Effectiveness of nonpharmacologic interventions on pregnancy-related low back pain: A network meta-analysis of randomized controlled trials

Abstract

Background:

Pregnancy-related low back pain (PLBP) is a prevalent clinical condition occurring antenatally and postnatally. Given limited evidence on conservative management, this study systematically assessed the efficacy of non-pharmacological interventions for pain relief and physical function improvement in females with PLBP.

Methods:

The Preferred Reporting Items for Systematic Reviews and Meta-analyses guidelines were followed. A systematic search was performed across 12 electronic databases from inception to March 30, 2024. Eligibility screening was performed according to predefined criteria, followed by the quality assessment utilizing the Cochrane Risk of Bias Tool. Data analysis was conducted using Stata 18.0. Network meta-analysis applied the node-cut method for the consistency test, and used the surface under the cumulative ranking curve to reflect the likelihood of each non-pharmacological intervention being the best intervention.

Results:

Thirty-six randomized controlled trials were examined (N = 4511 participants). The surface under the cumulative ranking curve identified music-relaxation as the highest-ranked therapy for pain reduction (97.1%), and manipulation-acupuncture as the highest-ranked therapy for physical function indicators (78.3%).

Conclusion:

Limited evidence indicated that music-relaxation therapy may be the most effective strategy for alleviating pain, while manipulation-acupuncture therapy may be optimal for enhancing physical function. Integrated non-pharmacological interventions demonstrated greater effectiveness compared to monotherapy in improving PLBP.

1. Introduction

Pregnancy-related low back pain (PLBP) is a condition frequently encountered during pregnancy and the postpartum phase, characterized by pain and dysfunction that occurs in the region of the twelve ribs and gluteal folds. The condition may present in 2 different patterns: low back pain and pregnancy-related pelvic girdle pain.^[1] In recent years, the prevalence of PLBP has escalated, with rates ranging from 20% to 90%. Approximately one-third of pregnant women endure severe PLBP, and over one-quarter experience pain that persists for 2 to 3 years postoperatively.^[2]

The symptoms of PLBP exacerbate with the progression of pregnancy, significantly restricting mobility and elevating the likelihood of problems and negative outcomes, profoundly affecting the long-term physical and mental health and quality of life of women.^[3] Pregnant women experiencing the early start of PLBP are at increased risk of developing chronic low back pain.^[4]

The prevalence of PLBP and its significantly detrimental effects on the health of pregnant women have garnered considerable attention globally, and researchers are diligently attempting to identify effective preventive and therapeutic strategies. Owing to the unique attributes of the pregnant population, several non-pharmacological therapies have been endorsed as the recommended initial treatment strategy in recent years.^[5] Non-pharmacological interventions are diverse, with varying mechanisms of action in alleviating PLBP. Acupuncture (AC) activates nerve endings via AC or trigger points (Ashi points) that regulate the flow of energy (qi) and blood circulation and balance the yin and yang in the body, thereby alleviating pain.^[6] Exercise (EX) can strengthen and stabilize the core muscles of the lower back and alleviate strain on the lumbar vertebrae.^[7] Most studies have compared single non-pharmacological interventions with standard care in enhancing PLBP.^[8–10] However, the absence of direct or indirect comparisons between non-pharmacological interventions limits comprehensive effectiveness assessments, hindering definitive conclusions regarding optimal strategies for PLBP management.

Network meta-analysis (NMA) is a new method that enables direct and indirect comparisons between 2 or more interventions, ranking them according to their effectiveness in improving outcome indicators to select the best intervention. Therefore, the main objective of this study was to utilize NMA to assess various non-pharmacological approaches aimed at alleviating pain and improving functional outcomes related to PLBP, by synthesizing and scrutinizing the available evidence. The objective was to provide a more scientific and comprehensive health management strategy for clinical care and reduce the incidence of PLBP.

2. Methods

2.1. Eligibility criteria

This study was designed according to the Preferred Reporting Items for Systematic Reviews and Meta-Analysis guidelines and was pre-registered at PROSPERO (CRD42024519877).

This study included randomized controlled trials (RCTs) enrolling pregnant women meeting the diagnostic criteria for PLBP. Eligible interventions included non-pharmacological therapies (e.g., EX, AC). The control groups encompassed routine care, placebo, no intervention, or alternative non-pharmacological modalities. The primary outcomes were pain intensity and functional status measures.

