Prenatal Alcohol Exposure Predicts Continued Deficitsin Offspring Size at 14 Years of Age

N. L. Day, S. L. Leech, G. A. Richardson, M. D. Cornelius, N. Robles, and C. Larkby

Background: Growth deficits are among the cardinal features for the diagnosis of fetal alcohol syn-drome. Growth deficits have also been noted among those who were exposed to alcohol prenatally but whodo not have fetal alcohol syndrome. Few studies have observed subjects past early and middle childhood,however, to evaluate the longer-term effects of prenatal alcohol exposure on growth in adolescence. Thisis a report of the effects of alcohol exposure during gestation on the size of the offspring at 14 years of age.

Methods: Women were recruited in their fourth prenatal month. These women were interviewed in thefourth and seventh months of pregnancy and at delivery. The women and their children were seen when theoffspring were 14 years of age.

Results: Growth deficits associated with prenatal alcohol exposure were identified among the offspringat 14 years of age. Weight, height, head circumference, and skinfold thickness continued to be significantlyaffected by prenatal alcohol exposure after controlling for other significant predictors of size. These effectsexhibited a dose-response pattern, and significant effects were found at levels below one drink per day. Forexample, first trimester alcohol exposure predicted weights of 152 lbs for the offspring of abstainers, 149 lbsfor the offspring of light drinkers (�0 and �0.2 drinks per day), 143 lbs for the offspring of moderatedrinkers (�0.2 and �0.89 drinks per day), and 136 lbs for the offspring of heavy drinkers (�0.89 drinks perday).

Conclusions: Prenatal alcohol exposure continues to affect size at age 14 years in this cohort of childrenfollowed since their fourth month of gestation.

Key Words: Alcohol, Prenatal, Growth, Adolescence.

GROWTH DEFICITS ARE one of the three character-istics required for the diagnosis of fetal alcohol syn-

drome (FAS) (Sokol and Clarren, 1989). The effects ofprenatal alcohol exposure on growth, however, are notlimited to those who have FAS. Partial expressions of thesyndrome, or fetal alcohol effects, have also been found inoffspring who are exposed to alcohol during gestation.

The Maternal Health Practices and Child DevelopmentProject (MHPCD) is a longitudinal study of children whowere prenatally exposed to alcohol. Analyses of data fromthe MHPCD project at birth, 8 and 18 months, and 3 and6 years of age demonstrated that deficits in height, weight,head circumference, palpebral fissures, and skinfold thick-ness are detectable at exposure levels of below one drinkper day. These analyses also showed that associations be-tween exposure and size are best modeled as a dose-response relation (Day et al., 1989, 1990, 1991a,b, 1994).

At age 10 years, in the MHPCD cohort, there was asignificant association between prenatal alcohol exposureand the child’s weight after controlling for significant co-variates such as other substance exposure during preg-nancy. There was a 4-lb decrease in weight for a change infirst trimester alcohol exposure from zero to an averagedaily volume (ADV) of one drink. The comparable statis-tics for the second and third trimesters were 7 and 7.1 lbs,respectively (Day et al., 1999). Alcohol exposure also sig-nificantly predicted reduced height among the offspringwhile controlling for all of the significant covariates. Theeffect sizes were 0.53, 0.8, and 0.64 inches for a change inADV from zero to one for the first, second, and thirdtrimesters, respectively. Smaller head circumference at age10 years was predicted by exposure during the first andthird trimesters of pregnancy. A reduction in skinfold thick-ness was predicted by third trimester alcohol exposure.

Most studies have reported deficits in growth at birth dueto prenatal alcohol exposure, although there are inconsis-tent findings. In a parallel study of teenage mothers fromthe MHPCD project, second trimester alcohol exposurewas significantly related to smaller head and chest circum-ference and lower birth weight (Cornelius et al., 1999).Alcohol use, measured in the fifth month of gestation,predicted reduced birth weight, length, and head circum-ference in the Seattle Longitudinal Prospective Study (Stre-

From the Western Psychiatric Institute and Clinic of the University ofPittsburgh Medical Center, Pittsburgh, Pennsylvania.

