The Effects of Age on the Morphometry of the Cervical Spinal Cord and Spinal Column in Adult Rats: An MRI-Based Study

Authors

  • Andrew C. Laing,

    Corresponding author
    1. Injury Biomechanics and Aging Laboratory, Department of Kinesiology, University of Waterloo, 200 University Ave West, Waterloo, Ontario, Canada
    2. International Collaboration on Repair Discoveries (ICORD), University of British Columbia, Vancouver, British Columbia, Canada
    • Correspondence to: Andrew Laing; Department of Kinesiology, University of Waterloo, 200 University Ave, West, Waterloo, ON N2L 3G1, Canada. E-mail: actlaing@uwaterloo.ca

    Search for more papers by this author
  • Elora C. Brenneman,

    1. Injury Biomechanics and Aging Laboratory, Department of Kinesiology, University of Waterloo, 200 University Ave West, Waterloo, Ontario, Canada
    Search for more papers by this author
  • Andrew Yung,

    1. MRI Research Centre, University of British Columbia, Vancouver, British Columbia, Canada
    Search for more papers by this author
  • Jie liu,

    1. International Collaboration on Repair Discoveries (ICORD), University of British Columbia, Vancouver, British Columbia, Canada
    Search for more papers by this author
  • Piotr Kozlowski,

    1. International Collaboration on Repair Discoveries (ICORD), University of British Columbia, Vancouver, British Columbia, Canada
    2. MRI Research Centre, University of British Columbia, Vancouver, British Columbia, Canada
    Search for more papers by this author
  • Thomas Oxland

    1. International Collaboration on Repair Discoveries (ICORD), University of British Columbia, Vancouver, British Columbia, Canada
    2. Orthopaedics and Injury Biomechanics Group, Departments of Orthopaedics and Mechanical Engineering, University of British Columbia, Vancouver, British Columbia, Canada
    Search for more papers by this author

ABSTRACT

Rat models are commonly used to investigate the pathophysiological pathways and treatment outcomes after spinal cord injury (SCI). The high incidence of fall-induced SCI in older adults has created a need for aging models of SCI in rats to investigate potential age-related differences in SCI severity and outcomes. The aims of this study were to determine the influences of age and vertebral level on the geometries of the cervical spinal cord and spinal column in a rat model. Three young (3 months) and three aged (12 months) Fischer 344 rats were imaged in a high field (7 T) small-animal magnetic resonance imaging system. All spinal cord geometry variables (including depth, width, and axial cross-sectional area) and one spinal canal variable (depth) were significantly larger in the aged specimens by an average of 8.1%. There were main effects of vertebral level on all spinal cord variables and four spinal canal variables with values generally larger at C4 as compared to C6 (average increases ranged from 5.7% to 12.9% in spinal cord measures and 5.4% to 6.8% in spinal canal measures). High inter-rater reliability between two measurers was observed with a mean intraclass correlation of 0.921 and percent difference of 0.9% across all variables measured. This study clearly demonstrates that cervical spinal cord geometry changes between the ages of 3 and 12 months in Fischer 344 rats. This information can aid in the planning and interpretation of studies that use a rat model to investigate the influence of age on cervical SCI. Anat Rec, 297:1885–1895, 2014. © 2014 Wiley Periodicals, Inc.

Spinal cord injuries (SCI) are a major burden to the health care system with approximately 12,000 new cases per year in the United States with each case costing upward of $1.5 million over a lifetime (Priebe et al., 2007). Motor vehicle accidents and falls continue to be the two leading causes of injury (NSCISC, 2013). Of substantial concern is the increasing trend of fall-induced SCI in older adults over the age of 65 (Kannus, 2000; O'Connor, 2006; Ho et al., 2007; Kannus et al., 2007). The prevalence of comorbidities prior to SCI and prevalence of secondary complications after SCI are significantly higher for older adults compared to their younger counterparts (McKinley et al., 1999; Kannus et al., 2007). Therefore, investigations into potential age-related differences in SCI severity is warranted to determine whether treatment methods are optimized across the lifespan.

