Introduction

Spine surgery is a common adjunct treatment for degenerative spinal diseases, which has increased significantly in recent years [1]. However, this is associated with a growth in the frequency of inferior postoperative outcomes [2, 3]. Studies have revealed that the degeneration of paraspinal muscles, which can be generally found among elderly patients, is implicated in multiple degenerative lumbar pathologies [4,5,6].

Currently, the value of paraspinal muscle morphometry on image examination serving as a prognostic factor for several surgical disciplines including metastatic disease, trauma, and fracture is being unearthed [7,8,9]. Multifidus (MF), erector spinae (ES), and psoas major (PS), serving as primary extensor and flexor muscles, are generally quantified by the cross-sectional areas (CSA) and fat infiltration (FI) on magnetic resonance image (MRI). Although abundant work has been carried out to identify potential factors for the prognosis of spine surgery [10, 11], few unequivocal predictive factors related to paraspinal muscle have come to light.

Preoperative assessment of paraspinal muscle morphology may be conducive to identifying patients who tend to have unsatisfactory clinical outcomes and thus making precautionary measures in advance. Two systematic reviews have concluded the degree of preoperative paraspinal muscle degeneration in relation to several complications after spinal surgery [12, 13]. However, both of them could not conduct a meta-analysis because of high heterogeneity. Besides, they did not focus on the patients’ functional status and re-operation after surgery. Moreover, there exist conflicting results on whether paraspinal muscle morphology was associated with clinical outcomes in patients with lumbar surgery [14, 15]. In consequence, our systematic review and meta-analysis primarily aim to elucidate the predictive value of preoperative paraspinal muscle morphology on functional status, symptoms, and re-operation in patients with surgery for degenerative lumbar diseases.

Materials and methods

Search strategy and eligibility criteria

The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement was used to structure this systematic review and meta-analysis. To retrieve interrelated articles, we conducted a search in the following three databases: PubMed, EMBASE, and Web of Science databases through September 2022. All fields were searched for these terms: “paraspinal muscle,” “paravertebral muscle,” “multifidus,” “erector spinae,” or “psoas major”; and “surgery,” “operative,” “clinical outcome,” or “functional status”; and “lumbar” or “lumbosacral.” Two authors assessed all abstracts and titles to rate adherence to review criteria. Inclusion criteria consisted of the following: (1) articles including adults with degenerative lumbar diseases; (2) assessment of any lumbar paraspinal muscle characteristic on MRI or computed tomography preoperatively; (3) assessment of any clinical outcomes after lumbar surgery; (4) analyzed the relationship between preoperative imaging data and postoperative outcomes; (5) articles were published in English. Studies were excluded if they included subjects < 18 years of age; assessed lumbar muscle through nonconventional MRI (such as functional MRI, MRS, and chemical-shift MRI) and ultrasonography; included only postsurgical subjects. Studies included randomized controlled trials, cohort studies, case–control studies, and case series.

Assessment of risk of bias

We used Quality in Prognosis Studies (QUIPS), a widely accepted tool for evaluating the risk of bias in prognostic studies [16]. All articles meeting review criteria were evaluated independently for risk of bias by two authors, with any differences in assessment resolved by discussions until consensus was reached. The QUIPS contained 6 separate domains: study participation, study attrition, prognostic factor measurement, outcome measurement, study confounding, and statistical analysis and reporting. Each domain was judged as low, moderate, or high risk of bias, according to published criteria [17]. The overall risk of bias for an included study was defined as low risk with ≥ 4 low- and no high-risk domains, moderate risk with < 4 low- and no high-risk domains, and high risk with ≥ 1 high-risk domain [18].

Data extraction

Two authors independently extracted the following information from included studies: study design, participant characteristics, details of assessments of preoperative lumbar muscle characteristics, and study results that were relevant to our research question. Any disagreement would be adjudicated by a third author.

Measures and outcomes

The FI and atrophy of paraspinal muscles were evaluated by MRI or computed tomography. The parameters of high FI covered the increased percentage of the fat area or signal intensity of muscles, and the parameters of muscle atrophy included the declined area of total or lean paraspinal muscles. We defined the paraspinal extensor muscle (PEM) group as the integrity of MF and ES [19]. Oswestry disability index (ODI) scores were adopted to evaluate the postoperative functional disability [20]. Visual analogue scale (VAS) and numerical rating scale (NRS) were used to determine the postoperative pain at low back or legs [21]. The rate of revision surgery was also recorded.

For the applicability of synthesis, we have conducted subgroup analyses with a consistently measured paraspinal muscle and its method of morphology measure to reduce the possible heterogeneity among studies. Similar subgroup analyses were conducted in previous studies [19, 22].

