Effectiveness of nonpharmacological interventions for reducing postpartum fatigue: a meta-analysis

The protocol has been registered in the International Prospective Register of Systematic Reviews (PROSPERO [RRID: SCR_019061]); registration number CRD42021234869.

This meta-analysis has been reported following the Preferred Reporting Items for Systematic Reviews and Meta-Analysis statement (PRISMA [RRID: SCR_018721]). The PRISMA checklist can be seen in Additional file 1.

The focus of the review was to examine nonpharmacological interventions for fatigue reduction in postpartum women. Seven electronic databases, including the Cochrane Library, PubMed, Embase, Web of Science, PsycINFO, CINAHL and ProQuest, were searched for articles published from inception until June 2021 without time restrictions. An additional search of OpenGrey was conducted for grey literature. First, we identified main concepts including “postpartum”, “fatigue” and “clinical trial” using the PICOS (Population, Intervention, Comparison, Outcome and Study design) search tool. Then for each concept, a combination of Medical Subject Heading, Emtree terms and free terms [e.g., (postnatal OR postpartum OR delivery OR childbirth OR birth OR parturition OR labour OR pregnancy) AND (fatigue OR mental fatigue OR lassitude OR exhaust* OR tiredness OR tired*) AND (clinical trial OR randomized controlled trial OR controlled clinical trial OR cohort)] was used. The final search strategies applied are shown in Additional file 2. Moreover, a manual search of reference lists was performed to thoroughly identify relevant studies that were missed. Two authors (JLQ and SWS) performed the search process independently.

The inclusion criteria of this review were identified based on PICOS: (1) Participants: women aged 18 years or over who had a healthy pregnancy without postnatal complications; (2) Intervention: nonpharmacological interventions conducted during 0 ~ 78 weeks postpartum with the primary or secondary aim of decreasing fatigue symptoms. The intervention setting, frequency, timing and duration were not limited; (3) Comparison: usual care, placebo, waitlist or no interventions; (4) Outcomes: the primary efficacy outcome was postpartum fatigue estimated as the rate or mean severity of fatigue; The second efficacy outcome was psychological variables, such as depression, anxiety or stress; (5) Study design: Clinical trials adopting randomized controlled trials (RCTs), quasi-experimental, before-and-after or prospective cohort study designs; and (6) Original articles published in English. The exclusion criteria were as follows: (1) duplicated publications (only the one with the most participants was included); and (2) studies without sufficient data to be extracted.

After removing duplicates, two authors (JLQ and SWS) independently screened the studies according to the inclusion criteria in 2 steps: 1) title and abstract screening and 2) full-text screening. A third author (LL) was consulted to reach a consensus when there was any uncertainty about the inclusion of an article.

A standardized data extraction sheet was used to extract important information from the included studies. The extracted data included first author, publication year, country, study design, population, sample size (trial/control), intervention details (e.g., type, start time, duration and frequency), control, evaluation time points, assessment tools and outcomes. Two authors (JLQ and SWS) independently extracted the data, and any inconsistencies were resolved by a third author (LL).

For RCTs, the Cochrane risk-of-bias tool was used for the quality assessment [16]. Studies were assessed based on seven criteria (random sequence generation, allocation concealment, blinding of participants and researchers, blinding of outcome assessor, incomplete outcomes data, selective reporting and other bias). Bias was evaluated as a judgement (high or low or unclear), and then each included study was rated as having a high, moderate or low risk of bias. The Risk of Bias in Nonrandomized Studies of Interventions was used to assess the risk of bias of non-RCTs [17]. Seven domains are evaluated in this tool (confounding, selection of participants into the study, classification of intervention, deviation from the intended interventions, missing data, measurements of outcomes and selection of the reported results). An inter-rater reliability score using cohen’s kappa [18] for risk of bias was used. Cohen’s kappa coefficient is a statistical measure of inter-rater agreement (< 0: poor agreement; 0.0–0.20: slight agreement; 0.21–0.40: fair agreement; 0.41–0.60: moderate agreement; 0.61–0.80: substantial agreement; 0.81–1.0: almost perfect agreement). Two authors (JLQ and SWS) assessed the risk of bias and evidence quality separately.