Despite the trend of increasing incidence of SCI in older persons in most developed nations (O'Connor, 2006), there exist only a handful of studies using a rat model (a common approach in SCI research) that have incorporated aging specimens in their investigations. For example, following clip compression, Genovese et al. (2006) found that 18-month old rats (approximately middle-aged (Thurman et al., 1994)) demonstrated significantly higher neutrophil infiltration during the secondary inflammation phase, increased immunoreactivity for nitrotyrosine, and a higher mortality rate than the 3-month-old rats (analogous to a young adult human population). Although statistical analysis was not performed between young and old specimen groups, Chaovipoch et al. (2006) demonstrated that treatment of a midthoracic crush SCI with 17-beta-estradiol decreased locomotor deficits, increased white matter sparing, and reduced gray matter apoptotic cell death. Siegenthaler et al. (2008a) showed that there was an age-associated increase in the area of pathology accompanied with decreases in motor function and bladder recovery for aged (18 months) and geriatric (30 months) rats compared to young rats. Additionally, Sheilds et al. (1999) demonstrated that remyelination following lysolecithin injection occurred at a reduced rate in old versus young rat specimens. These studies provide evidence of the significant effects that age can have on SCI outcomes and lead to questions about the mechanisms underlying these age-related differences.

Age-related differences in the spinal cord and spinal canal have been reported for rat models. Laing et al. (2011) used µCT to characterize age-related morphometric differences in spinal canal dimensions in young (3 months), aged (18 months), and geriatric (30 months) Fisher 344 rats. They observed that variables including spinal canal depth and width, vertebral body depth and height, and canal pinch diameter were significantly larger in the aged compared to young group, but that few significant differences existed between the aged and geriatric rats. Though the use of µCT is a useful tool for measuring spinal canal variables, magnetic resonance imaging (MRI) is the preferred method when examining soft tissues, such as the spinal cord (Chandra et al., 2012). Although MRI has been used to characterize spinal cord geometry in rats (Weber et al., 2006; Scholtes et al., 2008; Byrnes et al., 2010) to date it has not been used to assess potential age-related changes in the cervical spinal cord of rats. Furthermore, no studies have compared spinal cord and spinal canal geometry from the same rat specimens. Such information would assist in interpreting studies that assess age-related differences in tissue damage and functional impairments following SCI.

The objectives of this study were threefold. Our primary aim was to use a custom-developed small-animal MRI data collection and analysis procedure to determine the influence of age and vertebral level on the geometry of the cervical spinal cord in a rat model. The secondary aim was to determine the influence of these factors on spinal canal geometry. Our final aim was to assess the inter- and intrarater reliability of our image analysis approach.

MATERIALS AND METHODS

Specimens

For this investigation, the cervical spine region of six female Fisher 344 rats were examined in vivo using magnetic resonance imaging. Three rats were 3 months of age and were classified as “young” (mean (SD) mass = 182 (4) g). The other three specimens were 12 months of age and comprised the “aged”' group (mass = 230 (25) g). Age group classifications were made on the basis of longevity studies that report 50% and 90% mortality ages of 26 and 33 months, respectively, for female Fisher 344 rats dying of natural causes (Thurman et al., 1994), and correspond to age–group names used in previous rat models of age and SCI (Siegenthaler et al., 2008a,b; Laing et al., 2011).

