smoking during pregnancy and offspring externalizing ... 20081.pdf · with offspring adhd. the...

Smoking during pregnancy and offspringexternalizing problems: An explorationof genetic and environmental confounds

BRIAN M. D’ONOFRIO,a CAROL A. VAN HULLE,b IRWIN D. WALDMAN,c

JOSEPH LEE RODGERS,d K. PAIGE HARDEN,e PAUL J. RATHOUZ,b AND

BENJAMIN B. LAHEYb

aIndiana University; bUniversity of Chicago; cEmory University; dUniversity of Oklahoma; andeUniversity of Virginia

Abstract

Previous studies have documented that smoking during pregnancy (SDP) is associated with offspring externalizingproblems, even when measured covariates were used to control for possible confounds. However, the association may bebecause of nonmeasured environmental and genetic factors that increase risk for offspring externalizing problems. Thecurrent project used the National Longitudinal Survey of Youth and their children, ages 4–10 years, to explore the relationsbetween SDP and offspring conduct problems (CPs), oppositional defiant problems (ODPs), and attention-deficit/hyperactivity problems (ADHPs) using methodological and statistical controls for confounds. When offspring werecompared to their own siblings who differed in their exposure to prenatal nicotine, there was no effect of SDP onoffspring CP and ODP. This suggests that SDP does not have a causal effect on offspring CP and ODP. There was asmall association between SDP and ADHP, consistent with a causal effect of SDP, but the magnitude of the associationwas greatly reduced by methodological and statistical controls. Genetically informed analyses suggest that unmeasuredenvironmental variables influencing both SDP and offspring externalizing behaviors account for the previouslyobserved associations. That is, the current analyses imply that important unidentified environmental factors accountfor the association between SDP and offspring externalizing problems, not teratogenic effects of SDP.

Smoking during pregnancy (SDP) has beenconsistently linked with externalizing problemsin offspring, particularly in males (reviews inCnattingius, 2004; Huizink & Mulder, 2006;Wakschlag & Hans, 2002; Wakschlag, Pickett,Cook, Benowitz, & Leventhal, 2002). It hasbeen associated with parent-reported conductproblems (CPs; Ernst, 2001), arrest history fromnational crime registries (Brennan, Grekin, &Mednick, 1999; Rasanen et al., 1999), contact

with police obtained by city police records(Gibson, Piquero, & Tibbets, 2000), opposi-tional defiant disorder (Wakschlag & Keenan,2001), conduct disorder (Fergusson, Woodward,& Horwood, 1998; Wakschlag & Hans, 2002;Wakschlag & Keenan, 2001; Wakschlag et al.,1997; Weissman, Warner, Wickramaratne, &Kandel, 1999), and attention-deficit / hyperactivitydisorder (Mick, Biederman, Faraone, Sayer, &Kleinman, 2002; Rodriguez & Bohlin, 2005).

Reviews of the literature note that the asso-ciation is consistent with a causal connectionbecause the association is specific to externaliz-ing problems, has been found across diversesamples and measures, demonstrates a dose–response relationship, and is consistent withfindings from basic research (Cnattingius,

Address correspondence and reprint requests to: BrianD’Onofrio, Department of Psychological and Brain Scien-ces, Indiana University, 1101 E. 10th Street, Bloomington,IN 47405; E-mail: [email protected].

This research was supported by Grant R01-MH070025 toBenjamin B. Lahey. We thank Valerie Knopik for her help-ful comments on an earlier version of the manuscript.

Development and Psychopathology 20 (2008), 139–164Copyright # 2008 Cambridge University PressPrinted in the United States of AmericaDOI: 10.1017/S0954579408000072

139

2004; Wakschlag et al., 2002). The interest inSDP as a cause of externalizing problems isdriven, at least in part, by the possibility ofidentifying a risk factor that is amenable to in-tervention. Many researchers, however, havenoted that methodological considerations ren-der causal interpretations impossible at thistime because children cannot be randomly as-signed to nicotine exposure conditions; the asso-ciation between SDP and offspring behaviormay be because of risk factors associated withSDP rather than the influence of prenatal nico-tine exposure (Fergusson, 1999; Wakschlag etal., 2002). Although associations between SDPand offspring characteristics may be becauseof more than just nicotine (Huizink & Mulder,2006), we use the phrase “prenatal nicotine ex-posure” to refer to SDP throughout the articlefor ease of presentation.

It is important to note that SDP is correlatedwith many factors that are also correlated withexternalizing problems is children. SDP is cor-related with low socioeconomic status (Mat-thews, 2001; Zimmer & Zimmer, 1998), earlyage of pregnancy (Cnattingius, 2004; Zimmer& Zimmer, 1998), race (Zimmer & Zimmer,1998), prenatal care (Zimmer & Zimmer, 1998),maternal depression (Breslau, Kilbey, & An-dreski, 1993), history of maternal delinquency(Fergusson, 1999), paternal antisocial behavior(Maughan, Taylor, Caspi, & Moffitt, 2004),and the family environment (Brook, Brook, &Whiteman, 2000), just to name a few possibleconfounds to the association between SDPand offspring externalizing problems. Moststudies have relied on statistical controls ofpossible confounds by including measured co-variates, such as parental antisocial behavior,parenting practices, family cohesion, and socio-economic status, in the analyses. Researchershave also utilized case–control studies (Mick etal., 2002). Overall, associations between SDPand offspring externalizing behaviors have re-mained significant in most of the extant SDPstudies when the measured covariates are includedin the analyses (Cnattingius, 2004; Wakschlaget al., 2002). Some parental variables, however,such as ongoing exposure to secondhand smoke(Maughan, Taylor, Taylor, Butler, & Bynner,2001), maternal report of conduct disorder asa teenager (but not adult externalizing prob-

lems; Silberg et al., 2003), and antisocial be-havior of both parents (Maughan, Taylor,Caspi, & Moffitt, 2004), mediate a majorityof the relation between SDP and CP.

In addition to the numerous potential envi-ronmental risk factors associated with SDP, anumber of researchers have noted that geneticconfounds may mediate the relationship betweenSDP and offspring externalizing problems(D’Onofrio et al., 2003; Fergusson, 1999;Moffitt, 2005; Silberg et al., 2003; Wakschlaget al., 2002). Mothers who smoke during preg-nancy may pass down genetic risk for exter-nalizing problems to their offspring, a form ofpassive gene–environment correlation (r GE;Scarr & McCartney, 1983). Passive r GE occurswhen genetic factors common to both the par-ent and the offspring are correlated with oneor more measures of the family environment.Genetic risk for antisocial behavior may con-tribute to a mother’s likelihood of SDP andmay confer vulnerability to externalizing prob-lems when inherited by offspring. Standard cor-relational research approaches, in which envi-ronmental and genetic risks are confounded,are ineffective in delineating whether associa-tions between externalizing problems andSDP are because of environmental causationor passive r GE (D’Onofrio et al., 2003; Mof-fitt, 2005; Rutter, Pickles, Murray, & Eaves,2001). Moreover, most of the genetically in-formed studies of SDP and externalizing prob-lems in offspring (Button, Thapar, & McGuf-fin, 2005; Knopik et al., 2005; Maughanet al., 2004; Thapar et al., 2003) have used anapproach, including a measure of SDP in a stan-dard twin study, that assumes the relation ispurely environmental (Purcell & Koenen, 2005;Turkheimer, D’Onofrio, Maes, & Eaves, 2005).Therefore, previous studies have been unableto explore the possibility that passive r GEaccounts for observed relations between SDPand externalizing.

The standard twin design may be of limitedutility in differentiating between causation andpassive r GE (Thapar et al., 2003), but alterna-tive behavior genetic designs are effective inexamining the processes underlying intergen-erational associations (see review by Rutter,Pickles, Murray, & Eaves, 2001). In particular,the children of twins (CoT) design is well

B. M. D’Onofrio et al.140

suited for studying SDP because the approachcan delineate between environmental pro-cesses related specifically to a risk factorshared by siblings, genetic transmission fromparents to their offspring, and environmentalconfounds that vary between families (reviewsin D’Onofrio et al., 2003, 2005; Gottesman &Bertelsen, 1989; Heath, Kendler, Eaves, &Markell, 1985; Rutter et al., 2001). D’Onofrioand colleagues (2003) explored the associationbetween SDP and offspring birth weight as amodel system and found that genetic con-founds did not mediate the intergenerationalassociation. Recently, Knopik et al. (2006)utilized the CoT design to explore the associa-tion between parental alcoholism and SDPwith offspring ADHD. The analyses suggestthat the association between parental alcoholdependence and offspring ADHD is geneti-cally mediated and that genetic risk transmit-ted from parents to their offspring accountsfor a significant portion of the SDP-offspringADHD relation.

The current article explores the associationof SDP and offspring externalizing in a sampleof women from the National Longitudinal Sur-vey of Youth (NLSY79) and their children(CNLSY), a sample that confers many advanta-ges. The sample is a nationally representativehousehold sample of women and their children,and the survey included measures related toCPs, oppositional defiant problems (ODPs),and attention-deficit/hyperactivity problems(ADHPs). Previous analyses in the NLSY79have documented characteristics of families re-lated to SDP (Zimmer & Zimmer, 1998), andthe associations between SDP and offspringhealth, behavior problems, and academic achieve-ment have been illustrated in offspring of theNLSY79 (Li & Poirier, 2003).

Finally, the sample provides the opportu-nity to delineate genetic and environmentalprocesses related to SDP and offspring exter-nalizing. There is a history of using theNLSY79 to explore causal mechanisms usingdifferences within and between families, thetheoretical basis of the approach used in thisarticle. For example, researchers have used thedata set to explore the relation between birthorder and intelligence (Wichman, Rodgers, &MacCallum, 2006), maternal alcohol and illicit

drug use and offspring psychopathology (Chat-terji & Markowitz, 2001), teenage childbearingand the women’s later adjustment (Geronimus& Korenman, 1992), and maternal age at firstbirth and offspring adjustment (Turley, 2003).

Multiple adult females from the same house-hold and multiple offspring per mother were in-cluded in the NLSY79 and CNLSY studies.This clustering of data permits the associationbetween SDP and offspring problems to be de-composed in two ways. First, children exposedto greater SDP can be compared to their ownsiblings exposed to more or less SDP. Thiswithin-mother effect controls for genetic andenvironmental factors shared by children withthe same mother, whether they are full or halfsiblings. If SDP causes offspring externalizingproblems, the association would be presentwhen siblings who differ in their exposure toprenatal nicotine are compared (e.g., Rodgers,Cleveland, van den Oord, & Rowe, 2000). Sec-ond, the externalizing problems of children ofadult siblings who differ in their average levelof SDP can be compared. The adult NLSY79sample was originally based on householdsand included sibling pairs, referred to as theNLSY household level in the analyses. Theclustering allowed children exposed to moreSDP to be compared to their cousins with lessexposure to prenatal nicotine. This comparisonis a within-adult–sibling effect. Furthermore,because the NLSY79 has adult sibling pairswho differ in their genetic relatedness, thiscousin comparison can be analyzed contingenton the genetic risk associated with SDP. TheNLSY is a very powerful design for such pur-poses because it contains information on ge-netic relatedness across two generations: forthe original NLSY79 youth (Rodgers, Buster,& Rowe, 2001; Rodgers, Rowe, & Buster,1999) and the children of the mothers of theNLSY (CNLSY; Rodgers, Rowe, & Li,1994; Rodgers, Rowe, & May, 1994). Theseanalyses are an extension of the CoT Design,taking advantage of multiple genetically infor-mative groups in the NLSY79 and CNLSY.The methodological controls provided byfamily relationships in this design were alsocombined with traditional statistical controlsprovided by the inclusion of measured familycharacteristics, thus providing a rigorous,

Smoking during pregnancy and offspring externalizing problems 141

quasiexperimental test of whether associationsbetween SDP and externalizing problems arecausal.

