Antidepressants are they a safe and effective choice for the treatment of postnatal depression?

This review assessed the evidence concerning the effectiveness and safety of antidepressants in the management of postnatal depression. This would facilitate evidence-based clinical decisions in the treatment of patients. Data was sourced from several electronic Athens-based and free databases covering the psycho-biomedical and nursing literature.

Studies found included randomised clinical trials, case- and cohort-controlled studies, questionnaire surveys, and qualitative/exploratory research. Previous reviews were also appraised. Outcomes from over 1200 mothers, mother-infant pairings, or infants, exposed to antidepressants were considered. Antidepressants appear to significantly alleviate depressive symptoms. Furthermore, the reported side effects are generally benign and clinically insignificant. However, methodological and analytic flaws negate conclusive inferences. Many studies fail to account for important covariates that may explain effects attributed to antidepressants. Furthermore, most studies fail to account for interactions between antidepressants and patient characteristics, which may reveal more severe adverse effects. Additionally, there is a paucity of literature on long-term effects. Finally, a lack of randomised clinical trials precludes inferences of causality. Given these constraints it is recommended that antidepressants are used as a last resort, and patients are closely monitored to identify unexpected side effects, or recovery induced by covariates rather than antidepressants. 

CHAPTER ONE

INTRODUCTION, RATIONALE, AIMS

Introduction

Antidepressants

1.    Tricyclics (TCAs): amitriptyline, imipramine, nortriptyline.
2.    Selective Serotonin Reuptake Inhibitors (SSRIs): sertraline, paroxetine, fluoxetine, citalopram, venlafaxine, moclobemide.
3.    Serotonin and Noradrenaline Reuptake Inhibitors (SNRIs): venlafaxine, reboxetine.
4.    Monoamine Oxidase Inhibitors (MAOIs): tranylcypromine, moclobemide, phenelzine.

The RCP posits that following three months of treatment 50% to 65% of people given an antidepressant show improvements in mood, compared with 25% to 30% of people administered a placebo. Thus, even after accounting for placebo effects, antidepressants still facilitate further recovery from depressive symptoms. TCAs are generally older than SSRIs and are considered to produce more side effects, especially if there is an overdose. However, all four classes of antidepressants are considered to have by-products, such as high blood pressure, anxiety, indigestion, dry mouth, heart tremor, and sleepiness. Most of the adverse effects are considered mild and expected to dissipate after a few weeks.

The RCP cites evidence of withdrawal symptoms in infants shortly after birth, especially with paroxetine (RCP, 2006). Babies can also receive a minute concentration of antidepressants via breastfeeding (Kohen, 2005), albeit the risk of pathology is considered small due to the rapid development of kidneys and livers in infants. Overall, use of antidepressants during breastfeeding is not discouraged. Some pregnant women suffer a recurrence of depressive symptoms, and therefore may need to take antidepressants continually.

The National Institute for Clinical Excellence (NICE, 2004) has published guidelines for the treatment of depression. However, there is no special emphasis on pregnancy-related depression. Antenatal and postnatal guidelines are due to be published by 2007 (Green, 2005).

Postnatal Depression

Rationale

Limited scope

Identification of gaps and inconsistencies

Verification of current claims

CHAPTER TWO

LITERATURE REVIEW

Value of conducting a literature review

Effects of anti-depressants

A one-way analysis of variance was used to compare outcome measures across the three groups. Correlational and regression tests were used to assess the contribution of confounding variables. Results revealed no group differences in child’s global IQ, language development, or behaviour (see Figure 1). The authors concluded, “Exposure to tricyclic antidepressants or fluoxetine throughout the gestation period does not appear to adversely affect cognition, language development, or the temperament of preschool and early-school children. Although regression was used to account for the contribution of confounding factors, such as verbal comprehension and expressive language, the variance explained by these variables was not in fact partialled out before testing for group differences. This would have required a multivariate analysis of covariance in which adjustments for covariates are built into the analysis. More importantly, the observed similarity in outcomes across the three groups may reflect simple or complex interactions with other variables. This issue is discussed in greater detail in Chapter 3.

Figure 1 Cognitive outcomes (mental and psychomotor development, and cognitive abilities) across antidepressant and control groups (Nulman et al, 2002). Differences are not significant.

Wisner et al (2001) performed a double-blind randomised control trial to assess the effect of nortriptyline on the rate of reoccurrence of postpartum depression in non-depressed women who had previously had at least one depressive episode. Women were randomly exposed to nortriptyline or a placebo immediately after childbirth. Outcome data was collected over a 5-month period using the Hamilton Rating Scale for Depression, and Research Diagnostic Criteria for depression. No group differences emerged, suggesting that nortriptyline was no more effective than a placebo in treating PND. This study was followed up with another RCT (Wisner et al, 2004), this time evaluating the effect of sertraline on the rate of and time to reoccurrence of postpartum depression. They highlighted a paucity of clinical trials on the impact of antidepressants in women who have previously had a depressive episode, and hence may be prone to experience a reoccurrence. Participants were pregnant women with gestation periods of 9 months or less, and at least one episode of postpartum depression that fits that the DSM-IV definition of major depression. Women with other forms of psychopathology (e.g. psychosis, or bipolar disorder) were excluded. Participants were randomly assigned to a treatment (sertraline) or placebo group. The drug was administered immediately after birth, beginning with a 50mg/day dose, which was later dropped to 25mg/day to minimise side effects (e.g. headache). Data analysis using Fisher’s exact test showed a significant group difference in rate of reoccurrences, during a 17-week preventive treatment period.

Reoccurrences occurred in 4/8 women assigned to the placebo group, and 1/14 women in the treatment condition, translating into a 0.43 difference in reoccurrence rates. All women had adhered to the treatment regime, thus minimising the confounded effect of non-compliance. There was also a significant group difference in time to reoccurrence, with first reoccurrence beginning much earlier for the placebo group (at 5 weeks, followed by more reoccurrences) compared with the treatment group (at 17 weeks, followed by more reoccurrences). However, the treatment group reported more side effects (e.g. dizziness, drowsiness). This RCT clearly demonstrates the effectiveness of an SSRI in preventing the reoccurrence of postpartum depression, albeit the conclusiveness of these findings is constrained by the failure to control for key background variables, such as previous and recent history of psychopathology, and drug effect expectations. For example, lingering symptoms of a distant depressive episode may help precipitate a quicker reoccurrence.

