DOI: 10.1111/1471-0528.13078
Commentary
www.bjog.org
Prenatal exposure to alcohol and the developing fetus: methodological issues J Niclasen Department of Psychology, University of Copenhagen, Copenhagen K, Denmark Correspondence: Dr J Niclasen, Department of Psychology, University of Copenhagen, Øster Farimagsgade 2A, 1353 Copenhagen K, Denmark. Email [email protected] Accepted 2 August 2014. Published Online 11 September 2014. Please cite this paper as: Niclasen J. Prenatal exposure to alcohol and the developing fetus: methodological issues. BJOG 2015;122:770–772.
Forty years ago, The Lancet published Jones and Smith’s now classic article on ‘Recognition of the fetal alcohol syndrome in early infancy’.1 Back then the focus was on investigating prenatal exposures to excessive amounts of alcohol and child neurodevelopment. Since that time, the focus of research has shifted towards investigations evaluating exposures to much lower doses of alcohol, and the ‘hot-topic’ today is to identify whether there is a safe level below which drinking is not associated with any harm to the developing fetus. Most studies investigating this make use of a group comparison design, i.e. child neurodevelopmental outcomes of women with different self-reported average intakes are compared. The findings from such observational studies have been somewhat contradictory. Some studies have indeed reported negative associations between prenatal exposure to low doses of alcohol and neurodevelopmental outcomes in childhood,2,3 whereas other studies have reported no such associations.2,4 Many studies have even reported a J-shaped association, such that exposure to low doses of alcohol has an apparently protective effect on the developing fetus.2,5 These inconsistencies observed in the literature may be a consequence of a number of methodological limitations, including extraneous factors in the statistical analyses, shortcomings regarding the alcohol exposure categories, and problems regarding the use of the outcome measures. The state-of-the-art articles in the epidemiological field today report findings from large-scale cohort studies where multiple regression is the standard procedure for analysing data. In such analyses a number of factors are typically controlled for statistically, e.g. maternal age, education and smoking. However, if studies do not sufficiently control for confounding factors then unmeasured and residual confounding may bias the results. One study reported substantial differences on socio-demographic, psychological and
770
lifestyle factors, i.e. potential confounding factors, between pregnancy abstainers and intakers.6 This study reported very large differences on a long list of lifestyle factors including education, psychiatric diagnoses, smoking, body mass index and exercising. The authors concluded that residual confounding factors may mask potential differences between exposure groups in observational studies. Thus, these results suggest that the contradictory findings from observational studies may reflect residual and unmeasured confounding factors. Whereas there is a long tradition within epidemiology of controlling statistically for prenatal factors, postnatal, psychosocial, mediating factors have often been little considered.7 Postnatal factors including attachment, personality and IQ on part of the child, are strong predictors of psychopathology.7 For example, attachment, i.e. the quality of the mother–child relationship, has a very substantial impact on the development of cognitive and mental health outcomes.8 Children engaging in what is known as a ‘secure’ attachment style (as opposed to an insecure attachment style) are those with a history of sensitive and responsive maternal care. These children later in life are better at emotional regulation, have a higher self-esteem, and have better developed coping skills. This, in turn, makes them more able to handle stressful or challenging situations and conversely lowers their risk for poorer mental health outcomes later in life.9 Unfortunately, ‘attachment style’ is rarely controlled for in the large-scale epidemiological studies. Going back to the observational studies on prenatal exposure to low doses of alcohol, a ‘small’ biological effect is expected. However, such a ‘small’ effect may easily be masked by the ‘large’ effect of, for example, attachment. The importance of psychosocial factors is demonstrated in a study by Srikartika and O’Leary in which they examine fetal outcomes of non-aboriginal and aboriginal mothers
ª 2014 Royal College of Obstetricians and Gynaecologists
Prenatal exposure to alcohol and child development
with an alcohol-related diagnosis in Western Australia.10 Aboriginal mothers are younger, less likely to be married, more likely to live in remote areas, are socio-economically less advanced and are more likely to smoke. By means of a frequency-matched case–control design these authors move on to present population-attributable fractions for low birthweight, preterm birth and APGAR scores. They show marked differences in the fetal outcomes between aboriginal and non-aboriginal women with alcohol-related diagnoses. The study illustrates one alternative design that to a large extent overcomes the problem of confounding and mediating factors and shows the importance of rethinking the group-comparison design. Apart from these issues regarding residual confounding and mediating factors, shortcomings regarding the alcohol exposure categories also should be considered. ‘Dose’ (how much is