Method

Sample

Mother generation sample. The NLSY79 sur-vey was funded by the Bureau of Labor Statis-tics as a comprehensive study of the future USworkforce. It included a nationally representa-tive sample of 6,111 14- to 21-year-old youthswho were not in the military, as well as a sup-plemental oversample of 3,652 African Ameri-can and Hispanic youth. The current study used6,283 females from the combined sample, be-cause there is cross-generational informationonly for the females and their children. Theprobability sample for the NLSY79 was se-lected using a stratified and clustered design.Primary sampling units consisting of standardmetropolitan statistical areas and countieswere selected randomly, proportionate to popu-lation size. Smaller units (census block groupsor enumeration districts) within each primarysampling unit were then randomly selected,with households randomly selected in the thirdor fourth step. In the initial NLSY79 assess-ment, the response rate was 90%. Participantswere reinterviewed annually from 1979 through1994 and every 2 years since then. Retentionrates during follow-up assessments were 90%or better during the first 16 waves and havesince stayed above 80%.

Of the 6,283 women in the NLSY79 sample,4,886 had given birth to at least one child by the2002 report. The sample was racially diverse:

1,002 mothers were Hispanic (16.0%), 1,561were Black (24.8%), and 3,720 were non-Black,non-Hispanic (59.2%). The characteristics of thesample with children are presented in Table 1.The women reported the highest grade theycompleted, with 20 years being the maximum.For income, one outlier over $350,000 wasremoved from the data set so that the valuewould not improperly skew the results. How-ever, the woman was not removed from themodels because the analyses used analyticaltools that accounted for missing values. No othervariables had outliers that were found to skewthe results.

Kinship links. The NLSY79 mother generationis genetically informative because data werecollected on all qualified individuals who re-sided in the sampled households, which includedboth full sibling, half sibling, and other types ofkinship pairs. Approximately 60% of all withinhousehold kinship links have been found to beclassifiable as twins, full siblings, half siblings,and cousins using the data available up to the1992 survey. Consistent with previous researchusing the NLSY, the following estimates of thegenetic relatedness (R coefficients) were used inthe study: R ¼ 0.125 for cousins, R ¼ 0.25 forhalf siblings, R¼ 0.375 for ambiguous siblings,R ¼ 0.50 for full siblings, and R ¼ 0.75 forsame-gender twins of unknown zygosity (Rod-gers, 1996). Most of these values are the stan-dard measures of correlation between differentlevels of kinship relatedness that can be derivedfrom the quantitative genetic model (e.g., Fal-coner, 1981). There are two exceptions. Thesame-gender twins of unknown zygosity con-tain half monozygotic (MZ) and half dizygotic

Table 1. Sample characteristics for National Longitudinal Survey of Youth parents

Maternal Variables N M SD Min. Max.

Age at first birth 4886 23.20 5.29 10.76 42.78Intellectual ability 4637 37.92% 27.21% 1% 99%Years of education 4879 12.77 2.57 0 20Income at 30 years old 3972 $32,070 $26,186 $18 $189,918Delinquency (1980) 4610 0.01 1.48 21.50 9.00

Note: Delinquency is the number of delinquent activities during the previous year regressed on mother’s age when she com-pleted the Self-Reported Delinquency Interview.


(DZ) twins in the population, so that we as-signed an R coefficient halfway between thoseof MZ twins (R ¼ 1.0) and DZ twins (R ¼0.5). For the siblings that are either full or halfsiblings (but who could not be classified bythe linking algorithm into either category), wealso assigned an R coefficient halfway betweenthat of full siblings (R ¼ 0.50) and half siblings(R¼ 0.25). Although there are more full siblingsthan half siblings in the population, past researchusing these kinship links has suggested that thereare only minor differences between using themidrange values and others that account for theunequal numbers of full and half siblings.

The kinship links in the NLSY79 samplehave been validated using several differentmechanisms. First, in the original developmentof these kinship links, a validity study was runby estimating genetic influences on gender-standardized adult physical height. Consistentwith a large literature that indicates heritabilityfor height (h2) of about .90, the h2 for adultheight using these kinship links was .88 in theNLSY79 sample; both meta-analytic and thevalidity study estimates of x 2 were quite small(Rodgers, 1996). Second, a number of pub-lished studies using these kinship links containsensitivity analyses to ascertain the sensitivityof biometrical estimates to the assumptions in-volved in using these links (e.g., the use ofR ¼ 0.375 described above), which informsthe use of these kinship links in this and othermore recent studies. Third, we have evidenceof concurrent validity by using findings basedon these kinship links and comparing them tothose from other studies. For example, compareresults on age at first intercourse from Rodgerset al. (1999), to the molecular genetic resultsfrom Miller et al. (1999); compare Rodgerset al. (2001) delinquency patterns to those inMiles and Carey (1997); or compare the multi-variate patterns across education and IQ in Neiss,Rowe, and Rodgers (2002) to those fromTambs, Sundet, Magnus, and Berg (1989).

Offspring generation sample. Biannual assess-ments of the children of women in theNLSY79 began in 1986, with an initial re-sponse rate of 95% and subsequent average re-sponse rate of 90% (Chase-Lansdale, Mott,Brooks-Gunn, & Phillips, 1991). In 1986,

95% of the biological offspring of NLSY79mothers were assessed. This response rate hasstayed high, with an average response rate of90%. The interview rate in 2000 was 77%, pri-marily because approximately 40% of the eligi-ble 15- to 19-year-olds in the African Americanand Hispanic oversample groups were not as-sessed because of budgetary limitations. Thelow response was not a major concern in thecurrent analyses because we explored external-izing problems in 4- to 10-year-old offspring.The full offspring sample was eligible to be re-assessed in 2002, however, and the overallresponse rate returned to 93%.

Three types of interviews were conductedconcerning the offspring generation. Motherswere asked to report on their children’s charac-teristics, including behavior, temperament, andhome environment. Beginning in 1988, chil-dren between the ages of 10 and 14 yearswere interviewed directly about their interac-tions with parents, responsibilities, use ofleisure time, relationships with peers, expecta-tions, and delinquent activities. Finally, begin-ning in 1994, adolescents aged 15 years andover were interviewed extensively on family in-teractions, substance abuse, delinquent activ-ities, and other issues relating to transitions toadulthood. The current analyses are based onmaternal report of offspring ages 4–10 years.

The analyses in the current paper weredrawn from 11,192 children of the mothersfrom the 2002 NLSY79. The offspring in-cluded 4,886 (43.66%) who were first borns,3,705 (33.10%) who were second borns,1,747 (15.61%) who were third borns, 637(5.69%) who were fourth borns, and 217(1.94%) who were fifth borns. Fifty-one percent(N ¼ 5,703) of the offspring were male, with49% being female (N ¼ 5,484). Of the 11,192children, 8,889 offspring were of appropriateage to have been assessed with the BehaviorProblem Index (BPI; see described later) andthus were used in the current analyses. Nineteenpercent of the offspring included in the analyseswere assessed once with the BPI, 22% were as-sessed twice, 38% were assessed three times,and 22% were assessed four times. In general,mothers of children for whom data are availableare slightly older, more educated, wealthier, andhave higher aptitude scores, as measured by the


Armed Services Vocational Aptitude Battery(see below) than mothers of children withno available data. Therefore, mother’s age atfirst birth, highest level of education, income,and intellectual abilities were controlled in theanalyses.

It should be noted that a consistent andproblematic source of selection bias is rapidlydisappearing from research using the childrenof the NLSY79 mothers. Past research hasbeen plagued by the fact that, until all child-bearing by the NLSY79 cohort is completed,the children were necessarily born to theyounger mothers. Because well over 95% ofall childbearing was completed by the 2002 sur-vey, and because we used relatively young chil-dren in our analyses, this form of selection biaswas relatively small (though certainly worthnoting). Further, the bias reported above inthe opposite direction reported above may ap-proximately compensate.

Measures

Maternal characteristics. In every wave since1983, the NLSY79 females were asked abouttheir frequency of smoking and alcohol use dur-ing 12 months prior to their most recent preg-nancy and during pregnancy. Table 2 includesthe response categories for smoking and alco-hol use during pregnancy, as well as the conver-

sions to packs/day of smoking and days/monthof drinking. Of the 8,889 offspring analyzed thecurrent study, the mothers reported their SDPfor 6,503 (73%) within approximately 1 yearof the birth. Because few children in the currentanalyses were born before 1979, mothers re-ported their SDP for 90% of their childrenwithin 4 years of each birth. Unfortunately,we have no measure of offspring’s exposureto secondary smoke. When such exposure hasbeen assessed in previous studies, however,maternal SDP almost always been found toaccount for unique variance in predicting off-spring CPs (Brook et al., 2000; Day, Richardson,Goldschmidt, & Cornelius, 2000; Griesler, Kan-del, & Davies, 1998; Weissman et al., 1999),with one exception (Maughan et al., 2001).

Because previous studies using the NLSYidentified characteristics of families related toSDP (Zimmer & Zimmer, 1998), a number ofmaternal characteristics were included in theanalyses. When they were 15–22 years old,the NLSY79 mothers (and future mothers)were asked about their engagement in 12 delin-quent behaviors during the previous year usinga version of the Self-Reported DelinquencyInterview (Elliott & Huizinga, 1983). Thismeasure, which was also administered to theoffspring generation, is reliable and valid andis the benchmark measure used in contempo-rary delinquency research (Loeber, Farrington,

Table 2. Smoking and drinking during pregnancy responses for each pregnancy

Frequency Percentage Calc. Packs/Day

Smoking responseNone 7177 64.13 0�1 pack/day 2066 18.46 0.51–2 packs/day 787 7.03 1.5�2 packs/day 85 0.76 2.5Missing 1077 9.62 —

Alcohol responseNever 6965 62.23 0,1/month 1538 13.74 0.51/month 739 6.60 13–4 days/month 434 3.88 3.51–2 days/week 344 3.07 63–4 days/week 66 0.59 14Nearly every day 19 0.17 22Every day 25 0.22 28Missing 1062 9.49 —


Stouthamer-Loeber, & Van Kammen, 1998;Moffitt, 1990; Moffitt, Caspi, Dickson, Silva,& Stanton, 1996). For more details see Rodgerset al. (2001). Symptom counts were regressedon the woman’s age at which she completedthe survey.

