maternal–fetal blood incompatibility and the risk of schizophrenia in offspring
TRANSCRIPT
www.elsevier.com/locate/schres
Schizophrenia Research
Maternal–fetal blood incompatibility and the risk of schizophrenia
in offspring
Beverly J. Insel *, Alan S. Brown, Michaeline A. Bresnahan,
Catherine A. Schaefer, Ezra S. Susser
Mailman School of Public Health of Columbia University Epidemiology, 722 W. 168 St., New York 10032, United States
Received 18 November 2004; received in revised form 26 May 2005; accepted 6 June 2005
Available online 11 July 2005
Abstract
Objective: Predicated on a maternal immune response to paternally inherited foreign fetal blood antigens, we hypothesized that
maternal–fetal blood incompatibility increases susceptibility to schizophrenia in the offspring. The relation between schizo-
phrenia and maternal–fetal blood incompatibility, arising from the D antigen of the Rhesus (Rh) and the ABO blood group
antigens, was examined in a cohort of live-births.
Method: The data were drawn from the Prenatal Determinants of Schizophrenia Study, a cohort of births occurring between
1959 and 1967 to women enrolled in a Kaiser Permanente Plan-Northern California Region (KP). Adult offspring belonging to
the KP from 1981 to 1997 were followed for the incidence of schizophrenia spectrum disorder (SSD). Cox proportional hazards
regression was the primary analytic technique.
Results: Among second and later born offspring, the adjusted incidence rate ratio (RRadj) of SSD was 1.80 (95% CI=0.71–
4.58) for the Rh incompatible offspring compared with the Rh compatible offspring; with the males exhibiting higher rate ratio
(RRadj=2.37; 95% CI=0.82–6.86) than the females (RRadj=0.93 95% CI=0.12–7.01). Among all offspring, the RRadj for ABO
incompatibility was lower and the elevated rate ratio was similarly limited to the males (RRadj=1.68; 95% CI=0.76–3.70). For
Rh and/or ABO incompatibility, the RRadj was 1.57 (95% CI=0.87–2.82). A statistically significant result was detected only for
the male offspring (RRadj=2.22; 95% CI=1.10–4.47).
Conclusion: Although the results should be interpreted with caution given the few events of SSD, the findings extend the line
of evidence that maternal–fetal blood incompatibility is a risk factor for schizophrenia spectrum disorder; with the strongest
evidence to date implicating that the susceptibility pertains only to male offspring.
D 2005 Elsevier B.V. All rights reserved.
Keywords: Schizophrenia; Schizophrenia spectrum disorder; Maternal–fetal blood incompatibility; Rh incompatibility; ABO incompatibility
0920-9964/$ - see front matter D 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.schres.2005.06.005
* Corresponding author.
E-mail address: [email protected] (B.J. Insel).
1. Introduction
Mounting empirical evidence for a neurodevelop-
mental origin of schizophrenia is fostering research on
80 (2005) 331–342
B.J. Insel et al. / Schizophrenia Research 80 (2005) 331–342332
the possible prenatal contribution to the risk of schizo-
phrenia (Weinberger, 1987; Susser et al., 1999; Mur-
ray and Fearon, 1999; McGrath et al., 2003). While
epidemiologic findings for individual prenatal risk
factors are not definitive, the evidence collectively
suggests that prenatal insults are of etiological signif-
icance in schizophrenia.
In the present study, we hypothesized that offspring
of a maternal–fetal blood incompatible pregnancy are
at increased risk for schizophrenia, based on biolog-
ical plausibility and prior reports (see below). In
maternal–fetal blood incompatibility, an elevated
risk of schizophrenia in the offspring is postulated to
arise from a maternal immune response to the pater-
nally inherited foreign fetal red blood cell antigens.
Disruption of fetal neurodevelopment may evolve
from multiple immunologic pathways leading to sev-
eral pathological states including hemolytic induced
chronic fetal hypoxia (Mollison et al., 1993), the
accumulation of neurotoxic bilirubin within the devel-
oping brain (Cashore, 1990), as well as more specu-
lative pathological processes involving the cross-
reaction of maternal IgG antibodies with fetal brain
antigens (Laing et al., 1995), and the breakdown of
maternal immune tolerance of the fetus (van Gent et
al., 1997).
Rh and ABO maternal–fetal blood incompatibil-
ity both can induce the production of maternal IgG
antibodies against the fetal red blood cell antigens.
The IgG antibodies can in turn transverse the pla-
centa, enter the fetal circulation and attack fetal red
blood cells (Bowman, 1999), and also gain access
to the developing brain due to the immaturity and
the permeability of the blood brain barrier (Adinolfi,
1985). A significant maternal antibody response can
provoke hemolytic disease of the fetus. An Rh-
negative woman requires a sensitizing exposure to
Rh-positive cells. As a result, firstborn Rh incom-
patible offspring rarely experience hemolytic dis-
ease, while second and later Rh incompatible
offspring are more likely to suffer its immunological
sequalae. In contrast, since most group O women
have anti-A and anti-B antibodies predating preg-
nancy, ABO hemolytic disease can occur in first
pregnancies (Rawson and Abelson, 1960; Ozolek et
al., 1994). ABO and Rh incompatibility are the two
most common causes of hemolytic disease of the
fetus. Rh disease is associated with serious morbid-
ity and mortality while ABO disease is consistently
milder.
Rh incompatibility, but not ABO incompatibility,
has been examined as a risk factor for schizophrenia
in several studies. A historical cohort study, con-
ducted by Hollister et al. (1996) within a Danish
sample born between 1959 and 1961, was the first
study to explore the hypothesis. The authors restrict-
ed the analyses to males due to the small number of
females diagnosed with schizophrenia within the
cohort. They found that among the men, the rate
of schizophrenia was significantly greater in the Rh
incompatible group than in the Rh compatible group
(RR=2.78, 95% confidence interval=1.2– 6.6). The
risk was confined to second and later-born offspring
from the Rh incompatible pregnancies. The study
had some limitations: the restriction to male subjects,
the use of hospital registry diagnoses for schizophre-
nia, and subject eligibility that was contingent on
neonatal blood typing, which was not a routine
procedure at that time. To further explore this asso-
ciation, Palmer et al. (2002) conducted a family-
based candidate-gene study using patient–parent(s)
pairs and trios comprising the youngest affected
male or female sibling and at least one biological
parent. The families were drawn from a Finnish
sample born prior to the widespread use of anti-Rh
(D) prophylaxis. Their estimated relative risk for Rh
maternal–fetal genotype incompatibility was 2.6
(90% CI=1.1, 6.0). Kraft et al. (2004) extended
the work of Palmer et al. (2002) using the same
Finnish sample and included all affected siblings
within a family. They reported a relative risk of
1.7 (90% CI=1.1, 2.5) in second and later born Rh
incompatible offspring, based on the assumption that
firstborn incompatibles were not at increased risk.