Received for publication March 27, 2002; accepted August 2, 2002.Supported by Grants AA06666, DA03874, and HD336890 (N. L. Day,

principal investigator).Reprint requests: Nancy L. Day, PhD, Western Psychiatric Institute and

Clinic, 3811 O’Hara St., Pittsburgh, PA 15213-2593; Fax: 412-681-1261;E-mail: nday@pitt.edu.

Copyright © 2002 by the Research Society on Alcoholism.

DOI: 10.1097/01.ALC.0000034036.75248.D9

0145-6008/02/2610-1584$03.00/0ALCOHOLISM: CLINICAL AND EXPERIMENTAL RESEARCH

Vol. 26, No. 10October 2002

1584 Alcohol Clin Exp Res, Vol 26, No 10, 2002: pp 1584–1591

issguth et al., 1981). Smith et al. (1986) found a significantrelation between duration and amount of alcohol use dur-ing pregnancy and birth weight, length, and head circum-ference. Fried and O’Connell (1987) reported a significantreduction in birth weight and length among the offspring ofwomen who drank more than two drinks per day, averagedacross pregnancy, compared with the remainder of thesample. Jacobson et al. (1994) found that exposure toalcohol, averaged across pregnancy, was associated withdecreased birth weight, length, and head circumference,although only among offspring of women older than 30years. Two studies (Ernhart et al., 1985; Russell and Skin-ner, 1988) found no association between prenatal alcoholexposure and birth size. These differences are likely due tomethodological differences among the studies, particularlyin measurement of alcohol use, as well as the use of cohortswith different sociodemographic characteristics.

At older ages, the effects of prenatal alcohol use ongrowth deficits have not been as consistent. In the Atlantastudy, 68 children from the original cohort were assessed atages ranging from 5 to 8 years (Coles et al., 1991). Theoffspring who were exposed continuously to alcoholthroughout pregnancy had smaller head circumferencesthan children who were not exposed to alcohol and thanthose who were exposed only in the first and second tri-mester, although they did not differ in weight or height. Inthe Buffalo study, at 6 years, head circumference was neg-atively associated with alcohol use in the year before preg-nancy, but there were no effects of prenatal alcohol expo-sure on weight or height (Russell et al., 1991). Sampson etal. (1994) reported that growth deficits due to alcoholexposure were not detectable from the age of 8 monthsthrough 14 years of age.

This analysis explores the relations between prenatalalcohol exposure during pregnancy and offspring size at 14years of age. Women were recruited from an outpatientprenatal clinic between May 1983 and July 1985 and werefirst interviewed in their fourth prenatal month. The assess-ment of the 14-year-olds began in July 1998 and was com-pleted in June 2001. The Institutional Review Board of theUniversity of Pittsburgh has approved each phase of theprotocol.

METHODS

Sample Selection and Study Design

Women who were in their fourth month of pregnancy and were 18 yearsor older were approached for interview. Those who were not interviewedat their fourth-month prenatal visit were contacted again at their fifth-month visit. If the interview was not completed at this point, they were notcontacted further. The refusal rate at recruitment was 15%.

The initial interview was administered to 1360 women. The studycohort was selected from this group. Women who averaged three or moredrinks per week in the first trimester and a random sample of one third ofthe women who drank alcohol less often or not at all were selected.Another cohort, selected to study the effects of marijuana use duringpregnancy, included women who used marijuana during the first trimesterat the rate of two or more joints per month and a random sample of

women who used less than this amount or not at all. Sampling was donewith replacement, so women could be in either or both of the cohorts. Thecohorts were combined for this report. The two studies were conducted inparallel, and all data collection techniques and assessment instrumentswere identical.

In addition to the fourth-month assessment, the women selected for thestudy cohort were interviewed at 7 months, at delivery, at 8 and 18 months,and at 3, 6, 10, and 14 years. The children were evaluated at each phaseafter birth. An assessment at age 16 years is ongoing. The core datasetused at each phase includes demographic characteristics, maternal phys-ical and psychological status, and an assessment of the mother’s use ofalcohol, tobacco, marijuana, and other drugs, including other illicit drugs,and prescribed and over-the-counter medications. At each study phase,growth, morphological abnormalities, and developmental, cognitive, andbehavioral characteristics of the offspring are assessed. Multiple aspects ofthe mother’s and child’s environments are also evaluated.

