Evidence reviews for the US Preventive Services Task Force (USPSTF) use an analytic framework to visually display the key questions that the review will address to allow the USPSTF to evaluate the effectiveness and safety of a preventive service. The questions are depicted by linkages that relate interventions and outcomes. Refer to the USPSTF Procedure Manual for interpretation of the analytic framework.14 HrQoL indicates health-related quality of life.
KQ indicates key question.
aReasons for exclusion: Aim: Study aim was not relevant. Setting: Study was not conducted in a country relevant to US practice, or not conducted in, recruited from, or feasible for primary care or a health system. Comparative effectiveness: Active comparator (eg, liquid-based cytology vs conventional cytology alone). Outcomes: Study did not have relevant outcomes or had incomplete outcomes. Population: Study was not conducted in an included population. Intervention: Intervention was out of scope. Design: Study did not use an included design. Language: Publication not in English. Quality: Study was poor quality. Unable to locate: Review staff was unable to locate article.
bStudies may appear in more than 1 KQ.
Weights are from random-effects analysis. BDI-II indicates Beck Depression Inventory II; CBT, cognitive-behavioral therapy; EPDS, Edinburgh Postnatal Depression Scale; IPT, interpersonal therapy; LQ, Leverton Questionnaire; MT, motivational therapy; NR, not reported; RR, risk ratio.
a”g” indicates during gestation and “p” indicates postpartum; thus, for example, g37 indicates 37 weeks’ gestation and p12 indicates 12 weeks postpartum.
bStudy-reported adjusted analysis was not statistically significant, although effect size shown in the forest plot, based on unadjusted data, is statistically significant. Tourne indicates a psychosomatic humanist group intervention developed by Dr Claude-Emile Tourne.
Weights are from random-effects analysis. BDI-II indicates Beck Depression Inventory II; CES-D, Center for Epidemiologic Studies Depression Scale; DHA, docosahexaenoic acid; EPA, eicosapentaenoic acid; EPDS, Edinburgh Postnatal Depression Scale; NR, not reported; PA, physical activity; PPD, postpartum depression; RR, risk ratio.
bStudy-reported adjusted analyses were statistically significant, although effect size shown in the forest plot, based on unadjusted data, are not statistically significant.
eTable 1. Characteristics of Included Studies, by Intervention Type
eTable 2. Summary of Intervention Characteristics, by Intervention Type
eFigure 1. Forest Plot of Standardized Mean Differences in Depression Symptom Scores for Counseling Interventions, Sorted by Follow-up Time
eFigure 2. Forest Plot of Unstandardized Mean Differences in Depression Symptom Scores for Counseling Interventions, Sorted by Follow-up Time
Customize your JAMA Network experience by selecting one or more topics from the list below.
O’Connor E, Senger CA, Henninger ML, Coppola E, Gaynes BN. Interventions to Prevent Perinatal Depression: Evidence Report and Systematic Review for the US Preventive Services Task Force. JAMA. 2019;321(6):588–601. doi:10.1001/jama.2018.20865
Depression during pregnancy and the postpartum period is relatively common and can have adverse effects on both mother and child.
To systematically review benefits and harms of primary care–relevant interventions to prevent perinatal depression, a major or minor depressive episode during pregnancy or up to 1 year after childbirth, to inform the US Preventive Services Task Force.
MEDLINE, PubMED (for publisher-supplied records only), PsycINFO, and the Cochrane Central Register of Controlled Trials; surveillance through December 5, 2018.
Randomized clinical trials (RCTs) and nonrandomized controlled intervention studies of interventions (eg, behavior-based, antidepressants, dietary supplements) to prevent perinatal depression in general populations of pregnant and postpartum individuals or in those at increased risk of perinatal depression. Large cohort studies were considered for harms of antidepressant use only.
Data Extraction and Synthesis
Two investigators independently reviewed abstracts and full-text articles and quality rated included studies. Random-effects meta-analysis was used to estimate the benefits of the interventions.
Main Outcomes and Measures
Depression status; depression symptoms; maternal, infant, and child health outcomes.
