Is bigger better? Macrosomia and psychopathology later in life

Authors

  • R. J. Van Lieshout,

    1. Department of Psychiatry and Behavioural Neurosciences and the Offord Centre for Child Studies, McMaster University, Hamilton, Ontario, Canada.
    Search for more papers by this author
  • M. H. Boyle

    1. Department of Psychiatry and Behavioural Neurosciences and the Offord Centre for Child Studies, McMaster University, Hamilton, Ontario, Canada.
    Search for more papers by this author

Dr RJ Van Lieshout, Department of Psychaitry and Behavioural Neurosciences, McMaster University, Chedoke Division, Box 2000, Central Building Rm 304, 1200 Main Street West, Hamilton, Ontario, Canada L8N 3Z5. E-mail: vanlierj@mcmaster.ca

Summary

Evidence suggests that a curvilinear relationship may exist between birth weight and later psychopathology. Increases in the prevalence of macrosomia and of two of its risk factors (maternal pre-pregnancy obesity and diabetes mellitus) and their amenability to intervention argue for a critical review of the association between macrosomia and mental illness. Of the nine studies in adults and six studies in youth that have examined associations between macrosomia and psychiatric disorders, seven have provided evidence suggestive of a link. Significant methodological variability and an inability to adjust for important confounders limit the findings of these studies. As a result, it remains unclear if individuals born macrosomic are at increased risk for psychopathology later in life. Future work should attempt to examine a broader range of psychiatric outcomes, use validated measures, include data on putative confounders and utilize genetically sensitive designs to assess associations between macrosomia, its precursors and later psychological and emotional functioning.

Introduction

As early as the mid-1950s, it has been suggested that suboptimal intrauterine environments can affect brain development in ways that may increase the risk for psychiatric disorder later in life (1). Given that the majority of the neuroanatomical mediators of cognition and emotion form in utero(2), it is reasonable to expect that exposure to physiological and psychological stresses in gestation could increase one's risk for psychopathology.

Understanding the impact of prenatal factors on the risk of developing psychiatric problems is important for two reasons. One, identification of causal risk factors could provide the scope for prediction and primary and secondary preventive interventions.Two, an understanding of the prenatal factors relevant to the etiology of psychiatric disorders can inform hypotheses regarding the pathophysiologies of these syndromes and guide the development of therapeutics.

Most of the research that has examined the relationship between prenatal environments and psychopathology in humans has focused on those born at low birth weight (LBW). However, increases in the prevalence of macrosomia (being born large for gestational age or at high birth weight) and two of its major risk factors, maternal pre-pregnancy obesity and diabetes mellitus (DM) (3–5), and the ease of identification and amenability to intervention of these risk factors argue for a critical examination of the association between macrosomia and later mental health problems. Emerging research suggests that an increased susceptibility to psychopathology may not be restricted to infants born small. Reverse J-shaped associations have been observed between birth weight and a number of disorders including DM (6). Despite this, few studies have examined associations between macrosomia and psychopathology. Most research that has examined the relationship between LBW and psychopathology has relegated macrosomic infants to control groups, possibly underestimating the strength of the association with LBW and masking possible links to macrosomia.

Interest in the association between LBW and later pathology is based on the fact that LBW reflects a stressful intrauterine milieu. However, macrosomia may also be a marker of exposure to intrauterine stress. In animal models, crowding stress (7) and adrenocorticotropic hormone administration (8) in late gestation are associated with increased birth weight. Macrosomia is also associated with an elevated risk of complications at birth including a higher incidence of obstetric trauma and maternal haemorrhage (9) which are also associated with an increased risk of certain forms of psychopathology (10). Pre-pregnancy obesity and DM, can be marked by increased levels of pro-inflammatory cytokines (11) and oxidative stress (12), states that are also associated with an elevated risk of certain types of psychiatric disorders (13,14). While multiple studies have reported associations between LBW and later psychopathology (10,15), the development of primary preventive efforts targeting the prenatal period require that causal relationships be demonstrated (16). Unfortunately, the presence of residual confounders such as the familial risk of psychopathology and the post-natally persisting suboptimal psychosocial environments have undermined the demonstration of causality in studies of LBW. Such confounders are also relevant to work that examines macrosomia.

In this paper, we will review the studies that have contained data on the association between macrosomia and later mental illness. Subsequently, we explore the pathways through which macrosomia and its antecedents could increase the risk of later psychopathology and make recommendations for studies aimed at clarifying empirically the strength and nature of this association, should it exist.

