[1] 11w Family Psychiatric History, Cerebrospinal Fluid Monoamine Metabolites, and Temperament in Infants

[1] 165w Background: Variations in cerebrospinal fluid (CSF) levels of the monoamine metabolites 5-hydroxyindoleacetic acid, 3-methoxy-4-hydroxyphenylglycol, and homovanillic acid have been associated with behavioral abnormalities in nonhuman primates, and with psychopathology in studies of children and adults. Methods: We assayed monoamine metabolites in "leftover" spinal fluid from 167 neurologically normal newborn infants (0 -3 months of age), and later (at age 18 -21 months of age) obtained their family psychiatric histories and assessed their temperament using the Colorado Childhood Temperament Inventory (CCTI). Results: Family history of antisocial personality disorder predicted significantly lower scores for soothability (p ϭ .003) at 18 -21 months. There were no statistically significant associations between newborn monoamine metabolite levels and any aspect of temperament on the CCTI. Conclusions: These findings suggest complex relationships between genetic liability for psychiatric disorders and CSF monoamine metabolite levels; those relationships do not seem to be mediated by infant temperament. It appears likely that interindividual differences in monoamine metabolite levels change over the course of development in humans.

[1] 177w S everal years ago, in an editorial published in Biologi- cal Psychiatry (Belmaker and Biederman 1994), it was proposed that studies linking genetic markers, temperament, and psychopathology might prove very useful in elucidating the causes of psychiatric disorders. Since many heritable psychologic traits appear to be the result of influences of multiple genes (Lesch et al 1996), it was reasoned that each gene contributes only a small share of the variance to a given outcome. Presumably, the more complex the psychologic outcome, the more genes may be involved. Therefore, it seemed to make sense to study links between genes and "simpler" heritable dimensions of behavior (temperament), which would reduce the gene: outcome ratio, increase the share of the variance that each gene contributes, and thereby increase the power of research studies to find the relevant genetic markers. Basic temperament traits are known to predict some aspects of psychopathology (Caspi et al 1995;Cloninger et al 1993;Rende 1993), and knowledge of the genetic basis of relevant dimensions of temperament could lead to better understanding of genetic influences on psychopathology.

[2] 207w This study represents a variation on this methodologic theme by examining genetic risk for psychiatric disorders, suspected biological correlates of those disorders, and temperament. Although the genetics of monoamine neurotransmitters in humans have not been well worked out, cerebrospinal fluid (CSF) concentrations of the metabolites of those neurotransmitters are thought to reflect brain monoaminergic activity (Stanley et al 1985), and have been associated with a variety of psychopathologic outcomes in humans. For example, CSF levels of the serotonin metabolite 5-hydroxyindoleacetic acid (5-HIAA) have been consistently associated (correlation coefficients on the order of .4 -.8) with violent behavior and impulsive suicide (Roy et al 1991;Virkkunen et al 1994Virkkunen et al , 1995)); CSF levels of the dopamine metabolite homovanillic acid (HVA) have been associated with major depression (Salzman et al 1993), substance abuse (Gabel et al 1994), psychosis (Maas et al 1997), response to treatment in attention deficit disorder (Castellanos et al 1996), and early mortality in psychiatric patients (Faustman et al 1993); CSF levels of the norepinephrine metabolite 3-methoxy-4-hydroxyphenylglycol (MHPG) have been associated with major depression (Samson et al 1992;Sharma et al 1994). Variations in the relationships between these CSF metabolites (within individuals) have been associated with phenotypic variations in patients with schizophrenia (Hsaio et al 1993).

[3] 155w Studies in nonhuman primates have demonstrated considerable stability of interindividual differences in CSF monoamine metabolite levels over time (Higley et al 1992;Shannon et al 1995), and have provided evidence for both genetic and early rearing influences on those levels (Clarke et al 1995(Clarke et al , 1996;;Higley et al 1993). In humans, however, it is not yet known whether interindividual differences are preserved over the life course or to what extent they are genetically determined. In a study of adult twins and unrelated individuals, Oxenstierna et al (1986) found evidence for familial influences on the order of .6 for HVA and 5-HIAA, but reported that shared environmental influences were greater than genetic influences for both metabolites. The sample size in that study (n ϭ 30) severely limited the power of the study to confirm genetic influences comprising less than 50% of the total variance. For MHPG, genetic influences on the order of .75 were reported.

