Age- and Level-Dependence of Fatty Infiltration in Lumbar Paravertebral Muscles of Healthy Volunteers

BACKGROUND AND PURPOSE: Normative age-related decline in paravertebral muscle quality is important for reference to disease and risk identification in patients. We aimed to establish age- and vertebral level–dependence of paravertebral (multifidus and erector spinae) muscle volume and fat content in healthy adult volunteers. MATERIALS AND METHODS: In this prospective study multifidus and erector spinae fat signal fraction and volume at lumbar levels L1–L5 were measured in 80 healthy volunteers (10 women and men per decade, 20–62 years of age) by 2-point Dixon 3T MR imaging. ANOVA with post hoc Bonferroni correction compared fat signal fraction and volume among subgroups. Pearson and Spearman analysis were used for correlations (P < .05). RESULTS: Fat signal fraction was higher in women (17.8% ± 10.7%) than men (14.7% ± 7.8%; P < .001) and increased with age. Multifidus and erector spinae volume was lower in women (565.4 ± 83.8 cm3) than in men (811.6 ± 98.9 cm3; P < .001) and was age-independent. No differences in fat signal fraction were shown between the right and left paravertebral muscles or among the L1, L2, and L3 lumbar levels. The fat signal fraction was highest at L5 (women, 31.9% ± 9.3%; men, 25.7% ± 8.0%; P < .001). The fat signal fraction at L4 correlated best with total lumbar fat signal fraction (women, r = 0.95; men, r = 0.92, P < .001). Total fat signal fraction was higher in the multifidus compared with erector spinae muscles at L1–L4 for both sexes (P < .001). CONCLUSIONS: Lumbar paravertebral muscle fat content increases with aging, independent of volume, in healthy volunteers 20–62 years of age. Women, low lumbar levels, and the multifidus muscle are most affected. Further study examining younger and older subjects and the functional impact of fatty infiltrated paravertebral muscles are warranted.

M R imaging is the criterion standard for evaluating the size and structure of soft-aqueous skeletal muscles. 1,2 While T1WI is commonly used for qualitative assessment of muscle fat infiltration (MFI), 3,4 chemical-shift-based imaging sequences allow quantification, [5][6][7] which correlates with clinical symptoms 3,8 and histology. 2,9 The Dixon technique is a robust chemical-shift imaging application producing water-and fat-only images from dual-echo ac-quisitions. 10,11 Excellent accuracy for MFI quantification is shown in comparison with T1WI, 12 spectroscopy, 5 and histology in different animal species. 9 Accordingly, Dixon MR imaging was used for evaluating muscle fat content in several clinical studies including patients with low back pain (LBP), 5,13 acute-to-chronic whiplash, 14 and neuromuscular disorders. 15 Quantification of degeneration (fat infiltration and atrophy) of lumbar paravertebral muscles has attracted interest in understanding their biologic influence on persistent LBP. Atrophy and fat infiltrates are identified in patients [16][17][18][19][20][21] and following experimentally induced lesions in a porcine model. 22 However, human studies describing lumbar MFI report inconsistent findings: An association with LBP was demonstrated in some 18,[23][24][25][26] but not in others. 13,16 One explanation for discrepant findings is the influence of age on muscle composition. 17,18,[27][28][29] Spinal degeneration is known to occur early and increasingly throughout the life span, 30 yet to our knowledge, no study has assessed age-related morphologic changes to lumbar paravertebral muscles in healthy volunteers, which would provide a crucial supplement for future comparative studies.
We sought to quantify lumbar paravertebral muscle volume and fat content by 2-point Dixon MR imaging in healthy adult volunteers spanning 4 decades of life, and we aimed to establish an age-and level-dependent reference of lumbar paravertebral muscle volume and fat content as a reflection of natural aging history. We hypothesize greater MFI with age for both sexes and lumbar MFI level-dependence with an increasing craniocaudal trend.

