Effects of maternal smoking upon neuropsychological development in early childhood: importance of taking account of social and environmental factors

Pnediutric nnd Pninntnl Epidemiology 1992, 6, 403415

Original articles

Effects of maternal smoking upon neuropsychological development in early childhood: importance of taking account of social and environmental factors

Peter A. Baghurst', Shi-lu Tongt, Alistair Woodward$ and Anthony J. McMichaelS *CSlRO Division of Human Nutrition, Adelaide, Australia, tDepartment of Special Programmes, National Health Education Institute, Beijing, China, and $Department of Community Medicine, Univenity of Adelaide, Australia

Summary. Data from a prospective study of 548 children followed from birth to 4 years of age were analysed to determine whether maternal smoking during and/or after pregnancy affects children's neuropsychological development. The differences in mean develop- mental test scores between children whose mothers smoked and those of mothers who did not smoke were slight, with subscale scores only 2.4 to 4.1% lower in children whose mothers smoked. These differences were not statistically significant after adjustment for socio-economic status, quality of home environment and mother's intelligence, suggesting that the social and environmental factors are major confounders of the associ- ation of exposure to maternal smoking and neuropsychological de- velopment in childhood. In order to gain a better understanding of this area, more precise measures of exposure to environmental tobacco smoke and comprehensive consideration of confounders will be re- quired.

Introduction Although it is generally accepted that exposure of tobacco products from the mother's smoking during pregnancy is associated with both a reduction in

Address for correspondence: Peter A. Baghurst, CSIRO Division of Human Nutrition, PO Box 10041 Gouger St, Adelaide, South Australia 5000, Australia.

403

404 P.A. Bughurst et al. birthweight and an increase in perinatal mortality, uncertainties exist about the functional consequences of maternal smoking on offspring. Increasing attention has been paid to studying the relationship between exposure to maternal smoking and neuropsychological development in childhood, extending into early adult- hood. Several epidemiological studies have indicated that neuropsychological development in children may be impaired by such exposure, but a justified criticism of the investigations is that few prenatal and postnatal factors, other than social class, have been taken into account in these studies. Thus, the neuro- psychological differences between the children of smokers and non-smokers may be attributable to other plausible explanations.1,2

In cohorts followed in Britain, Canada and Finland, associations have been reported between prenatal exposure to maternal smoking and deficits in neuropsychological functioning of school-aged children.= A clinical trial of smoking cessation intervention during pregnancy6 has also indicated that cess- ation of smoking after becoming pregnant may prevent some cognitive damage to the fetus, which seems to confirm a possible effect of prenatal exposure to tobacco on neuropsychological development in early childhood. Nevertheless, the quality of the studies has vaned widely. Few of the earlier studies have controlled for potential confounders, such as maternal intelligence, psychological status of the parents and postnatal exposure to environmental tobacco smoke.

The US Collaborative Perinatal Project,’ which followed 29 889 children from birth to 7 years of age, has provided equivocal evidence of an association of cognitive deficit in children with exposure to maternal smoking. Although, in univariable analysis, the children of mothers who smoked 20 or more cigarettes per day had an increased risk for learning difficulties (LD), hyperkinetic-im- pulsive behaviour (HI) and neurological ‘soft signs’ (NS), no significant relation- ship between prenatal exposure to maternal smoking and LD or NS in children was found in the final multiple discriminant model. Mother‘s smoking during pregnancy, in this study, was identified as one of the significant discriminators between children with HI and those who did not have HI. However, the effect of maternal smoking on child’s hyperkinetic-impulsive behaviour was not as large as other factors (such as occurrence of siblings with HI, and head circumference at birth).

Some participating centres in the US Collaborative Perinatal Project reported inconsistent results. The data collected at the Boston Lying-In Hospital indicated that hyperactivity, short attention span, and lower scores on spelling and reading tests were more frequent for children whose mothers had smoked throughout pregnancy.8 The cognitive decrements were mild, with achievement test scores only 2 to 4% lower in children whose mothers had smoked during pregnancy.

