Risks and benefits of psychotropic medication in pregnancy: cohort studies based on UK electronic primary care health records

Onset of psychoses and psychotropic treatment

The onset of psychoses (schizophrenia and bipolar disorder) in women usually occurs within childbearing age and long-term treatment is often required, including a mixture of psychotropic medication such as antipsychotics, lithium and anticonvulsant mood stabilisers, for example valproate, lamotrigine and carbamazepine (see Table 1 for trade names and manufacturers). 1–6 In 2007 in the UK, women with a diagnosis of schizophrenia aged between 18 and 44 years received antipsychotic treatment for more than 50% of the time they were registered with a general practice. However, antipsychotics are being increasingly prescribed not just for schizophrenia, but also for bipolar disorder and severe depression. 2,7,8 A second UK study revealed that in 2009, 233 out of 682 (34%) women of childbearing age who had bipolar disorder received two or more prescriptions of valproate. 2 Atypical antipsychotics in combination with lithium are also often prescribed to women of this age group. 2,4

Pregnancy and psychotropic treatment dilemma

Although many women treated with psychotropic medication become pregnant or plan pregnancy,9–11 no psychotropic medication has been licenced for use in pregnancy. 12,13 This leaves women and their health-care professionals in a treatment dilemma as they need to balance the health of the women with that of the unborn child. 1,14–16 Advice on treatment varies across countries and in some instances standard psychiatric advice is that women should maintain pharmacological treatment across the perinatal period,17 however, some psychotropic medications are known to have teratogenic and adverse neurodevelopmental effects. 4,12,18 Thus, the 2014 National Institute for Health and Care Excellence (NICE) guidelines12 for antenatal and postnatal mental health clearly state that valproate should not be offered for acute or long-term treatment of a mental health problem in women of childbearing potential. Likewise, the guidelines suggest that lithium should not be prescribed to women who are planning a pregnancy or who are pregnant, unless there has been a poor response to antipsychotic medication. 12 The evidence base for adverse effects of other psychotropic medications is sparse. Although antipsychotic drugs are often used in the treatment of both schizophrenia and bipolar disorder in pregnancy, several reviews conclude that there is a paucity of information on the risks and benefits of pharmacological treatment of psychoses in pregnancy in the absence of large and well-designed prospective studies. 1,14,17,19,20 This is further supported by a Cochrane review from 2004,21 updated in 2009,22 which concluded that no randomised controlled trials (RCTs) have been conducted to establish whether the benefits of taking antipsychotic drugs outweigh the risks for pregnant or postpartum women. Similarly, limited information is available for anticonvulsant mood stabilisers other than valproate,18 even though the prescribing of drugs such as lamotrigine has been on the rise for more than a decade.

In recognition of the lack of evidence on the risks and benefits of psychotropic medication in pregnancy and the difficulties encountered in evaluating this issue using a traditional RCT design, in 2001, the National Institute for Health Research Health Technology Assessment (HTA) programme commissioned research utilising information derived from established databases. The commissioned call was titled ‘What are the risks and benefits of psychotropic drugs in women treated for psychosis who become pregnant?’ (HTA reference number 11/35/06). The ‘health technology’ to be evaluated in this call was psychotropic medications that included antipsychotics, lithium and anticonvulsant mood stabilisers prescribed to women with psychosis (bipolar disorder or schizophrenia or overlap syndromes) and whose symptoms are controlled on treatment and who become pregnant. The focus of the investigation was to compare the relative benefits and harms of these different drugs on the mother and the child, when prescribed both during pregnancy and when discontinued.

This project was hence designed in response to this commissioned call. We used data from two large UK clinical databases – The Health Improvement Network (THIN) and the Clinical Practice Research Datalink (CPRD) – to study ‘real-life’ prescribing of psychotropic medication just before, during and after pregnancy and to examine the absolute and relative risks of adverse maternal and child outcomes in women who use psychotropic medication in pregnancy.

Pregnancy cohort

Our pregnancy cohort was based on the recorded date of delivery of the women, the postnatal care record, the first day of last menstrual periods (LMPs) and the estimated delivery dates (EDDs).

The Health Improvement Network includes records that are made as a part of clinical management in primary care; therefore, some pregnancy and antenatal records may not represent actual pregnancies, but they represent historical information. Furthermore, some pregnancies may result in early terminations (either selective or spontaneous abortions/miscarriages) and in these instances little information is recorded in the electronic health records, making it impossible to determine the start and duration of the pregnancy. We therefore derived a set of rules for the inclusion of pregnancies in our cohort. Every pregnancy was ascertained using two different types of information as follows:

• LMP date

• antenatal record

• delivery record

• postnatal care record

• child whose GP record could be matched to the current pregnancy.

Further, we ensured that if we had information only on LMP and antenatal records, the latest antenatal record should be at least 105 days after the first date of the LMP (i.e. equivalent to 15 weeks’ gestation). Only a very small proportion (1%) of the pregnancy cohort was identified from LMP and antenatal records alone.

In total, we identified 495,953 pregnancies in 365,138 women who were permanently registered with one of the general practices that contributed data to THIN in the period between 1 January 1995 and 31 December 2012. This cohort was used as the basis for selecting the target population for our examination of psychotropic medication prescribed before pregnancy, during pregnancy and up to 15 months after delivery; changes in severity of illness during the course of pregnancy and in the period after delivery; and for assessing the absolute and relative risks of adverse maternal pregnancy outcomes.

Duration of pregnancy

In accordance with clinical practice in the UK, the first day of the LMP was considered as the start of pregnancy. As the clinical records do not always hold direct information about the pregnancy length, we estimated the duration of the pregnancy based on information on gestation and maturity of the fetus and/or baby as entered on the electronic record. For women where there was no information available on length of pregnancy or no indications that suggested a child was born pre or post term, we made the assumption that the pregnancy lasted the normal course of 280 days (40 weeks).

Linked mother–child cohorts

Pregnant women and their potential children were linked if they were both registered with the same general practice and shared the same family/household identifier. Furthermore, the date of delivery and the child’s month of birth were required to be near to each other (within 6 months) and the child should have been registered with the general practice within 6 months of birth.

We excluded mother–child pairs when several possible mothers could be linked to a child (< 0.2%). This could have occurred if two women from a block of flats (who would have shared the same household/family identifier) were pregnant at the same time. In cases where the child was registered with another general practice, linkage with the mother was not possible. This was also not done in instances where the mother and child moved to a different practice shortly after the birth of the child. We also excluded pregnancies from the mother–child cohort where there were two or more children associated with the same delivery.

We first identified mother–child pairs in THIN. They were combined with records from the CPRD and together used as the target populations to examine the absolute and relative risks of adverse effects of discontinuation compared with continuation of psychotropic medication on child outcomes.

Studies

We conducted five studies to evaluate utilisation and recording of mental health in primary care; these were (1) the prevalence, initiation and termination of prescribing of psychotropic medication at 6 months before pregnancy, during pregnancy and up to 15 months after delivery; (2) the patterns of recording that indicate worsening of mental health at 18 months before pregnancy, during the course of pregnancy and up to 15 months after delivery; (3) the time trends in prescribing of psychotropic medication around and during pregnancy over the calendar period 1995–2012; (4) the extent of discontinuation and factors associated with discontinuation of psychotropic medication in pregnancy; and (5) the extent of restarting and factors associated with restarting psychotropic medication in pregnancy. Below we provide further details on each study in turn.

