Obstetric Complications and the Risk of Schizophrenia:  A Longitudinal Study of a National Birth Cohort FREE

THE ORIGIN of schizophrenia is largely unknown. A vulnerability originating from disturbances in the central nervous system (CNS) during the embryonal, fetal, or neonatal period may possibly explain some of the reason for the occurrence of schizophrenia.¹

Numerous epidemiological studies have been published on the association between obstetric complications and schizophrenia. In a meta-analysis² of 18 studies, an overall odds ratio of 2.03 (95% confidence interval [CI], 1.6-2.4) was found for schizophrenia following obstetric complications of any kind. Interestingly, the only 2 historical cohort studies^3- 4 in that analysis failed to find a significant association. However, the number of schizophrenic or psychotic patients was small in those studies.

Many studies have used the summary scale of obstetric complications by Lewis et al,⁵ which is unspecific and thus gives little clue to the underlying cause. McNeil et al⁶ showed that the kind of scale used has consequences for the results obtained, with higher risk estimates if a weighted scale is used.⁷

Several studies found 1 or 2 (rarely 3) significant risk factors that are seldom confirmed in other studies. Suggested risk factors are preeclampsia,^8- 10 small head circumference,^11- 12 low birth weight,^13- 17 Rh incompatibility,¹⁸ fetal distress,¹⁹ body weight heavy for length (weight/length, ≥1 SD),²⁰ and abnormal presentations.^21- 23 Most of these studies are relatively small, including fewer than 100 patients, and, thus, the results are not precise regarding risk estimates. In some studies,^13,24 maternal recall has been used instead of case records, which implies an information bias risk.

Other types of evidence for prenatal risk factors for schizophrenia are the increased occurrence of minor physical anomalies (for example, epicanthal fold and syndactylia) in individuals with schizophrenia^25- 26 and findings from the Dutch Famine Study by Susser et al,²⁷ in which malnutrition was found to be related to neurodevelopmental disturbances including schizophrenia.

Men have an earlier onset of illness, lower premorbid functional level, and poorer outcome than women.^28- 29 Thus, an interesting question is whether there is a sex difference in the risk of schizophrenia following obstetric complications; the results of previous studies are inconsistent in this respect.^11,30- 33 Another question that has been raised is whether obstetric complications are more common in cases with an early age of onset.^8,22,30

Central nervous system disturbances can arise through several mechanisms. In this study, we considered the following 3 possible etiologic mechanisms that might explain the association between obstetric complications and schizophrenia. In particular, we wanted to distinguish these mechanisms since they require different preventive approaches and also correlate, with an overlap, to different types of neuropathologic characteristics.³⁴

Although all these etiologic mechanisms are plausible, there is also an association between a psychotic disorder in the mother and an increased risk of obstetric complications (shown by Sacker et al³⁸ and also in our data). Therefore, the occurrence of psychotic disorders in mothers should be taken into account.

The possibility of linking the Swedish National Birth Register with the National Inpatient Register^39- 40 enabled us to address several of the issues mentioned herein. Data on obstetric complications were available for all Swedish children born between 1973 and 1977 (507,516 children). The children were observed with regard to diagnosis of schizophrenia to a maximum age of 22 years. Furthermore, the birth records could be linked to data on psychiatric illness in the mother, which enabled us to control for maternal psychotic illness in the analyses. We wanted to study specific risk factors as well as sets of risk factors that were indicators of the 3 etiologic mechanisms mentioned herein.

By using the National Birth Register,^41- 42 we obtained data on obstetric complications in all Swedish children born between 1973 and 1977 (507,516 children). The National Birth Register has been in operation since 1973 and contains information on all children born in Sweden, around 100,000 every year. Data are based on forms completed by midwives and the pediatricians in charge and sent to the National Board of Health and Welfare with information on risk factors in pregnancy, events during delivery, anthropometric data on the child, and conditions present in the neonatal period. Diagnoses were given according to International Classification of Diseases, Eighth Revision (ICD-8).⁴³ The proportion of missing data is around 1%.