Image Acquisition

A 7 T small animal MRI scanner (Bruker Biospin, Ettlingen, Germany) was used for image acquisition, using a quadrature surface coil for transmission and reception. The specimens were injected with 50 µL of 40 mM Gadodiamide in the cisterna magna using a stereotactic frame under ketamine/xylazine anesthesia, and left to diffuse for 30–60 min before image acquisition. This contrast agent does not cross the spinal cord–blood barrier, and was used to increase signal intensity in the CSF space (by shortening its T1 relaxation time) to increase contrast for delineation of the spinal cord and canal. All specimens were maintained by isoflurane anesthesia delivered via tooth bar and heater water jacket, and monitored by a rectal temperature probe and a pneumatic respiration sensor. The animals were placed supine on a plastic scanning platform, using surgical tape for moderate restraint to limit motion artifact. A fast spin echo axial sequence was used to acquire axial images with moderate T2 weighting (RARE sequence, echo time = 23.5 ms, repetition time = 5 sec, 256 × 256 matrix size, FOV = 2.56 × 2.56 cm, inplane resolution = 0.1 mm, slice thickness = 0.7 mm, 16 slices, respiratory triggering on, RARE factor = 8, 6 averages, approx scan time = 16 min), starting from the C2 level and progressing 11.2 mm caudally. Sagittal images of the cervical spinal column (7 slices) were also acquired using identical sequence parameters.

The signal acquired with MRI is dependent on the physical properties of tissue, with the “T1 relaxation time” being the tissue property most relevant for our contrast enhancement technique. The T1 relaxation time describes how quickly the tissue magnetization recovers from excitation, with a lower T1 relaxation time resulting in a larger amount of signal due to faster recovery. At 7 T magnetic field strength, the cord T1 is on the order of 1.5 sec (Meyerand et al., 1998) whereas the CSF T1 is greater than 4 sec (Bluestein et al., 2012). The unenhanced CSF signal is therefore attenuated relative to the cord signal, often to a level which makes it indistinguishable from the spinal column bone. The gadolinium-based contrast agent used in this study is designed to reduce the T1 relaxation time of its environment; in this way, the CSF signal was increased and therefore its contrast relative to bone and cord was enhanced.

Image Analysis

Freely available software (ImageJ, version 1.43, http://www.rsbweb.nih.gov/ij/) was used for image analysis. The binary data import parameters were 16-bit signed, 256 width, 256 height, 0 offset, little-endian byte order and number of images = 16 (axial scans) or 7 (sagittal scans). The fourth (C4), fifth (C5), and sixth (C6) cervical vertebrae levels were selected for analysis purposes as they were clear in all specimens, and the C4/C5 region is a commonly injured location (Pickett et al., 2006). The axial slices were carefully examined to determine the images that best represented the C4, C5, and C6 levels (slices were selected where the vertebral body was fully evident; typically C6 was identified first by it is characteristic laminae ventralis). Similarly, the sagittal image that most clearly displayed the midsagittal spinal cord (typically identified by the central canal or minimal gray matter) was selected from each specimen.

We used ImageJs standard evaluation tools to directly measure nine geometric parameters from each specimen. The straight line tool was used to measure linear dimensions, with the rater selecting the outer edge of pixels for start and end points of each line measurement. The freehand and wand (tracing) tools were used for the area measurements. Specfically, the freehand tool provided an initial areal measurement utilizing the interface between pixel boundaries. The wand tool was subsequently used to fine-tune this initial areal measurement. Pixel counts were converted into appropriate units via calibration constants based on pixel dimensions.

Six variables were measured from the axial images including: spinal cord depth (SCOD), spinal canal depth (SCD), spinal cord width (SCOW), spinal canal width (SCW), spinal cord area (SCOA), and spinal canal area (SCA). Figure 1 illustrates how each was measured. In brief, SCOD was defined as the distance between the ventral and dorsal edges of the spinal cord along the midsagittal plane through the central canal. SCD was defined as the distance between the ventral and dorsal edges of the spinal canal along the same midsagittal plane. SCOW was defined as the largest distance between the lateral edges of the spinal cord along a line perpendicular to the midsagittal plane. SCW was measured as the distance between the lateral edges of the spinal canal along the line used to define SCOW. Spinal cord area (SCOA) and spinal canal area (SCA) were measured by tracing the perimeters of the spinal cord and spinal canal, respectively (aided by the contrast agent within the cerebrospinal fluid). Three variables were measured from the midsagittal image from each specimen (Fig. 2). Vertebral body height (VBH) was defined as the distance between the most rostral and caudal points on the dorsal side of the vertebral body (Laing et al., 2011). Secondary measures of spinal canal depth (SCDsag) and spinal cord depth (SCODsag) were also measured along a line perpendicular to the dorsal surface of the vertebral body and that extended through the midpoint of VBH (Fig. 2).