Hardy and Mellits reported that, in a matched-pair prospective study con- ducted in east Baltimore, there was no significant difference in intellectual func- tioning between children of mothers who smoked during pregnancy and children

Maternal smoking and neuropsychological development 405 whose mothers did not smoke.9 As Donovan pointed out, however, the sample size of the investigation was small (88 pairs), and the statistical power of the investigation was 10w.l~ Thus, the results of this study do not exclude the existence of harmful effects of smoking while pregnant.

Since the possible long-term effects of maternal smoking on childhood neuropsychological development are of great theoretical and practical import- ance, the question has emerged of whether the reported lower neuropsycho- logical functioning of children can be attributed to exposure to maternal smoking or to other coexistent aspects of the child's social and environmental circumstances.

The results reported in this paper come from a follow-up study of 548 4-year- old children whose mothers were recruited during pregnancy in the Port Pirie Cohort Study. The analyses focus on the associations of the neuropsychological outcome in childhood with maternal smoking, and a number of key social and environmental factors. Evidence is provided that the decrements in children's neuropsychological functioning associated with antenatal or postnatal exposure to maternal smoking are not as large as those attributable to the child's social and environmental factors, and that appropriate adjustment for these factors may explain the mild association between exposure to maternal smoking and neuropsychological development in children.

Methods

The Port Pirie Cohort Study is a large collaborative epidemiological study in which the primary objective has been to examine the relation between exposure to environmental lead and early childhood growth and development. The study population presented an unusual opportunity to evaluate children's neuro- psychological development in relation to exposure to maternal smoking, while controlling for other covariates. The study has obtained detailed and formal measures of neuropsychological development in children, as well as measure- ments of the major social and environmental confounders, and details of the mother's smoking during and after pregnancy.

All women living in Port Pirie (population 16000), South Australia, or in the immediately adjoining region, who attended for the first time for antenatal care between May 1979 and May 1982, were encouraged to contact the study co- ordinator (community health nurse) to discuss participation. The participants completed a baseline questionnaire that included informa tion about their current and pre-pregnancy smoking habits, reproductive histories, sociodemographic and clinical characteristics, intake of alcohol, and diet. The women were defined as non-smokers if they had never smoked or had smoked no more than five cigarettes (total) during this pregnancy. Any changes in smoking status during the pregnancy were recorded in an interview at 32 weeks of gestation. The

406 P.A. Bughurst et al.

information on the child's postnatal exposure to maternal smoking was also obtained by questionnaire-interview. All interviews were carried out by one of four trained nurse-interviewers.

A total of 831 pregnant women were recruited into the study. These repre- sented an estimated 90% of all newly occurring pregnancies during this period. Of these 831 pregnancies, 723 were followed to single live births. The number of children remaining in the cohort was 548 at the age of 4 years. The majority (approximately 80%) of the children lost to follow-up were in the families that left the Port Pirie district; a smaller number of families (about 20% of those lost to follow-up) simply discontinued their participation.

The developmental status of each child was assessed with the use of the Bayley Scales of Infant Development (BSID) at age 2 years and McCarthy Scales of Children's Abilities (MSCA) at age 4 BSID comprises a Mental Scale of 163 items and Motor Scale of 81 items. The Mental Scale assesses sensory- perceptual acuities and discriminations, and ability to respond to these; it also assesses acquisition of 'object constancy', memory, learning, problem-solving ability, early language and speech development, and ability to form general- isations and classifications. Results are expressed as a single standard score-the Mental Development Index (MDI). The Motor Scale measures gross and fine motor co-ordination, which is expressed as Psychomotor Development Index (PDI) .

The MSCA consists of General Cognitive Index (GCI), and two additional subscales of memory and motor performance. The GCI is constructed by combin- ing scores on the verbal, perceptual-performance and quantitative subscales. Both BSID and MSCA are among tests that best predict subsequent neuro- psychological development in childhood.11,13

A research psychologist, who was unaware of the child's exposure to maternal smoking, conducted all testing sessions in a clinic setting. Although the psycholo- gist had also assessed the child's abilities (using BSID) at the age of 2 years, he was not aware of that earlier result when the MSCA was conducted.