Methods

Studies for this review were identified using MEDLINE, EMBASE and PsycINFO searched from their inceptions to September 2010 using the terms ([fetal macrosomia OR large for gestational age OR high birth weight] AND [mental disorders ‘OR’ mental disease]) A targeted search of recent reviews of the prenatal programming of psychiatric disorders (10,15,17–19) was also conducted and studies in the reference lists of these papers were checked. Studies that reported data on all types of psychopathology in those born macrosomic were eligible and reviewed.

Results

Table 1 summarizes the 15 studies (20–34) that have examined associations between macrosomia and psychopathology and includes information on study design, findings and limitations. Nine of these studies reported outcomes in adults (20–28) and six in youth (29–34). Seven of these suggested that macrosomia is a significant predictor of later psychiatric disorder – five were studies of adults (20,24–27) and two, youth (33,34). However, those that reported positive results were not unequivocal.

Table 1.  Published data reporting on associations between macrosomia and psychopathology
ReferenceSampleDesignExposure macrosomia definitionReference groupOutcome (means of assessment)Findings (support/refute; risk, 95% CI)Limitations
  1. AGA, appropriate for gestational age; BW, birth weight; CBCL, Child Behavior Checklist; CI, confidence Interval; HBW, high birth weight; HR, hazard ratio; LBW, low birth weigh; LGA, large for gestational age; MR, mental retardation; NBW, normal birth weight; NS, not significant; OR, odds ratio; PI, Ponderal Index (a variably defined index of infant mass over length); RR, relative risk; SD, standard deviation; SES, socioeconomic status; TCP, Teacher's Checklist of Psychopathology.

