We examined Rhesus D incompatibility as a sex-dependent risk factor for schizophrenia using a candidate gene approach in a large sample and by performing a meta-analysis of the sex-specific incompatibility effects produced by datasets from three studies. The candidate gene approach yielded several lines of evidence to support a male incompatibility effect. First, we rejected the null hypothesis of no male incompatibility effect when comparing a model that estimated the sex-specific incompatibility effects to a model that constrained the male incompatibility effect to its null value; whereas we failed to reject the null hypothesis of no female incompatibility effect with an analogous model comparison. Second, we demonstrated that a model that constrains the female effect to one but allows for a male effect had the best fit by AIC of the models we examined, further supporting that Rhesus D incompatibility is a risk factor for schizophrenia only in males. In our sample, the relative risk of schizophrenia in male Rhesus D incompatible pregnancies was greater than for female incompatible pregnancies (1.44 and 1.12, respectively). However, when we compared a model with these estimates of the male and female incompatibility effects to a model that constrained the sex-specific effects to be equal, we found that the two models were not statistically different, thus failing to provide additional support for the hypothesis that the incompatibility effect is exclusively for male offspring.
By combining three datasets in the meta-analysis, we increased the sample size for affected males to 358 and affected females to 226 (a 12–18% increase from the Finnish sample). The results of our meta-analysis yielded the strongest evidence thus far that the incompatibility effect is limited to male offspring. The point estimate of the pooled female incompatibility effect is not only smaller than that for males [μfemale = 1.07, 90% CI 0.72–1.58; μmale = 1.64, 90% CI 1.25–2.17], it also does not fall within the 90% CI for the male incompatibility effect, and it is contained within a 90% CI that includes the null value. Furthermore, the 90% CI for the male incompatibility effect does not cover the null value of one and the effect estimate for males does not fall within the 90% CI for the female effect.
The results of our candidate gene analysis and meta-analysis suggest that the effect of Rhesus D incompatibility on the development of schizophrenia is limited to male offspring. However, because the sample size for females was ~35% less than that of the males, our study may have lacked sufficient statistical power to detect a small female effect. On the other hand, the estimated pooled relative risk for females in the meta-analysis is 1.07, which is very small in magnitude, leaving in question its substantive meaningevenif found to be statistically significant in a larger sample. A null or very small female effect relative to males is an important finding that allows for hypothesis generation and testing to determine why schizophrenia effects of Rhesus D incompatibility are so much greater for male offspring compared to female offspring.
One hypothesis is that male fetuses are at greater risk because clinical manifestations of Rhesus D incompatibility are much more likely to occur in pregnancies with male fetuses than with female fetuses. Although there is some literature that suggests that Rhesus D negative mothers are more likely to be immunized
by male fetuses than female fetuses (Renkonen and Timonen, 1967
; Woodrow and Donohoe, 1968
), there is no evidence that its related condition of Rhesus D incompatibility HDN is more likely to occur in subsequent pregnancies with male fetuses than with female fetuses. Hence, this is unlikely to explain male–female differences in the incompatibility effect.
A related hypothesis is that clinical manifestations of Rhesus D incompatibility must surpass a threshold of severity before they increase risk for schizophrenia, and male fetuses are more likely to experience sequelae that surpass this threshold than female fetuses. This hypothesis predicts that the clinical manifestations of Rhesus D incompatibility would be more severe in pregnancies with male fetuses than with female fetuses. Evidence to support this hypothesis comes from Ulm et al. (1999)
,where they found that compared to incompatible female fetuses, incompatible male fetuses had a significantly lower relative hemoglobin and hematocrit at their initial intravenous transfusion, required more intravenous transfusions, and were more likely to develop hydrops. In addition, more male fetuses than female fetuses died in utero or during the neonatal period (Ulm et al., 1999
). An implication of this hypothesis is that specific schizophrenia-effects of Rhesus D incompatibility (i.e., hypoxia, hyperbilirubinemia) can affect female fetuses but that females are less likely to surpassthe threshold compared to male fetuses.
A third hypothesis is that the timing of manifestations of Rhesus D incompatibility during gestation coincides with sex-specific maturational events such that the developing male brain is more vulnerable to effects of Rhesus D incompatibility than the developing female brain. Research supports differences in brain maturational rates in males and females (Cahill, 2006
), where the pace of cerebral development is slower in males than females (Castle and Murray, 1991
). Hence, the slower-maturing brains of males may be vulnerable to environmental insults for a longer period of time than the more quickly maturing brains of females (Castle and Murray, 1991
). Placental transfer of maternal anti-D antibodies has been observed as early as the 18th week of gestation (Adinolfi, 1985
) and evidence of hemolytic disease in the fetus has been demonstrated in the second trimester (Mollison, 1993
), providing a timeframe for considering the effects of Rhesus D incompatibility. Of note, male fetuses have been found to require intravenous transfusion at an earlier gestational age than female fetuses (male median gestation age 24.5 weeks, range 18–35 weeks; female median gestational age 31.0 weeks, range 22–36 weeks) (Ulm et al., 1999
), suggesting the possibility of a sex-dependent vulnerability to the timing of Rhesus D incompatible effects, with male fetuses experiencing an earlier and longer window of vulnerability during neurodevelopment than female fetuses. An implication of this hypothesis is that males and females may be equally vulnerable to the specific effects produced by Rhesus D incompatibility, but that these effects must occur at sex-dependent times during development. If this is the case, it would mean that female fetuses may be at increased risk for schizophrenia when subject to obstetric complications that produce the same downstream effects as HDN, e.g., hypoxia, hyperbilirubinemia, but that these effects must occur earlier in gestation to increase risk for schizophrenia in females.
A fourth hypothesis is that male fetuses are at greater risk because male developing brains are more vulnerable to one or both specific effects of the clinical manifestations of Rhesus D incompatibility, i.e., hypoxia, hyperbilirubinemia, than female developing brains. Data bearing on this hypothesis are conflicting with some studies finding higher rates of hypoxic-related obstetric complications in males with schizophrenia (Hultman et al., 1999
), and others finding higher rates in females with schizophrenia than in males with schizophrenia (Goldstein et al., 2000
Our findings suggest that the effects of Rhesus D incompatibility, and hence, adverse prenatal environments, on the development of schizophrenia may depend on offspring sex. To further bolster this conclusion, one would expect to observe an increased risk among 2nd or later born incompatible males, but not for females. However, because our datasets were not conducive for jointly analyzing the effect of prior Rhesus D incompatibility exposure and offspring sex in either the candidate gene or meta-analysis approach (due to the relatively small sample sizes of incompatible affected offspring and other study design limitations), we are unable to provide this additional piece of evidence. Nevertheless, our sex-dependent finding is particularly interesting in light of recent work demonstrating that HLA-B matching as a schizophrenia risk factor is limited to females (Palmer et al., 2006
). Not only do these findings suggest that sex differences during fetal neurodevelopment should be investigated to fully elucidate the etiology of schizophrenia, but they allow investigators to generate several testable hypotheses regarding the effects of the prenatal environment on schizophrenia risk which may lead to further understanding of the basic processes underlying schizophrenia and preventative strategies as well.