While the sample of Palmer et al. (2002) and Kraft
et al. (2004) appeared to contain a sufficient sample
of females (respectively 38% and 42% of the entire
sample), the effect estimates were not differentially
reported by sex of affected offspring.
The Rh incompatibility–schizophrenia association
has also been tangentially included in epidemiologic
studies that investigated the etiologic contribution of
obstetric complications. Two meta-analyses (Geddes
et al., 1999 and Cannon et al., 2002) examined the
association within mixed-sex samples, in which both
excluded the study of Hollister et al. (1996). Assem-
B.J. Insel et al. / Schizophrenia Research 80 (2005) 331–342 333
bling ten studies, Geddes et al. (1999) calculated a
summary odds ratio of 1.40 (95% CI=0.54– 3.63) for
Rh incompatibility and Cannon et al. (2002), aggre-
gating three other studies, derived a summary odds
ratio of 2.00 (95% CI=1.01– 3.96) for the effect of
their Rhesus variable. (The studies evaluated different
Rhesus related factor. Sacker et al. (1995), Kendell et
al. (2000), and Byrne et al. (2000) estimated respec-
tively the effect of maternal Rh negativity, Rh anti-
bodies, and Rh incompatibility).
The epidemiologic literature on schizophrenia has
consistently shown that males and females differ in
terms of premorbid functioning, age at onset, symp-
tomatology, treatment response, and course (Gold-
stein, 1997). Similarly, male infants are more
severely affected by Rh hemolytic disease and kernic-
terus than their female counterparts (Walker and Mol-
lison, 1957; Ulm et al., 1998). Based on these
observations and from the study of Hollister et al.
(1996), we posited that sex of the offspring would
modify the effect of maternal–fetal blood incompati-
bility on the risk of schizophrenia.
We examined the effect of Rh and ABO incompat-
ibility in offspring from the Prenatal Determinants of
Schizophrenia (PDS) study (Susser et al., 2000), a
prospective study drawn from a large birth cohort
born before the availability of anti-Rh (D) prophylax-
is. The PDS study featured comprehensive measures
of prenatal and perinatal exposures, including the
blood types of mother and infant, and implemented
rigorous diagnostic assessments of adult schizophre-
nia outcome in the offspring. In addition, the cohort
had an appreciable number of female offspring, which
enabled us to examine interaction between Rh and
ABO incompatibility and sex of offspring on the risk
of schizophrenia.
2. Methods
2.1. Description of the cohort
The analytic sample was derived from the Prenatal
Determinants of Schizophrenia (PDS); which has
been described in detail elsewhere (Susser et al.,
2000). The PDS study originated from the Child
Health and Development Study, (CHDS), a birth co-
hort implemented to investigate associations between
pregnancy events and offspring development (van den
Berg, 1979; van den Berg et al., 1988). From 1959 to
1967, the CHDS assembled 19044 live births of
women who received prenatal care from Kaiser Per-
manente Plan-Northern California Region (KP). The
cohort was racially, educationally and occupationally
diverse and reflected the population of KP and of the
population of Alameda County during the course of
the study.
The PDS study cohort consisted of the 12094
liveborn offspring who belonged to KP anytime
from January 1, 1981 (the year in which computerized
registries became available) to December 31, 1997.
The cohort was followed for seventeen years and
consequently, the ages of the offspring ranged from
13 to 38 over the course of the PDS study follow-up.
All subjects in the PDS study provided written
informed consent for human investigation. The
study protocol was approved by the Institutional Re-
view Boards of the New York State Psychiatric Insti-
tute and the KP.
2.2. Follow-up
The KP registries provided a record of each off-
spring’s membership status and treatment status over
the 17 year follow-up. Offspring who remained in KP
and those lost to follow-up were similar to one another
on sociodemographic factors, prenatal exposures, and
birth outcomes (Susser et al., 2000). Computerized
psychiatric and medical utilization records were used
to track offspring who were treated for psychiatric
disorders while they were members of the KP.
2.3. Diagnoses
Schizophrenia spectrum disorders were defined as
schizophrenia, schizoaffective disorder, delusional
disorder, psychotic disorder not otherwise specified,
and schizotypal personality disorder (Kendler et al.,
1995). Potential cases among the cohort were identi-
fied by means of computerized record linkage be-
tween CHDS and KP inpatient, outpatient, and
pharmacy registries. Offspring with schizophrenia or
potential schizophrenia spectrum disorders were
screened from ICD-9 registry diagnoses of 295–299
or medical records indicating antipsychotic medica-
tions, followed by psychiatrist review of all records.
B.J. Insel et al. / Schizophrenia Research 80 (2005) 331–342334
Among those who screened positive (N =183), 13
were deceased.
Contact was achieved with 146 (86%) of the sur-
viving 170 potential cases and 107 completed in-
person diagnostic assessment. Clinically experienced
interviewers were trained to administer the Diagnostic
Interview for Genetic Studies (DIGS) (Nurnberger et
al., 1994). Diagnoses were assigned using DSM-IV
criteria by consensus of three psychiatrists based on
review of the written DIGS narrative, the medical
records, and discussions with the interviewers. Over-
all, 71 PDS offspring met the criteria for SSD, 44 as
result of consensus diagnosis following interview and
27 from chart reviews (90% of cases had either
schizophrenia or schizoaffective disorder).
2.4. Assessment and definition of blood
incompatibility
As stated above, a comprehensive interview was
conducted at a woman’s enrollment into the CHDS. In
addition to the maternal interview and medical chart
extraction, blood samples were drawn. The primary
source for the maternal–fetal blood incompatibility
exposure was the assay results derived from the ma-
ternal and newborn blood specimens collected at de-
livery and analyzed at the same laboratory.