Studies in languages other than Chinese or English, those with no full text or complete data, non-RCTs, low-quality studies, and duplicate publications were excluded.

2.2. Search strategy

Comprehensive literature searches from inception to March 30, 2024, were conducted using 12 electronic databases: PubMed, Cochrane Library, Scopus, Embase, CINAHL, Web of Science, CNKI, CBM, VIP, Wanfang Data, Clinical trials, and the China Clinical Trials Registry, focusing on RCTs evaluating non-pharmacological interventions for improving PLBP (Fig. S1, Supplemental Digital Content, https://links.lww.com/MD/P702).

2.3. Study selection and data extraction

All retrieved records were input into (EndNote 20, Yanji, Jilin Province, China) 2 researchers screened and extracted the data based on the inclusion and exclusion criteria and cross-checked the screening results. Disputes were addressed through collaboration with an external researcher. The data extraction parameters encompassed the authors, publication date, country/region, sample size, age, intervention modality, duration, and outcome indicators.

2.4. Quality assessments

Two scholars individually assessed the quality of the research utilizing the RCT bias risk evaluation tool as recommended by the Cochrane Handbook 5.1.0.^[11]

2.5. Statistical analyses

2.5.1. Methods of analysis

The risk of bias was assessed using RevMan 5.4, and NMA was performed using (Stata 18.0, Yanji, Jilin Province, China). This study utilized continuous variables and multiple measurement tools, with network estimates from all outcome variables expressed as standardized mean differences (SMD) and 95% confidence intervals (CI). Evidence network diagrams were generated for each outcome, where nodes represented interventions and the connecting line thickness corresponded to the number of direct comparisons between intervention modalities. The surface under the cumulative ranking curve (SUCRA) was calculated to rank intervention efficacy. Publication bias was evaluated by visually inspecting the funnel plots.

2.5.2. Assessment of inconsistency

We applied local and global network analyses to ascertain the extent of heterogeneity accounted for by inconsistency. The overall data was subjected to an incongruity test, indicating minimal heterogeneity if the P > .05. The inconsistency test is performed via the node splitting method when a closed loop exists in the network diagram. If the disparity between the direct and indirect comparison outcomes is not statistically significant (P > .05), the analysis employs the consistency model; alternatively, the inconsistency model is utilized when the disparity is considerable.

3. Results

3.1. Study selection

The initial screening obtained 906 research, and 7 research by the Snowballing method, totaling 913 documents; 431 duplicate studies were rejected, 334 studies were eliminated based on the titles and abstracts, and 112 were discarded following full-text assessment, resulting in the inclusion of 36 RCTs^{[2,4,12–45]} (Fig. 1).

Figure 1.

Literature screening flow chart. RCTs = randomized controlled trials.

3.2. Study characteristics

The 36 studies were published from 2008 to 2023, comprising 6 publications in Chinese and 30 in English, including 4511 patients, 2164 in the intervention group and 2347 in the control group. Furthermore, 17 interventions were conducted, varying in duration from 1 to 12 weeks (Table 1).

Table 1.

Characteristics of studies included in the meta-analysis.