Assessment of socio-economic status, using Daniel's Scale of Prestige of Occupations in A~stralia, '~ was made based on the information obtained from the mother at 16 weeks of gestation. On this Scale of Prestige of Occupation a lower Daniel's score represents greater prestige. Maternal intelligence was estimated with use of the Wechsler Adult Intelligence Scale-Revi~ed.~~ This test was administered in full by the research psychologist while the children were in the age range of 3 to 5 years. The care-giving environment, using the Home Observation for Measurement of the Environment (HOME) inventory,16 was assessed during a visit to the children's homes when they were 3 years of age, The detailed description of the measurement procedures has been reported else- where. 17,18

Maternal smoking and neuropsychological development 407

Statistical methods Statistical analysis was done by conventional regression techniques included in Genstat (A General Statistical Program). l9 In simple and multiple regression analyses, a dichotomous variable for maternal smoking was used. The effect of maternal smoking was obtained from the regression coefficient on a dummy independent variable which took on a value of zero if the mother was a non- smoker and one if the mother was a smoker. The regression analyses were conducted only on those children with complete data for all the variables con- sidered.

Results

Maternal smoking and children’s abilities The analyses of children’s scores on the Bayley and McCarthy scales of children’s abilities with postnatal exposure to maternal smoking (Table 1) showed that the children of smokers performed at 2.4 to 4.1% lower level in most of the testing sessions by comparison with those of non-smokers (PG 0.03). The mean values of MDI and GCI for children with postnatal exposure to mother‘s smoke were 107.7 (s.e.=l.O) and 105.3 (s.e.=1.1) compared to 110.4 (s.e.=0.8) and 108.8 (s.e.=0.8) for children whose mothers did not smoke. The decrements in MDI and GCI scores in the children of smokers were statistically significant (P=0.03 and 0.01, respectively). The children with postnatal exposure to maternal smoking also had lower scores for some other subscales of MSCA: the verbal, perceptual-perform- ance and motor subscales ( P S 0.03).

Table 1. children‘s abilities with exposure to maternal smoking

Children‘s mean scores (standard error) on the Bayley and McCarthy scales of

Prenatal Postnatal

Yes No P Yes No P

BSID MDI PDI

MSCA GCI

Verbal Perceptual- performance Quantitative

Memory Motor

n=160 108.4(1.2) 105.2( 1.1)

n=148 106.0(1.3) 52.7(0.8)

56.1(0.8) SO.O(O.7) 47.8(0.6) 52.9(0.7)

n=420 109.7(0.8) 105.3(0.7)

n=373 108.2(0.8) 53.9( 0.5)

57.7(0.4) Sl.O(O.5) 48.5(0.5) 54.3(0.4)

0.34 0.94

0.13 0.19

0.07 0.26 0.46 0.10

n=232 n=348 107.7(1.0) 110.4(0.8) 104.8(0.9) 105.7(0.8)

n=190 n=331 105.3( 1.1) 108.8( 0.8) 52.4(0.7) 54.3(0.6)

55.7(0.7) 58.1(0.5) sO.O(O.7) 51.1(0.5) 47.5(0.6) 48.8( 0.5) 52.6(0.6) 54.6(0.4)

0.03 0.46

0.01 0.03

co.01 0.17 0.10

co.01

408 P.A. Bughurst et al.

However, no significant differences were found between the scores of chil- dren with antenatal exposure to maternal smoking and those of children with no antenatal exposure.

Socio-environmental factors and children’s abilities An examination of the relationship of children’s scores on the Bayley and McCarthy scales of children’s abilities to soao-economic status, HOME scores and maternal intelligence indicated that the child’s neuropsychological development was significantly associated with soda1 and environmental factors (Tables 2, 3

Table 2. children’s abilities with socio-economic status

Children’s mean scores (standard error) on the Bayley and McCarthy scales of

- Daniel’s scores

<40 40-50 51-60 >a P

BSID MDI PDI

n=72 n=195 n=139 n=169 116.0(1.6) 109.6(1.1) llO.O(l.4) 105.7(1.1) <0.01 104.9(1.8) 104.3(1.0) 107.3(1.4) 105.0(1.1) 0.29