Figure 1.

Illustration of measurement approaches from MRI images for spinal cord depth (SCOD), spinal cord width (SCOW), spinal canal depth (SCD), spinal canal width (SCW), spinal cord area (SCOA; outlined in yellow), and spinal canal area (SCA; outlined in blue). Slices are oriented with the ventral aspect at the top of the image. The spinal cord appears as gray, the cerebral spinal fluid (CSF) as white, while the calcified tissues of the spinal column present as dark gray/black.

Figure 2.

Illustration of measurement approaches from MRI images for vertebral body height (VBH), sagittal spinal cord depth (SCODsag), and sagittal spinal canal depth (SCDsag). Slices are oriented with the ventral aspect to the left of the image. The spinal cord appears as gray, the cerebral spinal fluid (CSF) as white, while the calcified tissues of the spinal column present as dark gray/black.

We calculated five additional parameters using the variables described above. The area filled by cerebrospinal fluid and connective tissues (CSFarea) was calculated by subtracting SCOA from SCA. The ratio of spinal cord depth to width was calculated as SCOD/SCOW. Finally, the percentage of the spinal canal filled by the spinal cord was calculated from our depth (SCOD/SCD × 100%), width (SCOW/SCW × 100%), and area (SCOA/SCA × 100%) measurements.

Statistical Analysis

Except where noted, all statistical analyses were performed with a software package using an α of 0.05 (SPSS Version 17.0, SPSS, Chicago, USA). For each dependent variable, a two-way mixed model ANOVA was performed to test the effect of age (between factor: young vs. aged) and vertebral level (repeated factor: C4, C5, and C6). When main effects of vertebral level were observed, least significant difference pairwise comparisons were used to test for specific differences between the C4, C5, and C6 levels. To further investigate the strength of the age factor, for each dependent variable at each vertebral level, we calculated an effect size using Cohen's d approach (mean difference divided by pooled standard deviation). We used the following criteria to interpret the effect sizes: “small” = 0.20; “medium” = 0.50, “large” = 0.80 (Cohen, 1988). Finally, the statistical power associated with the age–group comparison for each variable at each vertebral level was calculated with the G*Power 3.1 software application (Faul et al., 2007).

Image analysis was initially performed by a single assessor across all specimens (the results of which are presented in this manuscript). To address the study's third aim, a second assessor independently measured the same variables. In addition, a single rater repeated the same measurements twice. Intraclass correlation coefficients using an alpha, absolute agreement, two-way random effects model were used to characterize inter-rater reliability (i.e., between rater-ICCb) and intrarater reliability (i.e., within rater-ICCw).

RESULTS

Figures 3 and 4 provide axial and sagittal plane images (respectively) for representative young and aged specimens. Tables 1-3 present the overall results for the spinal cord, spinal canal, and spinal cord/canal variables, respectively. Table 4 indicates the ANOVA test statistic (F) and probability (p) values from the statistical analyses. Only one significant interaction effect was observed (for SCOA/SCA). For all other variables, main effects were interpreted and are presented below.

Figure 3.

Axial MRI images for representative young and aged specimens across the fourth, fifth, and sixth cervical vertebral levels. Slices are oriented with the ventral aspect at the top of the image. The spinal cord appears as gray, the cerebral spinal fluid (CSF) as white, while the calcified tissues of the spinal column present as dark gray/black.

Figure 4.

Midsagittal MRI images for representative young and aged specimens indicating the fourth (C4) to seventh (C7) cervical vertebral levels. Slices are oriented with the ventral aspect to the left of the image. The spinal cord appears as gray, the cerebral spinal fluid (CSF) as white, while the calcified tissues of the spinal column present as dark gray/black.