Adults       
 Hultman et al., 1997 (20)Consecutively admitted in-patients vs. matched controls 82 cases, 164 controlsCase–controlHBW for length (≥1 SD)NBW for lengthSchizophrenia (in-patient psychiatrist diagnosis)Support
OR = 4.42 (1.97–9.91)
Sampling: hospitalization data only
Control of error: no examination of familial risk of psychopathology or post-natal environment
 Hultman et al., 1999 (21)Linked birth and psychiatric registries 167 cases, 835 controlsHistorical cohortLGA (≥2 SD)
PI (≥95th percentile)
AGA
PI (6th–94th percentile)
Schizophrenia (in-patient psychiatrist diagnosis)Refute
OR = 0.5 (0.2–1.6)
OR = 1.0 (0.4–2.2)
Sampling: hospitalization data only, followed to 22 years old, three cases ≥4500 g
Measurement: high BW cut-off used
Control of error: no examination of familial risk of psychopathology or post-natal environment
 Dalman et al., 1999 (22)Linked birth and psychiatric registries (up to age 22) 507 516 eligible, 238 casesHistorical cohort≥4500 g4000–4499 gSchizophrenia (in-patient psychiatrist diagnosis)Refute
RR = 0.7 (0.3–1.8)
Sampling: hospitalization data only, followed to 22 years old, five cases ≥4500 g
Control of error: no examination of post-natal environment
 Gunnell et al., 2003 (35) & 2005 (23)Linked birth and psychiatric registries 719 476 eligible, 736 casesHistorical cohort≥4000 g3500–3999 gSchizophrenia (in-patient psychiatrist diagnosis)Refute
HR = 1.02 (0.62–1.69)
Sampling: hospitalization data only, followed up to 27 years old, 19 cases ≥4000 g
 Bersani et al., 2007 (24)40 consecutively admitted male psychiatric clinic patients and 73 of their brothersCase–control≥4000 g3001–3999 gSchizophrenia (psychiatrist diagnosis)Support
OR = 4.52 (1.00–20.48)
Sampling: males only, small sample
 Cheung, 2002 (25)5567 members of the 1970 British Cohort StudyBirth cohort≥1.5 SD above the meanMean birth weight 3380 gPsychological symptoms (at age 26; malaise inventory)Mixed (supported when based on continuous outcome data but not when outcome categorized)Sampling: sample attrition
Measurement: implications of outcome used unclear
Control of error: no examination of familial risk of psychopathology
 Colman et al., 2007 (26)Members of 1946 British Birth Cohort (sample size varied by age)Birth cohort≥4500 g<4500 gSymptoms of anxiety and depression over the lifespan (a variety of measures were used, depending on age)Support (no OR reported in paper)Sampling: sample attrition
Measurement: use of multiple scales; psychometric properties not reported
Reporting: exact nature of relationship and magnitude not described in paper
Control of error: no examination of familial risk of psychopathology
 Herva et al., 2008 (27)8339 members of the 1966 North Finland Birth CohortBirth cohort4500–4999 g
PI ≥ 95th percentile
3000–3499 g
50–75th percentile
Depression (at age 31; Hopkins Symptom Checklist)Mixed (supports in females but not males; on HSC, not when reported as physician diagnosed)
OR = 2.02 (1.20–3.39)
OR = 1.53 (1.05–2.22)
Sampling: Sample attrition
 Favaro et al., 2006 (28)Cases (n = 187) and controls (n = 540) sampled from a birth cohort of Italian women with case sample augmented by the addition of eating disorders clinic patientsCase–control>4000 g2500–4000 gAnorexia or bulimia nervosa (structured clinical interview for the DSM)Refute
Anorexia
OR = 1.3 (0.5–3.3)
Bulimia OR = 0.6 (0.2–2.7)
Sampling: some ascertainment bias in the collection of cases
Control of error: no examination of familial risk of psychopathology or post-natal environment
Youth       
 Eaton et al., 2001 (29)Linked birth and psychiatric registries yielding 3325 cases and 102 905 controlsHistorical cohort>4000 g and LGA3000–4000 g and AGAAdolescent psychiatric in-patients (various diagnoses up to age 15; in-patient psychiatrist diagnosis)Refute
Autism:
RR = 1.07 (0.5–2.1)
MR: RR = 1.1 (0.6–1.9)
Asperger's: RR = 1.1 (0.6–1.9)
Learning disorder: RR = 0.97 (0.7–1.4)
Eating disorder: RR = 0.28 (0.0–2.1)
Sampling: hospitalization data only
Control of error: no examination of familial risk of psychopathology or post-natal environment
 Hultman et al., 2002 (30)Linked birth and psychiatric registries yielding 408 cases and 2040 controlsHistorical cohort≥4500 g
LGA (>2 SD above mean)
2500–4499 g
AGA
Autism (by age 10; in-patient psychiatrist diagnosis)Refute
Unadjusted OR = 1.5 (0.9–2.5)
Unadjusted OR = 1.7 (1.0–2.7)
Adjusted OR = 1.6 (0.9–2.8)
Sampling: hospitalization data only, restricted to those under age 10
Control of error: no examination of familial risk of psychopathology or post-natal environment
 Larsson et al., 2005 (31)Linked birth and psychiatric registries yielding 698 cases and 17450 controlsHistorical cohort>4500 g
>90th percentile for GA
3001–3500 g
10th–90th percentile
Autism (in-patient psychiatrist diagnosis)Refuted for both definitions
RR = 0.99 (0.59–1.67)
RR = 0.90 (0.67–1.22)
Sampling: mainly hospitalization data
Control of error: no examination of familial risk of psychopathology
 Linnett et al., 2005 (32)Linked birth and psychiatric registries yielding 834 cases and 20100 controlsHistorical cohort>4000 g3000–3999 gHyperkinetic disorder (psychiatrist diagnosis)Refute
RR = 1.0 (0.9–1.3)
Sampling: mainly hospitalization data
 Buschgens et al., 2009 (33)2230 Dutch pre-adolescentsCohort>4500 g2501–4499 gExternalizing symptoms (hyperactive/ impulsive, aggression, delinquency at age 10–12 using CBCL (parent), TCP (teacher)Support (continuous outcomes examined only; slightly different pattern of risk present depending on informant)
β values
Parent-rated (CBCL)
Inattention: NS
Aggression: 0.06
Delinquency: NS
Teacher-rated (TPC)
Inattention: 0.07 Hyperactivity/impulsivity: 0.09
Aggression: 0.08 Delinquency: 0.09
Control of error: no adjustment for SES
 Alati et al., 2009 (34)Prospective birth cohort of 4971 Australian childrenBirth cohortHighest BW quintileMiddle quintileSocial problems, internalizing and externalizing symptoms at age 14 (CBCL youth self-report)Mixed (support for social problems but not for internalizing or externalizing symptoms)
β values
Anxious/depressive: 0.39 (−0.01–0.78)
Social problems: 0.21 (0.02–0.41)
Measurement: self-report only

Six studies examined associations between being born macrosomic and schizophrenia (20–25) and all but two of these (20,24) were based on linked birth and psychiatric hospitalization registries. The two (20,24) that supported an association used a case–control design, included a relatively small number of cases and failed to adjust for parental diagnosis of psychotic disorder. Another study (35) that reported a positive association between macrosomia and schizophrenia in males who were followed up to age 22 (23) failed to report a link when follow-up was extended to age 27.

Studies of associations between macrosomia and more common psychiatric outcomes such as depression in adults are less common. Data on two population-based birth cohorts (26,27) found positive associations between high birth weight and elevated symptoms of depression in adulthood but the former was only positive for females and only when defined by a psychometric scale, not when assessed by self-report of a physician diagnosis. Cheung reported on the link between birth weight for gestational age and psychological symptoms assessed by the malaise inventory in a British birth cohort (25). In this study, more symptoms were reported among individuals born less than 0.5 standard deviations (SD) and more than 1.5 SD above the mean but only when these were reported on a continuous rather than a dichotomous scale.