Maternal–fetal Rh incompatibility is conventional-
ly defined as a pregnancy in which an Rh-negative
woman has an Rh-positive fetus. However, predicated
on the fact that a first Rh incompatible pregnancy
immunizes the mother but rarely suffers from its
immunologic sequelae and findings of Hollister et
al. (1996), we restricted our definition of Rh exposure
to second and later Rh incompatible pregnancy (re-
ferred to as strict Rh). Due to the lack of data on prior
Rh incompatible pregnancies, the number of past Rh
incompatible pregnancies was approximated by the
number of all past pregnancies. This surrogate mea-
sure is reasonable because among Rh-negative women
whose indexed pregnancy was Rh incompatible, the
probability that a past pregnancy was also Rh incom-
patible is over 70%, assuming that the father of the
indexed pregnancy also fathered at least one previous
pregnancy. (Under the assumption that the frequency
of Rh-positive blood in a sample is 85%, 45% of Rh-
positive persons are homozygous for D and 55% are
heterozygous. If a father is homozygous, all of his
children are Rh-positive and if he is heterozygous,
each child has 50% chance of being Rh-positive. As a
result, 0.45+0.5�0.55=0.725 of offspring are Rh-
positive. A higher frequency of Rh positive yields a
higher proportion of Rh-positive offspring).
ABO incompatibility was defined as a pregnancy
in which a pregnant woman has type O blood and the
fetus has blood type A or B. We did not adjust for the
existence of a prior ABO incompatible pregnancy.
We also combined the two distinct exposures,
strict Rh and ABO incompatibility, into a single
exposure variable (referred to as composite blood
incompatibility). Our rationale was that both types
of maternal–fetal blood incompatibility give rise to
the production of transplacentally transmitted mater-
nal antibodies, which can cause hemolytic disease of
the fetus, our primary hypothesized causal mecha-
nism for schizophrenia.
2.5. Analytic strategy
The construction of the PDS analytic sample is
described in detail elsewhere by Susser et al. (2000).
Since the CHDS birth cohort contained siblings, only
one sibling from each family was randomly selected
to maintain independence of observations in the anal-
yses. Due to the limited number of offspring diag-
nosed with SSD during the course of the PDS study
follow-up, if a sibship contained an affected sibling,
the affected sibling was retained and the unaffected
siblings were excluded. However, if a sibship did not
include an affected sibling, then one unaffected sib-
ling was randomly selected for inclusion into the
sample. This selection process resulted in 7796 off-
spring. In the present analyses, five offspring diag-
nosed with SSD were subsequently excluded; four
who were diagnosed prior to January 1, 1981 (the
start date of the PDS study) and one who had an
affected sibling who was a member of the PDS cohort.
Analyses that included these five subjects resulted in
equivalent findings (Insel, 2003).
Rh and ABO compatibility status were unavailable
respectively for 758 (9.7%) and 998 (12.8%) of the
7791 PDS offspring. Missing blood types were highly
correlated (kappa=0.75), which reduced the possibil-
ity of differential ascertainment of Rh incompatible
compared with ABO incompatible pregnancies. Prior
to the main analysis, we evaluated the impact of
B.J. Insel et al. / Schizophrenia Research 80 (2005) 331–342 335
missing blood type data by examining its correlation
with maternal health conditions and birth outcomes
such as low birth weight and prematurity; no correla-
tions were observed. Thus, it is unlikely that substan-
tial bias would have been introduced by the exclusion
of offspring with missing blood compatibility status.
Different subsamples were constructed for each of
the blood compatibility analyses. The Rh analyses
focused on only second and later born unrelated off-
spring. From among the 7791 PDS offspring, 7033
offspring had identifiable Rh compatibility status. The
exclusion of the 1832 firstborns produced an Rh
analytic sample of 5201 offspring in which 48 were
diagnosed with SSD. The ABO analytic sample
contained 6793 offspring in which 59 offspring were
diagnosed with SSD. To construct the sample for the
composite blood incompatibility analysis, firstborns
were retained since ABO incompatible firstborns
were regarded as exposed due to lack of the necessity
of a prior sensitizing pregnancy to produce hemolytic
disease. We regarded the Rh incompatible firstborns
to be unexposed. However, to control for the possi-
bility that the Rh incompatible firstborns were to some
extent exposed; we inserted an additional exposure
category reflecting Rh incompatible firstborns in the
composite blood analytic models. Among the 7791
offspring, 681 were missing the composite blood
compatibility status, 77 ABO status, and 317 Rh
status; yielding a composite blood sample of 6716
offspring in which 59 offspring were diagnosed with
SSD (refer to Table 1). Nevertheless, we also ran the
analyses on a composite blood incompatible sample
excluding all firstborns regardless of compatibility
status and obtained essentially the same results.
Cox Proportional Hazards regression (Cox and
Oakes, 1984) was used to analyze the data since this
statistical technique takes into account varying dura-
tions of follow-up, while simultaneously adjusting for
Table 1
Distribution of schizophrenia spectrum disorder cases
Analytic sample Total sample
size
Unexposed
offspring
Exposed
offspring
Noncases Cases Noncases Cases
Strict Rh 5201 4808 43 345 5
ABO 6793 5818 49 916 10
Composite blood
incompatibility
6716 5448 44 1209 15
multiple covariates. For offspring diagnosed with
SSD, the onset date of SSD was approximated by
the date of first hospital admission or first outpatient
visit. Thus, the length of the follow-up for offspring
with SSD was quantified in days elapsed from the
beginning of observation until the onset date of SSD.
Analogously, for unaffected offspring, it was calculat-
ed as days since the start of the observation until the
KP membership termination date or until the end of
the PDS study, whichever date came first.
To strengthen the evidence for causality, confound-
ing needs to be excluded via examination of the
association within the context of potentially related
covariates (i.e., those correlated with blood incompat-
ibility and independently related to the diagnosis of
SSD). Based on the uniqueness of the blood incom-
patibility exposure, maternal ethnicity and maternal
age were selected as potential confounders. We also
assessed whether other variables such as maternal
education could modify the results and none had an
appreciable impact. This analytical process was re-
peated for the ABO and the composite blood incom-
patibility samples.
The current tenet of epidemiologic textbooks is
that an additive model is the most appropriate method
to identify biological interaction. Accordingly, we
adapted an approach derived from Darroch (1997)
and Rothman and Greenland (1998) to investigate
the biological interaction between maternal–fetal
blood incompatibility and sex of the offspring using
an additive model. Interaction is supported if the
combined effect of blood incompatibility and sex of
the offspring on the rate of schizophrenia is greater
than the sum of their independent effects. To statisti-
cally assess whether the interaction was significant,
we employed a test devised by Ten Have et al. (2002)
that identified an interaction effect as being significant
if the expected combined rate ratio is less than the
lower limit of the confidence interval of the observed
combined effect.