MATERIALS AND METHODS
This study was approved by the University Hospital Zurich institutional review board, participants provided written informed consent, and no authors had financial interests.
Participants were recruited via advertisements on the intranet and clinical trials Web site of the University Hospital Zurich. Volunteers were screened by using the "Swiss Olympics' First Sports Medicine Interview" health survey, including comprehensive questions determining medical history, handand leg-dominance, weight change (yes/no), smoker (yes/no), alcohol and drug use, and exercise (moderate/high or low). Musculoskeletal pain or injury was determined by the following question: "Do you currently or have you had injuries/ symptoms/operations of the musculoskeletal system?" In response to this question, 16 body parts (including low back), plus an option of "other," were marked yes or no; only responders answering no for all 16 body parts were included.
Subjects progressed to clinical examination when the selfreported height-and weight-registered body mass index (BMI) was between 18 and 26 kg/m 2 by using the following formula: Weight/Height 2 (kg/m 2 ). 31 Exclusion criteria were the following: contraindications for MR imaging (claustrophobia, metal, pacemaker, pregnancy); previous arthrodesis surgery; systemic diseases (chronic obstructive pulmonary diseases, diabetes, metabolic diseases, rheumatologic disorders, tumors, chronic pain syndrome); vascular problems (coronary heart disease, peripheral vascular disease); and alcohol or drug abuse.

Clinical Examination Pre-MR Imaging
Barefoot standing height was measured with a wall-mounted stadiometer (nearest 0.1 cm); bodyweight (kilograms) was collected via a standard scale; and BMI was derived from these 2 parameters. Two bioelectric impedance instruments were used to reflect body fat: Tanita UM-081 (foot sensor pads on a standing scale; Tanita, Arlington Heights, Illinois) (BIA foot ), and Omron BF300 (hand sensor pads with a hand-held device; Omron Healthcare, Kyoto, Japan) (BIA hand ). Leg dominance was confirmed with a balance recovery test. 32

Imaging Acquisition
Whole-body scans were performed in a 3T MR imaging scanner (Ingenia; Philips Healthcare, Best, the Netherlands). A table-integrated 15-element dS head coil, automatically centering on imaged anatomy; a 16-channel posterior coil; and two 16-channel anterior coils were used for signal reception; the dual-transmit body coil of the scanner was used for radiofrequency transmission. Subjects were supine (arms alongside the body).

Image Analysis
Image analysis was performed with dedicated software providing semiautomatic segmentation with linear interpolation (Myrian; Intrasense, Paris, France). An experienced musculoskeletaltrained reader (R.J.C.), skilled in MR imaging postprocessing and volumetric analyses, segmented muscles at all vertebrae between the superior endplate of L1 to the superior endplate of the sacrum. Each lumbar level was defined from the superior endplate of the upper vertebra to the superior endplate of the vertebra (or sacrum) immediately below, including the intervening disc. We drew left and right paravertebral muscle ROIs every third section of the axial water image, while scout-referencing equivalent fat images. The multifidus (MF), longissimus, and iliocostalis were identified as medial-to-lateral paravertebral muscles, respectively; the longissimus and iliocostalis muscles were segmented as a single region labeled "erector spinae" (ES); the border between the MF and ES was scouted from the mammillary processes. The software calculated muscle volume and mean signal intensity values of all pixels. ROIs were copied from the water images to fat images of the same sequence; mean signal intensities (Signal water ϩ Signal fat ) within these ROIs were recorded (On-line Figure). Fat signal fraction (FSF) was calculated as FSF (%) ϭ (Signal fat /[Signal water ϩ Signal fat ])ϫ100. 33

Statistical Analysis
Statistical analyses were performed by using commercial software (GraphPad Prism; GraphPad Software, San Diego, California; and SPSS, Version 20; IBM, Armonk, New York). Mean (ϮSD) FSF and volume for MF, ES, and MF ϩ ES were calculated for each subject. The Kolmogorov-Smirnov test verified normal distribution. The ANOVA with post hoc Bonferroni correction compared FSF and volume values among subgroups. Except for the comparison between the right and left sides as part of ANOVA, FSF and volume values of both sides were combined for statistical analysis (On-line Figure). Distributions of leg and hand dominance were compared by using the 2 test. Correlations between linear parameters (FSF, age, volume, BMI, BIA foot , and BIA hand ) were assessed with the Pearson correlation coefficient (r). Correlations of FSF and BMI with categoric variables (weight change, diet, smoking, alcohol, leg/hand dominance) were evaluated by using the Spearman r. Correlations between 1.0 and 0.5 were considered strong; 0.5-0.3, moderate; and 0.3-0.1, weak. Statistical significance was P Ͻ .05.

