Effects of Exercise on Lumbopelvic Pain During Pregnancy: A Systematic Review and Meta-analysis

Abstract

Background: 

At least 50% of women experience lumbopelvic pain during pregnancy. Physical activity has moderate health benefits and no adverse outcomes for women with uncomplicated pregnancies. Physically active pregnant women have lower pain intensities than sedentary women. It is not clear whether common exercise types are equally beneficial for the management of lumbopelvic pain during pregnancy.

Objective: 

The aim of this systematic review is to determine the effect of exercise on reducing lumbopelvic pain intensity in pregnant women.

Study Design: 

A systematic review and meta-analysis.

Methods: 

Seven databases (CINAHL, Medline, ScienceDirect, SportDiscus, ProQuest, Google Scholar, and Cochrane) were searched in October 2020 and again in February 2022. Studies investigating the effect of exercise on low back or pelvic pain in pregnant women were included, with no limitation on publication date. Primary data collected for quantitative analysis included pain intensity. Bias was assessed using the Cochrane bias tool. A meta-analysis was performed using individual study effect sizes (ES) using a random-effects model.

Results: 

Initial search yielded 1,771 results and 21 articles were identified as meeting the inclusion criteria. One article utilized 2 intervention groups and was counted as separate studies. Meta-analysis on 22 studies yielded a statistically significant and large ES, indicating exercise intervention during pregnancy results in significantly lower pain reports compared with usual prenatal care (overall ES = 2.07; 95% confidence interval = 1.35-2.78; P < .001).

Conclusion: 

Exercise is beneficial in the management of lumbopelvic pain in pregnant women.

INTRODUCTION

Lumbopelvic pain during pregnancy may be defined most simply as continuous or recurrent pain from the lumbar spine or pelvis for more than 1 week.¹ Although lumbopelvic pain can be further differentiated into low back and pelvic girdle pain, women experiencing both have reported greater consequences on their health and function.² A variety of causative factors for pregnancy-related lumbopelvic pain have been proposed, including hormonal factors and biomechanical changes related to spinal load, pelvic girdle stability, abdominal diameter, and fetal weight; however, these continue to be disputed in the literature and the true etiology of such pain is poorly understood.^1,3 Estimates of the prevalence of lumbopelvic pain in pregnant women vary widely, from 25% to more than 80%, with many sources estimating that 50% of pregnant women experience lumbopelvic pain at some point during their pregnancy.^3–6 It has been proposed that lumbopelvic pain is a normal part of pregnancy, yet one-third of these women report pain that is significant enough to impact their daily activity, quality of life, and time away from work.^1,3,4,6 Such impairment should be considered a complication of pregnancy and should not be ignored. Pain intensity varies widely among individuals, but generally increases over the course of pregnancy.⁶ Significant differences in pain intensity have been found; however, between active and sedentary pregnant women, with sedentary women being 30% more likely to report severe pain.⁷

The American College of Obstetricians and Gynecologists states that exercise, including a variety of aerobic and resistance-type exercises, is safe and recommended for pregnant women as a means of facilitating overall maternal and fetal health.⁸ Exercise has also been recommended as a frontline, nonpharmacologic treatment of lumbopelvic pain in pregnant women, as there is minimal risk of adverse reactions in terms of maternal and fetal morbidity or mortality.⁹ Recommended exercises are similar to those that are recommended for nonpregnant individuals with low back pain, with modifications such as avoiding a prone or prolonged supine position made as needed to accommodate pregnancy.⁸

Previous systematic literature reviews have evaluated the impact of a variety of conservative interventions, such as pelvic support belts,^10,11 acupuncture,^10–12 postural education,¹³ pharmacology,¹⁴ manual therapies,^11–13 and exercise^11,12,15 on pregnancy-related lumbopelvic pain. These reviews have found benefits to pain, disability, and quality of life, ranging from none to significant with moderate- to low-quality evidence.^10–15 Some studies have compared exercise to other conservative interventions, which does not allow for analysis of efficacy against routine prenatal care.¹³ In 2015, Belogolovsky et al¹⁶ performed a meta-analysis of land-based exercise programs with level of evidence 2b or higher from the previous 10 years and found statistically significant improvements in low back pain among pregnant women with exercise, though again this was limited to the previous 10 years and did not include any aquatic exercise interventions.