MSCA n=63 n=170 n=130 n=154 GCI 1 17.8( 1.6) 109.0( 1.0) 106.3( 1.4) 103.0( 1.3) <O. 01

Verbal 59.4(1.2) 54.3(0.7) 53.1(0.8) 50.9(0.8) cO.01 Perceptual-performance 62.4(0.8) 58.4(0.6) 56.6(0.8) 54.5(0.8) <0.01 Quantitative 55.7(1.0) 51.3(0.6) 49.9(0.8) 49.0(0.8) CO.01

Memory 52.7(1.1) 49.4(0.6) 47.4(0.8) 46.1(0.7) CO.01 Motor 57.7(0.8) 54.8(0.6) 53.3(0.8) 51.8(0.7) CO.01

Children’s mean scores (standard error) on the Bayley and McCarthy scales of Table 3. children’s abilities by home observation for measurement of the environment (HOME)

HOME scores

<35 35-40 41-45 >45 P

BSID n=50 n=127 n=219 n=152 MDI 96.6(2.0 104.7(1.2) 110.9(1.0) 115.6(1.2) cO.01 PDI 97.0(1.9) 105.1(1.4) 106.3(0.9) 106.6(1.2) <0.01

MSCA n=45 n=119 n-203 n=140 GCI 94.4(2.3) 101.5(1.4) lm.l(O.9) 114.4(1.3) CO.01

Verbal 46.4(1.4) 50.4(0.9) 54.2(0.6) 57.5(0.8) <0.01 Perceptual-performance 49.3(1.6) slr.O(O.8) 58.6(0.6) 60.6(0.7) CO.01 Quantitative 46.1(1.6) 47.9(0.8) 50.8(0.5) 54.4(0.8) cO.01

Memory 43.3(1.2) 45.9(0.8) 48.1(0.5) 51.8(0.8) cO.01 Motor 49.0(1.6) 50.9(0.7) 54.7(0.5) 56.7(0.7) cO.01

Maternal smoking and neuropsychological development 409 Table 4. Children’s mean scores (standard error) on the Bayley and McCarthy scales of children’s abilities with maternal intelligence

~~

Mother’s IQ

s80 81-90 91-100 >lo0 P

BSID n=43 n=108 n=129 n=93 MDI 98.0(2.5) 8.9(1.3) 110.3(1.2) 113.5(1.6) c0.01 PDI 101.3(2.1) 105.6(1.3) 107.8(1.3) 107.7(1.6) <0.01

GCI 93.1(2.8) 107.7(1.4) 107.7(1.3) 150.0(1.3) c0.01 Verbal 45.4( 1.5) 53.6(0.9) 53.8( 0.9) 57.4( 1 .O) cO.01 Perceptual-performance 48.5(1.6) 57.3(0.8) 57.6(0.8) 61.3(0.7) cO.01 Quantitative 45.8(1.8) 49.8(0.8) 50.6(0.7) 55.1(0.8) cO.01

Memory 42.3(1.4) 48.7(0.8) 47.9(0.8) 51.6(1.0) cO.01 Motor 46.2(1.4) 54.8(0.7) 54.0(0.7) 55.9(0.8) cO.01

MSCA n=37 n=103 n=127 n=90

Table 5. Estimated coefficients (P-value) of postnatal exposure to maternal smoking from simple and multiple regression analyses of children’s scores on Bayley and McCarthy scales of children’s abilities.

Adjusted for Adjusted for socio-economic socio-economic

Adjusted for status and status, HOME HOME so ci o - e c o n o m i c scores and

Unadjusted status scores mother‘s IQ

Bayley MDI

PDI

MSCA GCI

Verbal

Perceptual- performance

Quantita t- ive

Memory

Motor

-2.74

-0.90 (0.48)

(0.04)

-3.54

-1.92 (0.01)

(0.04)

-2.38 (CO.01)

-1.12 (0.18)

-1.30 (0.10) - 1.97

(<om)

-1.76 (0.18)

-0.82 (0.52)

-2.04 (0.13) - 1.02 (0.24)

-1.55 (0.06)