All spinal cord measures were significantly larger for the aged specimens (Tables 1, 4). Averaged across vertebral levels, the average percent increase in the aged group was 5.9% for SCOW (p = 0.011), 7.2% for SCOD (p = 0.001), 10.8% for SCODsag (p = 0.016), and 10.4% for SCOA (p = 0.014) (Table 1). There were significant main effects of vertebral level for SCOW (p = 0.004), SCOD (p = 0.008), SCODsag (p < 0.001), SCOA (p = 0.003), and SCOD/SCOW (p = 0.012). SCOW measurements increased from C4 to C5, then decreased from C5 to C6. In contrast, SCOD, SCODsag, SCOA, and SCOD/SCOW decreased in size from C4 to C6 (Fig. 5).

Table 1. Average (SD) values for spinal cord geometry variables across age groups and vertebral levels
 Spinal cord geometric variables
 SCOD (mm)SCOW (mm)SCOA (cm2)SCODsag (mm)SCOD/SCOW
 YoungAged% diffYoungAged% diffYoungAged% diffYoungAged% diffYoungAged% diff
  1. Percent (%) difference values are relative to the young group with negative (−) values representing a decrease.

C42.56 (0.12)2.72 (0.07)5.83.95 (0.04)4.26 (0.03)7.20.088 (0.004)0.101 (0.003)12.42.69 (0.05)2.90 (0.07)7.20.65 (0.03)0.64 (0.02)−1.5
C52.52 (0.04)2.72 (0.12)7.04.08 (0.08)4.33 (0.11)5.70.089 (0.003)0.098 (0.003)9.22.50 (0.04)2.87 (0.12)12.90.62 (0.02)0.63 (0.02)1.6
C62.30 (0.08)2.52 (0.07)8.83.97 (0.07)4.17 (0.11)4.80.083 (0.002)0.092 (0.004)9.62.42 (0.10)2.75 (0.18)12.20.58 (0.01)0.61 (0.01)5.2
Avg7.25.910.410.81.7
Figure 5.

Average (±1 SD) measures at the fourth, fifth, and sixth cervical vertebrae for: (a) spinal cord width (SCOW) and spinal canal width (SCW). A significant effect of age was seen only for SCOW. Significant effects of vertebral level were observed for SCOW for C4 versus C5 and C5 versus C6. (b) spinal cord depth (SCOD) and spinal canal depth (SCD). Significant effects of age were observed for both variables. Significant effects of vertebral level were observed for SCOD for C4 versus C6 and C5 versus C6, and for SCD between all three levels. (c) Spinal cord area (SCOA) and canal area (SCA). There was a significant effect of age for SCOA. Significant vertebral effects were observed for SCOA for C4 versus C6 and C5 versus C6.

For spinal canal geometry variables (Tables 2, 4), only SCD demonstrated a significant effect of age with an average increase of 6.1% for older compared to younger specimens (p = 0.025). Three other spinal canal variables were associated with nonsignificant increases for the aged compared to young group (SCW, SCDsag, and SCA). We observed significant effects of vertebral level for SCD (p < 0.001), SCDsag (p = 0.001), CSFarea (p = 0.013), and VBH (p = 0.006). SCD and VBH variables decreased in size from C4 to C6, while CSFarea demonstrated the opposite trend.

Table 2. Average (SD) values for vertebral body geometry variables across age groups and vertebral levels
 Spinal canal geometric variables
 SCD (mm)SCW (mm)SCA (cm2)SCDsag (mm)CSFarea (cm2)VBHD (mm)
 YoungAged% diffYoungAged% diffYoungAged% diffYoungAged% diffYoungAged% diffYoungAged% diff
  1. % difference values are relative to the Young group with negative (−) values representing a decrease.