Figure 2 Rate of recurrence of postpartum depression in placebo and SSRI women (Wisner et al, 2004)

Oberlander et al (2005) tested the effect of SSRI exposure on biobehavioural responses to acute procedural pain in newborn babies at 2 months of age. Previous research has suggested altered behavioural and physiological reactions to a routine painful event in infants, after prenatal exposure to SSRI antidepressants. There is paucity of literature on the long-term effects of SSRIs on neurobehavioural variables, such as cognitive, language and motor development. Given that SSRIs work by inhibiting the reuptake of serotonin (5-hydroxytrypamine [5HT], a neurotransmitter that regulates cardiovascular function and pain signals in the developing brain), and given that SSRIs easily pass through the placenta, it is possible that regions of the brain associated with pain reactivity may be affected. Participants were recruited from a cohort of mothers and their infants during pregnancy, as part of a longitudinal study of prenatal medication use. Only Mothers/infants with no psychotropic or antidepressant use during pregnancy, whose pregnancy was 9 to 10 weeks, and no history of maternal mental illness, were eligible to be assigned to the control group.

Three groups of infants were compared: (a) infants exposed to prenatal SSRI (fluoxetine); (b) infants exposed postnatal via breastfeeding (paroxetine, fluoxetine, sertraline); and (c) control infants. Behavioural (facial activity), physiological (variations in heart rate [HR], often used as a measure of pain reactivity in infants), and pharmacological (analysis of blood and breast milk samples) data was collected. Results showed impaired facial reactions in infants exposed to prenatal SSRI. Altered pain reactivity was observed in both prenatal and postnatal exposed infants, suggesting enduring neurobehavioural SSRI effects that extend beyond the newborn phase. Oberlander et al’s (2005) study was constrained by low power and generalisability (limited sample size), and lack of a non-medicated control group with depressive symptomatology. They were uncertain about the clinical implications of these findings, suggesting that use of SSRIs for treating maternal depression was appropriate pending further research on the sustained effects of SSRIs.

Marcus et al (2005) screened prenatal depression in pregnant women attending an obstetrics clinic. The study aimed to assess the rates of anti-depressant use and its association with depression, measured by the Center for Epidemiological Studies Depression Scale (CES-D). Overall, 390 women who had used antidepressants within two years of conception were screened. Average age was 28.6 years, and most women were married and Caucasian (73%). Screening took place at around 24 gestation weeks. Data was collected regarding the use of antidepressants during the past two years, and discontinued use following pregnancy, in addition to the CES-D data. The standard CES-D cut-off of 16 was used to establish the presence of depressive symptomatology.

A t-Test was used to compare two groups: women who reported they stopped using anti-depressants and hence were not currently on medication (n=248); and women who continued to use antidepressants during pregnancy (n=68). The dependent/outcome variable was total CES-D scores. Chi-square was also used to assess use/non-use of antidepressant medication and CES-D groupings (i.e. <16 versus.  ≥16 scores). Chi-square revealed no reliable differences in depression scores between women taking and those not taking antidepressants during pregnancy. The t-Test also revealed no group differences in actual CES-D scores. Thus, antidepressant use seemed to have no bearing on depression levels.

Figure 3 CES-D data for women who did and those who did not use antidepressants during pregnancy (Marcus et al, 2005). Observed differences are not significant.

The authors attributed the null results to poor treatment adherence, and inadequate prescribing/monitoring. Furthermore, they suggested that group differences might have been more pronounced if the study focused on unmediated women (i.e. those who had not used antidepressants at all, rather discontinued use). This study was unique because it assessed antidepressant use around the time of conception. However, the findings are compromised by several analytic constraints. Firstly, the use of a t-Test is questionable. This test makes no provision for controlling for covariates (i.e. important background variables, such as patient preference, compliance history, side-effect profile, social support, quality of marital relationship, prior history depression) that may confound significant group differences, although this concern is less important given the null results.

A more serious problem is the possibility that certain assumptions which underlie use of the t-Test were violated, notably homogeneity of variance. The huge disparity in group sizes (268 versus 68) hugely increases the possibility of significant differences in group variances, which in turn would obscure reliable differences in CES-D scores. The authors do not report Levene test results, which would have addressed the homogeneity issue. Perhaps a non-parametric test (e.g. Mann-Whitney) may have been more appropriate. Furthermore, it is not clear why the authors conducted a chi-square test! Collapsing the CES-D scores into a dichotomy reduces the quality of the data because it obscures subtle differences between scores. Overall, the chi-square analyses amounted to a less precise duplication of the t-Test results! Finally, this study was entirely based on women’s self-reports of medication use, with no familial, clinical, or other verification. It is therefore unclear to what extent the null results are attributable to self-report bias.

Several review articles on antidepressants and postnatal depression have been published. These range from limited commentaries (e.g. Goldstein & Sundell, 1999; Yoshida et al, 1999; Misri & Kostaras, 2002; Hendrick, 2003; Bennett et al, 2004; Kohen, 2005; Marcus et al, 2005) to comprehensive and systematic appraisals.

Goldstein and Sundell (1999) reviewed literature on the safety of SSRIs during pregnancy. Their work was based on the premise that although antidepressants may be necessary during pregnancy it is essential identify and weigh the risks against the benefits in order to make an informed choice as to whether or not to use the drugs. Due to the paucity of randomised controlled trials on the topic, the review focused on evidence obtained from cohort/case-controlled studies, patient surveys, retrospective studies, and anecdotal reports. Electronic databases searched included Medline, EMBASE, Derwent Drug File, and PsychINFO. Four cohort-controlled and 5 prospective studies were found which evaluated the impact of SSRI exposure. One study compared fluoxetine, TCA, and non-teratogen (e.g. antibiotics) exposed groups of non-depressed females. SSRI and TCA exposure produced no significant malformations, or differences in birth weight and infant prematurity. However, there was a greater tendency for fluoxetine- and tricyclic-exposed women to miscarry compared with controls. However, this effect was not significant and hence may simply have occurred by chance.

Goldstein and Sundell (1999) report another study which compared early exposed (prior to 25 weeks), late exposed (continuing after 24 weeks), and a non-teratogen control group. Again findings revealed no adverse effects in the treatment groups, albeit infants exposed to fluoxetine early showed a higher prevalence of minor anomalies that have little or no clinical importance. Furthermore late exposure to fluoxetine seemed to increase the rates of admission to special care nurseries and impaired fetal development. However, these findings were inconclusive due to prior group differences on previous psychotropic drug use, and failure to control for depression levels. Still other research suggests no effect of SSRIs (sertraline) on the prevalence of stillbirth, prematurity, mean birth weight and gestational age. Evidence suggests no statistically significant differences between SSRI exposed and control groups on IQ, language development, height, and head circumference. Of the prospective studies reviewed three assessed paroxetine, and fluoxetine, and two tested sertraline. All studies reported no significant increase in the rate of malformations and spontaneous abortion, although there was some evidence of lower birth weight given protracted use of antidepressants.