Fifty studies (N = 22 385) that met inclusion criteria were identified. Counseling interventions were the most widely studied interventions. Compared with controls, counseling interventions were associated with a lower likelihood of onset of perinatal depression (pooled risk ratio [RR], 0.61 [95% CI, 0.47-0.78]; 17 RCTs [n = 3094]; I2 = 39.0%). The absolute difference in the risk of perinatal depression ranged from 1.3% greater reduction in the control group to 31.8% greater reduction in the intervention group. Health system interventions showed a benefit in 3 studies (n = 5321) and had a pooled effect size similar to that of the counseling interventions, but the pooled effect was not statistically significant using a method appropriate for pooling a small number of studies (restricted maximum likelihood RR, 0.58 [95% CI, 0.22-1.53]; n = 4738; I2 = 66.3%; absolute risk reduction range, −3.1% to −13.1%). None of the behavior-based interventions reported on harms directly. A smaller percentage of participants prescribed sertraline had a depression recurrence compared with those prescribed placebo (7% vs 50%, P = .04) at 20 weeks postpartum in 1 very small RCT (n = 22 analyzed) but with an increased risk of adverse effects to the mother.
Conclusions and Relevance
Counseling interventions can be effective in preventing perinatal depression, although most evidence was limited to women at increased risk for perinatal depression. A variety of other intervention approaches provided some evidence of effectiveness but lacked a robust evidence base and need further research.
Perinatal depression is a common condition that was estimated in 2012 to affect more than 180 000 new mothers (11.5%) annually in the United States and that can have a devastating effect on the mother as well as the infant.1 Perinatal depression is defined as the occurrence of a major or minor depressive episode during pregnancy or up to 1 year after childbirth.2 In addition to the typical symptoms of depressive disorders (eg, feeling hopeless, loss of interest in activities that used to be enjoyed, withdrawing from friends and family), other symptoms in the perinatal period may include persistent doubt of the ability to take care of the infant, trouble bonding with the infant, and thoughts of self-harm or harm of the infant.3
Risk factors that can be used to identify individuals at risk for perinatal depression include a history of depression,4-7 history of physical or sexual abuse,5 unplanned or unwanted pregnancy,8 stressful life events,1,5,9 intimate partner violence,10,11 and complications during pregnancy.12 Additionally, low socioeconomic status, lack of social support, and bearing children during adolescence have been associated with a greater risk of developing perinatal depression after delivery.5,6,8,13 Numerous interventions have been proposed to prevent perinatal depression; however, there is no commonly agreed-on method of prevention. Thus, there is likely substantial variation in clinical practice. Although there are risk factors for perinatal depression and interventions exist that may help prevent perinatal depression, the effectiveness of these interventions and the subpopulations who could most benefit need further evaluation.
There are currently no clinical guidelines on how to prevent perinatal depression and no prior US Preventive Services Task Force (USPSTF) recommendation on this topic. This systematic review was conducted to synthesize the evidence related to the effectiveness of interventions in preventing perinatal depression to support a new USPSTF recommendation.
The USPSTF commissioned this review to evaluate direct evidence from trials and large cohort studies (for harms of antidepressant use only) on interventions to reduce the risk of perinatal depression initiated during pregnancy or the first year postpartum. Specifically, the 2 key questions (KQs) (Figure 1) aimed to identify the benefits (KQ1) and harms (KQ2) of interventions to prevent perinatal depression for pregnant or postpartum individuals and their children. Additional methodological details regarding the review search strategies, detailed study inclusion criteria, quality assessment, excluded studies, and description of data analyses are publicly available in the full evidence report at http://www.uspreventiveservicestaskforce.org/Page/Document/UpdateSummaryFinal/perinatal-depression-preventive-interventions.
Comprehensive literature searches were performed for primary literature in MEDLINE, PubMed (for publisher-supplied records only), PsycINFO, and the Cochrane Collaboration Registry of Controlled Trials from January 2012 through February 6, 2018. Database searches were supplemented with suggestions from preidentified experts in the field and by reviewing reference lists from other relevant systematic reviews. After February 2018, ongoing surveillance continued through article alerts and targeted searches of high-impact journals to identify major studies published in the interim that could affect the conclusions or understanding of the evidence and affect the related USPSTF recommendation. The last surveillance was conducted on December 5, 2018, and resulted in the addition of no new studies.
