Kyphosis and paraspinal muscle composition in older men: a cross-sectional study for the osteoporotic fractures in men (MrOS) research group

The Osteoporotic Fractures in Men (MrOS) cohort consists of 5,994 men aged 65 or older who were recruited from 6 academic medical centers in Birmingham, AL; Minneapolis, MN; Palo Alto, CA; Pittsburgh, PA; Portland, OR; and San Diego, CA [23, 24]. The primary purpose of MrOS has been to describe the risk factors for osteoporotic fractures in older men. Funding for MrOS began in July 1999 and the study is now its 13th year. At baseline, all men were referred for plain thoracic and lumbar radiographs and 3786 men were referred for quantitative computed tomography (QCT) scans of the hips, lumbar spine and abdomen. Abdominal body composition measures were performed for a random sample of 667 participants in the sample of 3786 who had been referred for QCT scans at baseline [25]. The analytic cohort for our study included the 475 randomly selected participants from the MrOS study with both baseline abdominal body composition measures from QCT scans and plain thoracic radiographs.

Approval of the conduct of the MrOS study was obtained from the Institutional Review Board of each of the participating medical centers and written informed consent was obtained from all study participants.

Baseline abdominal QCT scans were used to obtain abdominal body composition measurements of abdominal wall muscle and adipose tissue distribution. Abdominal scans were obtained using a standardized protocol which specified scanning from the mid-L3 to the mid-L5 vertebra at 5-mm slice thickness with participants in the supine position. We completed abdominal body composition measurements on the scan at the L4-L5 intervertebral disc level. Scans were processed by two specialists at Oregon Health and Science University (OHSU) using a standardized protocol as previously described [26]. Briefly, a spline tool was used to trace the outer and inner fascial borders of the abdominal wall musculature comprising 6 bilateral muscle groups - the rectus abdominis, external oblique, internal oblique, transverse abdominis, psoas and paraspinals. The paraspinal muscle group included the muscles on each side of the vertebral body in the lumbar region located lateral to the lumbar vertebral spinous process and posterior to the lumbar vertebral transverse processes comprising the longissimus, iliocostalis and multifidus muscles. Intermuscular adipose tissue (IMAT) included adipose tissue located around muscle bundles within each muscle group (Figure 1). Attenuation ranges used to estimate volumes (cm³) of muscle were 0 to 100 HU [15, 18] and of IMAT were -190 to -30 HU [16, 17]. Thus, for all muscle groups studied, we had three volume variables: skeletal muscle only, IMAT, and total volume which is the sum of skeletal muscle plus IMAT volume. Muscle attenuation (HU) was obtained for the total volume and the skeletal muscle only volume. To account for variability in image attenuation across scanners, the HU in the scans were scaled to the calibration standard for water density (0 mg/cm³). Inter- and intra-rater reliability was monitored throughout image processing with intraclass correlation coefficients (ICC). Final intra-rater ICCs for all measures were ≥0.98 for rater 1 and ≥0.95 for rater 2, and final inter-rater ICCs were all ≥0.96.

The Cobb angle of kyphosis was measured from supine lateral spine radiographs taken during baseline visit 1. We used the modified Cobb method [27] with a fixed cut-off of T4 and T12, largely because T1 to T3 are usually not well visualized on lateral spine films due to the overlying shoulders and scapulae interfering with the projection. Technicians placed 6 points corresponding to the 4 corners of the vertebral body and the midpoints of the endplates on each vertebral body from T4 to T12. From the superior surface of T4 and inferior surface of T12, a computerized digitization program erected perpendicular lines whose intersection was the kyphotic angle (Figure 2). If for any reason T4 or T12 were not visible, the next adjacent visible vertebra (T5 if T4 not visible or T11 if T12 not visible) was used as an alternative. The ICC for Cobb angle of kyphosis has been previously reported 0.99 [27].