-0.48 (0.56)

-0.58 (0.47)

-1.34 (0.07)

-0.67 (0.60)

-0.20 (0.88)

-0.60

-0.25 (0.76)

(0.64)

0.13 (0.88)

-0.02 03-97)

,-0.75 (0.30)

-0.55

-0.17 (0.89)

(0.66)

-0.45

-0.17 (0.73)

(0.W

-0.67 (0.38)

0.21 (0.80) 0.06

(0.93) -0.68 (0.35)

410 P.A. Baghurst et al. Table 6. Estimated coefficients of the factors from multiple regression analyses of children's scores on Bayley and McCarthy scales of children's abilities

Bayley MSCA Factors (reference categories) Categories MDI PDI GCI Memory Motor

Mother's smoke (no)

Father's smoke (no)

Gender (boys)

Sodo-economic status (lower)

HOME scores (lower)

Mother's IQ (C80)

Mother's education (C3 years)

Mother's job (lower)

Birthweight (S2500 g)

Feeding style (breast)

Number of siblings (none)

Yes

Yes

Girls

Middle Higher Gighest

Middle Higher Highest

81-90 91-100 >lo0

>3 years

Middle Higher

>2500

Mixed Bottle

One 2Two

1.74 -0.15

-1.44 -2.07

6.43** 0.93

1.25 1.59 0.77 -1.26 4.83' -1.22

6.08' 6.73' 10.21" 8.31'' 12.96"' 8.30*+

5.73' 3.89 7.42'* 5.71' 9.26"' 4.12

-1.34 1.25

8.63 -0.72 0.83 1.07

8.22'- 5.95

-2.37 1.32 -3.39' 2.73

-0.27 -0.25 -3.63' 0.48

1.86 1.26 0.20

-1.06 -0.78 0.38

2.78' 0.96 2.59''

-2.16 -1.36 0.08 1.25 1.05 1.66 6.58'' 2.89 2.69'

3.90 1.23 -0.56 10.28" 2.42 2.65 12.80" 5.26'' 3.78'

8.83" 3.72' 6.85** 9.71'* 3.14 6.44*'

10.19" 3.79' 6.79**

0.60 0.26 0.18

7.92 2.68 0.09 1.38 -0.07 -2.75

3.69 2.08 5.68''

0.19 0.28 2.93 -1.78 -1.84' -0.23

2.73 1.01 1.45 0.26 0.72 0.96

~~ ~~~ ~ ~~

*, *' statistically significant at P<0.05 and PcO.01, respectively.

and 4). The higher the socio-economic status, HOME scores, and mother's IQ, the higher the child's neuropsychological abilities.

Effects of socio-environmental factors on association between maternal smoking and deuelopment of child's abilities The multiple regression analysis (Table 5) showed that, after taking the socio- economic status, HOME scores and maternal intelligence into account, the

Maternal smoking and neuropsychological development 411 estimated coefficients of postnatal exposure to maternal smoking in MDI, GCI and Motor subscales were changed from -2.74, -3.54 and -1.97 to -0.55, -0.45 and -0.68, respectively. The absolute values of estimated coefficients in other sub- scales were also reduced. None of the estimated coefficients was statistically significant (Pa 0.35) after the covariate adjustment.

In the final regression model (Table 6) , the variables which were identified as having independent effects on children’s abilities included gender, socio-econ- omic status, HOME scores, maternal intelligence, birthweight, feeding style and number of siblings in the household.

Discussion

These results reveal a statistically significant inverse association between ma- ternal smoking and neuropsychological development which becomes quite insig- nificant when other putative determinants of development are taken into account. Interpretation of these findings is therefore difficult, and must take into account both the reliability of the exposure measures, and the more general problems of selection bias, and ’over-adjustment’.

Information on maternal smoking was obtained by questionnaire, and there are two sources of uncertainty affecting the estimates of prenatal and postnatal exposure to mother’s smoke. These are, first, the factors other than maternal smoking which influence in utera and postnatal exposure to environmental tobacco smoke (ETS) and, secondly, the reliability of self-reported maternal smoking. In this study, the estimates of exposure to ERS are based solely on mother‘s reporting. Since the time-activity patterns of children, which indicate the time spent in the home and other environments containing ETS, vary widely, the extent of exposure to ETS is probably highly variable among children at a given point in time, and the variability is likely to increase with age.20 Little is known about the variation in exposure of the same individual at different points in time.