Two reviewers independently reviewed abstracts and full-text articles against specified inclusion criteria (Figure 2). Studies were eligible if they were published in English, conducted in countries ranked as having “very high” human development according to the World Health Organization, and included pregnant persons or mothers up to a maximum of 1 year postpartum. Studies limited to persons with mental health symptoms or disorders (eg, anxiety disorders) were eligible; however, studies limited to perinatal individuals currently experiencing or being treated for a depressive episode were excluded, as were studies limited to persons with psychotic or developmental disorders. In addition, studies limited to persons with a medical condition (eg, HIV/AIDS), and those limited to persons in institutions (eg, psychiatric inpatients) or long-term care or residential facilities were excluded because of generalizability concerns. Studies that included a subset of these types of participants were included; however, it was required that the number not exceed 50% of the total sample to be considered for inclusion.
Studies were required have a primary or secondary aim to prevent perinatal depression. The following interventions were included: counseling (eg, cognitive-behavioral therapy [CBT], interpersonal therapy [IPT]), psychoeducation, or other supportive interventions (eg, peer mentoring); care delivery models targeting improved mental health outcomes; prophylactic use of antidepressants; widely available physical activity or complementary and alternative therapies; and hormonal therapy. Pharmacotherapy harms were only to be evaluated for medications that were found to support the prevention of perinatal depression and were to be examined only during the phase (pregnancy or postpartum) in which the evidence was identified. Interventions composed of general parenting education without a mental health component (eg, prenatal or infant care classes) were excluded.
Depression diagnosis (determined through a clinical interview) or symptoms (measured using a validated instrument) were a required outcome for included studies. Other maternal health outcomes, infant and child outcomes, birth outcomes, and any information reported on harms were also abstracted. Relevant outcomes reported at least 6 weeks after the baseline assessment or intervention initiation were included, although harms outcomes reported any time after the intervention was initiated were considered.
Interventions that were conducted in or recruited from primary care or a health care system, or that could potentially be implemented in or referred from primary care, were included. This included interventions taking place in primary care clinics; prenatal clinics; obstetrics/gynecology clinics; pediatric clinics; family planning clinics; military health clinics; school-based health clinics; mental health clinics; and research settings, homes, or other community settings, including electronic or computer-based interventions. Studies conducted in correctional facilities, school classrooms, worksites, and emergency departments were excluded.
Two reviewers applied USPSTF design-specific criteria14 to assess the methodological quality of all eligible studies. Each study was assigned a quality rating of “good,” “fair,” or “poor.” Discordant quality ratings were resolved by discussion or by a third reviewer and adjudicated as needed. Studies were rated as poor quality and excluded if there was a major flaw such as very high attrition (generally >40%), differential attrition between intervention groups (generally >20%); substantial lack of baseline comparability between groups without adjustment; or major concerns about the trial conduct, analysis, or reporting of results. One investigator extracted study-level data using data entry forms developed in DistillerSR (Evidence Partners) and a second investigator confirmed the accuracy of the data.
Summary tables showing study, population, intervention characteristics, and outcomes were created. Studies were examined overall and grouped according to intervention type: counseling, health system, physical activity, education (without counseling, extensive skills practice, or other supportive interventions), support (without counseling or skill-building), infant sleep, debriefing (exploring the events and emotions of the birth experience, with a counselor providing normalization and education), other behavior-based approaches, antidepressants, and supplements. The intervention categories were developed post hoc, and some trials were difficult to categorize and could possibly have fit into more than 1 category. The one that appeared to have the best fit was chosen by the primary investigator.
Because of its clinical utility, depression status was chosen as the primary outcome. Most trials reported a related dichotomous depression outcome: cumulative incidence of depression, prevalence, or the proportion scoring above a cutoff on a symptom severity scale. Since most trials excluded women with depression or high symptom levels at baseline, it was assumed that most cases of depression identified after the start of the study would be new-onset cases, but not necessarily first-onset cases, since many women had previous episodes of depression.
Strength of evidence was rated for each key question, based on consistency (similarity of effect direction and size), precision (degree of certainty around an estimate), reporting bias (potential for bias related to publication, selective outcome reporting, or selective analysis reporting), and study quality (ie, study limitations).
Random-effects models on both the main outcome of depression status and continuous measures of depression symptom severity were conducted, both overall and separately by intervention. The DerSimonian and Laird model for pooling was used, and I2 and Q statistic were calculated to test for heterogeneity. In addition, because the DerSimonian and Laird method is prone to insufficient coverage of the full 95% confidence intervals when the number of studies is small and statistical heterogeneity is high, restricted maximum likelihood models with the Knapp-Hartung correction for small samples were used when fewer than 10 trials were pooled and the DerSimonian and Laird model showed a statistically significant effect. For the full body of evidence, a funnel plot was generated and the Egger test was performed to explore small-study effects, which can be related to publication bias.15 Additionally, meta-regression and sensitivity analyses were conducted to explore factors associated with effect size for the dichotomous depression status outcome.