Study	Country	Sample size		Intervention		Duration	Outcome	Tool
		Intervention	Contrast	Intervention	Contrast
Aparicio et al[20]	Spanish	49	44	EX	RN	8 weeks	a; b	①④
Barbier et al[22]	French	24	28	EX	RN	8 weeks	a	②
Mamipour et al[4]	Iranian	18	17	EX	RN	10 weeks	b	④
Yildirim et al[45]	Istanbul	17	17	EX	RN	12 weeks	b	⑤
Sonmezer et al[39]	Istanbul	20	20	EX	RN	8 weeks	a; b	①④
Backhausen et al[21]	Denmark	240	230	EX	RN	12 weeks	b	⑤
Ozdemir et al[37]	Istanbul	48	48	EX	RN	4 weeks	a; b	①④
Miquelutti et al[36]	Brazilian	97	100	EX	RN	20 weeks	a	①
Kordi et al[33]	Iranian	34	31	EX	EI	6 weeks	b	④
			31		RN
Eggen et al[26]	Norway	129	128	EX	RN	4 weeks	a; b	②⑤
Kluge et al[32]	South Africa	26	24	EX	CON	10 weeks	b	⑤
Peterson et al[38]	America	22	20	EX	MA	6 weeks	a; b	②⑤
			15		PSY
Stafne et al[40]	Norway	397	395	EX	RN	12 weeks	a	①
Elden et al[27]	Sweden	114	109	EX	AC	6 weeks	a	①
			105		RN
Cheng et al[24]	China	53	53	AC	RN	2 weeks	a; b	①④⑤
Vas et al[43]	Spanish	55	55	AC	RN	2 weeks	a; b	①⑤
Bishop et al[23]	England	42	41	AC	RN	8 weeks	b	④
Keskin et al[31]	Istanbul	20	19	ES	EX	3 weeks	b	⑤
			21		CON
Svahn et al[41]	Sweden	54	54	ES	AC	5 weeks	a; b	②④
Li et al[2]	China	38	38	MA	EX	3 weeks	a; b	①④
Vaidya[42]	India	15	15	MA	ES	2 weeks	b	⑤
Melkersson et al[35]	Sweden	35	34	MA	PLA	6 weeks	a	①
Close et al[25]	England	30	30	MA	RN	6 weeks	a; b	①⑤
					PLA
Lee et al[34]	China	30	30	MA	RN	5 days	a	①
Elden et al[28]	Sweden	63	60	MA	RN	6 weeks	a; b	①④
Yazdanpanahi et al[44]	Iranian	50	50	CU	MA	2 weeks	a	③
Akbarzadeh et al[18]	Iranian	50	50	CU	RN	2 weeks	a	①
Heydari et al[30]	Iranian	50	50	EI	CON	3 weeks	a; b	①④
Chen et al[12]	China	31	35	A	ES	6 weeks	a	①
			30		CON
Guo[15]	China	23	23	B	MA	8 weeks	a; b	②⑤
George et al[29]	America	66	71	B	RN	10 weeks	a; b	②⑤
Cheng et al[13]	China	42	44	C	MA	4 weeks	a; b	①④
Shi[16]	China	40	40	D	MA	4 weeks	a; b	②⑤
Akmese and Oran[19]	Istanbul	33	33	E	RN	8 weeks	a	①
Fan[14]	China	28	28	F	EX	2 weeks	a; b	①⑤
Zhang[17]	China	81	81	G	AC	2 weeks	a	①

① Visual Analog Scale (VAS), ② Numerical Rating Scale (NRS), ③ McGill Pain Questionnaire (MPQ), ④ Roland-Morris Lumbar Dysfunction Questionnaire (RMDQ), ⑤ Oswestry Dysfunction Index. EX: exercise; AC: acupuncture; CU: cupping; MA: manipulation; EI: external immobilization therapy; ES: electrical stimulation therapy; PLA: placebo; CON: blank control; RN: routine nursing care; PSY: psychotherapy; A: electrical stimulation-psychotherapy; B: manipulation-exercise; C: manipulation-acupuncture; D: manipulation-electrical stimulation; E: music-relaxation; F: exercise-external immobilization; G: acupuncture-exercise; a: Pain outcome indicators; b: Function outcome indicators.

3.3. Quality assessment

The results of the qualitative evaluation showed that 3 RCTs^[22,28,39] had a quality assessment grade of A, and the remaining 33 studies^{[2,4,12–21,23–27,29–38,40–45]} had a grade of B. All 36 RCTs reported details of random sequence generation; 18 RCTs^{[4,21,22,24–29,31,35–42]} described allocation concealment methods; 5 RCTs^{[22,23,28,35,39]} reported implementing a double-blind method, 14 RCTs^{[4,14,19,22–25,27–29,39,42,43,45]} reported implementing a blinded method for assessors, and 15 RCTs^{[4,18,20–23,26–28,30,33,36–38,40,43]} reported registration plans. All studies assured data completeness without other biases (Fig. 2; Fig. S2, Supplemental Digital Content, https://links.lww.com/MD/P702).

Figure 2.

Graph showing the risk of bia.