C42.95 (0.03)3.12 (0.03)5.44.79 (0.08)4.87 (0.09)1.60.121 (0.002)0.129 (0.010)5.73.09 (0.08)3.34 (0.08)7.40.033 (0.002)0.028 (0.002)−18.42.62 (0.09)2.67 (0.16)2.0
C52.83 (0.03)3.04 (0.10)6.84.74 (0.11)4.88 (0.09)2.80.119 (0.002)0.128 (0.01)7.62.93 (0.08)3.22 (0.20)8.80.029 (0.002)0.030 (0.004)2.42.54 (0.09)2.45 (0.13)−3.7
C62.74 (0.08)2.91 (0.13)6.04.80 (0.11)4.81 (0.15)0.10.119 (0.003)0.124 (0.010)4.02.83 (0.04)3.09 (0.25)8.40.036 (0.004)0.032 (0.006)−12.22.45 (0.07)2.36 (0.05)−3.7
Avg6.11.55.88.2−9.4−1.8

Results for the spinal cord to spinal canal percentage measures are provided in Tables 3 and 4. SCOW/SCW was significantly larger (by an average of 4.7%) for the aged compared to young group (p = 0.021). A significant main effect of vertebral level was observed for SCOD/SCD, characterized by smaller values at C6 compared to C5. Finally, a significant interaction was observed for SCOA/SCA, with the largest value observed at C5 for the young group compared to a progressive decrease from C4 to C6 for the aged group (p = 0.019).

Table 3. Average (SD) values for spinal cord and spinal canal comparison variables across age groups and vertebral levels
 Spinal cord/canal geometric variables
 SCOD/SCD (%)SCOW/SCW (%)SCOA/SCA (%)
 YoungAged% diffYoungAged% diffYoungAged% diff
  1. % difference values are relative to the Young group with negative (−) values representing a decrease.

C486.80 (3.37)87.14 (3.11)0.482.53 (0.97)87.51 (1.38)6.072.80 (2.36)78.40 (0.90)7.7
C589.26 (1.36)89.42 (1.76)0.286.06 (3.08)88.72 (1.07)3.175.10 (1.94)76.49 (1.82)1.9
C684.13 (1.20)86.77 (1.69)3.182.68 (3.07)86.74 (1.10)4.969.85 (3.00)74.30 (2.58)6.4
Avg1.24.75.3
Table 4. Analysis of variance test statistic (F) and probability (p) values associated with main and interaction effects for age and vertebral level variables, in addition to pairwise comparison posthoc test results (with Bonferroni's corrections) across vertebral levels
 FpPost hoc p
 AgeVert levelAge × levelAgeVert levelAge × levelC4 vs. C5C4 vs. C6C5 vs. C6
  1. a

    p < 0.05.

SCOD71.49.50.20.001a0.008a0.8600.7330.027a0.006a
SCOW20.012.21.70.011a0.004a0.7520.036a0.2490.006a
SCOA17.720.21.30.014a0.003a0.3160.0880.011a0.009a
SCOD/SCOW0.98.20.90.4040.012a0.4280.1970.022a0.041a
SCODsag15.822.93.70.016a<0.001a0.0730.003a0.005a0.045a
          
SCD12.225.40.20.025a<0.001a0.7920.020a0.005a0.006a
SCW1.00.21.10.3840.8440.393NANANA
SCA3.32.40.90.1430.1730.423NANANA
SCDsag5.922.70.10.0720.001a0.8670.015a0.007a0.008a
CSFarea1.09.44.90.3750.013a0.0540.410.0610.014a
VBH0.510.41.20.5310.006a0.2510.0570.009a0.15
          
SCOD/SCD0.75.70.70.4510.029a0.5240.1440.2680.016a
SCOW/SCW13.63.70.60.021a0.0740.580NANANA
SCOA/SCA5.324.96.70.083<0.001a0.019a0.6040.012a0.002a

The inter-rater reliability (ICCb) analyses generally revealed excellent agreement between the measurements completed by the two researchers (Landis and Koch, 1977) (Table 5). 81% of comparisons (34 of 42) had ICCbs of greater than 0.80. At vertebral levels C4, C5, and C6, average ICCb values were 0.852, 0.869, and 0.951, respectively. The variables associated with the highest and lowest ICCbs were SCW (0.982) and CSFarea (0.752), respectively. The overall average ICCb, which includes all of the variables at all of the vertebral levels, was 0.921. Intrarater reliability (ICCw) was also high, with 31 of 42 comparisons having ICCws of greater than 0.80 (Table 6). The overall average (SD) ICCw was 0.871 (0.184), and average (SD) percent difference between subsequent measures was less than 1% (average = 0.93%, SD = 1.29%).