Stata version 15.1 (StataCorp LP) was used for all analyses. All significance testing was 2-sided, and results were considered statistically significant if the P value was .05 or less.
Two reviewers independently assessed 1036 abstracts and reviewed 247 full-text articles. In total, 50 studies (8 good quality, 42 fair quality; N = 22 385; 49 randomized clinical trials [RCTs]16-64 and 1 nonrandomized controlled intervention study65) were included (Figure 2; eTable 1 in the Supplement). Of the 50 included studies, 20 (40%) were conducted in the United States, and most recruited women from primary care or obstetrics/gynecology practices (33/50 [66%]) or from other clinical settings (13/50 [26%]) such as in the hospital postdelivery, through electronic medical records, or in clinic- or hospital-based childbirth education classes (Table 1; eTable 2 in the Supplement). Twenty-six of the included studies (52%) recruited pregnant women, 22 (44%) recruited postpartum women, and 2 (4%) recruited women who were pregnant as well as those up to 26 weeks postpartum.16,17 Most studies (42/50 [84%]) were limited to women 18 years or older, but 1 was limited to adolescents18 and 7 had no age restrictions.19-25 The studies assessed the effect of a wide range of intervention approaches, including counseling, health system–level interventions, physical activity, supportive interventions, education, infant sleep advice, birth-experience postpartum debriefing, expressive writing, yoga, omega-3 fatty acids, sertraline, and nortriptyline. For all KQs, additional descriptive and outcome data are available in the full report.
Twenty-seven studies (54%) selected women at increased risk for perinatal depression, such as having a personal or family history of depression (or perinatal depression), elevated depressive symptoms, or socioeconomic (eg, low income, single, young, recent intimate partner violence) or mental health (eg, elevated anxiety symptoms) risk factors (Table 1). The most common approach was to select women on the basis of depression symptoms or history.16,17,26-35 The Edinburgh Postnatal Depression Scale (EPDS; range, 0-30; higher score indicates greater distress) and the Center for Epidemiologic Studies Depression Scale (CES-D; range, 0-60; higher score indicates greater distress) were the most widely used tools for identifying women at risk for developing postpartum depression in the included studies.
Although the majority of participants in the included studies were non-Hispanic white (69% of all participants in trials that reported race/ethnicity), 2 trials were limited to Latina women,27,28 and 8 had majority black and Latina samples.16-18,36-40 In addition, 13 studies (26%) were primarily or entirely composed of economically disadvantaged women.16,17,22,27,28,37-44
Key Question 1. Do interventions to prevent perinatal depression improve health outcomes in pregnant or postpartum women or their children?
Key Question 1a. In trials that limit enrollment to high-risk women, how are participants identified as being at high risk of developing perinatal depression?
Twenty RCTs (2 good quality, 18 fair quality) of counseling interventions were identified (n = 4107). Seventeen of these reported incidence, prevalence, and exceeding symptom cutoff and are included in this pooled estimate; the 3 studies43,45,46 that did not report these dichotomous measures all reported continuous measures of depressive symptoms, with mixed findings. The pooled risk ratio (RR) for counseling interventions was 0.61 when the outcomes of incidence, prevalence, and exceeding symptom cutoff were combined (95% CI, 0.47 to 0.78; 17 trials [n = 3094]; I2 = 39.0%) (Figure 3). The proportion of participants with depression according to any of the dichotomous depression outcomes at the main time point of 26 weeks postpartum (or the closest to this time point) ranged from 0% to 34% in the intervention groups, compared with 6% to 50% in the control groups, with absolute risk differences (ARDs) ranging from 1.3% greater reduction in the control group to 31.8% greater reduction in the intervention group. Of these 17 trials, 13 (76%) reported an outcome of major depressive disorder diagnosis based on a clinical interview. Trials reported depression outcomes over a wide range of follow-up time points, ranging from 6 to 52 weeks postpartum.