We performed a meta-analysis utilizing Review Manager 5.3 (RRID: SCR_003581). For continuous outcomes, we calculated mean differences (MDs) and 95% confidence intervals (CIs) if the outcomes were measured using the same tool. We used standardized mean differences (SMDs) and 95% CIs to combine studies when the same outcome was measured by adopting different tools [19]. For the SMD, ≤0.20, =0.50, and ≥ 0.80 are designated as small, moderate and large effect sizes, respectively. The heterogeneity among the analysed trials was examined by standard chi-square and I-square statistics. If the P value was > 0.1 or I² < 50%, it indicated that there was no observed heterogeneity, and the researchers employed a fixed-effects model to combine the study results. If not, a random-effects model analysis was used [20]. Publication bias was examined using Egger’s test because it is not appropriate to use funnel plots when analysing fewer than 10 studies [21].

The study characteristics are summarized in Table 1. Studies were published between 2003 ~ 2020. The 17 included articles were conducted in different countries: China (n = 5), Iran (n = 4), the USA (n = 2), Canada (n = 2), Turkey (n = 2), Australia (n = 1) and the UK (n = 1). A total of 2142 participants were included in this review. The sample size ranged from 27 to 356 participants in each trial. Eleven of the 17 included studies employed a randomized controlled trial design.

Postpartum interventions were grouped into four broad categories: exercise [12, 22‐25] (n = 5), psychoeducational intervention [13, 15, 26‐30] (n = 7), drinking tea [31, 32] (n = 2) and physical therapy [10, 11, 33] (n = 3). The intervention duration ranged from 1 day to 3 months. The intervention start time varied from immediately postpartum care to 1 year after delivery. The corresponding finish time of intervention ranged from parturition day to 1 year and 3 months postpartum. All included studies described the baseline assessments, with the scores for the intervention and control groups comparable at baseline. Additionally, the studies reported assessment scores immediately after the intervention and follow-up assessment scores at 1 month postintervention [31, 32], 6 weeks [15, 27, 29] and 9 weeks postintervention [15], 2 months postintervention [22], 3 months postintervention [23, 25, 27, 29] and 6 months [25] postintervention. These trials adopted the MFI-20, MFSC, PFS, VAS-F, FAS and FSC to evaluate the level of postpartum fatigue; the EDPS, DASS-21, CES-D, VAS for depression; and the PANAS and PSS were used to assess psychological variables such as anxiety and stress.

The quality of the study designs was low to moderate overall. Several methodological limitations were observed in the critical appraisal. Quality assessments of the 11 included studies using an RCT design, with a risk of bias graph and risk of bias summary, are described in Fig. 2 and Fig. 3. All of the studies reported using randomization; however, two articles did not provide detailed information about the randomization method. Six studies reported sufficient details about allocation concealment. Blinding of the participants and researchers who delivered the interventions was not feasible because the interventions were easy to identify. Therefore, all of the studies are at a high risk of performance bias. Regarding detection bias, only four studies provided sufficient explanations. Except for one study, all RCT studies gave clear information about the incomplete outcome data. No reporting bias was found in the included RCTs. The results of the quality appraisal of the six nonrandomized studies are displayed in Table 2. Bias due to confounding factors, selection of participants into the study and classification of interventions were low for all included nonrandomized studies. Three articles were at moderate risk of bias due to deviations from the intended intervention. All of the studies were reasonably reported and addressed missing data. In regard to bias in the measurement of outcomes, considering that the interventions were not blinded to the participants, all studies were judged as at moderate risk of bias. Bias in selection of the reported results was not observed. According to the assessment results, the calculated cohen’s kappa was 0.74, which suggested that there was a substantial agreement between two raters.

We did not conduct a meta-analysis on the physical therapy interventions. Because there are only three studies and the types of intervention are completely different in each study. For the purposes of meta-analysis, three types of interventions were undertaken: exercise, psychoeducational interventions and drinking tea. Only one trial [28] reported the postpartum fatigue rate. Three trials [15, 23, 25] without sufficient original data for meta-analysis were thus not included. Ultimately, ten studies that reported the mean fatigue scores and standard deviations were included in the meta-analysis. For the secondary outcomes, since more than two studies reported depression, a statistical combination was performed regarding this psychological outcome.

Three studies examined the effects of exercise on fatigue among postnatal women [12, 22, 24]. Pilates exercise, aerobic exercise, yoga were chosen in this kind of intervention. The intervention start time varied from 72 h after delivery to 6 weeks postpartum, and the intervention duration ranged from 3 to 12 weeks.

