Women with no identified risk factors

Four RCTs met the eligibility criteria for this review: HARRISONHOHNER2001 (Harrison-Hohner et al., 2001), LLORENTE2003 (Llorente et al., 2003), MAKRIDES2010 (Makrides et al., 2010), MOKHBER2011 (Mokhber et al., 2011). All of these studies were published in peer-reviewed journals between 2001 and 2011. Further information about the included studies can be found in Appendix 18. All studies included sufficient data to be included in the statistical analysis. Of these, there were two studies (N=2,537) involving a comparison of omega-3 and placebo, one study (N=166) that compared selenium and placebo and one study (N=374) that compared calcium and placebo (see Table 264).

Table 264

Study information table for trials included in the meta-analyses of pharmacological interventions compared with placebo in women with no identified risk factors.

Women with identified risk factors

Two RCTs met the eligibility criteria for this review: HARRIS2002 (Harris et al., 2002), LAWRIE1998B (Lawrie et al., 1998b). Further information about included studies can be found in Appendix 18.

There was one study (N=180) that compared norethisterone with placebo and one study (N=446) involved a comparison between thyroxine and placebo (see Table 265). In one study participants had psychosocial risk factors (low income) and in one study participants were positive for thyroid antibodies (although this was not one of the pre-specified risk factors, women included in this study were at risk of postnatal depression, and therefore included in the review for risk factors).

Table 265

Study information table for trials included in the meta-analyses of any pharmacological intervention versus placebo comparison in women with identified risk factors.

Prophylaxis of mental health problems

Two RCTs met the eligibility criteria for this review: WISNER2001 (Wisner et al., 2001); WISNER2004B (Wisner et al., 2004b). In addition seven studies were excluded from the review. The reasons for exclusion were that the studies did not have a control group or were not RCTs. Further information about both included and excluded studies can be found in Appendix 18. One study compared TCAs (nortriptyline) with placebo, and one study compared SSRIs (sertraline) with placebo (see Table 266).

Table 266

Study information table for trials included in the meta-analyses of any pharmacological intervention versus placebo comparison for prophylaxis of mental health problems.

Depression outcomes (by intervention)

Omega-3 versus placebo

There was no evidence for clinically or statistically significant benefits (p=0.18-1.00) associated with omega-3 for mean depression scores, depression symptomology or diagnosis at endpoint or at intermediate follow-up (Table 267).

Table 267

Summary of findings table for effects of omega-3 compared with placebo on preventing depression outcomes in women with no identified risk factors.

Selenium versus placebo

There was low quality, single study (N=85) evidence in favour of a preventative benefit of selenium on reducing mean depression scores at endpoint however this effect did not reach statistical significance (p=0.07) and failed to reach a threshold indicative of clinically significant benefits (Table 268).

Table 268

Summary of findings table for effects of selenium compared with placebo on preventing depression outcomes in women with no identified risk factors.

Calcium versus placebo

The evidence for calcium as a preventative intervention was inconsistent (Table 269). There was low quality, single study (N=374) evidence for a moderate preventative benefit of calcium on depression symptomology ay endpoint, however this effect was not statistically significant (p=0.13) and there was very serious imprecision (due to the small event rate and the 95% CIs included both no effect and appreciable benefit). There was some discrepancy between dichotomous and continuous measures of depression at short term follow-up. There was moderate quality, single study (N=247) evidence for a large beneficial effect of calcium on preventing depression symptomology (p=0.02), however there was serious imprecision of this effect estimate due to the low number of events. In addition, there was no statistically or clinically significant preventive benefit on mean depression scores at short-term follow-up (p=0.13).

Table 269

Summary of findings table for effects of calcium compared with placebo on depression outcomes in women with no identified risk factors.

Compliance outcomes (by intervention)

Selenium versus placebo

There was low quality, single study (N=85) evidence for a large beneficial effect of selenium on compliance (as measured by serum selenium concentration) post-treatment (p<0.00001, Table 270). However, confidence that this is a true measure of the effect is low due to the small population size and unclear risk of selection bias (unclear method of randomisation and allocation concealment).

Table 270

Summary of findings table for effects of calcium compared with placebo on compliance outcomes in women with no identified risk factors.

Quality of life outcomes (by intervention)

Calcium versus placebo

There was no statistically or clinically significant benefit of calcium on positive (p=0.16) or negative (p=0.48) life events (Table 271).

Table 271

Summary of findings table for effects of calcium compared with placebo on quality of life outcomes in women with no identified risk factors.

Infant outcomes (by intervention)

Omega-3 versus placebo

There was no evidence for a statistically or clinically significant benefit of omega-3 on any of the of the Bayley scales of Infant and toddler development subscales (p=0.14-0.95) at long-term follow-up. There was moderate quality, single study (N=726) evidence for a statistically significant benefit on cognitive performance using an ITT analysis (p=0.05), however this effect was just under the threshold indicative of clinically significant benefits. There was no statistically or clinically significant effect on language performance (p=0.91, Table 272).

Table 272

Summary of findings table for preventative effects of omega-3 compared with placebo on infant outcomes in women with no identified risk factors.

Leaving the study early for any reason (by intervention)

Omega-3 versus placebo

There was no evidence for a statistically or clinically significant benefit of omega-3 on leaving the study early for any reason (p=0.25, Table 273)

Table 273

Summary of findings table for effects of omega-3 compared with placebo on leaving the study early in women with no identified risk factors.

Adverse events and service utilisation (by intervention)

Omega-3 versus placebo

There was moderate quality, single study (N=2,399) evidence for moderate beneficial effect (p=0.04) of omega-3 on preventing infant admission to neonatal intensive care (Table 274). However, the imprecision of this effect estimate was serious due to the small number of events. There were no statistically or clinically significant differences between omega-3 and placebo on maternal hospitalisation for serious adverse events (p=1.00) or major congenital abnormalities of the infant at long-term follow-up (p=0.43).

Table 274

Summary of findings table for effects of omega-3 compared with placebo on adverse events and service utilisation.

Depression outcomes (by intervention)

Thyroxine versus placebo

There was no evidence for a statistically or clinically significant benefit (p=0.44-0.98) of thyroxine on depression symptomology or diagnosis at the end of intervention (Table 275).

Table 275

Summary of findings table for effects of thyroxine compared with placebo on depression outcomes in women with identified risk factors.

Norethisterone compared with placebo

There was moderate quality, single study (N=163) evidence for a non-beneficial effect of norethisterone on depression outcomes at the end of intervention (Table 276). There was a statistically significant effect on mean depression scores favouring the placebo group compared to the norethisterone group (p=0.004), although this effect failed to reach a threshold indicative of clinically significant benefits. There was a moderate effect favouring placebo on depression symptomology (p=0.01), however there was serious imprecision (due to the small sample size). Moreover, this effect was not maintained at short-term follow-up, with no statistically or clinically significant difference in effect on mean depression scores or symptomology.

Table 276

Summary of findings table for effects of norethisterone compared with placebo on adverse events.

Compliance outcomes (by intervention)

Thyroxine versus placebo

There was no evidence for a statistically or clinically significant benefit of thyroxine on compliance post-treatment (p=0.44, Table 277).

Table 277

Summary of findings table for effects of thyroxine compared with placebo on adverse events.

Mother-infant interaction outcomes (by intervention)

Norethisterone versus placebo

There was no evidence for a statistically or clinically significant benefit of norethisterone on breastfeeding outcomes at the end of intervention (p=0.30) or at short-term follow-up (p=0.28, Table 278).

Table 278

Summary of findings table for effects of norethisterone compared with placebo on mother-infant interaction outcomes.

Leaving the study early (by intervention)

Norethisterone versus placebo

There was low to moderate quality, single study (N=180) evidence for large beneficial effect of norethisterone on leaving the study early at the end of intervention (p=0.03) and short-term follow-up (p=0.09), however the imprecision of this effect estimate was serious due to the small population and the 95% CIs were wide (Table 279).

Table 279

Summary of findings table for effects of omega-3 compared with placebo on adverse events.

Adverse event outcomes (by intervention)

Norethisterone versus placebo

There was low quality evidence for a moderate to large effect on days of vaginal bleeding in favour of placebo compared with norethisterone at the end of intervention (p<0.0001) and short-term follow-up (p<0.0001), and for troublesome bleeding at the end of intervention (p=0.002), however the imprecision of these effect estimates was serious due to the small population and number of events (Table 280). There was no statistically or clinically significant effect of norethisterone on return of sexual interest.

Table 280

Summary of findings table for effects of omega-3 compared with placebo on adverse events.

Recurrence of depression outcomes (by intervention)

SSRIs (sertraline) versus placebo

There was low quality, single study (N=22) evidence for a large beneficial effect of SSRIs on preventing recurrence of depression at post-treatment (p=0.06). However, the imprecision of this effect estimate was very serious due to the very small population size and large 95% CIs (Table 281).

Table 281

Summary of findings table for effects of SSRIs (sertraline) compared with placebo on depression outcomes.

TCAs (nortriptyline) versus placebo

There was no evidence for a statistically or clinically significant benefit of nortriptyline on recurrence of depression at post-treatment (p=0.94) or long-term follow-up (p=0.63, Table 282)

Table 282

Summary of findings table for effects of TCAs (nortriptyline) compared with placebo on depression outcomes.

Adverse events outcomes (by intervention)

SSRIs (sertraline) versus placebo

There was very low quality, single study (N=22) evidence for a statistically significant increased risk drowsiness with SSRIs (sertraline, p=0.002), however the imprecision of this effect estimate was very serious due to the very small population size and large 95% CI (Table 283). There was no evidence for an effect of SSRIs (sertraline) on dizziness.

Table 283

Summary of findings table for effects of SSRIs (sertraline) compared with placebo on adverse events.