Effects were largest for CBT- and IPT-based interventions (Table 2), and the 2 most commonly used approaches were the CBT-based Mothers and Babies program (used in 4 studies16,17,27,28) and the IPT-based Reach Out, Stand Strong, Essentials for New Mothers (ROSE) program (used in 5 studies18,37-40). The Mothers and Babies program involved 8 to 17 group sessions during pregnancy and postpartum, with a goal of helping participants create a healthy physical, social, and psychological environment for themselves and their infants. The ROSE program involved 4 to 6 sessions during pregnancy and postpartum, covering topics such as stress management, development of a social support system, role transitions and changes associated with role transitions, and types of interpersonal conflicts common around childbirth.
When limited to trials that only included women at increased risk of perinatal depression, the pooled RR was 0.55 (95% CI, 0.44 to 0.68; 14 trials [n = 1411]; I2 = 0%) (Table 2). There was a statistically significant small-studies effect for the counseling trials (Egger test, −1.52; P = .01). Smaller trials were more likely to limit inclusion to populations selected for increased risk of perinatal depression, which may have been a major source of the association between effect size and study size.
Overall, counseling interventions were associated with a small beneficial effect in symptom score measures, amounting to a pooled standardized effect size of 0.2, which would generally be considered a small effect,66 or a 1.5-point greater reduction in depression symptom severity than control conditions (standardized mean difference, −0.21 [95% CI, −0.40 to −0.02]; 13 trials [n = 1367]; I2 = 57.2%) [eFigure 1 in the Supplement]; weighted mean difference in change between groups, −1.51 [95% CI, −2.84 to −0.18]; 13 trials [n = 1367]; I2 = 61.3% [eFigure 2 in the Supplement]). This analysis combined a variety of instruments with 30- to 63-point ranges. Thirteen trials reported continuous symptom score measures and showed a wide range of results; however, group differences were statistically significant in only 5 trials.16,17,29,37,42
Most of the counseling trials also reported other maternal or child outcomes; however, there was a wide variety of outcome measures and little consistency across studies. Stress and anxiety were the most commonly reported other maternal or child outcome. For example, 4 trials42,46-48 reported a measure of stress, but most did not show statistically or clinically important differences between groups. Other outcomes, generally reported by only 1 or 2 studies, included measures of functioning (general,19,38 maternal,21 and family21,42,48), quality of life,48 social support,16,42 trauma symptoms,39 mental health treatment,40 breastfeeding,38 child development,49 child attachment,49 birth weight,42 and preterm birth.42 Of these, 1 trial (n = 184) showed statistically significant benefits on birth weight (between-group difference, 283 g; P = .01) and incidence of preterm birth (RR, 0.19 [95% CI, 0.06 to 0.65]).42 No other outcomes were statistically significant.
Fifteen of the 20 trials of counseling interventions (75%) were limited to women who were known to be at increased risk of perinatal depression, owing to depression history or symptoms (6/20 [30%]), non–depression-related risk factors (3/20 [15%]), or either depression-related or other risk factors (6/20 [30%]). Thirteen of the 20 trials (65%) excluded women who met diagnostic criteria for current major depression or scored above a prespecified cutoff on a symptom severity scale. The trials that did not exclude women with a depression diagnosis or high symptom level used either unselected populations or selected participants based on non–depression-related criteria, so the proportion with depression was estimated or reported to be well below 50%.
Counseling interventions lasted a median of 8 weeks (range, 4-70 weeks), included a median of 8 sessions (range, 4-20 sessions), and had an estimated median of 12 hours of contact (range, 4-23.3 hours). Fifteen (75%) included group sessions, 11 (55%) included individual sessions, and 3 intervened with couples (eTable 2 in the Supplement).19,45,50 Most of the interventions used CBT or IPT approaches. Information on adherence for counseling and other intervention approaches is available in the full evidence report.
Three fair-quality studies (n = 5321; 2 RCTs,23,51 1 nonrandomized controlled intervention study65) examined the effects of health system–level interventions. All 3 programs showed beneficial effects on depression, with RRs ranging from 0.33 (95% CI, 0.19 to 0.58; n = 433; ARD, −13.1%) to 0.71 (95% CI, 0.53 to 0.95; n = 2241; ARD, −3.1%) for exceeding a specified depression symptom level at follow-up. The pooled RR for health system interventions was 0.60 (95% CI, 0.43 to 0.83; 3 studies [n = 4738]; I2 = 66.3%) (Figure 3). However, this association was not statistically significant with the restricted maximum likelihood analysis, which better accounts for the small number of studies pooled (RR, 0.58 [95% CI, 0.22 to 1.53]). None were conducted in the United States, so applicability to this country may be limited. Additionally, none limited their sample to women at increased risk of depression, suggesting that some universal prevention programs may be effective.