Data analysis and levels of evidence

The literature presented data in various forms, including the means with/without standard deviations (SDs) for continuous outcomes, odds ratios (ORs) for dichotomous outcomes, and correlation coefficient (r) in correlation analysis. In view of that, a standardized mean difference (SMD) was calculated from the reported means and SDs [23]. The SDs could also be estimated from t test value or p value with the degree of freedom, if not directly reported [24]. Other different metrics (ORs and correlation coefficients) were transformed into SMDs, for the feasibility of overall comparison among various metrics [25]. The effect size was defined as small, medium, large, or very large with a SMD at 0.10–0.34, 0.35–0.64, 0.65–1.19, or > 1.20 respectively [26].

Meta-analyses were performed to show the weighted effect size for homogenous comparisons when required metrics could be calculated in ≥ 3 studies according to previous studies [27, 28]. The analysis was completed by Revman 5.4.1 and Rstudio software [29]. The forest plots were used to report SMDs and 95% CIs for separate studies and weighted effects. The heterogeneity was evaluated by Cochran’s Q and I2 statistics tests. The fixed-effects model was selected when the heterogeneity was acceptable with p > 0.10 and I2 < 50%; otherwise, a random-effects model was fitted [27]. Publication bias was displayed by funnel plots and Egger’s test [23, 27].

When the meta-analysis was inappropriate due to the lack of metrics reported in < 3 studies, a vote-counting model was conducted according to previous studies, with the direction of effect [30,31,32]. The model followed the Cochrane and Synthesis without meta-analysis (SWiM) reporting guidelines [33, 34]. We performed a qualitative summary of evidence for lumbar muscle characteristics as predictors of postoperative outcomes. The model followed the definition for levels of evidence in previous reviews [35,36,37]: “strong” evidence was defined as consistent findings (≥ 75%), among which ≥ 2 studies were at low risk; “moderate” evidence was defined as consistent findings (≥ 75%), among which one study was at low-risk and ≥ 1 studies were at moderate- or high-risk; “limited” evidence was defined as consistent findings (≥ 75%), among which one study was at low- or moderate-risk or ≥ 2 studies were at high-risk; “conflicting” evidence was defined as inconsistent findings irrespective of study quality. One high-risk study was not considered in the vote-counting model. The effect direction plot displayed the direction of effect in the model [28, 34].

Results

For the review of preoperative paraspinal muscle characteristics and their predictive value in clinical outcomes after spinal surgery, 6917 studies were identified through database searching. After the removal of duplicate records, 3482 studies were screened in our study. Of these articles, 140 were deemed to be eligible for full-text review. Finally, a total of ten articles were included. Of them, five studies with required metrics were included in the meta-analysis. The search flow diagram is shown in Fig. 1. The meta-analyses for the relationships between preoperative MF FI, postoperative ODI, and postoperative back pain were performed in a fixed-effects model since the heterogeneities were both acceptable (p = 0.13, I2 = 47% for ODI; p = 0.43, I2 = 0% for back pain).

Fig. 1
figure 1

Flowchart of the study selection and inclusion process. A total of 6917 articles were identified from databases through September 2022. After the removal of duplicate records, 3482 studies were screened. Of these articles, 140 were eligible for full-text review. A total of 10 articles were included in the review and 5 articles were included in the meta-analysis

Full size image

Study characteristics and risk of bias

Of ten included studies, six studies investigated participants with lumbar disc herniation (LDH, n = 1) [38] or lumbar spinal stenosis (LSS, n = 5) [14, 15, 21, 39, 40], one reported spondylolisthesis [41], and three reported multiple lumbar degenerative diseases [42,43,44]. All articles examined the relationship between preoperative paraspinal muscle and postoperative symptoms and functional status and two articles examined the re-operation. All studies assessed the muscle morphology by MRI. There were five cohort studies, four case series, and one secondary analysis in the randomized controlled trial. After the assessments by QUIPS, there were six, one, and three studies considered to have a low, moderate, and high risk of bias, respectively (Fig. 2 and Supplementary table 1). The characteristics of the included studies are summarized in Table 1. In addition, publication bias was not observed in meta-analyses based on funnel plots and Egger’s test (Supplementary Figs. 1 and 2).

Fig. 2
figure 2

Risk of bias for the included studies. A = study participation, B = study attrition, C = prognostic factor measurement, D = outcome measurement, E = study confounding, F = statistical analysis and reporting, O = overall risk of bias. The overall risk of bias for an included study was defined as low risk with ≥ 4 low- and no high-risk domains, moderate risk with < 4 low- and no high-risk domains, and high risk with ≥ 1 high-risk domains

Full size image
Table 1 Study characteristic of the included studies
Full size table

Postoperative functional status and symptoms

Postoperative ODI

Eight studies investigated the relationship between preoperative paraspinal muscle morphology and postoperative ODI [14, 15, 38,39,40,41,42,43]. A meta-analysis including four studies with sufficient data in the fixed-effects model was performed for comparison in postoperative ODI scores between groups with high or low FI in MF [39, 40, 42, 43]. The meta-analysis in the fixed-effects model revealed that patients with a high grade of preoperative MF FI had higher postoperative ODI scores, compared to those with a low grade of MF FI (SMD = 0.33, 95% CI 0.16–0.50, p = 0.0001; Fig. 3).