TCAs (nortriptyline) versus placebo

There was low quality, single study evidence (N=51) for a large effect of nortriptyline on the number of participants reporting constipation at post-treatment (p<0.01), however imprecision of this effect estimate was very serious due to the small population size and large 95% CI (Table 284). There was no statistically or clinically significant effect of nortriptyline on discontinuation due to adverse effects at post-treatment (p=0.48).

Table 284

Summary of findings table for effects of TCAs (nortriptyline) compared with placebo on adverse events.

Leaving the study early (by intervention)

SSRIs (sertraline) versus placebo

There was no statistically or clinically significant difference between of SSRIs (sertraline) and placebo on leaving the study early for any reason except for recurrence at post-treatment (p=0.17, Table 285).

Table 285

Summary of findings table for effects of SSRIs (sertraline) compared with placebo on leaving the study early.

TCAs (nortriptyline) versus placebo

There was statistically or clinically significant difference between TCAs (nortriptyline) and placebo on leaving the study early for any reason except for recurrence at post-treatment (p=0.63,)

Table 286

Summary of findings table for effects of TCAs (nortriptyline) compared with placebo on adverse events.

Systematic literature review

No studies assessing the cost effectiveness of pharmacological interventions for the prevention of mental health problems in pregnancy or the postnatal period were identified by the systematic search of the economic literature undertaken for this guideline. Details on the methods used for the systematic search of the economic literature are described in Chapter 3.

Non-response to treatment (by intervention)

Omega-3 versus placebo

There was very low quality, single study (N=36) evidence for moderate beneficial effects of omega-3 on response to treatment from both an available case and an ITT analysis approach at endpoint (Table 290). However these effects did not reach statistical significance (p=0.09-0.11) and there was very serious imprecision due to the small number of participants and 95% CIs including estimates of no effect and clinically meaningful benefit. There was no statistically or clinically significant benefit of omega-3 on non-remission using either an available case or an ITT (WCS) analysis approach.

Table 290

Summary of findings table for treatment effects of omega-3 versus placebo on response outcomes.

SSRIs (sertraline/paroxetine) versus placebo

There were mixed results for treatment effects on response outcomes associated with SSRIs (Table 291). Adopting an available case analysis approach, there was very low quality, single study evidence (N=33) for a large benefit of SSRIs (sertraline) on non-response at endpoint (p=0.05), however there was very serious imprecision due to the small number of participants and events. Using an ITT (LOCF) analysis very low quality evidence from two studies (N=106) found no statistically significant effect on non-remission (p=0.28) although the effect just met the threshold for a clinically appreciable benefit. There was low to very low quality evidence for a statistically significant and moderate effect of SSRIs on non-remission at endpoint using both an available case (p=0.05) and an ITT (LOCF, p=0.04) analysis, however the quality of evidence was very low due to serious imprecision and high risk of attrition bias.

Table 291

Summary of findings table for treatment effects of SRRIs compared with placebo on response outcomes.

SSRIs in combination with psychological interventions compared with placebo in combination with psychological interventions

There was low quality, single study (N=42) evidence for a moderate effect of SSRIs combined with brief dynamic psychotherapy on response and remission using an ITT (LOCF/WCS) analysis approach (Table 292), however this was not statistically significant (p=0.2-0.22) and the confidence in the estimate was low due to number of events being less than 300 and 95% CI crosses both line of no effect and measure of appreciable benefit or harm.

Table 292

Summary of findings table for effects of SRRIs in combination with psychosocial interventions compared with placebo in combination with psychosocial interventions on response outcomes.

SSRIs versus TCAs

There were inconsistent results for response outcomes associated with SSRIs compared with TCAs. There was no evidence of a statistically or clinically significant effect of SSRIs compared with TCAs on non-response or non-remission using an ITT (LOCF) analysis approach at post-treatment (Table 293). At intermediate follow-up there was a large effect in favour of TCAs on response using an available case analysis, however the confidence in this effect estimate is very low due to very serious imprecision (small event rate and the 95% CI included both no effect and appreciable benefit).

Table 293

Summary of findings table for effects of SRRIs compared with TCAs on response outcomes.

Depression outcomes (by intervention)

SSRIs versus placebo

There was very low quality, single study (N=31) evidence for a moderate beneficial effect of SSRIs (paroxetine) on mean depression scores at the end of intervention using an available case analysis (p=0.10, Table 294). However, the quality of this evidence was very low due to very serious imprecision (with small number of participants and 95% CIs including estimates of no effect and clinically meaningful benefit) and a high risk of attrition bias.

Table 294

Summary of findings table for effects of SRRIs compared with placebo on depression outcomes.

SSRIs versus TCA

There was no evidence for a statistically significant benefit of SSRIs compared with TCAs on mean depression scores using an available analysis approach at post-treatment or at intermediate follow-up (p=0.6-0.88, Table 295).

Table 295

Summary of findings table for effects of SRRIs compared with TCAs on depression outcomes.

SSRIs in combination with psychological interventions compared with placebo in combination with psychological interventions

There was low quality evidence for a moderate beneficial effect of SSRIs combined with psychosocial interventions on mean depression scores post-intervention using both an available case (p=0.03) and an ITT (LOCF, p=0.02) analysis (Table 296). However the quality of this evidence was low due to serious imprecision (with small number of participants) and high and unbalanced attrition rates.

Table 296

Summary of findings table for effects of SRRIs in combination with psychosocial interventions compared with placebo in combination with psychosocial interventions on depression outcomes.

Antidepressants versus general supportive care

There was very low quality, single study (N=254) evidence for a moderate beneficial effect of antidepressants on depression symptomology at post-treatment using both an available case (p=0.0001) and an ITT (P= 0.0006) analysis (Table 297). There was also a statistically significant beneficial effect favouring antidepressants on mean depression scores using an available case analysis (p=0.0004). However the quality of evidence was very low due to high risk of bias and serious imprecision.

Table 297

Summary of findings table for effects of antidepressants compared with placebo in combination with general supportive care on depression outcomes.

Omega-3 versus placebo

There was no evidence for a statistically or clinically significant effect of omega-3 on mean depression scores using an ITT analysis approach at the end of intervention (Table 298), however there was substantial heterogeneity between the effect sizes of the four studies.

Table 298

Summary of findings table for effects of omega-3 compared with placebo on depression outcomes.

Hormones (transdermal oestrogen) versus placebo

There was moderate quality, single study (N=64) evidence for a large beneficial effect of hormones (transdermal oestrogen) on mean depression scores using an available case analysis (p<0.001) and on symptomology using an ITT analysis (p<0.0007) at the end of intervention (Table 299). However there was serious imprecision due to the small number of participants and events.

Table 299

Summary of findings table for treatment effects of hormones compared with placebo on depression outcomes.

General mental health outcomes (by intervention)

SSRIs versus TCAs

There was low quality, single study (N=29) for a moderate effect in favour of SSRIs on global severity and improvement symptomology at endpoint using an available case analysis (Table 300). However this effect estimate is low due to very serious imprecision (very small event rate and the 95% CI included both no effect and appreciable benefit, p=0.72). There was no statistically or clinically significant evidence in any effect of SSRIs compared with TCAs on all other general mental health outcomes using an available case analysis at the end of intervention or at intermediate follow-up (p=0.69-0.93,).

Table 300

Summary of findings table for effects SSRIs compared with TCAs on general mental health outcomes.

SSRIs combined with psychosocial interventions versus placebo combined with psychosocial interventions

There was moderate quality, single study evidence (N=40) for a large beneficial effect of SSRIs combined with psychosocial interventions on mean global severity scores (p<0.01) using an ITT analysis post-intervention (Table 301). However there was no statistically or clinically significant benefit on mean global improvement, mean distress or mean well-being scores post-treatment (p=0.36-0.63).

Table 301

Summary of findings table for effects of SSRIs combined with psychological interventions compared with placebo combined with psychological interventions on general mental health outcomes.

SSRIs compared with placebo

There was low quality, single study evidence (N=31) for a large beneficial effect of SSRIs on mean global severity and improvement scores (p=0.02) using an available case analysis at the end of intervention (Table 302). However the precision was poor and there are risk of bias concerns with this study due to high rate of attrition.

Table 302

Summary of findings table for effects of SSRIs compared with placebo on general mental health outcomes.

Service utilisation outcomes

SSRIs combined with psychosocial interventions compared with placebo combined with psychosocial interventions (by intervention)

There was no evidence for a clinically or statistically significant benefit of SSRIs combined with psychosocial interventions relative to placebo combined with psychosocial interventions on lorazepam use post-treatment (p=0.34; Table 303).

Table 303

Summary of findings table for effects of SSRIs combined with psychological interventions compared with placebo combined with psychological interventions on service utilisation outcomes.

Omega-3 compared with placebo

There was low quality, single study (N=118) evidence for a moderate effect favouring placebo relative to omega-3 on antidepressant use post-treatment (p=0.31; Table 304). However the confidence in this effect is low due to poor precision (small population and number of events and the 95% CI crosses both line of no effect and measure of appreciable benefit or harm).

Table 304

Summary of findings table for effects of omega-3 compared with placebo on service utilisation outcomes.

SSRIs compared with placebo

There was low quality, single study (N=36) evidence for increased benzodiazepine use associated with SSRIs (sertraline) at the end of intervention (p=0.14; Table 305). However confidence in this effect is low due to very serious imprecision (the population size and number of events was low and the 95% CI crosses both line of no effect and measure of appreciable benefit or harm).

Table 305

Summary of findings table for effects of SSRIs compared with placebo on service utilization.