While evidence supports an exercise benefit in overall maternal and fetal health, as well as in reducing lumbopelvic pain, it is not clear from current literature which type of exercise should be recommended to reduce lumbopelvic pain severity in pregnant women. Several new randomized controlled trials (RCTs) on this topic have been published in recent years, which have yet to be included in comparative analyses.^17–20 Additionally, previous reviews have been inconsistent as to how they define, differentiate, and include pelvic pain when evaluating the impact of exercise or other conservative therapies on pregnancy-related low back pain. With many definitions of pelvic pain, some of which overlap with the definition of low back pain, drawing a distinction between the 2 when analyzing prior literature is difficult. Due to the lack of consistent definition and reporting, this review takes a broad view of lumbopelvic pain as defined previously, including pain in the pelvic region or lumbar spine. Therefore, the primary aim of this systematic review is to include the full history of evidence to determine the benefits of exercise on reducing lumbopelvic pain intensity in pregnant women. A secondary aim is to identify potential moderators of response to exercise.

METHODS

Search Strategy and Inclusion and Exclusion Criteria

A systematic literature review was performed in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidelines. This study was not prospectively registered with PROSPERO. The authors searched 7 databases (Medline Complete, CINAHL, ScienceDirect, SportDiscus, ProQuest, Google Scholar, and Cochrane Library) in October 2020 and again in February 2022, in consultation with a librarian. No restrictions on language, publication date, or study design were imposed on the search. Reference lists of relevant review articles and the included studies were hand searched to identify additional studies. The search terms and strategy for Medline were as follows: “Pregnancy” [MeSH] AND “Exercise” [MeSH] AND “Low Back Pain” [MeSH] OR “Pelvic Pain” [MeSH]. Search terms and strategies were modified as needed for all other databases. A full list of search terms is included in Appendix 1.

Studies meeting the following criteria were considered for review: participants were pregnant women at least 18 years of age with any number of past pregnancies and carrying any number of fetuses, a single exercise type was used as the intervention, pain intensity was reported as an outcome measure, and outcomes were measured prior to delivery. In most studies, pain intensity was primarily reported through the visual analog scale (VAS) or numeric rating scale, with some instead reporting Quebec Back Pain Questionnaire or Smith's Pregnancy Discomfort Index scores. Exercise types could include aerobic, aquatic, resistance training, mobility exercise, core stabilization, or relaxation exercises. In studies with multiple intervention groups, each intervention was compared independently to the control. Intervention could include either supervised or unsupervised exercise. Studies were excluded for the following reasons: participants had a history of chronic low back or pelvic pain, participants had a complicated pregnancy, exercise was combined with other interventions in the treatment group (such as manual therapies or modalities), no control group was used, or the control group had an intervention other than usual daily activity or usual prenatal care.

Study Selection

A total of 1771 articles were identified through database searching, with 670 remaining after duplicates were removed. Three additional studies were included based on hand searching of reference lists. Of those, 496 were excluded based on title and abstract reviews. The remaining 174 were evaluated through review of the full text. After full-text review, 153 articles were excluded based on the inclusion and exclusion criteria listed earlier, and inability to access the full text or gain necessary data from authors. Two of the present study's authors independently reviewed articles at each stage of the process. Any disagreements were resolved by a third author. The selection and review processes are summarized in Figure 1.

Figure 1.: PRISMA flowchart for selection of studies in the systematic review. PRISMA indicates Preferred Reporting Items for Systematic Reviews and Meta-analyses.

Data Processing

For the meta-analysis, data extraction included sample size, numeric pain values, mean participant age, type of exercise intervention, intervention frequency (sessions/week), intensity (minutes/week), and duration (total weeks). Data were extracted as means, standard deviations (SDs), and sample sizes for most of the studies. Pre-/post-data were collected for the remaining studies. When data were presented in a format other than means, SDs, and sample sizes, authors were contacted at least 2 times to obtain relevant data. If no response was received after 2 attempts or necessary data were not available, those articles were excluded.