Figure 4 shows the effect sizes of the exercise intervention in terms of fatigue and depression. For the primary outcome, three trials showed a postassessment of fatigue. The heterogeneity was I² = 95% (P < 0.0001), so a random-effects model was applied. The pooled SMD was − 1.74 (95% CI = -2.61 to − 0.88, Z = 3.94, P < 0.0001), demonstrating that the exercise intervention had a better effect on decreasing fatigue symptoms. For the secondary outcome, depression data were presented in two trials. No significant effects on reducing depression were indicated. Significant heterogeneity was not detected (I² = 46%, P = 0.17), so a fixed-effect model was used. The pooled SMD was − 0.05 (95% CI = -0.33 to 0.24, Z = 0.31, P = 0.75).

Five studies used psychoeducational intervention [13, 26, 27, 29, 30]. Psychoeducational intervention was most frequently conducted via home visits, leaflet/booklet and home calls in order to give women instructions of relaxation, sleeping exercises, nutrition and so on. The intervention start time varied from immediately postpartum care to 6 month postpartum, and duration ranged from 3 to 12 weeks.

Figure 5 displays the effect sizes of psychoeducational interventions according to fatigue and depression at postintervention and at the 8-week follow-up. Five studies displayed the results of psychoeducational interventions on fatigue. We found no significant difference between the two groups regarding fatigue. The heterogeneity was high (I² = 94%, P < 0.0001), so a random-effects model was applied. The pooled SMD was − 0.41 (95% CI = -0.92 to 0.11, Z = 1.55, P = 0.12). Similarly, no significant effects on depression were noted in two trials presenting the depression data. We used a fixed-effects model because the heterogeneity was insignificant (I² = 41%, P = 0.18). The pooled SMD was − 0.04 (95% CI = -0.23 to 0.15, Z = 0.39, P = 0.69).

Eight-week follow-up fatigue and depression assessments were presented in two trials. No significant effects on relieving fatigue and depression were indicated. For fatigue, we used a random-effects model because the heterogeneity was I² = 58% (P = 0.05). The pooled SMD was − 0.08 (95% CI = -0.33 to 0.18, Z = 0.60, P = 0.55). In regard to 8-week follow-up depression scores, compared with the control groups, the participants in the intervention group did not experience significant changes in depression. There was no substantial evidence of high heterogeneity (I² = 0%, P = 0.61); therefore, a fixed-effects model was used. The pooled SMD was − 0.01 (95% CI = -0.21 to 0.18, Z = 0.12, P = 0.90).

Two studies examined the effectiveness of drinking tea among participants who had poor sleep quality [31, 32]. Participants in the intervention group were required to drink one cup of chamomile tea every day for 2 weeks.

The effectiveness of drinking tea on fatigue and depression at postintervention and at the 2-week follow-up were presented in two trials (Fig. 6). At postintervention, significant differences were noted between the intervention and control groups in regard to fatigue and depression. No significant heterogeneity was found in terms of fatigue and depression (I² = 0%, P = 0.44; I² = 0%, P = 0.74). Therefore, a fixed-effects model was used. The pooled MD was − 3.12 (95% CI = -5.44 to − 0.80, Z = 2.64, P = 0.008) and − 2.89 (95% CI = -4.30 to − 1.49, Z = 4.04, P < 0.0001).

At the 2-week follow-up, the results suggested that there were no significant differences between the intervention and control groups in regard to fatigue. There was no evidence of significant heterogeneity (I² = 0%, P = 0.59). The pooled MD was − 1.38 (95% CI = -3.58 to 0.81, Z = 1.24, P = 0.22). The results showed that tea consumption had significant effects on depression at the 2-week follow-up. We used a fixed-effects model because the heterogeneity was I² = 0% (P = 0.77). The pooled MD was − 2.05 (95% CI = -3.47 to − 0.62, Z = 2.82, P = 0.005).

Three studies [10, 11, 33] involving 492 participants examined the effect of physical therapy interventions on postpartum fatigue. Mother-infant skin-to-skin contact, taking warm showers and breathing lavender oil aroma were approaches used as physical therapies. The intervention started within 1 to 2 days postpartum. The intervention frequency ranged from 1 to 3 times and each intervention time varied from 10 to 30 min. Funda et al. [10] investigated the efficacy of mother-infant skin-to-skin contact on postpartum fatigue. The infant was left in the prone position on the mother’s chest for 30 min in the first hour after the delivery. However, no significant differences between groups were found on postpartum fatigue. Hsieh et al. [33] instructed healthy postpartum women to take warm showers for approximately 20 min on the second day after delivery. The results indicated that warm showers significantly decreased postpartum fatigue in the intervention group compared to the control group. Vaziri et al. [11] tested breathing lavender oil aroma on managing postpartum fatigue among primiparous women. A significantly positive effect of this intervention on fatigue, distress, positive and negative moods was observed between groups.