[4] 110w Since the development and stability of monoamine neurotransmitter systems may be influenced by complex gene-environment interactions over the course of the life span, behavioral correlates of one-time samplings of monoamine metabolites in mature individuals need to be interpreted with caution. Multiple samplings of CSF over time, however, are not readily obtainable in human populations, since ethical and practical considerations generally preclude invasive procedures to gather this kind of data. This is particularly true for children, despite the fact that knowledge about monoamine metabolites in developing individuals is especially important, given its implications for understanding cause and effect in the observed associations between monoamine systems, psychopathology, and resilience later in life.

[5] 162w One exception to the difficulty in obtaining human samples is the fact that CSF is frequently drawn and examined in newborn babies (birth to 10 weeks of age) with minor febrile illnesses, since fever, in a small minority of infants, is the sole presenting symptom of bacterial meningitis (which is imminently life-threatening). The vast majority of these infants turn out not to have meningitis; most have ordinary viral infections. We recently demonstrated that monoamine metabolite levels could be reliably assayed from "leftover" samples of human newborn CSF, that the levels were largely unaffected by variations in the infants' physiologic condition (including febrile status and behavioral state) at the time of lumbar puncture, and that interindividual differences in HVA and 5-HIAA appeared to be stable over the first year of life (Constantino and Murphy 1996). We subsequently reported that there was a modest but statistically significant inverse correlation between newborn CSF 5-HIAA and family history of antisocial personality disorder (Constantino et al 1997).

[6] 146w We undertook this study to determine if either genetic liability for psychiatric disorders or newborn monoamine metabolite levels predicted temperament, defined as human personality traits that are identifiable in infancy and that are presumed to be heritable (Goldsmith et al 1987). We hypothesized that if variations in monoamine neurotransmission are causal factors in psychiatric disorders (particularly those with which they have been previously associated in clinical studies of older individuals), we might observe a relationship between newborn monoamine metabolite levels and: a) genetic liability for those disorders; and b) temperament dimensions that have been proven to be heritable (Plomin et al 1993) and that might mediate the relationship between genetic liability and behavioral (psychopathologic) outcomes (as suggested by Rende 1993;Caspi et al 1995 and others). We know of no previous study attempting to correlate CSF monoamine metabolite levels with infant temperament in an epidemiologically ascertained sample.

[1] 194w Lifetime prevalence rates for psychiatric disorders among the 2068 adult first-and second-degree relatives of the 167 infants, as ascertained by the FHAM, were as follows: antisocial personality disorder (ASPD)-.02; alcoholism-.04; drug abuse-.02; major depression-.05; lifetime history of suicide attempt-.01; schizophrenia-.001. After Bonferroni corrections, there were no statistically significant correlations between levels of any of the monoamine metabolites and prevalence of psychiatric disorders among first-and second-degree relatives. The correlation between 5-HIAA and prevalence of ASPD among first-and second-degree relatives was weak (r ϭ Ϫ.13) but in the expected direction, as previously reported (Constantino et al 1997). There was a similar modest correlation between MHPG and depression in the expected direction (r ϭ .17). Of the 3 infants with documented family histories of schizophrenia (all in second-degree relatives), 2 were among a group of 9 high outliers in the MHPG distribution (Fisher's Exact Test p ϭ .02). It should be noted that in contrast to CSF levels for 5-HIAA and HVA, MHPG levels in infancy are distributed in a fairly narrow range (15-200 pmol/mL) except for approximately 5% of the population, which constitute high outliers in the distribution (300 -900 pmol/mL; see Constantino and Murphy 1996).

[2] 149w Among 52 infants for whom temperament assessments were completed at both 18 and 21 months, we observed moderate stability of CCTI temperament dimensions [r s ϭ .45-.65, p s Ͻ .001, highly comparable to what was reported by Plomin et al (1993) using the same instrument] with the exception of the dimension "activity," as shown in Table 1. Correlations between monoamine metabolite levels and CCTI temperament dimensions were all statistically nonsignificant after correction of p values using the Bonferroni method. Analyses were repeated using subsamples that included narrower age ranges than the entire 18 -21-month range, but this did not reveal any statistically significant correlations between monoamine metabolite levels and temperament. Correlations between monoamine metabolites and within-individual changes in temperament from 18 to 21 months were also calculated (for the 52 subjects who had temperament assessments done at 18 and 21 months); none were found to be statistically significant.

