RESULTS

Table 1 summarizes the study population and the interventions. More detailed information about the setting and the country, number of participants, participation proportions, method of selection for selective and indicated interventions, method of randomization, attrition, participation in the intervention, loss to follow-up and the interventions themselves are summarized from data in the original publications on the following Web site:www.latrobe/csmch/prism.

There are five trials of universal antenatal interventions. The first was a group intervention, developed by Gordon and Gordon with a strong didactic and advisory focus, implemented within a childbirth education program (39–41). Antenatal classes rather than individual women were randomized, with stratification for partner attendance. The second universal intervention, implemented in three trials, one in the Scotland (88;95) and two in Australia (8;9;96;97), involved a comparison between standard maternal and perinatal hospital and community care through the whole pregnancy, labor, and postnatal period with new programs of team midwifery care planned to enhance continuity of care and to increase the chance that there would be continuity of carer as well. The third universal intervention was a multifaceted psychoeducational intervention about perinatal depression developed for women having their first child, delivered on a one-to-one basis by an experienced midwife (49).

Characteristics of the seven selective antenatal interventions, all of which followed screening to select women at increased risk of postnatal depression, are summarized next. Elliott and colleagues (31;32;63) carried out the first controlled trial, a psychoeducational group intervention to normalize the experience and empower participants. The intervention was implemented in groups in four other trials (Buist [16], Brugha and colleagues [14;15;99;100], Stamp and colleagues [92;93], Zlotnick and colleagues [105]). Although they had many features in common, these trials differed in the number of sessions provided, the theoretical basis for the intervention, and the balance between education, support, problem-solving, and the acquisition of skills. Attendance at the groups also varied as did the completeness of follow-up. The other two trials were implemented on a one-to-one basis. One, carried out by Marks and colleagues (65) involved enhanced continuity of midwifery care, similar to those described among the universal antenatal trials, offered to women with a history of major depressive disorder. This was designed to provide social support, enhance early detection of symptoms, and speed referral and treatment if that became necessary. The final trial in this group had an education and additional support focus, providing information to the women identified as at-risk for postnatal depression, informing her primary care providers about her increased risk, and offering referral within antenatal care to psychiatry or social work (98).

Only one trial of an indicated antenatal intervention for women with antenatal depression was found (91). This compared individual interpersonal psychotherapy with individual parenting education.

Interventions in labor, evaluated in four trials, all involved the provision of continuous labor support. In three of them, support was provided by a “doula” or lay female companion, usually from the local community, to be with the woman during labor. One trial was completed in South Africa (80;104) and the other two in the United States (38;57). The fourth compared continuous support from a trained labor support nurse with standard care in a large study of Canadian and U.S. hospitals (55).

Six diverse trials of potentially universal postnatal interventions have been reported. Two, carried out during the postnatal stay in hospital, involved postnatal “debriefing” or “listening and discussion” by a midwife (61;85). The other four were changes to the provision of standard posthospital postnatal care. The first changed the timing of the postnatal check from six weeks after birth to a week after hospital discharge (44). The second offered a substantial increase in practical support at home (72;73). The third, using a factorial design, provided an information package, an invitation to a new mothers' group, both interventions or neither (87). The fourth was a complex design (cluster-randomized) and a complex intervention—evidence-based, re-designed postnatal care—compared with the usual care provided by community midwives. It involved randomization of the general practices to which the midwives were attached (67).

Four trials of very different selective postnatal interventions were identified. One was a trial of midwifery-led debriefing in hospital offered to women who had experienced an operative birth (90). The second provided a home-visiting support and educational intervention for women with major social risk factors (3;4). The third involved screening women in hospital with the Edinburgh Postnatal Depression Scale (EPDS) to detect those with moderate levels of depressive symptoms and offering them an immediate intervention before hospital discharge with educational, supportive, and cognitive-behavioral components provided by a trained and supervised therapist (18;19). The fourth also screened women with the EPDS after birth. Those with scores >9 were recruited to a trial of telephone-based peer support, provided by trained volunteers who had themselves recovered from an episode of postnatal depression (28).

