2.1. Overview

This study followed a systematic review and meta-analysis recommendation protocol of PRISMA-P [19]. The protocol was registered under the identification number CRD42023484380 in PROSPERO before the start of the review.

2.2. Study Eligibility

2.3. Inclusion Criteria

Studies that used observational (prospective and retrospective cohorts) and interventional (randomized controlled trials, RCTs) epidemiological designs, comparing the incidence of back pain after delivery with and without neuraxial anesthesia (spinal anesthesia, epidural anesthesia, and combined spinal-epidural anesthesia), were eligible for the study.

2.4. Exclusion Criteria

Studies that reported in non-English languages and does not compare the incidence of back pain after delivery and neuraxial anesthesia.

2.5. Data Sources

Systematic literature search was conducted from a database: Cochrane Reviews, The Virtual Health Library, National Library of Medicine PubMed, and Google Scholar.

2.6. Search Strategy

Search was conducted by author ST and AB starting from May 1^st 2024, encompassing studies conducted at any year.

3.1. The Virtual Health Library

(Neuraxial anesthesia OR spinal anesthesia OR epidural anesthesia) AND (back pain OR backache OR low back pain) AND (labor OR delivery OR caesarean section OR childbirth).

3.2. Google Scholar

Back pain childbirth neuraxial spinal and epidural anesthesia “labor” OR “caesarean section” OR “back pain” OR backache OR “spinal anesthesia” OR “epidural anesthesia” OR “epidural analgesia” OR “neuraxial anesthesia” “back pain”.

3.3. Cochrane Library

3.4. Study Selection

All citations identified by our search strategy were exported to EndNote, bibliographic management software, and duplicates were removed. The identified articles were read thoroughly by two reviewers SA and AB independently to make sure they are suitable for data extraction. A disagreement between the two reviewers was solved by a third reviewer GW. A hand search was performed on the reference lists of selected articles in order to include studies that were not identified by the search strategy. The search process is presented in a PRISMA flowchart [20].

3.5. Data Extraction

Extraction was done using a standardized data extraction format with Microsoft Excel 2010, which was adopted from the Joanna Briggs Institute (JBI) data extraction format, which includes author, publication year, study area, study design, response rate, follow up period, data collection method (retrospective or prospective), region the study was conducted in, sample size, incidence of back pain in exposed, and nonexposed group and the quality score of each study. The extraction form was performed by two independent reviewers SA and AB. Disagreement was fixed through consensus or inviting third neutral reviewer, WM.

3.6. Outcome of the Study

The outcome variable was incidence of back pain after delivery. Back pain after delivery is defined by a pain or discomfort localized at the lower back that starts two to 3 hours after neuraxial anesthesia and delivery lasting up to 6 weeks and more [9, 10, 16].

3.7. Quality Appraisal

The JBI check list was used to assess the quality of researches based on the listed criteria, and two reviewers assessed for those studies. Based on the quality assessment studies which had seven and more result were identified as qualified to be included in the study.

3.8. Data Analysis

Data were analyzed using Stata: Statistical software for data science version 17. Data were presented from eligible studies in evidence table and summaries using descriptive statistics. A forest plot generated to show the individual study effect size with odds ratio (OR) and 95% CI using the random effect Restricted maximum—likelihood model, displaying the author’s name, publication year and study weights.

The heterogeneity in the analysis was assessed by the using Cochrane Q test statistic with p value less < or = 0.1 considered statistical significant heterogeneity and tau² index test expressed as, together with 90% confidence intervals (CI). I² was tabulated to assess the inconsistency between studies. For I² values of 0%–25% represented minimal inconsistency, 26%–75% represented moderate inconsistency, and values greater than 75% represented substantial inconsistency. Since there was substantial heterogeneity further subgroup analysis was done. In order to detect the outlier studies that resulted in significant heterogeneity Galbraith plot was performed. For small study effect, funnel plot was done with associated Begg’s regression tests.

The meta-analyses were performed by calculating ORs from the incidence of postpartum back pain in delivery groups with and without neuraxial anesthesia, extracted from the studies. The random effect restricted maximum—likelihood model was used to assess the effect of neuraxial anesthesia on occurrence of postpartum back pain. Weighted ORs and 95% CI were calculated to pool study and control groups in each publication for analysis.

3.9. Subgroup Analysis

Subgroup analysis was conducted to minimize the possible source of heterogeneity by study design (cohort and RCT), data collection (prospective and retrospective), region (Europe, North America, Australia, and Asia), and follow up period (1 year, 6 months, 3 months, and 6 weeks).

