Background
Deficits in cognitive function are associated with a variety of adverse health and social outcomes [
1]. While many determinants of cognitive function are not remediable through public health interventions, prenatal exposure to maternal cigarette smoking (PEMCS) is common and potentially preventable. Public health campaigns target PEMCS in order to prevent adverse outcomes in the offspring, including low birth weight and preterm birth [
2]. PEMCS has been associated with medical problems [
3] and with psychiatric and behavior problems in adolescence [
4‐
6] and in adulthood [
7‐
9].
In addition to these problems, a significant body of literature suggests that PEMCS may be harmful to cognitive functioning and development [
10‐
17]. However, not all research supports this perspective [
18‐
22] and a literature review of studies from 1972 to 1992 suggested that study limitations, such as poor control for covariates, prevented firm conclusions from being made [
23]. More recent studies have addressed some of these limitations and in their systematic review of studies from 2000 to 2011, Clifford, Lang and Chen suggested more recent evidence supports a relationship, though one that is not straightforward [
24], with the association most consistent for measures of academic achievement and intellectual ability. For example, in the Northern Finland 1966 Birth Cohort, PEMCS was strongly associated with poorer educational attainment in adulthood [
25].
Considering a more nuanced association bewteen PEMCS and cognition, it has been suggested that PEMCS is more specifically associated with “hot” cognitive tasks that involve stress or frustration [
26,
27]. This would be consistent with findings linking this exposure to childhood externalising behavior [
28,
29]. Huijbregts et al. found that children with PEMCS showed poorer inhibitory control in a more frustrating task (but not in a regular task), while also showing higher conduct problems and hyperactivity-inattention scores [
27]. Difficulties with self-regulation appear to be more pronounced in boys than in girls [
30].
Despite the number of studies examining PEMCS and cognition, to our knowledge only two have investigated the association in adulthood [
12,
15]. Among 18-year old male Swedish conscripts, Lundberg et al. found that intellectual impairment was associated with PEMCS, though to a reduced degree after controlling for parental factors [
15]. A study by Mortensen et al. found that smoking during pregnancy (measured during the third trimester) was associated with lower IQ score in 18-year old males in a dose-response manner, while controlling for parental social status and education and other factors [
12].
Clifford et al. have suggested that stronger findings in male-only samples may indicate that any association between PEMCS and cognition in adulthood could be sex-specific with vulnerability in males possibly masked in other analyses by inclusion of a high number of females [
24]. Two among the more recent studies showed a significant association between PEMCS and intelligence in males only [
12,
14]. There are biological and developmental reasons to suspect sex-specific effects on cognition. Rodent studies have noted differences in neurodevelopment in response to prenatal nicotine according to sex [
31,
32]. Studies examining PEMCS in humans have also noted more marked imaging differences in females [
33,
34] and behavioral findings that differed between males and females [
34‐
37]. This has included differences in gene by exposure interaction in one study [
38]. In addition to these imaging and laboratory findings, males are more at risk for inattention and hyperactivity and other externalizing behaviors that are associated with poorer cognitive outcomes [
39].
The relationship between PEMCS and cognition is complex and the potential role of confounders, mediators and moderators of the relationship needs to be considered. Heritable traits such as parental cognitive ability and parental psychiatric vulnerability may particularly confound associations between PEMCS and cognition [
24], while maternal use of alcohol in pregnancy may co-occur with PEMCS [
11,
19]. Other factors, for example birth weight and family functioning [
21], have been suggested as potential mediators or moderators of observed effects.
To date, there has been limited exploration of how adolescent mental health influences the relationship between PEMCS and adult cognition. A range of adolescent psychiatric and behavioral problems has been associated with both PEMCS and with poorer cognitive outcomes, though not in all studies. PEMCS has been associated with adolescent psychotic-like experiences (PLEs) [
4] and with schizophrenia [
40], though not in all studies [
41]. Similarly, PEMCS has been associated with inattention and hyperactivity and other externalizing behaviors [
5,
6], though not in all studies [
42]. These psychiatric and behavioral problems are also associated with problems in cognition. Young people with PLEs have shown deficits across multiple domains [
43]. In addition to poorer performance on tasks requiring attention [
44], adults with attention-deficit hyperactivity disorder (ADHD) perform more poorly on tests of general intellectual ability [
45]. These findings raise the possibility that PEMCS may affect later cognition through an effect on adolescent emotional and behavioral problems. These problems, including the presence and degree of PLEs, adolescent inattention and hyperactivity and other externalizing behaviors in adolescence, could potentially affect the longitudinal course of cognitive development into early adulthood.