The extracted pain intensity data were converted to a standard format by calculating the standardized mean difference, which will be referred to as effect size (ES). The ES calculation was set so that a positive ES indicates lower pain ratings in the exercise intervention groups compared with control groups. Independent group studies in which means, SDs, and sample sizes were reported (ie, the most common scenario [n = 15 studies]), standardized mean differences were calculated.²¹ We assumed an intertrial correlation of 0.5 for all outcome measures in these calculations since no individual subject data were available. The remaining studies reported pre-/post-means, P values, and sample sizes (n = 1), pre-/post-means, and P value (n = 1), difference in means with confidence intervals (CIs) (n = 2), and groups means, sample size, and P value (n = 3), with the standardized mean difference calculated as previously described.²¹

The meta-analysis was conducted using a random-effects model to account for experimental variability across the included studies (eg, type of exercise, duration of program, and outcomes assessed). The extent of heterogeneity (ie, between-study variation) was assessed using both the I² value and a χ² test of the Q value. Because large heterogeneity was observed, we sought to determine the role of various factors in explaining that heterogeneity. Meta-regression, using a method-of-moments model, was used to assess continuous moderator variables, that is age of mother, program duration (weeks), treatment frequency (days/week), and dose of exercise (minutes/week). Subgroup meta-analysis was used to assess nominal moderator variables in which subsets of studies are compared using Q tests based on analysis of variance. This test along with the common assumption of study variance across subgroups was used to assess several factors such as type of exercise and study design. Subgroups were required to include 2 or more studies to be analyzed.

Meta-analyses were performed using Comprehensive Meta-Analysis software (version 3.0, Biostat, Inc), with significance of P < .05. ESs of 0.2, 0.5, and 0.8 were considered to be small, moderate, and large, respectively.^22,23I² values of 25%, 50%, and 75% were considered to indicate low, medium, and large degrees of heterogeneity, respectively. Publication bias was assessed using a funnel plot of ES versus standard error.

Risk of Bias Across Studies

A Cochrane risk of bias assessment was conducted to determine methodological quality of the included studies as recommended by the Cochrane Collaboration²⁴ (Table 1). This assessment evaluated each study for risk of bias in areas of participant randomization practices, effects of adhering to the intervention, missing outcome data, methods of measuring outcomes, and selection of reported results. An algorithm defined by the tool was then used to determine an overall risk of bias for each study, with possible outcomes being low risk of bias, some concerns, or high risk of bias. Data extraction and risk of bias assessments for each study were performed independently by 2 of the present study's authors, with disagreements settled by a third author.