Leaving the study early (by intervention)

SSRIs combined with psychological interventions compared with Placebo combined with psychological interventions

There was low quality, single study (N=128) evidence for a large effect of leaving the study early due to adverse events in favour of SSRIs combined with psychological interventions (Table 306), however the imprecision is very serious due to very small number of events and the 95% CI crosses both line of no effect and measure of appreciable benefit or harm. There was no statistically or clinically significant effect on leaving the study early due to any other reasons.

Table 306

Summary of findings table for effects of SSRIs combined with psychological interventions compared with placebo combined with psychological interventions on leaving the study early.

SSRIs compared with placebo

There was no statistically or clinically significant effect of SSRIs on leaving the study early due to any reason at endpoint (Table 307).

Table 307

Summary of findings table for effects of SSRIs compared with placebo on leaving the study early.

SSRIs compared with TCAs

There was low quality, single study (N=109) evidence in favour of TCAs for leaving the study early (Table 308, p=0.06). However the quality of evidence is low due to very serious imprecision (small number of events and 95% CI crosses both line of no effect and measure of appreciable benefit or harm).

Table 308

Summary of findings table for effects of SSRIs compared with TCAs on leaving the study early.

Hormones versus placebo

There was low quality, single study (N=64) evidence for fewer participants leaving the study early for any reason in favour of oestradiol (Table 309, p=0.14), however the quality of evidence is low due to very serious imprecision (small number of events and 95% CI crosses both line of no effect and measure of appreciable benefit or harm).

Table 309

Summary of findings table for effects of hormones compared with placebo on service utilization.

Omega-3 versus placebo

There was low quality evidence from four studies (N=239) for a moderate beneficial effect of omega-3 on leaving the study early (Table 310, p=0.09). However the quality of evidence is low due to very serious imprecision (small number of events and 95% CI crosses both line of no effect and measure of appreciable benefit or harm).

Table 310

Summary of findings table for effects of SSRIs compared with placebo on service utilization.

Adverse events outcomes (by intervention)

SSRI combined with psychosocial interventions compared with placebo compared with psychosocial interventions

There was no statistically or clinically significant effect of SSRIs combined with psychological interventions on mean side effect scores (Table 311). There were two cases of hypomanic switching in the SSRIs combined with psychological intervention group and none in the placebo combined with psychological intervention group.

Table 311

Summary of findings table for treatment effects of SSRIs combined with psychological interventions compared with placebo combined with psychological interventions on adverse events.

Omega-3 versus placebo

There was no statistically or clinically significant effect of omega-3 on mild or transient side effects post-treatment (Table 312, p=0.64). There was one case of hypomanic side effects in the omega-3 group and one case of suicide in the placebo group.

Table 312

Summary of findings table for effects of SSRIs compared with placebo on service utilization.

SSRIs versus placebo

The evidence for effects of SSRIs on adverse events was very low (Table 313) due to very serious imprecision (very small number of events and 95% CI crosses both line of no effect and measure of appreciable benefit or harm), however there were moderate effects for decreased appetite and dizziness associated with SSRIs (p=0.65-0.3), and there was a large effect for dry mouth associated with SSRIs (p=0.14).

Table 313

Summary of findings table for effects of SSRIs compared with placebo on adverse events.

Systematic literature review

No studies assessing the cost effectiveness of pharmacological interventions for the treatment of mental health problems in pregnant and breastfeeding women were identified by the systematic search of the economic literature undertaken for this guideline. Details on the methods used for the systematic search of the economic literature are described in Chapter 3.

Inclusion criteria

Study design: cohort and case-control study designs were included in the review. Both ‘retrospective’ cohort studies, that is those that identify subjects from past records describing the exposure received and follows them from the time of those records, and ‘prospective’ cohort studies, that is those that recruit participants before any exposure and follows them into the future, were included.Case-control designs compare subjects who have an outcome, and look back retrospectively to compare how frequently an exposure is present in each group.

Comparison group: only studies with an unexposed comparison group were included in the review

Outcome data: in order to be included studies needed to report data on at least one of the malformations and at least one of the drugs of interest, and provide sufficient data to calculate an effect size. Congenital malformations are defined as any birth defect. Congenital malformations have traditionally been divided into categories of “major” and “minor”. In this review, and in accordance with the majority of the studies included, the term congenital malformations is used to refer to both major and minor malformations combined. Evidence for the subordinate group of major malformations was also analysed separately. There is no standard major-minor definition or distinction. However, commonly used distinguishing features include that a major malformation has an adverse effect on the individual's health, functioning or social acceptability while a minor malformation is generally considered to be of limited social or medical significance. Where studies made this major-minor distinction, data was extracted using the categorisation as defined in the paper. Where categorisation was not applied in the original paper, the malformations were categorised consistently with the other studies in the review.

Review criteria

The following criteria were considered when assessing studies reporting on harms associated with specific drugs in pregnancy:

Study design: results from different study designs are expected to differ systematically, resulting in increased heterogeneity. Therefore, cohort and case-control study designs were not combined in a meta-analysis, but conducted separately for each study design.

Comparison group: a distinction was made between disorder specific comparison groups, that is, studies which used as a comparison group, those who were unexposed to the drug of interest but had the same disorder as the exposed group, and a comparison group that consisted of women from the general population. Each study was used in only one analysis, and the disorder specific comparison group was prioritised where studies reported data for both. Where possible, subgroup analyses were conducted for disorder specific comparison groups.

Reporting on specific drugs: the class of drugs was used as a start point. The GDG decided to look at individual drugs where data existed and where there was reason to suspect that there may be an issue with an individual drug. However caution was taken in singling out individual drugs where there was limited data, in order to avoid making risky interpretations.

Timing of exposure: to maximise available data, results were pooled for studies reporting exposure during any trimester (however the majority of studies reported at least first trimester use). A sensitivity analysis for timing was not possible due to inconsistent reporting on specific trimester exposure.

Type of exposure: data for drugs taken in monotherapy were prioritised because this was most meaningful in terms of attributing the specific drug to harm, rather than the use of the drug in combination. However caution was taken when interpreting the data; for many mental health problems (for example bipolar disorder) taking drugs in combination is the norm.

Outcome reporting: the highest order class of harms was used as the main analysis, for instance, where studies report congenital malformations (all malformations), major malformations and minor malformations, primary review of outcomes would focus on congenital malformations as the superordinate class. Where there was a priori evidence for specific adverse events, these were reported, however it was noted that these were not independent of the main class of harms. For instance, in the case of antidepressants there is a priori evidence for septal defects and this evidence will be reviewed, but the GDG were mindful that the different classes of harms were not necessarily independent from each other so that in this example septal defects are a subgroup of cardiac malformations which form a subgroup for major malformations which form a subgroup for congenital malformations. Only studies where an appropriate definition of the class of harms was provided were included in the meta-analysis. Only outcomes which had more than one study or a substantial sample size (equivalent to the sample sizes in the multiple study meta-analyses) were included in the review.

Type of data: unadjusted, rather than adjusted data was used for the following reasons: there is considerable variability over what each study adjusts for; unadjusted data is most consistently reported and allows the maximisation of available data; the use of unadjusted data allows for absolute rates to be calculated from the raw event rates.

Statistical analysis: for dichotomous outcomes the effect size was reported as an odds ratio. The odds ratiois the measure of association between an exposure and an outcome, where the odds ratio represents the odds that an outcome will occur given a particular exposure, compared to the odds of the outcome occurring in the absence of that exposure (Szumilas, 2010). However the GDG were cautious of over-interpretation of odds ratios when the actual event rate is low. Therefore absolute risk for exposed and unexposed groups and the absolute risk differences were reported. The absolute risk refers to the actual event rate; the number of cases in each group divided by the total number in the group and multiplied by 1000, resulting in a value out of 1000. The absolute risk difference was calculated by working out the difference between the absolute rates in the exposed group compared to the unexposed group. The reason for including both odds ratios and absolute risks is that where the actual event rate is low in the general population, an effect may be statistically significant, however the absolute risk difference very low, which may lead to an over-interpretation of the effect size. It was not appropriate to calculate absolute risk values for studies using a case-control design because of the inflated prevalence of the cases in the population. Therefore, where possible, odds ratios were interpreted along-side the absolute values, which were used to inform the recommendations. Continuous outcomes were reported as standard mean differences.

Antidepressants

From the existing systematic reviews, 30 studies met the eligibility criteria for the review of antidepressants: BOUCHER2008 (Boucher et al., 2008), CALDERON-MARGALIT2009 (Calderon-Margalit et al., 2009), CASPER2003 (Casper et al., 2003), CHAMBERS1996 (Chambers et al., 1996), COSTEI2002 (Costei et al., 2002), DAVIS2007 (Davis et al., 2007) DIAV-CITRIN2008A (Diav-Citrin et al., 2008a), EINARSON2009 (Einarson et al., 2009), FERREIRA2007 (Ferreira et al., 2007), GALBALLY2009 (Galbally et al., 2009), KALLEN2004 (Kallen et al., 2004), KALLEN2007 (Kallen et al., 2007), KIELER2012 (Kieler et al., 2012), KORNUM2010 (Kornum et al., 2010), KULIN1998 (Kulin et al., 1998), LAINE2003 (Laine et al., 2003), LEVINSONCASTIEL2006 (Levinson-Castiel et al., 2006), MALM2011 (Malm et al., 2011), MASCHI2008 (Maschi et al., 2008), OBERLANDER2006 (Oberlander et al., 2006), OBERLANDER2008 (Oberlander et al., 2008), PEDERSEN2009 (Pedersen et al., 2009), RAI2013 (Rai et al., 2013), SIMON2002 (Simon et al., 2002), SIVOJELEZOVA2005 (Sivojelezova et al., 2005), SURI2007 (Suri et al., 2007), WEN2006 (Wen et al., 2006), WICHMAN2009 (Wichman et al., 2009), WISNER2009 (Wisner et al., 2009), WOGELIUS2006 (Wogelius et al., 2006). Six studies were included in the existing systematic reviews, however did not provide the relevant data for the current review, or reported on single study outcomes: ALWAN2007 (Alwan et al., 2007), BAKKER2010 (Bakker et al., 2010), CHAMBERS2006 (Chambers et al., 2006), EINARSON2008 (Einarson et al., 2008), RAMOS2008 (Ramos et al., 2008), WILSON2011 (Wilson et al., 2011). In addition 13 studies were excluded from the analysis as that did not meet the criteria for this review. Further information about both the included and excluded studies can be found in Appendix 18 and the full methodological checklists can be found in Appendix 17.