[3] 116w There was an inverse correlation (r ϭ .26, n ϭ 167, Bonferroni corrected p ϭ .04) between family psychiatric history of ASPD and soothability at 18 -21 months. There were otherwise no significant associations between family psychiatric history and CCTI temperament dimensions. A linear regression analysis for soothability was performed, in which uncorrelated factors (including 5-HIAA, MHPG, family psychiatric history, birthweight, gender, race, and age at follow-up) were entered as independent variables. When controlling for these variables, the inverse association between family history of ASPD and soothability remained statistically significant [F ϭ 1.97; df ϭ 157,8; p ϭ .05; R 2 ϭ .09; regression coefficient (ASPD) ϭ Ϫ3.6 Ϯ 1.2; t ϭ Ϫ2.974; p ϭ .003].

[4] 24w Separate analyses were conducted to include the 2 subjects with developmental delay, but the findings described above were not significantly changed by their inclusion.

[1] 69w We hypothesized that if variations in monoamine neurotransmission are causal factors in psychiatric disorders (particularly those with which they have been previously associated in clinical studies of older individuals), we might observe a relationship between newborn monoamine metabolite levels and: a) genetic liability for those disorders; and b) temperament dimensions that have been proven to be heritable and that might mediate the relationship between genetic liability and psychopathologic outcomes.

[2] 91w In this study, which had sufficient statistical power to detect correlations on the order of .3, we found little evidence for either relationship. Monoamine metabolite levels were not associated with any dimension of temperament or with change in any dimension of temperament in a subsample of individuals studied at both 18 and 21 months. We previously reported an association between 5-HIAA and sociability at 9 months in a small sample (Constantino and Murphy 1996), but there was no association between 5-HIAA and sociability when temperament was measured at 18 -21 months.

[3] 143w Two of the 3 infants with schizophrenic second-degree relatives were high outliers in the MHPG distribution, but this finding should be interpreted with caution given the nature of the MHPG distribution in infants and the very low prevalence of schizophrenia ascertained in this sample. Aside from trends for a positive correlation between MHPG and depression and a negative correlation between 5-HIAA and family history of ASPD, there were no significant correlations between monoamine metabolite levels and family history of psychiatric disorder. We previously reported a statistically significant relationship between 5-HIAA and family history of ASPD in a study that was designed to examine that relationship a priori (Constantino et al 1997). The fact that the association was not statistically significant after Bonferroni corrections in the present analysis reflects and highlights what we previously emphasized: that the actual association is probably small in magnitude.

[4] 362w We did find an intriguing negative correlation between family history of ASPD and 18 -21-month soothability, which has never previously been reported and warrants further study. It is possible that some of the genetic liability for ASPD is mediated by "difficult to soothe" temperament, but this component does not appear to be mediated by 5-HIAA. Although infant temperament is a weak predictor of antisocial behavior in childhood (Sanson et al 1991), there have been few studies that have carefully measured temperament in infancy and followed individuals to adulthood. A recent behavioral genetic study involving several thousand twin pairs (Lyons et al 1995) has suggested that despite the known stability of antisocial patterns of behavior over the life span, heritability of juvenile antisocial traits is markedly lower than that for adult antisocial traits. In this regard, the association between ASPD and soothability is particularly interesting; it is possible that low soothability might predict adult antisocial behavior even though it does not seem to predict childhood antisocial behavior. We found that CCTI dimensions were moderately stable over the period from 18 to 21 months; our results were very similar to those obtained for time 1-time 2 correlations between 14 and 20 months using the same instrument in a large twin sample (Plomin et al 1993); in that study, genetic factors influenced both continuity and discontinuity in temperament within individuals over the period from 14 to 20 months. It remains to be seen whether genetically determined discontinuities in temperament during early childhood might correlate with other biological markers or with psychopathologic outcomes. The development of new methods of measuring temperament that reflect the influences of underlying neurotransmitter systems might prove more useful than the CCTI (heritable though its dimensions may be) in elucidating the complex relationships between genetic liability, neurotransmitter systems, temperament, and psychopathology. Similarly, the availability of better measures of the function of neurotransmitter systems using direct genetic studies of those systems (Murphy 1990), neuroimaging of their components (Bremner et al 1997), or pharmacologic challenges directed toward specific components of those systems may make it possible to identify relationships between neurotransmitter function and risk for psychopathology that could not be identified in this study.