RESULTS
Descriptive statistics (mean Ϯ SD, range) of demographics are found in Table 1. No significant differences in paravertebral muscle FSF or volume were found between the right and left sides (P ϭ .30) or hand (P ϭ .52) or leg (P ϭ .29) dominance. FSF (percentage) and volume (cubic centimeters) (mean Ϯ SD) for age group, sex, and lumbar level are presented in Online Tables 1 and 2 and are illustrated in Figs 1 and 2.

Sex
FSF for MF and ES alone and combined was higher in women than in men (P Ͻ .001), whereas the volume for MF and ES alone and combined was higher in men (P Ͻ .001).

Muscles
Total FSF (L1-L5) was higher in the MF than in ES (both sexes, P Ͻ .001). For MF and ES separately, FSF (L1-L5) was higher in women (P Ͻ .001). According to muscle and level, FSF was higher in the MF thanintheESatL1,L2,L3,andL4(bothsexes,PϽ.001each)andhigher in the ES at L5 in women (P Ͻ .05) (On-line Table 1 and Fig 1).

Associations
Associations between variables are presented in Table 2.
No significant correlation occurred between BMI and FSF. A strong correlation occurred between BMI and BIA foot or BIA hand ( Table 2).
No association was seen between FSF or BMI and consumption-related parameters (weight change, smoking, alcohol/drug use, exercise) or hand/leg dominance (data not shown).

DISCUSSION
We quantified lumbar paravertebral muscle volume and fat content by using 2-point Dixon whole-body MR imaging of 80 healthy adult volunteers (20 -62 years of age) and showed that the lumbar paravertebral muscle fat increased with age, independent of volume. Women, low lumbar levels, and multifidus muscles were most affected; these findings align with degenerative features of the spinal column that are highly prevalent in asymptomatic individuals. 30 Lumbar paravertebral muscle degeneration occurs in LBP [16][17][18][19][20]24 and in response to induced lesions, 22 17,24,25 In disagreement, Fortin et al 18 found no correlation with age and FSF cross-sectional areas at L3-L4 or L5-S1 derived from T2WI in men 35-69 years of age. This likely reflects methodologic differences and may relate to their lack of cases representing the 20-to 30-year age group, further highlighting the need for consistency in quantifying paravertebral muscle quality.