Eleven trials of indicated postnatal interventions for women diagnosed with depression or probable depression were found. Seven were counseling interventions (a range of modalities), provided by a trained primary care provider such as a Health Visitor (58); a Child Health Nurse, (101); by a psychologist, therapist, or psychiatrist (2;18;19;71;82); or both (24). The trial of Appleby and colleagues had a complex factorial design involving one or six sessions of cognitive-behavioral counseling and either fluoxetine or placebo, with breastfeeding women excluded (2). The trial of Cooper and colleagues compared three different modalities (nondirective counseling, cognitive-behavioral therapy, and psychodynamic therapy) with routine primary care and also compared their implementation by specialists or generalists with additional training (24;75). In the one trial where the intervention was provided by a psychiatrist, the comparison arm was standard psychiatric follow-up for six weeks with the intervention involving the partner's attendance at the second and fourth visits and a focus on increasing the partner's support (71). Three trials implemented a group intervention (20;36;59). Like the trials described earlier in relation to selective antenatal interventions the groups combined educational and supportive components. The final trial by Field and colleagues compared massage therapy with relaxation therapy for disadvantaged and depressed young mothers (35).

Four trials—three in which women were selected as being at increased risk of depression, and one for women currently depressed—focused on mother-infant interaction. The first, which extended over a year, was a weekly home-visiting, relationship-based intervention offered to women characterized by poverty and a lack of support, identified in late pregnancy and having their first child (50). The second identified women two to four weeks after birth with mild to severe depressive symptoms and randomized them to receive a coached behavioral intervention designed to promote maternal-infant responsiveness (60). The third offered a sleep management plan and advice on “controlled crying” from a senior pediatric trainee, to mothers of six- to twelve-month-old babies with severe sleep problems. Women in both the intervention and comparison arms received information about normal sleep (51). The fourth trial compared a support group plus infant massage classes with a support group alone for women identified as depressed four weeks after birth (83).

The mental health outcomes of these trials are given in Tables 2 through 8. Among the universal antenatal interventions the trial by Gordon and Gordon (40) appeared to be highly effective and the trial was very influential for the design of subsequent trials. Although its design, with randomization of classes is very appropriate, the analysis is more complex, because the assumption that all the participants are independent no longer holds. In addition, the outcome measure, although well designed and measured, is not sufficiently similar to depression and probable depression for the trial to be included. Hayes' one-to-one intervention was not effective, despite its careful design, thoughtful intervention, and implementation (49). The three team midwifery care trials designed to improve continuity of care and carer all had very similar results with overlapping confidence intervals, with no consistent evidence of a beneficial effect on postnatal depression (8;88;96).

Table 3 provides outcome data on selective and indicated antenatal interventions. The trial of Elliott and colleagues (31), like the Gordons' trial, has been very influential, and like that trial reported an intervention that was effective. However, the trial was not randomized but had historical controls and, thus, has been excluded. The three group interventions that followed were not effective, two of them reporting problems with participation in the program (14;15;92;93). The fourth had promising results, although the trial was very small, with only thirty-five participants in total (105). The enhanced continuity of care trial had no effect on depressive illness either in pregnancy or after birth: 23 percent of women in both arms were depressed in the first three months after birth (65). Webster's intervention, which combined careful screening during antenatal care with a low-key, low-cost intervention, was not shown to be effective (98). The only trial of an intervention to reduce current major antenatal depression (91), was effective in reducing the severity of depression and improving depressive symptoms, measured with the Clinical Global Impression. Drop-outs and low levels of participation were high in this study where the women, many of them immigrants, lived in highly adverse social circumstances.

The mental health benefits of doula care in labor, found at short-term follow-up in one South African trial (104), were not sustained some weeks later (80) and have not been replicated either in the other doula trials or in a large trial of continuous labor support by a trained labor support nurse (38;55;57; Table 4).

The universal postnatal interventions summarized in Table 5 include two debriefing interventions. The very large trial of Priest and colleagues virtually ruled out any substantial effect of this intervention on depression in the first year after birth (85). The trial of Lavender and Walkinshaw (61) assessed outcomes very soon after birth (three weeks) and also found extremely high levels of anxiety and depression in the standard care arm, both of which make it unlikely that the very large reported benefit can be expected elsewhere; this trial is a true outlier. The other universal trials had several points in common: all were interested in physical health as well as mental health outcomes, all used the SF-36 and the EPDS as standard outcome measures, and all measured outcomes at several time points after birth. The striking finding is that three of the four trials had no effects on mental or physical health (44;72;87), but the fourth demonstrated a large mental health benefit in terms of mean EPDS and SF-36 mental health scores and probable depression (67). MacArthur and colleagues developed a very careful and relatively intensive intervention, embedded in existing community-based midwifery postnatal services.