Risk of autism was not included as an adverse event in any of the systematic reviews identified, however the GDG felt this was an important outcome to consider. An additional search was therefore conducted for studies reporting on risk autism associated with antidepressant exposure in pregnancy. One study met the eligibility criteria for this review: ELMARROUN2014 (El Marroun et al., 2014). In addition four studies were excluded because they did not have a disorder specific comparison group. Table 314 provides summary information for studies included in the meta-analysis. Further information about both the included and excluded studies can be found in Appendix 18 and the full methodological checklists can be found in Appendix 17.

Table 314

Study information table for studiesincluded in the meta-analysis for adverse events associated with antidepressant exposure.

Antipsychotics

Of the eligible studies, 12 met the inclusion criteria. Due to the limited available evidence for antipsychotics, all studies were included in the meta-analysis (including those with single study outcomes), to maximise the available data: AUERBACH1992 (Auerbach et al., 1992), BODEN2012A (Boden et al., 2012a) BODEN2012B (Boden et al., 2012b), DIAV-CITRIN2005 (Diav-Citrin et al., 2005), HABERMANN2013 (Habermann et al., 2013), JOHNSON2012 (Johnson et al., 2012), LIN2010 (Lin et al., 2010), MCKENNA2005 (McKenna et al., 2005), NEWHAM2008 (Newham et al., 2008), PENG2013 (Peng et al., 2013) REIS2008 (Reis et al., 2008), SADOWSKI2013 (Sadowski et al., 2013).

Table 315 provides summary information for the studies included in the meta-analyses. In addition five studies were excluded from the review. The main reason for exclusion was that the studies did not have an unexposed control group. Further information about both the included and excluded studies can be found in Appendix 18 and the full methodological checklists can be found in Appendix 17. Two studies provided disaggregated data for first generation and second generation antipsychotics, however the GDG felt that there was generally very little drug specificity, therefore the analyses were conducted for all antipsychotics as a class.

Table 315

Study information table for studies included in the meta-analysis for adverse events associated with antipsychotic exposure.

Anticonvulsants

Of the eligible studies, there were 34 which met the inclusion criteria: ADAB2004/VINTEN2005 (Adab et al., 2004; Vinten et al., 2005), ARTAMA2005 (Artama et al., 2005), ARTAMA2013 (Artama et al., 2013), BODEN2012A, BORTHEN2011 (Borthen et al., 2011), BROSH2011 (Brosh et al., 2011), BURJA2006 (Burja et al., 2006), CANGER1999 (Canger et al., 1999), CASSINA2013 (Cassina et a., 2013), CHARLTON2011 (Charlton et al., 2011), CHRISTENSEN2013 (Christensen et al., 2013), CUNNINGTON2011 (Cunnington et al., 2011), DIAV-CITRIN2001 (Diav-Citrin et al., 2001), DIAV-CITRIN2008B (Diav-Citrin et al., 2008b), DOLK2008 (Dolk et al., 2008), ERIKSSON2005 (Eriksson et al., 2005), GAILY2004 (Gaily et al., 2004), HERNANDEZ-DIAZ2012 (Hernandez-Diaz et al., 2012), HOLMES2001 (Holmes et al., 2001), HOLMES2008 (Holmes et al., 2008), HVAS2000 (Hvas et al., 2000) JENTINK2010 (Jentink et al., 2010), KAAJA2003 (Kaaja et al., 2003), KANEKO1999 (Kaneko et al., 1999), KINI2007 (Kini et al., 2007), MOLGAARD-NIELSEN2011 (Molgaard-Nielsen et al., 2011), MORROW2006 (Morrow et al., 2006), ORNOY1996 (Ornoy et al., 1996), RIHTMAN2013 (Rihtman et al., 2013), RODRIGUEZ-PINILLA2000 (Rodriguez-Pinilla et al., 2000), SAMREN1999 (Samren et al., 1999), STEEGERS-THEUNISSEN1994 (Steegers-Theunissen et al., 1994), VAJDA2007 (Vajda et al., 2007), VEIBY2013 (Veiby et al., 2013), WERLER2011 (Werler et al., 2011). Summary information for the studies included in the meta-analysis can be found in Table 316. In addition 12 studies met the inclusion criteria however could not be included in the meta-analysis as the relevant data could not be extracted, the outcomes could not be combined or the data was not disaggregated for individual drug: ADAB2001 (Adab et al., 2001), ALMGREN2009 (Almgren et al., 2009), BROMLEY2013 (Bromley at al., 2013), CUNNINGTON2011, FONAGER2000 (Fonager et al., 2000), FORSBERG2011 (Forsberg et al., 2011), KAAJA2002 (Kaaja et al., 2002), KULAGA2011 (Kulaga et al., 2011), NULMAN1997 (Nulman et al., 1997), LIN2009 (Lin et al., 2009), RODRIGUEZ-PINILLA2008 (Rodriguez-Pinilla et al., 2008), THOMAS2008 (Thomas et al., 2008), VAJDA2004 (Vajda et al., 2004). In addition 25 studies were excluded from the review. The main reason for exclusion was that the studies did not have an unexposed control group. Data was disaggregated for carbamazepine, lamotrigine and valproate as the magnitude of risks and specific abnormalities varies for each anticonvulsant and have different properties. Further information about both the included and excluded studies can be found in Appendix 18 and the full methodological checklists can be found in Appendix 17.

Table 316

Study information table for studies included in the meta-analysis for adverse events associated with anticonvulsant exposure.

Lithium

There were six studies which met the inclusion criteria for the review: BODEN2012A, CORREA-VILLASENOR1994 (Correa-Villasenor et al., 1994), CZEIZEL1990 (Czeizel et al., 1990), JACOBSON1992 (Jacobson et al., 1992), KALLEN1983 (Kallen et al., 1983), REIS2008 (Reis et al., 2008). Summary information for the studies included in the meta-analyses can be found in Table 317. In addition four studies were excluded from the review, the reasons for exclusion were that the studies did not have an unexposed control group and that no cases of lithium exposure were found in case-control studies. Further information about the included and excluded studies can be found in Appendix 18 and the full methodological checklists can be found in Appendix 17.

Table 317

Study information table for studies included in the meta-analysis for adverse events associated with lithium exposure.

Benzodiazepines and related drugs¹⁷

There were 18 studies which met the inclusion criteria, however only nine studies provided sufficient data to be included in the meta-analysis: BAN2014 (Ban et al., 2014), CZEIZEL1987 (Czeizel et al., 1987), LAEGREID1990 (Laegreid et al., 1990), LAEGREID1992 (Laegreid et al., 1992), LEPPEE2010 (Leppee et al., 2010), OBERLANDER2008 (Oberlander et al., 2008), ORNOY1998 (Ornoy et al., 1998), PASTUSZAK1996 (Pastuszak et al., 1996), WIKNER2007 (Wikner et al., 2007). Nine studies met the criteria but were not included in the meta-analysis because no relevant data could be extracted or the study only reported single outcomes: BONNOT2001 (Bonnot et al., 2001), CORREA-VILLASENOR1994 (Correa-Villasenor et al., 1994), CZEIZEL1999 (Czeizel et al., 1999), CZEIZEL2003 (Czeizel et al., 2003), CZEIZEL2004 (Czeizel et al., 2004), DIAV-CITRIN1999 (Diav-Citrin et al., 1999), EROS2002 (Eros et al., 2002), KJAER2007 (Kjaer et al., 2007), WANG2010 (Wang et al., 2010). A summary of the studies included in the meta-analysis can be found in Table 318. One study (BAN2014) was in press at the time of the review. Further information about the included and excluded studies can be found in Appendix 18 and the full methodological checklists can be found in Appendix 17.

Table 318

Study information for studies included in the meta-analyses for benzodiazepines and related drugs.

Stimulants

Of the eligible studies, only one met the inclusion criteria: POTTEGARD2014 (Pottegard et al., 2014). In addition four studies were excluded from the review as they did not have an unexposed control group. Summary information for this study can be found in Table 319. Further information about the included and excluded studies can be found in Appendix 18 and the full methodological checklists can be found in Appendix 17.

Table 319

Study information for studies included in the meta-analyses for stimulants.

Teratogenic harms

The results of the meta-analyses for antidepressants split by individual drug are summarised for congenital malformations (Table 320), major congenital malformation (Table 321), cardiac malformations (Table 322) and septal heart defects (Table 323).

Table 320

Summary of findings table for effects of exposure to antidepressants in pregnancy compared with no exposure to antidepressants on congenital malformations.

Table 321

Summary of findings table for effects of exposure to antidepressants in pregnancy compared with no exposure to antidepressants on major congenital malformations.

Table 322

Summary of findings table for effects of exposure to antidepressants in pregnancy compared with no exposure to antidepressants on cardiac malformations.

Table 323

Summary of findings table for effects of exposure to antidepressants compared with no exposure to antidepressants on septal defects (including both atrial septal defects and/or ventral septal defects).