3. Results
3.1. Rh Sample characteristics
The Rh compatible group contained 5201 offspring and
the Rh incompatible group had 350 offspring. The Rh
Table 2
Sociodemographic characteristics of study sample for the Rh analysis
Preliminary Rh
sample (n =7033)
Final Rh samplea
(n =5201)
Unexposed Rh
samplea (n =4851)
Exposed Rh
samplea (n =350)
n % n % n % n %
Maternal characteristics
Age at delivery
b20 years 385 5.5 135 2.6 128 2.6 7 2.0
20–29 4113 58.5 2765 53.2 2578 53.1 187 53.4
30–39 2202 31.3 1988 38.2 1856 38.3 132 37.7
40+ 333 4.7 313 6.0 289 6.0 24 6.9
Mean age (SD) 27.9 (6.2) 29.2 (6.0) 29.2 (6.0) 29.2 (5.9)
Race/ethnicity
White 3935 56.0 2783 53.5 2547 52.5 236 67.4
Black 2028 28.8 1621 31.2 1536 31.7 85 24.3
Other 1070 15.2 797 15.3 768 15.8 29 8.3
Number of prior pregnancies
None 1832 26.0 0 0.0 0 0.0 0 0.0
One 1690 24.0 1690 32.5 1583 32.6 107 30.6
Two or more 3510 50.0 3511 67.5 3268 67.4 243 69.4
Infant characteristics
Sex
Male 3602 51.2 2643 50.8 2462 50.8 181 51.7
Female 3431 48.8 2558 49.2 2389 49.2 169 48.3
Low birth weight (b=2500 g)
Absent 6657 94.7 4923 94.7 4594 94.7 329 94.0
Present 376 5.3 278 5.3 257 5.3 21 6.0
Mean birth weight in oz (SD) 117.1 (18.2) 117.9 (18.4) 117.9 (18.4) 117.9 (18.6)
Prematurity (b37 weeks)
Absent 6522 92.7 4809 92.5 4484 92.4 325 92.9
Present 511 7.3 392 7.5 367 7.6 25 7.1
Mean gestational age (SD) 278.9 (14.2) 278.4 (14.3) 278.4 (14.2) 278.8 (15.0)
ABO status
ABO compatible 5790 82.3 5492 82.2 3990 82.3 298 85.1
ABO incompatible 926 13.2 874 13.1 654 13.5 52 14.9
Missing 317 4.5 317 4.7 207 4.3 0 0
a Excludes all firstborns.
B.J. Insel et al. / Schizophrenia Research 80 (2005) 331–342336
incompatible offspring in the PDS study were similar to the
Rh compatible offspring, except that Rh incompatible off-
spring were more likely to be offspring of Caucasian women
than offspring of African-American and other minority eth-
nic women (see Table 2).
Table 3
Rate ratios of schizophrenia spectrum disorder and strict Rh incompatibil
SSD count (exposed) All offspring (n =5201)
48 (5)
RR 95% CI
Unadjusted 1.63 0.65–4.12
Adjusteda 1.80 0.71–4.58
a Adjusted for maternal age and maternal ethnicity.
3.2. The effects of Rh incompatibility
The unadjusted rate ratio for strict Rh incompatibility
and SSD was 1.63 (95% CI=0.65–4.12) (Table 3). Adjust-
ing for maternal age and ethnicity modestly increased the
ity
Male offspring (n =3602) Female offspring (n =2558)
30 (4) 18 (1)
RR 95% CI RR 95% CI
2.02 0.71–5.79 0.90 0.12–6.74
2.37 0.82–6.86 0.93 0.12–7.01
Table 4
Rate ratios of schizophrenia spectrum disorder and strict Rh incom-
patibility: effect modification for sex of offspring
Study sample (n =5201)
Combined effect of Adjusted RRa 95% CI
Being male Strict Rh incompatible
No No 1.00
Yes No 1.51 0.82–2.78
No Yes 0.94 0.13–7.11
Yes Yes 3.50 1.17–10.49
a Adjusted for maternal age and maternal ethnicity.
B.J. Insel et al. / Schizophrenia Research 80 (2005) 331–342 337
rate ratio to 1.80 (95% CI=0.71–4.58). Stratification on sex
produced an adjusted rate ratio of 2.37 for strict Rh incom-
patibility among the male offspring, but did not achieve
statistical significance (95% CI=0.82–6.86). In contrast,
no elevated effect was observed for the female offspring.
Although not shown in the tables, for comparative purpose
with the existing literature, we conducted analyses using an
alternative definition for the Rh incompatibility exposure,
which considered all Rh incompatibles as exposed regard-
less of birth order. This resulted in slightly lower adjusted
RR=1.67 (95% CI=0.71–3.88) than for the strictly defined
Rh incompatibility exposure. Analogous to our prior finding
for the male offspring, we derived an adjusted rate ratio of
2.33 (95% CI=0.90–6.02).
3.3. Interaction between Rh incompatibility and sex of
offspring
To examine interaction between sex of offspring and Rh
incompatibility, we computed the combined effects of being
male and being strictly Rh incompatible on the incidence of
SSD, as shown in Table 4. The adjusted rate ratio for the
independent effect of male sex alone (i.e., the effect among
Rh compatibles) was 1.51 (95% CI=0.82–2.78) and for the
independent effect of strict Rh incompatibility (i.e., the
effect among women) was 0.94 (95% CI=0.13–7.11). The
adjusted rate ratio for the observed combined effect of male
sex and strict Rh incompatibility was 3.50 (95% CI=1.17–
10.49). We calculated the expected effect for the scenario of
Table 5
Rate ratios of schizophrenia spectrum disorder and ABO incompatibility
SSD count (exposed) All offspring (n =6793)
59 (10)
RR 95% CI
Unadjusted 1.31 0.66–2.59
Adjusteda 1.28 0.65–2.53
a Adjusted for maternal age and maternal ethnicity.
no interaction within an additive model, which is the sum of
the two independent effects minus the reference value of 1
(1.51+0.94�1.00=1.45). The observed combined effect of
3.50 was 2.4 times the expected combined effect of 1.45,
suggesting the presence of additive interaction. Nonetheless,
the data did not satisfy the criterion of Ten Have et al.