Study participants

We used data from women in the pregnancy cohort derived from THIN. As a minimum inclusion criterion, we required that women were registered with the practice throughout their pregnancy. For each of the individual studies, we introduced further inclusion criteria, detailed in the analysis section.

For the studies on discontinuation and restarting psychotropic medication, we randomly selected one pregnancy from the women who had records of more than one eligible pregnancy (antipsychotics, n = 19; lithium, n = 1; and anticonvulsant mood stabilisers, n = 2). Pregnancies that ended in miscarriage or termination were excluded from all four studies as it was impossible to determine the start and end dates of these pregnancies.

Psychotropic medication, prescription intervals and dose

Most psychotropic prescribing occurred over monthly intervals. For > 98% of women prescribed antipsychotics in the year prior to pregnancy, the median gap between prescriptions was < 3 months (91 days) and similar patterns were observed for other psychotropic medication. We considered a new episode of treatment if a woman had not received psychotropic medication prescriptions in the past 3 months (91 days). Likewise, if a woman received no further prescriptions after 3 months, she was deemed to have discontinued an episode of treatment. The date of initiation was considered as the date the first prescription was issued for that episode. The date of termination was considered as the date of the last prescription issued in the episode.

For antipsychotics we also calculated the average daily dose during the period from 4–6 months before the start of pregnancy by dividing the total amount of drugs prescribed over the period by the expected total duration of the relevant prescriptions. Durations were estimated with the help of the enhanced dosage determination method developed by the University of Nottingham Division of Epidemiology and Public Health (further details can be obtained from the data providers of THIN, IMS Health). The mass of each antipsychotic drug was standardised into units of the defined daily dose (DDD) for maintenance treatment of psychosis. 44

Prevalence, initiation and termination of psychotropic treatment: study participants, outcomes and data analysis

Patterns of recording that indicate worsening of mental health: 18 months before pregnancy, during the course of pregnancy and up to 15 months after delivery

Time trends in prevalence of psychotropic medication treatment around and during pregnancy

Discontinuation and factors associated with discontinuation of psychotropic medication in pregnancy

Restarting and factors associated with restarting psychotropic medication in pregnancy

Changes to the protocol

We originally planned to undertake an evaluation of changes in severity of mental illness from the period commencing 18 months before the start of pregnancy up to 15 months after the delivery of the baby. This analysis was deemed infeasible, as it was not possible to ‘grade’ the severity of mental illness in an individual merely from their Read code entries. Instead we decided to explore how the entries varied more generally over this time period by choosing a number of outcomes suggestive of adverse mental health, deterioration or relapse and we then estimated 3-month (trimester in pregnancy) prevalence.

Prevalence, initiation and terminations of psychotropic treatment: 6 months before pregnancy, during pregnancy and up to 15 months after delivery

Overall, 495,953 pregnancies were included in the study from 1 January 1995 to 31 December 2012. Below we describe the results of our studies on the prevalence, initiation and termination of psychotropic treatment. Parts of this work have been published elsewhere. 46,47

Prevalence of psychotropic treatment before, during and after delivery

In the 1–3 months before the start of pregnancy 1051 out of 495,624 (0.21%) women were prescribed antipsychotics (Table 2), 78 out of 495,624 (0.015%) were prescribed lithium and 2046 out of 495,624 (0.41%) were prescribed anticonvulsant mood stabilisers (Table 3). Only 165 out of 52,998 (0.31%) of the women prescribed an anticonvulsant drug had a record of psychosis or depression (see Table 3).

TABLE 2 - Numbers of women prescribed antipsychotic drugs before, during and after pregnancy in the total pregnancy cohort

Prescribed antipsychotic	4–6 months before pregnancy	1–3 months before pregnancy	First trimester	Second trimester	Third trimester	1–3 months after delivery	4–6 months after delivery	7–9 months after delivery	10–12 months after delivery	13–15 months after delivery
Any antipsychotic	999	1051	992	554	509	987	1157	1127	1092	1065
Any typical	550	594	596	301	258	533	582	558	536	519
Chlorpromazine	120	134	185	110	81	114	117	109	98	101
Trifluoperazine	118	111	110	61	50	103	98	100	100	94
Flupentixol	112	125	86	28	15	108	123	123	115	130
Promazine	24	26	88	30	41	17	18	16	36	36
Thioridazine	96	97	52	20	17	83	114	97	89	77
Haloperidol	36	37	36	30	36	59	42	47	39	32
Sulpiride	15	11	9	10	10	30	33	24	23	17
Pericyazine	6	20	11	4	3	10	15	14	15	11
Zuclopenthixol	6	8	3	5	5	7	5	4	7	6
Fluphenazine	15	20	7	1	0	11	11	17	13	14
Perphenazine	11	10	5	2	2	6	10	14	9	8
Levomepromazine	2	2	6	2	0	2	3	5	5	4
Droperidol	1	4	2	0	0	1	3	3	3	2
Loxapine	1	1	1	0	0	0	0	0	0	0
Trifluperidol	0	0	1	0	0	0	0	0	0	0
Pimozide	0	0	0	0	1	1	1	0	0	0
Pipotiazine	0	0	0	0	0	0	1	0	0	1
Paliperidone	0	0	0	0	0	0	1	1	1	1
Any atypical	473	485	427	265	263	485	604	597	579	573
Quetiapine	173	200	171	113	112	177	209	211	209	214
Olanzapine	154	157	149	101	108	199	243	229	219	207
Risperidone	90	85	70	30	25	76	107	125	118	111
Aripiprazole	40	45	37	20	18	37	42	35	41	40
Amisulpride	21	17	9	4	3	13	28	19	17	19
Clozapine	0	1	1	0	0	0	0	0	1	0
Total pregnancies	476,270	495,624	495,953	495,953	493,672	495,894	473,358	456,040	439,313	423,192

TABLE 3 - Numbers of women prescribed the three main anticonvulsant mood stabilisers and lithium before, during and after pregnancy in the total pregnancy cohort

ACMS, anticonvulsant mood stabilisers.

Prescribed drug	4–6 months before pregnancy	1–3 months before pregnancy	First trimester	Second trimester	Third trimester	1–3 months after delivery	4–6 months after delivery	7–9 months after delivery	10–12 months after delivery	13–15 months after delivery
Lithium	84	78	58	33	37	69	98	110	104	105
Any of the three ACMS	1991	2046	1991	1782	1741	1965	1957	1922	1862	1819
Carbamazepine	830	859	822	706	706	799	780	795	776	749
Lamotrigine	687	727	748	707	696	718	722	687	666	639
Valproate	629	607	556	472	436	562	577	555	545	556
Total pregnancies	476,270	495,624	495,953	495,953	493,672	495,894	473,358	456,040	439,313	423,192
Women with a record of psychosis or depression
Any of the three ACMS	160	165	135	57	55	121	151	159	149	152
Carbamazepine	77	82	65	23	26	59	61	68	61	53
Valproate	68	66	47	19	17	50	68	66	66	77
Lamotrigine	19	21	24	16	15	16	26	28	26	26
Total pregnancies	51,718	52,998	53,012	53,012	52,736	53,006	50,487	48,489	46,546	44,711

During pregnancy the prevalence of antipsychotics and lithium prescribing fell dramatically. Hence, 554 out of 495,953 (0.11%) were prescribed antipsychotics and 33 out of 495,953 (0.006%) lithium in the second trimester. The corresponding PRs of antipsychotics and lithium prescribing relative to the period from 1–3 months before pregnancy were 0.53 (95% CI 0.47 to 0.58) and 0.42 (95% CI 0.27 to 0.64), respectively (Table 4 and Figure 1). In the case of anticonvulsant mood stabilisers there was only a small decline in the prevalence of prescribing in pregnancy, 1782 out of 495,953 (0.36%) were prescribed in the second trimester and the PR was 0.87 (95% CI 0.82 to 0.93) (see Table 4 and Figure 1). In the case of those prescribed anticonvulsant mood stabilisers who had a record of psychosis or depression the level of prescribing in the second trimester fell dramatically, with 57 out of 53,012 (0.11%) being prescribed an anticonvulsant mood stabiliser in the second trimester. The PR in the second trimester was 0.35 (95% CI 0.25 to 0.47) (see Table 4 and Figure 1).