Anthropometric data and malformations are classified in the register. Small for gestational age and large for gestational age are classified according to a method used by the National Board of Health and Welfare based on birth weight in relation to gestational age (based on information regarding the last menstruation). Cutoff points for these and other anthropometric measures are defined according to Maršál et al⁴⁴ (standard curves ± 2 SDs based on mean measurements of 759 ultrasonic estimates of Scandinavian children). Malformations are classified into types 1 and 2 in the register. Type 1 includes all registered malformations except preauricular appendix, retractile testis, hydrocele testis, hip-joint luxation, and nevus. Type 2 includes malformations that are known to be more accurately and completely recorded than those in type 1. These are malformations in the CNS, lungs, and intestinal canal, serious malformations of the heart, ears, or eyes, choanal atresia, facial clefts, esophageal atresia, intersexuality, and hypospadias.

By means of the unique personal identity number, the cohort was followed up in the National Inpatient Register that covers between 97% and 98% of all episodes of inpatient care. They were given clinical diagnoses according to the Nordic version of International Classification of Diseases, Ninth Revision (ICD-9)⁴⁵ during 1987 to 1996. We recorded all admissions with a diagnosis of schizophrenia (ICD-9 code 295) between 1987 and 1995, which included 238 persons (99 women and 139 men).

Data on the mother in the birth records enabled us to search for data on psychiatric care in the National Inpatient Register from 1973 onward. We collected data on psychotic illness (codes 295-299 according to ICD-8 until 1986; codes 295-298 according to ICD-9 from 1987) in the mothers for the period from 1973 to 1995.

Variables reflecting the 3 types of etiologic mechanisms described herein were grouped to identify states reflecting the following:

We calculated relative risk (RR) estimates with the Mantel-Haenszel test and 95% CIs.⁴⁷ Logistic regression analysis was used to take into account other obstetric complications as confounders for each other. By logistic regression analyses, we estimated odds ratios (ORs) for particular variables, controlling for the other variables included in each hypothesized etiologic mechanism. We also used logistic regression analysis to study the impact of each mechanism while controlling for the other 2 hypothesized mechanisms. Finally, we constructed a logistic regression model with the most important variables from all 3 mechanisms.

The RRs associated with the major risk factors are presented in Table 1. These factors are controlled for in the analyses presented below. There was an 8-fold increased risk of schizophrenia among offspring to mothers who had had a psychotic disorder (including schizophrenia) during their adult life. The risk of obstetric complications, defined as any of the complications included as indicators of the 3 mechanisms, was also increased (RR, 1.3 [95% CI, 1.2-1.3]) in this group of mothers. Children born to mothers younger than 19 or older than 39 years had an increased risk of schizophrenia, although this risk was not statistically significant, except for males born to mothers older than 39 years (RR, 3.5 [95% CI, 1.4-8.9]) and females born to young mothers (RR, 2.3 [95% CI, 1.1-4.7]). The risk was also increased, although it was not statistically significant, if the mother was unmarried or if her marital status was unknown at the time of delivery.

Table 2 shows the RRs of schizophrenia among children born with various obstetric complications. We found that several risk factors during pregnancy, as well as during the time around delivery, significantly increased the risk of schizophrenia: preeclampsia, gestational age younger than 33 weeks, inertia of labor, vacuum extraction, a ponderal index less than 20, respiratory illness, and type 2 malformations. Adjustments for the variables presented in Table 1 did not dramatically alter the results. There was a clear difference between the sexes. Being the fourth child and small for gestational age were associated with a significantly increased risk of schizophrenia in males, while a birth weight of less than 1500 g significantly increased the risk of schizophrenia among females.