Table 1. - Included Study Characteristics^a

Study	Study Design	Calculated Effect Size	Exercise Type	Sample Size Group, n	Participant Age (Weighted Mean)	Intervention Duration, wk	Outcome	Overall Bias
Akmeşe et al25	RCT	4.73	Relaxation	Exercise (33) Control (33)	N/A	8	VAS	Some concerns
Backhausen et al26	RCT	0.18	Aquatic	Exercise (258) Control (258)	31	12	NRS	Some concerns
Banpei et al27	RCT	0.62	Core stability	Exercise (57) Control (55)	28	Not specified, follow-up at 12 wk	VAS	Some concerns
Eggen et al28	RCT	0.08	Core stability	Exercise (103) Control (107)	21	16-20	NRS	High
Figueira et al29	RCT	1.81	Mobility	Exercise (10) Control (10)	25	9	VAS	Some concerns
Filipec et al30	RCT	16.1	Core stability	Exercise (207) Control (201)	33	6	VAS	Some concerns
Gil et al31	RCT	4.52	Mobility	Exercise (17) Control (17)	26	8 sessions (frequency not specified)	NRS	Some concerns
Kamali et al32	RCT	2.34	Mobility	Exercise (30) Control (30)	30	8	VAS	Some concerns
Kihlstrand et al33	RCT	0.16	Aquatic	Exercise (123) Control (118)	29	17-20	VAS	Some concerns
Kluge et al34	RCT	−0.55	Core stability	Exercise (24) Control (22)	28	10	BPI	High
Lamadah et al19	RCT	2.00	Relaxation	Exercise (20) Control (20)	N/A	8	VAS	High
Nilsson-Wikmar et al35	RCT	−0.05	Core stability	Exercise (41) Control (40)	29	14 (average)	VAS	High
Nilsson-Wikmar et al35	RCT	−0.05	Core stability	Exercise (37) Control (40)	29	14 (average)	VAS	High
Ozdemir et al36	RCT	0.79	Aerobic	Exercise (48) Control (48)	30	4	VAS	Some concerns
Podebradska et al37	RCT	0.50	Foot strengthening	Exercise (6) Control (6)	25	Not specified	VAS	Some concerns
Sayed et al38	RCT	0.57	Core stability	Exercise (29) Control (29)	26	4	VAS	Some concerns
Sedaghati et al39	RCT	1.01	Aerobic	Exercise (40) Control (50)	23	8	QBPDS	Some concerns
Shim et al40	Pre-post	0.61	Core stability	Exercise (29) Control (27)	N/A	12	VAS	High
Smith et al41	Pre-post	1.63	Aquatic	Exercise (20) Control (20)	25	6	SPDII	High
Sonmezer et al20	RCT	1.46	Core stability	Exercise (20) Control (20)	29	8	VAS	Some concerns
Suputtitada et al42	RCT	5.40	Core stability	Exercise (31) Control (34)	N/A	8	VAS	Some concerns
Yousefabadi et al43	RCT	3.52	Mobility	Exercise (26) Control (25)	28	6	QBPDS	High

Abbreviations: BPI, Brief Pain Inventory; NRS, numeric pain rating scale; QBPDS, Quebec Back Pain Disability Score; RCT, randomized controlled trial; SPDII, Smith's Pregnancy Discomfort Intensity Index; VAS, visual analog scale.

^aN/A in the age column means authors reported ages as a range or percent, which could not be calculated as a weighted mean. Effect sizes of 0.2, 0.5, and 0.8 were considered small, moderate, and large, respectively.

RESULTS

Description of Included Studies

In total, 21 studies published between 1999 and 2021 were included in this analysis. Studies included RCTs (n = 19) and pre-/post-design (n = 2). A total of 2379 participants were included in this review, with 1170 in control groups and 1209 in intervention groups. The mean age of participants was 27.6 years, with an overall range of 18 to 42 years. No study reported a significant difference in age between control and intervention groups.

Intervention exercises were grouped into the following categories: core stability (n = 10), mobility (n = 4), aquatic (n = 3), aerobic (n = 2), foot strengthening (n = 1), and relaxation (n = 2). Due to the inconsistent reporting of exercise, intervention studies were grouped into the categories by the present study's authors. Core stability included isometric and concentric exercises to strengthen the abdominal and pelvic musculature (bridges, curl-ups, Kegels, straight leg raise, quadruped, and transverse abdominis engagement), exercises targeting lumbopelvic stabilization during functional activities such as squatting and lifting, and Pilates training. Mobility included pelvic tilts and stretching exercises targeting the hamstrings, hip adductors, and thoracic and lumbar musculature. Aquatic exercise included any program taking place in a pool environment. Aerobic exercise included walking and cycling. Relaxation included breathing exercises and progressive muscle relaxation. All intervention groups fell into only one of the identified categories. The VAS was the primary method for reporting pain intensity across studies (n = 15), with remaining studies alternatively using the Quebec Back Pain Questionnaire (n = 2), Smith's Pregnancy Discomfort Index (n = 1), Brief Pain Inventory (n = 1), and a numeric pain rating scale (n = 3).

Seven studies were found to have a high risk of bias overall, with the remaining 14 studies found to have some concerns. All studies were determined to have at least some concerns in the measurement of outcome domain, as pain intensity was self-reported by the participants. Most studies (n = 14) were also found to have high risk or some concerns within the randomization domain due to assigning participants to groups using alternation or allowing participants to choose their group.