The initial sample of 829 women was reduced to 763 by delivery. The66 losses before delivery included 21 women who moved out of thePittsburgh area, 16 who were lost to follow-up, 8 who refused the deliveryinterview or newborn examination, 2 multiple births, 18 deaths, and 1adoption. At 14 years, we were unable to interview six children who wereplaced for adoption or in foster care or were institutionalized. Forty-eightmothers refused to participate, 69 were lost to follow-up, 51 moved out ofstate, 3 lost custody and the children could not be traced, and 6 childrendied. Thus, 580 mother/child pairs were assessed at 14 years. Thefollow-up rate, for those who were eligible for follow-up (n � 697), was83%, which represented 76% of the live-born singleton births (n � 763).We monitor the effects of subject loss, particularly with regard to prenatalalcohol use, other substance use, race, and child’s health. At age 14, therates of attrition did not differ by any of these characteristics. There werealso no significant differences in attrition between the mothers who drankthree or more drinks a week during the first trimester compared withwomen who drank less than that amount in the first trimester.

Thirteen children were removed from the analysis due to conditionsthat would affect growth, including major physical anomalies and disor-ders such as Down’s syndrome, sickle cell anemia, and cerebral palsy. Twoother teenagers were ineligible, one was too old at the assessment, and thegrowth measurements were missing in another. This resulted in a cohortof 565 for the analyses.

At 14 years, 11.5% (n � 64) of the children were not living with theirmothers. In these cases, the current caregiver provided information aboutthe child and the child’s environment. For simplicity, we have used theterminology of mother and maternal characteristics throughout.

Sample Description

The women were selected from a prenatal clinic. They were healthy, oflower socioeconomic status, and 18 years of age or older. Forty-six percentof the women were Caucasian; the remainder were African American,reflecting the distribution of the clinic population. At the fourth prenatalmonth interview, 86% of the women had completed high school, theirmean age was 23 years (range, 18 to 42 years), and 62% had a familyincome of less than $400 per month. A majority (68%) of the women werenot married, and 30% were primigravidous. The mean alcohol use in thefirst trimester was 0.6 drinks per day (range, 0 to 20). On average, thewomen smoked 8.0 cigarettes per day (range, 0 to 50), and mean mari-juana use was 0.4 joints per day (range, 0 to 9) in the first trimester.Cocaine use was reported by 3.2% of the women, and 9.0% reported illicitdrug use other than marijuana or cocaine.

At birth, 7.8% of the offspring were premature (�37 weeks gestation),and 11% were small for gestational age (less than the 10th percentile forgestational age). The average birth weight in the sample was 3197 g(range, 1150 to 4990 g), and 48% of the children were male. At age 14, thechildren weighed an average of 146 lbs (range, 78 to 304 lbs). Their meanheight was 65 inches (range, 56 to 76 inches), the mean head circumfer-ence was 559 mm (range, 503 to 610 mm), and the average skinfoldthickness was 15.2 mm (range, 2 to 50 mm). The children were, on

PRENATAL ALCOHOL EXPOSURE PREDICTS SIZE DEFICITS AT AGE 14 1585

average, 14.8 years of age (range, 13.9 to 16.2 years). Two of the girls werepregnant at the time of the assessment.

Alcohol was commonly consumed in the home at 14 years postpartum;25.0% of the children’s mothers drank one or more drinks per day, and24.5% were abstainers. Most of the mothers were light to moderatedrinkers (Fig. 1). The mean educational level of the mothers was 12.5years, and they were, on average, 39 years old, although the wide range inage (20 to 74 years) reflects the fact that 11.5% of the children were livingwith siblings, foster parents, or grandparents. The average income level inthe households was $1915 per month. In 75% of the households, themother worked or attended school. Fifty-two percent of the women cur-rently had an adult male living in the household; 43% of the women weremarried at 14 years postpartum. Fifty-seven percent of the men currentlyin the household were the biological fathers of the target offspring.

