The characteristics and health needs of pregnant women with schizophrenia compared with bipolar disorder and affective psychoses

We identified a cohort of 456 women with SMI pregnant from 2007–2011. In this cohort , women with non-affective SMI had significantly higher levels of recent acute psychiatric morbidity, less social support (more likely to be single), were younger and had a higher prevalence of smoking and substance misuse during pregnancy compared with women with affective SMI. This suggests that women with non-affective SMI will need particularly high levels of support to optimise both psychiatric and obstetric adverse outcomes. Pregnant women in both diagnostic groups were from relatively deprived backgrounds, had significantly impaired levels of functioning and had relatively high levels of domestic violence in pregnancy recorded in the clinical record. Statistics from the local area show smoking in pregnancy to be at about 4.4% in the local catchment area [28] suggesting our cohort of women with SMI were more likely to smoke in pregnancy. However the prevalence in our sample of women with SMI was much lower than the 38% prevalence reported in a similar clinical population in a retrospective case note review [17]; this is likely to reflect under-reporting by patients and under-recording by clinicians. Some of these risk factors, notably smoking, substance misuse and domestic violence, are potentially modifiable but, although there is evidence for the effectiveness of interventions such as smoking cessation programmes and domestic violence advocacy [29], these are not currently tailored for women with SMI and more evidence is needed on how to best support women with SMI to reduce these risks. The data we have provided on trauma (child abuse and domestic violence) are of particular significance as they could impact on child protection outcomes and the need for trauma informed obstetric care.

Following the onset of pregnancy, a relatively high proportion of women, around 40% in both diagnostic groups, either stopped or switched medication in the first trimester of pregnancy, particularly mood stabilisers. Women in the affective group were more likely than women in the non-affective group to stop medication in the first trimester, but less likely to switch medication. Similar levels of changes in antipsychotic use have been described in data from USA health plans [8] and women in contact with a German teratology service [30]. Around fifteen percent of pregnancies in this study involved exposure to mood stabilisers, mainly sodium valproate and lithium, with 50% discontinuation rates in the first trimester; similar discontinuation rates have been described in the USA [8] and in UK primary care [31]. Our data on the extent of medication exposure in early pregnancy therefore appears to be consistent with other international studies and reflects concerns by both clinicians and patients of teratogenicity. As teratogenic effects are most likely in the first few weeks of pregnancy, stopping mood stabilisers when women know they are pregnant may be too late to prevent these effects, although the impact on neurodevelopment continues for valproate exposure throughout pregnancy [32], so discontinuation would still be indicated. Nevertheless, as pregnancy is unplanned in 50-70% of pregnancies in women with SMI it may be safer not to prescribe valproate to women of childbearing age other than when other medication options have failed – and NICE and SIGN guidelines do indeed recommend this avoidance [33,34]. NICE 2007 guidelines also recommend considering an antipsychotic as an alternative. This appears to be reflected in clinical practice here as there were increases in antipsychotic exposures and decreases in mood stabilisers in the changes in exposure to medication during pregnancy over time from 2007–2011.

The impact of switching and discontinuation of medication is not clear as the limited literature to date has been based on selected clinical samples of pregnant women with bipolar disorder in whom discontinuation of mood stabilisers was associated with a marked increased risk of relapse during pregnancy after adjusting for potential confounders [9]. Outside of the perinatal period switching antipsychotics in people with bipolar disorder [35] or schizophrenia [36,37] appears to be well tolerated but discontinuation is associated with increased risk of relapse, particularly where discontinuation is carried out abruptly [9,38]. Women may be unaware that most psychotropic medication prescribed is associated with very small absolute risks in pregnancy while discontinuation, particularly abrupt discontinuation, may lead to relapse. Follow up of this cohort of women will enable us to investigate the extent to which discontinuation or switching of medication is associated with relapse in both women with affective or non-affective psychoses but this data suggests that psycho-education on the risks of abrupt discontinuation in pregnancy is needed. We will also be able to investigate obstetric outcomes.

Given that most women with SMI have children, and fertility rates have been shown to be increasing [1-3], our findings imply that more attention should be paid to reproductive health when prescribing medications for SMI in women of child bearing age. A high proportion of women changed medication once pregnant. It is noteworthy that in women taking the same medications for epilepsy, far lower rates of discontinuation occur [31], suggesting different attitudes to risk and benefit, despite the fact that maternal deaths are more commonly associated with mental disorders than epilepsy [4]. It is therefore important for psychiatric services to implement recommendations for pre-conception counselling for women with pre-existing chronic conditions so that women can make better informed decisions about medication and support with other risk factors before conception to optimise pregnancy outcomes.

A major strength of this data is the size and likely generalisability of the sample. The mental health provider (South London and Maudsley NHS Foundation Trust) provides near monopoly coverage of its geographical area enabling us to have established a cohort of representative women from an ethnically diverse population with SMI in the perinatal period. However it includes only women in receipt of secondary care, thus missing cases managed exclusively in primary care. Although these data come from the medical notes of patients in one secondary mental health care provider and may thus simply reflect the patterns of prescribing of clinicians in a particular health care trust, the similar rates of medication discontinuation in pregnancy to other national and international data suggests these data are more generalizable to other areas and services.

In addition the CRIS database reflects a dynamic cohort of women who may be referred in and out of services and move from the catchment area; the use of Hospital Episode Statistics, with full coverage of England, allowed follow up for women who moved away or were discharged and enabled us to collect data for women who moved into the area.

Multiple statistical comparisons were made suggesting some associations could have occurred by chance. However findings were in the expected direction suggesting face validity and increasing evidence that our women are a representative cohort of pregnant women with SMI.

The linkage to maternity data also provided a robust method of identifying pregnancies including the whole of England; however, maternity information has varying levels of completeness, some systematic errors in its transfer [39] and it does not include births taking place at home, which was 2.4% for England in 2011 [40]. Finally by using the full clinical care records, a high level of clinical detail can be collected. In particular we collected more detailed data regarding medication use in pregnancy than in many previous studies which have used prescription registries and health plan data only. However the quality of clinical data depends on the accuracy and comprehensiveness of clinicians’ records and some important data are missing such as Body Mass Index and detailed descriptions of social support.