Health outcomes were sparsely reported in these studies and showed mixed results (see full evidence report for more details). No infant or child outcomes were reported in these trials.
A wide variety of other intervention approaches were identified. Some of these approaches showed benefit in some trials, but most trials did not find statistically significant group differences (Figure 4, for studies reporting sufficient data to plot). The intervention focus of these studies included physical activity (3 RCTs [n = 1200]),52-54 education (without counseling or extensive support; 6 RCTs [n = 2949]),20,30,41,55-57 supportive interventions (without formal counseling; 7 RCTs [n = 4569]),22,24,31,44,58,59 infant sleep advice (3 RCTs [n = 980]),36,60,61 birth-experience postpartum debriefing (2 RCTs [n = 2786]),25,62 expressive writing (1 RCT [n = 120]),63 antidepressants (2 RCTs [n = 80]),33,34 supplements (2 RCTs [n = 227]),35,64 and yoga (1 RCT [n = 46]).26
Of these approaches, the physical activity interventions consistently reported point estimates in the direction of benefit (ARDs ranged from −1.3% to −12.5%), but only 1 trial found statistically significant group differences.53 Birth-experience debriefing25,62 and omega-3 fatty acid supplementation35,64 showed no benefit (study RRs ranged from 0.99 [95% CI, 0.87 to 1.11] [n = 1745] to 2.70 [95% CI, 0.56 to 13.09] [n = 79]). Additional sensitivity and subgroup analyses exploring effect sizes by study and intervention characteristics are available in the full report.
The 2 trials of antidepressants to prevent perinatal depression assessed the effects of nortriptyline (n = 58)33 and sertraline (n = 22).34 Both trials were conducted in the United States. The trial of nortriptyline found that the drug offered no preventive benefits compared with placebo (Figure 4).33 Neither the rates of recurrence between participants taking nortriptyline and those taking placebo (23% vs 24%; between-group statistics not reported; P > .99), nor the time to postpartum recurrence (detailed findings not reported; exact log-rank ≤0.00; P = .83) differed between the 2 groups. The trial of sertraline found that a smaller percentage of participants taking sertraline had a depression recurrence compared with those taking placebo (7% vs 50%; difference in recurrence rates, 0.43 [95% exact CI, −0.01 to 0.84]; P = .04) at 20 weeks postpartum.34 Further, the time to recurrence was faster in those receiving placebo (hazard ratio, 0.11 [95% exact CI, 0.02 to 1.02]; exact Wilcoxon-Gehan P = .02). In these trials, women with a history of postpartum depression in the previous 5 years were randomized to receive either an antidepressant (nortriptyline [75 mg/d]33 or sertraline [50 mg/d]34) or placebo for 17 weeks, starting as soon as possible after birth, followed by a 3-week tapering phase. Neither trial of antidepressants reported other maternal health outcomes or child outcomes.
Key Question 2. What harms are associated with interventions to prevent perinatal depression in pregnant or postpartum women?
None of the nonpharmaceutical studies reported any global harms outcomes for either mothers or infants. Across all of these studies, none of the outcomes reported showed any pattern of increased risk of harms, based on group means.
Both antidepressant trials systematically collected adverse event information. In the nortriptyline trial (n = 58),33 the authors stated that participants tolerated nortriptyline well, reporting no differences in withdrawals attributable to adverse effects, with only 1 person withdrawing from each group. They reported only the number of events for 1 of the 11 adverse effects collected; constipation differed between groups (78% of the women taking nortriptyline vs 22% taking placebo). In the sertraline trial (n = 22),34 participants receiving sertraline were more likely than those receiving placebo to report dizziness (57% vs 13%, P = .05) and drowsiness (100% vs 50%, P = .02) but did not differ in rates of other adverse events (but data were not shown). Three women stopped taking sertraline because of adverse effects (21%), compared with none in the control group; however, this difference was not statistically significant. One participant taking nortriptyline and 1 taking sertraline converted to mania or hypomania, while no women taking a placebo did so; this difference was not statistically significant for either agent, although the studies were not powered for this outcome. There were no additional studies that addressed harms of sertraline in postpartum women. Additional harms studies of nortriptyline were not searched for because efficacy was not demonstrated for this treatment in the included studies.