Measurement of Substance Use

Substance use was assessed at fixed time points for each trimester ofpregnancy. Usual, maximum, and minimum quantity and frequency weredetermined for beer, wine, liquor, and wine and beer coolers. The meth-odology for assessment has been published elsewhere (Day and Robles,1989). At 14 years, maternal substance use was reported over the prioryear. Alcohol use was expressed as the average number of drinks per day,or ADV. ADV was categorized to allow analysis of covariance (AN-COVA). Four categories were defined: abstainer, no alcohol use duringthe trimester (ADV � 0); light, fewer than 1.5 drinks per week (ADV �0and �0.2); moderate, 1.5 drinks per week to less than one drink per day(ADV �0.2 and �0.89); and heavy, one or more drinks per day (ADV�0.89). ADV is calculated as

�number of drinks/week) � �4 weeks/month)/31 days/month

The cut-point of ADV �0.89 defines the level of one drink per day (7drinks/week � 4 weeks/31 � 0.90). A variable was also created to repre-sent heavier quantities of drinking per occasion. Frequent heavy drinking(FHD) was defined as the frequency with which a woman drank four ormore drinks per occasion. This measure was also dichotomized to repre-sent binge versus nonbinge drinking.

Marijuana was assessed by using the same measures as were used foralcohol. Hashish and sinsemilla were added to the marijuana measure byusing a ratio of three joints of marijuana for each use of hashish and twojoints for each use of sinsemilla. These values represent the relativeamounts of �-9 tetrahydrocannabinol in the three substances (Hawks andChiang, 1986; Julien, 1997). The summary marijuana measure was parallelto that defined for alcohol and was labeled average daily joints (ADJ).Tobacco use was measured as the average number of cigarettes per day.For these analyses, cocaine use was expressed as a dichotomous variable(use/no use). Other illicit drugs, such as amphetamines and barbiturates,were combined and expressed as a dichotomous variable for each prenatalassessment period.

Measurement of Size

Assessors were trained to reliability and were unaware of the prenatalexposure status of the children. The child’s weight (lbs), height (inches),and head circumference (mm) were measured at each assessment. Tricepsskinfold thickness (mm) was measured with calipers.

Child Variables

Puberty was measured with the Pubertal Development Scale (Petersenet al., 1988), a self-report measure of pubertal status that reflects thesequence of pubertal development. Scores can be used as a continuousvariable or grouped into five categories of pubertal status; prepubertal;early, mid, or late puberty; or postpubertal. We used the categoricalvariable in the analyses.

Two other variables, number of illnesses and number of hospitaliza-tions, were ascertained from the mother at each assessment. These wereused as direct counts in the analysis.

Measures of the Current Environment

Maternal education, current work status, and family income were usedto assess socioeconomic status. The psychological environment includedmeasures of maternal depression, by using the Center for EpidemiologicStudies-Depression Scale (Radloff, 1977), and anxiety and hostility, byusing the Spielberger State-Trait Anxiety Inventory (Spielberger et al.,1970). Items from the PERI Life Events Scale (Dohrenwend et al., 1978)were adapted to measure stressful life events. Maternal social support wasdetermined by using a measure developed by the Human PopulationLaboratory (Berkman and Syme, 1979).

Household structure was measured as the number of and distance inage between siblings and the presence of a male in the household. Thequality of the home environment was assessed with an interview version ofthe HOME (Home Observation for Measurement of the Environment)questionnaire (Baker and Mott, 1989). Current maternal substance usewas measured with the same instruments that were used in the prenatalassessments.

Statistical Methods

The response variables and the independent variables were evaluatedby using histograms to explore the shape of the distributions and toidentify extreme values. Height, head circumference, and skinfold thick-ness were approximately normally distributed. Weight was skewed, and alog transformation was used to normalize it for the regression analyses.Scatter plots were used to investigate the bivariate relations between eachoutcome variable and the independent variables. Separate stepwise re-gression models were developed for each growth outcome and for expo-sure at each trimester to identify the significant independent variables. Inthe selection process, the significance level for entry was set low to identify

Fig. 1. Maternal drinking patterns during pregnancy and at 14 years. GT, greater than; LE; less than or equal to.

