Impact of lumbar spinal canal stenosis on paravertebral muscle condition: an observational study in Japan

Ethics

Written general consent for the use of clinical data for scientific purposes was obtained from each patient. Institutional Review Board approval or approval from any similar entity was not required because of the retrospective use of anonymized data.

Study population

This retrospective study included 78 patients (51 men and 27 women) who underwent posterior decompression or posterior decompression with fusion for LSS. The mean age at surgery was 71.3±7.4 years. The mean height was 161.8±9.8 cm, the mean weight was 63.0±13.0 kg, and the mean body mass index (BMI) was 23.9±4.0 kg/cm². Table 1 presents the age distribution.

Table 1

Profile of 78 patients

Inclusion and exclusion criteria

The inclusion criteria were as follows: (1) Patients who underwent posterior decompression or posterior decompression with fusion for LSS and were resistant to comprehensive conservative treatment in our department between 2015 and 2018. (2) Preoperative magnetic resonance imaging (MRI) was performed using a GE medical system (General Electric Company, Chicago, IL, USA), SIEMENS (Siemens Healthineers AG, Erlangen, Germany), or TOSHIBA_MEC scanner (Toshiba Medical Systems Corp., Otawara, Japan).

Patients were excluded if they had a history of lumbar surgery, lumbar vertebral fracture, neuromuscular disease, accompanying neurological disorders originating from the cervical spine or thoracic spine, degenerative lumbar scoliosis with a Cobb angle >10°, or sagittal imbalance (sagittal vertical axis [SVA] >4 cm) to eliminate confounding effects unrelated to spinal canal stenosis.

Measurements of Paraspinal muscles

Quantitative image analysis was performed using ImageJ ver. 1.53a (National Institutes of Health, Bethesda, MD, USA). The total muscle cross-sectional area (TMCSA), muscle cross-sectional area (MCSA), fat cross-sectional area (FCSA), and fat percentage (FP) of the multifidus, longissimus, and psoas muscles at each disc level were measured on preoperative axial T2-weighted images from T12 to S1 (Fig. 1). The TMCSA was obtained by manually outlining each muscle’s boundary. For the longissimus multis and longissimus dorsi, the cross-sectional area (CSA) of the fat component was determined using the “binarization (Auto)” function, and the resulting value was recorded as the FCSA, as previously described [11] (Fig. 2).

Fig. 1

The measurement of the total muscle cross-sectional areas (TMCSA) on axial T2-weighted magnetic resonance images. It was measured by manually drawing regions of interest around the boundaries of each muscle. The areas labeled 1 and 7 in the figure correspond to the multifidus muscle, while the areas labeled 2 and 8 correspond to the longissimus muscle. The total cross-sectional area of the left and right multifidus and longissimus muscles, including the fat, is defined as TMCSA.

Fig. 2

Assessment of fatty muscle infiltration of the paraspinal muscles of the lower spine was conducted using ‘binarization’ as derived from ImageJ software. Images were converted to black-and-white based on differences in signal intensity, and the area with high signal intensity was calculated and defined as the fat cross-sectional area.

The MCSA was calculated as TMCSA minus FCSA, and FP was calculated as FCSA/TMCSA. The iliocostal muscles were excluded from the study due to the limited number of images with adequate visualization. For the psoas major, only TMCSA was measured because the fat boundary was indistinct and could not be reliably extracted. The CSA of the spinal canal at each disc level and sagittal spinal alignment were also measured. Both sides were analyzed using ImageJ ver. 1.53a (National Institutes of Health), and mean values were used for statistical analysis.

Evaluation

Using MRI horizontal section images, the degree of spinal canal stenosis at each level (L1/2–L5/S1) was classified into the following three grades according to the grading system by Ogura et al. [12]: (1) Grade 1 (none–mild): No compression or mild reduction in canal size with compression of the disc (or end plate); the anterior part of the dural canal is convex or straight, and the dural sac is circular or oval. (2) Grade 2 (moderate): Reduced canal size with compression from the disc (or end plate) and ligamentum flavum; the anterior part of the dural canal is concave, and the dural sac is triangular or semicircular, or more deformed than grade 1. (3) Grade 3 (severe): Minimal cerebrospinal fluid or visible canal space.

Patients were divided into two groups based on the number of stenotic levels: (1) the single stenosis group, comprising patients with grade 2 or 3 stenosis at a single level; and (2) the multiple stenosis groups, comprising patients with grade 2 or 3 stenosis at two or more levels. Additionally, patients were categorized into three subgroups based on the anatomical level of their stenosis: (1) the upper lumbar group (L3/4 or above), (2) the lower lumbar group (L4/5 or below), and (3) the extensive group (stenosis involving both L3/4 or above and L4/5 or below).