Goldstein and Sundell (1999) found one study, which showed that fluoxetine exposure during the first trimester did not increase the risk of malformations. Rates of spontaneous abortion and prematurity were no greater than historical patterns. One study assessed the effects of third trimester exposure to depressants, and up till delivery, indicating that fluoxetine intake during the third trimester produces no significant adverse effects. Goldstein and Sundell (1999) emphasise the importance of acknowledging the limitations of case/cohort control studies, surveys, retrospective designs, and other uncontrolled correlational methods, compared with randomised clinical trials. It is important that findings from less rigorous designs are interpreted with caution. Overall, this review summarises outcomes from 1000+ pregnancies exposed to fluoxetine, and 300 pregnancies exposed to other SSRIs. All report little or no increased risk from using antidepressants.

Yoshida et al (1999) appraised existing case reports and studies on breastfeeding in relation to antidepressants, specifically TCAs and SSRIs. The review revealed that the preponderance of studies were single case studies - there was a lack of randomised controlled trials or prospective controlled studies. Comparison of findings across different studies has hampered by methodological differences or absence of essential information. Most of the literature focused on TCAs, but these studies in total only tested a limited (n=66) number of mother-infant pairs. By contrast there was less evidence concerning the impact of SSRIs. Yoshida et al (1999) surmised that the benefits of taking TCAs probably outweigh the risks provided recommended TCAs are being taking, and at the right dosage. Furthermore, the infant must be healthy. It is suggested that an accumulation of case control studies will ultimately provide a platform for launching clinical trials designed to identify severe toxic and long-term health outcomes of antidepressant use during breastfeeding.

Simpson and Noble (2000) reviewed the literature on the effects of the SSRI fluoxetine on depression in women, noting the paucity of literature in this area. Studies confirmed that fluoxetine is secreted into breast milk and that breast-fed infants ingest from 3% to 10.8% of their mothers fluoxetine intake. However, traces of the drug can accumulate in infants to a significant degree. Evidence from randomised controlled trials suggests that fluoxetine has no adverse surgical, medical or cosmetic effects on the fetus. The rate of malformations is no greater for pregnancies exposed to fluoxetine compared with non-exposed pregnancies. Simpson and Noble (2000) also considered evidence from the manufacturer, who retained accumulated records of over 3000 fluoxetine-exposed pregnancies since the drug became available on the open market. Only a minority (3.5%) of pregnancies with verified exposure during the first trimester developed major abnormalities.

Only 0.2% developed minor problems, compared with 2.3% and 14% for controls. However, at least one study (prospective, cohort) found a higher incidence of minor abnormalities in first-trimester fluoxetine-exposed pregnancies, compared to control pregnancies. Crucially, this group difference remained significant even after controlling for other psychotherapeutic drugs (benzodiazepine) also consumed by mothers taking flouxetine. Furthermore, the probability of premature birth, admission to special care nurseries, and poor neonatal development was higher during the third trimester compared with early-exposed (first or second trimester) or control pregnancies (see Figure 4). Furthermore, late-exposed infants had significantly lower birth weight and length compared with early exposed or control infants (see Figure 5).

Figure 4 Adverse effects of fluoxetine as a function of trimester of exposure (Simpson & Noble, 2000)

 Figure 5 Effects of fluoxetine on weight as a function of trimester of exposure (Simpson & Noble, 2000)

Other research found no adverse effects for fluoxetine compared with TCA-exposed or control pregnancies. Studies on breastfeeding were also considered. Simpson and Noble (2000) noted that fluoxetine is not recommended for breastfeeding mothers, albeit the drug may be useful in this arena. A mixture of case reports and retrospective studies painted a confusing picture. On the one hand fluoxetine has been found to produce no significant malformations in breast-fed infants. On the other hand there have been exceptions. One 3-week old infant appeared to develop vomiting and watery stools; another experienced ‘seizure-like’ fits, while a third became irritable. Fluoxetine has also been implicated in impaired growth rate for breast-fed infants. However, one study monitored the development of several breast-fed infants for over a year and found no abnormalities upon neurological assessment.

It was noted that the long half-life of fluoxetine means that traces of the drug ingested during the third trimester may persist in breast milk and be passed on to an infant, even if the mother did not take fluoxetine during breastfeeding. Nevertheless, fluoxetine is considered to play an important role in postnatal depression. Simpson and Noble (2000) note that antidepressants can be recommended when depressive symptoms become severe and start to retard the mother’s well-being. Moreover, use of antidepressants, including SSRIs, is often inevitable due to the significant number of unexpected pregnancies and high prevalence of depression in the general population . Overall, Simpson and Noble (2000) conclude that the existing evidence reveals no “significant association between the use of fluoxetine during the first trimester and major malformations in the fetus… Information on third trimester exposure is very limited, making it impossible to draw definitive conclusions” (p.321).

Simpson and Noble (2000) also reviewed the literature on postpartum depression. No other antidepressant other than fluoxetine appeared to have been tested using randomised controlled (double-blind) trials. Findings suggest symptom improvement in treatment compared with placebo subjects. However, methodological and analytic constraints negate any conclusive inferences. TCAs are more commonly prescribed in this setting, although there is a paucity of conclusive evidence concerning the clinical viability of TCAs over SSRIs. Data on the impact of fluoxetine specifically is limited and Simpson and Noble (2000) declined to draw any firm conclusions. Overall, they surmised that use of fluoxetine by pregnant women requires thorough cost-benefit appraisals, albeit taking the drug during the first trimester appears to entail less risk.

Hoffbrand et al (2002) report a cochrane review (also cited by the Royal College of Psychiatrists [RCN, 2004]), on the safety and efficacy of antidepressant treatment for PND. The review was based on a search of clinical trials registered by the Cochrane Depression, Anxiety, and Neurosis Group, and the Cochrane Pregnancy and Childbirth Group, and other databases. Only randomised trials in which women reporting depressive symptoms within the first 6 months after childbirth were randomly assigned to receive an antidepressant alone, or in combination with other interventions, or receive a placebo, were eligible to be included. Only one trial (Appleby et al, 1997) was eligible to be reviewed. This study (already reviewed elsewhere in this chapter) showed that fluoxetine was more effective than a placebo and as effective as cognitive-behavioural therapy. Overall, this review highlighted the need for more trials incorporating a longer follow-up period. 