Vertebral fractures were adjudicated from baseline lateral lumbar and thoracic spine radiographs based upon a previously developed protocol [28]. The SpineAnalyzer™ (Optasia Medical Ltd., Cheadle, UK) workflow tool was used to automate placement of 6-point morphometric points on each radiograph in order to efficiently identify the vertebral bodies for evaluation. An expert radiologist then graded each vertebra as normal (0), mild (1), moderate (2), severe (3) using the well-established, highly reliable semi-quantitative Genant method [29]. Prevalent vertebral fractures were defined as grade 2 or 3.

Body weight and height were taken at baseline by an examiner using standard equipment, including a Harpenden stadiometer and balance beam scale while wearing indoor street clothing without shoes [23]. Body mass index (BMI) was calculated as weight divided by the square of height in meters (kg/m²). Bone mineral density (BMD) was also measured at baseline in the proximal femur and lumbar spine using DXA measured by Hologic QDR 4500 densitometers.

We built models to control for potential confounding using an empirical approach. We examined the association between an a priori group of potential confounders and Cobb angle and based on the change in the regression coefficient, we used p value <0.1 to select confounders including age, height, vertebral fractures, and total hip bone mineral density (BMD). Loess plots were used initially to determine the shape of the relations of muscle volume, adipose volume and Cobb angle. Because associations were not linear, muscle mass measurements were divided into quintiles, quartiles and tertiles to investigate non-linear trends in kyphosis, but due to the limited sample size, we used tertiles for our analysis. Multivariate linear models were used to investigate the association between the muscle variables and kyphosis using total muscle volume, as well as individual components of the total muscle volume that included the IMAT and skeletal muscle compartments alone. All multivariable analyses controlled for age, height, vertebral fractures, and total hip BMD. These associations were also investigated separately for the total abdominal (rectus abdominis, transversus abdominis, internal obliques and external obliques), psoas and paraspinal muscle groups. We examined these associations among those with no prevalent vertebral fracture (n = 429) and among those with BMI < 30 kg/m² (n = 394). We conducted sensitivity analyses with different adjustment variables as well as an alternate definition of hyperkyphosis >40 degrees.

This study has a number of strengths including that the MrOS cohort includes community-dwelling older men residing in multiple geographic areas of the United States who were not preselected for hyperkyphosis. The kyphosis and the paraspinal muscle measurements were assessed at baseline with high reliability in a subgroup of study subjects. However, this study also has a number of limitations including possible misclassification of muscle measurements depending upon the exact location of the QCT scan slice. The paraspinal muscle group does increase in size closer its attachment sites, therefore our paraspinal muscle volume estimates measured at the L4-L5 intervertebral disc space may be underestimated, as would any association of paraspinal muscle volume and kyphosis found in this study. Another potential limitation is that we could not measure muscle volumes in the thoracic spine due to the CT technology used. However, our results suggest that lumbar paraspinal muscles serve as a proxy for muscle degeneration in the upper spine as well. We also used BMD measurements from the hip even though the muscle and kyphosis measurements were made in the spine. Hip BMD is often used in place of spine BMD because there is less artifact in the hip and the lumbar spine BMD from DXA can be overestimated because of extravertebral calcification (e.g. osteophytes). Hip BMD thus provides a more accurate measurement, particularly in elderly subjects. Moreover, when we replaced hip BMD with volumetric measurements of spine BMD with QCT, there was no appreciable difference in the results. The MrOS population may have degenerative spine conditions that could increase kyphosis and our regression analysis did not include adjustment for these degenerative changes. However, this is beyond the scope of this study. We chose to measure global Cobb angle using T4 and T12 endplates rather than other methods such as the centroid method that is less affected by endplate tilt [45] because the global method using the endplates of T4 and T12 is reliable and one of the most commonly used radiographic methods for measuring thoracic kyphosis. In future studies, we may consider using other methods that reflect changes in the middle of the curve. Finally, supine measurements of Cobb angle of kyphosis underestimate the degree of kyphosis compared to standing measurements, particularly among those with greater degree of kyphosis [27]. However, this would serve to underestimate any effects we found.