.74)b	0.31 (−3.76, 4.39)
SDP
SDP (CS)		−0.18 (−0.32, −0.03)c	−0.08 (−0.25, 0.08)	−0.17 (−0.32, −0.02)c	−0.02 (−0.22, 0.18)
SDP (FA)				−0.18 (−0.51, 0.16)	−0.18 (−0.51, 0.15)
Controls
Child			0.50 (−0.19, 1.19)		−0.31 (−1.62, 1.01)
Mother education (CS)			−0.10 (−0.29, 0.10)		−0.03 (−0.48, 0.41)
Mother education (FA)					−0.11 (−0.33, 0.11)
Mother age (CS)			−0.02 (−0.09, 0.05)		0.31 (−0.07, 0.69)
Mother age (FA)					−0.02 (−0.09, 0.05)
Mother marital status			−0.48 (−1.48, 0.52)		−0.44 (−1.47, 0.59)
Food stamps (CS)			−0.39 (−1.49, 0.71)		−1.36 (−3.34, 0.62)
Food stamps (FA)					0.13 (−1.22, 1.48)
Second-hand smoke exposure (CS)			0.03 (−0.20, 0.26)		0.05 (−0.37, 0.47)
Second-hand smoke exposure (FA)					0.07 (−0.21, 0.34)
IQ (CS)			0.09 (0.06, 0.12)b		0.09 (0.04, 0.14)b
IQ (FA)					0.1 (0.07, 0.14)b
Mother inhibitory control			0.18 (0.05, 0.30)b		0.19 (0.06, 0.32)b
Sex (CS)			0.52 (−0.08, 1.13)		0.79 (−0.03, 1.6)
Sex (FA)					0.19 (−0.74, 1.13)
Variance
Family level	1.74b	0	0	1.77b	1.03
Individual-level	6.58b	5.93b	5.66b	6.56b	5.20b
% Within-family variance explained beyond unconditional model		9.88%	13.98%	0.3%	20.97%

Note. SDP=smoking during pregnancy; b= unstandardized beta weights; CI=95% confidence interval; CS= child specific; FA= family average.

^bp<.01;

^cp<.05.

In order to quantify how much of the total within-family variance is explained by each conditional model above and beyond the unconditional model, we computed the percentage of within-family variance explained (unconditional individual child-level variance–conditional [e.g., standard model with covariates] individual child-level variance)/unconditional individual child-level variance; Singer, 1998). SDP alone explained 9.88% and 0.3% of the variance in IC in the standard and sibling comparison models without covariates, respectively (Table 2). The sibling comparison with covariates model explained approximately 21% of the variance in IC above and beyond the unconditional model, indicating that child- and family level confounders explain the most additional variance in IC.

To examine if the pattern of findings was consistent in families where fathers did and did not participate, all analyses were also conducted in the subsample of families (n=96) where both parents participated (and included father IC as a covariate). The pattern of findings for SDP were consistent with those presented here (see Appendix 2), however, maternal IC was no longer a significant predictor of child IC. Further, in this subsample of families, the sibling comparison model with covariates model explained less of the variance in IC than in the full sample (19 vs 21%).

APPENDIX 2.

Parameter estimates (95% confidence intervals) and variances in subsample of families in which dad participated

	Model
	Unconditional	Standard without covariates	Standard with covariates	Sibling-comparison without covariates	Sibling-comparison with covariates
	b (CI)	b (CI)	b (CI)	b (CI)	b (CI)
Intercept	10.21 (9.71, 10.71)b	10.80 (10.14, 11.46)b	−3.0 (−8.26, 2.26)	9.93 (8.34, 11.53)b	−3.63 (−9.84, 2.57)
SDP
SDP (CS)		−0.21 (−0.41, −0.01)c	−0.04 (−0.20, 0.28)	−0.22 (−0.42, −0.01)c	<.01 (−0.29, 0.30)
SDP (FA)				0.07 (−0.42, 0.56)	0.14 (−0.37, 0.65)
Controls
Child			0.79 (−0.17, 1.75)		−0.30 (−2.13, 1.54)
Mother education (CS)			−0.11 (−0.38, 0.15)		−0.19 (−0.74, 0.36)
Mother education (FA)					−0.13 (−0.46, 0.19)
Mother age (CS)			−0.03 (−0.07, 0.14)		0.31 (−0.19, 0.81)
Mother age (FA)					0.03 (−0.09, 0.14)
Mother marital status			−0.90 (−2.61, 0.81)		−1.55 (−3.52, 0.42)
Food stamps (CS)			−1.29 (−3.04, 0.45)		−1.10 (−3.75, 1.56)
Food stamps (FA)					−1.18 (−3.71, 1.35)
Second-hand smoke exposure (CS)			−0.01 (−0.33, 0.30)		−0.26 (−0.81, 0.30)
Second-hand smoke exposure (FA)					0.19 (−0.22, 0.60)
IQ (CS)			0.11 (0.07, 0.15)b		0.09 (0.02, 0.17)c
IQ (FA)					0.13 (0.08, 0.19)b
Mother inhibitory control			0.07 (−0.12, 0.26)		0.11 (−0.10, 0.31)
Father inhibitory control			0.15 (−0.01, 0.31)		0.11 (−0.06, 0.28)
Sex (CS)			−0.02 (−0.84, 0.81)		0.36 (−0.78, 1.50)
Sex (FA)					−0.13 (−1.43, 1.17)
Variance
Family level	2.27c	0	0	2.51b	1.33
Individual-level	7.29b	6.01b	5.77b	6.97b	5.91b
% Within-family variance explained beyond unconditional model		17.56%	20.85%	4.39%	18.93%

Note. SDP=smoking during pregnancy; b= unstandardized beta weights; CI=95% confidence interval; CS= child specific; FA= family average.