Studies examining associations between macrosomia and psychopathology in children and adolescents are rarer. Three of these studies (29–31) used linked birth and hospitalization data to explore the association between macrosomia and autism. None reported an association. Another examined the link between perinatal variables and hyperkinetic disorder [attention-deficit hyperactivity disorder (ADHD)] in 834 cases and 20 100 register-based controls and found no link with macrosomia (32). However, two population-based studies of adolescents reported that macrosomia was a significant predictor of emotional and behavioural problems. In one study, significant associations were noted between being born >4500 g and five out of seven parent- and teacher-rated externalizing scales (e.g. inattention, hyperactivity-impulsivity, aggression and delinquency) in 10- to 12-year-old Dutch youth (33). A study of 14-year-old Australians reported that those born at weights greater than the 80th percentile for gestational age had significantly increased levels of self-reported social problems and a trend towards elevated levels of internalizing (depressive/anxiety symptoms) and externalizing problems although these latter two scales did not reach statistical significance (34).

Six studies defined macrosomia as being born at 4500 g or more (22,26,27,30,31,33) and four used ≥4000 g as a cut-off (23,24,28,29,32). The remainder (21,25,34) defined macrosomia according to the distribution of birth weights in that study or incorporated gestational age into their definition. Studies by Hultman and colleagues' (30) and Larsson et al. (31) examined both birth weight and birth weight adjusted for gestational age as predictors of psychopathology. The birth weights to which macrosomic infants were compared varied but the most common definition was 2500–4499 g. Absence of a common definition of macrosomia and normal birth weight complicates comparisons between and synthesis of studies in this area. The way in which psychopathology was defined and measured was also highly variable. Seven studies relied on in-patient hospitalization data to define diagnosis (20–23,29–31) and three used out-patient diagnoses24,28,32. Five studies used scales to define disorder (25–27,33,34). Unfortunately, a number of studies failed to adjust for putative confounders of the link between macrosomia and mental illness. Only seven (20,21,26,28–31) attempted to adjust for familial risk of psychopathology or the presence of maternal mental illness in pregnancy, and six failed to adjust for post-natal environmental confounders (20–22,28–30).

While significant variability existed in the criteria used to classify macrosomia, there was no apparent link between the definition used and positive or negative findings. However, the outcomes examined, the ways in which they were measured, and the means of identifying and selecting cases did influence the likelihood of observing significant associations. Studies that used hospital referrals and classified disorder according to in-patient discharge diagnoses were less likely to report a link than studies that selected their cases from general population samples containing all possible individuals with disorder (25–27,33,34). Interestingly, these five studies not only all reported positive results, but also used scales to measure mental illness as opposed to psychiatrist-derived discharge diagnoses. Moreover, studies of disorders less likely to result in hospitalization (e.g. depression vs. schizophrenia) were more likely to report a positive association. However, studies of disorders more likely to require hospitalization most frequently used in-patient discharge data and so it is difficult to determine if these associations are due to true etiological links or study-specific methodological features.

Discussion

It is not yet possible to come to firm conclusions about the association between macrosomia and psychopathology at the present time given the current mixed findings and methodological and clinical heterogeneity inherent in these studies. Contributing to this heterogeneity are variability in the disorders assessed, developmental stages examined, sample sizes and sampling frames available, measurement tools and the variables available to address confounding.

The rarity of and difficulties associated with recruiting and retaining individuals with illnesses such as schizophrenia and bipolar I disorder in general population cohorts frequently necessitates the use of hospital records based on in-patient stays for sampling. Unfortunately, such registry-based strategies exclude individuals who are not hospitalized. While this may not greatly affect the representativeness of populations of individuals with the above conditions, utilization of in-patient data alone will miss the majority of sufferers of most depressive and anxiety disorders. While sampling through the general population produces more representative samples of individuals suffering from these problems, they are not without limits. Such studies can be affected adversely by high rates of attrition (25–27), the psychometric properties of their measures (34) and missing data. Unfortunately, while case–control studies may be an efficient way to examine rare disorders, they are more susceptible to bias than cohort studies given their separate sam macrosomia is a significant predictor of later mental disorders.