Statistical analysis

To adjust for individual differences, muscle area and FP were analyzed using a multiple regression model with sex, age, height, and weight as explanatory variables. Comparisons between the two groups categorized by the number of stenosis were conducted using the Student t-test. Tukey’s honestly significant difference test was used to compare the three groups categorized by the level of stenosis. A p-value <0.05 was considered indicative of statistical significance. The Wilcoxon’s rank-sum test was used as a nonparametric test for comparisons of Japanese Orthopedic Association Back Pain Evaluation Questionnaire (JOABPEQ) scores. To ensure measurement reliability, inter- and intraobserver agreements for MRI-based CSA measurements and stenosis grading were evaluated using intraclass correlation coefficients (ICC). Ten images were measured by the same observer and by another observer 2 weeks after the initial evaluation. The ICC values were ICC (1,1)=0.9952 and ICC (2,1)=0.9873, indicating excellent reliability. Statistical analyses were performed using JMP ver. 16.0.0 (SAS Institute Inc., Cary, NC, USA; 1989–2021).

Patient characteristics

Patient characteristics, including age, sex, height, weight, and spinal alignment parameters (thoracic kyphosis [TK], thoracolumbar kyphosis [TLK], lumbar lordosis [LL], SVA), as well as preoperative Roland-Morris Disability Questionnaire (RDQ), Japanese Orthopedic Association (JOA) score, and JOABPEQ score, were compared among the groups (Table 2). The JOABPEQ is a disease-specific instrument to assess low back pain-related quality of life. It consists of 25 items covering five subscales: low back pain (four items), lumbar function (six items), walking ability (five items), social function (four items), and mental health (six items). Each subscale is scored from 0 to 100, with higher scores indicating better function or condition [13].

Table 2

Patients background

When comparing the upper lumbar and lower lumbar groups, LL was significantly smaller in the upper lumbar group (24.0°±2.5°) than in the lower lumbar group (34.9°±1.9°). SVA was significantly larger in the upper lumbar group (72.2±9.5 cm) than in the lower lumbar group (43.7±7.2 cm). No other parameters showed significant differences between the groups.

Comparison of paraspinal muscle measurements between the multiple stenosis and single stenosis groups

The results of TMCSA, MCSA, and FP for each muscle are summarized in Table 3. The total MCSA of the multifidus and longissimus muscles was significantly smaller in the multiple stenosis group compared to the single stenosis group (single 56.5±2.1 cm² vs. multiple 51.0±1.5 cm², p=0.0384). This tendency was consistent across the L3/4–L5/S1 levels. The FP of the multifidus and longissimus muscles was higher in the multiple stenosis group compared to the single stenosis group (multifidus: single 25.5%±1.4% vs. multiple 30.3%±1.0%, p=0.0081; longissimus: single 14.3%±0.9% vs. multiple 17.2%±0.7%, p=0.0123), and this pattern was consistent at all levels of L1/2 to L5/S1. At the L4/5 level, the MCSA of the multifidus, longissimus, and their combined value were all significantly smaller in the multiple stenosis group (multifidus: single 6.3±0.3 cm² vs. multiple 5.6±0.2 cm², p=0.0293; longissimus: single 4.7±0.2 cm² vs. multiple 4.2±0.1 cm², p=0.0283; multifidus+longissimus: single 11.0±0.4 cm² vs. multiple 9.8±0.3 cm², p=0.0215). At the L3/4 level, the FP of the multifidus, longissimus, and their combined value was significantly higher in the multiple stenosis group (multifidus: single 23.3%±1.6% vs. multiple 27.8%±1.2%, p=0.0271; longissimus: single 12.5%±1.1% vs. multiple 17.1%±0.8%, p=0.0007; multifidus+longissimus: single 17.3%±1.3% vs. multiple 22.1%±0.9%, p=0.0031). For the psoas major muscle, the TMCSA at the L1/2 level was significantly smaller in the multiple stenosis group (single 3.5±0.1 cm² vs. multiple 3.0±0.1 cm², p=0.0003), while no significant differences were observed at other levels.