Table 5. Between-rater intraclass correlations (ICCb) and % difference (% diff) results across the two raters
 C4C5C6Avg
 ICCb% diffICCb% diffICCb% diffICCb% diff
  1. ICCb values ranged between 0.514 and 0.993, with the average across all variables and vertebral levels equal to 0.921.

SCOD0.9421.20.9440.80.9700.20.9520.8
SCOW0.9010.80.9101.40.9530.40.9210.9
SCOA0.9330.10.9070.10.9490.30.9300.2
SCOD/SCOW0.8161.10.8490.20.8120.40.8260.6
SCODsag0.7790.20.9921.20.9830.80.9180.7
         
SCD0.9272.30.9341.10.9910.80.9511.4
SCW0.9661.50.9931.80.9881.40.9821.6
SCA0.7284.70.9440.50.9620.70.8782.0
SCDsag0.9380.00.9910.40.9500.50.9600.3
CSFarea0.5141.30.7490.70.9921.60.7521.2
VBH0.9883.30.9910.80.9120.10.9641.4
         
SCOD/SCD0.9150.30.4512.20.9080.20.7580.9
SCOW/SCW0.7840.30.7961.20.9600.50.8470.6
SCOA/SCA0.8020.70.7130.80.9830.60.8330.7
Table 6. Within-rater intraclass correlations (ICCw) and % difference (% diff) results from comparisons of two measurements from a single rater
 C4C5C6Avg
 ICCw% diffICCw% diffICCw% diffICCw% diff
  1. Average (SD) ICCw values across all variables and vertebral levels equalled 0.871 (0.184). Average percentage difference values were less than 1% (average = 0.93%, SD = 1.29%).

SCOD0.4732.10.8961.70.9970.00.7891.297
SCOW0.9960.00.9100.50.9670.40.9580.283
SCOA0.9930.40.9900.90.9520.20.9780.487
SCOD/SCOW0.9981.00.7500.50.9090.60.8860.720
SCODsag0.9910.80.9891.10.9720.50.9840.808
         
SCD0.9420.50.9970.10.9980.20.9790.263
SCW0.8350.30.6571.30.9980.10.8300.533
SCA0.9310.40.9461.30.9430.50.9400.736
SCDsag1.0000.10.9910.60.9920.10.9940.292
CSFarea0.9060.30.7847.20.6982.50.7963.362
VBH0.9371.10.9810.30.7833.60.9001.666
         
SCOD/SCD0.2372.70.7610.10.9830.10.6600.979
SCOW/SCW0.7410.40.2621.30.9670.30.6570.664
SCOA/SCA0.9560.00.8652.00.7250.80.8490.922

DISCUSSION

The objectives of this study were to characterize age-related geometric differences in the spinal cord (primary aim) and canal (secondary aim) of young and aged Fischer 344 rat specimens and to determine the inter-rater reliability of the image analysis approach used (tertiary aim). Concerning our primary aim, we observed that spinal cord depth, width, and area were significantly larger in the aged group by an average of 8.6%. Additionally, there were main effects of vertebral level on these variables, with values generally larger (by values ranging from 5.7% to 12.9%) at C4 compared to C6. Regarding our secondary aim, only a single spinal canal variable was significantly larger in the aged specimens (spinal canal depth (SCD) by an average of 6.1%). Similar to the spinal cord variables, there were main effects of vertebral level for SCD, SCDsag, CSFarea, and VBH with values generally larger (ranging from 5.4% to 6.8%) for C4 compared to the C6 levels. Finally, our image analysis approach was found to have high inter-rater and intrarater reliability with mean intraclass correlations across all measures of 0.921 and 0.871, respectively (tertiary aim). These data will assist in interpreting and planning studies which use a rat model to investigate the influence of aging on injury outcomes following SCI.