Mother’s age at first birth was calculated bysubtracting the mother’s date of birth from herfirst child’s date of birth. Maternal age at eachbirth was not calculated because previous re-search has noted that age at first birth, and notmaternal age specific to each birth, predictsoffspring delinquency in the NLSY (Turley,2003). Total net family income reported bymothers at age 30, which includes included in-come from all adults in the household at thattime, and highest degree attained by mothers(as of the 2002 survey) were used as measuresof families’ socioeconomic status. Total netfamily income was a summary variable calcu-lated from all income received in the household,including government support and food stamps,by the mother and her spouse. Income from co-habitating partners was not included.

During the summer and fall of 1980,NLSY79 respondents completed the ArmedServices Vocational Aptitude Battery, whichmeasured knowledge and skill in 10 areas. Acomposite score derived from select sectionsof the battery (word knowledge, paragraphcomprehension, math knowledge, and arith-metic reasoning) was used to construct an ap-proximate and unofficial Armed Forces Qualifi-cations Test score for each participant, anapproximate IQ equivalent. Raw scores werestandardized, summed, and converted to a per-centile for a measure of maternal intellectualability. Multiple imputation (Little & Rubin,1987) was used to account for the maternalcharacteristics that were missing (see below).

Offspring CP, ODP, and ADHP. At each wave,mothers rated their 4- to 10-year-old children’sbehavior problems using the BPI. The BPI wascreated by selecting items from the Child Be-havior Checklist (CBCL; Achenbach, 1978)that had the strongest correlations with CBCLfactor scores (Peterson & Zill, 1986). Mothersrated each of their children in each assessmentwave using a 3-point scale for each item: 3 ¼often true, 2 ¼ sometimes true, and 1 ¼ not

true. We used BPI items to create three a prioriscales based on DSM constructs. However, con-firmatory factor analysis of the 13 externalizingitems were conducted with MPlus (Muthen &Muthen, 1998–2005) to account for the dichot-omous variables, and the results support thethree factor solution (results available upon re-quest). The items measuring CP included thefollowing: (a) cheats or lies, (b) breaks thingson purpose or deliberately destroys his/herown or another’s things, (c) disobedient athome, (d) disobedient at school, (e) has troublegetting along with teachers, (f) does not feelsorry after misbehaving, and (g) bullies otherchildren. ODP included the following: (a)argues too much, (b) is stubborn, sullen or irri-table, and (c) has a very strong temper and losesit easily. ADHP included the following: (a) hasdifficulty concentrating, (b) impulsive or actswithout thinking, and (c) restless or overly ac-tive, cannot sit still. We obtain standardizedZ scores for CP, ODP, and ADHP within eachage and, for children assessed repeatedly, calcu-lated the average score. Therefore, each mea-sure represents the average number of problemsbetween the ages of 4 and 10 years.

The child CP items used in the CNLSY over-lap substantially with those used in previous pop-ulation-based longitudinal studies (Fergusson &Horwood, 2002; Moffitt et al., 1996). Moreover,the ADHP and ODP items are very similar tothose used in several studies of the developmentof CPs that yielded results very similar to studiesthat used full DSM measures of ADHD and ODD(Lahey, McBurnett, & Loeber, 2000). The aver-age CP for males (M ¼ 0.16, SD ¼ 1.09, N ¼4,531) was larger than for females (M ¼ 20.17,SD ¼ 0.87, N ¼ 4,358; t ¼ 15.63, p , .0001).Similar results were found for ODP (Mmales ¼

0.05, SDmales ¼ 1.02, Mfemales ¼ 20.06,SDfemales ¼ 0.97; t ¼ 5.28, p , .0001) andADHP (Mmales ¼ 0.18, SDmales ¼ 0.88, Mfemales

¼ 20.18, SDfemales ¼ 0.92; t ¼ 17.70, p ,

.0001).

Results

Maternal characteristics associated with SDP

Estimates of the relations between SDP and ma-ternal intellectual abilities, years of education,


income, delinquency, age at first birth, and meanalcohol consumption across all pregnancies arein Table 3. Correlations and unstandardized re-gression weights are presented. The regressionweights were calculated based on the data setwith no missing values (using listwise deletionfor those with missing values) and with five mul-tiply imputed data sets for the sample of womenwho had children so that the entire sample, in-cluding those with missing values, could be ana-lyzed (Little & Rubin, 1987; Schafer, 1997). Themultiply imputed data sets were based on aver-age maternal smoking and alcohol consumptionacross pregnancies, intellectual abilities, years ofeducation, income at the age of 30, delinquency,and age at first birth.

Mothers who smoked more on average dur-ing their pregnancies were more likely to havelower intellectual abilities. For every unit in-crease in packs/day, maternal intellectual abil-ities were over seven percentage points lower.Every pack/day was associated with a decreasein over 1.5 years of completed education andmore than $10,000 in yearly income at theage of 30. The average packs/day was also as-sociated with maternal delinquency (an in-crease in .70 reported delinquent activities inthe past year), a decrease of 2.32 years in ageat first birth, and a .69 increase in days of alco-hol consumption per month during pregnancy.Average SDP was also related to maternalrace, F (2, 4816) ¼ 78.40, p , .0001. The dif-ferences among the three groups were all signif-icant, using the Tukey Studentized procedure:Hispanic (M ¼ 0.10, SD ¼ 0.25, N ¼ 823);Black (M ¼ 0.18, SD ¼ 0.35, N ¼ 1,249),and non-Black, non-Hispanic (M ¼ 0.29, SD ¼

0.48, N ¼ 2,747). The findings are consistentwith previous reports using the NLSY (Zimmer& Zimmer, 1998).

Association between SDP and child behaviorproblems

Mean comparisons. For descriptive purposes,Table 4 presents the mean of CP, ODP, andADHP by number of packs smoked duringthe pregnancy and the sample of offspring.The left column shows the mean (partialedfor gender of offspring) and sample size forall of the offspring in the entire sample. The re-sults for CP illustrate a dramatic rise in behaviorproblems as the number of packs/day in-creased. The second column presents resultsfor a subset of offspring of women for whomthere was variation in SDP within their preg-nancies (e.g., the mother did not smoke duringone pregnancy and smoked during another, orshe varied the amount of cigarettes she smokedamong her pregnancies). This subset includes704 mothers and 1,752 offspring. Offspringwho were not exposed to prenatal nicotine, butwhose mother smoked during other pregnancies(M ¼ 0.12) had an increase relative to the esti-mate in the entire sample (M¼20.08). Further-more, offspring who were exposed to largeamounts of prenatal nicotine (approximately2.5 packs/day), but whose mother smoked lessduring other pregnancies (M ¼ 0.37), had fewerbehavior problems than children in the entiresample exposed to the same level of SDP (M ¼0.55). The results suggest that at least part ofthe association between SDP and offspring CPis not because of prenatal nicotine exposure.

Table 3. Relations with mean maternal smoking during pregnancy across all pregnancies

Maternal Variables r b N ba

Intellectual abilities 2.11 27.37 4581 27.07Years of education 2.27 21.60 4809 21.60Income at 30 years old 2.17 210,729 3939 210,605Delinquency (1980) .20 0.70b 4552 0.70b

Age at first birth 2.19 22.33 4816 22.32Mean alcohol consumption .15 0.67 4817 0.67

Note: The regression coefficients represent the mother’s average smoking during pregnancy across all of her preg-nancies regressed on each maternal characteristic. All parameters are significant at p , .001.aRegression estimate based on five multiply imputed data sets for the women who had children.bEstimate based on maternal income standardized with M ¼ 0 and SD ¼ 1.


The comparison between the subset ofmothers with variation in their SDP to the entiresample, however, is largely a between-familycomparison. To partially control for these be-tween-family differences, mean offspring ex-ternalizing problems are presented for threegroups of women in the subset, based on theirmean level of smoking across all of theirpregnancies: a low level (average mother SDPM ¼ 0), a medium level (M ¼ 0.5), and ahigh level (M ¼ .1.0). In the first subset(low mean mother SDP), offspring specificallyexposed 0 or 0.5 packs/day did not differ greatlyin their CP (0.08 vs. 0.11, respectively). Inthe medium mean mother SDP group, exposureto increased individual-level SDP was somewhatassociated with more CP (M0 ¼ 0.09, M0:5 ¼

0.16, M1:5 ¼ 0.28), but this trend was not statis-tically significant. Furthermore, offspring in thehigh mean mother SDP group did not exhibit

greater CP if they were exposed to more prenatalnicotine (M0 ¼ 0.43, M0:5 ¼ 0.41, M1:5 ¼ 0.37,M2:5 ¼ 0.37). The results suggest that offspringCP were strongly associated with a mother’smean SDP across all pregnancies but only mini-mally, if at all, related to individual-level expo-sure to SDP.

Similar patterns can be found for associa-tions of SDP with ODP and ADHP. In the entiresample externalizing problems increased withhigher levels of SDP; however, the relationwas attenuated in the total subset of offspringwhose mothers who had variation in SDPamong pregnancies. Furthermore, offspringODP and ADHP appeared to be significantlyrelated to the average level of mother SDP inall pregnancies; yet, within the three subgroups,offspring externalizing problems were notmore prevalent as the individual-specific mea-sure of SDP increased. We caution that the

Table 4. Mean behavior problems by smoking during pregnancy and maternal sample

Within-Mother Variationa

TotalSubset

Mean Mother

Entire Sample SDP ¼ 0 SDP ¼ 0.5 SDP � 1.0

Packs/Day M N M N M N M N M N

Conduct Problems

0 2.08 5896 .12 617 .08 248 .09 317 .43 520.5 .15 1672 .23 745 .11 113 .16 400 .41 2321.5 .40 608 .36 344 — — .28 37 .37 3072.5 .55 64 .37 46 — — — — .37 46

Oppositional Defiant Problems

0 2.08 5896 .10 617 .02 248 .11 317 .39 520.5 .16 1672 .17 745 2.02 113 .11 400 .38 2321.5 .44 608 .35 344 — — .16 37 .37 3072.5 .45 64 .24 46 — — — — .24 46

ADHD Problems

0 2.09 5896 .13 617 2.01 248 .21 317 .29 520.5 .20 1672 .20 745 .03 113 .17 400 .32 2321.5 .40 608 .40 344 — — .28 37 .41 3072.5 .31 64 .26 46 — — — — .26 46

Note: All means are presented in Z scores.aSample only includes women with multiple pregnancies during which they reported different levels of smoking duringpregnancy.


comparisons in Table 4 still include betweenfamilies effects, as the table does not specifi-cally contrast offspring within the same nuclearfamily. More advanced statistical approachesare required to compare within-mother varia-tion in SDP and behavior problems. The meananalyses, nevertheless, imply that the associa-tion between SDP and offspring externalizingproblems may largely occur only betweenmothers. Again, if SDP were to cause offspringexternalizing problems, the association wouldbe found at all levels of analysis.

Hierarchical linear models (HLMs). The off-spring of cousins/siblings/twins represents anested three-level design: the offspring level,the mother level, and the NLSY householdlevel (for more details, see D’Onofrio et al.,2005; Lynch et al., 2006; Mendle et al., 2006).Because offspring are nested under mothers,who are nested under households, observationsare not independent. The association betweenSDP and offspring behavior problems can bestudied within regression analyses using HLMsto account for the nested nature of the data, aswell as provide appropriate standard error esti-mates and significance testing (Raudenbush& Bryk, 2002).