Findings from other studies support the use of maternal smoking as an indirect measure of exposure to ETS. A recent study of 433 healthy neonates showed that 75% of smoking mothers smoked near their offspring, and urine cotinine concentrations in the infants were correlated with household member smoking, particularly with mother’s smoking.2’ Other studies in similar settings have confirmed the accuracy of self-reported smoking data.6,z Validity of the information about maternal smoking used in this study has also been strength- ened by the consistency of our results with other previous investigations on maternal smoking and children’s abilities.

This study might also be criticised for an apparent loss of information in pooling all smokers into one group and contrasting them with non-smokers. This

412 P.A. Baghurst et al. was done as a consequence of statistical power considerations (there were only 29 women who smoked more than 20 cigarettes a day). The developmental index which showed the strongest crude 'dose-response' association with ETS was the Bayley MDI, which averaged 110.4 in non-smokers; 108.2 in light smokers (a20 cigarettedday); and 103.3 in heavy smokers (> 20 agarettedday). The addition of only the one covariate-socio-economic status-to the analysis reduced the esti- mated decrement in MDI among the children of heavy smokers to only 3.2 (s.e. =2.9). With maternal IQ and HOME score also in the model, the decrement became an increment of 0.94 (s.e.=2.9). It seems improbable that the conclusions would change with more refined exposure categories.

A methodological issue associated with studies of this nature is that of selec- tion bias. Since this is a prospective study of the relationship between maternal smoking and neuropsychological development within a motherkhild pair, selec- tion bias is unlikely to be important. Losses to follow-up may distort general summary results (such as the proportion of mothers who smoke), but it would be necessary to invoke a mechanism by which the relationship between maternal smoking and childhood development differed among those lost, in order to invalidate this study on the grounds of selection bias. In any event, the proportion of participants lost to follow-up was not large: 17% from birth to age 2, and 9% from age 2 to 4 (the characteristics of children remaining in the cohort and those lost to follow-up have been compared elsewherp).

Another issue is the problem of 'over-control'. For example, a relationship between birthweigh t and child's neurobehavioural status has been well estab-

and, in this study, birthweight was again identified as one of the variables which was able to explain a significant proportion of variation in neuropsycho- logical development. However, birthweight is itself influenced by maternal smok- ing, and therefore may act as an intervening factor in the association between prenatal exposure to smoking and childhood development-which raises the concern of whether it is legitimate to allow for birthweight in the analysis. The question becomes more complex when other variables for which less obvious mechanistic relationships are considered. A measure like socio-economic status will help disentangle problems such as the increased prevalence of smoking among subjects of lower status (not observed in this study), but may at the same time act as a surrogate measure of other less identifiable factors which may or may not intervene in the association between exposure to tobacco smoke and child- hood development. The problem has been discussed in many contexts, including this cohort study,= and the ultimate decision on which variables to include and exclude must depend on implied mechanisms rather than purely statistical arguments.z6

Table 5 demonstrates that socio-economic status, the home environment provided by the parents and maternal IQ all make significant inroads into the crude association of poor development with maternal smoking, and corroborates

Maternal smoking and neuropsychological development 413

the results of earlier ~ t u d i e s . ~ , ~ , ~ ’ While it may still be argued that smoking reduces a woman’s IQ and renders her less able to provide a highly stimulatory home environment for her children, there is little published evidence on which to base such speculation.

Although maternal consumption of alcohol and other drugs (such as mari- juana) during pregnancy has not been considered in this analysis, it is likely that their inclusion would have further eroded the inverse association of smoking with neuropsychological development, since smoking is typically more common among consumers of recreational drug^.^*,^^

An additional potential criticism of this work is that the measures of neuropsychological assessments used are more indicators of development than of long-term cognitive ability, and hence an even longer-term follow-up period may be required to assess the effects of ETS adequately.