Fig. 3
figure 3

Forest plot of postoperative ODI between patients with high and low MF FI for lumbar surgery. Patients with high-grade preoperative MF FI had higher postoperative ODI scores, compared to those with low-grade MF FI (SMD = 0.33, 95% CI 0.16–0.50, p = 0.0001)

Full size image

In the vote-counting model, whether MF atrophy was related to postoperative ODI remained indistinct (Table 2). Zotti et al [14] found that greater tCSA (< 8.5 cm2) of MF at L4-5 predicted a larger improvement in ODI (SMD = 0.85, p = 0.006). Chen et al [41] reported that a higher muscle/vertebra CSA ratio (> 0.60) of MF at L4-5 was associated with more improvement in ODI (SMD = 0.55, p = 0.010). On the contrary, Wang et al [15] found that the lower scores and better improvement in postoperative ODI were not correlated to a higher tCSA of MF at L4 in patients with LSS after PLIF. As for ES, only one single study by Wang et al assessed the prognostic effect of CSA and FI on ODI [15]. Limited evidence revealed that FI of ES could predict the postoperative ODI after lumbar surgery, while atrophy of ES could not (Table 2 and Fig. 4). Moreover, PS thickness could be a possible indicator for change in ODI with limited evidence (Table 2). One study by Verla et al [44] measured the thickness of PS at each segment and revealed that more improvement (> 50%) in ODI was associated with greater psoas muscle thickness at L3-5 (p = 0.017) in patients with multiple degenerative lumbar diseases after PLIF.

Table 2 Levels of evidence for paraspinal muscle characteristics as predictors of postoperative ODI after lumbar surgery in vote counting model
Full size table
Fig. 4
figure 4

Forest plot of postoperative back pain between patients with high and low MF FI for lumbar surgery. Patients with higher preoperative FI of MF had greater postoperative LBP compared to those with low FI of MF (SMD = 0.17, 95%CI 0.02–0.31, p = 0.03)

Full size image

Postoperative pain

Eight studies evaluated the association between postoperative pain and FI of paraspinal muscles [21, 38,39,40,41,42,43,44]. A meta-analysis including five studies with sufficient data in the fixed-effects model was conducted for comparison in low back pain (LBP) between groups with high or low FI in MF [21, 39, 40, 42, 43]. The meta-analysis in the fixed-effects model reported that the patients with higher preoperative FI of MF had greater postoperative LBP compared to those with low FI of MF (SMD = 0.17, 95%CI 0.02–0.31, p = 0.03; Fig. 4).

In the vote counting model, whether FI of MF could predict persistent leg pain remained conflicting (Table 3). Although some studies by Hong et al [38], Liu et al [40], and Chen et al [41] showed that MF FI failed to predict the improvement in VAS for leg pain, Tiago et al [42] uncovered the predictive evidence of lower MF FI for greater relief of self-reported leg pain post surgeries (grades 3–4: 35/43 vs. grades 1–2: 31/32, SMD =  − 1.09, p = 0.04). Han et al [39] also discovered a higher VAS for leg pain in LSS patients with higher MF FI (≥ 25% tCSA: n = 160, mean = 2.8, SD = 2.4) than lower MF FI (< 25% tCSA: n = 160, mean = 2.3, SD = 2.5) after PLIF (SMD = 0.20, p = 0.039). Besides, limited evidence showed that the atrophy of PS could predict persistent LBP (Table 3). Verla et al [44] found that more improvement (> 50%) in VAS was associated with a greater PS thickness at L2-3 (p = 0.032), L3-4 (p = 0.043), and L4-5 (p = 0.022) in patients with multiple degenerative lumbar diseases after PLIF.

Table 3 Levels of evidence for paraspinal muscle characteristics as predictors of postoperative pain after lumbar surgery in vote counting model
Full size table

Re-operation

In the vote counting model, conflicting evidence existed in the prognostic value of MF and ES on the re-operation rate (Table 4). Betz et al [21] quantified the preoperative quality of the PEM group by the degree of fatty degeneration (according to the Goutallier classification) at L3. They found that there was no difference in revision surgery between symptomatic LSS patients who underwent revision surgery with good or bad muscle quality (SMD =  − 0.17, p = 0.55). Whereas Hong et al [38] confirmed that MF atrophy (OR = 2.095, SMD = 0.41, p = 0.01) at the involved level according to a three-scale grading system on MRI was a statistically significant predictor of a need for re-operation in patients with microendoscopic discectomy for single-level LDH.