FIG 2.
Age-group-averaged fat signal fraction (percentage) and muscle volume for both sexes. Multifidus and erector spinae FSFs and muscle volumes are given separately and combined (MF ϩ ES); FSF is shown as a volume-weighted average. FSF and volumes are represented by the sum of both sides (left ϩ right). Significant differences of the means between age groups are indicated by an asterisk (P Ͻ .05) and triple asterisks (P Ͻ .001). FSF was higher in women than in men at each age group, lumbar level, and muscle (MF ϩ ES, P Ͻ .001; MF, P Ͻ .01; ES, P Ͻ .01). Volume was lower in women than in men as indicated by a dagger (P Ͻ .01) and double daggers (P Ͻ .001). We found an age-related increase of MF and ES MFI for both sexes, suggesting progressive worsening in muscle quality, even in healthy individuals. Male participants in each older age group had significantly more MFI in both MF and ES muscles than the youngest men; the third decade of life is an effective baseline for paravertebral muscle degeneration in healthy men. On the contrary, increased MFI in our female participants only occurred in the fifth and sixth decades. These findings may indicate earlier onset of MFI in men and a sex-dependent decline in muscle quality.
We showed an increasing craniocaudal trend for FSF between L1 and L5. Levels L4 -L5 had higher MFI compared with the supradjacent level; the upper lumbar spine showed a trend in women (women, P ϭ .07; men, P ϭ .84). This finding agrees with the longitudinal male population study of Fortin et al, 17 in which L5-S1 had higher MFI than L3-L4. D'Hooge et al 34 showed more MFI bilaterally in subjects with LBP compared with healthy controls at both L4 endplates, yet equivalence at the L3 superior endplate. While D'Hooge et al 34 did not report interlevel comparisons, agreement exists with respect to a caudal trend for increasing MFI. Our difference between sexes in the upper lumbar spine is interesting. We speculate that muscle quality relates to the intrinsic shape and function of the thoracolumbosacral spine, which may differ between sexes. High paravertebral muscle volume occurs in people with greater lordotic angulation. 28 While there is conjecture about sex differences in lumbar curvature, 35 determining spinal shape in relation to MFI could offer clarification.
Plausible explanations behind normative muscle atrophy and MFI exist and include a combination of disuse and denervation secondary to the degenerative cascade and concomitant altered tensile properties of lumbar myofascial and neural tissues. Disuse-related muscle decline purportedly relates to deconditioning, local tensile unload, and altered muscle recruitment. 16,19,22 Paravertebral muscle denervation occurs in asymptomatic individuals, 36 and the multifidus muscle is susceptible to the effects of neural stretching after disc height loss and subsequent listhesis. 37 However, the extent of paraspinal muscle atrophy is not explained by matching denervation; this finding indicates the potential for reversal through activity. 38 We describe no association of FSF with BMI or exercise, perhaps not supporting either theory. Instead, a local disuse mechanism dependent on paravertebral muscle morphology and proximity to the vertebra as shown in the cervical spine 39 might better explain the etiology.
Conflicting with several studies reporting laterality in healthy individuals, 40 those with LBP, 19,20,27,28 and the general population, 17 we agree with another study 18 in showing no differences between sides or hand/leg dominance in our healthy volunteers. Valentin et al 29 described asymmetry in MF and ES MFI in their healthy subjects, yet no asymmetry for ES volume. The discrepancy among studies may be due to methodologic differences relating to the types of measurement techniques (quantitative versus qualitative, MR imaging versus CT/sonography, volume versus cross-sectional area), defined paravertebral ROIs, and study samples.
Our valuable methodologic finding promotes more time-effi-cient data collection by using imaging at L4 to generalize for total lumbar paravertebral muscle fat content. MFI was higher in women than in men at each age group in our series, despite control for BMI in subject selection, resulting in a lower female mean (P ϭ .05). Age-related change to skeletal muscle quality differs for sex-dependence. 41 Generally, men lose more muscle with aging, yet women have greater functional consequences. 41 Therefore, the influence of general body fat on healthy paravertebral MFI cannot be ignored. 29,34 While our study showed no correlation between BMI and FSF, our bioimpedance was strongly associated. Despite the questionable validity of bioimpedance in measuring body fat, 42 further investigation appears warranted in determining the suitability of BMI or bioimpedance in identifying modifiable risks.
While our study has several strengths, it is limited by being cross-sectional, though it is feasible in assessing a wide age range. Identifying a baseline age group from which to reference the natural history of change represents a valuable contribution to the literature to which longitudinal studies can be directed. We only included participants 20 -62 years of age; this age range may not be generalizable to other age groups. Our sample included 10 cases per sex and decade, offering an improvement to the findings in the literature, but was potentially inadequately powered for reference as normative data. Whole-body MRI may not be generalizable to higher resolution clinical scans. Additionally, 2-point Dixon sequences are prone to underestimating FSF; however, we limited the influence by reduction of the flip angle. 5 Our healthy volunteers evidenced declining muscle quality as a normal process of aging from the twenties into mature adulthood. Investigating whether serial decline continues into healthy older adulthood would be valuable. Furthermore, that poorer muscle quality as determined by increased MFI affects muscle function when noncontractile tissue replaces muscle fibers is implied. Whether this speculation is true should be investigated with applied research examining function.

CONCLUSIONS
Lumbar multifidus and erector spinae fat content increased with age in healthy adults 20 -62 years of age, was higher in women than men, and was more prevalent in the low lumbar levels and in the multifidus compared with erector spinae muscles. These muscle-based degenerative features are part of healthy aging and, relating to LBP, should be interpreted within the individual context. Measurements at L4 are surrogates for the whole lumbar spine. Further studies examining adolescents and adults older than 60 years and in applied research examining function may be helpful in directing interventions aimed at reducing muscle degeneration in the lumbar spine.