There was some evidence for a statistically significant association between all SSRIs and congenital malformations (p=0.04) with an absolute risk difference of 9 more per 1000. The association between major congenital malformations and all SSRIs was not statistically significant, however the absolute risk difference was 12 more per 1000. Paroxetine was statistically associated with congenital (p=0.05), major congenital (p=0.04) and cardiac (p=0.006) malformations, and fluoxetine with major congenital (p=0.008) and cardiac (p=0.02) malformations with absolute risk differences ranging from 3 to 8 more per 1000. There was some evidence for a statistically significant association between citalopram and escitalopram and ventral septal defects with absolute risk difference of 4 and 9 more per 1000, respectively. It is noteworthy that the association between congenital malformations and TCAs was in favour of the exposed group (absolute risk difference, 20 fewer per 1000), however the baseline rate in the unexposed group was unexpectedly high (137 per 1000).

Course of pregnancy, obstetric and neonatal complications

The results of the meta-analysis for antidepressants split by individual drug are summarised in Table 324. There was some evidence for a statistically significant association between SSRIs in late pregnancy and persistent pulmonary hypertension (p=0.00001), but the actual risk difference was low with only 2 more per 1000 in the SSRI exposed group. However, larger effect sizes were found for an association between any antidepressant and poor neonatal adaptation syndrome, respiratory distress and tremor with absolute risk differences ranging from 34 more to 333 more per 1000. There was also some evidence for greater risk of preterm delivery (17 more per 1000) and miscarriage (12 more per 1000) associated with the SSRI group.

Table 324

Summary of findings table for effects of exposure to antidepressants compared with no exposure to antidepressants on obstetric and neonatal complications.

Neurodevelopmental outcomes

There was limited evidence for long-term neurodevelopmental outcomes associated with antidepressants. Risk of autism was not considered in the existing systematic reviews, and additionally searched for. There is evidence that parental mental health problems are themselves associated with autism spectrum disorders in the offspring (Daniels et al., 2008), therefore only studies which used a disorder specific comparison group were included. One study (ELMARROUN2014) found children prenatally exposed to SSRIs had more autistic traits (B= 0.15 [0.08, 0.22]) and were at a higher risk for developing pervasive developmental problems, OR=1.91 (1.31, 3.47) but not affective problems compared with children who were only exposed to depressive symptoms in pregnancy.

Teratogenic harms

A summary of the meta-analyses for major congenital malformations and congenital malformations is found in Table 325. There was some evidence for a statistically significant association between antipsychotics and congenital and major congenital malformations, with absolute risk differences of 36 more and 13 more per 1000, respectively. When restricting the analysis to one study where the comparison group had a disorder specific comparison group (bipolar disorder), the effect size remained similar, although was no longer statistically significant.

Table 325

Summary of findings table for effects of exposure to antipsychotics compared with no exposure to antipsychotics on congenital and major congenital malformations.

Course of pregnancy, obstetric and neonatal complications

The results of the meta-analysis for course of pregnancy, obstetric and neonatal complications are summarised in Table 326. There was some evidence for a statistically significant association between antipsychotics and gestational diabetes with an absolute risk difference of 19 more per 1000. However the association was no longer statistically significant and the risk difference reduced to only 1 more per 1000 with a disorder specific comparison, although the sample size was substantially smaller. There was evidence for a significant association between antipsychotics and small for gestational age and low birthweight babies, with large absolute risk differences. However when the control group had a psychiatric diagnosis, the association for small for gestational age was no longer statistically significant and the risk difference reduced. There was evidence for a statistically significant association with preterm delivery and caesarean section with large absolute risk differences of 57 and 103 more per 1000, respectively.

Table 326

Summary of findings table for effects of exposure to antipsychotics compared with no exposure to antipsychotics on course of pregnancy, obstetric and neonatal complications.

Neurodevelopmental complications

The results from the meta-analysis for neurodevelopmental outcomes are summarised in Table Table 327. There was no evidence for a statistically significant difference between the exposed and unexposed groups on the mean score or delayed development on the Bayley scales of Infant and Toddler development at 52 weeks from a single study. There was a single study evidence for statistically significant effect in favour of the unexposed group on the Infant Neurological Battery, however the GDG were mindful that the impact of maternal mental health on the long term development of the infant or child was not adjusted for.

Table Table 327

Summary of findings table for effects of exposure to antipsychotics compared with no exposure to antipsychotics on neurodevelopmental outcomes.

Teratogenic harms

The results of the meta-analysis for congenital and major congenital malformations are summarised in Table 328 and for specific teratogenic malformations in Table 329. There was some evidence for a statistically significant association between carbamazepine and congenital malformations and major congenital malformations with absolute risk differences of 62 more and 15 more per 1000. This remained significant when performing a sensitivity analysis for studies with a disordered comparison. The results from the meta-analysis suggested an event rate of 3.5% for major malformations, broadly in line with registry data event rates which range from 2.6% to 5.6%. There was no evidence for a statistically significant association between lamotrigine and major congenital malformations. The absolute risk from the meta-analysis suggested an event rate of 2.8% also in line with existing registry data. There was strong evidence for a statistically significant association between valproate and congenital and major congenital malformations, with a risk difference 20 more per thousand (50 more per thousand when using a disordered comparison). The event rate from the meta-analysis suggests a prevalence of 7.7%, broadly in line with registry data which ranges from 6.7% to 9.7%. There was evidence for a statistically significant association with cleft lip and/or palate and valproate (absolute risk difference 11 more per 1000), and for carbamazepine, although the absolute risk difference was low (4 more per 1000). There was also evidence for a statisitacally significant association between neural tube defects and valproare (absolute risk difference 11 more per 1000), which was not found for either carbamazepine or lamotrigine.

Table 328

Summary of findings table for effects of exposure to anticonvulsants compared with no exposure to anticonvulsants on congenital and major congenital malformations.

Table 329

Summary of findings table for effects of exposure to anticonvulsants compared with no exposure to anticonvulsants on specific teratogenic malformations.

Course of pregnancy, obstetric and neonatal complications

The results of the meta-analysis for course of pregnancy, obstetric and neonatal complications are summarised in Table 330. There was limited evidence for neonatal and obstetric complications, however the data suggested no statistically or clinically significant evidence for an increased risk of still birth or perinatal death with carbamazepine. There was an increased risk of preterm birth and carbamazepine but this was not statistically significant. There was limited evidence for neonatal and obstetric complications associated with lamotrigine, but available data suggests there does not appear to be any increased risks. There was some evidence for increase in preterm birth for valproate, although not statistically significant.

Table 330

Summary of findings table for effects of exposure to anticonvulsants compared with no exposure to anticonvulsants on course of pregnancy, obstetric and neonatal complications.

Neurodevelopmental outcomes

The results of the meta-analysis for neurodevelopmental outcomes are summarised in Table 331. The data did not suggest evidence for an increased risk of longer-term neurodevelopmental complications with carbamazepine or lamotrigine. However, there was evidence for a statistically significant association with valproate and low IQ (particularly verbal IQ), and also with autism at 9 year follow-up.

Table 331

Summary of findings table for effects of exposure to anticonvulsants compared with no exposure to anticonvulsants on neurodevelopmental outcomes.

Teratogenic harms

The results of the meta-analysis for teratogenic harms are summarised in Table 332. There was limited evidence for lithium due to the small number of studies which provided extractable data. There was some evidence for a statistically significant increase for congenital malformations, however the absolute risk reduction was only 7 more per 1000. Rates of Ebstein's anomaly have previously been associated with lithium exposure. Two studies reporting on Ebstein's anomaly met the inclusion criteria for our review; however, estimates were unstable because of the low number of events, [1 in 20,000 in the general population (Cohen et al., 1994)]. This was similarly found in a recent systematic review of lithium safety which analysed six case-control studies (N=264) and measured the association between Ebstein's anomaly and lithium (McKnight et al., 2012). They found the odds of exposure to lithium did not differ significantly from controls; however, estimates were unstable because of the low number of events.

Table 332

Summary of findings table for effects of exposure to lithium compared with no exposure to lithium on teratogenic harms.

Course of pregnancy, obstetric and neonatal complications

There was insufficient evidence for course of pregnancy, neonatal and obstetric complication outcomes.

Neurodevelopmental outcomes

There was insufficient evidence for neurodevelopmental outcomes.

Teratogenic harms

The results of the meta-analysis for teratogenic harms are summarised in Table 333. The data did not suggest an increased risk of congenital, major congenital or cardiac malformations and benzodiazepines. Data from one cohort study and two case-control studies did not suggest an association with cleft lip or cleft palate.

Table 333

Summary of findings table for effects of exposure to benzodiazepines in pregnancy compared with no exposure to benzodiazepines on teratogenic harms.

Course of pregnancy, obstetric and neonatal complications

The results of the meta-analysis for course of pregnancy, obstetric and neonatal complications are summarised in Table 334. There was some evidence for an increased risk of caesarean delivery and miscarriage and some evidence of an increased risk of respiratory disorder.

Table 334

Summary of findings table for effects of exposure to benzodiazepines compared with no exposure to benzodiazepines on course of pregnancy, obstetric and neonatal complications.

Benzodiazepines: neurodevelopmental outcomes

There was insufficient evidence for neurodevelopmental outcomes.

Teratogenic harms

The results of the meta-analysis for teratogenic harms are summarised in Table 335. There was no statistically or clinically meaningful association betweenmethylphenidate and congenital and major congenital malformations.

Table 335

Summary of findings table for effects of exposure to stimulants compared with no exposure to stimulants on course of pregnancy, obstetric and neonatal complications.

Course of pregnancy, obstetric and neonatal complications

There was insufficient evidence for course of pregnancy, obstetric and neonatal complication outcomes.

Neurodevelopment outcomes

There was insufficient evidence for neurodevelopmental outcomes.