Aims and objectives
Using a cohort sub-group with groups selected based on PEMCS and matched by maternal education and geographic region, this study aims to clarify the long-term associations between PEMCS and cognition and whether any of these associations are sex-specific. The primary hypothesis is that PEMCS is associated with adverse cognitive outcomes in adulthood but that males are more vulnerable to these adverse outcomes. Where there was evidence in support of the primary hypothesis, we tested three secondary hypotheses: (1) observed associations are not explained by maternal alcohol use during pregnancy, birth weight or current smoking status; (2) observed associations are partly mediated by adolescent mental health status, specifically PLEs, inattention and hyperactivity, and other externalizing behaviors; (3) PEMCS interacts with adolescent mental health status in predicting these cognitive associations.
Discussion
In this large general population-based and well-controlled sample, PEMCS was not associated with poorer cognitive scores across most cognitive tests performed in males and all cognitive tests performed in females. In adult males, PEMCS was associated with a small effect in terms of poorer performance on vocabulary and matrix reasoning. These associations were partly mediated by adolescent inattention and hyperactivity and showed differences depending on the presence or absence of PLEs in adolescence.
Our finding that there were no significant associations between PEMCS and cognition in females and only limited and small associations in males is consistent with the observation of nuanced associations described elsewhere [
24]. Earlier studies have not controlled sufficiently for confounding factors, including socioeconomic status and use of other substances in pregnancy. This study was matched on maternal education and controlled for alcohol use during pregnancy, factors that may have influenced positive findings in other studies. Our finding of no associations in females across all tests, including both traditional pencil and paper tests and computer -based tests, adds to the consistency of the negative findings in females. Furthermore, the coefficients and confidence intervals in these tests were close to zero, supporting the absence of associations in these cases. This is consistent with suggestions that although PEMCS may exert structural and functional effects [
34], this may not result in any measurable cognitive effects [
19]. We know that PEMCS results in prenatal exposure to a variety of toxic compounds, of which nicotine is but one, along with reduced uterine blood flow with associated episodes of foetal hypoxia and malnutrition [
61]. However, significant brain development follows birth with associated multiple opportunities for brain adaptation to early insult, particularly if these insults are mild. This may explain the general lack of associations between PEMCS and cognition in our sample.
Our results regarding the negative effects of PEMCS on aspects of male cognition replicate those of Mortensen et al., who noted that heavy smoking reduced intelligence scores in young adult males by 0.41 standard deviations [
12]. Indeed, we have found that PEMCS was associated with poorer performance on aspects of intelligence in adult males: by 0.44 standard deviations in the case of vocabulary scores and by 0.38 standard deviations in the case of matrix reasoning scores. While it is more difficult to compare results to those found in children, Braun et al. also noted a differential effect on intelligence scores in 7-year old males [
14]. Our findings in relation to males are further supported by evidence for a dose response trend in the case of matrix reasoning and a trend towards this in the case of vocabulary.
Our findings must be viewed in the context of a growing body of evidence suggesting differences in male and female responses to PEMCS. Though many imaging studies have not noted sex differences in response to PEMCS [
62], studies of particular regions have noted more pronounced PEMCS-related differences in female than male offspring with reduced corpus callosum size [
33] and cortical thinning in several regions [
34]. The latter finding was also associated with differences in self-rated assessment of caring in females but not males. At a physiological level, rodent studies have found that prenatal exposure to nicotine was associated with reduced number of serotonin transporter binding sites in the cerebral cortex that is more marked in females than males [
32]. On the other hand, males but not females exposed to prenatal nicotine showed increased number of 5HT1A receptors in their cerebral cortex in another study [
31]. Studies of behavioral outcomes have also noted sex differences, with prenatal exposure to nicotine reducing locomotor activity in male but not female rodents [
63]. In human adolescents, PEMCS has been associated with more pronounced conduct problems in males than females [
35,
36]. Indeed, genetic studies have noted a sex-specific pattern of gene-by-exposure interaction in predicting conduct disorder symptoms [
38]. Our findings therefore add to existing evidence of differences in response to PEMCS by sex. Stronger associations in males between PEMCS and language and intelligence skills (adulthood) and conduct problems (adolescence) may be one manifestation of a more fundamental neural process that has yet to be identified. One potential mechanism is reward sensitivity and emotion regulation, which have been shown to be more vulnerable to the effects of PEMCS in boys than in girls [
30].