The problem of confounding and over-control are avoided in animal experi- ments where the study subjects are more homogeneous, and can be randomly allocated to their treatment groups. A study of the effect of nicotine adminis- tration in utero upon activity in the rat showed that offspring whose mothers were injected twice daily throughout gestation with 3.0 mgkg nicotine (about 26 times the dose of moderate smoking in human beings) differed little from saline- injected control offspring on an activity measure.30 Nevertheless, several other animal experiments have implicated nicotine as a behavioural t e r a t ~ g e n . ~ ’ ~ ~ The problems of extrapolating from laboratory animal models to human beings have been widely debated, and the levels of exposure and the conditions that applied in many of these studies were quite different from those experienced in human populations. There is also the problem that environmental tobacco smoke is a complex mixture, and nicotine is only one of 2000 toxic substances so far identified.

In this study, children whose mothers smoked during pregnancy had small decreases in mean scores on the Bayley and McCarthy scales, compared with the scores of children whose mothers did not smoke, but these differences were not statistically significant (P>0.05) even ignoring all the confounders. However, since numbers of pregnant women who smoked more than 20 cigarettes per day were small, the study may have lacked the statistical power required to detect effects in this group. Moreover, in view of the widespread publicity that smoking during pregnancy may be detrimental to the health of the unborn infant, it is conceivable that some women would have been untruthful about their prenatal smoking, and real effects may have been hidden by the misclassification. No study has yet been able to distinguish confidently the effect of prenatal exposure to maternal smoking from that of postnatal exposure. This is due chiefly to the very strong association between smoking behaviour in the prenatal and postnatal period: most women who smoke during pregnancy also smoke after their child is born.

414 P.A. Baghurst et al.

Conclusion

We conclude that there is, at present, no strong evidence that maternal smoking exerts an independent effect upon neuropsychological development in early childhood. A major reason for the inconsistent results observed in this area may be confounding due to social and environmental factors. If passive smoking does have an effect on the development of children’s abilities, it is likely to be difficult to detect in the presence of wide variations in these factors. In order to gain a clearer understanding of this problem, more precise measures of exposure to environ- mental tobacco smoke, both in utero and postnatally, may also be required.

References 1 Hunt, J.V. Environmental risks in fetal and neonatal life as biological determinants of infant intelligence. In: Origins of Intelligence. Editor: M. Lewis. New York: Plenum Press, 1983. 2 Yerushalmy, J. The relationship of parents’ cigarette smoking to outcome of preg- nancy; implications as to the problem of inferring causation from observed associations. American Journal of Epidemiology 1971; 93A43-456. 3 Fogelman, K.R., Manor, 0. Smoking in pregnancy and development into early adulthood. British Medical Journal 1988; 29212S1236. 4 Dunn, H.G., McBumey, A.K., Hunter, C.M. Maternal cigarette smoking during pregnancy and the child’s subsequent development: 11. Neurological and intellectual maturation to the age of 6% years. Canadian Journal of Public Health 1977; 68:43-50. 5 Rantakallio, P. A follow-up study up to the age of 14 of children whose mothers smoked during pregnancy. Acta Paediatrica Scandinavica 1983; 72:747-753. 6 Sexton, M., Fox, N.L., Hebel, J.R. Prenatal exposure to tobacco: 11. Effects on cognitive functioning at age three. International Journal of Epidemiology 1990; 19:72-77. 7 Nichols, P.L., Chen, T.C. Minimal Brain Dysfunction: A Prospective Study. New Jersey: Lawrence Erlbaum Associates, 1981; pp. 23-273. 8 Naeye, R.L., Peters, E.C. Mental development of children whose mothers smoked during pregnancy. Obstetrics and Gynecology 1984; 64.601-607. 9 Hardy, J.B., Mellits, E.D. Does maternal smoking duringpregnancy have a long-term effect on the child? Lancet 1972; ik1332-1336. 10 Donovan, J.W. Effect on child of maternal smoking during pregnancy. Lancet 1973; ik376. 11 Bayley, N. Bayley Scules of Infant Development. New York The Psychological Corpor- ation, 1969. 12 McCarthy, D. Manual for the McCarthy Scales of Children’s Abilities. New York: The Psychological Corporation, 1972; pp. 1-12. 13 Kaufman, A.S., Kaufman, N. Clinical Evaluation of Young Children with the McCarthy Scales. New York: Grune & Stratton, 1977. 14 Daniel, A. The measurement of social class. Community Health Studies 1984; 8:21&222. 15 Wechsler, D. Wechsler Adult Intelligence Scale-Revised. New York: The Psychological Corporation, 1981. 16 Caldwell, B., Bradley, R. Home Observation for Measurement of the Environment. New York Dorsey, 1985.