1586 DAY ET AL.

all possible significant predictors. Before determining the final model foreach outcome, we checked regression model assumptions by using resid-uals and fitted values. No violations were detected. As a final step ofmodel fitting, we identified the influential points and assessed the extentof their influence. The associations between prenatal alcohol exposure andsize did not change with the removal of influential points.

The covariates listed in Table 1 were used for model development.These covariates were selected from an initial bivariate screen of relationsbetween variables. They also correspond to a list of covariates recom-mended for consideration in studies of teratogenic exposures (Jacobsonand Jacobson, 1990). Maternal height was included to control for thegenetic contribution of the mother to the size of the child.

Prenatal alcohol, marijuana, and tobacco exposure were used as con-tinuous variables, expressed as ADV, ADJ, and number of cigarettes perday, respectively. In the regression analyses, ADV and ADJ were trans-formed to reduce skewness.

RESULTS

There was a substantial decrease in alcohol use duringpregnancy (Fig. 1). Although 38% of the women drank oneor more drinks per day before pregnancy, by the firsttrimester, only 18.9% of the women drank at this rate. Fourpercent of the women reported drinking at this level in thethird trimester.

There were significant differences in the level of alcoholconsumption by race, income level, and presence of a malein the household. In the first trimester, whites were more

likely to drink heavily than were African Americans. How-ever, the white women were more likely to decrease orabstain from drinking later in pregnancy, so that by thethird trimester, heavy drinkers were predominately AfricanAmerican. Monthly income level was lowest among theheavier drinkers ($1663) and abstainers ($1836) and high-est among the light and moderate drinkers ($2078 and$2076, respectively; p � 0.05). Light drinkers had the high-est proportion of males in the household (64%) comparedwith abstainers (48%).

Stepwise linear regression analyses were used to explorethe relations between the growth parameters and prenatalexposure to alcohol. These analyses were initially run withpubertal status included. However, missing data on thismeasure meant that 50 cases would be excluded. Regres-sions were performed with pubertal status included andexcluded: the results were identical. Thus, the relationsbetween alcohol and size are independent of pubertal sta-tus. Because there were no differences, we have presentedthe results with puberty excluded to take advantage of thelarger sample size. Two of the girls were pregnant at the14-year assessment. The analyses were performed withthese cases included and excluded, and the results did notdiffer. These subjects were left in the cohort.

In the regression analyses, first and second trimesterexposures to alcohol predicted significant reductions inweight at 14 years of age. Third trimester exposure did nothave a significant effect on the offspring’s weight at 14years. The number of siblings; maternal height; child’s age,gender, race, and number of hospitalizations; and maternaltobacco use during pregnancy were also significant predic-tors of weight (Table 1). The effects of alcohol exposurereported previously were found while controlling for theseother significant predictors of growth.

Height was significantly and negatively associated withprenatal alcohol exposure during the first trimester. Secondand third trimester exposures did not predict decrements inheight at 14 years of age. Other significant covariates ofheight at age 14 were number of siblings, maternal heightand current maternal tobacco use, and child’s age, gender,and race (Table 1).

Head circumference was also significantly predicted byfirst trimester alcohol exposure. At age 14, second and thirdtrimesters were not significant predictors of head circum-ference. The other significant covariates of head circum-ference included number of siblings, maternal education,maternal height, gender, race, and prenatal tobacco expo-sure (Table 1).

We found that skinfold thickness was significantly asso-ciated with second trimester alcohol exposure, although itwas not related to exposure during either the first or thirdtrimester. Skinfold thickness was also predicted by familyincome, maternal social support, and gender.

The regressions were also run with the FHD measure,the frequency of drinking four or more drinks per occasion(binge drinking). More than one third of the women re-

Table 1. Variables Considered in the Development of the Model

Covariates in the initial model

Outcomes predicted byeach covariate

(p � 0.05)

Current environmental variablesNo. siblings wt, ht, hcAge span, oldest to youngest siblingAdult male in the householdFamily income sfHOME scoreLife events

Current maternal variablesEducation hcWork statusAgeHeight wt, ht, hcSocial support sfDepressionHostilityAnxietyTobacco use htAlcohol useMarijuana useCocaine useIllicit drug use

Child variablesAge wt, htGender wt, ht, hc, sfRace wt, ht, hcNo. hospitalizations since last assessment wtNo. illnesses since last assessmentPubertal status wt, ht, hc, sf

Prenatal substance useTobacco use during each trimester wt, hcMarijuana use during each trimesterIllicit drug use during each trimesterCocaine use during each trimester

wt, weight; ht, height; hc, head circumference; sf, skinfold thickness.