Misri and Kostaras (2002) considered the literature on the various risks and benefits to mother and child of using antidepressants to treat postnatal depression. It is recognised that many newly delivered mothers who require antidepressants may also intend to breastfeed. SSRIs and TCAs are identified as the most commonly used antidepressants, with very scant literature available regarding the use of MAOIs. The literature search was conducted using MEDLINE, and dating back 20 years.

Figure 6 shows the number of infants assessed and adverse events reported by existing studies of SSRIs. Norfluoxetine, the active component of fluoxetine, may accumulate in the serum of breastfeeding infants due to its long half-life. Studies have reported various adverse effects including colic, crying, lower body weight, fussiness, and seizures. However, they reported no abnormal effects although most studies indicate low concentrations of the drug in breast milk and infant serum. Low concentration levels may be too weak to stimulate malformations. The evidence for sertraline, paroxetine, fluvoxamine, and citalopram is mixed. Like fluoxetine, these SSRIs were generally detected in low concentrations, with generally no adverse effects. Fluvoxamine has been found to be an effective remedy for postnatal depression. Overall, the evidence base for SSRIs other than fluoxetine was too scant to warrant firm conclusions about their impact on postnatal depression.

Figure 6 Number of infants and adverse events reported for various SSRIs (Misri & Kostaras, 2002)

Figure 7 Number of infants and adverse events reported for various TCAs (Misri & Kostaras, 2002)

Research findings on TCAs are summarised in Figure 7. Like SSRIs, TCAs tend to be found in low concentrations in breast milk and serum. Doxepin has been shown to produce abnormalities (respiratory depression, muscle hypotonia, jaundice, vomiting, poor sucking, and drowsiness). However other TCAs – amitriptyline, clomipramine, desipramine, imipramine and nortriptyline - do not appear to produce any adverse health outcomes. The negative effects of doxepin are attributed to the long half-life of its active metabolite, so use of this TCA is generally not recommended. Literature on other depressants was also considered.

Hendrick (2003) comments on literature concerning the effectiveness of antidepressants and psychotherapeutic interventions. She argues that postnatal depression is often overlooked in paediatric clinics. Her review highlights the paucity of randomised controlled trials on the impact of antidepressants on depressive symtomatology. One study was found in this regard (see Appleby et al, 1997). This investigation showed that an antidepressant – fluoxetine – was no more effective than cognitive behaviour therapies. Other studies suggest that sertraline, paroxetine, venlafaxine, and nortriptyline generally have no adverse health outcomes, although some reports suggest fluoxetine can induce sleeplessness and irritability. Furthermore, there is a paucity of literature on the long-term effects of antidepressants. By contrast, psychotherapeutic treatments have been found to be both highly acceptable and effective. Nevertheless Hendrick (2003) concluded that antidepressants can be helpful and need not necessitate suspension of breastfeeding. Overall the review is rather limited in scope (only 12 studies are cited) and hence the reliability of her deductions may be questionable. However, the conclusions seem to concur with more comprehensive literature reviews discussed here.

Bennett et al (2004) reviewed evidence on the use of antidepressants for treatment of post-natal depression. The review focused on clinical outcomes for untreated postnatal depression. Data from twenty-nine studies was considered, including cohort studies of fluoxetine, venlafaxine, fluvoxamine, sertraline, paroxetine, trazodone/nefazodone, SSRIs in general and tricyclic antidepressants (TCAs), which generally indicate no adverse clinical effects on fetal development. Longitudinal studies reported no adverse effects of antidepressants on intellectual, language, or behavioural development in children exposed to TCAs or fluoxetine while in the uterus. By contrast, there was a paucity of randomised controlled trials (RCTs) focusing on pregnant women, due to concerns about possible adverse health outcomes. Thus, most clinical trials have used male participants. Consequently, little is known about the causal relationship between use of antidepressants and postnatal depression.

Bennett et al (2004) note that, in the absence of conclusive evidence, health warnings against the use of antidepressants during pregnancy may have little practical value since many pregnant women use psychotropic drugs for several days or weeks after conception before realising they are pregnant. Several studies suggest that antidepressants may in fact produce negative health outcomes. Infants exposed to antidepressants such as fluoxetine and paroxetine during the third trimester have been found to develop neonatal withdrawal symptoms (e.g. excessive crying, difficulty eating), and other complications (e.g. jaundice), not to mention impaired psychomotor development, pain response, premature delivery, miscarriages, and other clinical maladies, and a higher probability of admission to special care nurseries. It is not entirely clear whether these effects were caused by antidepressants or other confounding ‘third’ variables, such as the impact of other antenatal drugs. There is a lack of research controlling for such extraneous variables. However, until such evidence accumulates, caution is necessary in prescribing antidepressants during pregnancy.

The picture is further obscured by non-recognition of depressive symptoms (by both clinicians and patients), under-prescribing, patient non-compliance with drug regimes, and other factors. Furthermore, there is a lack of evidence on correct dose requirements during pregnancy – various antenatal physiological changes (e.g. decrease in gastrointestinal activity) may require specific dose alterations. SSRI dose requirements have been found to vary across the three trimesters. The lack of clinical guidelines about antidepressants dose specifications during pregnancy increases the probability of under- and over-prescribing, with significant implications for fetus development. Overall Bennett et al’s (2004) review highlights the uncertainty regarding the use of antidepressants. More importantly it documents evidence suggesting that antidepressants can have adverse maternal and fetal health outcomes.    

A more recent review by Boath et al (2005) focused on the findings of randomised controlled trials on the effectiveness of interventions, including antidepressants, in the treatment of postnatal depression. The paper is based on a comprehensive review of multiple electronic databases, including PsychINFO, CINAHL, EMBASE, and Medline, in addition to searches of various electronic journals (e.g. British Journal of Psychiatry, the Journal of Affective Disorders). The literature dated back to the mid 1960s. Over 100 articles were found of which 25 were randomised controlled trials, and 21 actually reviewed. A wide range of interventions have been evaluated including psychological interventions, interpersonal therapy, postnatal stress debriefing, reconfiguring midwifery and other service provision, home based care, hormonal prevention and use of antidepressants. Unsurprisingly only one RCT was found that tested the effect of antidepressants (Wisner et al, 2005; this study is reviewed in detail elsewhere in this chapter). Noting the limited efficacy of other interventions, none of which demonstrated long-term success, Boath et al (2005) argued that there is a need for an ‘integrated approach’, in which antidepressants and other interventions are combined. The main constraint of Boath et al’s (2001) review is its scope. The focus on RCTs meant that reliable data from prospective and cohort studies was not considered. Such designs are especially relevant for testing the effect of antidepressants given the ethical constraints associated with conducting clinical trials.