In Table 3, we analyzed by logistic regression the impact of the variables included in each etiologic group. Among the hypoxia or ischemia variables, low Apgar scores at 5 minutes and vacuum extraction or cesarean section on indication of fetal distress were associated with the highest risks, although these risks were not statistically significant. Among the indicators of malnutrition, preeclampsia and low ponderal index (leanness) were the strongest risk factors. The third group consisted only of 1 factor, extreme prematurity, which was associated with a significantly increased risk. The separate ORs for each mechanism, without adjustments for the 2 other mechanisms, are also shown in Table 3.

In the next set of regression models, we studied the impact of each etiologic mechanism adjusted for the other 2 mechanisms. The results are shown at the bottom line of the specific mechanisms in Table 3. There is an increased risk, although the risk was not statistically significant, among those who had been exposed to indicators of fetal malnutrition (a total of 24 individuals with schizophrenia) and a gestational age younger than 33 weeks (5 individuals). In the hypoxia or ischemia group, the risk estimate was lower.

By comparing the unadjusted and the adjusted ORs for each of the 3 mechanisms in Table 3, we were able to study the effects of confounding between the different mechanisms. The risk estimate for prematurity was reduced, which could be interpreted as if some of the risk was explained by factors of malnutrition during pregnancy and hypoxia or ischemia at birth. The ORs of malnutrition and hypoxia or ischemia were less reduced in this procedure. Notably, the risks were still increased, which indicates independent effects.

The risks were unevenly distributed between the sexes (Table 3). Indicators of malnutrition were associated with higher risk in males (OR, 2.1 [95% CI, 1.3-3.4]), while the females had increased risk following prematurity (OR, 4.3 [95% CI, 1.0-18.1]). Proportionally more males were exposed to the obstetric complications (Table 3).

In a final model, the strongest risk factors from each etiologic group in Table 3 were entered in a logistic regression model to assess the separate effects of the most important risk factors of the 3 etiologic mechanisms (Table 4). The risk associated with a low Apgar score, low ponderal index, and prematurity was reduced, especially among the females, compared with the findings in Table 3. However, the point estimates were increased for complications of all 3 mechanisms, although only preeclampsia remained significantly associated with an increased risk of schizophrenia.

In this population-based cohort study, controlling for psychotic illness in mothers and other known risk factors, we found that several complications during pregnancy and delivery were associated with an increased risk of schizophrenia. A possible explanation of the fact that we found more obstetric risk factors than those found in previous studies is that our study included a relatively large and homogeneous group of patients with an early age at onset of schizophrenia. Two previous studies^8,22 focused on patients with early age at onset and suggested that obstetric complications might be a stronger risk factor in this group. This finding has been supported by other authors.^19,29,31,48

When reviewing previous findings regarding obstetric risk factors, the patterns of causes are difficult to discern. Attempts have been made to understand the underlying mechanisms. McNeil et al⁶ found an increased risk when using a weighted scale⁷ compared with other scales.^5,23 In a cohort study,⁴ there was no association with specific subcategories of exposure (chronic fetal hypoxia, prematurity, or other complications), but the number of patients with psychoses was small.

To understand more about the underlying mechanisms, we grouped possible risk factors based on previous clinical knowledge and studies of the disease. This resulted in 3 sets of risk factors: malnutrition during fetal life, extreme prematurity, and hypoxia or ischemia around birth. The etiologic mechanisms have some features in common but still seem to be independent of each other. Hypoxia is a factor in common for the mechanism of malnutrition during fetal life as well as hypoxia or ischemia around birth. Nevertheless, we made this classification based on clinical practice and the knowledge that brain damage following chronic hypoxia during fetal life differs from that following an acute episode of hypoxia at birth. Prematurity is associated with a number of different conditions, such as intraventricular hemorrhages, periventricular leukomalacia, interstitial respiratory distress syndrome, infections, as well as hypoxia or ischemia around birth and fetal malnutrition. Despite this, the results of the logistic regression analysis favor our hypothesis that all 3 subcategories represent different possible mechanisms although some features are mutual.