3.4. Results of the meta-analysis

3.4.1. Pain scores

Twenty-eight RCTs^{[2,12–20,22,24–30,34–41,43,44]} comprising 3733 participants evaluated the impact of 17 non-pharmacological treatments on pain in women with PLBP. The pain evidence network map indicated that EX, manipulative, and AC therapies were more prevalent than other interventions, establishing a closed loop among EX, psychotherapy (PSY), manipulation (MA), cupping (CU), and routine nursing care (RN) for pain indicators (Fig. 3). Figure 5 displays the key results of the network meta-analyses. Ten non-pharmacological interventions demonstrated statistically significant efficacy in alleviating pain sensations compared to RN: music-relaxation (SMD, 3.58%; 95% CI, 2.21–4.95), MA-ES (SMD, 2.61%; 95% CI. 1.26–3.95), MA-AC (SMD, 2.55%; 95% CI, 1.25–3.85), AC-EX (SMD, 2.00%; 95% CI, 0.66–3.33), MA-EX (SMD, 1.64%; 95% CI, 0.73–2.54), EX-EI (SMD, 1.49%; 95% CI, 0.18–2.80), CU (SMD, 1.28%; 95% CI, 0.40–2.17), AC (SMD, 0.88%; 95% CI, 0.23–1.53), MA (SMD, 0.64%; 95% CI, 0.14–1.15), and EX (SMD, 0.51%; 95% CI, 0.12–0.90). The SUCRA values, from highest to lowest, were as follows: music-relaxation (97.1%) > MA-ES (86.0%) > MA-AC (85.4%) > AC-EX (74.5%) > MA-EX (66.9%) > ES-PSY (64.8%) > EX-EI (61.1%) > CU (56.7%) > ES (46.4%) > EI (45.9%) > AC (42.9%) > MA (34.9%) > EX (29.3%) > PSY (18.1%) > blank control (14.9%) > placebo (13.6%) > RN (11.5%) (Fig. S3, Supplemental Digital Content, https://links.lww.com/MD/P702).

Figure 3.

Network plot of the pain included studies.

Figure 5.

Network meta-analysis league table of the effect of different interventions on pregnancy-related lower back pain. (Note: lower left: function indicators; upper right: pain indicators; –: indicates not involved; *: indicates P < .05.)

3.4.2. Function state scores

A total of 25 RCTs^{[2,4,13–16,20,21,23–26,28–33,37–39,41–43,45]} and 12 non-pharmacological therapies were evaluated for their efficacy regarding the functional states. A network diagram was created, showing a closed loop among AC, RN, MA, PSY, EX, and ES (Fig. 4). The NMA showed that MA-AC (SMD, 1.96%; 95% CI, 0.08–3.85), MA-EX (SMD, 1.63%; 95% CI, 0.36–2.91), EI (SMD, 1.30%; 95% CI, 0.03–2.56), AC (SMD, 1.23%; 95% CI, 0.45–2.01), ES (SMD, 1.23%; 95% CI, 0.15–2.32), and EX (SMD, 0.86%; 95% CI, 0.31–1.41) improved the patients’ lower lumbar function compared to RN (Fig. 5). The SUCRA values, from highest to lowest, were as follows: MA-AC (78.3%) > MA-ES (73.9%) > MA-EX (72.3%) > EX-EI (63.1%) > EI (60.3%) > AC (58.8%) > ES (58.4%) > PSY (44.0%) > EX (40.1%) > MA (34.2%) > RN (8.4%) > blank control (8.1%) (Fig. S4, Supplemental Digital Content, https://links.lww.com/MD/P702).

Figure 4.

Network plot of the function included studies. (Note: EX: Exercise; AC: Acupuncture; CU: Cupping; MA: manipulation; EI: External immobilization therapy; ES: Electrical stimulation therapy; PLA: Placebo; CON: Blank control; RN: Routine nursing care; PSY: psychotherapy; A: electrical stimulation-psychotherapy; B: manipulation-exercise; C: manipulation-acupuncture; D: manipulation-electrical stimulation; E: music-relaxation; F: exercise-external immobilization; G: acupuncture-exercise.)

3.5. Heterogeneity and consistency analysis

The results of the overall data inconsistency test showed that the pain (P = .429) and physical function (P = .337) indicators were > .05, indicating good overall consistency. The pain and physical function outcome indicators formed a closed loop, and the results of the local inconsistency test showed P > .05, indicating no significant inconsistencies.

3.6. Publication bias analysis

Funnel plots showed signs of symmetry, indicating a relatively low likelihood of publication bias (Figs. 6 and 7).

Figure 6.

Funnel chart of pain indicators.

Figure 7.

Funnel chart of functional indicators.