Physical activity versus treatment as usual

There was low quality, single study (N=74) evidence for a large beneficial preventative effect of physical activity on mean depression scores at the end of the intervention (p=0.0006, Table 338). In addition, there was low quality, single study (N=737) evidence for a large preventative effect of physical activity on depression symptomology (above threshold), p=0.16. However there was very serious imprecision due to the small number of events and the 95% CI included both no effect and the measure of appreciable benefit.

Table 338

Summary of findings tables for the preventative effects of physical interventions on depression outcomes.

Physical activity combined with psychoeducation versus psychoeducation

There was no statistically or clinically significant effect of physical activity combined with psychoeducation on mean depression scores (p=0.17) at the end of intervention from low quality, single study (N=135) evidence (Table 339). However there was a trend (p=0.06) towards a preventative beneficial effect at short term follow-up using an ITT (LOCF) analysis, however the effect size failed to reach the threshold for a measure of clinically appreciable benefit.

Table 339

Summary of findings tables for the effects of physical interventions on preventing depression outcomes in women who are pregnant or postpartum.

Systematic literature review

No studies assessing the cost effectiveness of physical interventions for the prevention of mental health problems in pregnancy or the postnatal period were identified by the systematic search of the economic literature undertaken for this guideline. Details on the methods used for the systematic search of the economic literature are described in Chapter 3.

Acupuncture versus placebo acupuncture

There was no statistically or clinically significant effect of acupuncture on mean anxiety scores (p=0.14) or cortisol levels (p=1.00) at the end of intervention (Table 341).

Table 341

Summary of findings tables for the effects of acupuncture on preventing anxiety outcomes in women who are pregnant or postpartum.

Response outcomes (by intervention)

Acupuncture versus massage

There was no statistically or clinically significant difference in effect for acupuncture compared with massage on depression outcomes at post-treatment (p=0.27, Table 346).

Table 346

Summary of findings tables for treatment effects of acupuncture versus massage on response outcomes.

Depression specific acupuncture versus non-depression specific acupuncture

There was very low quality evidence from two studies (N=121) for a moderate beneficial effect of depression-specific acupuncture post-treatment (p=0.009, Table 347). However, the confidence in this estimate was very low due to serious imprecision (small number of events) and risk of bias in several domains.

Table 347

summary of findings tables for effects of depression-specific acupuncture versus non-depression-specific acupuncture on response outcomes.

Bright light therapy versus placebo

The results for response to treatment for bright light therapy were inconsistent. There was very low quality, single study (N=27) evidence for a large beneficial effect on response (p=0.06) and remission (p=0.10) to treatment at the end of intervention using an available case-analysis, however the effect was not statistically significant (Table 348). Moreover, the confidence in this estimate was very low due to serious imprecision (small number of events and the 95% CI included both no effect and measure of appreciable benefit) and risk of bias in several domains.

Table 348

Summary of findings tables for treatment effects of bright light therapy versus placebo on response outcomes.

Depression outcomes (by intervention)

Physical activity versus treatment as usual

There was no evidence for a statistically or clinically meaningful effect of physical activity on mean depression scores at the end of intervention (p= 0.11), although the effect favoured physical activity compared with control

Table 349

Summary of findings tables for treatment effects of physical interventions versus treatment as usual on depression outcomes.

Physical activity versus mutual support

There was very low quality, single study (N=19) evidence for a large beneficial effect of physical activity compared with mutual support on mean depression scores at post-treatment (p=0.04) and at short-term follow-up (p=0.03, Table 350). However, the confidence in this estimate was very low due to serious imprecision (very small population size) and risk of bias in several domains.

Table 350

Summary of findings tables for treatment effects of physical interventions versus mutual support on depression outcomes.

Acupuncture versus massage

There was no statistically or clinically significant difference in effect for acupuncture (depression and non-depression specific acupuncture combined) compared with massage on mean depression scores at post-treatment or short term follow-up (Table 351). There was very low quality evidence for a moderate beneficial effect of acupuncture compared with massage on depression diagnosis at short term follow-up (p=-0.31), but this was not statistically significant and the confidence in the estimate of the effect is low due to very serious imprecision (low number of events and the 95% CIs included both no effect and a measure of appreciable benefit.

Table 351

Summary of findings tables for treatment effects of acupuncture versus massage on depression outcomes.

Depression-specific acupuncture versus non-depression specific acupuncture

There was no statistically or clinically significant difference between depression-specific acupuncture and non-depression specific acupuncture on mean depression scores at post-treatment or short term follow-up (Table 352). However there was very low quality, single study (n=35) evidence for a moderate to large effect in the favour of depression-specific acupuncture on depression diagnosis at the end of intervention (p=0.33) and at short term follow-up (p=0.71), however these effects were not statistically significant and confidence in this estimate is very low due to very serious imprecision (very small number of events and the 95% CI crosses both the line of no effect and measure of appreciable benefit or harm).

Table 352

Summary of findings tables for treatment effects of depression-specific acupuncture versus non-depression specific acupuncture on depression outcomes.

Electroacupuncture versus non-invasive sham acupuncture

There was no statistically or clinically significant effect for electroacupuncture on mean depression scores at post-treatment (p=0.65, Table 353).

Table 353

Summary of findings tables for treatment effects of electroacupuncture versus non-invasive sham acupuncture on depression outcomes.

Massage combined with support versus support

There was very low quality, single study (N=25) evidence for a large beneficial effect of massage combined with support compared with support alone on mean depression scores post-treatment using an available case analysis (p=0.005, Table 354). However the confidence in this estimate was very low due to serious imprecision (very small population size) and there was a risk of bias in several domains.

Table 354

Summary of findings tables for treatment effects of electroacupuncture versus non-invasive sham acupuncture on depression outcomes.

Massage versus support

There was no statistically or clinically significant effect of massage compared with support on mean depression scores at post-treatment (p=0.20) or short term follow-up (p=0.70, Table 355).

Table 355

Summary of findings tables for treatment effects of massage versus support on depression outcomes.

Bright light therapy versus placebo

Although there was a trend towards a beneficial effect of bright light therapy on mean depression symptoms, it was not statistically or clinically significant at post-treatment as measured by atypical depression supplement score (p=0.26) or the HRDS (p=0.76, Table 356) and the quality of evidence was very low due to serious imprecision and risk of bias.

Table 356

Summary of findings tables for treatment effects of massage versus support on depression outcomes.

Anxiety outcomes (by intervention)

Physical activity versus control

There was no statistically or clinically significant effect of [physical activity on mean anxiety scores at post-treatment (p=0.43, Table 357).

Table 357

Summary of findings table for the effects of physical interventions on anxiety in pregnancy and the postnatal period.

Electroacupuncture versus non-invasive sham acupuncture

There was no statistically or clinically significant effect of electroacupuncture on mean anxiety scores at post-treatment (p=0.96, Table 358).

Table 358

Summary of findings table for the effects of electroacupuncture on anxiety in pregnancy and the postnatal period.

General mental health outcomes (by intervention)

Physical activity versus control

There was low quality, single study (N=75) evidence for a statistically significant (p=0.05) beneficial effect of physical activity on mean sleep disturbance score at post- treatment, however the effect size failed to reach a threshold indicative of clinically significant benefits (Table 359). In addition the quality of evidence was low due to the serious imprecision (small sample size) and unclear risk of bias in several domains.

Table 359

Summary of findings table for the effects of physical interventions on anxiety in pregnancy and the postnatal period.

Systematic literature review

No studies assessing the cost effectiveness of interventions for the treatment of mental health problems in pregnancy or the postnatal period were identified by the systematic search of the economic literature undertaken for this guideline. Details on the methods used for the systematic search of the economic literature are described in Chapter 3.

Systematic literature review

No studies assessing the cost effectiveness of ECT for women with mental health problems in pregnancy or the postnatal period were identified by the systematic search of the economic literature undertaken for this guideline. Details on the methods used for the systematic search of the economic literature are described in Chapter 3.

Antidepressants (TCAs, SSRIs, (S)NRIs)

In reviewing the evidence and developing the recommendations on the harms associated with antidepressant use in pregnancy, the GDG was mindful of the serious nature of the outcomes reviewed, which could have a profound effect on the life course of any individual who is born with a major congenital defect. They were also concerned about the potentially increased rate of a number of the outcomes considered in women with depression who had not been exposed to antidepressant drugs. The GDG were cautious when it came to interpreting the data on individual drugs given the variation in the size of the datasets. Finally, although absolute rate differences were small in most cases, the GDG was aware of the high level of prescribing of antidepressant drugs and the potential impact on a large number of women and babies.

The GDG agreed that there was a small but significant increase in a number of congenital abnormalities (in particular cardiac abnormalities) for a number of important outcomes. However, because of the limitations of the comparator groups (not all contained women with a depressive disorder and where this was the case it was not often known if the severity of the disorder in the comparator group was similar), the GDG was uncertain whether all of this increase could be accounted for by the drug. The GDG considered the possibility that the potential harms arising for the women having the mental health problem may possibly increase as the severity of the depressive disorder increased. The GDG also took into consideration that for many of these women there may be a prior history of depression and this may also be used to guide prescribing practice. Given the considerable uncertainty surrounding the evidence, the GDG adopted a cautious approach in developing its recommendations and also took account of what is known about the effective treatment of depression in non-pregnant women. No data were available to the GDG on the cost effectiveness or impact on resource use of the interventions considered in pregnancy. However, the guideline meta-analysis of clinical evidence points to similar levels of harms across the antidepressants reviewed. Most of the drugs reviewed are off patent and available in generic form. In the case of newer drugs the lack of any greater effect than older drugs makes the added cost potentially not worthwhile. Again the GDG took into account what is known about the cost effectiveness of treatment of depression in non-pregnant women. The GDG also took into account that many women (up to 90%) stop taking medication when they discover that they are pregnant, often without consulting a healthcare professional.