TABLE 4 - Prevalence ratios (95% CI) for the prevalence of prescribing psychotropic drugs in the total pregnancy cohort

ACMS, anticonvulsant mood stabilisers; SMI, severe mental illness.

Prescribed antipsychotic	4–6 months before pregnancy	1–3 months before pregnancy	First trimester	Second trimester	Third trimester	1–3 months after delivery	4–6 months after delivery	7–9 months after delivery	10–12 months after delivery	13–15 months after delivery
Any antipsychotic	0.99 (0.91 to 1.08)	1	0.94 (0.86 to 1.03)	0.53 (0.47 to 0.58)	0.49 (0.44 to 0.54)	0.94 (0.86 to 1.02)	1.15 (1.06 to 1.25)	1.17 (1.07 to 1.27)	1.17 (1.08 to 1.28)	1.19 (1.09 to 1.29)
Any typical	0.96 (0.86 to 1.08)	1	1.00 (0.89 to 1.13)	0.51 (0.44 to 0.58)	0.44 (0.38 to 0.51)	0.90 (0.80 to 1.01)	1.03 (0.91 to 1.15)	1.02 (0.91 to 1.15)	1.02 (0.90 to 1.15)	1.02 (0.91 to 1.15)
Any atypical	1.01 (0.89 to 1.15)	1	0.88 (0.77 to 1.00)	0.55 (0.47 to 0.64)	0.54 (0.47 to 0.63)	1.00 (0.88 to 1.14)	1.30 (1.16 to 1.47)	1.34 (1.18 to 1.51)	1.35 (1.19 to 1.52)	1.38 (1.22 to 1.56)
Lithium	1.12 (0.81 to 1.55)	1	0.74 (0.52 to 1.06)	0.42 (0.27 to 0.64)	0.48 (0.31 to 0.71)	0.88 (0.63 to 1.24)	1.32 (0.97 to 1.79)	1.53 (1.14 to 2.08)	1.50 (1.11 to 2.04)	1.58 (1.17 to 2.14)
Any of the three ACMS	1.01 (0.95 to 1.08)	1	0.97 (0.91 to 1.03)	0.87 (0.82 to 0.93)	0.85 (0.80 to 0.91)	0.96 (0.90 to 1.02)	1.00 (0.94 to 1.07)	1.02 (0.96 to 1.09)	1.03 (0.96 to 1.09)	1.04 (0.98 to 1.11)
Any of the three ACMS limited to women with SMI or recent depression	0.99 (0.79 to 1.24)	1	0.82 (0.65 to 1.03)	0.35 (0.25 to 0.47)	0.33 (0.24 to 0.46)	0.73 (0.58 to 0.93)	0.96 (0.77 to 1.21)	1.05 (0.84 to 1.32)	1.03 (0.82 to 1.29)	1.09 (0.87 to 1.37)

FIGURE 1.

Overall prevalence of antipsychotics (a–d), lithium (e–h) and anticonvulsant mood stabilisers (i–l) (women with a record of psychosis or depression) prescribing in the total pregnancy cohort along with levels of starting and ending a prescribing episode and single prescriptions.

For the period after delivery the prevalence of both antipsychotic and lithium treatment were higher in the period between 4 months and 15 months after delivery than before pregnancy. In the case of anticonvulsant mood stabilisers the prevalence after pregnancy remained similar to that before pregnancy irrespective of whether or not the sample was limited to women with psychoses or depression (see Table 4 and Figure 1). The three most commonly prescribed typical antipsychotics in pregnancy were chlorpromazine, trifluoperazine and flupentixol and for atypical antipsychotics, it was quetiapine, olanzapine, and risperidone (see Table 2). For anticonvulsant mood stabilisers, carbamazepine was primarily prescribed in pregnancy followed by lamotrigine and valproate (see Table 3), but this has changed over time, see Time trends in prevalence of psychotropic medication treatment around and during pregnancy over the calendar period 1995–2012.

Initiation of psychotropic treatment, termination of psychotropic treatment and isolated psychotropic prescriptions

Although very few women were prescribed a new episode of antipsychotics, lithium or anticonvulsant mood stabilisers immediately before and during pregnancy (see Figure 1), there was a sharp rise in the proportion of women initiating new episodes in the first 6 months after delivery (see Figure 1). In contrast, a large number of women terminated treatments before and during pregnancy. Thus, the highest proportion of women terminating treatments were found in the 1–3 months before pregnancy and during the first pregnancy trimester (see Figure 1).

In general there were few women who received a single isolated prescription of psychotropic medication and there were even fewer women who received such during the second and third pregnancy trimester (see Figure 1).

Patterns of recording that indicate worsening of mental health: 18 months before pregnancy, during the course of pregnancy and up to 15 months after delivery

Overall, 495,953 pregnancies were included in the study from 1 January 1995 to 31 December 2012.

Below, we describe the annual prevalence of each of the three sets of outcomes: (1) suicide attempts, overdose or deliberate self-harm; (2) mental health hospital admission or examination in relation to the Mental Health Act;45 and (3) psychosis, mania or hypomania before, during and after delivery over the calendar period 1995–2012.

In general, relatively few women had any entries in their records that suggested deterioration or change of severity of mental illnesses in the period before and during pregnancy as well as after delivery (see Figure 2). The recording of suicide attempts, overdose or deliberate self-harm was relatively constant in the 18 months prior to pregnancy (Figure 2 and Table 7). During pregnancy the prevalence declined and relative to the period of 1–3 months before pregnancy the PR was 0.11 (95% CI 0.08 to 0.14) in the third trimester. It rose after pregnancy, but was still only half of what it was prior to pregnancy (see Table 7). The entries made of mental health hospital admissions or invoking of the Mental Health Act45 more than tripled just after delivery in comparison to the period of 1–3 months before pregnancy (PR 3.16, 95% CI 1.86 to 5.60) (see Table 7). Records of psychosis, mania or hypomania followed similar patterns with a doubling just after delivery in comparison to the 1–3 months before pregnancy (PR 2.02, 95% CI 1.53 to 2.69) (see Table 7).

FIGURE 2.

Prevalence recording of (a) suicide attempts, overdose or self-harm; (b) mental health hospital admission; and (c) psychosis, mania or hypomania, from 18 months before the start of pregnancy up to 15 months after delivery.

Time trends in prevalence of psychotropic medication treatment around and during pregnancy over the calendar period 1995–2012

Overall, 495,953 pregnancies were included in the study from 1 January 1995 to 31 December 2012. Below we describe annual prevalence for each of the three classes of psychotropic medication (antipsychotics, lithium and anticonvulsant mood stabilisers) before, during and after delivery over the calendar period 1995–2012. The work on antipsychotics and lithium has been published in part elsewhere. 46,47

Overall, annual prescribing in the 6 months before pregnancy and during pregnancy of both antipsychotics and anticonvulsant mood stabilisers was relatively stable from 1995 to 2006, but increased from around 2007 (Figure 3). By 2011/12, just under 0.4% were prescribed antipsychotics in the 6 months before pregnancy and just under 0.3% received antipsychotic treatment in pregnancy by 2011/12, suggesting a more than 50% increase since 1995/6. For anticonvulsant mood stabilisers the prevalence figures for women with a record of psychoses or depression for 2011/12 were 0.6% before pregnancy, 0.26% during pregnancy and 0.36% after delivery. Hence the treatment prevalence has almost doubled since 1995/6 (see Figure 3).