4. Discussion

The SUCRA showed that music-relaxation therapy, as part of the combination treatment, was considered the best intervention option. Music can activate inhibitory nerve reflexes and reduce pain signal transmission.^[46] Furthermore, it influences the hippocampus and sympathetic nervous system, lowers cortisol levels, and enhances the patient’s physiological condition.^[47] Additionally, music activates the brain’s reward circuitry, releasing dopamine and endorphins, alleviating tension and anxiety, and facilitating relaxation, thereby contributing to pain relief.^[48] Relaxation therapy encompasses the acquisition and application of diverse techniques, including deep breathing, and progressive muscle relaxation, etc,^[49] primarily aimed at directing the patient’s focus toward muscle contraction and relaxation dynamics, thereby diminishing the body’s stress response, relieving localized muscle tension and stiffness, and reducing the sensation of pain.^[50] Music-relaxation therapy effectively merges the benefits of both approaches for a more pronounced effect. This strategy offers excellent practicality, safety, and cost-effectiveness, making it acceptable for implementation in pain management in patients with PLBP. The efficacy hinges on the patient’s ability to adhere to treatment in the long term, which is frequently inadequate when they undergo treatment in isolation at home. Given the prevalence of mobile devices like smartphones, patients can monitor their training and connect with physicians instantaneously through applications in the future; thus, it is imperative to conduct and enhance pertinent high-quality RCTs to validate the findings of this study.

The SUCRA ranking results demonstrated that MA-AC therapy was the most effective combined treatment. From the perspective of traditional Chinese medicine, lumbar pain belongs to the “pain paralysis” category, primarily resulting from meridian obstruction, impaired qi and blood circulation, and blood stasis in the lumbar area. AC therapy utilizes the principles of evidence-based traditional Chinese medicine to stimulate specific AC points, such as the Shen Shu and Large Intestine Shu.^[17] AC signals are conveyed through the nervous system to the brain, where they integrate with pain signal feedback, activating pain regulatory mechanisms within the brain.^[51] Moreover, local AC stimulates the brain to release opioids, including beta-endorphin and enkephalin, alongside neurotransmitters such as 5-hydroxytryptamine and dopamine, significantly diminishes pain perception, promotes blood circulation, and improves overall functionality.^[6] Critically, relevant meta-analyses confirm that AC’s superiority over pharmacotherapy alone in restoring functional capacity in low back pain patients.^[52]

Physiologically, the uterus expands during pregnancy, causing the body’s center of gravity to shift forward, resulting in reduced lumbar spine stability.^[8] Manipulative therapy employs noninvasive techniques or integrates mechanical devices to deliver treatment directly to the affected region or relevant acupoints, aimed at stimulating the meridians, rectifying anatomical positional abnormalities, and enhancing the functionality of involved joints,^[53] such as massage, Tui Na, and traction. However, the central effects caused by different types of MA are different. Traction can expand the lumbar intervertebral space and facilitate the restoration of the intervertebral foramen to its normal configuration, thereby alleviating tension on the nerve roots and diminishing intervertebral pressure.^[54] Tui Na, wherein external forces exerted through kneading and pressing are directed at acupoints adjacent to the spine, such as the Vital Gate, Lumbar Yang Gate, and Kidney Shu, address lumbar and back ailments, fortify the skeletal structure, and alleviate local muscle spasms.^[55] MA-AC is characterized by high safety and significant efficacy; however, skilled therapists must adeptly modify the intensity and frequency of interventions based on the patient’s specific condition to attain an optimal intervention effect.

This study considers the extensive sample and tighter CIs in this NMA, and the results are trustworthy. However, the study has some limitations. First, several non-pharmacological interventions that could not be formed into a closed loop were excluded, which may have reduced comprehensiveness. Second, several studies did not implement allocation concealment or blinding; hence, the results may have some potential bias. Third, the same type of interventions differed in duration and frequency, which may have weakened the accuracy of the results.

5. Conclusion

In conclusion, non-pharmacological interventions for PLBP demonstrated considerable effectiveness, with music-relaxation therapy being the most effective in alleviating pain, while MA-AC therapy exhibited the highest likelihood of enhancing physical function. In addition, combination therapy had greater efficacy than monotherapy in this trial, possibly due to the effect of multiple treatments working together. Caregivers are encouraged to select an appropriate combination of therapies to manage PLBP effectively, tailoring interventions to the patient’s circumstances and needs.

Acknowledgments

Everyone who contributed significantly to the work has been listed and has given their consent.

Author contributions

Data curation: Haijie Zhang, Guangjian Zhang.

Formal analysis: Suyue Wang.

Funding acquisition: Guangjian Zhang.

Software: Suyue Wang.

Writing – original draft: Suyue Wang.

Writing – review & editing: Lihua Jin.