PRENATAL ALCOHOL EXPOSURE PREDICTS SIZE DEFICITS AT AGE 14 1587

ported drinking four or more drinks per occasion at sometime during their pregnancy: 34% during the first trimesterand 4% and 5% during the second and third trimesters,respectively. There were no significant associations be-tween FHD and any of the growth parameters at anytrimester. When FHD was dichotomized to binge/no binge,there were significant associations between binge drinkingin the first trimester and offspring weight and head circum-ference, and between second trimester exposure and off-spring weight. Within the group of women who binged,however, there was not a dose-response relation betweenthe level of exposure and outcome.

ANCOVA was used to illustrate the dose-response na-ture of the relations and to estimate the effect sizes. Thiswas necessary because we used the log of alcohol in theregression analyses, and, as a result, these analyses couldnot be used to estimate the effect size. The significantcovariates from the regression models were used to adjustthe means within categories in the ANCOVAs. The resultsfrom these analyses were the same as the regression anal-yses. The only exception was that the significant relationfound in the regression analysis between height and firsttrimester alcohol exposure was not significant in the AN-COVA, given the specific categories we used for analysis.

In the first trimester, children whose mothers were lightdrinkers during the first trimester weighed 2.5 lbs less thanthe offspring of abstainers (Fig. 2). The effect size formoderate drinking was 9.1 lbs compared with abstinence,and the difference between abstinence and heavy drinkingwas 16 lbs. The differences were even larger when the effectof drinking in the second trimester on weight was evaluated(Fig. 2). Differences of 9.9, 8.5, and 19 lbs were found whenthe differences were calculated between abstinence andexposure to light, moderate, and heavy drinking, respec-tively, during the second trimester. All of the previousfindings reflect the effect sizes after adjusting for the sig-nificant covariates. It is notable that the data demonstrate

a dose-response relation across all three trimesters, al-though the relation is not significant for the third trimester.

The same dose-response relation can be seen for heightat 14 years (Fig. 3). After adjustment for the covariates,there is a pattern of decreasing height as the amount ofprenatal alcohol exposure increases. Only first trimesterresults are statistically significant, however.

For head circumference, the difference between drinkingone drink or more per day in the first trimester and absti-nence was 6.6 mm (Fig. 4). The difference between expo-sure to one drink or more per day in the second trimesterand abstinence was 5.76 mm in skinfold thickness (Fig. 5).As with weight and height, these results for head circum-ference and skinfold thickness demonstrate a dose-response pattern.

DISCUSSION

Growth deficits related to prenatal alcohol exposure arestill detectable at 14 years of age in this cohort. The ado-lescents are significantly smaller in weight, height, headcircumference, and skinfold thickness. These differences,although generally small, have been consistently noted inthis cohort at each assessment. Importantly, the growthdeficits have a dose-response relation to gestational expo-sure, and effects on growth are detectable at exposures thatare considerably below one drink per day. As presentedpreviously, for example, first trimester alcohol exposurepredicted weights of 152, 149, 143, and 136 lbs for theoffspring of abstainers and light, moderate, and heavydrinkers, respectively. These results remained significantafter controlling for other significant covariates, includingthe use of other substances during pregnancy.

At this phase, as in our previous analyses, we found thatthe relation between prenatal alcohol exposure and sizewas linear, or dose-response. The ADV of drinking, ratherthan FHD, best predicted this association. There is an

Fig. 2. Association between alcohol exposure by trimester and offspring weight after controlling for significant covariates of weight. *Significant in regressionanalysis. GT, greater than; LE; less than or equal to.

1588 DAY ET AL.

obvious association between these two measures, becausewomen with a high ADV are also likely to be the frequentheavy drinkers. The significant association between thedichotomized binge drinking measure and size was a reflec-tion of this relation. This implies that, for growth, a patternof continuous use, rather than episodic heavy use, is themost important.

