Fat infiltration of paraspinal muscles as an independent risk for bone nonunion after posterior lumbar interbody fusion

Posterior lumbar interbody fusion (PLIF) is a commonly surgical treatment for degenerative spinal diseases so as to stabilize the motion segment, restore lordosis and correct deformity [1]. However, bone nonunion as a surgical complication can be observed during follow-up. Most recent studies have reported that the rate of bone nonunion ranged from 7 to 20% [1‐3]. The process of bone union can be affected by multiple factors including increased fused level, osteoporosis and obesity [4, 5].

The predictive value of paraspinal muscle morphometry on operative complications has been investigated [6‐9]. Some studies reported that decreased cross-sectional area (CSA) of paraspinal muscles was correlated to bone nonunion rate in patients with lumbar surgery [3, 7]. However, the effect of paraspinal muscles fat infiltration (FI) on bone nonunion remains indistinct. We hypothesized that the patients with higher FI were inclined to occur bone nonunion. This study aimed to examine the relationship between FI of paraspinal muscles and bone nonunion in patients with lumbar spinal stenosis (LSS) after PLIF.

Hospitalized patients undergoing PLIF for LSS between July 2011 and December 2015 were reviewed. Inclusion criteria were: (1) aged ≥50 years, (2) underwent lumbar magnetic resonance imaging (MRI) and lumbar computed tomography (CT) within 3 months before the index surgery, (3) underwent follow-up of ≥12 months. Exclusion criteria were (1) previous spinal surgery, (2) patients with bone tumor, ankylosing spondylitis, diffuse idiopathic skeletal hyperostosis, rheumatoid arthritis, tuberculosis, or secondary osteoporosis, and (3) patients with scoliosis (> 10°).

A total of 351 patients were identified. Among them, 244 patients with PLIF were LSS and 107 were LSS combined with degenerative lumbar spondylolisthesis. Degenerative lumbar spondylolisthesis was defined as displacement of 1 vertebra over subjacent vertebra using Meyerding grading system [10]. All surgical strategies and approaches were discussed and decided before surgery. For PLIF procedures, using the posterior midline approach, meticulous exposure of the spine and posterior decompression fusion and fixation with pedicle screw was performed. After pedicle screws had been implanted, the neural decompression by laminectomy and discectomy was performed. A polyetheretherketone (PEEK) cage packed with autogenous bone was placed into the interbody space for all patients. Posterolateral fusion was also performed simultaneously. The autograft was harvested from decompression. No bone morphogenetic protein has been used in these patients.

Our study showed that the patients who did not achieve union had a higher MF FI and a higher ES FI than those of the patients with union. In Lee et al’s study, it was reported that the union rate decreased as fat content of extensor muscles increased, which was accordant to our findings [17]. However, they only used a semiquantitative scale to quantify the FI and did not investigate the MF and ES separately. Considering that MF is in the deep attaching to the lumbar vertebrae while ES is more superficial spanning more sections of the spine, evaluating them separately is reasonable [18]. Our results were also consistent to Katsu et al’s study that focused on the patients with osteoporotic fractures [19]. They found that FI of MF and ES were both higher in insufficient union group than in union group. Several studies have revealed that increased muscle FI was correlated to poorer muscle strength [20, 21]. Previous study indicated that incremental bending moment transmitted by the internal fixation device would increase the risk of bone nonunion [22]. Consequently, paraspinal muscles with higher FI might be less effective on reducing the bending moment. Besides, desired paraspinal musculature could provide important vascular ingrowth into the fusion site and accelerate the bone healing [23]. It is suggested that severe muscular degeneration might impede this process.

Of note, multifactor analysis demonstrated that FI of MF rather than ES, had an effect on nonunion. Liu et al. investigated 118 LSS patients and found that the postoperative improvement of Oswestry Disability Index (ODI) was significantly less in MF FI ≥ 25% group than in MF FI < 25% group [24]. Besides, Hong et al. found that MF FI also contributed to superior clinical outcomes including less improvement in ODI, greater postoperative pain and higher reoperation rate [25]. As MF is the innermost and largest one of the paraspinal muscles and provides two-thirds of spinal segmental stability [26], MF might have a more remarkable effect on clinical outcomes compared to ES.

Our analysis showed that the nonunion group had a greater MF atrophy than that of union group, but no statistical significance was seen in multivariate analysis. Choi et al’s study reported that CSA of MF, ES and PS in the nonunion group were all smaller than those in the union group [7]. Furthermore, they found that only PS TCSA was correlated to fusion rate in multivariate analysis. The reason why the TCSA did not show an arresting effect on bone nonunion might be that the relationship between TCSA and muscle strength was not as significant as that of FI. A study demonstrated that FI of paraspinal muscles, not CSA, remained a significant predictor of extensor strength in multivariate regression [20]. Our study indicated that surgeons should pay more attention to the FI of paraspinal muscles rather than atrophy when evaluating the risk of bone nonunion preoperatively.

Multivariable analysis showed that the number of levels fused significantly affected the incidence of nonunion. In previous studies, the number of fusion levels was considered to be a crucial factor in achieving solid fusion in degenerative lumbar diseases [17, 22]. In view of the possible correlation between fusion length and the degree of preoperative FI, we performed a subgroup analysis. The results exhibited that the effect of FI only existed in the patients with one or two fused levels, not in the patients with three or more fused levels. We speculated that in the patients with long-segment fusion, the fusion length had a more notable effect on the bone nonunion over paraspinal muscles FI, hence a severe fatty degeneration might not notably increase the risk of nonunion. Nevertheless, in the patients with shorter fused levels, FI of paraspinal muscles began to take effect.

In addition, we found that the patients with fusion to S1 had a higher rate of bone nonunion, which was consistent with previous studies [17, 22, 27]. In subgroup analysis, the significant difference of FI between the union group and the nonunion group only existed in the patients with lumbosacral fusion, not in the patients without lumbosacral fusion. It could be interpreted by that as a great mechanical load could be applied to the sacrum in lumbosacral fusion, patients need stronger paraspinal muscles to counteract this negative effect [17]. Once patients have a higher FI of muscles preoperatively, lumbosacral fusion will highlight the effect of muscles and then the risk of bone nonunion will increase.

Our findings indicated that surgeons should pay attention to the FI of paraspinal muscles when performing posterior surgery for patients who need short-segment fusion or to extend fusion to S1. In the above cases, preventive measures such as the use of materials to facilitate bone grafting or screws with greater fixation strength should be considered.

We recognize some limitations in the study. First, there were no postoperative MRIs to evaluate the condition of muscle injury during the operation, which might reduce the predictive value of paraspinal muscles on bone nonunion. Besides, we did not perform CT to evaluate the fusion status which may reduce the reliability of bone nonunion. While in our study, 2 observers evaluated union independently with strict criteria for defining nonunion in order to increase the accuracy. Third, the heterogeneity of cases and the small number of cases for subgroup analysis might produce bias. In addition, we have not taken the size and length of the screw into account, which might be related to bone healing [28]. Moreover, it may be impractical to actually measure FI in clinical practice since automated measurement software is not yet available.

This is the first study focusing on the prognostic value of back muscles FI to predict bone nonunion after PLIF. We demonstrated that higher fatty degeneration of MF was an independent factor of nonunion. Furtherly, the effect of FI only existed in the patients with one or two fused levels, not in the patients with three or more fused levels. Besides, the significant difference of FI between the union group and nonunion groups only existed in the patients with lumbosacral fusion, not in the patients without lumbosacral fusion. In cases with higher MF FI during preoperative evaluation, we considered that more rigid fixation or more graft bone might be necessary.