(2002) for a significant interaction effect since the expected
combined effect for the absence of interaction was greater
than 1.17, the lower limit of the 95% confidence interval of
the observed combined effect.
3.4. The effects of ABO incompatibility
The sociodemographic characteristics for the ABO sam-
ple were almost identical to those of the Rh sample. The rate
ratios for ABO incompatibility were consistently lower than
the corresponding rate ratios for Rh incompatibility (see
Table 5). Among the male offspring, the adjusted rate
ratio for ABO incompatibility was 1.68 (95% CI=0.76–
3.70). Analogous to the female Rh incompatibility results,
no elevation of risk for ABO incompatibility for the female
offspring was ascertained.
We also examined interaction between sex of offspring
and ABO incompatibility. The adjusted rate ratio for the
independent effect of male sex alone (i.e., the effect among
ABO compatibles) was 1.33 and for the independent effect
of ABO incompatibility (i.e., the effect among women) was
0.62. The adjusted rate ratio for the observed combined
effect of male sex and ABO incompatibility was 2.29
(95% CI= 1.01–5.17) and the expected combined effect in
absence of interaction was 0.95. These results satisfied the
criterion of Ten Have et al. (2002) for statistical significant
interaction effect since the expected combined effect in
absence of interaction was less than the lower limit of the
95% confidence interval for the observed combined effect.
3.5. The effect of composite blood incompatibility
We next examined whether having the composite (either
Rh or ABO) blood incompatibility increased the risk for
SSD. The adjusted rate ratio for composite blood incompat-
ibility was 1.57 (95% CI=0.87–2.82), as shown in Table 6.
Male offspring (n =3467) Female offspring (n =3326)
36 (8) 23 (2)
RR 95% CI RR 95% CI
1.83 0.83–4.02 0.61 0.14–2.61
1.68 0.76–3.70 0.62 0.14–2.63
Table 6
Rate ratios of schizophrenia spectrum disorder and composite blood (strict Rh and/or ABO) incompatibility
SSD count (exposed) All offspring (n =6716) Male offspring (n =3425) Female offspring (n =3291)
59 (15) 36 (12) 23 (3)
RR 95% CI RR 95% CI RR 95% CI
Unadjusted 1.55 0.86–2.79 2.24 1.12–4.48 0.70 0.21–2.34
Adj. for covariatesa 1.57 0.87–2.82 2.22 1.10–4.47 0.70 0.21–2.35
a Adjusted for maternal age, maternal ethnicity, and Rh incompatible firstborns.
B.J. Insel et al. / Schizophrenia Research 80 (2005) 331–342338
Stratification on sex of offspring revealed a statistically
significant two-fold rate ratio for blood incompatibility
among the male offspring, (Adjusted RR=2.22, 95%
CI=1.10–4.47).
We also evaluated the presence of interaction between
sex of offspring and composite blood incompatibility. The
corresponding adjusted rate ratio for the independent effect
of male sex was 1.20, for the independent effect of blood
incompatibility was 0.71, and the adjusted rate ratio for their
observed combined effect was 2.67 (95% CI=1.30–5.48)
(see Table 7). The expected combined effect in the absence
of interaction was 0.91. The observed combined effect was
almost three times the corresponding expected combined
effect and met the criterion of Ten Have et al. (2002) for
statistical significance.
7
Material-Fetal Blood Incompatibility RateRatios (RRadj) for SSD
io
4. Discussion
The study has extended the line of evidence that Rh
incompatibility may be a risk factor for schizophrenia.
We found that strict Rh incompatibility was associated
with a two-fold elevated rate of schizophrenia spectrum
disorder and the effect was limited to male offspring.
We also explored the effect of ABO incompatibility and
observed a weaker effect, also limited to male off-
spring. When Rh and ABO incompatibility were ana-
Table 7
Rate ratios of schizophrenia spectrum disorder and composite blood
(strict Rh and/or ABO) incompatibility: effect modification for sex
of offspring
Study sample (n =6716)
Combined effect of Adjusted RRa 95% CI
Being male Incompatible
No No 1.00
Yes No 1.20 0.66–2.17
No Yes 0.71 0.21–2.40
Yes Yes 2.67 1.30–5.48
a Adjusted for maternal age and maternal ethnicity.
lyzed separately, statistical significance was not
achieved. However, the composite Rh and/or ABO
incompatibility exposure produced a significant, great-
er than two-fold rate of schizophrenia among the male
offspring. The findings are summarized in Fig. 1.
Schizophrenia is currently regarded as a complex
disorder that arises from heterogeneous pathways in-
volving interactions between many different genes and
environmental factors, each with varying causal effec-
tivity. Thus, it may not be surprising that a modest
effect size was found for maternal–fetal blood incom-
patibility. The magnitude does not preclude a potential
causal association; rather the results are compatible
with etiological heterogeneity within the disorder.
Moreover, the effect size may reflect the prevalence
within the study population of other necessary factors
that operate in conjunction with maternal–fetal blood
incompatibility to cause schizophrenia.
We considered whether the findings were internally
valid or were the result of ascertainment bias, uncon-
trolled confounding, or chance. The prospective de-
1.68 2.220.93 0.65 0.702.370
1
2
3
4
5
6
Rh ABO Rh/ABO
Blood Incompatibility
Ad
just
ed R
ate
Rat
RRadj
Fig. 1. Gender differences in the effects of blood incompatibility for
SSD. Rate ratios and 95% confidence intervals are adjusted for mater-
nal age and ethnicity. Dashed line represents the null value of 1. Only
the male adjusted 95% CI for Rh/ABO is above the null value line.
B.J. Insel et al. / Schizophrenia Research 80 (2005) 331–342 339
sign of the PDS study makes it unlikely that the
findings could be accounted for by ascertainment
bias. The offspring were identified prior to their
birth from a well-characterized population of pregnant
women. The incidence of schizophrenia spectrum
disorder among the PDS cohort was ascertained me-
thodically and independently of prenatal and birth
outcomes. Furthermore, maternal and fetal exposure
data were collected uniformly and consistently at each
woman’s enrollment in the study, throughout the preg-
nancy, and during labor and delivery.