FIGURE 3.

Prevalence of prescribing over calendar period 1995–2012. Prescribing of antipsychotics (a) in the 6 months before pregnancy; (b) during pregnancy; (c) 6 months after delivery; lithium (d) in the 6 months before pregnancy; (e) during pregnancy; (f) 6 months after delivery; anticonvulsant mood stabilisers (g) in the 6 months before pregnancy; (h) during pregnancy; and (i) 6 months after delivery.

Prescribing of typical antipsychotics has been declining since 1997/8, whereas prescribing of atypical antipsychotics has been increasing. Thus, atypical antipsychotics were more commonly prescribed before pregnancy, during pregnancy and after delivery after 2007/8 (see Figure 3). For anticonvulsant mood stabilisers, prescribing of carbamazepine has declined, whereas both valproate and lamotrigine have gradually increased before pregnancy (see Figure 3). By 2011/12, carbamazepine was superseded by lamotrigine before, during and after pregnancy and valproate was the most commonly prescribed anticonvulsant mood stabiliser before pregnancy (see Figure 3).

Lithium was rarely prescribed; before pregnancy the annual prevalence of lithium prescribing ranged between 0.01% and 0.03%, and during pregnancy between 0.003% and 0.018%. The annual prevalence in the 6 months after delivery declined from 0.048% in 1995/6 to 0.015 in 1999/2000 (see Figure 3).

Discontinuation and factors associated with continuation of psychotropic medication in pregnancy

We identified 207 women receiving typical antipsychotics, 279 receiving atypical antipsychotics, 52 receiving lithium and 93 with a record of psychoses or depression receiving anticonvulsant mood stabilisers in the 4–6 months before the start of their pregnancy.

Although many women discontinued psychotropic medication either before or early in pregnancy the proportion varied between psychotropic treatments. Women prescribed atypical antipsychotics were least likely to discontinue treatment in pregnancy and 150 out of 279 (54%) received further prescriptions after 6 weeks of pregnancy (when the woman is likely to become aware of the pregnancy). In contrast, only 73 out of 207 (35%) women received further prescriptions of typical antipsychotics, 17 out of 52 (33%) lithium and 34 out of 93 (37%) anticonvulsant mood stabilisers after 6 weeks of pregnancy. By the start of the third trimester the figures were 107 out of 279 (38%) for atypical antipsychotics, 39 out of 207 (19%) for typical antipsychotics, 14 out of 52 (27%) for lithium and 13 out of 93 (14%) for anticonvulsant mood stabilisers.

We report below, additional results from studies on discontinuation of psychotropic medication in pregnancy and factors associated with continuation for each of the classes of psychotropic medication (antipsychotics, lithium and anticonvulsant mood stabilisers) Parts of this work have been published elsewhere. 46,47

Discontinuation of antipsychotics

Pregnant and non-pregnant women prescribed atypical antipsychotics discontinued at similar rates up to the start of pregnancy (or index date) (Figure 4). However, pregnant women were more likely to discontinue atypical antipsychotics than non-pregnant women (see Figure 4b).

FIGURE 4.

Discontinuation of antipsychotic drugs in pregnant and non-pregnant women by (a) typical; and (b) atypical antipsychotics.

For women on typical antipsychotics there was a substantial difference in the rates of discontinuation between pregnant and non-pregnant women even before the pregnancy (see Figure 4a) and the gap became wider in early pregnancy (see Figure 4a). The comparisons with non-pregnant women, however, suggest that awareness of the pregnancy may not be the only reason for stopping antipsychotics. About 75% of non-pregnant women continued both typical and atypical antipsychotics throughout the follow-up period (see Figure 4).

The rates of discontinuation differed by dose and type of antipsychotics (Figure 5). Among women receiving prescriptions of less than one-quarter of the DDD of typical antipsychotics, only 29 out of 118 (25%) continued prescriptions after 6 weeks. For women on atypical antipsychotics the figure was 24 out of 52 (46%) after 6 weeks (see Figure 5).

FIGURE 5.

Discontinuation of antipsychotic drugs in pregnant women, by dose, in women who were prescribed (a) typical; and (b) atypical antipsychotics.

Women on a high dose (DDD > 1) of typical antipsychotics were highly likely to discontinue prescriptions prior to pregnancy in contrast to women on a high dose (DDD > 1) of atypical antipsychotics (see Figure 5). Three out of 15 women on high dose typical antipsychotics were on depots prior to the start of pregnancy.

Switch between typical and atypical antipsychotic treatment

In general, few women switched between typical and atypical antipsychotic treatment just before or in pregnancy. Only 5 out of 207 (2.4%) women switched from typical to atypical antipsychotics and 9 out of 279 (3.2%) switched from atypical to typical antipsychotics. However, among the more frequently used antipsychotics, switching levels were high for two drugs: 12 out of 50 women (24.0%) switched from risperidone to another antipsychotic, while 9 out of 48 (18.8%) switched from trifluoperazine to another antipsychotic.

Discontinuation of lithium

For lithium, there was a substantial difference in the rates of discontinuation between pregnant and non-pregnant women (Figure 6). Only 14 out of 52 (27%) continued lithium treatment after the start of the third trimester. Of the non-pregnant women, 80 out of 104 (77%) continued lithium treatment beyond this period (see Figure 6).

FIGURE 6.

Discontinuation of lithium in pregnant and non-pregnant women.

Switch from lithium to antipsychotic treatment

Of the 39 women who discontinued lithium before the end of follow-up at 220 days, six received at least two prescriptions for an antipsychotic in the 91 days after lithium discontinuation. However, five of these were already receiving an antipsychotic prior to lithium discontinuation, so only one could be classed as having ‘switched’ treatment. We cannot exclude the possibility that some of the other five may have started a new antipsychotic while gradually tapering off lithium.

Discontinuation of anticonvulsant mood stabilisers

We identified 1175 women receiving anticonvulsant mood stabilisers in the period of 4–6 months before the start of their pregnancy. Of these, 1007 had a record of epilepsy, 62 had a record of psychosis, 31 had a record of depression and 75 did not have a record of any of these indications. Among the 93 women with a record of psychosis or depression, there was a substantial difference in the rates of discontinuation between pregnant and non-pregnant women (Figure 7). This was the case even before the pregnancy started (see Figure 7). As for antipsychotics and lithium, the comparisons with non-pregnant women suggest that awareness of the pregnancy may not be the only reason for stopping treatment, as about 70% of non-pregnant women continued anticonvulsant mood stabilisers throughout the follow-up period (see Figure 7).

FIGURE 7.

Discontinuation of anticonvulsant mood stabilisers in pregnant and non-pregnant women with a record of psychosis or depression.

Although the numbers were small, there appeared to be no substantial difference in the rate of discontinuation between women who were prescribed lamotrigine, carbamazepine or valproates (Figure 8).

FIGURE 8.

Discontinuation of anticonvulsant mood stabilisers in pregnant women with a record of psychosis or depression, by drug.