actually consumed), ‘pattern’ (e.g. binge drinking or consumption spread over several days), and ‘timing’ [at what time(s) during the pregnancy the exposure took place] are all important.2 However, previous studies have generally failed to incorporate all of these three factors into the exposure categories. Most often no more than two factors are considered simultaneously. Furthermore, most cohort studies use self-reported information on alcohol intake, despite the fact that this may be quite inaccurate.11 These issues raise important questions regarding the validity of our measures of exposure and in turn the extent of misclassification. In the study by Srikartika and O’Leary the scope was fetal outcomes among women with alcohol-related diagnoses.10 In this case the validity of the exposures, i.e. the diagnoses, is obviously quite solid. The use of register-based diagnoses is therefore one potential way to overcome the problem of misclassification. This in turn will lower the extent of misclassification. However, when the aim is to investigate very low doses of alcohol the use of alcoholrelated diagnoses is obviously not a possibility. Instead, observational studies focusing on low doses of alcohol should ideally be based on objective measurements. One possibility is to subdivide the women on the basis of meconium alcohol biomarkers rather than on the basis of self-reported intakes. The third limitation relates to the children’s outcome measures employed in large-scale cohort studies. Brief screening instruments are often employed. On the positive side such questionnaires often show excellent psychometric properties, and may suit the purpose of some risk factor studies. However, when the aim is to investigate potential effects of very low doses of alcohol, screening tools may not be sensitive enough to detect the potential damage from prenatal exposure to alcohol. Instead, more sensitive neuropsychological assessments are probably needed.12 Furthermore, the ages for the follow-up evaluation of children
ª 2014 Royal College of Obstetricians and Gynaecologists
are not chosen for the specific purpose of individual risk factor studies. This may be problematic because, with respect to neurodevelopment, the ages for the follow-up evaluations are vital. Previous studies with fetal alcohol syndrome (FAS) children have, for example, reported that behavioural effects may not be present until late childhood. These children had normal behavioural scores in early childhood but above clinical cut-offs scores in middle childhood.12 Therefore it is very likely that previous ‘low-dose’ studies have been conducted with children at too early ages, when potential effects are not yet evident. On this basis, future studies should not only make use of self-reported questionnaires, but should combine these with register-based information on, for example, school performance, psychiatric diagnoses and prescribed psychotropic medication. In summary, I believe that we will never resolve whether there is a safe level below which drinking is not associated with any harm to the developing foetus. With regard to residual confounding and mediating factors, one possibility would be to apply a matched case–control design as was done by Srikartika and O’Leary.10 However, no one design will provide an adequate solution on its own. Instead, a whole range of natural experiments and research designs are needed. We must recognise that each study design have its own strengths, limitations and biases.13 Such research designs include studies that compare data from different cultural settings in which drinking in pregnancy is observed within different socio-economic groups. Another way could be to make use of a sibling design; i.e. compare children of mothers who changed their drinking habits between pregnancies. In addition, neurodevelopmental outcomes in children whose fathers drank during the mothers’ pregnancy, but whose mothers themselves did not drink, should be compared with children whose fathers did not drink during the mothers’ pregnancy but whose mothers did. If we assume that mothers’ drinking acts as a biological teratogen, then similar effects in the two groups would imply that maternal drinking is not causative.13 These different research designs are all better than the standard group comparison design most often applied today in which children of women with different average intakes are simply compared. Future studies should evaluate in more depth sociodemographic and lifestyle factors of the applied samples. The descriptive study by Niclasen6 cited above investigating such factors in depth, reported substantial differences between intake groups. Therefore, more such studies are needed to get an understanding of the extent to which the women differ across intake groups. Because questions regarding who drinks, why and how much, are a culturally dependent issue, it needs to be investigated within different cultural populations. Even then, we may never really know
771
Niclasen
if there is a safe level of alcohol consumption during pregnancy.
Disclosure of interests None.
Details of ethics approval Not applicable.
Funding This work was supported by the Department of Psychology, University of Copenhagen; Ludvig og Sara Elsass Foundation; Aase og Ejnar Danielsens Foundation; Carl J. Becker’s Foundation; the Lundbeck Foundation; Børneog Ungdomspsykiatrisk Selskab i Danmark; Dagmar Marshalls Foundation; The A.P. Møller Foundation for the Advancement of Medical Science; Direktør Jakob Madsens Legat.