The association is also unlikely to be explained by
uncontrolled or residual confounding. We adjusted for
potential confounders, of which only two covariates,
maternal ethnicity and maternal age at delivery, had a
slight impact on the rate ratios. In addition, maternal–
fetal blood incompatibility is an immutable, biological
phenomenon that is determined at fertilization. Thus,
maternal behavior or lifestyle before, during, or after a
pregnancy cannot influence maternal–fetal blood in-
compatibility. Moreover, as predicted, adjusting for
other maternal and infant characteristics that were
observed in past studies to be risk factors for schizo-
phrenia did not alter the effect estimates for maternal–
fetal blood incompatibility.
Lastly, we considered the contribution of chance.
We observed elevated but not statistically significant
rate ratios for Rh and ABO incompatibility among the
male offspring. However, the associations were sup-
ported by the fact that the hypotheses were precisely
and a priori formulated. In addition, biological plau-
sibility, the Rh incompatibility findings of Hollister et
al. (1996) and Palmer et al. (2002), and our statisti-
cally significant findings among the males for the
composite blood incompatibility exposure strengthen
the validity of the associations. Nevertheless, chance
remains a possible explanation.
We also contemplated whether the lack of an ele-
vated risk for maternal–fetal blood incompatibility for
the female offspring was real or reflected the inade-
quacy of the study power. During the PDS study
follow-up, substantially more males than females di-
agnosed with SSD were ascertained. Using the multi-
national WHO-Determinants of Outcome-Study,
Hambrecht et al. (1992) estimated that approximately
90% of men are less than age 40 at diagnosis for
schizophrenia, while only 70% of women are less
than age 40 at diagnosis. The lack of power within
the PDS study sample may reflect the fact that females
are often diagnosed with schizophrenia three to five
years later than their male counterparts. Although we
found a later average age at diagnosis for SSD among
the Rh incompatible group compared with the Rh
compatible group in the PDS cohort, in which the
age difference was 2.7 years, the independent sample
t-test was not statistically significant ( p-value=0.24).
Thus, it is unlikely that Rh incompatibility was asso-
ciated with a later onset of schizophrenia in women
and limited our ability to detect an elevated effect for
the females. We also conducted sensitivity analyses
and the resultant female rate ratios for Rh incompat-
ibility (ranged from 0.90 to 1.27) never approached
the two-fold effect observed for the male offspring
and never achieved marginal statistical significance
(results available on request from first author).
Biological and empirical observations corroborate
the observed sex differences in our findings, which
revealed that only males showed the effect of blood
incompatibility for schizophrenia. Moreover, sex dif-
ferences have been consistently observed in the inci-
dence and course of numerous types of psychopathol-
ogy (Rutter et al., 2003). These findings indicate that
biological mechanisms contributing to causal hetero-
geneity may underlie the sex differences. Divergent
results may reflect the consequences of sexual dimor-
phism in brain structure and function and the interplay
of estrogen. A slower rate of cerebral development in
males than in females (Seeman, 1989) may augment
the susceptibility of the male brain to environmental
and perinatal insults. Male infants are also more se-
verely affected by maternal Rh-D immunization, in-
cluding kernicterus, than female infants (Ulm et al.,
1998; Walker and Mollison, 1957). In vitro and ani-
mal observations provide additional, although specu-
lative, support for the sex differential effect of
maternal–fetal blood incompatibility on schizophre-
nia. In cultured neurons, estrogen attenuated the neu-
rotoxic effects of oxidative damage from prolonged
exposure to hemoglobin (Sawada et al., 1998; Regan
and Guo, 1997). Similarly, in vivo findings showed
that male gerbils were more susceptible to ischemic
cortical and hippocampal injury than female gerbils
(Hall et al., 1991). To summarize, the cumulative
evidence from basic science and clinical research
suggests that male fetuses and infants may be more
vulnerable to certain early adverse environmental
B.J. Insel et al. / Schizophrenia Research 80 (2005) 331–342340
exposures of maternal–fetal blood incompatibility and
less able also to compensate for aberrant neurodeve-
lopment than their female counterparts.
This is the first study to explore the effect of ABO
incompatibility in relation to schizophrenia. Although
both Rh and ABO blood incompatibility induce trans-
placentally transmitted maternal antibodies, they dif-
fer with respect to both biological sequelae and
clinical significance. An ABO incompatible pregnan-
cy occurs more frequently than an Rh incompatible
pregnancy, but is rarely accompanied by hemolytic
disease of the fetus. When ABO hemolytic disease
does occur, it is relatively mild. In contrast, to Rh
hemolytic disease, ABO hemolytic disease can also
arise in firstborn infants (Mollison et al., 1993; Wal-
dron and de Alarcon, 1999). The ABO results are
preliminary, but suggest that further research on
ABO incompatibility as a possible risk factor for
schizophrenia also warrants investigation.
4.1. Limitations of the study
The study has some limitations that deserve con-
sideration. The first limitation is the modest number of
offspring diagnosed with SSD during the course of the
PDS follow-up.
A second limitation is that approximately 10% of
the mother–infant dyads were missing blood incom-
patibility data. However, the likelihood of a bias
introduced from the exclusion of offspring with miss-
ing blood types is slight. Moreover, most blood col-
lection occurred routinely during delivery and missing
blood types were not correlated with prenatal factors
or birth outcomes.
A third limitation is the absence of information on
whether a mother experienced a prior Rh incompati-
ble pregnancy. As a result, we utilized the number of
all prior pregnancies to estimate the number of prior
Rh incompatible pregnancies. Based on the distribu-
tion of the Rh-negative frequency in the sample,
approximately 23% of presumed second and later
Rh incompatible offspring may actually have been
Rh incompatible firstborns since a prior pregnancy
may actually have been Rh compatible. Thus, our
operationalized measure may have overestimated the
number of second and later Rh incompatible pregnan-
cies and would bias the results toward the null by
misclassification of some unexposed offspring (first-
born Rh incompatible) as exposed offspring (second
and later Rh incompatible).
The lack of study power did not allow us to further
examine whether firstborn Rh incompatible offspring
are at elevated risk for schizophrenia since the PDS
study cohort contained only one Rh incompatible
firstborn offspring diagnosed with SSD. Past findings
concerning the risk for later schizophrenia among
firstborn Rh incompatible offspring were inconclu-
sive, including the findings of Hollister et al. (1996),
which observed an elevated but statistically insignif-
icant risk. The data of Kraft et al. (2004) equally
supported two scenarios, one in which all Rh incom-
patible offspring are at increased risk and the second,
in which only subsequent Rh incompatible offspring
are at increased risk for schizophrenia. The biological
plausibility is less robust for firstborn Rh incompati-
ble offspring compared with second and later Rh
incompatible offspring. Generally, during a first Rh
incompatible pregnancy, an appreciable maternal im-
munologic response and ensuing hemolytic disease of
the fetus does not occur. Nevertheless, the association
between firstborn Rh incompatible offspring and
schizophrenia deserves investigation.