Table 3

Comparison of paraspinal muscle measurements between the multiple stenosis group and the single stenosis group

Comparison of paraspinal muscle measurements between the upper-level, lower-level, and extensive groups

The results of MCSA and FP of each muscle are shown in Tables 4 and 5. When comparing groups based on the level of stenosis, there were no significant differences in the total MCSA of the multifidus and longissimus between the upper lumbar, lower lumbar, and extensive groups. This trend was consistent across all levels from L1/2 to L5/S1. At the L3/4 level, the FP of the multifidus and longissimus muscles was lower in the upper lumbar group than in both the lower lumbar and extensive groups (multifidus+longissimus: upper 16.8%±1.5% vs. lower 21.4%±1.1%, p=0.0452; upper 16.8%±1.5% vs. extensive 22.2%±1.4%, p=0.0244). No significant differences were observed at any other level.

Table 4

Comparison of MCSA between the upper-level group, the lower-level group, and the extensive group

Table 5

Comparison of FP between the upper-level group, the lower-level group, and the extensive group

Correlation between JOA score and paraspinal muscle

We next examined the correlation between JOA score and the total MCSA of the multifidus and longissimus muscles. Spearman’s rank correlation analysis revealed a significant positive correlation between the JOA score and the combined MCSA of the multifidus and longissimus muscles (p=0.0010, r=0.3665). In contrast, a significant negative correlation was observed between the JOA score and the FP of these muscles (p<0.0001, r=−0.4307).

Atrophy of paraspinal muscles

In this study, multiple stenosis was significantly associated with paraspinal muscle atrophy. One possible mechanism is neurogenic muscle atrophy. Paraspinal muscle atrophy is believed to result from entrapment neuropathy caused by chronic compression of the spinal cord or cauda equina, leading to neurogenic degeneration of the innervated muscles [14,15]. Consequently, paraspinal muscle atrophy tends to be more pronounced in patients with multiple nerve root compressions due to multilevel stenosis. The paraspinal muscles, which attach to a single lumbar spinous process, typically originate from three lower transverse processes and are innervated segmentally by the posterior and medial branches at the same level as the spinous process [16]. Therefore, as the numbers of stenotic levels increases, a greater number of innervating nerves are affected, which may result in more extensive atrophy of the paraspinal muscles.

Another contributing factor may be disuse muscle atrophy. This mechanism involves the differentiation of fibro/adipogenic progenitors (mesenchymal stromal cells) into adipocytes as a result of reduced muscle activity. Such inactivity often arises from pain-related postural limitations and decreased mobility in daily life [17].

Accordingly, patients with multiple stenotic levels may experience more severe pain, further restricting movement and accelerating disuse atrophy. To explore the relationship between the number or level of stenoses and pain severity, functional assessments such as the RDQ and JOA score are useful tools [18]. In the present study, no statistically significant differences were observed in RDQ, JOA score, or JOABPEQ between single and multiple stenosis groups, or among the groups categorized by stenosis level. This may indicate that the number or anatomical level of stenosis is not directly associated with clinical symptom severity in patients with LSS. However, given the limited sample size and interindividual variability in symptom presentation, the study may not have had sufficient power to detect subtle differences. We observed a positive correlation between JOA score and paraspinal muscle area, and a negative correlation between JOA score and FP. In contrast, no significant associations were detected with RDQ or JOABPEQ. This discrepancy likely reflects the multidimensional nature of these assessment tools, as RDQ and JOABPEQ encompass broader domains such as pain, social functioning, and quality of life, which are not solely determined by structural changes in paraspinal muscles. Moreover, confounding factors such as psychological status, comorbidities, and physical activity levels may have attenuated the observed associations between muscle condition and patient-reported outcomes. Collectively, these findings suggest that while paraspinal muscle degeneration contributes to motor dysfunction in LSS, its relationship with broader symptomatology is likely influenced by multiple interacting variables.

Although scores for some JOABPEQ subscales, such as walking ability and mental health, showed numerical differences between subgroups, these did not reach statistical significance. This may be attributed to the limited sample size and high variability in patient-reported outcomes. Nevertheless, these numerical trends may indicate subtle clinical differences that could become significant in a larger cohort.

In summary, it remains unclear whether the decline in JOA score is primarily due to neural compression or secondary degeneration of the paraspinal muscles. Previous studies have demonstrated that fatty degeneration of paraspinal muscles is independently associated with low back pain and functional disability [19]. The present findings suggest that paraspinal muscle changes may contribute to functional decline independently of neural compression severity. Therefore, it is plausible that both mechanisms—direct neural compression and secondary muscle degeneration—interact to influence clinical outcomes in patients with LSS.