Six HLMs were fit for each externalizingmeasure. Model 1 included SDP specific toeach child (variable name under the headingSDP 2 packs ¼ offspring), child gender, andthe interaction of the SDP and offspring genderbecause previous research has indicated the ef-fect of SDP is larger for male offspring (Wak-schlag et al., 2002). The first model comparesoffspring whose mothers smoked during theirpregnancy with unrelated (e.g., not siblings orcousins) offspring whose mothers did notsmoke. Because the measures of externalizingwere standardized, the coefficients representthe effect size (Cohen d ) associated with eachincrease in packs smoked/day. Each HLM in-cludes three variance parameters, which repre-sent the variance in externalizing problems at-tributable to the three levels in the analysis.Model 2, represents the standard approach tocontrol for differences between mothers whodiffer in their SDP by including measuredcovariates to help statistically account for con-founds. The model included maternal intellec-

tual ability, years of education, income, delin-quency, and age at first birth.

Model 3 calculated the average number ofpacks/day a mother smoked during all of herpregnancies (SDP – packs ¼ mother) and in-cluded it in the model as a second-level vari-able. The parameter associated with the sec-ond-level variable estimated whether motherswho smoke more on average have childrenwho have more externalizing problems on aver-age. The offspring-level SDP variable (SDP –packs¼ offspring[C]) was calculated as the dif-ference between the SDP during the specificpregnancy and the average maternal SDP. Ifthere was no variation in SDP within a mother,the value was zero. Therefore, the offspring-level SDP variable in Model 3 compared sib-lings in the same family in which the mothersmoked more during one pregnancy than theothers, while holding constant the averageSDP for each mother. The parameter associatedwith the offspring-level SDP variable in Model3, the within-mother effect, provided a strongertest of the causal connection between SDP andoffspring externalizing, because it was not con-founded by factors that vary between mothers.Model 4 included the measured covariates tocombine the statistical and methodologicalapproaches.

Model 5 included the third level of the datato calculate the average number of packs/dayall women from a household smoked duringall of their pregnancies (SDP – packs ¼ house-hold). The parameter associated with this third-level variable measures whether mothers andaunts, women from the original NLSY house-holds who smoked more on average have chil-dren who had more externalizing problems onaverage. This variable compared childrenwhose mother and aunts smoked more duringtheir pregnancies with unrelated children whosemother and aunts smoked less. In Model 5, thematernal level SDP (SDP – packs¼mother[C])was the deviation between the third level aver-age household SDP and average maternalSDP. The parameter associated with this vari-able compared cousins who differed in their ex-posure to prenatal nicotine, holding constant theaverage SDP of women in the NLSY house-hold. The parameter is the within-adult siblingeffect and is free of confounds that vary


between unrelated adult siblings. The off-spring-level SDP variable in Model 5 comparedsiblings who differed in their prenatal nicotineexposure, while holding constant the averagematernal and NLSY average household SDPconstant. Finally, Model 6 added the statisticalcovariates to the methodological controls fromModel 5. Algebraic representations of the modelscan be found elsewhere (D’Onofrio et al., 2005).

All analyses were conducted on five multi-ply imputed data sets to analyze all availabledata and avoid bias introduced by individualsor families with missing data. The imputeddata were based on the maternal characteristicsincluded as covariates and offspring-specificmeasures of maternal SDP and alcohol con-sumption. Any lack of precision because ofmissing values is represented by larger standarderrors around the parameters. Unstandardized

regression parameters were utilized because ofthe difficulty comparing standardized coeffi-cients when exploring causal processes (Kim &Ferree, 1981; Kim & Mueller, 1976). A compar-ison of the between and within parameters canonly be made with unstandardized parameters.

The parameters from the HLMs for CP arepresented in Table 5. The results of Model 1suggest that SDP (offspring) was associatedwith offspring CP, with the relation being largerfor male (b ¼ .29) offspring than female off-spring (b¼ .29 2 .11¼ .18). Because the inter-action between offspring gender and SDP wassignificant in the first model, the interactionof offspring gender and SDP at each level wasincluded in the models. Model 2 illustrateshow statistically controlling for measured co-variates slightly reduced the effect of SDP(bmales ¼ .21 and bfemales ¼ .21 2 .12 ¼ .09).

Table 5. Parameter estimates from hierarchical linear models for conduct problems

Model

1 2 3 4 5 6

Variables b SE b SE b SE b SE b SE b SE

SDP packsOffspring .29 .03 .21 .03Offspring�Gender 2.11 .03 2.12 .04Offspring(C) .06 .05 .06 .05 .06 .05 .06 .05Offspring�Gender(C) 2.07 .05 2.07 .05 2.07 .05 2.07 .05Mother .49 .04 .35 .04Mother�Gender 2.24 .07 2.25 .07Mother(C) .48 .14 .41 .13Mother�Gender(C) 2.43 .18 2.45 .18Household .49 .05 .33 .05Household�Gender 2.21 .08 2.22 .08

Child gender 2.29 .02 2.29 .02 2.29 .02 2.29 .02 2.29 .02 2.30 .02Maternal

Intellectual abilities .00 .00 .00 .00 .00 .00Education (years) 2.02 .00 2.02 .00 2.02 .00Incomea 2.11 .01 2.11 .01 2.11 .01Delinquency .06 .01 .06 .01 .06 .01Age at first birth 2.02 .00 2.02 .00 2.02 .00

Intercept .07 .02 .69 .13 .04 .02 .64 .13 .03 .03 .64 .13Covariances

Household .10 .03 .06 .03 .10 .03 .06 .03 .10 .03 .06 .03Mother .26 .03 .27 .03 .27 .03 .27 .03 .26 .03 .27 .03Offspring .58 .01 .57 .01 .57 .01 .57 .01 .57 .01 .57 .01

Note: SDP, smoking during pregnancy. All parameters are unstandardized. Parameters in bold are significant at p , .05.Child gender is coded male ¼ 0 and female ¼ 1. Offspring(C) represents the within-mother effect of SDP. Mother(C)represents the within-adult sibling effect.aIncome was converted to a Z score so that the parameter could be accurately estimated in standard units.


It is difficult to interpret the coefficients associ-ated with the maternal covariates because thecoefficients were the result of a simultaneousregression analysis. In Model 3, the mean mater-nal SDP across all pregnancies (mother) was as-sociated with offspring CP (bmales ¼ .49 andbfemales ¼ .49 2 .24¼ .25). However, when off-spring were compared to their siblings whodiffered in prenatal nicotine exposure (off-spring[C]), there was no association (bmales ¼

.06 and bfemales ¼ .06 2 .07 ¼ 2.01). Thesame pattern of results occurred in Model 4when the measured covariates were included.The mean maternal SDP was associated withoffspring CP (bmales ¼ .35 and bfemales ¼ .35 2 .25¼ .10), but there was no effect within mothers(bmales ¼ .06 and bfemales ¼ .06 2 .07 ¼ 2.01).In Model 5, the average NLSY Household SDP(household) was associated with offspring CP(bmales ¼ .49 and bfemales ¼ .49 2 .21¼ .28). Fur-thermore, there was an association with mater-nal SDP corrected for average household SDP(mother[C]), (bmales ¼ .48 and bfemales ¼

.48 2 .43 ¼ .05), but there was no associationwhen siblings were compared (bmales ¼ .06 andbfemales ¼ .06 2 .07 ¼ 2.01). In Model 6, theaverage SDP at the household level (bmales ¼

.33 and bfemales ¼ .33 2 .22 ¼ .11) was signifi-cantly associated with offspring CP when statis-tical covariates were included in the analyses.Average maternal SDP, holding constant averagehousehold SDP (bmales ¼ .41 and bfemales ¼

.41 2 .45 ¼ 2.04.) was significantly associatedfor males. In contrast, offspring SDP, holding con-stant the average household and maternal SDP(bmales ¼ .06 and bfemales ¼ .06 2 .07 ¼ 2.01),was not significantly associated for either gender.

Figure 1 presents unstandardized regressionparameters for Level 1 SDP from Models 1–4,for each measure of offspring externalizing.Figure 1 illustrates how the effect sizes associ-ated with SDP drop for both males and femalesin each successive model. The association be-tween SDP and offspring CP was evidentwhen comparing unrelated children (Model 1),was slightly reduced when including statisticalcontrols (Model 2), and was further attenuatedfor boys and nonexistent for girls when com-paring siblings who differ in their exposure toprenatal nicotine, with or without statisticalcontrols (Models 3 and 4). In other words, the

average amount of SDP in adult siblings was re-lated to the average CP in all their children (cou-sins), and the average amount of SDP for amother across all of her pregnancies was associ-ated with the average CP in all her children.However, siblings who were exposed to moreprenatal nicotine did not have higher CP thantheir brothers and sisters exposed to less prena-tal nicotine. This pattern of results is not consis-tent with the hypothesis that SDP causes CP. IfSDP truly caused CP, the association would befound at all three levels of the analysis, particu-larly when comparing siblings (e.g., Rodgerset al., 2000).

Table 6 presents results for ODP. In Model1, SDP (offspring) was associated with ODP(b ¼ .29), but there is no interaction with off-spring gender. Therefore, subsequent modelsdid not include the interaction between SDPand offspring gender. In Model 2, controllingfor measured characteristics of the mothersslightly reduced the association (b ¼ .20). InModel 3, however, only the mean maternalSDP across all of her pregnancies (mother)was associated with offspring ODP (b ¼ .41).There was no relation when siblings were com-pared (offspring[C]; b¼2.02). In Model 4, themeasured covariates reduced the associationwith average maternal SDP. In Model 5, off-spring ODP was associated with average house-hold (household; b ¼ .43) and maternal SDP(mother[C]; (b¼ .21) but not with the offspringlevel SDP (b ¼ 2.02). Finally, Model 6 indi-cated that when covariates are added to themodel, offspring ODP was only associatedwith household-level SDP (b ¼ .33), which isa comparison of unrelated offspring. Again,Figure 1 shows how the use of different com-parison groups drastically influences the effectsizes associated with SDP.

The HLM results for ADHP are presented inTable 7. The pattern is similar to that of ODP.SDP was associated with offspring ADHP inModel 1 (offspring; b ¼ .27), and offspringgender did not moderate the association. Themagnitude of the relation was slightly reducedin Model 2 (b ¼ .17). In Models 3 and 4, aver-age maternal SDP (mother) was associated withADHP, but the sibling comparison (off-spring[C]) suggested a small association thatwas not statistically significant (bs ¼ .07). In


Figure 1. Associations between smoking during pregnancy (SDP) and offspring externalizing problemsusing different methodological and statistical controls. Note there was no interaction between SDP and off-spring gender for oppositional defiant problems and attention-deficit/hyperactivity problems.