One other study has observed offspring to this age andthis study did not find an association between prenatalalcohol exposure and size at age 14 years (Sampson et al.,1994). One explanation may be that that study had only onemeasure in mid pregnancy, whereas we were able to assessdrinking at each trimester. Given that many of our findingsare first trimester effects, this may have been a significantdifference. Another explanation may be that the Sampsonet al. study had a middle-class sample, whereas our cohortis low-income. The women in our cohort were light to

moderate drinkers, but their environments were stressful,and they and their children have high rates of medical,economic, psychiatric, social, and legal problems. It may bethat these additional stressors made the growth deficitsmore apparent among these exposed children. Alterna-tively, the less privileged environment may decrease thechildren’s ability for catch-up growth, given the teratogenicinsult. Our findings parallel case reports on individuals withFAS, in whom growth deficits persist into adolescence.

It is an advantage that our cohort represents mostly lightto moderately exposed offspring. This has allowed us tolook at the effects of lower levels of exposure in contrast toother studies that have focused on the offspring of heavydrinking or alcoholic women. Our cohort is relatively ho-mogeneous with respect to demography, however, and wecannot extrapolate our results to a more advantaged pop-ulation. The cohort is racially balanced, and the data rep-

Fig. 3. Association between alcohol exposure by trimester and offspring height after controlling for significant covariates of height. *Significant in regression analysis.GT, greater than; LE; less than or equal to.

Fig. 4. Association between alcohol exposure by trimester and offspring head circumference after controlling for significant covariates. *Significant in regressionanalysis. GT, greater than; LE; less than or equal to.

PRENATAL ALCOHOL EXPOSURE PREDICTS SIZE DEFICITS AT AGE 14 1589

resent the findings on a cohort of lower-income AfricanAmerican and Caucasian women and their 14-year-oldchildren.

Prenatal exposure to alcohol is associated with significantdeficits in growth through 14 years of age in our cohort.These deficits are detectable at levels considerably belowone drink per day, and the association is a dose-responserelation. In addition, the deficits are associated with expo-sure early in pregnancy, which has considerable implica-tions for the timing of prevention efforts. The clinical sig-nificance of these growth deficits is not apparent, anddeficits of this magnitude would not be noticed in a smallersample of exposed children. This does not mean, however,that they are not important. These continued growth defi-cits serve as a potentially permanent marker of a terato-genic exposure, and we need to understand better themechanisms that maintain these deficits.

REFERENCES

Baker P, Mott F (1989) National Longitudinal Study of Youth Child Hand-book. Ohio State University Center for Human Resource Research,Columbus, OH.

Berkman LF, Syme SL (1979) Social networks, host resistance, and mor-tality: a nine-year follow-up study of Alameda County residents. Am JEpidemiol 109:186–204.

Coles C, Brown R, Smith I, Platzman K, Erickson S, Falek A (1991)Effects of prenatal alcohol exposure at school age. I. Physical andcognitive development. Neurotoxicol Teratol 13:357–367.

Cornelius M, Goldschmidt L, Taylor P, Day N (1999) Prenatal alcohol useamong teenagers: effects on neonatal outcomes. Alcohol Clin Exp Res23:1238–1244.

Day N, Goldschmidt L, Robles N, Richardson G, Cornelius M (1991a)Prenatal alcohol exposure and offspring growth at eighteen months ofage: the predictive validity of two measures of drinking. Alcohol ClinExp Res 15:914–918.

Day N, Jasperse D, Richardson G, Robles N, Sambamoorthi U, Taylor P,Scher M, Stoffer D, Bloom M (1989) Prenatal exposure to alcohol:effect on infant growth and morphologic characteristics. Pediatrics 84:536–541.

Day N, Richardson G, Geva D, Robles N (1994) Alcohol, marijuana andtobacco: the effects of prenatal exposure on offspring growth andmorphology at age six. Alcohol Clin Exp Res 18:786–794.

Day N, Richardson G, Robles N, Scher M, Stoffer D, Jasperse D, Cor-nelius M (1990) The effect of prenatal alcohol exposure on growth andmorphology of the offspring at eight months of age. Pediatrics 85:748–752.