Fat infiltration of paraspinal muscles

Fat infiltration of the paraspinal muscles can occur through various mechanisms, including the replacement of skeletal muscle stem cells with adipocytes due to disuse muscle atrophy. Local inflammation has also been reported to disrupt the balance of cytokines such as interleukin (IL)-1β, tumor necrosis factor-α, and IL-10, promoting adipose tissue accumulation [20]. Handschin and Spiegelman [21] further suggested that physical inactivity and obesity trigger persistent, low-grade systemic inflammation, which contributes to the development of various chronic diseases.

In the present study, fat infiltration of paraspinal muscles was more pronounced in patients with multiple stenoses. Although this study was not designed to directly assess inflammatory activity, chronic neural compression and impaired perfusion associated with multilevel LSS may create a localized proinflammatory milieu. Such low-grade inflammation has been proposed to induce adipogenic differentiation of fibro-adipogenic progenitor cells, thereby contributing to fatty degeneration of muscle in chronic musculoskeletal disorders [22]. These findings suggest that, in addition to disuse and denervation, inflammation-related metabolic alterations may contribute to paraspinal muscle degeneration.

Interestingly, the results indicated that the overall muscle area did not decrease significantly, while the proportion of fat increased. This observation suggests that muscle atrophy may not be caused by fatty infiltration per se, but that fatty replacement may occur as a reactive process secondary to muscle degeneration.

No significant differences in the paraspinal muscle atrophy or fat infiltration were observed among the groups categorized by the level of stenosis. As noted earlier, the paraspinal muscles are innervated by the posterior medial branch at the same level as the spinous process to which they attach [16]. Therefore, neuropathy resulting from stenosis would be expected to affect muscles innervated at or below the stenotic level, implying that cranial-level stenoses, such as those at L1–3, could have a substantial impact. In the present study, however, only a small number of patients presented with stenosis limited to the upper lumbar levels (L1–L3). Most cases in the upper lumbar group had involvement at L3/4 rather than at more cranial levels. Consequently, the analysis may have been underpowered to detect level-specific differences in paraspinal muscle atrophy and fat infiltration. These findings should therefore be interpreted with caution, and future studies with larger samples and greater representation of true upper lumbar stenosis are warranted to clarify potential level-specific effects.

Most previous studies included mixed cohorts of men and women; therefore, sex-specific analyses were not conducted in this study. To minimize the influence of individual variability, multiple regression analysis was conducted with adjustments for age, sex, height, and weight. Although the sample size was limited, subgroup analysis by sex demonstrated similar trends.

A better understanding of the relationships among stenosis severity, paraspinal muscle degeneration, and symptom development, particularly low back pain, may contribute to more informed clinical decision-making regarding surgical indications and postoperative management. Furthermore, evaluating paraspinal muscle condition preoperatively, especially in patients with marked paraspinal muscle atrophy or fat infiltration, may help predict functional recovery and guide individualized rehabilitation strategies or earlier surgical intervention.

Limitations

This study has several limitations. First, the absence of a nonsurgical or age-matched healthy control group limits the ability to distinguish paraspinal muscle changes specifically attributable to LSS from those related to normal aging. Because paraspinal muscle atrophy and fat infiltration can also occur as part of the aging process, future studies should include age- and sex-matched individuals with low back pain but without radiological evidence of LSS. Incorporating such control groups and adjusting for confounding variables such as age, sex, BMI, physical activity, and comorbidities will improve the accuracy and generalizability of future findings. Differentiating LSS-related muscle degeneration from age-associated sarcopenia remains a major clinical challenge when determining surgical indications and rehabilitation strategies. Establishing multicenter MRI databases that include asymptomatic volunteers or nonsurgical low back pain patients could enable more robust age- and sex-matched comparisons.

Second, this study included patients referred from multiple healthcare facilities, and preoperative MRIs were obtained using scanners from different manufacturers. Variations in imaging parameters may have introduced measurement bias. Standardized MRI protocols and quantitative imaging methods should be used in future studies to enhance reproducibility. The T2-weighted images employed here are useful for detecting denervation changes and fatty infiltration; however, edema and inflammation may also appear as high-signal images, potentially confounding interpretation.

Finally, the FP of the psoas major muscle was not evaluated because its characteristically low-fat content made reliable segmentation using binary thresholding infeasible. Therefore, only the total CSA was measured. Future studies employing more advanced imaging modalities or segmentation methods may allow for more accurate quantification of fat content in the psoas muscle.