151

Table 6. Parameter estimates from hierarchical linear models for opposition defiant problems

Model

1 2 3 4 5 6


SDP packsOffspring .29 .03 .20 .02Offspring�Gender 2.03 .04Offspring(C) 2.02 .04 2.01 .04 2.02 .04 2.02 2.04Mother .41 .03 .32 .03Mother(C) .21 .10 .16 .10Household .43 .03 .33 .03


Intellectual abilities 2.002 .001 2.002 .00 2.001 .000Education (years) 2.02 .01 2.02 .01 2.02 .01Incomea 2.09 .02 2.08 .02 2.08 .02Delinquency .05 .01 .04 .01 .04 .01Age at first birth 2.01 .00 2.01 .00 2.01 .00

Intercept 2.02 .02 .49 .13 2.05 .02 .39 .13 2.05 .01 .38 .13Covariances


Note: SDP, smoking duirng pregnancy. All parameters are unstandardized. Parameters in bold are significant at p , .05. Child gender is coded male¼ 0 and female¼ 1. Offspring(C) represents thewithin-mother effect of SDP. Mother(C) represents the within-adult sibling effect.aIncome was converted to a Z score so that the parameter could be accurately estimated in standard units.

152

Table 7. Parameter estimates from hierarchical linear models for attention-deficit/hyperactivity disorder problems

Model

1 2 3 4 5 6


SDP packsOffspring .27 .03 .17 .03Offspring�Gender .02 .04Offspring(C) .07 .05 .07 .05 .07 .05 .07 .05Mother .37 .03 .22 .03Mother(C) .05 .10 2.04 .10Household .41 .03 .24 .03


Intellectual abilities 2.001 .001 2.001 .001 2.001 .001Education (years) 2.02 .01 2.02 .01 2.02 .01Incomea 2.10 .01 2.10 .01 2.10 .01Delinquency .04 .01 .04 .01 .04 .01Age at first birth 2.03 .00 2.03 .00 2.03 .00

Intercept .13 .02 1.05 .11 .11 .02 1.01 .11 .10 .02 .99 .11Covariances


Note: SDP, smoking during pregnancy. All parameters are unstandardized. Parameters in bold are significant at p , .05. Child gender is coded male¼ 0 and female¼ 1. Offspring(C) represents the within-mother effect of SDP. Mother(C) represents the within-adult sibling effect.aIncome was converted to a Z score so that the parameter could be accurately estimated in standard units.

153

Models 5 and 6, the association between SDPand offspring ADHP was primarily found atthe household level (household; b ¼ .41 and.24). The comparison of cousins (mother[C])and siblings (offspring[C]) who differed inexposure to prenatal nicotine found small asso-ciations between ADHP and SDP.1 Figure 1notes the small association between SDP andoffspring ADHP when siblings who differedin their exposure to prenatal nicotine werecompared.

Structural equation modeling. The HLM re-sults suggested that the association betweenSDP and each measure of externalizing is be-cause of unmeasured confounds that are sharedby family members. Consequently, we usedmultilevel structural equation models (SEMs)to examine whether the relevant confoundswere genetic or environmental in origin (Hardenet al., 2007). The two-level SEM is illustrated inFigure 2. The first member of an adult sibshipand her respective children are represented onthe left side of the figure; the second memberand her children are represented on the rightside of the figure. The figure is divided intotwo portions, representing relations betweenSDP and externalizing problems within chil-dren with the same mother and between unre-lated children (described in more detail below).This graphical convenience is not meant to im-ply that these portions of the model are esti-mated separately; analogous to the estimation oflevel-specific regressions in HLM, both por-tions are estimated simultaneously. The readeris referred to Mehta and Neale (2005) for amore complete didactic on multilevel SEMand its equivalence to HLMs.

The bottom portion of the model (in brokenlines) reflects relationships within children withthe same mother. The externalizing problems ofthe ith child in the jth nuclear family of the kthtwin family (EXTijk) is predicted by his or herown exposure to prenatal nicotine (SDPijk), asrepresented by the path labeled w. To the extentthe child exposed to more SDP than his or her

siblings also reports more externalizing prob-lems than his or her siblings, this is reflectedin the w path.

The top portion of the model (in unbrokenlines) reflects relationships between relatedmothers and their children. Average maternalSDP across all pregnancies are representedhere by the latent variables labeled SDP0jk. Inaddition, similar to the standard twin designpath model, the variance in average SDP is de-composed into three components: variance be-cause of additive genetic influences (Va), var-iance because of other environmental influencesthat make the adult siblings more similar (Vc),and variance because of environmental influ-ences that make the adult siblings in the samehousehold different (Ve). The covariance pa-rameters between the latent genetic parameters(Acov) were constrained so that the genetic cor-relations were appropriate for each group. Thecovariance of the shared environmental factors(Ccov) was constrained such that the correlationequaled 1.0, the standard correlation for theshared environmental latent factors (Neale &Cardon, 1992). Readers may be more familiarwith a twin model parameterization in whichthe A, C, and E components are standardizedto a variance of 1.0 and the paths from themare freely estimated. The current model is sim-ply a reparamaterization with the paths fixedto one and the variances freely estimated. Be-cause the paths are fixed to 1, the scale ofeach component is defined by the maternalSDP variable, and the total variance in maternalSDP is the sum of the estimated variances of theA, C, and E components. Dividing the esti-mated variance of each component by the totalvariance of average maternal SDP yields thefamiliar proportions of the heritability (h2),shared environment (c2), and nonshared envi-ronment (e2). For more detail concerning varia-tions on twin and family models, see Neale andCardon (1992).

The average number of externalizing prob-lems in children with the same mother(EXT0jk) is then regressed on the variance com-ponents of SDP: A, C, and E. The model is sim-ilar to the standard bivariate genetic analysis(Neale & Cardon, 1992), except that the modelestimated the unstandardized variance compo-nents in order to compare the intergenerational

1. Fixed effect models, a common econometric approach toclustered data (Greene, 2003), were also fit to the data atthe nuclear family level. The parameters associated withSDP were comparable to the HLM results.


Figure 2. The structural equation model for the offspring of siblings and twins; SDP, smoking EXT, offspring externalizing. (—)The maternal and National Longitudinal Survey of Youth household level of the analysis and (- - -) the offspring level (within maternal).See Harden et al. (2007) for a comparison to the specific children of twins model.

155

paths (ba, bc, and be) on the same scale. The ba

parameter estimates the effect of genetic factorscommon to both maternal SDP and child exter-nalizing (passive rGE). The bc parameter esti-mates the effect of environmental factors thatboth make adult siblings similar for SDP andinfluence offspring externalizing. The model,therefore, controls for environmental factorsthat make adult siblings similar and influenceoffspring externalizing. Finally, the be pa-rameter estimates whether mothers who differin their levels of SDP for nongenetic reasons(as reflected in variance component Ve) havechildren who differ in their average number ofexternalizing problems.

If SDP caused offspring externalizing, thewithin-mother (w) and nonshared environ-mental intergenerational path (be) would belarge. These parameters estimate how differ-ences between child siblings in prenatal nico-tine exposure are associated with externalizingat the offspring level (i.e., if a child sibling ex-posed to more prenatal nicotine had more exter-nalizing problems than his/her sibling exposedto less), as well as how differences betweenadult siblings in SDP are associated with exter-nalizing at the maternal level (i.e., if one mothersmoked more than her adult sibling and hadchildren who had more externalizing problemson average). If the relation is not causal, a com-parison of the genetic (ba) and common envi-ronmental (bc) parameters would elucidate thesource of the confounds. A residual correlationbetween the cousins in the offspring generation(rco) was also estimated to account for the co-variance among cousins not accounted for bythe SDP status of their mothers. The correlationwas estimated separately in the different groupsto account for differing levels of genetic andenvironmental relatedness.

The SEM were initially fit with data from thecousin, half sibling, ambiguous sibling, and fullsibling families.2 Thirty-seven sibling pairs and

their offspring were dropped from the analysesbecause the NLSY household included multi-ple pairs of varying genetic relatedness (e.g.,some participants were related as full siblingsbut others were related as cousins, even thoughthey lived in the same household). The analysisof clustered data in which participants within asingle cluster (NLSY household) belong tomultiple groups (e.g., full siblings and cousins)was not permitted in Mplus (Muthen & Mu-then, 1998–2005). Pairs of individuals inhouseholds of unknown genetic relatednesswere also not included in the first model be-cause these pairs could not help separate the es-timates of genetic and environmental influ-ences. The decomposition of the variance inSDP is based on the correlations among the dif-ferent adult sibling pairs in the NLSY, found inTable 8. The parent-level correlations were rela-tively stable as the genetic relatedness of thesiblings increase, suggesting that the sharedenvironmental factors account for most of thecovariance between siblings. Similarly, SEMsestimated the variance attributable to additive ge-netic factors (Va) to be zero. Moreover, the inter-generational parameter associated with geneticvariance (ba) was unstable (i.e., ba was largebut had very large standard errors), a result thatfrequently occurs when one latent variance com-ponent is negligible (D’Onofrio et al., 2003).

Therefore, the parameters associated withgenetic factors (Va and ba) were constrained tobe zero in the subsequent models, making theSEMs a between and within level of SDP at

Table 8. Correlations between adultsiblings in average smoking duringpregnancy across all pregnancies

Relatedness r N

Unknown .41 297Cousins .39 31Half-sibs .27 16Ambiguous sibs .54 85Full sibs .34 472Twins — 0

Note: Correlations in bold are significant at p ,

.05. The correlations were not corrected for multi-ple sibling pairs per family. There were no twinpairs where both female twins had reports ofsmoking during pregnancy.

2. The SEMs were fit using all of the available sibling rela-tionships, including multiple relationships from thesame NLSY household. Because the SEMs assumedthat each adult sibling pair was independent, all of themodels were reanalyzed only using the first siblingpair from each family. The results were consistent withthe models based on the complete sample.


the nuclear family and NLSY household levels.The subsequent SEMs included all adult siblingpairs, including those in the NLSY where thegenetic relatedness of the siblings was un-known. All sibling pairs were included in thismodel because they were informative aboutenvironmental factors that make individualsgrowing up in the same household similar.The models were estimated using MLR, a max-imum likelihood estimator with standard errorsthat are robust to nonnormality (Muthen &Muthen, 1998–2005). The SEM indicated thatshared environmental factors accounted for61% of the variance (Vc ¼ .083, SE ¼ .01) ofmean SDP at the maternal level, and nonsharedenvironmental influences accounted for 39%(Ve ¼ .054, SE ¼ .01).