Maternal smoking and neuropsychological development 415 17 McMichael, A.J., Baghurst, P.A., Wigg, N.R., et al. Port Pirie Cohort Study: environ- mental exposure to lead and children’s abilities at the age of four years. New EnglandJoumd of Medicine 1988; 319:468-475. 18 Wigg, N.R., Vimpani, G.V., McMichael, A.J. et al. Port Pirie Cohort Study: childhood blood lead and neuropsychological development at age two years. Journal of Epidemiology and Community Health 1988; 42:213-219. 19 Payne, R.W., Lane, P.W., Ainsley, A.E. et al. Genstat 5 Reference Manual. New York: Oxford University Press, 1988; pp. 303-360. 20 Wu-Williams, A.H., Samet, J.M. Environmental tobacco smoke: exposure-response relationship in epidemiologic studies. Risk Analysis 1990; 10:3948. 21 Greenberg, R.A., Bauman, K.E., Glover, L.H. ef al. Ecology of passive smoking by young infants. Journal of Pediatrics 1989; 114774-780. 22 Woodward, A., Douglas, R.M., Graham, N. et al. Acute respiratory illness in Adelaide children: breast feeding modifies the effect of passive smoking. Journal of Epidemiology and Community Health 1990; 44:224-230. 23 McMichael, A.J., Baghurst, P.A., Tong, S.L. et al. Maternal smoking and childhood respiratory illnesses: a seven-year cohort study. Znternational Journal of Environmental Health Research 1991; 1:192-203. 24 Mussen, P.H., Conger J.J., Kagan J. Child Development and Personality 3rd edn. New York: Harper & Row, 1969; pp. 64-95. 25 Baghurst, P.A., Robertson, E.R., McMichael, A.J. et al. The Port Pine Cohort Study: lead effects on pregnancy outcome and early childhood development. Neurotoxicology 1987; 8:395-402. 26 Kiely, J.L. Some conceptual problems in multivariable analyses of perinatal mortality. Paediatric and Perinatal Epidemiology 1991; 5:243-257. 27 Butler, N.R., Goldstein, H. Smoking in pregnancy and subsequent child develop- ment. British Medical Journal 1973; 4:573-575. 28 Martin, J., Martin, D.C., Lund, C.A. et al. Maternal alcohol ingestion and cigarette smoking and their effects on newborn conditioning. Alcoholism: Clinical and Experimental Research 1977; 1:24%247. 29 Fried, P.A., Watkinson, B. 36- and 48- month neurobehavioral follow-up of children prenatally exposed to marijuana, cigarettes, *and alcohol. Iournal of Developmental and Behavioural Pediatrics 1990; 11:49-58. 30 Martin, J.C., Becker, R.F. The effects of nicotine administration in utero upon activity in the rat. Psychonomic Science 1970; 1959-60. 31 Martin, J.C., Becker, R.F. The effects of maternal nicotine absorption or hypoxic episodes upon appetitive behavior of rat offspring. DeueIopmental Psychobiology 1971; 4 1 S 1 4 7 . 32 Baer, D.S., McCleam, G.E., Wilson, J.R. Fertility, maternal care, and offspring be- havior in mice prenatally treated with tobacco smoke. Developmental Psychobiology 1980; 13:643-652. 33 Genedani, S., Bernardi, M., Bertolini, A. Sex-linked differences in avoidance learning in the offspring of rats treated with nicotine during pregnancy. Psychopharmacology 1983; 80:93-95.