The potential publication bias was assessed by performing Egger’s linear regression test of exercise and psychoeducational interventions. For the primary outcome fatigue at postintervention, Egger’s test suggested no indication of publication bias of exercise (t = − 1.84, p = 0.317) and psychoeducational intervention (t = − 2.28, p = 0.052).

This review identified a total of seventeen published clinical trials examining nonpharmacological interventions (including exercise, psychoeducational interventions, drinking tea and physical therapy) on postpartum fatigue in postnatal women. Ten studies involving 1194 participants were included in the meta-analysis to examine the respective effectiveness of each type of intervention.

The findings of the meta-analysis suggest that exercise intervention significantly reduces symptoms of postpartum fatigue (SMD = − 1.74, 95% CI = -2.61 to − 0.88, Z = 3.94, P < 0.0001). This is consistent with the findings of previous studies [34, 35], which clarified the necessities and benefits of physical exercise for the overall health of postpartum women. A meta-analysis [36] indicated that postpartum exercise was beneficial for decreasing postpartum fatigue. However, postpartum women are still at high risk for physical inactivity and rarely understand how to engage in postpartum exercise [35]. The earliest start time of exercise intervention was 72 h after delivery among included studies and the exercise intervention usually continue for at least 3 weeks. The American College of Obstetricians and Gynaecologists (ACOG) suggested that physical activity, including performing muscle-toning exercises, could be restarted 6 weeks after childbirth if the delivery was uncomplicated [37]. Daily 20 ~ 30 min of regular moderate-to-intense exercise is recommended [38]. A previous study reported that supervised postpartum exercise lasting more than 8 weeks is suggested for reducing postpartum fatigue [14]. Therefore, clinical staff should give evidence-based instructions involving detailed start time, duration and frequency of postnatal exercises to help reduce women’s postpartum fatigue and facilitate their postpartum rehabilitation.

For depression, the results showed that there were no significant differences between the two groups (SMD = -0.05, 95% CI = -0.33 to 0.24, Z = 0.31, P = 0.75). This is inconsistent with the findings of a previous meta-analysis [39], which reported that exercise reduced symptoms of depression (SMD = -0.81, 95% CI = -1.53 to − 0.10) of mothers who had been diagnosed with depression. The insignificant results may be explained by the fact that the two studies [12, 24] in our review did not include participants with obvious depressed symptoms at baseline. Therefore, a significant reduction in depression was difficult to observe after the exercise intervention. It is worth noting that in Yang et al. [12]‘s study, women reported feelings of mood relaxation and pressure relief after the intervention, which suggested that exercise may have potential efficacy for women’s mental health. More research conducting exercise interventions among puerperae is needed to confirm the true effect of exercise on mothers’ psychological well-being.

Five studies performing psychoeducational interventions showed no obvious differences in postpartum fatigue and depression between the two groups at either the postintervention or 8-week follow-up. It’s worth noting that studies with different psychoeducational aims and strategies were combined in the meta-analysis, which may be the reason for significant heterogeneity. Psychoeducational interventions were offered to women immediately postpartum and were aimed at resolving various problems associated with postnatal health. This type of intervention was mostly delivered via home visits, leaflet/booklets and home calls. It is difficult to guarantee the good compliance of participants due to a lack of sufficient supervision and guidance. Therefore, it may be difficult to perform psychoeducational interventions effectively. With the development of rapid electronic technology, people’s requests for health services have also facilitated instant communication and promoted efficiency in the transmission of information [40]. A previous study reported that web-based interventions had better effects on improving postnatal depression than home-based postnatal psychoeducational interventions [41] and suggested that web-based interventions should be introduced to mothers for better postnatal care. Hence, psychoeducational interventions could be combined with internet technology [42] or smartphones [43] in the future to better manage the intervention implementation, observe the compliance of the participants and improve the intervention efficacy.