Another limitation is that we lacked information on
both post-natal bilirubin levels and the incidence of
hemolytic disease in the infants. At some later point,
we hope to obtain access to this data. Thus, our
present findings are applicable only to Rh and ABO
blood incompatibility. If hemolytic disease goes un-
treated, it may progress to neonatal kernicterus; irre-
versible brain damage accompanied by long-term
neurological deficits, and is likely to be associated
with a substantial risk of later schizophrenia, if the
infant survives. Presumably, hemolytic disease of the
fetus would confer a greater risk for schizophrenia
than blood incompatibility alone and could also pro-
duce an elevated effect for female offspring. Conse-
quently, the association between hemolytic disease of
the fetus and newborn and schizophrenia deserves
future consideration.
A final limitation concerns the inherent character-
istics of the PDS study population. The PDS offspring
were born between 1959 and 1967 to women who
were enrolled in a prepaid health care plan. Although
anti-Rh (D) prophylaxis was not available until 1968
to prevent sensitization in Rh-negative women, mul-
tiple methods of exchange transfusion and photother-
B.J. Insel et al. / Schizophrenia Research 80 (2005) 331–342 341
apy were in use to treat infants born with hemolytic
disease. As such, the Rh incompatible infants from the
PDS study cohort had access to these treatment mo-
dalities, which may have controlled the severity of the
neurologic sequelae associated with hemolytic dis-
ease. However, women living in low income countries
have limited access to perinatal health care, particu-
larly anti-Rh (D) prophylaxis, may experience more
unfavorable outcomes associated with Rh and ABO
incompatibility than our study population. Thus, our
results may have underestimated the association be-
tween maternal–fetal blood incompatibility and
schizophrenia.
5. Conclusion
Maternal–fetal blood incompatibility is a unique
prenatal exposure that incorporates genetics, the dis-
cordant maternal and fetal blood genotypes, and the
intra-uterine environment, ensuing from the maternal
immunologic responsiveness and fetal compensatory
mechanisms. We observed that the most pronounced
effect on later schizophrenia was exhibited among
male offspring who were products of second and
later Rh incompatible pregnancies, but the findings
raise the possibility that other blood group incompat-
ibilities may be etiologically important in schizophre-
nia. Our maternal–fetal blood incompatibility results
are congruent with the epidemiologic literature of
schizophrenia and consistent with past studies that
showed that immunological disturbances of the fetal
environment, particularly the induction of maternal
antibodies, might lead to aberrant fetal neurodevelop-
ment, manifesting in later psychosis in the offspring.
Likewise, the observed differential effect by sex of
offspring for maternal–fetal blood incompatibility ex-
posure on the risk of schizophrenia mirrors the sex
differences observed for hemolytic disease of the fetus
and newborn.
Acknowledgements
This manuscript was supported by the following
grants: NIMH 1R01MH 63264-01A1 (A.S.B.), NIMH
1K02MH65422-01 (A.S.B.), NICHD NO1-HD-1-3334
(B. Cohn), NICHD NO1-HD-6-3258 (B. Cohn).
We also wish to acknowledge the following indi-
viduals for their contributions to this work: Barbara
van den Berg, M.D., Barbara Cohn, Ph.D., and the
late Jacob Yerushalmy, M.D. We also wish to thank
the Lieber Center for Schizophrenia Research, the
National Institute for Child Health and Development,
and the Public Health Institute, Berkeley, CA.
References
Adinolfi, M., 1985. The development of the human blood–CSF–
brain barrier. Dev. Med. Child Neurol. 27, 532–537.
Bowman, J.M., 1999. Hemolytic disease (Erythroblastosis fetalis).
In: Creasy, R.K., Resnick, R. (Eds.), Maternal–Fetal Medicine,
4th ed. WB Saunders Company, Philadelphia, pp. 736–767.
Byrne, M., Browne, R., Mulryan, N., Scully, A., Morris, M.,
Kinsella, A., Takei, N., McNeil, T., Walsh, D., O’Callaghan,
E., 2000. Labour and delivery complications and schizophrenia.
Case-control study using contemporaneous labour ward records.
Br. J. Psychiatry 176, 531–536.
Cannon, M., Jones, P.B., Murray, R.M., 2002. Obstetric complica-
tions and schizophrenia: historical and meta-analytic review.
Am. J. Psychiatry 159, 1080–1092.
Cashore, W.J., 1990. The neurotoxicity of bilirubin. Clin. Perinatol.
17, 437–447.
Cox, D.R., Oakes, D., 1984. Analysis of Survival Data. New York
Chapman and Hall Ltd., London.
Darroch, J., 1997. Biologic synergism and parallelism. Am. J.
Epidemiol. 145, 661–668.
Geddes, J.R., Verdoux, H., Takei, N., Lawrie, S.M., Bovet, P.,
Eagles, J.M., Heun, R., McCreadie, R.G., McNeil, T.F., O’Cal-
laghan, E., Stober, G., Willinger, U., Murray, R.M., 1999.
Schizophrenia and complications of pregnancy and labor: an
individual patient data meta-analysis. Schizophr. Bull. 25,
413–423.
Goldstein, J.M., 1997. Sex differences in schizophrenia: epidemi-
ology, genetics and the brain. Int. Rev. Psychiatry 9, 399–408.
Hall, E.D., Pazara, K.E., Linseman, K.L., 1991. Sex differences in
postischemic neuronal necrosis in gerbils. J. Cereb. Blood Flow
Metab. 11, 292–298.
Hambrecht, M., Maurer, K., Hafner, H., Sartorius, N., 1992. Trans-
national stability of gender differences in schizophrenia? An
analysis based on the WHO study on determinants of outcome
of severe mental disorders. Eur. Arch. Psychiatry Clin. Neu-
rosci. 242, 6–12.
Hollister, J.M., Laing, P., Mednick, S.A., 1996. Rhesus incompat-
ibility as a risk factor for schizophrenia in male adults. Arch.
Gen. Psychiatry 53, 19–24.