Day N, Robles N (1989) Methodological issues in the measurement ofsubstance use. Ann NY Acad Sci 562:8–13.

Day N, Robles N, Richardson G, Geva D, Taylor P, Scher M, Stoffer D,Cornelius M, Goldschmidt L (1991b) The effects of prenatal alcohol useon the growth of children at three years of age. Alcohol Clin Exp Res15:67–71.

Day NL, Zuo Y, Richardson GA, Goldschmidt L, Larkby CA, CorneliusMD (1999) Prenatal alcohol use and offspring size at 10 years of age.Alcohol Clin Exp Res 23:863–869.

Dohrenwend BS, Krasnoff K, Askenasy AR, Dohrenwend P (1978) Ex-emplification of a method for scaling life events: the Peri Life EventsScale. J Health Soc Behav 19:205–229.

Ernhart C, Wolf A, Linn P, Sokol R, Kennard M, Filipovich H (1985)Alcohol-related birth defects: syndromal anomalies, intrauterine growthretardation, and neonatal behavioral assessment. Alcohol Clin Exp Res9:447–453.

Fried P, O’Connell C (1987) A comparison of effects of prenatal exposureto tobacco, alcohol, cannabis, and caffeine on birth size and subsequentgrowth. Neurotoxicol Teratol 9:79–85.

Hawks R, Chiang C (1986) Examples of specific drug assays. NIDA ResMonogr 73:84–112.

Jacobson J, Jacobson S, Sokol R, Martier S, Ager J, Shankaran S (1994)Effects of alcohol use, smoking, and illicit drug use on fetal growth inblack infants. J Pediatr 124:757–764.

Jacobson JL, Jacobson SW (1990) Methodological issues in humanbehavioral teratology, in Advances in Infancy Research (Rovee-Collier C, Lipsitt LP, Hayne H eds), vol 6, pp 111–148. Ablex,Westport, CT.

Julien RM (1997) A Primer of Drug Action. A Concise, Nontechnical Guideto the Actions, Uses, and Side Effects of Psychoactive Drugs. 8th ed, pp319–349. WH Freeman, New York.

Petersen AC, Crockett L, Richards M, Boxer A (1988) A self-reportmeasure of pubertal status: reliability, validity, and initial norms. JYouth Adolesc 17:117–133.

Fig. 5. Association between alcohol exposure by trimester and offspring skinfold thickness after controlling for significant covariates. *Significant in regressionanalysis. GT, greater than; LE; less than or equal to.

1590 DAY ET AL.

Radloff L (1977) The CES-D Scale: a self-report depression scale forresearch in the general population. Appl Psychol Meas 1:385–401.

Russell M, Czarnecki D, Cowan R, McPherson E, Mudar P (1991) Mea-sures of maternal alcohol use as predictors of development in earlychildhood. Alcohol Clin Exp Res 15:991–1000.

Russell M, Skinner J (1988) Early measures of maternal alcohol misuse aspredictors of adverse pregnancy outcomes. Alcohol Clin Exp Res 12:824–830.

Sampson PD, Bookstein FL, Barr HM, Streissguth AP (1994) Prenatalalcohol exposure, birthweight and measures of child size from birth toage 14 years. Am J Publ Health 84:1421–1428.

Smith I, Coles C, Lancaster J, Fernhoff P, Falek A (1986) The effect ofvolume and duration of prenatal ethanol exposure on neonatal physicaland behavioral development. Neurobehav Toxicol Teratol 8:375–381.

Sokol R, Clarren S (1989) Guidelines for use of terminology describingthe impact of prenatal alcohol on the offspring. Alcoholism 13:597–598.

Spielberger C, Gorsuch R, Lushene R (1970) Manual for the State-TraitAnxiety Inventory. Consulting Psychologists Press, Palo Alto, CA.

Streissguth AP, Martin DC, Martin JC, Barr HM (1981) The SeattleLongitudinal Prospective Study on alcohol and pregnancy. NeurobehavToxicol Teratol 3:223–233. .0

PRENATAL ALCOHOL EXPOSURE PREDICTS SIZE DEFICITS AT AGE 14 1591