Discontinuation of anticonvulsant mood stabilisers in women with a record of epilepsy compared with women with other or unknown indications

There was a sharp contrast in rates of discontinuation between women with a record of epilepsy who were prescribed anticonvulsant mood stabilisers and women who had a record of psychosis or depression or unknown indications (Figure 9). Hence, among women who had a record of epilepsy 795 out of 1007 (79%) continued to be prescribed anticonvulsant mood stabilisers beyond 6 weeks of pregnancy (see Figure 9) compared with 34 out of 93 (37%) for women with records of psychoses or depression (see Figure 9).

FIGURE 9.

Discontinuation of anticonvulsant mood stabilisers in pregnant women with a record of epilepsy, psychosis, depression or unknown indications. SMI, severe mental illness.

Restarting and factors associated with restarting psychotropic medication in pregnancy

We identified 134 women who discontinued typical antipsychotics and 129 women who discontinued atypical antipsychotics before 6 weeks of pregnancy. The figures were 35 for lithium and 59 for anticonvulsant mood stabilisers prescribed to women with a record of psychosis or depression. The proportion of women who restarted treatment (i.e. received additional prescriptions after > 91 days of a gap) varied between psychotropic treatments. At 15 months after delivery, the proportion of restarting treatment was highest for lithium (0.76) and lowest for typical antipsychotics (0.40). We report below additional results from studies on restarting of psychotropic medication in pregnancy and factors associated with restarting for each of the classes of psychotropic medication (antipsychotics, lithium and anticonvulsant mood stabilisers).

Restarting of antipsychotics

Of the 134 women who discontinued typical antipsychotics, the proportion of women who had restarted at the EDD, 6 months and 15 months after delivery was 0.17, 0.33 and 0.40, respectively (Figure 10). Of the 129 women who discontinued atypical antipsychotics, the proportion of women who had restarted treatment at the EDD was 0.23. The proportion of women who had restarted at 6 and 15 months after delivery was 0.44 and 0.52, respectively (see Figure 10).

FIGURE 10.

Proportions of women restarting antipsychotics among those who discontinued by 6 weeks of pregnancy (Kaplan–Meier failure functions). (a) Typical antipsychotics and (b) atypical antipsychotics.

Restarting of lithium

Of the 35 women who discontinued lithium, the proportion of women who restarted treatment by the EDD was 0.26. The proportion of women who had restarted at 6 and 15 months after delivery was 0.64 and 0.76, respectively (Figure 11).

FIGURE 11.

Proportions of women restarting lithium among those who discontinued by 6 weeks of pregnancy (Kaplan–Meier failure functions).

Restarting of anticonvulsant mood stabilisers

Of the 59 women with a record of psychosis or depression who discontinued anticonvulsant mood stabilisers the proportion of women who restarted treatment at the EDD was 0.29. The proportion of women who had restarted at 6 and 15 months after delivery was 0.58 and 0.64, respectively (Figure 12).

FIGURE 12.

Proportions of women restarting anticonvulsant mood stabilisers among those who discontinued by 6 weeks of pregnancy (Kaplan–Meier failure functions).

Prevalence, initiation and termination of psychotropic treatment around and during pregnancy

Overall, the patterns of psychotropic medication prescribing before and during pregnancy and after delivery were remarkably similar, although the absolute prevalence estimates varied between classes of medication. Following a broad ‘u shape’ the prevalence was relatively constant before pregnancy, decreased sharply in early pregnancy and increased after delivery to the level of before or even higher. Hence, 0.11% were prescribed antipsychotics, 0.006% lithium and 0.11% were prescribed anticonvulsant mood stabilisers in the second trimester. The prevalence was higher after delivery for atypical antipsychotics and lithium, but remained at the same level before and after pregnancy for typical antipsychotics and anticonvulsant mood stabilisers.

Chlorpromazine, trifluoperazine and flupentixol were the three most commonly prescribed typical antipsychotics in pregnancy and quetiapine, olanzapine, and risperidone were the most commonly prescribed atypical antipsychotics.

We observed a peak in the starting of new prescribing episodes just after delivery for all three classes of psychotropic medication and likewise a peak in termination of prescribing episodes during the first trimester. Single isolated prescriptions were at the lowest during the second and third trimester and that was true of all three classes of psychotropic medications.

A US study based on pharmacy dispensing data from nearly 600,000 deliveries demonstrated a sharp increase in the use of atypical antipsychotics in pregnant women between 2001 and 2007, but estimated that atypical antipsychotics were, on average, dispensed to 26.7 out of 10,000 pregnancies in the second trimester, whereas typical antipsychotics were dispensed to 4.8 out of 10,000 pregnancies in the second trimester. 37 Our estimates are much lower, but in line with another study,36 also based on UK primary care data. This study, which covered the period from 1989 to 2010, found that 0.08% were prescribed antipsychotics in the second trimester. A Swedish study identified 570 women who reported use of antipsychotics out of 958,729 pregnancies (equivalent to 0.06%) from the Swedish birth registry between July 1995 and the end of 2005. 48 Maternal drug use in early pregnancy was recorded from interviews performed by the midwife at the first antenatal care visit, usually before the end of the first trimester. It is likely, however, that this study may have underestimated the prevalence of antipsychotic use in pregnant women in the Swedish population. 49 Another study also using Swedish registry data, but from 2005 to 2009 estimated atypical antipsychotic use during pregnancy to be approximately 0.1%. 50

We found that quetiapine, olanzapine and risperidone were the three most commonly prescribed atypical antipsychotics both before and after pregnancy. This mirrors the patterns of atypical antipsychotics dispensed in the USA37 and the general prescribing pattern of antipsychotics in the UK. 7

The utilisation of anticonvulsant mood stabilisers was examined in another US study also based on pharmacy dispensing data from nearly 600,000 pregnancies. 40 This study included benzodiazepines as one of the ‘older’ anticonvulsant mood stabilisers and the prevalence estimates were primarily driven by the dispensing of these drugs. They found that 0.9% received benzodiazepines at any time in pregnancy while 0.2% received valproic acid derivatives. 40 Of those who were dispensed an old anticonvulsant mood stabiliser 4024 out of 9001 (45%) had an indication of psychiatric disorder, while 2115 out of 3515 (60%) of those dispensed new anticonvulsant mood stabilisers had an indication of psychiatric disorder. Only between 21% and 25% had an indication of epilepsy. 40 A study by Kulaga et al. 41 identified 349 epileptic pregnant women within the Quebec Pregnancy Registry and divided these into three groups based on maternal use of anticonvulsant mood stabilisers during pregnancy. Like our study, Kulaga et al. 41 also found that the frequency of exposure to anticonvulsant mood stabilisers declined substantially during pregnancy, and in the second trimester was estimated to be 0.22% (95% CI 0.19 to 0.26%). 41 Most women were dispensed anticonvulsant mood stabilisers as monotherapy (79.6%) during pregnancy and the three most prevalent anticonvulsants were carbamazepine (29.9%), valproate (19.7%) and phenytoin (11.5%). 41

To our knowledge there are no recent estimates of lithium usage or prescribing in pregnancy. However, the study by Reis and Källén48 identified 79 lithium users among 958,729 pregnancies (< 0.01%) from the Swedish birth registry between July 1995 and the end of 2005. Another study on adverse birth outcome following lithium exposure in pregnancy identified 83 lithium-exposed pregnancies among women who contacted the Israeli Teratology Information Service between 1999 and 2010. 51 Both studies support our findings that lithium is still used by pregnant women, but rarely.