[5] 188w If one were to assume relative stability of interindividual differences in monoamine metabolites over the life course, these findings would raise questions about causal relationships that have been inferred from associations between monoamine metabolite levels and psychopathology in one-time samplings of older affected individuals; previously reported associations might be attributable to the effects of disorders on monoamine metabolites, rather than the other way around. An alternative assumption is that interindividual differences in newborn monoamine metabolites are not preserved over the life course: that genetic or environmental factors exert correlated influences on monoamines and behavior over the course of later development. Whether interindividual differences change as a result of disorder or as a result of normal development, our findings suggest that there is change, since correlations between infant monoamine metabolites and family psychiatric history were less than proportionate with correlations between monoamine metabolites and psychiatric disorders (of known genetic influence) that have been reported in adults. It is noteworthy that a small previous behavioral genetic study of monoamine metabolites in adults (Oxenstierna et al 1986) suggested significant family environmental influences on 5-HIAA and HVA (on the order of .25).

[6] 187w This study has a number of limitations. Ascertainment of psychiatric disorders by family history resulted in a prevalence rate for schizophrenia that was an order of magnitude lower than the established point prevalence. Data were not available on whether or not infants were breast-fed, or whether mothers of breast-fed infants were using medications or illicit drugs. Given the pronounced main effects of race and socioeconomic status (SES) on the initiation and maintenance of breast-feeding (Visness and Kennedy 1997), however, we would have expected monoamine metabolite levels to vary as a function of race and SES if breast-feeding were influencing the metabolite levels. Other potential sources of error include the CSF sampling method, the fact that CSF monoamine metabolite levels may only approximate levels of CNS monoaminergic neurotransmission, unassessed factors related to whether infants presenting to emergency rooms with fever might or might not be representative of the population, and the possibility of dominant rather than additive genetic influences on monoamine metabolite levels. Any of these factors could have led to underestimation of the magnitude of genetic influences on monoamine metabolites linked to temperament or family psychiatric history.

[7] 114w In general, there are a myriad of influences on CSF monoamine metabolite levels in nonhuman primates and human adults, some genetically mediated, others a function of environmental influences. Recent studies have just begun to elucidate the complex nature of the genetic influences and suggest that a variety of genes each contribute a fraction of the inherited variance in various components of monoamine neurotransmitter systems and their function (Karayiorgou et al 1997;Lesch et al 1996). It is quite possible that there are developmental effects on monoamine systems-not only on the levels of each respective monoamine metabolite, but on the relationships between monoamine neurotransmitters-which result in changes in interindividual differences over time (Kraemer and Clarke 1996).

[8] 122w Thus, environmental insults and the psychopathologic outcomes with which they have been associated [including antisocial behavior (Cadoret et al 1995) and depression (Kendler and Karkowski-Shuman 1997)] may not be reflected by unidirectional changes in a specific monoamine or its metabolites; rather they may manifest themselves through abnormalities in how the monoamine systems are integrated with one another over the course of development. The elucidation of such complex relationships would require research studies that simultaneously control for the effects of genetic and environmental influences, and that track the longitudinal course of monoamine systems and behavior within a developmental framework. To be feasible and ethical, such studies will probably require newly developing technologies for noninvasive assessment of central nervous system monoamine function in humans.

[1] 185w This study was an extension and follow-up of a previous study (Constantino et al 1997) that examined newborn 5-HIAA and its relation to family history of antisocial personality disorder. The rationale and procedure for securing and analyzing newborn CSF samples for this study have been described elsewhere (Constantino and Murphy 1996). Briefly, leftover portions of CSF drawn from febrile infants (n ϭ 646, age birth to 3 months) in the emergency department at Saint Louis Children's Hospital during the calendar year 1995 were obtained from the Clinical Chemistry laboratory and stored in plastic vials at Ϫ70°C until assayed for monoamine metabolites using high-performance liquid chromatography with electrochemical detection (Murphy et al 1991). The standard procedure for clinical analysis of spinal fluid is that the second of three 0.5-1.0-mL aliquots obtained at lumbar puncture is the portion that is sent to the Clinical Chemistry laboratory for measurement of glucose and protein. Harvesting the leftover samples exclusively from this laboratory served to minimize potential confounding influences of gradient effects on CSF monoamine metabolite levels, which have previously been observed for 5-HIAA in newborns (Constantino et al 1997).