After considering these factors and the significant limitations of the evidence, the GDG decided that for antidepressants (and for most other drugs used for the treatment of mental health problems in pregnancy) that the primary focus of the recommendations should be on a set of principles to guide prescribing rather than a set of recommendations for individual drugs. These principles are as follows:

• All women of childbearing potential should be informed of the limited evidence and consequent uncertainty regarding the harms to the fetus associated with the use of antidepressant medication in pregnancy and the postnatal period, including breastfeeding.
• All women of childbearing potential should be informed of the benefits and side effects associated with the use of antidepressants in pregnancy and the postnatal period (including breastfeeding) if such drugs are being considered.
• All women of childbearing potential should be informed of the background risks associated with depression in pregnancy and the postnatal period.
• All risks should be made clear to women in a manner which is understandable and is based on an assessment of each woman's needs.
• Non-specialists should seek advice or refer onto specialists if they are uncertain about the benefits and harms associated with the use of a particular drug.
• Given the uncertainty about the risks, the threshold for the prescribing of antidepressants should be adjusted in comparison to that for non-pregnant women and that there should be an increased level of monitoring and support for women taking antidepressants in pregnancy and the postnatal period.
• Considerable caution should be exercised when changing or stopping antidepressant drugs in pregnancy and the postnatal period.
• Babies should be monitored for the effects of medication taken in pregnancy and a drug offered that enables the woman to breastfeed if she chooses.
• Specific drugs should only be named where there was evidence to support this.
• That the recommendations for all psychotropic drug use in pregnancy as far as possible should be based on a common set of principles as long as they are supported by the available evidence.

Antipsychotics

In reviewing the evidence and developing the recommendations on the harms associated with antipsychotic use in pregnancy the GDG was, as with antidepressants, mindful of the serious nature of the outcomes reviewed, which could have a profound effect on the life course of any individual who is born with a major congenital defect. They were also aware of the potentially increased rate of a number of the outcomes considered in women with psychosis and bipolar disorder who had not been exposed to antipsychotic drugs. There was some indication from studies of women with a psychotic disorder who were not exposed to medication to support this view. The GDG was also cautious when it came to interpreting the data on individual drugs given the very limited data available and the variation in the size of the datasets.

The GDG agreed that there was a small but significant increase in a number of congenital abnormalities for a number of important outcomes. However, while this rate was reduced and not significant in a comparator group with the disorder there was still a small increase in the absolute rate and the GDG remained uncertain as to whether the increased rate of abnormality could be accounted for by the drug. For a number of neonatal and obstetric outcomes there was evidence of an increased rate of babies being small for gestational age and increased rates of gestational diabetes, preterm delivery and Caesarean section. Again where data were available for disorder-specific comparisons there was a reduction in the absolute rates of these complications. The GDG was also aware that for many of these women there might be a prior history of psychosis or bipolar disorder and this might also be used to guide prescribing practice. Given the uncertainty surrounding the evidence, the GDG adopted a cautious approach in developing its recommendations and also took account of what is known about the effective treatment of psychosis and bipolar disorder in non-pregnant women. In developing the recommendations the GDG considered, in particular, the potential protective function of antipsychotics in reducing the likelihood of postpartum psychosis. The GDG was of the view that it was particularly important to inform women of the risk of not taking medication in pregnancy if they have a history of severe mental illness. In addition the GDG took into account the evidence elsewhere in this chapter on the risks of harm associated with the use of other drugs (notably lithium and valproate) in developing the recommendations. Given the evidence on gestational diabetes and possible related risk for the fetus, the GDG considered it important that additional and careful monitoring for diabetes and offer of an oral glucose tolerance test should be provided for all pregnant women taking an antipsychotic. No data were available to the GDG on the cost effectiveness or impact on resource use of the interventions considered in pregnancy. The GDG considered the potential resource use implications and high costs associated with the management of congenital abnormalities, neonatal and obstetric complications (that is, babies being small for gestational age and increased rates of gestational diabetes, preterm delivery and Caesarean section); however the GDG was also aware that for many of these women there might be a prior history of psychosis or bipolar disorder and the potential protective function of antipsychotics in reducing the likelihood of costly postpartum psychosis. The GDG found it difficult to judge the net effect to NHS costs. Again the GDG took into account what is known about the cost effectiveness of treatment of antipsychotics in non-pregnant women. Moreover, as with depression, the GDG also took into account that many women stop taking medication when they discover that they are pregnant, often without consulting a healthcare professional.

After considering these factors and the significant limitations of the evidence, the GDG decided that as for antidepressants, the primary focus of the recommendations for the use of antipsychotics in pregnancy should be on a set of principles to guide prescribing. These are as follows:

• All women of childbearing potential should be informed of the limited evidence and consequent uncertainty regarding the harms to the fetus associated with the use of antipsychotic medication in pregnancy and the postnatal period including breastfeeding.
• All women of childbearing potential should be informed of the benefits and side effects associated with the use of antipsychotics in pregnancy and the postnatal period (including breastfeeding) if such drugs are being considered.
• All women of childbearing potential should be informed of the background risks associated with psychotic disorders in pregnancy and the postnatal period.
• All risks should be made clear to women in a manner which is understandable and is based on an assessment of each woman's needs.
• Non-specialists should seek advice or refer onto specialists if they are uncertain about the benefits and harms associated with the use of a particular drug.
• Given the uncertainty about the risks associated with antipsychotics (for example, gestational diabetes) there should be an increased level of monitoring and support (for example, help with drug-induced weight gain and offer an oral glucose tolerance test) for women taking antipsychotics in pregnancy and the postnatal period.
• Considerable caution should be exercised when changing or stopping antipsychotic drugs in pregnancy and the postnatal period.
• Babies should be monitored for the effects of medication taken in pregnancy and a drug offered that enables the woman to breastfeed if she chooses.
• Specific drugs should only be named where there was evidence to support this.

Anticonvulsants

In reviewing the evidence and developing the recommendations on the harms associated with anticonvulsant use in pregnancy the GDG was mindful of the serious nature of the outcomes reviewed, which could have a profound effect on the life course of any individual who is born with a major congenital defect. They were also aware of the potentially increased rate of a number of the outcomes considered in women with bipolar disorderwho had not been exposed to anticonvulsant drugs. There was some indication from studies reviewed of women with a disorder who were not exposed to medication to support this view. The GDG were aware that the dataset was primarily drawn from women with epilepsy but they did not think that this invalidated the evidence, which was still seen as relevant to women with bipolar disorder. The small number of anticonvulsant drugs used in bipolar disorder and the relatively large number of studies also meant that the GDG was able to consider the evidence for the three drugs (carbamazepine, valproate and lamotrigine) separately. This is important as there is a clear indication of different patterns of harm associated with each drug. The GDG was of the view that the evidence of significant harms (both congenital and neurodevelopmental) to the fetus associated with valproate was such that it should not be used in the acute or long-term treatment of a mental health problem in women of childbearing potential. There was also evidence of an increased rate of congenital harms associated with carbamazepine albeit not at the same level as valproate and not at a level which would preclude the use of carbamazepine in women of childbearing potential. However, the risk for harm associated with carbamazepine was greater than that observed for lamotrigine and the GDG recommended that carbamazepine is not offered for the treatment of a mental health problem to women who are planning a pregnancy, pregnant or considering breastfeeding. For a woman who is already taking carbamazepine and is planning a pregnancy or becomes pregnant, the GDG recommended that a discussion with the woman should be had about stopping the drug given the possible risk of adverse drug interactions and fetal malformations. The review of lamotrigine did not suggest that there was any significant increase in risk associated with its use in pregnancy. However, the GDG were mindful of the need to check lamotrigine levels frequently during pregnancy and into the postnatal period because they vary substantially at this time.

In considering recommendations concerning valproate, the GDG also took into account a recent audit in an East London psychiatric service (Jones et al., 2014) that shows that standards for prescribing sodium valproate to women of childbearing age, in keeping with NICE guidelines, were not being met. Jones and colleagues (2014) reported that in a trust-wide audit of female inpatients of childbearing age, 13% were prescribed sodium valproate in April 2013, and in a re-audit eight months later there were still 9% of female inpatients of childbearing age prescribed sodium valproate. In addition, and even more concerningly, there was evidence that NICE guidance was not being followed in terms of informing service users about the potential for harm associated with valproate. NICE guidance recommends that where there are no alternatives to sodium valproate, women of childbearing potential must be made aware of the teratogenic properties, have a documented pregnancy test and be prescribed folic acid. However, an audit (and a re-audit eight months later) in this East London NHS Trust revealed that despite recommendations these precautions for prescribing valproate (where there were no alternatives) were not being followed. For instance, 86% (April 2013)/83% (December 2013) of female inpatients of childbearing age prescribed sodium valproate had no documented discussion regarding the teratogenicity and other effects on the developing fetus of sodium valproate exposure, 57% (April 2013)/25% (December 2013) did not have a documented pregnancy test, 86% (April 2013)/92% (December 2013) were not prescribed folic acid, and 86% (April 2013)/58% (December, 2013) had not been provided with contraception. The GDG took the view that assuming that the prescribing practices in this trust were representative of other NHS Trusts (and there was no reason to think otherwise), and the potential for harm that these results imply, there were grounds to strengthen the valproate recommendations that were made in the previous 2007 guideline. This was also consistent with recommendations made in the bipolar guidance.