Meta-analysis

To determine whether exercise is better than no treatment or usual prenatal care for reducing pregnancy-related low back and pelvic girdle pain, 22 ESs were calculated from 21 studies. One study included 2 intervention groups. Each intervention was considered a separate study in the analysis. Therefore, one average ES was calculated for 22 studies. There was a considerable range of ESs among the 22 studies included, −0.06 to 16.2 (Figure 2). The first 3 studies in Figure 2 exhibited negative ESs, indicating the control groups reported lower pain intensity than the intervention groups. Conversely, 19 studies exhibited positive, beneficial effects of exercise on reported pain intensity compared with control. Overall meta-analysis on the combined 22 studies yielded a statistically significant and large ES, indicating exercise intervention during pregnancy results in significantly lower-reported pain intensity compared with usual prenatal care (overall ES = 2.07; 95% CI = 1.35-2.78; P < .001) (Figure 2).

Figure 2.: Forest plot of effect sizes (ESs) from the 22 studies that assessed effect of exercise on pregnancy-related low back and pelvic pain. A square in the plot represents the ES for a given study, with the size representing the weight of the study in the meta-analysis. Each horizontal line represents the 95% confidence interval (CI) for each study. The plot is arranged from the lowest to highest ES. The diamond at the bottom represents the overall ES calculated using a random-effects model. The width of the diamond is the overall 95% CI.

A funnel plot was created to visualize potential publication bias within the primary meta-analysis. Minimal asymmetry was noted with the Filipec and Matijevic¹⁸ article published in 2021 noted as the outlier. Removal of this study still produced significant results (overall ES = 1.36; 95% CI = 0.93-1.80; P < .001) and reduced the asymmetry in the funnel plot. Because of the substantial heterogeneity observed, it was not feasible to calculate an overall ES adjusted for publication bias.

Between-study variance in ES was determined to be large (I² = 98.1%) and statistically significant (Q-df = 21, P < .001). The large heterogeneity justifies using subgroup meta-analysis and metaregression in an attempt to explain the between-study variance. Table 2 summarizes the findings of the subgroup meta-analysis investigating the role of 2 nominal moderator variables, study design and exercise type. There was a significant difference between the 2 identified study types, RCT and pre-/post-design, P < .001. Removing either type continued to produce a significant overall ES; therefore, we chose to include both study design types in our overall analysis. Subgroup analysis of a single exercise type compared with all other exercise types was conducted. All subgroup analyses of exercise type were significant, P < .001 (Table 2).

Table 2. - Summary of Nominal Moderator Variable Subgroup Meta-analyses That Might Explain Between-Study Variance in Effect Size

Moderator Variable	Comparison	P Value
Intervention type: Studies that included aquatic exercise vs those that did not	Aquatic exercise (n = 3, ES = 0.49 [0.007-0.971]) vs no aquatic exercise (n = 19, ES = 2.332 [1.37-3.30])	<.001
Intervention type: Studies that included core stability exercise vs those that did not	Core stability exercise (n = 10, ES = 2.35 [0.86-3.83]) vs no core stability exercise (n = 12, ES = 1.85 [1.19-2.50])	<.001
Intervention type: Studies that included aerobic exercise vs those that did not	Aerobic exercise (n = 2, ES = 0.89 [0.59-1.19]) vs no aerobic exercise (n = 20, ES = 2.20 [1.40-2.99])	<.001
Intervention type: Studies that included mobility exercise vs those that did not	Mobility exercise (n = 4, ES = 2.99 [1.95-4.03]) vs no mobility exercise (n = 18, ES = 1.86 [1.09-2.63])	<.001
Intervention type: Studies that included relaxation exercise vs those that did not	Relaxation exercise (n = 2, ES = 3.35 [0.68-6.02]) vs no relaxation exercise (n = 20, ES = 1.94 [1.21-2.67])	<.001
Study type	Randomized controlled trials (n = 19, ES = 2.25 [1.46-3.04]) vs pre-post (n = 2, ES = 1.09 [0.10-2.09])	<.001

Abbreviation: ES, effect size.