Acknowledgements I would like to thank Associate Professor Tom Teasdale for reading through and commenting on the manuscript prior to submission. &
References
3 Sayal K, Heron J, Golding J, Emond A. Prenatal alcohol exposure and gender differences in childhood mental health problems: a longitudinal population-based study. Pediatrics 2007;119:426–34. 4 Kelly YJ, Sacker A, Gray R, Kelly J, Wolke D, Head J, et al. Light drinking during pregnancy: still no increased risk for socioemotional difficulties or cognitive deficits at 5 years of age? J Epidemiol Community Health 2012;66:41–8. 5 Niclasen J, Andersen AM, Teasdale TW, Strandberg-Larsen K. Prenatal exposure to alcohol, and gender differences on child mental health at age seven years. J Epidemiol Commun Health 2014;68:224–32. 6 Niclasen J. Drinking or not drinking in pregnancy: the multiplicity of confounding influences. Alcohol Alcohol 2014;49:349–55. 7 Baron RM, Kenny DA. The moderator-mediator variable distinction in social psychological research: conceptual, strategic, and statistical considerations. J Pers Soc Psychol 1986;51:1173–82. 8 Bowlby J. Research into the Origins of Delinquent Behaviour. BMJ 1950;1:570–3. 9 Sroufe LA. Attachment and development: a prospective, longitudinal study from birth to adulthood. Attach Hum Dev 2005;7:349–67. 10 Srikartika V, O’Leary C. Adverse fetal effects in the offspring of mothers with an alcohol-related diagnosis: a Western Australian population-based cohort study 1983–2007. BJOG 2015;122:795–804. 11 Abel EL. Fetal alcohol abuse syndrome. New York: Plenum, 1998. 12 Astley S, Grant T. Another perspective on ‘the effect of different alcohol drinking patterns in early to mid pregnancy on the child’s intelligence, attention, and executive function’. BJOG 2012; 119:1672–5. 13 Gray R, Mukherjee RA, Rutter M. Alcohol consumption during pregnancy and its effects on neurodevelopment: what is known and what remains uncertain. Addiction 2009;104:1270–3.
1 Jones KL, Smith DW. Recognition of the fetal alcohol syndrome in early infancy. Lancet 1973;302:999–1001. 2 Gray R, Henderson J. Review of the Fetal Effects of Prenatal Alcohol Exposure. Oxford: National Perinatal Epidemiology Unit, University of Oxford, 2006.
772
ª 2014 Royal College of Obstetricians and Gynaecologists
Copyright of BJOG: An International Journal of Obstetrics & Gynaecology is the property of Wiley-Blackwell and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.
Commentary
www.bjog.org
Prenatal exposure to alcohol and the developing fetus: methodological issues J Niclasen Department of Psychology, University of Copenhagen, Copenhagen K, Denmark Correspondence: Dr J Niclasen, Department of Psychology, University of Copenhagen, Øster Farimagsgade 2A, 1353 Copenhagen K, Denmark. Email [email protected] Accepted 2 August 2014. Published Online 11 September 2014. Please cite this paper as: Niclasen J. Prenatal exposure to alcohol and the developing fetus: methodological issues. BJOG 2015;122:770–772.
Forty years ago, The Lancet published Jones and Smith’s now classic article on ‘Recognition of the fetal alcohol syndrome in early infancy’.1 Back then the focus was on investigating prenatal exposures to excessive amounts of alcohol and child neurodevelopment. Since that time, the focus of research has shifted towards investigations evaluating exposures to much lower doses of alcohol, and the ‘hot-topic’ today is to identify whether there is a safe level below which drinking is not associated with any harm to the developing fetus. Most studies investigating this make use of a group comparison design, i.e. child neurodevelopmental outcomes of women with different self-reported average intakes are compared. The findings from such observational studies have been somewhat contradictory. Some studies have indeed reported negative associations between prenatal exposure to low doses of alcohol and neurodevelopmental outcomes in childhood,2,3 whereas other studies have reported no such associations.2,4 Many studies have even reported a J-shaped association, such that exposure to low doses of alcohol has an apparently protective effect on the developing fetus.2,5 These inconsistencies observed in the literature may be a consequence of a number of methodological limitations, including extraneous factors in the statistical analyses, shortcomings regarding the alcohol exposure categories, and problems regarding the use of the outcome measures. The state-of-the-art articles in the epidemiological field today report findings from large-scale cohort studies where multiple regression is the standard procedure for analysing data. In such analyses a number of factors are typically controlled for statistically, e.g. maternal age, education and smoking. However, if studies do not sufficiently control for confounding factors then unmeasured and residual confounding may bias the results. One study reported substantial differences on socio-demographic, psychological and
770