Patterns of recording that indicate worsening of mental health; 18 months before pregnancy, during the course of pregnancy and up to 15 months after delivery

In this exploratory study we observed that recording of suicide attempts, overdose or deliberate self-harm was relatively constant in the 18 months before pregnancy, but declined during pregnancy. Recording of psychosis, mania and hypomania was also slightly lower in the second and third trimester. However, entries increased substantially in the immediate period after delivery, while hospital admissions and the Mental Health Act45 examinations tripled compared with before pregnancy. Recording of psychosis, mania and hypomania followed similar patterns with a doubling just after delivery.

Comparing the results of this study with the prevalence of psychotropic medications around pregnancy there appears to be a strong correlation in terms of the timing of psychosis, mania and hypomania and the start of new prescribing episodes just after delivery (see Figure 1). In contrast, the low level of entries for suicide attempts, overdose or deliberate self-harm during pregnancy may suggest that there were fewer such events. However, it is also possible that some women were less likely to reveal these events to their GP during pregnancy, in particular if they were not in receipt of psychotropic medication.

Time trends in prevalence of psychotropic medication treatment around and during pregnancy

The overall annual prevalence of prescribing in the 6 months before and during pregnancy of both antipsychotics and anticonvulsant mood stabilisers was relatively stable from 1995 to 2006, but increased from around 2007. The total prevalence of antipsychotic treatment has increased by > 50% before and during pregnancy since 1995/6 and the prevalence of anticonvulsant mood stabilisers prescribing in women with a record of psychosis or depression has almost doubled since 1995/6. There has been a shift from typical to atypical antipsychotics in the study period. Likewise, for the anticonvulsant mood stabilisers, carbamazepine has recently been superseded by valproate and lamotrigine. Lithium was rarely prescribed and prescribing fluctuated over time with annual prescribing after delivery almost halved between 1995/6 and 2011/12. We observed a sharp increase in the prevalence of atypical antipsychotic prescribing in pregnancy although this only really ‘kicked off’ after 2007/8. However, the prescribing of atypical antipsychotics after delivery superseded typical antipsychotics by 2005/6.

Our study findings are consistent with an increase in atypical antipsychotics usage in the general population8 and an expansion of indications for usage including bipolar disorder and treatment-resistant depression. 2,52 However, there seems to be an even faster growth in dispensing of antipsychotics to pregnant women in the USA, which has increased 2.5-fold between 2001 and 2007. 37

As mentioned in Prevalence, initiation and termination of psychotropic treatment around and during pregnancy, there seem to be no recent estimates of lithium usage or prescribing in pregnancy. However, a UK study also based on primary care data suggests that lithium continues to be prescribed to women of childbearing age with bipolar disorder2 and that women were prescribed the drug for 30% of the time they were registered with the general practice, which remained constant between 1996 and 2009. 2

Pharmacy dispensing of anticonvulsant mood stabilisers before and during pregnancy in the USA increased by approximately 40% between 2001 and 2007, with 15.7 women receiving anticonvulsant mood stabilisers per 1000 pregnancies in 2001 and 21.9 per 1000 pregnancies in 2007. 40 As mentioned earlier these estimates include benzodiazepines, which account for a large proportion of the ‘older’ anticonvulsant mood stabilisers. However, it appears that our overall prevalence estimates for prescribing of anticonvulsant mood stabilisers are roughly the same for women prescribed anticonvulsant mood stabilisers with a record of psychosis or depression.

For some time there has been concern about potential teratogenic and neurodevelopmental effects of valproate. 15,53–58 We were therefore surprised to observe that valproate appears to be the most commonly prescribed mood stabilising anticonvulsant in the 4–6 months before pregnancy in women with a record of psychosis or depression in 2011/12. The NICE guidelines for antenatal and postnatal mental health and for management of bipolar disorders both issued in 201412,59 state clearly that valproate should not be prescribed to girls and women of childbearing potential. An American study based on Florida Medicaid beneficiaries suggested that use of valproate in pregnancy has declined for women with epilepsy over the period 1999 –2009, but not for other indications60 and a Danish registry study suggested that dispensing of lamotrigine has been sharply rising during pregnancy in the period 1996–2006 (the time period of the study). 61 Future investigations should monitor the impact of these guidelines on prescribing.

Discontinuation and factors associated with discontinuation of psychotropic medication in pregnancy

The overall patterns of discontinuation of psychotropic medication around pregnancy were remarkably similar. Thus, many women were not prescribed further psychotropic medication after 6 weeks of pregnancy, suggesting that pregnancy is a major determinant for stopping psychotropic prescribing. By the time the prescription would be due for renewal many women would have been aware of their pregnancy and have decided to stop the medication. Women prescribed lithium were most likely to discontinue treatment with only 17 out of 52 (33%) receiving further prescriptions after 6 weeks of gestation. For the other psychotropic medications the figures were 73 out of 207 (35%) for typical antipsychotics, 34 out of 93 (37%) for anticonvulsant mood stabilisers (in women with a record of psychosis or depression) and 150 out of 279 (54%) for atypical antipsychotics. By the start of third trimester the figures were 13 out of 93 (14%) for anticonvulsant mood stabilisers, 39 out of 207 (19%) for typical antipsychotics, 14 out of 52 (27%) for lithium and 107 out of 279 (38%) for atypical antipsychotics. However, women prescribed anticonvulsant mood stabilisers with an indication of epilepsy were far more likely to continue treatment in pregnancy than women with any other indications. Hence, of women with a record of epilepsy, 795 out of 1007 (79%) continued to be prescribed anticonvulsant mood stabilisers beyond 6 weeks of pregnancy.

Those most likely to continue antipsychotic treatment were those who had received continuous treatment for > 12 months prior to pregnancy. Other determinants for continuation of antipsychotic prescribing in pregnancy included women’s age and being on higher pre-pregnancy treatment doses. It appears that a greater proportion of those who continued lithium in pregnancy had also been prescribed antidepressants before they became pregnant, were pregnant with their first child and received lithium for < 6 months before pregnancy. However, the numbers were small and CIs were overlapping. Factors associated with continuation of anticonvulsant mood stabilisers in women with a record of psychosis or depression included duration of previous treatment and obesity.

Previous studies on utilisation of psychotropic medication in pregnancy suggest that many women discontinue treatment either just before or in early pregnancy. 36,37,39,40,49 As we describe in a Chapter 2, there was a peak in the termination of prescribing episodes for all three classes of psychotropic medications in the first pregnancy trimester.

The pregnant woman’s mental health team/consultant as well as the GP may play a pivotal role in advising an individual woman on continuation of psychotropic medication in pregnancy. This would be in keeping with the recommendations made by national formularies and the NICE guidelines. 12,32,59 As outlined earlier in this document, there are a few very specific recommendations in terms of prescribing of psychotropic medication to pregnant women, but most are non-specific. In many situations this leaves both health-care professionals and the women with a very difficult and complex decision. They will have to weigh up risks to the mother and child of continuation versus discontinuation of medication in each individual case.