Kohen (2005) also reviewed the antidepressant literature in relation to breast feeding, surmising from the outset that the concentrations in breast milk are minute, and that research on TCAs and SSRIs reveal “no clinical indication for women treated with either [TCAs or SSRIs] to stop breast-feeding, provided that the infant is healthy and its progress is monitored” (p.372). However, several methodological and analytic criticisms of the literature are highlighted, notably the preponderance of case studies and small series, and lack of power in statistical analysis. Evidence for individual TCAs and SSRIs are considered separately. Studies of TCAs in general show that these drugs do not accumulate in breastfed infants, although the short-term design of studies means that little is known about long-term effects. Thus, Kohen (2005) advises a full risk-benefit evaluation before prescribing TCAs to breastfeeding mothers. Studies have found little or no traces of amitriptyline, nortriptyline, or doxepin in infant’s breastfed by mothers taking these drugs. Levels of Clomipramine have been found to be high immediately after delivery but decline to its lowest detectable concentration after 5 weeks. Furthermore, studies have generally reported few if any side effects from using TCAs, although the longer acting metabolite of doxepin may induce respiratory depression and sleepiness. Kohen (2005) recommends amitriptyline and imipramine if TCAs are to be prescribed.

Regarding SSRIs, fluoxetine and its metabolite (norfluoxetine) have been found in extremely small concentrations (<10% clinical threshold) with no adverse health effects on infant development. However, this data may be confounded by sample heterogeneity and dosage, both of which may obscure significant treatment effects. Some single case studies of women on higher fluoxetine doses have reported adverse outcomes such as somnolence, fever, hypotonia, and weight loss. Nevertheless, the general consensus is that these side effects are clinically benign. Kohen (2005) recommends careful monitoring of mothers and infants on fluoxetine regimes. Paroxetine, citalopram, and venlafaxine, have all been found in very low concentrations (<10% threshold) and produce no adverse health effects, although studies suggest that citalopram can cause restlessness, irritability, and somnolence. No studies of mirtazapine, nefazodone, or bupropion were found. There is a paucity of research on the effects of MAOIs in infants, albeit no cases of adverse health outcomes have been reported. In view of this lack of empirical data and the availability of newer antidepressants, it is recommended that use of MAOIs be discontinued.          

Kohen (2005) concludes by highlighting the importance of conducting risk-benefit evaluations prior to recommending antidepressants to breastfeeding mothers. These appraisals should be based on details of the psychopathology (e.g. severity, frequency), availability of social support (e.g. friends, family), the patients’ prior attendance record and compliance with treatment regimes, and other factors. The following conclusions are reached:

1.    There is a clear treatment protocol for use of antidepressants with breast-feeding women;
2.    TCAs and SSRIs are generally safe to use while breastfeeding;
3.    Use of MAOIs should be discontinued. 

A number of studies have considered women’s perceptions of antidepressants. Negative attitudes towards this form of treatment may have implications for adherence to drug regimes, and even seeking treatment for postnatal depression in the first place.

Boath et al (2004) considered women’s experiences of taking antidepressants for post-natal depression. Research suggests that depressed pregnant women are sometimes prescribed no medication or medication dosages that are too low to be therapeutic. Participants were thirty-five women clinically diagnosed with depression who had been prescribed antidepressants. They were involved in a wider study on the cost-effectiveness on service provision for post-natal depression. A health visitor recruited women who had a 6-week to 1-year-old baby and scored over the Edinburgh Postnatal Depression Scale (EPDS) threshold of 12. A Standardised Psychiatric Interview (SPI) was conducted to verify the presence of clinical depression  and generate a diagnosis consistent with Research Diagnostic Criteria. A psychiatrist independently verified the diagnosis. A questionnaire assessing women’s experiences of using antidepressants was administered during each interview, incorporating both Likert style and open-ended questions.

Overall, 60 women fulfilled the eligibility criteria and 35 of these were prescribed antidepressants, of which 31 took their medication. Of these, most (25) had been prescribed tricyclic antidepressants (TCAs), while 5 were prescribed SSRIs, and 1 was on flupenthixol alone. Thematic analysis was used to identify common subjects in the data. Various themes emerged including ‘helpfulness of treatment’, ‘alternatives to medication’, ‘self-regulatory behaviours’, ‘information for health professionals’, and ‘suggestions for improvement’. These themes were further grouped into four categories: ‘women’s views on antidepressants’, ‘self-regulation of antidepressants’, ‘discontinuation of antidepressants’, and ‘communication and concordance’. Overall, the majority of women found antidepressants helpful. Nevertheless, data from some women indicated possible low compliance rates, highlighting the need for more reliable information on medication adherence. Low compliance can have significant clinical consequences if undetected.

One concern with this study is the lack of information regarding the reliability and validity of the findings. While the Kappa test was used to establish good inter-observer reliability this in itself does not demonstrate the authenticity of the emerging themes. Methods such as triangulation (i.e. using a different method [e.g. quantitative analysis] to verify observations), participant feedback, and re-checking of negative cases (i.e. observations which don’t fit the emerging themes) could have been used to verify the themes (Coolican, 1994). In the absence of such corroboration the current findings at best can be considered inconclusive.

Chabrol et al (2004) assessed the acceptability of antidepressants amongst 405 delivered mothers admitted in maternity clinics over a fixed period. Using obstetric clinics as the setting for this study allowed the implementation of health promotion initiatives. Three treatments were compared: antidepressants, and psychotherapy via consultations and home visits. The study assessed women’s views concerning the presence of anti-depressants in maternal breast milk and infant serum, which can adversely affect fetal development. It was expected that mothers would view antidepressants less favourably when presented with information that traces of these drugs may be present in breast milk. Furthermore, mother’s views on antidepressants and postpartum blues were also assessed.

During the 3 days post delivery mothers received information on the incidence and prevalence of postnatal depression, the impact on mother-child bonding, infant development, available treatments, and their efficacy. Participants then completed a questionnaire assessing their intention to breastfeed and acceptance of the three therapies, and also the EPDS. Analysis of variance and the t-Test were used to assess treatment acceptability pre- and post-information, and also across treatments.