The SEM for CP fit the data well (x 2 ¼

40.72, df ¼ 13, root mean square error of ap-proximation [RMSEA] ¼ .015). The within-mother parameter was minimal (w ¼ .07),consistent with HLM results. Environmentalfactors that vary between families (bc ¼ .48)and within-adult-sibling families (be ¼ .27)accounted for the association between SDPand CP. Overall, SDP accounted for 0.1% ofthe variance in offspring CP within mothersand 8% of the variance between mothers. Theintergenerational parameters from the SEMs,including unstandardized estimates, standarderrors, and standardized estimates, are foundin Table 9. The SEM for ODP also fit the datawell (x 2 ¼ 42.93, df ¼ 13, RMSEA ¼ .015).For ODP, the results indicated a negligible as-sociation with SDP within mothers (w ¼ .03).Environmental factors that make adult siblings

similar (bc ¼ .48) and unique (be ¼ .46) bothcontributed to the association between SDPand ODP. SDP accounted for no variance inODP within mother but accounted for 11% ofthe variance in ODP between mothers. Finally,the SEM for ADHP (x 2 ¼ 37.08, df ¼ 13,RMSEA ¼ .014) indicated a small associationwithin mothers (w ¼ .09). The parameter asso-ciated with nonshared environmental influ-ences was comparable to the within-mother ef-fect but was not statistically significant (be ¼

.10). Environmental factors that very betweenfamilies accounted for most of the intergenera-tional associations (bc ¼ .61). SDP accountedfor only 0.1% of the variance in ADHP withinmothers but 14.4% of the variance betweenmothers.

Discussion

The current article used the clustered design inthe NLSY sample to explore the association be-tween SDP and three measures of offspring ex-ternalizing problems (CP, ODP, and ADHP).The present comparisons of unrelated childrenwere consistent with the results of previousstudies (Wakschlag et al., 2002) in several re-spects: (a) CP, ODP, and ADHP problemswere significantly associated with SDP; (b)each association followed a dose–response rela-tionship; (c) the number of CP demonstrated bychildren exposed to SDP was higher for malechildren; and (d) each association remained sig-nificant after statistically controlling for associ-ated maternal characteristics. In addition to theuse of statistical covariates used in previous

Table 9. Structural equation modeling results for smoking during pregnancy offspringexternalizing

CP ODP ADHD

Intergenerational Parameters b SE b b SE b b SE b

Within mothers .07 .04 .03 .03 .04 .01 .09 .04 .04Genetic — — — — — — — — —Common environment .48 .14 .26 .48 .12 .26 .61 .13 .38Nonshared environment .27 .19 .12 .46 .19 .20 .10 .16 .05

Note: Because the initial structural equation modeling indicated that no variance in smoking during pregnancy was due toadditive genetic factors, the path from the genetic latent factor was dropped from the structural equation modeling. Parame-ters in bold are significant at p , .05.


studies, the current analyses utilized the multi-ple levels of the NLSY to account for unmea-sured confounds. If SDP caused higher conductand oppositional problems, the relation wouldhave been evident both when comparing related(e.g., within mothers) and unrelated children(e.g., Rodgers et al., 2000). However, whenchildren were compared with their siblingswho differed in their exposure to SDP, the off-spring did not differ with respect to conduct andoppositional defiant problems. SDP accountedfor no more than 0.1% of the within-nuclearfamily variance in offspring CPs. In contrast,SDP accounted for 8–14% of the between-nuclear family variance in offspring CPs. Theseresults suggest that previous studies found a re-lationship between SDP and offspring CPs notbecause SDP causes increased risk for conductor oppositional problems, but because environ-mental influences that vary between familiesconfound associations between SDP and off-spring externalizing. This finding is generally

consistent with another CoT study of SDP andADHD (Knopik et al., 2006), as well as withstudies that have included more precise mea-surement of adult characteristics that may con-found the relation, such as maternal and pater-nal antisocial characteristics (Maughan et al.,2004) and maternal delinquency during adoles-cence (Silberg et al., 2003).

The present findings indicate that environ-mental influences that are risk factors for bothmaternal SDP and offspring externalizing un-derlie the previously observed intergenerationalassociations between SDP and offspring CPs.This has important implications for futureresearch on the causes of CPs. The present find-ings imply that the salient aspects of the famil-ial or social environment that are robustly asso-ciated with both SDP and offspring CPs have notbeen the focus of the research on the etiology ofexternalizing problems. Because SDP is signifi-cantly related to offspring CPs after controllingfor essentially all of the parent and family

Table 10. Comparison of smoking during pregnancy parameter results from Model 3using different samples

Entire Samplea

Only FamiliesWith .1Childb

Only .1 Childand Within-Mother SDPVariationc

Parameter b SE b SE b SE

Conduct Problems

Offspring(C) .06 .05 .07 .05 .06 .06Offspring�Gender(C) 2.07 .05 2.07 .05 2.06 .06Mother .49 .04 .50 .05 .42 .08Mother�Gender 2.24 .07 2.27 .08 2.20 .11


Offspring(C) 2.02 .04 2.02 .05 2.02 .05Mother .41 .03 .41 .03 .35 .05

ADHD Problems

Offspring(C) .07 .05 .07 .04 .07 .05Mother .37 .03 .38 .03 .35 .05

Note: SDP, smoking during pregnancy; ADHD, attention-deficit/hyperactivity disorder. See Tables 5–7 formore details.aParameters are based on the complete analysis presented in the paper.bThe sample only included offspring from families with more than one child.cThe sample only included offspring from families with multiple children and where there was variation withinmothers in SDP.


characteristics that have been examined as riskfactors for offspring CPs (Wakschlag & Hans,2002), the present findings suggest that aspectsof the family or social environment that are unre-lated to maternal antisocial behavior, maternalage, and other well-studied risk factors are re-sponsible for the relations to SDP. Hopefully,the present findings will lead to an expandedsearch for these environmental risk factors. Theimportant clues offered by these findings isthat they are correlated with SDP, but notcorrelated with the risk factors that have beencontrolled in this and previous studies of SDP.

The present analyses, however, suggest aminimal role of SDP with offspring ADHPproblems, but the magnitude of the associationwas greatly reduced compared to previous ob-servations. A recent review has documented asmall association between SDP and ADHP aftercontrolling for various confounds (Linnet et al.,2003), and the small association, consistentwith a direct causal effect of SDP on offspringADHP, may be related to the influence of SDPon birth weight (Knopik et al., 2006).

The current article used two analytical ap-proaches to explore the associations betweenSDP and offspring externalizing because eachapproach has its own advantages and limita-tions. HLMs were used because they enabledthe inclusion of every family and offspring.Furthermore, interactions between variables(e.g., SDP and offspring gender) and the effectsof measured family characteristics can be moreeasily and directly included. Although theHLMs indicate that the relation between SDPand offspring externalizing is not causal, it isdifficult to identify whether the origin of rele-vant confounds are genetic or environmentalwithin the HLM approach. Interpreting the be-tween and within-family parameter estimatesfrom the multiple family groups in the NLSY(cousins, half-siblings, ambiguous siblings,full siblings, and twins) becomes quite compu-tationally burdensome (results not shown). Incontrast, the SEM approach provides a morestraightforward approach to identifying theintergenerational effects of genetic, commonenvironmental, and nonshared environmental

Table 11. Comparison of offspring externalizing by pack and detailed family structure

Sample

One Child/FamilyMultiple Children and NoWithin-Mother Variation

Multiple Children andWithin-Mother Variation

SDP Packs/Day M SD N M SD N M SD N

Conduct Problems

0 20.17 0.88 486 20.09 0.92 4793 0.12 1.03 6170.5 0.00 0.89 178 0.10 1.01 749 0.23 1.05 7451.5 0.58 1.38 42 0.42 1.19 222 0.36 1.12 3442.5 0.63 1.26 7 1.22 1.42 11 0.37 1.16 46


0 20.14 0.98 486 20.10 0.96 4793 0.10 1.00 6170.5 0.12 1.06 178 0.15 0.96 749 0.17 1.00 7451.5 0.67 1.29 42 0.54 1.12 222 0.35 1.08 3442.5 0.58 1.26 7 1.25 1.23 11 0.24 1.12 46

ADHD Problems

0 0.02 0.97 486 20.13 0.94 4793 0.13 0.98 6170.5 0.34 0.98 178 0.17 0.95 749 0.20 0.97 7451.5 0.47 1.12 42 0.39 1.14 222 0.40 1.07 3442.5 0.40 0.26 7 0.50 1.26 11 0.26 1.01 46

Note: SDP, smoking during pregnancy; ADHD, attention-deficit/hyperactivity disorder.


influences. The SEM approach, however, can-not include adult sibships that differed in ge-netic relatedness from other pairs within theirhousehold. Furthermore, the sibling correla-tions and SEM suggest no influence of geneticfactors on SDP, in contrast to other reports withlarger samples (D’Onofrio et al., 2003; Knopiket al., 2005). Accordingly, we conclude that theresults suggest that the confounds are environ-mental rather than genetic in origin. Analyseswith larger samples will be required to morespecifically elucidate the magnitude of geneticand common environmental processes respon-sible for the intergenerational association. Re-gardless of the nature of the confounds, theanalyses question whether associations betweenSDP and offspring externalizing in young chil-dren are causal.

There are a number of additional limitationsof the current study. First, all of the analyses arebased on self-report measures completed by themother. The retrospective report of SDP mayhave resulted in underreporting or greater mea-

surement error (Wakschlag et al., 2002), butthere is a high correlation between self-reportedsmoking status and serum cotinine measures(McDonald, Perkins, & Walker, 2005). The re-liability of retrospective reports is also similarto the recall of other substance use (Petitti,Friedman, & Kahn, 1981). Furthermore,mothers who report SDP may be more likelyto report higher externalizing in their children.The magnitude of intergenerational associa-tions in the current paper, however, are consis-tent with studies that used more sophisticatedmeasurement strategies (Wakschlag et al.,2002). Second, because of the complexity ofthe analyses we did not analyze the longitudinalresponses of the offspring across the age range(4–10 years old); rather, we relied on the aver-age across the years. Future analyses that ex-plore offspring characteristics associated withSDP would benefit from using analytical strate-gies, such as growth curve models, that can ac-count for individual differences in initial leveland change across time (e.g. Willett, Singer,

Table 12. Comparison of offspring externalizing in samples with no variation

Sample

One Child/Family2 Children and No

Within-Mother Variation3þ Children and No

Within-Mother Variation

SDP Packs/Day M SD N M SD N M SD N

Conduct Problems

0 20.17 0.88 486 20.17 0.88 2001 20.04 0.95 27920.5 0.00 0.89 178 0.09 0.95 357 0.10 1.07 3921.5 0.58 1.38 42 0.15 1.14 83 0.58 1.19 1392.5 0.63 1.26 7 0.45 1.71 5 1.86 0.77 6


0 20.14 0.98 486 20.09 0.95 2001 20.11 0.96 27920.5 0.12 1.06 178 0.27 0.94 357 0.05 0.97 3921.5 0.67 1.29 42 0.44 1.12 83 0.60 1.12 1392.5 0.58 1.26 7 0.78 1.45 5 1.64 0.99 6

ADHD Problems

0 0.02 0.97 486 20.12 1.14 2001 20.14 0.94 27920.5 0.34 0.98 178 0.21 0.91 357 0.14 0.98 3921.5 0.47 1.12 42 0.20 1.13 83 0.51 1.13 1392.5 0.40 0.26 7 20.20 0.97 5 1.09 1.22 6

Note: SDP, smoking during pregnancy; ADHD, attention-deficit/hyperactivity disorder.