Current approaches to classifying macrosomia may also undermine efforts to study its effects. At present, there is no consensus on its optimal definition and the appropriate reference group for predicting psychopathology. Defining macrosomia in a variety of ways and contrasting the predictive power of these cut-offs may benefit studies in this area. Variability in the quality of assessment data also poses problems for the interpretation of the above studies. Unfortunately, some failed to report on the psychometric properties of these measures, raising concerns about the attenuating effects of random measurement error.

Perhaps the biggest threat to establishing valid associations between macrosomia and psychopathology arises from the failure to control for potential confounders of the association, namely the familial risk of mental illness and the persisting effects of adverse psychosocial environments. Of the positive studies, only two (33,34) of seven adjusted for familial risk of psychopathology and three (25,26,34) controlled for post-natal psychosocial disadvantage. Studies that fail to adjust for these confounders may overestimate the strength of the association between macrosomia and psychopathology and further place into question the validity of the link. Disentangling genetic and environmental confounders in studies of prenatal programming is complicated; not only are maternal genes passed on to the fetus but they also shape pre- and post-natal environments. The use of genetically sensitive study designs may help us to better assess the relative contributions of these factors.

Certainly, it appears that individuals born macrosomic are exposed to a suboptimal intrauterine milieu that may in fact be stressful. However, it is not clear if this is sufficient to produce psychopathology. If prenatal factors are responsible for this increase in risk, it is clear that our current understanding of the mechanisms underlying it is not well developed. Indeed, it is possible that the aforementioned increases in pro-inflammatory cytokines and oxidative stress accompanying maternal diabetes and obesity could play a role. However, hormonal (e.g. cortisol, oestrogen, insulin) and dietary (increased glucose, folate deficiency) factors could also be involved.

It is also conceivable that macrosomia could affect the risk of psychopathology via post-natal mediating factors known to be associated with both macrosomia and psychopathology including problems with obesity and physical health in childhood (6). These illnesses or individuals' struggles with them could contribute to or be responsible for observed links. For example, it is known that macrosomic infants are more likely to become obese adolescents, and that such individuals are at elevated risk of bullying others and of being bullied themselves (36). It is also conceivable that macrosomia increases the risk of psychiatric problems through its putative effects on cognition (37,38), personality and/or stress sensitivity (39,40) although this has not yet been studied.

Even if stronger evidence of associations between macrosomia and mental disorders existed, it would still be unclear if this relation is causal or if macrosomia is merely a marker of putative upstream causal factors of later psychiatric problems including maternal obesity, weight gain during pregnancy and/or DM. Unfortunately, data on these predictors have either been lacking or ignored in studies examining the links between macrosomia and psychopathology. A limited number of studies have examined maternal obesity and DM for their association with psychological problems in the offspring of these gestations but none have taken macrosomia into account. A small number of studies suggest that individuals exposed to diabetic pregnancy are more likely to develop schizophrenia (10), ADHD (36) and anorexia nervosa (28). Moreover, some work also suggests that maternal pre-pregnancy obesity is associated offspring emotion regulation problems (37), ADHD symptoms (38) and perhaps even an increased risk of schizophrenia (39). It is not clear whether these links are mediated by intrauterine biological exposures, maladaptive maternal psychological characteristics that may put some women at risk for obesity in the first place, as well as affect their offspring post-natally (41), genetic factors or a combination of the three. This is certainly an area in need of further study.

Summary and conclusions

Despite theoretical arguments for expecting associations between macrosomia and later psychopathology, studies examining this link have produced mixed findings, which may reflect their vulnerability to a number of definitional and methodological limitations. To address these limitations, future studies might focus on the use of genetically sensitive designs, or, in their absence and depending on the outcome of interest, registry or population-based cohorts that contain adequately sized samples, incorporate measures of gestational age, perform assessments across development and use multiple informants and optimal measures of psychopathology.

If a link between macrosomia and psychiatric disorders is confirmed, work could then focus on identifying those at risk and on assessing the genetic and other biological antecedents (i.e. maternal pre-pregnancy obesity, weight gain during pregnancy and maternal DM), mediators and moderators of these links, and on modelling the trajectories of these individuals over time. Animal models of macrosomia, maternal obesity and DM are powerful tools that have the potential to elucidate the putative mechanisms by which these exposures could increase risk. While a considerable challenge, given the increasing prevalence of macrosomia and its risk factors, the demonstration of causal associations between it and psychiatric illness may be an important goal and could provide realizable targets for the prevention of mental disorders.

Conflict of Interest Statement

None.

Acknowledgements

Dr Van Lieshout is supported by a Canadian Institutes for Health Research Fellowship and Dr Boyle by a Canada Research Chair in the Social Determinants of Child Health. We would like to thank Dr Glenda MacQueen and Dr Valerie Taylor for their helpful comments on an earlier version of this manuscript.

Ancillary