4.1. Study Selection

In this systemic review and meta-analysis, a total of 392 articles were identified through search from Cochrane reviews, The Virtual Health Library, National Library of Medicine PubMed, and Google Scholar databases. After removing duplicates, 369 articles remained for further screening. The remaining articles were screened by their title based on which 294 articles were excluded. Abstract of the remaining 75 articles were evaluated for eligibility and, 21 studies were identified for full study evaluation (Supporting file 2). Five of the studies had no comparative group: one compared paramedian approach with median approach for spinal anesthesia without assessing the incidence of back pain in the absence of spinal anesthesia [21], another study assessed the factors associated with the incidence of back pain after delivery with epidural anesthesia [22], one was a study comparing spinal anesthesia performed versus 22-gauge and 25-gauge whitacre needles with 26-gauge quincke needles [23], another was comparison of 26-gauge Atraucan and 26-gauge Quincke Spinal Needles [24] and the last one assessed the prevalence of back pain in mothers who delivered under epidural and spinal anesthesia [25]. Two of the articles were excluded for short duration of follow up period 14 days [7] and 24 h [26], which were not adequate to assess persistent back pain. The last study excluded was a randomized control study that did not meet the required quality to be included [27].

From literature citation search, 11 studies were identified, after review 2 studies met the inclusion criteria and were included in the study. In the end, 15 studies were identified to be included in the final analysis Figure 1. The quality assessment showed that the studies were medium and high quality (Supporting file 1).

4.2. Characteristics of Included Studies

In this systematic review and meta-analysis, a total of 57711 parturient were included from the 15 studies selected, with sample size ranging from 198 to 40,057. Four of the studies were randomized controlled trials and 11 were cohort studies with both retrospective and prospective data collection methods. The publication dates of the studies were between 1990 and 2022.

Two of the studies were conducted in Asia (Pakistan and Taiwan), two of the studies were from North America (Canada and USA), one was conducted in Australia, and 10 of the studies were conducted in Europe (Turkey, England, Scotland, Sweden, and Lithuania) Table 1.

4.3. Meta-Analysis

15810 participants who delivered under neuraxial anesthesia developed back pain. The incidence of back pain after delivery in the neuraxial group of the included studies ranges from 3.9% [14] to 33.1% [18]. Based on the 15 studies included in this meta-analysis, the pooled OR according to random effect Restricted maximum—likelihood model was 1.2 (95% CI (0.77–1.86)) with p value = 0.43 Figure 2. There was a significant heterogeneity with I² = 97.76%, H² = 44.58, and Q statistics p value = 0.001 Figure 2.

A cumulative meta-analysis of the studies based on the year of publication was conducted to assess the pattern of relationship between postpartum back pain and neuraxial anesthesia over time Figure 3. Over the years, the relationship between neuraxial anesthesia and postpartum back pain has shifted from statistically significant relationship to no association.

4.4. Assessing Heterogeneity and Publication Bias

The studies showed significant heterogeneity and inconsistency using Cochrane Q test statistic (Q test p = 0.001) and I² = 97.76% which was substantial heterogeneity and an indicative for using random-effects model.

The possible small study effect across the studies was observed by a funnel plot and nonparametric rank correlation (Begg) test. The funnel plot and Begg’s test indicated small study effect was observed (Begg’s regression tests p values = 0.1494) Figure 4. The symmetry of the funnel plot also indicated that there was small study effect.

4.5. Subgroup Analysis

Subgroup analysis was conducted to diagnosis the possible source of heterogeneity by data collection method Figure 5, study design Figure 6, follow up Figure 7, and region Figure 8. The subgroup analysis showed that one possible cause for heterogeneity could be study design since the studies conducted with RCT design were homogeneous with I² = 0.00, H² = 1, and Q statistics p value = 0.51. Duration of follow up subgroup analysis may be another source of heterogeneity since the studies that had 6 months follow up were homogeneous with I² = 28.06, H² = 1.39, and Q statistics p value = 0.23. The rest of the variables used for subgroup analysis were unable to identify the source of heterogeneity.

In order to address the significant heterogeneity between studies we decided to remove three studies with a large sample sizes that were shown to be a source of heterogeneity from the Galbraith Plot Figure 9, Chia et al. (sample size = 40,957) [15] and MacArthur et al. (sample size = 11,701) [10], and MacLeod et al. (sample size = 984) [6]. This strategy resulted in a significant reduction in heterogeneity with I² = 56.7%, H² = 2.16, and Q statistics p value = 0.02, which was still a moderate heterogeneity. The pooled OR according to random effect Restricted maximum—likelihood model for the rest of the 12 studies was 0.98 (95% CI (0.78–1.24) p value = 0.87 Figure 10.