^bp<.01;

^cp<.05.

Discussion

In a sample of families specifically selected for sibling pairs discordant for exposure to SDP, we conducted the first genetically informed study of the association between SDP and child IC. An advantage of the current study is the inclusion of detailed covariates, particularly the ability to control for maternal IC. Additionally, the sibling comparison approach controls for unmeasured genetic and environmental influences that siblings share, enabling us to explore if SDP has a direct effect on child IC. The findings indicate that an initial SDP-IC association was fully attenuated after controlling for confounding factors.

Non-genetic studies yield mixed results with regard to the effects of SDP on IC (Cornelius et al., 2001; Huijbregts et al., 2008; Wiebe et al., 2015). An important consideration in studies of the SDP—IC association is that IC is often differentiated into ‘hot’ and ‘cool’ aspects (Zelazo & Müller, 2002). Hot IC is recruited under emotionally laden, motivated circumstances, whereas cold IC is elicited under decontextualized, neutral conditions. Because hot and cold IC are not entirely overlapping constructs (Zelazo & Müller, 2002), it is possible that SDP differentially impacts one IC process but not the other. The existing non-genetic studies that leverage the hot vs. cool IC distinction support this notion: hot IC problems were linked to SDP exposure, but cool IC problems were not. Our findings add further support to the notion that SDP is not associated with cool IC problems. It should be noted that when hot IC problems were observed in children who were exposed to SDP prenatally, they also presented for higher conduct problems and hyperactivity-inattention scores (Huijbregts et al., 2008). Emotion dysregulation and motivation deficits (i.e., those that are related to hot IC) are prevalent in individuals with ADHD (Shaw, Stingaris, Nigg, & Leibenluft, 2014; Volkow et al., 2010). Thus, it is possible that the SDP—hot IC relation may withstand control for confounding factors, as has been seen for some genetically informed assessments of the SDP—externalizing association (e.g., Gaysina et al., 2013; Knopik et al., 2016a), but this is an open empirical question. Because hot IC was not assessed in MO-MATCH, we were not able to examine the association between SDP and hot IC. Future genetically informed studies should explore if the associations between SDP and hot IC documented in non-genetic studies withstand control for child- and family level confounds. Further, as one of the most widely-invoked constructs across many literatures, the operationalization and measurement of EF is critical to the inferences that can be made about these competencies (Toplak, West, & Stanovich, 2013). Although the present study revealed confounding of the association between SDP and cool IC, the components of EF are not interchangeable. As such, future genetically informed studies should examine if these findings are consistent for different components of EF, such as set-shifting. Further, because IQ was a significant predictor of IC in each model that included covariates, a future direction is to explore the links between SDP and IQ in the MO-MATCH sample.

The following limitations should be considered when interpreting the present results. First, although retrospective report of SDP appears to be reliable and accurate in this and other samples (e.g., Estabrook et al., 2015; Knopik et al., 2016b), our results hinge on the SDP assessment to accurately reflect the amount of SDP exposure. There is a small but growing literature suggesting that retrospective recall of substance use, even over a number of years, is valid when better options (e.g., prospective cotinine levels) are not available, and are generally more accurate that the maternally reported information recorded on medical records (see Knopik et al., 2016b for a discussion). Second, the SDP severity measure assumes that smoking beyond the first trimester is more harmful than smoking only in the first trimester. Whereas this is supported in some preclinical and human studies (e.g., Dwyer et al., 2009; Hebel et al., 1988), there are mixed findings about the adverse effects of timing of exposure on measures of EF (see Micalizzi & Knopik, revise and resubmit for a discussion). However, sensitivity analyses revealed that the present findings were robust to measures of SDP that do not make assumptions about the timing of exposure (see footnote 1 in Knopik et al., 2016a for descriptions of and intercorrelations among SDP variables used in the present sensitivity analyses). Third, despite a carefully designed study that was built to target siblings discordant for SDP exposure, we could not measure or include all covariates that differ between siblings. As such, there are likely confounds that were not considered here that influence sibling comparison (see D’Onofrio et al., 2010). Fourth, because only 50% of the sample included fathers, the results of the main analysis did not include father confounds. Therefore, it is unknown if the findings would be consistent if the fathers of the other half of the sample were included. Because the pattern of findings for SDP was consistent in the subsample of families that included fathers and the Wilcoxon-Mann-Whitney tests did not reveal differences in child IC in families where fathers did and did not participate, it is expected that the findings would not change. Nonetheless, future studies should aim to identify and measure relevant father-specific confounds that may play a role in EF. Finally, only one measure of IC was included in the present study. Measures of EF are not interchangeable (Toplak et al., 2013). Therefore, it is unknown whether these findings would hold across different measures of IC.

In sum, our findings are consistent with and expand existing genetically informed literature that indicates familial confounding for most cognitive phenotypes (e.g., Ellingson et al., 2014). Co-occurring vulnerabilities appear to act as more salient risk factors for poorer child cool IC than SDP and may serve as effective targets of interventions seeking to improve children’s IC in families with nicotine dependent mothers. This should not diminish concern regarding SDP, but instead, advance our knowledge of what factors are and are not potential causes of EF problems (Knopik, 2009).