Results showed that mothers who were breastfeeding viewed antidepressants less favourably at post-information compared with mothers not breastfeeding. No group differences emerged for the other therapies. Furthermore, participants with an EPDS score ≥10 or <10 did not differ in their acceptance of antidepressants (or the other treatments) pre- or post-information. In essence postpartum depression did not appear to affect attitudes towards antidepressant drugs. However, t-Test results showed that antidepressants tended to be less acceptable than the other treatments, irrespective of breastfeeding status. Furthermore, endorsement of antidepressants (and other treatments) was significantly lower post-information.

Figure 8 Acceptance of treatments at pre- and post-information (Chabrol et al, 2004). Acceptance measured on visual analogue scale (0-10, from unacceptable to very acceptable)
 
Chabrol et al (2004) suggested that low acceptance of antidepressants may decrease patient adherence to prescriptions, especially amongst breastfeeding mothers. However caution is advised in interpreting the findings. The computation of multiple t-Tests increased the probability of a type I error (obtaining a significant result by chance). Moreover, the use of a less stringent significance level (p<.10, rather than p<.05) inspires less confidence in the findings - it increases the probability that the results occurred by chance. The authors do not indicate whether assumptions underlying use of the t-Test/analysis of variance were satisfied. More importantly, the study failed to control for covariates, meaning that group differences may be confounded by relevant background variables, such as past experience (use of antidepressants, pregnancy) and knowledge of postnatal depression prior to the study.

CHAPTER THREE

OVERALL SYNTHESIS, CRITICAL ANALYSIS OF THE LITERATURE

Overall, existing literature suggests that antidepressants can be effective in treating PND, and generally have few if any side effects. However, conclusive inferences are negated by various analytic and methodological constraints. These include:

1.    Failure to account for covariates
2.    Failure to account for moderator effects (which might reveal conditions under which significant treatment-control group differences may emerge)
3.    Lack of randomised clinical trials
4.    Low statistical power and liberal significance levels
5.    Lack of research on MAOIs and other antidepressants (e.g. mirtazapine, nefazodone, bupropion) (Kohen, 2005).

Covariates

Most studies reviewed here, including randomised controlled trials, failed to adequately control for key background variables that may confound observed effects of antidepressants on mood and other outcome variables. Green (2005) identified several important background variables including previous history of depression and other psychopathology (e.g. anxiety) during pregnancy, level of social support (e.g. from friends, family), state of marital relationship, stressful life events (e.g. financial and employment troubles) during pregnancy. Other factors include previous treatment history and response, side-effect profile, history of compliance with drug regimes, and patient preference for particular treatments. Any differences between antidepressant and placebo or control groups reported in the literature may be confounded by these variables – or covariates. Most prospective controlled and randomised clinical trials reviewed here made attempts to control for background variables, for example by using strict eligibility criteria (Wisner et al, 2004). Randomisation itself is a form of control, distributing extraneous influences evenly across experimental groups. However the majority of studies are in the form of cohort- or case-controlled designs in which self-selected groups (e.g. breastfeeding mothers, and/or mothers who used antidepressants) are compared. 

Covariates can be controlled during data analysis. Some studies employ simple analysis of variance, which is inappropriate because it fails to account for covariates. A more appropriate test would be an analysis of covariance, which controls for extraneous variables (Field, 2000). Other studies have employed regression procedures (e.g. multiple or logistic regression) but failed to use a hierarchical procedure, whereby the influence of background variables is first partialled out at an earlier step in the analysis before assessing the unique contribution of antidepressant versus placebo/control group membership (Field, 2000). Failure to account for covariates renders the findings inconclusive, because there is a possibility for example that reported effects of antidepressants are attributable to unknown prenatal conditions (e.g. better regime compliance rates in a fluoxetine compared with a placebo group).

Moderator effects

Inconsistencies in the literature may reflect third-variable moderate effects. A moderator variable specifies the conditions under which one variable (e.g. antidepressants) will have an effect on another (e.g. postnatal depression, health outcomes, cognitive development) (Baron & Kenny, 1986) (see Figure 9). This is especially relevant for studies reporting no adverse effects for antidepressants (Goldstein & Sundell, 1999). It is possible that SSRIs or TCAs do produce significant health outcomes but only under certain conditions. For example Boath et al (2004) found evidence for low compliance rates among women prescribed antidepressants. Low compliance may result in negligible if any differences between treatment and control or placebo groups. In this case compliance rate functions as a moderator, determining the conditions under which antidepressants will affect health outcomes. Similarly Chabrol et al (2004) found evidence of low acceptance of antidepressants, which in turn may obscure health outcomes: antidepressants may produce significant improvements in mood only in a sample of women who approve of this treatment (and hence may be more likely to adhere to prescriptions), denoting an interaction between antidepressants and acceptance levels (with the latter operating as the ‘moderator’).

Figure 9 Effects of antidepressants on outcome measures may be moderated by (i.e. interact with) other variables (Baron & Kenny, 1986)

Randomised Controlled Trials

This review has revealed a distinct lack of randomised controlled trials, consistent with the other reviews (e.g. Yoshida et al, 1999; Hoffbrand et al, 2002; Misri & Kostaras, 2002; Kohen, 2005). This negates causal inferences about the link between antidepressants and mood or adverse effects. Prospective studies may imply causality, based on the sequence of events (e.g. antidepressants were taken before the onset of infant malformations, therefore the latter could not have caused the former) but they do not actually demonstrate causality, which would require manipulation under controlled (randomised) conditions. However, randomised controlled trials may be difficult to implement due to ethical concerns. Ultimately, any assumptions of causality need to be made with caution, subject to verification.

MAOIs and other antidepressants

There is a paucity of recent literature on the effectiveness of MAOIs, especially in terms of long-term outcomes. Most studies have focused on TCAs and SSRIs. The RCP (2004) does not encourage the use of MAOIs due to concerns about its cardiovascular. Thus, the state of current literature seems to reflect a shift away from these antidepressants.   

Limits of Review

This review is limited by its lack of systematic and mathematical methodology. In other words, it’s failure to achieve the standards of a meta-analysis. This critique can also be levelled at previous reviews of this literature, most of which report conclusions based on the overall and subjective appraisals of the writer(s) (e.g. Goldstein and Sundell, 1999; Misri & Kostaras, 2002; Marcus et al, 2005). The use of a subjective (i.e. non-mathematical) approach has several disadvantages. Firstly, estimates of the overall magnitude of group differences (e.g. fluoxetine versus control groups) across studies are entirely subjective and hence imprecise. Thus, based on the current review, it would appear that in general use of SSRIs helps to alleviate depressive symptomatology, albeit there is no way of estimating the precise strength of this effect.