In terms of drinking tea, the results showed that there were significant differences in postpartum fatigue between the two groups at postintervention (MD = − 3.12, 95% CI = -5.44 to − 0.80, Z = 2.64, P = 0.008), but no significant differences were found at the 2-week follow-up. The positive effects of drinking tea on relieving depression were obvious at the postintervention (MD = − 2.89, 95% CI = -4.30 to − 1.49, Z = 4.04, P < 0.0001) and 2-week follow-up (MD = − 2.05, 95% CI = -3.74 to − 0.62, Z = 2.89, P = 0.005). Postnatal women were required to smell the aroma before drinking the tea. Aromatherapy has been used for pain and anxiety relief, relaxation, and creating a pleasant feeling in mothers [44‐46], which could help to relieve fatigue and depressive emotions. Women in the intervention group reported the benefits of drinking chamomile tea to be facilitating emotional stability and relaxation and having an aromatic fragrance, which could calm restlessness, facilitate the postnatal paternity relationship, and alleviate postpartum fatigue [31]. However, the effects of drinking tea on fatigue was immediate but not long term in our findings. A possible reason for the short-lived effect is that the employed women had completed their maternity leave and returned to work at 2-week follow-up (10 weeks postpartum) [31, 32]. Therefore, additional measurement points with shorter intervals may be essential to confirm the actual duration of the therapeutic effects of drinking tea. In addition, how behaviour change wasn’t monitored in these two studies [31, 32]. It may lead to decreased treatment adherence, thereby influencing the long-term efficacy of the intervention. Future study should strengthen the supervision of intervention implementation so as to improve the efficacy. The effectiveness of this intervention should be considered cautiously due to lack of blinding in allocation concealment, intervention implementation and outcome assessment. High-quality RCTs further investigating the effect of drinking tea are needed.

Interventions of physical therapy were started within 1 to 2 days after childbirth under the guidance of medical staff and completed in 1 day. Physical therapies have the advantages of a short intervention time and good controllability. Warm showers, as a comfort measure, are closely associated with increased relaxation and tension reduction [47]. It is a safe and effective measure for healthy, labouring women who are experiencing physical and psychological issues [47]. The benefits of lavender oil on postpartum fatigue were reported in another RCT conducted among pregnant women [48]. Participants in either the lavender and footbath or lavender alone group showed that fatigue was improved significantly at 6 weeks postpartum. Although the efficacy of mother-infant skin-to-skin on fatigue was not observed in Funda et al.’s study, a recently published meta-analysis demonstrated that mother-infant skin-to-skin was a cost-effective, simple and feasible approach for postpartum depression [49]. Therefore, more studies should be conducted to understand the effectiveness of a certain physical therapy. Among the three included studies, only Vaziri et al.’s study examined the efficacy of lavender oil aroma for psychological outcomes, including distress and mood, and positive effects were observed. Considering that physical therapies are relatively safe, physical therapies such as footbaths [48], reflexology [50], warm showers [33] and lavender oil [11], which show potential effects on fatigue reduction and mental health improvement, could be used in combination to enhance the intervention efficacy.

To the best of our knowledge, this is the first meta-analysis on nonpharmacological interventions for reducing postpartum fatigue. This review was rigorous and based on the Preferred Reporting Items for Systematic Reviews and Meta-Analysis statements as well as a prospective registered protocol. Another strength was that randomized controlled trials and controlled clinical trials were included in this review, which provides good standards for evidence-based research.

There were some limitations of this meta-analysis. A limitation of the review was that non-English electronic databases were not searched, which may cause language bias [51]. The number of included studies for each type of intervention and sample sizes were small. Our results showed high heterogeneity, and the reason may be that interventions were divided into broad groups and we have pooled very different types of studies into the same category. It may have influenced the pooled results, despite using a random-effects model. Further, given the conditions of the control groups are not time-matched, hence the mechanism of change on postpartum fatigue and depression is unclear. Another limitation lies in the risk of bias within the included studies. The potential biases may have influenced the reported effect estimates; therefore, caution is required when interpreting the findings of our study.

Only a part of included studies examined the effectiveness of fatigue-related interventions on depression. Given that there is a strong correlation between fatigue and depression among women in the first 2 years after child-birth [3], depression and other psychological outcomes could be included as outcomes in future research to see if fatigue-related interventions could both enhance postpartum women’s physical and psychological health. In regard to different nonpharmacological interventions for reducing postpartum fatigue, future research should work out effective and specific exercise programs for postpartum women with long term follow-up. Web-based or smartphone-based postnatal psychoeducational interventions need to be studied. Considering risk of bias within the included studies of drinking tea, risk biases existing in allocation of participants, blinding during intervention implementation and data analysis should be carefully considered. More rigorous RCTs of drinking tea should be conducted. The effectiveness of a certain physical therapy should be further investigated.