lifestyle factors, i.e. potential confounding factors, between pregnancy abstainers and intakers.6 This study reported very large differences on a long list of lifestyle factors including education, psychiatric diagnoses, smoking, body mass index and exercising. The authors concluded that residual confounding factors may mask potential differences between exposure groups in observational studies. Thus, these results suggest that the contradictory findings from observational studies may reflect residual and unmeasured confounding factors. Whereas there is a long tradition within epidemiology of controlling statistically for prenatal factors, postnatal, psychosocial, mediating factors have often been little considered.7 Postnatal factors including attachment, personality and IQ on part of the child, are strong predictors of psychopathology.7 For example, attachment, i.e. the quality of the mother–child relationship, has a very substantial impact on the development of cognitive and mental health outcomes.8 Children engaging in what is known as a ‘secure’ attachment style (as opposed to an insecure attachment style) are those with a history of sensitive and responsive maternal care. These children later in life are better at emotional regulation, have a higher self-esteem, and have better developed coping skills. This, in turn, makes them more able to handle stressful or challenging situations and conversely lowers their risk for poorer mental health outcomes later in life.9 Unfortunately, ‘attachment style’ is rarely controlled for in the large-scale epidemiological studies. Going back to the observational studies on prenatal exposure to low doses of alcohol, a ‘small’ biological effect is expected. However, such a ‘small’ effect may easily be masked by the ‘large’ effect of, for example, attachment. The importance of psychosocial factors is demonstrated in a study by Srikartika and O’Leary in which they examine fetal outcomes of non-aboriginal and aboriginal mothers
ª 2014 Royal College of Obstetricians and Gynaecologists
Prenatal exposure to alcohol and child development
with an alcohol-related diagnosis in Western Australia.10 Aboriginal mothers are younger, less likely to be married, more likely to live in remote areas, are socio-economically less advanced and are more likely to smoke. By means of a frequency-matched case–control design these authors move on to present population-attributable fractions for low birthweight, preterm birth and APGAR scores. They show marked differences in the fetal outcomes between aboriginal and non-aboriginal women with alcohol-related diagnoses. The study illustrates one alternative design that to a large extent overcomes the problem of confounding and mediating factors and shows the importance of rethinking the group-comparison design. Apart from these issues regarding residual confounding and mediating factors, shortcomings regarding the alcohol exposure categories also should be considered. ‘Dose’ (how much is actually consumed), ‘pattern’ (e.g. binge drinking or consumption spread over several days), and ‘timing’ [at what time(s) during the pregnancy the exposure took place] are all important.2 However, previous studies have generally failed to incorporate all of these three factors into the exposure categories. Most often no more than two factors are considered simultaneously. Furthermore, most cohort studies use self-reported information on alcohol intake, despite the fact that this may be quite inaccurate.11 These issues raise important questions regarding the validity of our measures of exposure and in turn the extent of misclassification. In the study by Srikartika and O’Leary the scope was fetal outcomes among women with alcohol-related diagnoses.10 In this case the validity of the exposures, i.e. the diagnoses, is obviously quite solid. The use of register-based diagnoses is therefore one potential way to overcome the problem of misclassification. This in turn will lower the extent of misclassification. However, when the aim is to investigate very low doses of alcohol the use of alcoholrelated diagnoses is obviously not a possibility. Instead, observational studies focusing on low doses of alcohol should ideally be based on objective measurements. One possibility is to subdivide the women on the basis of meconium alcohol biomarkers rather than on the basis of self-reported intakes. The third limitation relates to the children’s outcome measures employed in large-scale cohort studies. Brief screening instruments are often employed. On the positive side such questionnaires often show excellent psychometric properties, and may suit the purpose of some risk factor studies. However, when the aim is to investigate potential effects of very low doses of alcohol, screening tools may not be sensitive enough to detect the potential damage from prenatal exposure to alcohol. Instead, more sensitive neuropsychological assessments are probably needed.12 Furthermore, the ages for the follow-up evaluation of children
ª 2014 Royal College of Obstetricians and Gynaecologists