The partial mediation of our findings by inattention and hyperactivity has a number of possible explanations. The deficits in reward sensitivity and emotion regulation discussed above [
30] may play a role. On a related vein, poorer learning could arise from difficulties in performance of “hot” executive function tasks during childhood, difficulties that have been associated with PEMCS and externalizing behaviors [
64]. A further possibility is that inattention and hyperactivity is a proxy for another factor in its mediation. A recent sibling study found that PEMCS was not a strong causal factor for ADHD in childhood and adolescence within families, suggesting previous findings may be due to uncontrolled shared genetics or family environment [
42]. The role of family environment as a potential mediating factor in cognitive outcomes has also been suggested [
21]. Certainly difficulties with inattention and hyperactivity could affect educational achievement, reflected in testing for vocabulary and matrix reasoning. It remains to be seen if addressing childhood inattention and hyperactivity could prevent longer-term cognitive effects of PEMCS.
Previous studies have highlighted the value of identifying interactions in understanding how PEMCS may influence cognitive performance [
21] in order to identify who might benefit from intervention. We have identified interactions between PEMCS and both sex and adolescent PLEs that may prove useful at a public health level. The finding in relation to PLEs is somewhat counter-intuitive. The lack of an association between PEMCS and cognition in those with PLEs may reflect the exclusion of those later treated for serious mental illness from this study [
47]. Those with PLEs excluded from this study were likely to have poorer long-term outcomes and significant further risk factors that may affect cognition. The remaining individuals with PLEs included in this study may therefore be a particularly resilient group. The findings may also be due to chance, though the Chi-square test for interaction is a particularly conservative test.
Our study has a number of features that add weight to the findings above. Firstly, the sample was matched to reduce the possible role of parental socio-economic status or maternal education. Secondly, the sample has been followed prospectively from birth and we have been able to consider mental health status in adolescence, almost 10 years prior to cognitive testing. This allowed for the consideration of adolescent mental health in its temporal context, unaffected by recall bias. Thirdly, the sample includes both males and females, allowing for separate analysis of these groups to confirm or refute previous theories on susceptibility to neurodevelopmental insult.
Limitations of this study include that the size of our sample provided insufficient power to detect small effects on cognitive domains in males and females. However, our findings are consistent with previous studies. Study of smaller effects requires study in a sample of 550 per group [
57]. While the exclusion criteria in this study address many potential confounding factors, they also reduce the generalizability of the findings. Exclusion criteria meant that the sample were relatively healthy in terms of mental health and cognitive performance. For example, we cannot comment on the role of PEMCS on intellectual disability as individuals with an IQ < 70 were excluded and we cannot fully consider the role of mental health given the exclusion of those with more serious mental disorders. This is related to a further limitation, which is that over half of those invited did not participate. Those not completing may differ from those who completed. For example, we know that more without PEMCS than with PEMCS participated. This may contribute to a healthy participant bias, with findings not reflecting the experience of those with worse outcomes and results being potentially more conservative than would be the case at population level. A further limitation is that our measure of PEMCS is based on self-report of use rather than objective measures. This could result in inclusion of exposed participants who denied use in the non-exposed group with further potential to dilute any observed findings. Finally, though primary analyses were hypothesis driven, the number of tests performed in post-hoc analyses examining mediation and interaction mean that chance is a possible explanation for some of the findings. Indeed, most primary findings were negative and positive findings must be regarded as tentative and in need of replication in larger samples and different settings.
The findings of this study have several public health and research implications. Firstly, they add further weight to public health efforts to eliminate smoking in pregnancy as they suggest long-term adverse consequences of this behavior. From a research perspective, the results of this study add to findings suggesting the need for more nuanced analysis of the long-term effects of smoking, particularly in the area of cognition. Further analysis of potential mediating factors during development that may affect adult cognition, such as reward sensitivity and emotional regulation [
30] or difficulties with “hot” cognition tasks [
27], could provide targets to improve adult outcomes. Finally, these results highlight the need for further study on factors affecting cognition in those at risk for psychosis, including replication of these methods in a sample including individuals who progressed to develop significant mental illness in order to identify preventable causes of poorer functional outcomes in this population.
Acknowledgements
Not applicable.