In developing the recommendations the GDG considered, in particular, the potential protective function of anticonvulsants in reducing the likelihood of postpartum psychosis. The GDG was of the view that it was particularly important to inform the women of the risk of not taking medication in pregnancy if the women had a history of severe mental illness. In addition the GDG took into account the evidence elsewhere in this chapter on the risks of harm associated with the use of other drugs in developing the recommendations. The GDG considered it important that additional and careful monitoring of plasma drug levels should be undertaken for lamotrigine. No data were available to the GDG on the cost effectiveness or impact on resource use of anticonvulsants considered in pregnancy, however the GDG considered the increased rate of congenital and neurodevelopmental defects associated with valproate (when compared with carbamazepine and lamotrigine) and the potential increase to NHS costs. The GDG also noted that there is an increased rate of congenital harms associated with carbamazepine and potential for substantial resource use. Again the GDG took into account what is known about the cost effectiveness of treatment of anticonvulsants in non-pregnant women. As with other classes of drugs the GDG also took into account that many women stop taking medication when they discover that they are pregnant, often without consulting a healthcare professional.

After considering these factors and the significant limitations of the evidence, the GDG decided that as for antidepressants and antipsychotics, prescribing anticonvulsants should be guided by a set of principles, which are set out in the sections above.

Lithium

Lithium has been used in the treatment of bipolar disorder for over 50 years but the data on harm in pregnancy and the postnatal period is very limited. There was some evidence of a small (7 per 1000) increased risk of congenital abnormalities but it was not possible to obtain a clear picture on increased risk of heart defects despite previous concerns about an association of Ebstein's anomaly with the use of lithium in pregnancy. The GDG therefore felt that lithium could have role in the treatment of bipolar disorder in pregnancy but should only be offered in pregnancy and the postnatal period when an antipsychotic medication has not been effective and healthcare professionals should ensure that the woman knows that lithium levels may be high in breast milk with a risk of toxicity for the baby, lithium levels should be monitored more frequently throughout pregnancy and the postnatal period, and lithium should be stopped during labour and plasma lithium levels checked 12 hours after the woman's last dose.

In developing the recommendations the GDG considered, in particular, the potential protective function of lithium in reducing the likelihood of postpartum psychosis. The GDG was also of the view that it was particularly important to inform the woman of the risk of not taking medication in pregnancy if the woman has a history of severe mental illness. In addition the GDG took into account the evidence elsewhere in this chapter on the risks of harm associated with the use of other drugs in developing the recommendations. No data were available to the GDG on the cost effectiveness or impact on resource use of lithium considered in pregnancy. The GDG found it very difficult to judge the net effect on NHS costs associated with the use of lithium in the treatment of bipolar disorder in pregnancy (that is, the increased risk of congenital abnormalities and the reduced likelihood of postpartum psychosis). Again the GDG took into account what is known about the cost effectiveness of treatment of lithium in non-pregnant women. As with other drugs the GDG also took into account that many women stop taking medication when they discover that they are pregnant, often without consulting a healthcare professional.

After considering these factors and the significant limitations of the evidence, the GDG decided that as for antidepressants and antipsychotics, prescribing lithium should be guided by a set of principles, which are set out in the sections above.

Benzodiazepines and related drugs

Considering the limited evidence for congenital harms and the increase in obstetric complications associated with benzodiazepines and related drugs, the GDG did not consider there to be sufficient evidence of clinical benefit to justify their use in pregnancy and the postnatal period. Furthermore the GDG was of the view, given the potential for harm, that a woman who is taking a benzodiazepine when she becomes pregnant should be encouraged and supported in stopping the medication.

Stimulants

Considering the limited evidence on harms associated with stimulants (only data available for methyphenidate), and its limited use, the GDG did not wish to make any specific recommendations on the use of these drugs.

Treatment options for specific mental health problems

In addition to reviewing the evidence for harms of psychotropic medication, the GDG also reviewed the efficacy of pharmacological interventions in pregnancy and the postnatal period. The evidence for efficacy of psychotropic medication in pregnancy and the postnatal period was limited, both in terms of available studies and in the low quality of the evidence reviewed. The GDG was of the view that the evidence for omega-3 oils and transdermal oestrogen was weak in that there was no clear indication of any benefit. The GDG decided therefore to make no specific recommendations for omega-3 or transdermal oestrogen in the treatment of mental health problems in pregnancy and the postnatal period. The evidence for antidepressants was also limited but broadly in line with the evidence of the efficacy of this medication in non-pregnant populations. The GDG therefore was of the view that antidepressants had a role to play in the treatment of depression and anxiety disorders in pregnancy and the postnatal period.

The literature review was unable to identify any evidence for the efficacy of antipsychotic medication in pregnancy and the postnatal period. In line with the principles set out below, the GDG therefore referred to existing NICE guidelines. Again, in line with the principles for the reduction of harm, the GDG decided not to single out specific drugs.

The GDG developed an overarching principle regarding interventions to offer or consider for a specific mental health problem. This was arrived at by consensus and states that where a review of the evidence for efficacy of an intervention might be limited, but which contains no indication of a difference in efficacy or harm from the data in non-pregnant populations, then it is reasonable to extrapolate from evidence from non-pregnant populations to inform recommendations for this guideline (see Chapter 3, Section 3.5.6).

The consequence of this is that NICE guidelines for individual mental health problems should be followed other than where specifically indicated in this guideline. In making its recommendations regarding when and how interventions for mental health problems in pregnancy and the postnatal period might need to be modified, the GDG took into account the following evidence:

• reviews undertaken for this guideline update (in this chapter and also in Chapter 7 on psychological and psychosocial interventions)
• their own expert knowledge and opinion
• the recommendations and underlying evidence from the previous 2007 guideline
• NICE guidelines on specific mental health problems, most notably Depression in Adults (NICE, 2009a), Common Mental Health Disorders (NICE, 2011b), Bipolar Disorder (NICE, 2014b) and Psychosis and Schizophrenia in Adults (NICE, 2014a), Drug Misuse: Opioid Detoxification (NICE, 2007b) and Alcohol-use Disorders (NICE, 2011c).

For women with depression in pregnancy or the postnatal period, the GDG judged that for those with a history of severe disorder, but who present with mild depression when pregnant or after childbirth, an antidepressant should be considered as an option, but that healthcare professionals should take into account all of the recommendations regarding balancing risks and benefits. For women who have a moderate to severe episode of depression or an anxiety disorder that has its onset in pregnancy or the postnatal period, the GDG considered that the full range of options recommended in other relevant NICE guidelines should be available, including medication, psychological interventions, and a combination of both. But for women with pre-existing mild to moderate depression, the GDG considered the evidence reviewed for this guideline update in this chapter and in Chapter 7 and recommended that antidepressant medication should be discontinued and a psychosocial intervention (facilitated self-help) considered. Women with pre-existing mental health problems might be inclined to stop their medication when they know they are pregnant; but for women with moderate to severe depression or an anxiety disorder, the GDG advises changing to medication with lower risk of adverse effects and/or a structured psychological intervention. The GDG wished to emphasise that the clinician will have to carefully balance the need to ensure the woman is offered the optimal treatment against any risks associated with medication or untreated disorder for the fetus.

For women with severe mental illness (psychosis, schizophrenia or bipolar disorder), the GDG judged that an antipsychotic should be offered if a pregnant woman develops mania or psychosis and is not taking any psychotropic medication; the choice of antipsychotic will depend on full consideration of the risks and benefits. For women with pre-existing bipolar disorder, the GDG judged that an antipsychotic in line with the NICE guidelines on bipolar disorder (NICE, 2014b) might be a suitable drug to offer or continue with if a woman plans to breastfeed. Antipsychotic medication should also be offered if a woman with bipolar disorder if stopping the prophylactic use of lithium. If a pregnant woman develops mania while taking prophylactic medication, the GDG considered a number of options, including checking and, if necessary increasing the dose, of the existing medication, changing to antipsychotic medication, lithium if the mania is severe and there has been no response to other medication, and, finally, ECT if there is no response to lithium.

No studies were found that matched the inclusion criteria for the updated review of rapid tranquilisation. Therefore the recommendation from the previous 2007 guideline remains unchanged.

In making recommendations for pregnant women dependent on drugs and alcohol in the light of lack of evidence, the GDG drew on discussion with experts for this important area. They recommend that detoxification for pregnant women carried out in conjunction with specialist mental health and alcohol or substance misuse services, but highlight that women who do not wish to undertake a detoxification should be offered interventions to reduce their drug and alcohol intake. The GDG considered it important that healthcare professionals are aware of the possibility of accidental overdose in women who stop or reduce drug misuse in pregnancy but start misusing again soon after childbirth.

There was also a lack of high quality evidence for pharmacological and psychosocial interventions for sleep problems and insomnia in pregnancy and the postnatal period. However, the GDG was mindful that the previous 2007 guideline recommended low-dose chlorpromazine or amitriptyline for women with ‘serious and chronic problems’ and wished to amend this. The GDG considered the risks associated with low-dose chlorpromazine or amitriptyline and sedating drugs such as zopiclone, as well as the review of harms associated with both antidepressants and antipsychotics, and judged that promethazine was a suitable alternative in pregnancy.

Using and modifying NICE guidelines for specific mental health problems

Considerations for women of childbearing potential

8.9.1.2.

When prescribing psychotropic medication for women of childbearing potential, take account of the latest data on the risks to the fetus and baby. [new 2014]

8.9.1.3.

Do not offer valproate for acute or long-term treatment of a mental health problem in women of childbearing potential. [new 2014]

Treatment decisions, advice and monitoring for women who are planning a pregnancy, pregnant or in the postnatal period

Providing assessment and interventions for mental health problems in pregnancy and the postnatal period

8.9.1.34.

When offering psychotropic medication during pregnancy and the postnatal period, follow the principles in recommendations 8.9.1.6 to 8.9.1.33. [new 2014]

Treating specific mental health problems in pregnancy and the postnatal period

Considerations for women and their babies in the postnatal period