Moreover, it is clear from some studies that antidepressants do not always produce significant improvements in mood (Kohen, 2005). Similarly, while some research indicates that antidepressants have no side effects (e.g. Nulman et al, 2002), other studies suggest otherwise (Appleby et al, 1997) (also see reviews by Misri & Kostaras, 2002; Kohen, 2005). In essence, despite the preponderance of evidence suggesting that antidepressants are generally safe to use, the literature is littered with conflicting results. These inconsistencies are partly attributable to methodological and analytic differences across studies (e.g. sample size, analysis). Overall, interpretation of the literature is highly subjective, subject to theoretical biases, and fails to account for study characteristics that may explain inconsistencies. A meta-analysis (Howitt & Cramer, 2005) would have accounted for various methodological and analytic characteristics and generated effect sizes that statistically quantify the magnitude of group differences or strength of relationships between variables. Since the current literature review did not employ statistical (i.e. meta-analytic) techniques, any conclusions reached are entirely subjective and imprecise.  

Another limitation of this review is that the literature on ethics was not considered. The paucity of randomised controlled trials on this topic probably reflects the ethical constraints associated with withholding necessary (drug) treatment from study participants, merely to see if group differences emerge on various outcome measures. It would have been useful to appraise the various methods used in the few published clinical trials (e.g. Appleby et al, 1997; Wisner et al, 2004) to resolve ethical issues.

Reasons for excluding material

This review generally focused on studies of SSRIs and TCAs, with less emphasis on other less well known antidepressants, such as noradrenergic and specific serotonergic antidepressants (NaSSA), noradrenaline re-uptake inhibitor (NARI), reversible inhibitors of monoamine oxidase (RIMA), and St Johns Wort. Research evidence on these antidepressants is relatively patchy and much more uncertain (Smart & Corbett, 1998).  

CHAPTER FOUR

FINDINGS

The preceding literature review highlights several issues about the effectiveness and safety of antidepressants in PND:

1.    Antidepressants may be effective in reducing postnatal depression;
2.    Antidepressants may produce adverse health effects, but these are generally benign side effects;
3.    There is a paucity of randomised control trials, hampering causal inferences about the safety and effectiveness of antidepressants;
4.    The available evidence is inconclusive due to methodological and analytic constraints (e.g. failure to test for long-term effects, or control for background variables, or account for moderator effects).

Inconclusive evidence

Current literature is inconclusive. On the one hand the preponderance of evidence suggests that antidepressants are safe to use. They are found in very small concentrations in infant serum (Kohen, 2005), and produce little or no adverse effects on the cognitive, motor, and behavioural development of infants (Nulman et al, 1997, 2002). Furthermore, they are effective at alleviating depressive symptomatology (e.g. Wisner et a, 2002; Hendrick, 2003). These observations are generally supported by published reviews of the antidepressant/postnatal depression literature (e.g. Golstein & Sundell, 1999; Simpson & Noble, 2000; Misri & Kostaras, 2002). On the other hand no firm conclusions can be reached until additional research accumulates which address the limitations of existing literature. In particular there is a need to account for possible confounding and moderator variables that may obscure adverse health outcomes. One such factor is the rate of compliance with drug therapy (Boath et al, 2004) – non-compliance may attenuate the true effect of a drug, so that adverse effects may be too insignificant to worry clinicians. Furthermore, there is uncertainty about long-term or delayed effects (Hendrick, 2003). Overall, existing studies seem severely constrained by methodological and analytical weaknesses. Thus, only limited empirical value can be put on the current literature (see Below). 
 

1.    Degree of effectiveness cannot be estimated precisely without the computation of effect sizes
2.    Adverse effects may be obscured by methodological and analytic constraints
 

Implications for clinical practice

The current findings have direct implications for clinical guidelines concerning the use of antidepressants. One of the rationales for conducting this review was to verify certain claims made by the Royal College of Psychiatrists (RCP, 2004). Overall, the review yielded only partial support for these guidelines.  

1.    Firstly, the RCP states that people administered antidepressants show significant improvements over a placebo group. There was some evidence to support this view (e.g. Appleby et al, 1997; Simpson & Noble, 2000; Hendrick, 2003), although other publications reported no significant effects (e.g. Marcus et al, 2005). Overall, the beneficial effects of antidepressants are probable but not assured.

2.    TCAs are considered to be just as effective as SSRIs albeit the latter drug is thought to be safer. The literature is evidence is not clear on this point. There are studies and literature reviews supporting the efficacy of both types of antidepressants (e.g. Goldstein & Sundell, 1999; Kohen, 2005). However, a precise rather than subjective estimate of their relative effectiveness would require meta-analytic procedures, notably the computation of effect sizes. There is a paucity of meta-analytic evidence.

3.    The RCP suggests that MAOIs can induce high blood pressure under certain (dietary) conditions. This review highlights a paucity of literature on the effects of MAOIs on infants (Kohen, 2005). Thus the side effects of these antidepressants in babies in not clear, so it may be safer for mothers to rely on TCAs or SSRIs during pregnancy/breastfeeding.

4.    The RCP proposes that SNRIs are similar to SSRIs in effectiveness. This review yielded insufficient evidence to warrant conclusive inferences on this claim.

5.    The Royal College proposes that the babies of mothers who take antidepressants may experience withdrawal symptoms. Evidence suggests that any adverse effects on infants are generally benign (e.g. Nulman et al, 1997, 2002). However, there is a lack of studies assessing long-term effects (Simpson & Noble, 2000; Hendrick, 2003). Moreover there is evidence of sustained neurobehavioural outcomes in infants (Oberlander et al, 2005).  

6.    The RCP asserts that it is best to carry on taking antidepressants during breastfeeding, because only minute amounts will be transferred to the baby. There is evidence confirming that antidepressants, both TCAs and SSRIs, are found in very small amounts in the serum of breast-fed infants (Kohen, 2005). However caution is advised since overall the evidence base concerning breast-feeding is inconclusive (Simpson & Noble, 2000; Misri & Kostaras, 2002; Hendrick, 2003).

Overall, the general theme emerging from the present review is caution. The paucity of randomised controlled trials, which allow causal inferences, and methodological and analytical limitations of the available literature means that antidepressants must be prescribed guardedly, and probably only as a last resort (after other interventions [e.g. cognitive behavioural therapy] have been considered) and/or when patients present with severe rather than mild depressive symptoms. Any claims regarding the safety and efficacy of antidepressants should be regarded as tentative rather than conclusive. Only the accumulation of more clinical trials and meta-analytic reviews would permit definitive claims on the topic.