Meta-regression analysis was performed for the 4 continuous moderator variables (age, days per week, minutes per week, and total program duration) (Table 3). Total treatment duration (number of weeks) accounts for 5% of the variance, P < .001. The negative slope of the regression indicates an inverse relationship, meaning the longer the duration of treatment the more similar pain ratings between the control and intervention groups. Neither the mean age of the mother, frequency of interventions, days, or minutes per week were significant.

Table 3. - Summary of Metaregressions of Continuous Moderator Variables That May Explain Between-Study Variance in Effect Size

Moderator Variable	Slope	P Value
Mean age of mother, y	0.083	.657
Frequency of exercise, times/wk	−0.1901	.356
Duration, min/wk	−0.0046	.477
Total program duration, wk	−0.3381	.000

DISCUSSION

The main aim of this systematic review and meta-analysis was to determine the effects of exercise on reducing lumbopelvic pain in pregnant women. Overall, our results showed a very large, positive impact of exercise on reducing reported lumbopelvic pain intensity during pregnancy. Because of the large heterogeneity subanalyses were performed, yet no significant explanation of the variance was found. Together, these results indicate that, during pregnancy, exercise regardless of type is beneficial in reducing pain intensity.

Our results are consistent with the meta-analysis and evidence-based review performed by Belogolovsky et al¹⁶ in 2015. They included literature from the previous 10 years evaluating the impact of land-based exercise and found a statistically significant effect of exercise for reducing pregnancy-related lumbopelvic pain. Land-based exercises described in the review included yoga and low back strength and stabilization activities. Yoga and low back strengthening would be included in our categories of mobility and core stability exercise.

Although we combined low back and pelvic pain studies, our findings are similar to previous literature that investigated pain location independently of one another. Liddle and Pennick¹² found moderate-quality evidence that exercise reduced the number of women with pregnancy-related low back pain. They found low-quality evidence that exercise provided no benefit over usual prenatal care when it comes to pregnancy-related pelvic girdle pain. A 2014 systematic review by Van Benten et al⁵ found moderate-quality evidence to support the benefits of exercise in reducing low back pain in pregnant women. They also found that interventions, which included education, showed positive results on pain; however, it is not clear whether it is the education itself or the combination of interventions, which provides the benefit. Similarly, a 2008 guideline by Vleeming et al⁴⁴ found no evidence to support providing information alone as treatment for pregnancy-related pelvic girdle pain. Collectively, these studies support our conclusion that exercise reduces pregnancy-related lumbopelvic pain.

Consistent with our findings, many previous reviews and meta-analyses cite heterogeneity among included studies as a limitation in making clear conclusions and recommendations.^5,12,15,45 Colla et al⁴⁵ state that this heterogeneity makes it “impossible to reach a consensus or any conclusions about which protocol of therapeutic exercise is more effective... for pregnancy low back and pelvic pain.” Indeed, the results from our subgroup analyses failed to identify that one form of exercise was significantly better than another. Additional research is needed to provide larger pools of evidence from which to draw more clear conclusions related to the benefits of specific protocols over others.

This meta-analysis found that the ES of the benefit of exercise on pain decreased as the duration of the intervention increased. These results are consistent with the findings of Maia et al,¹³ which found that the difference in pain ratings during pregnancy with exercise compared with routine care was greater in the short term (up to 12 weeks) than in the long term (>12 weeks). Previous literature has discussed the pattern of increased pain with the progression of pregnancy,⁶ indicating that perhaps it is not the duration of the intervention that changes the ES, but rather the later stage of pregnancy. Regardless, our findings indicate that pain intensity is still lower among those who exercise, even in the later stages of pregnancy. This is consistent with de Sousa et al,⁷ who found in a prospective trial that sedentary women were 30% more likely to have high pain intensity when compared with active women, regardless of trimester or weight gain.