The greatest risk of discontinuation of psychotropic medication is the possibility of relapse of mental illness in pregnancy and postpartum. An observational study of mood disorders in 2252 pregnancies and postpartum periods demonstrated that women with bipolar disorders were at particularly high risk of developing major depression in the postpartum period (prevalence: 19% and 29% in women with bipolar I disorder and bipolar II disorder, respectively). 62 Likewise, it has been shown that women with a history of psychotic disorder are at higher risk of postpartum psychiatric illness, in particular non-psychotic anxiety and depressive disorders. 63 The severity of these illnesses is highlighted by case reports of suicide among these women. 16 Aside from the direct effects of discontinuation of psychotropic medication on the mother, the indirect impact on the fetus and child of severe depression and puerperal psychosis also needs to be taken into consideration. Women may have different reasons to discontinue psychotropic medication in pregnancy; some may discontinue because they fear the medication will harm the unborn child and deem that uncertainty of risk to be higher than the potential risk of relapse. Thus, an international survey of nearly 10,000 women reveals a substantial disparity between women’s perceived risks and the actual risks of prescribed medication in pregnancy. 64 In this survey women rated the risks of antidepressants on par with smoking and alcohol and almost as dangerous as thalidomide. 64 Indeed only about 20% of around 5000 women who were prescribed antidepressants before pregnancy continued to receive treatment after 6 weeks of gestation. 35 However, some women may discontinue psychotropic medication in this period because they no longer need it. We also observed that a number of women on high-dose typical antipsychotics ceased receiving prescriptions before they became pregnant. This may be a part of pregnancy planning, but could also be explained by the fact that typical antipsychotics are known to reduce fertility by inducing hyperprolactinaemia. Finally, the comparison of discontinuation between non-pregnant and pregnant women revealed that pregnancy is not the only reason for women to discontinue psychotropic medication. Other factors, such as experience of adverse effects and stigmatisation associated with psychotropic medication may influence continuous prescribing in pregnant as well as non-pregnant women.

As previously demonstrated by Man et al. ,34 using a similar approach to this study, the rate of discontinuation of anticonvulsant mood stabilisers in pregnancy is associated with the indications for treatment. Thus, women with a record of epilepsy were much more likely to continue anticonvulsant mood stabilisers in pregnancy compared with women with a record of psychosis or depression. There may be several reasons why we see these differences such as alternative treatment options and advice from specialists. However, a detailed investigation of this was outside the remit of this project.

Restarting and factors associated with restarting of psychotropic medication in pregnancy

Of the women who discontinued psychotropic medication either before or in early pregnancy those who were prescribed lithium were most likely to restart treatment. Thus, within 15 months of delivery three-quarters of the women who had discontinued were again receiving lithium treatment. For anticonvulsant mood stabilisers the proportion of women who had restarted within 15 months after delivery was 0.64. On the other hand approximately half of the women prescribed atypical antipsychotics had restarted by 15 months after delivery and a slightly smaller proportion of women prescribed typical antipsychotics had restarted. Overall, there were no clear predictors of restarting of treatment within 6 months of delivery, although for typical antipsychotics women who received an average daily dose greater than the DDD before pregnancy were about three times as likely to restart treatment as women receiving an average daily dose less than one-quarter of the DDD.

With these data we were unable to determine whether women restarted medication in response to relapse of their mental illness or whether treatments were given to prevent major psychoses or other relapses in the postnatal period, but it is likely that it is a mixture of the two. As discussed in Discontinuation and factors associated with discontinuation of psychotropic medication in pregnancy previous research has demonstrated that both women with a psychotic disorder and women with mood disorders are at a particularly high risk of developing major depression and affective psychosis in the postpartum period. 62,63 Some observational studies have sought to compare the risks of relapse of mental illnesses during and after pregnancy in relation to psychotropic treatment. 65–68 Newport et al. 65 observed 16 women who stopped lamotrigine treatment for bipolar disorder in pregnancy. All 16 women experienced a new illness episode during the course of the pregnancy in contrast to 3 out of 10 women who continued treatment. 65 In another study66 more than half (52%) of 42 women who discontinued treatment with lithium in pregnancy experienced a recurrence of bipolar disorder in pregnancy compared with 21% in the year before treatment was discontinued. Viguera et al. 66 observed 89 pregnant women with bipolar disorder (euthymic at conception) treated with one or more mood stabilisers. Of these women, 62 discontinued treatment and 86% experienced at least one recurrence in pregnancy. In contrast only 37% of the 27 women who maintained treatment experienced recurrences in pregnancy66 and similar findings were reported by Cohen et al. 42 for antidepressants. The relationship between drug treatment and mental illness is intrinsically difficult to disentangle as the women who continue psychotropic medication during pregnancy may be more likely to be those who are at the highest risk of relapse; this may result in spurious findings suggesting that psychotropic medication may increase risks of adverse mental outcomes or suggest no difference as in the study by Yonkers et al. 68 on antidepressant prescribing in pregnancy.

It is likely that some women titrate their medication (for example take one tablet instead of two tablets) and hence ‘stretch’ a prescription for much longer during pregnancy. In such cases what may have appeared as a restart of treatment (prescription after more than 91 days gap), was actually a delay in picking up a subsequent prescription because of this prescription ‘stretching’ behaviour.

Studies

For each class of psychotropic drugs (i.e. antipsychotics, lithium and anticonvulsant mood stabilisers) we performed two studies; one of which was based on the pregnancy cohorts to examine maternal outcomes, and another of which was based on the linked mother–child cohort to examine child outcomes.

Participants

Psychotropic medication exposures

For each class of psychotropic medication (i.e. antipsychotics, lithium and anticonvulsant mood stabilisers) we created four cohorts based on medication prescribed. Figure 13 provides a graphical representation of how the exposure cohorts were constructed.

FIGURE 13.

Schematic describing the derivation of the four exposure cohorts in the pregnancy and mother-child linked cohorts. Rx, prescription.

Cohort A contained women with records of psychotropic treatment between 4 and 24 months before the start of pregnancy and no evidence of prescriptions issued after 4 weeks prior to pregnancy start.

Cohort B1 contained women with records of psychotropic treatments between 4 and 24 months before the start of pregnancy and with evidence of prescriptions issued between 31 and 105 days (inclusive) after the start of pregnancy (which is the critical period for many major congenital malformations).

Cohort B2 contained women with records of psychotropic treatment between 4 and 24 months prior to the start of pregnancy and records of psychotropic treatment within 92 days prior to the delivery date. Thus, cohort B1 and cohort B2 are not mutually exclusive.

Cohort C contained women with no records of psychotropic treatment from 24 months before the start of pregnancy through to the delivery date. Start of pregnancy was defined as the first day of LMP or 280 days before delivery if no indication suggested a different duration of pregnancy.

These cohorts were based on prescribing of the psychotropic medication, irrespective of the underlying indication for prescribing.

For the study of anticonvulsant mood stabilisers we defined an additional two sets of subcohorts: (1) cohorts limited to women with a record of psychosis or depression and no records of epilepsy and (2) a cohort of women prescribed valproate in the beginning of pregnancy. For the latter cohort we made comparisons against other anticonvulsant mood stabilisers (lamotrigine and carbamazepine) within cohort B1 as well as against cohort A and C. The reason to select more than one comparison cohort was to investigate the potential issues of confounding. Thus, we anticipated that women who had discontinued treatment (cohort A) just before pregnancy would be more similar both in terms of their measured, but also their unmeasured, characteristics to women who continued treatment in pregnancy (cohort B1 and B2) than women in cohort C.

Characteristics of the women and information on prescribed medication

We extracted information from the women’s electronic primary care health records in order to include the following information in the analyses: age at delivery, social deprivation (Townsend scores), calendar year of delivery, body mass index (BMI), illicit drug use, alcohol problem, smoking status, ethnicity, pre-existing medical conditions (depression, epilepsy, psychosis, hypertension, diabetes), prescriptions from medication listed in the BNF chapter 4, including antidepressants, anxiolytics, hypnotics, anticonvulsant mood stabilisers, antipsychotics and lithium.