are not chosen for the specific purpose of individual risk factor studies. This may be problematic because, with respect to neurodevelopment, the ages for the follow-up evaluations are vital. Previous studies with fetal alcohol syndrome (FAS) children have, for example, reported that behavioural effects may not be present until late childhood. These children had normal behavioural scores in early childhood but above clinical cut-offs scores in middle childhood.12 Therefore it is very likely that previous ‘low-dose’ studies have been conducted with children at too early ages, when potential effects are not yet evident. On this basis, future studies should not only make use of self-reported questionnaires, but should combine these with register-based information on, for example, school performance, psychiatric diagnoses and prescribed psychotropic medication. In summary, I believe that we will never resolve whether there is a safe level below which drinking is not associated with any harm to the developing foetus. With regard to residual confounding and mediating factors, one possibility would be to apply a matched case–control design as was done by Srikartika and O’Leary.10 However, no one design will provide an adequate solution on its own. Instead, a whole range of natural experiments and research designs are needed. We must recognise that each study design have its own strengths, limitations and biases.13 Such research designs include studies that compare data from different cultural settings in which drinking in pregnancy is observed within different socio-economic groups. Another way could be to make use of a sibling design; i.e. compare children of mothers who changed their drinking habits between pregnancies. In addition, neurodevelopmental outcomes in children whose fathers drank during the mothers’ pregnancy, but whose mothers themselves did not drink, should be compared with children whose fathers did not drink during the mothers’ pregnancy but whose mothers did. If we assume that mothers’ drinking acts as a biological teratogen, then similar effects in the two groups would imply that maternal drinking is not causative.13 These different research designs are all better than the standard group comparison design most often applied today in which children of women with different average intakes are simply compared. Future studies should evaluate in more depth sociodemographic and lifestyle factors of the applied samples. The descriptive study by Niclasen6 cited above investigating such factors in depth, reported substantial differences between intake groups. Therefore, more such studies are needed to get an understanding of the extent to which the women differ across intake groups. Because questions regarding who drinks, why and how much, are a culturally dependent issue, it needs to be investigated within different cultural populations. Even then, we may never really know
771
Niclasen
if there is a safe level of alcohol consumption during pregnancy.
Disclosure of interests None.
Details of ethics approval Not applicable.
Funding This work was supported by the Department of Psychology, University of Copenhagen; Ludvig og Sara Elsass Foundation; Aase og Ejnar Danielsens Foundation; Carl J. Becker’s Foundation; the Lundbeck Foundation; Børneog Ungdomspsykiatrisk Selskab i Danmark; Dagmar Marshalls Foundation; The A.P. Møller Foundation for the Advancement of Medical Science; Direktør Jakob Madsens Legat.
Acknowledgements I would like to thank Associate Professor Tom Teasdale for reading through and commenting on the manuscript prior to submission. &
References
3 Sayal K, Heron J, Golding J, Emond A. Prenatal alcohol exposure and gender differences in childhood mental health problems: a longitudinal population-based study. Pediatrics 2007;119:426–34. 4 Kelly YJ, Sacker A, Gray R, Kelly J, Wolke D, Head J, et al. Light drinking during pregnancy: still no increased risk for socioemotional difficulties or cognitive deficits at 5 years of age? J Epidemiol Community Health 2012;66:41–8. 5 Niclasen J, Andersen AM, Teasdale TW, Strandberg-Larsen K. Prenatal exposure to alcohol, and gender differences on child mental health at age seven years. J Epidemiol Commun Health 2014;68:224–32. 6 Niclasen J. Drinking or not drinking in pregnancy: the multiplicity of confounding influences. Alcohol Alcohol 2014;49:349–55. 7 Baron RM, Kenny DA. The moderator-mediator variable distinction in social psychological research: conceptual, strategic, and statistical considerations. J Pers Soc Psychol 1986;51:1173–82. 8 Bowlby J. Research into the Origins of Delinquent Behaviour. BMJ 1950;1:570–3. 9 Sroufe LA. Attachment and development: a prospective, longitudinal study from birth to adulthood. Attach Hum Dev 2005;7:349–67. 10 Srikartika V, O’Leary C. Adverse fetal effects in the offspring of mothers with an alcohol-related diagnosis: a Western Australian population-based cohort study 1983–2007. BJOG 2015;122:795–804. 11 Abel EL. Fetal alcohol abuse syndrome. New York: Plenum, 1998. 12 Astley S, Grant T. Another perspective on ‘the effect of different alcohol drinking patterns in early to mid pregnancy on the child’s intelligence, attention, and executive function’. BJOG 2012; 119:1672–5. 13 Gray R, Mukherjee RA, Rutter M. Alcohol consumption during pregnancy and its effects on neurodevelopment: what is known and what remains uncertain. Addiction 2009;104:1270–3.
1 Jones KL, Smith DW. Recognition of the fetal alcohol syndrome in early infancy. Lancet 1973;302:999–1001. 2 Gray R, Henderson J. Review of the Fetal Effects of Prenatal Alcohol Exposure. Oxford: National Perinatal Epidemiology Unit, University of Oxford, 2006.
772
ª 2014 Royal College of Obstetricians and Gynaecologists
Copyright of BJOG: An International Journal of Obstetrics & Gynaecology is the property of Wiley-Blackwell and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.