Our review evaluates the impact of exercise on pain intensity during pregnancy; however, exercise performed during pregnancy has implications for reducing risk of pain as well. Shiri et al¹⁵ found reduced risk of pregnancy-related low back pain among women who exercised compared with those who engaged in usual daily activities and usual prenatal care. Additionally, exercise benefit can extend into the postpartum period. In a retrospective self-report, Mota et al⁶ report that 93% of women who had lumbopelvic pain during pregnancy did not seek treatment, expecting that the pain would resolve following delivery. While this is true for many women, Gutke et al⁴⁶ found that 33% of women continued to have pain at 3 months postpartum, and Norén et al⁴⁷ found that 20% of women who had lumbopelvic pain during pregnancy still had pain 3 years later. In a cohort study, Ha et al⁴⁸ found an inverse relationship between prenatal physical activity and postpartum low back pain. This evidence suggests that, in addition to our findings of the significant benefit of exercise on pregnancy-related lumbopelvic pain, the benefits of exercise during pregnancy extend beyond this stage.

Limitations

Limitations of our study include the following: (1) inclusion of studies with high risk of bias, (2) potential publication bias, (3) significantly large heterogeneity, and (4) lack of clarity and consistency of reported interventions in the included studies. Seven studies were identified as having high risk of bias. The studies identified as having high risk were mainly because they did not randomize participants into groups, meaning they scored as “high risk” on the Cochrane risk of bias assessment tool. However, if these studies were removed, the ES increased and remained significant (overall ES = 2.77; P < .001). In relation to risk of bias in measurement of the outcome, the Cochrane risk of bias tool specifies that any self-reported outcome has an inherent risk of bias.

Publication bias is when research that appears in the published literature is systematically unrepresentative of the population of completed studies. The tendency is for studies with nonsignificant findings to not be published. Meta-analyses tend to be biased toward published studies because of difficulty identifying and obtaining data from unpublished research. Our assessment of publication bias found some asymmetry in the funnel plot, which was found to be mainly due to one largely significant study. When that study was removed, our overall findings remained significant and the asymmetry in the funnel plot was reduced.

The large and significant amount of heterogeneity among studies is a limitation. The only subanalysis that would explain any of the heterogeneity was the total duration of the intervention. Although the subgroup meta-analyses performed for exercise type were all significantly different from one another, all exercise types indicated a reduction in pregnancy-related lumbopelvic pain. Because of the large heterogeneity among studies, the results of this analysis should be interpreted with caution.

Finally, the lack of clarity and consistency of reported interventions among the included studies is a moderate limitation. For example, in articles these authors grouped as core stability there was high variability in the detail of the intervention performed. Eggen et al²⁸ provided names and images of the exercises performed but did not specify the number of sets or repetitions or the intensity of the exercises. Kluge et al³⁴ simply reported their exercises “focused on the transverse abdominal and pelvic floor muscles” with progression to include cocontraction of various other muscle groups. No additional detail regarding specific exercises, positions, or intensity was provided. The lack of necessary detail in reporting of specific interventions, frequency, and intensity only allows for general conclusions and recommendations.

Clinical Implications and Future Research

The results of this study, in addition to prior literature, indicate that exercise can be beneficial in reducing pregnancy-related lumbopelvic pain. Since our findings did not identify a single type of exercise that was most beneficial in reducing pain intensity, women should be encouraged to participate in whatever type of physical activity they enjoy and can sustain, as any may be beneficial. As noted in a 2019 guideline for physical activity throughout pregnancy, Mottola et al⁹ conclude that the benefits are moderate, no harms have been identified, and it is a feasible intervention requiring minimal resources from both individuals and health systems.

Future research may provide more clarity around the time points during pregnancy during which exercise provides meaningful benefits for pain management as compared with usual prenatal care. Further clarification is needed around the differentiation between lumbar pain and pelvic girdle pain during pregnancy, and the respective benefits from exercise for pain management of each.

CONCLUSION

In summary, the findings from our systematic review and meta-analysis provide additional evidence that exercise during pregnancy is beneficial in reducing pregnancy-related lumbopelvic pain. Pregnant women who are experiencing symptoms should be provided education related to the benefits of exercise on reducing pain.