Details on how the characteristics and information on prescribed medications were defined and relevant Read codes and drug codes are provided in Appendix 1.

Outcomes

We separated the outcomes into maternal and child outcomes. Below we describe these in further detail.

Data analysis

For each class of psychotropic medications, characteristics of the women and the maternal and child outcomes were tabulated for cohort A, B1 and C. For continuous variables, the means and standard deviations (SDs) were estimated and for categorical variables, the numbers of individuals in each category and percentage were estimated. As there was an overlap between the individuals in cohort B1 and B2, and as cohort B2 was only used for a few specific sets of analyses the characteristics of these cohorts are described in Appendix 1.

For each outcome we first estimated and tabulated the number and percentage of events as well as risk difference with 95% CIs where there were more than five events. We then estimated RRRs using Poisson regression. Comparisons were made between cohort B1 (women who had continued psychotropic medication in the first part of pregnancy) and cohort A (women who had discontinued treatment before pregnancy) using the latter as a reference category. Likewise, comparisons were made between cohort B1 (women who had continued psychotropic medication in the first part of pregnancy) and cohort C (women who had not been treated with psychotropic medication) using cohort C as a reference category.

For transient poor birth outcomes (tremor, agitation, breathing and muscle tone problems) the comparisons were made between cohort B2 (women who had received psychotropic medication in the later part of pregnancy) and cohort A, as well as between cohort B2 and cohort C. Specifically for anticonvulsant mood stabilisers we compared outcomes between women in cohort B1 who were prescribed valproate against women in cohort A and C as well as women in cohort B1 who were prescribed other anticonvulsant mood stabilisers.

Poisson regression models were developed, thus providing RRRs with 95% CIs. For all analyses we adopted the following sequence of analysis: model (a): examining crude associations, that is, with no adjustment; model (b): examining associations with adjustment for maternal age; model (c): examining associations with adjustment for ‘health and lifestyle’ factors, that is, adjustment for smoking, obesity, records of alcohol and illicit drug problems; model (d): examining associations with adjustment for concomitant prescriptions issued in the same time window as for cohort B1. For example, for the analyses on antipsychotics we accounted for prescription of antidepressants and anticonvulsant mood stabilisers; and model (e): examining associations with adjustment for all covariates. The results of model (a) and model (e) were tabulated. Further, we tabulated the associations between the health, lifestyle factors, concomitant medication and the outcomes from model (e). These are reported in Appendix 1.

We only report results from analyses where there were more than five events in each exposure group.

Changes to the project protocol

Although we have maintained the same cohorts as planned in our original proposal, A, B1 and B2 (originally called C), we decided to also include a cohort of women not treated with psychotropic medication (cohort C) in our comparisons. We then made the comparisons between B1 (and B2 for specific outcomes) and A as well as B1 and C. This allowed us to compare our results with other studies where similar comparisons were made. In general we did not conduct analyses on specific drugs as the sample sizes became small. However, we did perform specific analyses to examine child outcomes in women prescribed valproate in pregnancy (cohort B1) in contrast to other anticonvulsant mood stabilisers prescribed in pregnancy (cohort B1) as well as cohorts A and C.

We decided to use standard regression analyses methods rather than propensity score-matched methods in order to have direct estimates available for the covariates. In terms of the outcomes we still use composite outcomes, but have changed our groupings from the original proposal to form three groups of composite child outcomes: (1) poor birth outcome, (2) transient poor birth outcome and (3) neurodevelopmental and behavioural outcomes as described further in Chapter 4, Child outcomes. We were unable to identify child and maternal renal problems in a coherent way and decided to drop this as an outcome.

Characteristics of the women in the pregnancy cohort

The characteristics of the women in cohort A, B1 and C are detailed in Tables 16–20. There was a substantial overlap between the women in cohort B1 and B2 (i.e. many of the women who received treatment in the beginning of pregnancy also received treatment towards the end). The details of the characteristics of women in cohort B2 can be found in Appendix 1.

Women who were prescribed antipsychotics, lithium or anticonvulsant mood stabilisers (with a record of psychosis or depression) in pregnancy were older {mean ages between 32 years (SD 5.6 years) and 35 years (SD 5.4 years) than women not prescribed psychotropic medication (cohort C) [mean age 30 years (SD 5.9 years)]}. Likewise, women who were prescribed antipsychotics and lithium in pregnancy were older than women who discontinued before pregnancy (cohort A). A large proportion of the women prescribed psychotropic medication in pregnancy were obese. For example, 72 out of 416 (17%) of the women who were prescribed antipsychotics were obese in contrast to 77 out of 670 (12%) in those who discontinued antipsychotics (cohort A) and 20,554 out of 318,434 (6.5%) in those not prescribed antipsychotics (cohort C) (see Table 16). Illicit drug use and alcohol problems were commonly recorded among women who continued psychotropic medication in pregnancy as well as those who discontinued. Hence, illicit drug use was recorded in 56 out of 416 (13.5%) of women who continued antipsychotics in pregnancy (see Table 16) and in 2–3% of women who continued anticonvulsant mood stabilisers and lithium. Less than 1% of the women not prescribed psychotropic medication had a record of illicit drug use or alcohol problems. More than 45% of women prescribed antipsychotics and anticonvulsant mood stabilisers in pregnancy (cohort B1) (with a record of psychosis or depression) were smokers. In women prescribed lithium in pregnancy (cohort B1) it was 37% and in the cohorts of women not prescribed psychotropic medication it was just under 20% (cohort C). Despite the fact that a large number of data on ethnicity was missing it appears that a relatively large proportion of women prescribed antipsychotics in pregnancy were black. Many women who continued antipsychotics, lithium or anticonvulsant mood stabilisers in pregnancy were also prescribed other medication listed in BNF chapter 4. For example, 238 out of 416 (57%) women who received antipsychotic treatment in pregnancy (cohort B1) also received antidepressant treatment in contrast to < 2% (5942/318,434) of the women in cohort C. Likewise, many received anticonvulsant mood stabilisers in the cohort of women receiving antipsychotics in pregnancy and vice versa.

Including all women, irrespective of indication, in anticonvulsant mood stabilisers cohorts slightly changed the characteristics of cohorts (see Table 18). In these cohorts the proportion of individuals with a record of obesity, alcohol problems and smoking was larger in those who discontinued treatment before pregnancy (cohort A) than in those who continued treatment (cohort B1).

Characteristics of the women in the mother–child cohorts

The mother–child cohorts were a subset of the pregnancy cohorts and included between 65% and 75% of the pregnancy cohorts. The characteristics of women in the mother–child cohorts varied slightly from the distribution in the pregnancy cohorts, but cohorts were overall similar for each class of psychotropic medication (Tables 21–25).

Characteristics of children in the mother–child study cohorts

The characteristics of the children in the mother–child cohorts are detailed in Table 26. The median days of follow-up varied within and between drug exposures from 657 days [interquartile range (IQR) 286–1351 days] in children of women who were prescribed antipsychotics in pregnancy (cohort B1) to 1197 days (IQR 396–1671 days) in children of women who were prescribed lithium in pregnancy (cohort B1). The median follow-up for children of women not prescribed psychotropic medications (cohort C) was around 740 days depending on the comparison cohorts. The male-to-female ratio varied between cohorts. For example, more females (55%) than males (45%) were born to women who continued antipsychotic treatment in pregnancy (cohort B1) (see Table 26). Major congenital malformations were recorded relatively soon after birth and most records of neurodevelopmental/behavioural disorders were made around the age of 2–3 years (see Table 26).

Maternal outcomes

Child outcomes