We tried to be strict in our selection of the indicators to avoid unnecessary dilution of the results. Nevertheless, the factors are more or less good indicators of the conditions they are defined as representing. For example, breech delivery does not automatically result in hypoxia or ischemia. The term small for gestational age used herein as an indicator of malnutrition is only a description independent of the cause, which could be genetic influence, chromosomal abbreviations, or infections, although it is often caused by malnutrition.

Preeclampsia was found to be the strongest risk factor for schizophrenia, which has been described in previous studies.^8- 10 Geddes et al⁴⁹ concluded in their meta-analysis that an association is found only when using birth records instead of maternal recall. In our study, data were based on ICD-8 and ICD-9 codes from the birth records. A possible cause for increased risk of schizophrenia might be that in preeclampsia there is an abnormal fetal blood flow, which is associated with reduced supply of nutrition to the fetus.^41,50 Interestingly, Ley⁵¹ found an increased occurrence of minor neurological dysfunctions and intellectual impairment at 7 years of age following abnormal blood flow during fetal life.

The risk of schizophrenia following obstetric complications was higher in males than in females. This is in accordance with findings by some authors^9,27- 32 but not with others.^8,11,22,33 The difference between the sexes is perhaps more pronounced for cases with an early age at onset of schizophrenia.

Because this longitudinal study included a national sample of mothers and their offspring, in other words, it was a population-based study, the risk for selection bias was reduced. The information is based on case records that are consecutively collected in the National Birth Register. Any misclassification should therefore be randomly distributed in the study. The register has been validated regarding errors of recording⁴¹ and proved to be of acceptable quality, except for information about methods of anesthesia during delivery and metabolic testing of the newborn.

Concerning potential confounders, we were able to adjust for mother's psychotic illness during her adult life (1973-1995) by linkage with the National Inpatient Register. To our knowledge, this has not been possible in other studies. The size of the hospital and age of the mother were controlled for, and they were also proxy factors for being a citizen of a large town at birth and parity, respectively, which is why we were able to omit these suspected confounders from the analysis. We did not have specific information on social class, which is a possible confounder in studies on obstetric complications. For example, are low birth weight and young gestational age more common in lower social classes?⁵² We collected data on marital status, which, in previous studies of the National Birth Register, proved to be a good indicator of social position.⁴²

Our conclusion from the findings of this national cohort study including cases of schizophrenia with an early age at onset supports the theory that some of the occurrence of schizophrenia could be associated with a vulnerability originating from obstetric complications. Although preeclampsia, which is a cause of fetal malnutrition, was the strongest individual risk factor, there was evidence of increased risk associated with the other 2 mechanisms that were studied: extreme prematurity and hypoxia or ischemia at birth. The increased risks associated with the different etiologic mechanisms remained, although at lower risk levels, in the logistic models, which indicates independent effects. Thus, it seems as if these etiologic mechanisms are acting in different ways but may result in the same future vulnerability for schizophrenia. One explanation could be that the different pathways result in the same kind of damage. Schizophrenia could result from many different disturbances that may cause an altered balance in the CNS of a rather unspecific nature. Nevertheless, knowledge of specific risk factors for schizophrenia is a public health interest, since at least some of the obstetric risk factors we confirmed can potentially be prevented.

Accepted for publication December 8, 1998.

The study was supported by grants from the Swedish Medical Research Council and the Söderberg-Königska Foundation, Stockholm, Sweden.

We thank Johan Gentz, MD, PhD, for pediatric advice and Anthony David, FRCPsych, and Glyn Lewis, FRCPsych, for fruitful discussions and for commenting on this article. We also thank Bengt Haglund, PhD, for statistical advice.

Corresponding author: Christina Dalman, MD, Södra Stockholms Sjukvårdsområde, Department of Psychiatry, Unit for Psychosis Research, PO Box 4402, S-102 68 Stockholm, Sweden (e-mail: christina.dalman@smd.sll.se).