No harms were associated with omega-3 fatty acids, although reports of adverse events were collected spontaneously rather than systematically through a validated instrument and adherence was by self-report. No significant differences in the proportion of participants reporting gastrointestinal adverse effects or adherence with the recommended intervention were reported.
The summary of evidence is reported in Table 3. This review found that counseling-based interventions, in particular depression-focused CBT and IPT, may be effective in preventing perinatal depression. This evidence was primarily limited to women at increased risk for perinatal depression, such as having current depressive symptoms, a history of depression, low socioeconomic status, and lack of support. The pooled RR of 0.61 for perinatal depression at up to 6 months postpartum for counseling interventions corresponds to a number needed to treat of 13.5 (95% CI, 9.9 to 23.9), assuming a 19% baseline risk of developing perinatal depression. Three different health system–level interventions were also effective in health care settings outside the United States, suggesting that similar interventions developed in US-based health care systems may have the potential to be effective and were not limited to women at increased risk of perinatal depression. In all 3 cases, usual care included home visitation, which may be a valuable intervention, suggesting the potential for even greater benefit compared with usual care in the United States. Additionally, other intervention approaches, such as physical activity or educational approaches, reported some positive findings but lacked robust evidence bases.
Only 2 studies of prophylactic use of antidepressants were identified, showing mixed results. Antidepressants can be an important tool for treatment of depression but have been associated with a number of rare but serious adverse events, including suicidality (in young adults), hyponatremia, seizures, gastrointestinal tract bleeding, and serotonin syndrome.67,68 The decision to use antidepressants in pregnant persons is complicated and is addressed by only a limited amount of evidence. The use of second-generation antidepressants during pregnancy has been associated with a small but increased risk of a number of serious pregnancy and neonatal outcomes.69 However, the clinical significance of these increased risks is unclear70 and most guidelines recommend antidepressants for severe depression, with a preference for sertraline,71 especially for those breastfeeding.72
The findings of this review are similar to those reported in other similarly scoped reviews. A 2013 review on psychosocial and psychological interventions to prevent perinatal depression found that women who received a psychosocial intervention were 22% less likely to develop perinatal depression, compared with usual care (pooled RR, 0.78 [95% CI, 0.66 to 0.93]; 20 studies [n = 14 727]).73 Another review of interventions designed to prevent perinatal depression found a pooled odds ratio of 0.67 for depressive episodes by 6 months postpartum, after excluding outliers (95% CI, 0.52 to 0.85; 26 studies; I2 = 46%).74 That review found no study or intervention characteristic that showed a statistically significant association with effect size.
An ideal method for determining which interventions would benefit persons with varying risk profiles is lacking, but the most common approach in the included trials was to include women with a history of depression or current depressive symptoms as measured by instruments such as the EPDS or CES-D. One study investigating the accuracy of the EPDS to predict future perinatal depression found that a cutoff of 9 or higher at 3 to 5 days postpartum had 82% sensitivity, 95% specificity, and a 43% positive predictive value for a diagnosis of major or minor depression at 8 weeks postpartum.75 The literature on predicting future perinatal depression includes a variety of patient- and clinician-administered tools, but results have been modest in many cases and need to be replicated.
There were a number of limitations in the studies underlying this review. There were relatively few good-quality trials, and approximately one-third of the trials within the scope of this review were excluded because of their poor quality. Many of these appeared to be pilot studies not designed to provide data on effectiveness of the intervention or that used intervention approaches that proved infeasible or ineffective and so were abandoned in the form studied. Some of these studies, however, could have provided useful information had they been conducted and reported in such a way that they met USPSTF quality standards.
The health system–level interventions in this review had limited applicability to health systems in the United States, especially since they involved enhancing home-visiting services, which are not routinely available in the United States. However, some home-visiting services are available in the United States, and these interventions also included other elements that would be relevant to US-based settings. Interventions designed for implementation in health care systems could involve clinician training, electronic medical records–based tools, and facilitated access to behavioral health specialists embedded in the primary care settings.
Another limitation of the evidence was the small number of trials examining several potentially valuable interventions, such as physical activity, infant sleep education, in-hospital perinatal depression education with follow-up, and peer counseling. In addition, larger-scale effectiveness trials of CBT and IPT approaches are needed to explore the degree to which these interventions can be scaled up, as well as their applicability to lower-risk, more general primary care populations. More research is also needed on the use of antidepressants and dietary supplements in the role of preventing perinatal depression.