CHAPTER FIVE

CONCLUSIONS

The aim of this review was to assess whether antidepressants are a safe and effective option for treating postnatal depression. A mixture of case-control and cohort studies, qualitative studies, and randomised controlled trials, as well as review papers, was considered. Regarding safety, the findings suggest that antidepressants may produce adverse side effects, but these are largely clinically benign. Nevertheless, antidepressants cannot be considered safe based on the available evidence. One reason is that the majority of studies have methodological and analytic constraints, notably failing to account for background variables, moderator effects, and long-term outcomes. Testing for interactions between treatment and other key variables (i.e. moderator effects) is crucial because it may reveal conditions (personal, environmental, clinical etc) under which a previously benign side effect suddenly becomes malignant. Moreover, there is evidence of enduring neurobehavioural and other clinically significant side effects in infants that necessitates further investigation.

Thus, until more research evidence accumulates the safety of antidepressants cannot be assumed.

The evidence suggests that antidepressants are effective at alleviating depression symptoms. However, findings are inconclusive. Some research found no differences between antidepressant and placebo/control groups (e.g. Marcus et al, 2005). Moreover, there is a lack of randomised trials, which would allow causal inferences, and many of the case/cohort-controlled and other studies failed to adequately control for important background variables that might ‘explain’ significant improvements in mood attributed to antidepressants.

Thus, there is a need for better controlled research before conclusive inferences on the effectiveness of antidepressants are justified.

Recommendations

The following recommendations must be considered provisional in view of the methodological and analytic constraints in the literature.

1.    Antidepressants should be used as a last resort. The side effects may be benign but this view is based on imperfect research methodologies that do not account for 3rd variable moderator effects. Moreover, there is a lack of evidence on the long-term effects.

2.    SSRIs and TCAs are preferable over MAOIs. Much of the most up-to-date research evidence relates to the first two classes of antidepressants.

3.    Clinicians should closely monitor patients on antidepressants, to establish whether any improvements are in fact attributable to antidepressants. Most studies do not adequately control for important covariates like patients treatment preferences, level of social support, and stressful life events, all of which may produce improvements in mood. Antidepressants should not be prescribed longer than necessary once evidence suggests that improvements have resulted from other factors.

4.    Clinicians should closely monitor patients to determine whether any serious side effects emerge for a particular patient group or patient characteristic. Current research on adverse effects often fails to account for third-variable moderator effects, whereby a negative health outcome, which is thought to be benign or non-existent, suddenly manifests and/or becomes clinically significant for patients with a particular characteristic.

5.    Fluoxetine should be prescribed if clinically viable. There appears to be least uncertainty about fluoxetine. There is a preponderance of evidence on this SSRI, compared with other antidepressants.

6.    Given the flaws in existing literature antidepressants should only be prescribed for more severe cases of PND. Midwifes can play an important role in identifying such cases (see below).

Implications for midwifes

Midwifes play a key role in screening for antenatal and postnatal depression (Tully et al, 2002). The current review suggests that screening is essential, not just to identify cases of PND, but also to differentiate between mild and more severe cases, and also obtain essential background information on the patient. Although the adverse effects of antidepressants, particularly SSRIs and TCAs, are generally benign, the paucity of randomised clinical trials and long-term assessments means that the safety of antidepressants cannot be taken for granted. Thus, the RCP (2004, 2006) recommends that antidepressants be used for severe or protracted (rather than milder) cases of PND. Midwifes have a responsibility for identifying such cases, primarily using the EPDS. However, it remains unclear the extent to which midwifes are trained for and involved in PND screening.

Consequently Tully et al (2002) conducted a study to assess the experience and role of midwifes in identifying and referring antenatal and postnatal depression cases, and current policies and practices in maternity units on this issue. Any deficiencies in training, practice, and policies would need to be addressed, in view of the importance of identifying mild and severe cases of PND. Questionnaire data was collected from 182 maternity units in England and Wales. A letter was sent to the head or senior midwife in each unit, with a request that the questionnaire be completed by the midwife in the unit with the most experience of antenatal and postnatal depression. The questionnaire contained items concerning policies, guidelines, and practices regarding screening and management of pre- and postnatal depression.

Regarding policies/guidelines, only a minority (8%) of units had regulations on the management of PND. Furthermore, just over half (58%) undertook universal screening of PND, with 42% not doing so. Of the units that performed universal screening, the vast majority (93%) used the EPDS, while the rest employed other screening tools. Additionally, screening was undertaken primarily by a health visitor in the bulk of units (91%). By comparison only a small minority of units (4%) had midwifes perform screening. However, almost half (46%) indicated that the supervisor of midwifes is routinely notified about any referrals when a midwife identifies a case of PND. When depression was detected, 35% of units left it entirely up to the midwife to refer the case to other professionals. Other units referred cases using a mixture of health professionals including midwifes.

Only a 16% of units provided midwifes with specific training in screening women for depression.

Midwifes and their colleagues reported a number of problems encountered including difficulties accessing referral services (some clinical consultants would not accept midwifery referrals) and lack of adequate training. The current review suggests that it is essential that midwifes receive the necessary training and support to effectively screen women for PND, and hence identify the more severe cases that may require antidepressants.

Furthermore, given the paucity of studies accounting for important confounding variables, it is essential that midwifes obtain comprehensive background information from each patient – this would provide an essential database to facilitate further randomised clinical trials and cohort/case-controlled studies which account for key covariates in testing assessing the efficacy and safety of antidepressants.

Final word

Scientific research findings are never conclusive. Even well designed clinical trials often raise more questions than they address. Empirical literature suggests that antidepressants, notably SSRIs and TCAs, can be effective in treating PND, with side effects that are mostly benign. However, the evidence is plagued by various methodological and analytic constraints, so that any conclusions reached are highly tentative at best. Thus, caution is advised in the use of antidepressants: consistent with national guidelines on the treatment of depression (e.g. DOH, 2004; NICE, 2004), it is recommended that antidepressants are not used unless PND symptoms are judged to be severe. Ultimately literature reviews such as this one can be highly subjective in the evaluation of studies and interpretation of findings. Thus, there is a need for meta-analytic reviews, which take account of methodological and statistical factors in each study and report effect sizes. This would provide a more accurate estimate of the clinical benefits of antidepressants and the seriousness of any side effects.