Insel, B.J., 2003. Maternal-fetal blood incompatibility and the risk
of schizophrenia in offspring. [dissertation]. New York. Mail-
man School of Public Health of Columbia University.
Kendell, R.E., McInneny, K., Juszczak, E., Bain, M., 2000. Obstet-
ric complications and schizophrenia. Two case-control studies
based on structured obstetric records. Br. J. Psychiatry 17,
516–522.
B.J. Insel et al. / Schizophrenia Research 80 (2005) 331–342342
Kendler, K.S., Neale, M.C., Walsh, D., 1995. Evaluating the spec-
trum concept of schizophrenia in the Roscommon family study.
Am. J. Psychiatry 152, 749–754.
Kraft, P., Palmer, C.G., Woodward, A.J., Turunen, J.A., Minassian,
S., Paunio, T., Lonnqvist, J., Peltonen, L., Sinsheimer, J.S.,
2004. RHD maternal–fetal genotype incompatibility and schizo-
phrenia: extending the MFG test to include multiple siblings and
birth order. Eur. J. Hum. Genet. 12, 192–198.
Laing, P., Knight, J., Wright, P., Irving, W., 1995. Disruption of fetal
brain development by maternal antibodies as an etiological
factor in schizophrenia. In: Mednick, S.A., Hollister, J.M.
(Eds.), Neural Development and Schizophrenia: Theory and
Research. Plenum, New York, pp. 215–245.
McGrath, J.J., Feron, F.P., Burne, T.H., Mackay-Sim, A., Eyles,
D.W., 2003. The neurodevelopmental hypothesis of schizophre-
nia: a review of recent developments. Ann. Med. 35, 86–93.
Mollison, P.L., Engelfriet, C.P., Contreras, M., 1993. Haemolytic
disease of the fetus and the newborn. Blood Transfusion in
Clinical Medicine, 9th ed. Blackwell Scientific, Oxford, Eng-
land, pp. 543–591.
Murray, R.M., Fearon, P., 1999. The developmental ’risk factor’
model of schizophrenia. J. Psychiatr. Res. 33, 497–499.
Nurnberger Jr., J.I., Blehar, M.C., Kaufmann, C.A., York-Cooler,
C., Simpson, S.G., Harkavy-Friedman, J., Severe, J.B., Mala-
spina, D., Reich, T., 1994. Diagnostic interview for genetic
studies. Rationale, unique features, and training. NIMH genetics
initiative. Arch. Gen. Psychiatry 51, 849–859.
Ozolek, J.A., Watchko, J.F., Mimouni, F., 1994. Prevalence and lack
of clinical significance of blood group incompatibility in
mothers with blood type A or B. J. Pediatr. 125, 87–91.
Palmer, C.G., Turunen, J.A., Sinsheimer, J.S., Minassian, S., Pau-
nio, T., Lonnqvist, J., Peltonen, L., Woodward, J.A., 2002. RHD
maternal–fetal genotype incompatibility increases schizophrenia
susceptibility. Am. J. Hum. Genet. 71, 312–319.
Rawson, A.J., Abelson, N.M., 1960. Studies of blood group anti-
bodies. IV. Physicochemical differences between isoanti-A, B
and isoanti-A or isoanti-B. J. Immunol. 85, 640–647.
Regan, R.F., Guo, Y., 1997. Estrogens attenuate neuronal injury due
to hemoglobin, chemical hypoxia, and excitatory amino acids in
murine cortical cultures. Brain Res. 764, 133–140.
Rothman, K.J., Greenland, S., 1998. Concepts of interaction. In:
Rothman, K.J., Greenland, S. (Eds.), Modern Epidemiology,
2nd ed. Lippincott Williams and Wilkins, Philadelphia, PA,
pp. 329–342.
Rutter, M., Caspi, A., Moffitt, T.E., 2003. Using sex differences
in psychopathology to study causal mechanisms: unifying
issues and research strategies. J. Child Psychol. Psychiatry
44, 1092–1115.
Sacker, A., Done, D.J., Crow, T.J., Golding, J., 1995. Antecedents
of schizophrenia and affective illness. Obstetric complications.
Br. J. Psychiatry 166, 734–741.
Sawada, H., Ibi, M., Kihara, T., Urushitani, M., Akaike, A., Shi-
mohama, S., 1998. Estradiol protects mesencephalic dopaminer-
gic neurons from oxidative stress-induced neuronal death.
J. Neurosci. Res. 54, 707–719.
Seeman, M.V., 1989. Prenatal gonadal hormones and schizophrenia
in men and women. Psychiatr. J. Univ. Ott. 14, 473–475.
Susser, E., Brown, A.S., Gorman, J.M., 1999. Prenatal Exposures in
Schizophrenia. American Psychiatric Press, Washington, DC.
Susser, E.S., Schaefer, C.A., Brown, A.S., Begg, M.D., Wyatt, R.J.,
2000. The design of the prenatal determinants of schizophrenia
study. Schizophr. Bull. 26, 257–273.
Ten Have, M., Vollebergh, W., Bijl, R., Ormel, J., 2002. Combined
effect of mental disorder and low social support on care service
for mental health problems in the Dutch general population.
Psychol. Med. 32, 311–323.
Ulm, B., Ulm, M.R., Panzer, S., 1998. Fetal sex and hemolytic
disease from maternal red-cell alloimmunization. N. Engl. J.
Med. 338, 1699–1700.
van den Berg, B.J., 1979. The California Child Health and Devel-
opment Studies: twenty years of research. World Health Stat. Q.
32, 269–286.
van den Berg, B.J., Christianson, R.E., Oechsli, F.W., 1988. The
California Child Health and Development Studies of the School
of Public Health, University of California at Berkeley. Paediatr.
Perinat. Epidemiol. 2, 265–282.
van Gent, T., Heijnen, C.J., Treffers, P.D., 1997. Autism and the
immune system. J. Child Psychol. Psychiatry 38, 337–349.
Waldron, P., de Alarcon, P., 1999. ABO hemolytic disease of the
newborn: a unique constellation of findings in siblings and
review of protective mechanisms in the fetal–maternal system.
Am. J. Perinatol. 16, 391–398.
Walker, W., Mollison, P.L., 1957. Haemolytic disease of the
newborns: deaths in England and Wales during 1953 and
1955. Lancet 1, 1309–1314.
Weinberger, D.R., 1987. Implications of normal brain development
for the pathogenesis of schizophrenia. Arch. Gen. Psychiatry 44,
660–669.