Another important deficit in the literature is a lack of good information on the best way to identify individuals at risk for perinatal depression. Measures of depression symptoms, such as the EPDS, likely provide the most direct association with future perinatal depression. However, evidence is lacking on whether and how to incorporate other risk factors, as well as on who is most likely to benefit from preventive interventions and how those individuals are best identified.
The review has several limitations. First, it was challenging to determine whether prevention of perinatal depression was truly an a priori aim; it is possible that some studies were missed that had depression as a specific aim, and some studies may have been included that added the depression prevention aim post hoc after determining that their intervention was effective in preventing depression. Second, both the overall body of evidence and the counseling intervention trials had statistically significant small-studies effects. Smaller trials also tended to use interventions that more directly addressed depression and to offer more intensive interventions, so the small-studies effect may be influenced by these and other study characteristics. However, it could not be determined to what extent the effect might be biasing results and overestimating the effect sizes.
Counseling-based interventions can be effective in preventing perinatal depression, although most evidence was limited to women at increased risk for perinatal depression. A variety of other intervention approaches provided some evidence of effectiveness but lacked a robust evidence base and need further research.
Corresponding Author: Elizabeth O’Connor, PhD, Kaiser Permanente Research Affiliates Evidence-based Practice Center, The Center for Health Research, Kaiser Permanente Northwest, 3800 N Interstate Ave, Portland, OR 97227 (Elizabeth.OConnor@kpchr.org).
Accepted for Publication: December 6, 2018.
Author Contributions: Dr O’Connor had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: O’Connor, Senger, Coppola, Gaynes.
Acquisition, analysis, or interpretation of data: O’Connor, Henninger, Coppola, Gaynes.
Drafting of the manuscript: O’Connor, Senger, Coppola, Gaynes.
Critical revision of the manuscript for important intellectual content: Henninger, Gaynes.
Statistical analysis: O’Connor, Gaynes.
Obtained funding: O’Connor, Gaynes.
Administrative, technical, or material support: Senger, Henninger, Coppola.
Conflict of Interest Disclosures: Dr Henninger reported receiving grants from Pfizer IGLC outside the submitted work. Dr Gaynes reported receiving personal fees from LivaNova and personal fees from Johnson and Johnson outside the submitted work. No other authors reported disclosures.
Funding/Support: This research was funded under contract number HHSA-290-2015-00017-I, Task Order 3, from the Agency for Healthcare Research and Quality (AHRQ), US Department of Health and Human Services, under a contract to support the USPSTF.
Role of the Funder/Sponsor: Investigators worked with USPSTF members and AHRQ staff to develop the scope, analytic framework, and key questions for this review. AHRQ had no role in study selection, quality assessment, or synthesis. AHRQ staff provided project oversight, reviewed the report to ensure that the analysis met methodological standards, and distributed the draft for peer review. Otherwise, AHRQ had no role in the conduct of the study; collection, management, analysis, and interpretation of the data; and preparation, review, or approval of the manuscript findings. The opinions expressed in this document are those of the authors and do not reflect the official position of AHRQ or the US Department of Health and Human Services.
Additional Contributions: We gratefully acknowledge the following individuals for their contributions to this project: the AHRQ staff; the US Preventive Services Task Force; Brittany U. Carter, DSc, MPH (Kaiser Permanente Care Management Institute), for her assistance with scoping; Meghan Soulsby Weyrich, MPH (University of California, Davis), for her assistance with data abstraction; and Smyth Lai, MLS, and Katherine Essick, BS, for technical and editorial assistance at the Kaiser Permanente Center for Health Research. USPSTF members, peer reviewers, and federal partner reviewers did not receive financial compensation for their contributions.
Additional Information: A draft version of this evidence report underwent external peer review from 3 content experts (Sherryl Goodman, PhD, Emory University; Michael O’Hara, PhD, University of Iowa; Barbara Yawn, MD, University of Minnesota) and 1 federal partner: the Centers for Disease Control and Prevention. Comments from reviewers were presented to the USPSTF during its deliberation of the evidence and were considered in preparing the final evidence review.
Editorial Disclaimer: This systematic review is presented as a document in support of the accompanying USPSTF Recommendation Statement. It did not undergo additional peer review after submission to JAMA.