& Martin, 1998). Third, our findings are limitedto measures of externalizing in young childrenand may not reflect the underlying mechanismsresponsible for the association between SDPand adult criminal activity (Brennan et al.,1999; Rasanen et al., 1999). Fourth, no fatherinformation was included in the analyses. Lim-ited information on the fathers is available inthe NLSY, and the association between SDPand offspring externalizing could be becauseof higher rates of paternal delinquency(Maughan et al., 2004). Methodological workon CoT design has highlighted its strengthwith characteristics of individuals (such asSDP), but inasmuch as SDP is influenced bythe women’s spouses, the interpretation of theparameters may be somewhat restricted (Eaves,Silberg, & Maes, 2005). Fifth, we only exploredwhether offspring gender moderated the influ-ence of SDP, but there may be vulnerability fac-tors (both environmental and genetic in origin)that make individuals more susceptible to theinfluence of SDP.

Sixth and finally, as is the case with all com-parisons using different levels of analyses, thedesign rests on a number of assumptions aboutwomen who have multiple offspring and whovary in the levels of SDP among the pregnan-cies. The design assumes that women whohave only one child are similar to women withmultiple offspring, because the within-motherparameter is based solely on women with multi-ple children. Furthermore, the analyses assumethat women who have multiple pregnancies andvariation in SDP are similar to women whohave multiple offspring but smoked at thesame level in each pregnancy, including theirability to accurately report SDP. Again, thewithin-mother parameter can only be estimatedbased on women with variation in SDP amongher pregnancies.3 Ultimately, the limitations of

the current article underscore the necessity ofreplicating the analyses in other samples andexploring the consequences of SDP in animalstudies, which have the advantage of randomassignment.

The study highlights the importance of usingresearch designs that can pull apart differentrisk mechanisms and account for nonmeasuredconfounds (Rutter et al., 2001). Furthermore,the results underline the limitations of solely re-lying on measured covariates to account forconfounds. Even though the initial HLMs thatcompared unrelated children included a mea-sure of maternal delinquency to control forthe intergenerational transmission of externaliz-ing problems, there was still an association be-tween SDP and offspring externalizing. If weonly relied on the measured covariates wewould have drawn the incorrect conclusion.Methodologically, these results support the im-portance of using design innovations as well asstatistical controls to study risk mechanismsrather than just identifying risk factors (Moffitt,2005; Rutter et al., 2001).

Similar to other genetically informed studiesof familial characteristics, such as parental di-vorce (D’Onofrio et al., 2006), the underlyingprocesses associated with a risk factor dependon the specific outcome measure being studied.

3. Given the seriousness of these assumptions, we con-ducted a series of additional analyses. First, all of theHLMs were conducted controlling for birth order, num-ber of offspring per mother, and the interaction betweenbirth order and number of offspring. The results of theHLMs did not change appreciably. We also ran the thirdHLM, which included within- and between-mother ef-fects of SDP, using two restricted subsets of the data:(a) a sample that only included families with multiplechildren and (b) a sample that only included offspring

of mothers who varied in their SDP among their multiplepregnancies. The comparison of the between-mothersSDP parameters in the two restricted subsets to the esti-mated between-mothers parameters in the entire sampleprovides a test of the assumptions underlying the currentanalyses (the within-mother parameter is based only onthe most restricted subset). The results are presented inTable 10. The between-mothers SDP parameters foreach measure of externalizing were only slightly re-duced, if at all, when the analyses were conducted onthe two restricted subsets. These results are consistentwith our assumptions of equality across the groups(e.g., women with variation in SDP vs. women withoutvariation in SDP). Tables 11 and 12 also present themeans for the offspring externalizing problems in morespecific subsets of the data. Finally, we explored the re-lation between SDP and offspring birth weight using thesibling comparison approach to see if our results repli-cated previous findings that controlled for genetic andenvironmental confounds (e.g., D’Onofrio et al.,2003). The results indicated that SDP within motherswas associated with decreased birth weight, consistentwith a causal connection and previous genetically in-formed research on the topic.


SDP appears to have a specific environmentalassociation with offspring birth weight, inas-much as genetic factors, common environ-mental, and measured covariates do not con-found the relation (D’Onofrio et al., 2003; seefootnote 3). However, CP and ODP in youngchildren associated with SDP are completelybecause of characteristics that lead to bothSDP and offspring externalizing behaviors,not the consequences of SDP. As a result,each risk factor and specific measure of adjust-ment needs to be studied independently withdesigns that can separate the risk mechanisms.

Finally, we must state that we were surprisedby the results of the analyses, given the existentresearch on SDP (although no previous researchhas allowed for control over unobserved hetero-geneity to the extent that ours has). We want tostress the need to replicate the current findings.Furthermore, we certainly concur with other re-searchers (e.g., Maughan et al., 2004) in stress-ing the importance of reducing SDP because ofits effects on a wide range of range of devel-

opmental characteristics, particularly neurobe-havioral functioning (Cnattingius, 2004; Hui-zink & Mulder, 2006). Animal research hasshown that prenatal nicotine exposure has neu-rotoxic effects (review in Wakschlag et al.,2002). Thus, the findings of the current analy-ses are limited to externalizing problems duringchildhood. The current article, though, empha-sizes that women who smoke during pregnancyare different than mothers who do not. Al-though the research literature has noted a num-ber of characteristics of mothers and familiesassociated with SDP, the analyses in the currentarticle suggest that unmeasured characteristicsaccounted for the associations between SDPand offspring externalizing problems. Perhapsthere are more differences than previouslyimagined between mothers who smoke duringpregnancy than those who do not. Therefore,addressing the myriad of risk factors associatedwith SDP may be a more effective approach ofminimizing externalizing problems in offspringthan specifically focusing on reducing SDP.

References

Achenbach, T. M. (1978). The Child Behavior Profile:I. Boys aged 6–11. Journal of Consulting and ClinicalPsychology, 46, 478–488.

Brennan, P. A., Grekin, E. R., & Mednick, S. A. (1999). Ma-ternal smoking during pregnancy and adult male criminaloutcomes. Archives of General Psychiatry, 56, 215–219.

Breslau, N., Kilbey, M., & Andreski, P. (1993). Vulnerabil-ity to psychopathology in nicotine-dependent smokers:An epidemiologic study of young adults. AmericanJournal of Psychiatry, 150, 941–946.

Brook, J. S., Brook, D. W., & Whiteman, M. (2000). Theinfluence of maternal smoking during pregnancy onthe toddler’s negativity. Archives of Pediatric and Ado-lescent Medicine, 154, 381–385.

Button, T. M. M., Thapar, A., & McGuffin, P. (2005).Relationship between antisocial behaviour, attention-deficit hyperactivity disorder and maternal prenatalsmoking. British Journal of Psychiatry, 187, 155–160.

Chase-Lansdale, P. L., Mott, F. L., Brooks-Gunn, J., &Phillips, D. A. (1991). Children of the National Longi-tudinal Survey of Youth: A unique research opportu-nity. Developmental Psychology, 27, 918–931.

Chatterji, P., & Markowitz, S. (2001). The impact of mater-nal alcohol and illicit drug use on children’s behaviorproblems: Evidence from the children of the NationalLongitudinal Survey of Youth. Journal of Health Eco-nomics, 20, 703–731.

Cnattingius, S. (2004). The epidemiology of smoking dur-ing pregnancy: Smoking prevalence, maternal charac-teristics, and pregnancy outcomes. Nicotine and To-bacco Research, 6, S125–S140.

D’Onofrio, B. M., Turkheimer, E., Eaves, L. J., Corey, L.A., Berg, K., Solaas, M. H., et al. (2003). The role

of the Children of Twins design in elucidating causal re-lations between parent characteristics and child out-comes. Journal of Child Psychology and Psychiatry,44, 1130–1144.

D’Onofrio, B. M., Turkheimer, E. N., Emery, R. E., Slutske,W., Heath, A., Madden, P. A. F., et al. (2005). A geneti-cally informed study of marital instability and its asso-ciation with offspring psychopathology. Journal of Ab-normal Psychology, 114, 570–586.

D’Onofrio, B. M., Turkheimer, E. N., Emery, R. E.,Slutske, W., Heath, A., Madden, P. A. F., et al.(2006). A genetically informed study of the processesunderlying the association between parental marital in-stability and offspring adjustment. Developmental Psy-chology, 42, 486–499.

Day, N. L., Richardson, G. A., Goldschmidt, L., & Corne-lius, M. D. (2000). Effects of prenatal tobacco exposureon preschoolers’ behavior. Journal of Developmentaland Behavioral Pediatrics, 21, 180–188.

Eaves, L. J., Silberg, J. L., & Maes, H. H. (2005). Revisitingthe children of twins: Can they be used to resolve theenvironmental effects of dyadic parental treatment onchild behavior? Twin Research, 8, 283–290

Elliott, D. S., & Huizinga, D. (1983). Social class and delin-quent behavior in a national youth panel. Criminology,21, 149–177.

Ernst, M. (2001). Behavioral and neural consequences of pre-ntal exposure to nicotine. Journal of the American Acad-emy of Child & Adolescent Psychiatry, 40, 630–642.

Falconer, D. S. (1981). Introduction to quantitative genet-ics. New York: Longman.

Fergusson, D. M. (1999). Prenatal smoking and antisocialbehavior. Archives of General Psychiatry, 56, 223–224.


Fergusson, D. M., & Horwood, L. J. (2002). Male and fe-male offending trajectories. Development and Psycho-pathology, 14, 159–177.

Fergusson, D. M., Woodward, L. J., & Horwood, L. J.(1998). Maternal smoking during pregnancy and psy-chiatric adjustment in late adolescence. Archives ofGeneral Psychiatry, 55, 721–727.

Geronimus, A. T., & Korenman, S. (1992). The socioeco-nomic consequences of teen childbearing reconsidered.The Quarterly Journal of Economics, 107, 1187–1214.

Gibson, C., Piquero, A., & Tibbets, S. (2000). Assessing therelationship between maternal cigarette smoking duringpregnancy and age at first police contact. Justice Quar-terly, 17, 519–542.

Gottesman, I. I., & Bertelsen, A. (1989). Confirming unex-pressed genotypes for schizophrenia. Archives of Gen-eral Psychiatry, 46, 867–872.

Greene, W. H. (2003). Econometric analysis. (5th ed.). Up-per Saddle River, NJ: Prentice–Hall.

Griesler, P., Kandel, D., & Davies, M. (1998). Maternalsmoking in pregnancy, child behavior problems, andadolescent smoking. Journal of Research on Adoles-cence, 8, 159–185.

Harden, K. P., Lynch, S. K., Turkheimer, E., Emery, R. E.,D’Onofrio, B. M., Slutske, W. S., et al.. (2007). A be-havior genetic investigation of adolescent motherhoodand offspring mental health problems. Journal of Ab-normal Psychology, 116, 667–683.

Heath, A. C., Kendler, K. S., Eaves, L. J., & Markell, D.(1985). The resolution of cultural and biological inheri-tance: Informativeness of different relationships. Behav-ior Genetics, 15, 439–465.

Huizink, A. C., & Mulder, E. J. H. (2006). Maternal smok-ing, drinking or cannabis use during pregnancy and neu-robehavioral and cognitive functioning in human off-spring. Neuroscience and Biobehavioral Reviews, 30,24–41.