[2] 250w Medical records of all 646 infants were reviewed, and 350 infants were excluded from the study on the basis of meningitis or evidence of neurologic abnormality (Constantino and Murphy 1996). Of the 296 remaining children, attempts were made to interview at least one parent during 1996. We were able to reach 199 of the families (67.2%), of whom all but four agreed to participate (total n ϭ 195). The group of children whose families could not be reached differed by race (65% black, 35% whitesee below), but not by mean monoamine metabolite levels, gender, age at the time of lumbar puncture, or birth weight. After complete description of the study to the subjects, written informed consent was obtained. Family psychiatric history, Denver Developmental Screen, and 18 -21-month temperament data was then obtained on all children who reached 18 months by April of 1997 (n ϭ 169). For 52 infants, we were able to obtain temperament assessments at both 18 and 21 months, to test the short-term stability of temperament [as measured by the Colorado Childhood Temperament Inventory (CCTI)] in our sample. Two children were excluded from the analysis based on significant developmental delays that were clinically not evident at the time of the first newborn assessment-when an initial medical record review excluded all neurologically compromised children; this brought the sample size to 167. Sample characteristics were representative of the population of the metropolitan St. Louis area and were as follows: 52% girls, 48% boys; 44% black, 54% white, 2% other.

[3] 159w 5-HIAA and HVA exhibited subtle circadian rhythms (Constantino and Murphy 1996;Constantino et al 1997) and were significantly associated with age at the time the lumbar puncture was performed (Constantino et al 1997). Therefore, for all subsequent analysis of data, the levels of these metabolites were age-adjusted using the slope of the respective regression lines from plots of age versus metabolite level (for 5-HIAA, the standardized regression coefficient was Ϫ.396, SE ϭ .065, p Ͻ .0001; for HVA the standardized regression coefficient was Ϫ.309, SE ϭ .075, p Ͻ .0001). Monoamine metabolite levels did not vary as a function of gender, race, or socioeconomic status, as was found in our previous studies of monoamine metabolites in newborn infants (Constantino and Murphy 1996;Constantino et al 1997). In the current follow-up sample, means for each metabolite were in keeping with those previously reported and were as follows: 5-HIAA: 675 Ϯ 235 pmol/mL; HVA: 825 Ϯ 256 pmol/mL; MHPG: 126 Ϯ 118 pmol/mL.

[4] 111w The Family History Assessment Module (FHAM) was conducted over the telephone with the child's mother or father after constructing a family tree of first-and second-degree relatives of the infant. The FHAM is a structured diagnostic instrument that is used to assess DSM-III-R psychiatric disorders among relatives of the informant, including major depression, antisocial personality disorder, schizophrenia, alcohol dependence, and substance abuse. The validity of the instrument was established in a study comparing family histories with direct diagnostic interviews of 2654 individuals (Rice et al 1995). Prevalence of each disorder in each infant's family was obtained by dividing the number of affected persons by the total number of adult firstand second-degree relatives.

[5] 39w The Denver Developmental Screening Test II (Frankenburg et al 1992) was also administered in the telephone interviews with parents. This instrument is used widely to screen for developmental delays in language and gross motor, fine motor, and social development.

[6] 124w The CCTI was administered via telephone interviews with parents. The CCTI is an amalgam of the emotionality-activitysociability dimensions described by Plomin and coworkers (Rowe and Plomin 1977;Plomin et al 1993), as well as the nine dimensions of temperament assessed by Thomas and Chess (1977) in the New York Longitudinal Study. Its psychometric properties have been established in studies of children aged 1-4 years. The emotionality and sociability dimensions predicted scores for anxiety, depression, and attentional problems in a sample of 164 children followed from 1 to 7 years of age (Rende 1993). The MacArthur Longitudinal Twin Study has demonstrated significant genetic influence for both continuity and within-individual change in temperament from 14 to 20 months of age using this instrument (Plomin et al 1993).

[7] 97w Pearson's coefficient of correlation was computed for the associations between a) monoamine metabolite levels and lifetime prevalence of FHAM disorders among first-and second-degree relatives; b) monoamine metabolite levels and CCTI temperament dimensions; and c) prevalence of FHAM disorders and CCTI temperament dimensions. The statistical significance of correlation coefficients was corrected using the Bonferroni method. Linear regression analysis was planned to explore further any correlation with temperament (headings b or c above) greater than .24 (Bonferroni corrected p Ͻ .1), to assess the relative contributions of monoamine metabolite levels and family psychiatric history to that specific temperament dimension.