Table 4 Levels of evidence for paraspinal muscle characteristics as predictors of re-operation after lumbar surgery in vote counting model
Full size table

In our analysis, all of the pieces of evidence in the vote-counting model were shown in an effect direction plot (Fig. 5).

Fig. 5
figure 5

Effect direction plot for vote counting model. Evidence remained conflicting in terms of whether MF atrophy could predict ODI, whether MF FI could predict leg pain, and whether PEM FI could predict revision surgery. Besides, evidence remained limited in terms of whether ES atrophy, ES FI, or PS atrophy could predict ODI, and whether PS atrophy could predict low back pain

Full size image

Discussion

Paraspinal muscle degeneration can be visibly characterized by a muscle atrophy and an increased FI. Multiple studies have considered paraspinal muscle degeneration as a prognostic factor for surgical disciplines [7,8,9]. However, whether paraspinal muscle degeneration is related to a poor improvement of functional status after lumbar surgery is still indistinct. This is the first study that investigated the predictive value of three main back muscles on postoperative functional status and re-operation.

This systematic review and meta-analysis included ten studies providing evidence for relationships between lumbar muscle characteristics and postoperative functional status and symptoms. First, the meta-analysis found an association between MF FI and postoperative ODI. Previous studies have also demonstrated the relationship between MF FI and postoperative complications, which was accordant to our findings [12, 13]. The reason why patients with higher MF FI have poorer ODI might be interpreted by the fact that increased muscle FI is correlated to poorer muscle function and low physical activity [45,46,47,48]. Besides, only two studies assessed ES and PS. As a result, the evidence for ES and PS as prognostic factors is limited.

In terms of postoperative pain, higher MF FI could predict persistent LBP after lumbar surgery in our meta-analysis. Compared to other paraspinal muscles, MF was more closely related to the vertebral plate and spinous process [49]. Thus, the ability of MF to stabilize and control the movement of the lumbar spine may be a crucial factor in the development of symptoms, since those with severer symptoms have lower activity in MF [50]. In addition, our findings suggested that FI, not atrophy, was a good predictor for postoperative outcomes, which would be helpful for clinical decision-making. The reason might be that the relationship between atrophy and muscle function was not as significant as that of FI. A study demonstrated that FI of paraspinal muscles, rather than CSA, remained a significant predictor of extensor strength in multivariate regression [45]. Moreover, we found limited evidence that showed that the atrophy of PS could predict persistent LBP. As PS is an indicator of sarcopenia and correlates to clinical outcomes in several surgeries [7, 9], it may be a potential factor for predicting LBP.

In our study, two articles examined the relationship between preoperative paraspinal muscle and the re-operation. There was conflicting evidence that MF and ES could predict the re-operation rate in lumbar surgery. The implementation of revision surgery may be due to the failure to achieve osseous fusion, complications resulting from surgical implants, adjacent segment diseases, and persistent pain, part of which were associated with paraspinal muscle degeneration [13, 51]. Consequently, considering the small amount of included studies, the relationship between paraspinal muscles and re-operation needs more high-quality studies.

There are some limitations in this systematic review and meta-analysis. First, the amount of included studies was relatively small, which might produce bias. However, the included studies have been more than previous systematic reviews that investigated factors of clinical outcomes in degenerative lumbar spine diseases [19, 23]. Besides, although the funnel plots and Egger’s test indicated no evidence for publication bias in our study, the publication bias assessment based on a small number of included studies might be unreliable. Second, heterogeneity existed in our study, such as diseases and operations. Previous reviews investigating paraspinal muscle degeneration could not perform a meta-analysis due to high heterogeneity [12, 13]. However, in our review, we made subgroup analyses to reduce the heterogeneity (I2 < 50%) and made it possible for meta-analyses.

In conclusion, FI of MF could be a predictive factor of postoperative ODI and LBP. Whereas, for postoperative leg pain, MF FI could not be a predictor with conflicting evidence. Besides, limited evidence was presented for the prognostic effects of ES and PS on postoperative functional status and symptoms. Additionally, there was conflicting evidence that FI of MF and ES could predict the incidence of revision surgery. The results suggested that the assessment of paraspinal muscle degeneration could be a viable method to stratify patients by risk of postoperative functional status and pain to some extent. A rehabilitation approach focused on training MF before and after surgery might improve clinical outcomes. Considering that limited studies were included and conflicting or limited evidence also existed in this study, more research needs to focus on this field for assisting surgeons to predict the prognosis.