Kim, J., & Ferree, G. D. (1981). Standardization in causal anal-ysis. Sociological Methods and Research, 10, 187–210.

Kim, J., & Mueller, C. W. (1976). Standardized and unstan-dardized coefficients in causal analysis: An expositorynote. Sociological Methods and Research, 4, 423–438.

Knopik, V. S., Heath, A. C., Jacob, T., Slutske, W. S.,Bucholz, K. K., Madden, P. A. F., et al. (2006). Mater-nal alcoholism and offspring ADHD: Disentanglinggenetic and environmental effects using a children-of-twins design. Psychological Medicine, 36, 1461–1471.

Knopik, V. S., Sparrow, E. P., Madden, P. A. F., Bucholz, K.K., Hudziak, J. J., Reich, W., et al. (2005). Contributionsof parental alcoholism, prenatal substance exposure, andgenetic transmission to child ADHD risk: A female twinstudy. Psychological Medicine, 35, 625–635.

Lahey, B. B., McBurnett, K., & Loeber, R. (2000). Are at-tention-deficit hyperactivity disorder and oppositionaldefiant disorder developmental precursors to conductdisorder? In A. Sameroff, M. Lewis, & S. Miller(Eds.), Handbook of developmental psychopathology(2nd ed., pp. 431–446). New York: Plenum Press.

Li, K., & Poirier, D. J. (2003). The roles of inputs andoutputs in predicting health, behaviour and test scoresin early childhood. Statistics in Medicine, 22, 3489–3514.

Linnet, K. M., Dalsgaard, S., Obel, C., Wisborg, K., Hen-riksen, T. B., Rodriguez, A., et al. (2003). Maternal life-style factors in pregnancy risk of attention deficit hyper-activity disorder and associated behaviors: review of

the current evidence. American Journal of Psychiatry,160, 1028–1040.

Little, R. J. A., & Rubin, D. B. (1987). Statistical analysiswith missing data. New York: Wiley.

Loeber, R., Farrington, D. P., Stouthamer-Loeber, M., &Van Kammen, W. (1998). Antisocial behavior and men-tal health problems: Explanatory factors in childhoodand adolescence. Mahwah, NJ: Erlbaum.

Lynch, S. K., Turkheimer, E., Emery, R. E., D’Onofrio, B.M., Mendle, J., Slutske, W., et al. (2006). A geneticallyinformed study of the association between harsh punish-ment and offspring behavioral problems. Journal ofFamily Psychology, 20, 190–198.

Matthews, T. J. (2001). Smoking during pregnancy in the1990s. National Vital Statistics Reports, 49, 1–14.

Maughan, B., Taylor, A., Caspi, A., & Moffitt, T. E. (2004).Prenatal smoking and early childhood conduct problems.Archives of General Psychiatry, 61, 836–843.

Maughan, B., Taylor, C., Taylor, A., Butler, N., & Bynner,J. (2001). Pregnancy smoking and childhood conductproblems: A causal association? Journal of Child Psy-chology and Psychiatry, 42, 1021–1028

McDonald, S. D., Perkins, S. L., & Walker, M. C. (2005).Correlation between self-reported smoking status andserum cotinie during pregnancy. Addictive Behaviors,30, 853–857.

Mehta, P. D., & Neale, M. C. (2005). People are variablestoo: Multilevel structuralequations modeling. Psycho-logical Methods, 10, 259–284.

Mendle, J., Turkheimer, E., D’Onofrio, B. M., Lynch, S.K., Emery, R. E., Slutske, W., et al. (2006). Familiystructure and age at menarche: A children of twins ap-proach. Developmental Psychology, 42, 535–542.

Mick, E., Biederman, J., Faraone, S. V., Sayer, J., & Klein-man, S. (2002). Case control study of attention deficithyperactivity disorder and maternal smoking, alcoholuse, and drug use during pregnancy. Journal of theAmerican Academy of Child & Adolescent Psychiatry,41, 378–385.

Miles, D. R., & Carey, G. (1997). Genetic and environ-mental architecture of human aggression. Journal ofPersonality and Social Psychology, 72, 207–217.

Miller, W. B., Pasta, D. J., MacMurray, J., Chiu, C., Wu, H.,& Comings, D. E. (1999). Dopamine receptors are asso-ciated with age at first sexual intercourse. Journal ofBiosocial Science, 31, 43–54.

Moffitt, T. E. (1990). Juvenile delinquency and attentiondeficit disorder: Boys’ developmental trajectoriesfrom age 3 to 15. Child Development, 61, 893–910.

Moffitt, T. E. (2005). The new look of behavioral geneticsin developmental psychopathology: Gene–environmentinterplay in antisocial behaviors. Psychological Bulle-tin, 131, 533–554.

Moffitt, T. E., Caspi, A., Dickson, N., Silva, P. A., & Stan-ton, W. (1996). Childhood-onset versus adolescent-onset antisocial conduct in males: Natural historyfrom age 3 to 18. Development and Psychopathology,8, 399–424.

Muthen, L. K., & Muthen, B. O. (1998–2005). Mplususer’s guide (3rd ed.) Los Angeles, CA: Author.

Neale, M. C., & Cardon, L. R. D. (1992). Methodology forgenetic studies of twins and families. Dordrecht:Kluwer.

Neiss, M., Rowe, D. C., & Rodgers, J. L. (2002). Does edu-cation mediate the relationship between IQ and age offirst birth? A behavioural genetic analysis. Journal ofBiosocial Science, 34, 259–275.


Peterson, J. L., & Zill, N. (1986). Marital disruption,parent–child relationships, and behavior problems inchildren. Journal of Marriage and the Family, 48,295–307.

Petitti, D. B., Friedman, G. D., & Kahn, W. (1981). Accu-racy of information on smoking habits provided on self-administered research questionnaires. American Jour-nal of Public Health, 71, 308–311.

Purcell, S., & Koenen, K. C. (2005). Environmental media-tion and the twin design. Behavior Genetics, 35, 491–498.

Rasanen, P., Hakko, H., Isohanni, M., Hodgins, S., Jarvelin,J., & Tiihonen, J. (1999). Maternal smoking duringpregnancy and risk of criminal behavior among adultmale offspring in the Northern Finland 1966 birth co-hort. American Journal of Psychiatry, 156, 857–862.

Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical lin-ear models: Applications and data analysis methods(2nd ed.) Thousand Oaks, CA: Sage.

Rodgers, J., Cleveland, H., van den Oord, E., & Rowe, D.(2000). Resolving the debate over birth order, familysize, and intelligence. American Psychologist, 55,599–612.

Rodgers, J. L. (1996). NLSY youth linking algorithm. Un-published manuscript.

Rodgers, J. L., Buster, M., & Rowe, D. C. (2001). Geneticand environmental influences on delinquency: DF anal-ysis of NLSY kinship data. Journal of QuantitativeCriminology, 17, 145–168.

Rodgers, J. L., Rowe, D. C., & Buster, M. (1999). Nature,nurture and first sexual intercourse in the USA: Fittingbehavioural genetic models to NLSY kinship data.Journal of Biosocial Science, 31, 29–41.

Rodgers, J. L., Rowe, D. C., & Li, C. (1994). Beyond nature vs.nurture: DF analyses of nonshared influences on problembehaviors. Developmental Psychology, 30, 374–384.

Rodgers, J. L., Rowe, D. C., & May, K. (1994). DF analysisof NLSY IQ/achievement data: Nonshared environ-mental influences. Intelligence, 19, 157–177.

Rodriguez, A., & Bohlin, G. (2005). Are maternal smokingand stress during pregnancy related to ADHD symp-toms in children? Journal of Child Psychology and Psy-chiatry, 46, 246–254

Rutter, M., Pickles, A., Murray, R., & Eaves, L. J. (2001).Testing hypotheses on specific environmental causaleffects on behavior. Psychological Bulletin, 127, 291– 324.

Scarr, S., & McCartney, K. (1983). How people make theirown environments: A theory of genotype–environmenteffects. Child Development, 54, 424–435.

Schafer, J. L. (1997). Analysis of incomplete multivariatedata. London: Chapman & Hall.

Silberg, J. L., Parr, T., Neale, M. C., Rutter, M., Angold, A.,& Eaves, L. J. (2003). Maternal smoking during preg-nancy and risk to boys’ conduct disturbance: An exam-

ination of the causal hypothesis. Biological Psychiatry,53, 130–135.

Tambs, K., Sundet, J. M., Magnus, P., & Berg, K. (1989).Genetic and environmental contributions to the covar-iance between occupational status, educational attain-ment, and IQ: A study of twins. Behavior Genetics,19, 505–522.

Thapar, A., Fowler, T., Rice, F., Scourfield, J., van denBree, M., Thomas, H., et al. (2003). Maternal smokingduring pregnancy and attention deficit hyperactivity dis-order symptoms in offspring. American Journal of Psy-chiatry, 160, 1985–1989.

Turkheimer, E., D’Onofrio, B. M., Maes, H. H., & Eaves, L.J. (2005). Analysis and interpretation of twin studieswith measured environments. Child Development, 76,1217–1233.

Turley, R. N. L. (2003). Are children of young mothers dis-advantaged because of their mother’s age or familybackground? Child Development, 74, 465–474.

Wakschlag, L. S., & Hans, S. L. (2002). Maternal smokingduring pregnancy and conduct problems in high-riskyouth: A developmental framework. Development andPsychopathology, 14, 351–369.

Wakschlag, L. S., & Keenan, K. (2001). Clinical signifi-cance and correlates of disruptive behavior symptomsin environmentally at-risk preschoolers. Journal ofClinical Child Psychology, 30, 262–275.

Wakschlag, L. S., Lahey, B. B., Loeber, R., Green, S. M., Gor-don, R. A., & Leventhal, B. L. (1997). Maternal smokingduring pregnancy and the risk of conduct disorder in boys.Archives of General Psychiatry, 54, 670–676.

Wakschlag, L. S., Pickett, K. E., Cook, E., Benowitz, N. L., &Leventhal, B. L. (2002). Maternal smoking duringpregnancy and severe antisocial behavior in offspring: Areview. American Journal of Public Health, 92, 966–974.

Weissman, M. M., Warner, V., Wickramaratne, P. J., &Kandel, D. B. (1999). Maternal smoking during preg-nancy and psychopathology in offspring followed toadulthood. Journal of the American Academy of Child& Adolescent Psychiatry, 38, 892–899.

Wichman, A. L., Rodgers, J. L., & MacCallum, R. C.(2006). A multilevel approach to the relationship be-tween birth order and intelligence. Personality and So-cial Psychology Bulletin, 32, 117–127.

Willett, J. B., Singer, J. D., & Martin, N. C. (1998). The de-sign and analysis of longitudinal studies of developmentand psychopathology in context: Statistical models andmethodological recommendations. Development andPsychopathology, 10, 395–426.

Zimmer, M. H., & Zimmer, M. (1998). Socioeconomic de-terminants of smoking behavior during pregnancy. TheSocial Science Journal, 35, 133–142.


smoking during pregnancy and offspring externalizing ... 20081.pdf · with offspring adhd. the...

Documents