The four selective postnatal interventions (Table 6) are quite diverse. Another large debriefing trial, by Small and colleagues, restricted to women who experienced an operative or assisted delivery, was unable to detect any benefit of the intervention (90); the same finding as that of Priest and colleagues above. Armstrong and colleagues' intensive home-visiting intervention provided to women as at-risk on social grounds had a large effect in the short term that was not sustained through to four months (3;4). In contrast, the postscreening psychological intervention of Chabrol and colleagues during the postnatal hospital stay was associated with a large reduction in the proportion of women having moderate to severe depressive symptoms four to six weeks later (18;19). Similarly, the telephone-based peer support provided to women after screening in Canada was effective in reducing probable depression (28).

The outcomes of indicated postnatal interventions summarized in Table 7 demonstrate a consistent and substantial reduction in depression/probable depression after the six counseling interventions, regardless of the model of therapy or the professional background of the therapist (18;19;24;58;71;82;101). Cooper and colleagues, in a recent more detailed report of their trial, show that while all three treatments had a significant impact on maternal depressive symptoms at 4.5 months, only psychodynamic therapy reduced clinically diagnosed depression (24). By nine months after birth, spontaneous recovery among those receiving routine primary care meant that women who had received the therapies no longer differed from those in primary care. These authors also point out that, despite the high acceptability of the treatments offered, a minority of women were reluctant to accept home visiting and they suggest that these women are likely to be particularly at-risk of adverse outcomes (75). The trial of Appleby and colleagues, which combined fluoxetine or placebo with one or six counseling sessions, reported highly significant improvement in all four groups, with a greater improvement in those receiving fluoxetine than placebo (2). The three group interventions had rather different interventions and different outcomes, with the latter two identifying a probable benefit of participation on depression (20;36). The trial comparing massage therapy with relaxation therapy found the former to be more effective on each occasion, but there were no effects sustained over time (35).

Table 8 summarizes the impact of four different interventions focused on mother-infant interaction. Those with a focus on mother-infant interaction itself had no effect on depression (50;60). Baby massage classes and a support group were reported to have a beneficial effect on one measure, but as with other group trials, there was no adjustment for within-group effects (83). The trial of Hiscock and colleagues that aimed to improve infant sleep patterns did not have a significant effect on the proportion of women depressed at two or four months, but the stratified analysis demonstrated a marked reduction in depressive symptoms associated with the intervention (51).

Fixed and random effects meta-analysis comparing interventions grouped by their timing and the nature of the intervention are shown in Table 9. Trials were excluded if they did not include a measure of depression or probable depression (2;35;49;50;60;83), if there were only summary statements on the outcomes (36;38;57), or the intervention was with the women's partner (71). Trials were included in the meta-analyses if the mental health outcome was depression or probable depression, if the intervention was compared with standard care or routine primary care, or if the comparison group received an intervention perceived as a placebo. Examples in the latter category were individual parent education (91), a leaflet about practical sources of help after operative birth (90), being put on a waiting-list (82), or an information sheet on normal sleep patterns (51).

The pooled data show that only the indicated postnatal interventions had a substantial impact on reducing postnatal depression and that there was no significant heterogeneity across the trials in this category. The one reservation to be made in relation to this inference is that graphical representation [not shown] of these trials in terms of study power suggests that there may be some publication bias with “missing” small negative trials. Both individual and group interventions were effective.

The marked heterogeneity in the universal postnatal trials reflects the individual trial outcomes (Table 5). Four large trials of different supportive interventions (44;72;85;87) were able to exclude any worthwhile effect, while another large trial (67) of a different and complex intervention was highly effective. (One small trial with several measurement problems also reported a very large effect [61]). Similarly, the selective postnatal interventions, where there was significant heterogeneity, included two small trials with promising results (19;28), one large trial showing no impact on depression (90), and another whose promising early effects were not sustained (3;4). It is also possible to calculate from the risk difference across the trials the number of women needing to be given an effective intervention (the number needed to treat, (NNT) for one “case” of depression or probable depression to be prevented or treated (1). For individual postnatal counseling in women diagnosed as depressed, NNT=3, with a 95 percent CI of 2 to 4. The NNT for women receiving re-designed postnatal care is fourteen, with a 95 percent CI of 9 to 33.