The age-related differences in spinal cord geometry we observed were statisically significant, and of sufficient magnitude to potentially influence outcomes in rat model studies that evaluate the effects of age on SCI. Due to the large mean difference across groups, and the small within-group variability, the effect sizes (which measure the strength of a phenomenon) were extremely large for SCOD (2.263), SCOW (4.512), SCOA (3.174), and SCODsag (3.285) (Table 7). The effect sizes for the spinal canal variables were generally lower but still above the 0.8 criteria Cohen (1988) suggests for “strong” relationships (SCW: 0.803; SCA: 1.011; SCDsag: 2.160; SCD: 3.362). However, the small sample size limited our power (Table 7) for some of the spinal canal comparisons (e.g., average power for SCW = 0.239; SCA = 0.277; SCDsag = 0.657), resulting in age-related differences that could not be classified as statistically significant. Armed with a larger sample size, Laing et al. (2011) used µCT (which is more suited to measuring calcified tissue geometry) to demonstrate that compared to young Fisher 344 rats, aged specimens have significantly larger SCW (by an average of 9.1%) and SCA (by an average of 19.7%). Accordingly, while the current study provides insufficient evidence to support significant effects of age on spinal canal width and area, based on the reasonably large effect sizes (Table 7) and the results of Laing et al. (2011), there is reason to believe that significant age-related differences in these secondary variables would have emerged given a larger sample size.

Table 7. Effect size (Cohen's d) and power values between the young and aged specimens for each variable at the fourth (C4), fifth (C5), and sixth (C6) cervical vertebral levels, and averages across these levels
 C4C5C6Avg
 dPowerdPowerdPowerdPower
  1. Power calculated with Gpower 3.1 (Faul et al., 2007) using a t test family, ‘Means: difference between two independent means’ option, one tailed test, and alpha = 0.05.

SCOD1.6280.5022.2360.7232.9260.8982.2630.708
SCOW8.7681.0002.5990.8302.1690.7054.5120.845
SCOA3.6760.9763.0000.9102.8460.8833.1740.923
SCOD/SCOW0.0000.0500.0500.0563.0000.9101.0170.339
SCODsag3.4520.9624.1360.9922.2660.7363.2850.897
         
SCD5.6661.0002.8440.8831.5750.4813.3620.788
SCW0.9390.2471.3930.4100.0760.0590.8030.239
SCA1.1090.3041.2480.3550.6770.1711.0110.277
SCDsag3.1250.9281.9030.6091.4520.4332.1600.657
CSFarea2.5000.8050.3160.0940.7840.2001.2000.366
VBH0.3850.1060.8050.2071.4800.4440.8900.252
         
SCOD/SCD0.2050.0760.3210.0951.6130.5000.7130.224
SCOW/SCW2.7030.8540.9300.2441.8160.5761.8160.558
SCOA/SCA2.0210.6530.5800.1471.7630.5551.4550.452

It is relevant to compare our results to the cervical spinal cord and canal geometry reported in the literature for rats. The 10.4% increase in cervical SCA for our aged animals corresponds to the 15.6% increase that Fontana et al. (2009) observed for very old (28 months) versus young (5 months) Sprague–Dawley rats. Our SCA magnitudes were also similar to theirs, with ranges of approximately 0.08–0.10 cm2 reported for both studies. In addition, both studies indicate that SCOD/SCOW decreases from C4 to C6 in both young and aged groups (although Fontana et al. reported their ratio as SCOW/SCOD). Our results also concur with observations that axial SCA decreases from C4 to C6 levels in rats (Portiansky et al., 2004). Regarding spinal canal measures, this study agrees with Laing et al. (2011) who used µCT to show that SCD was significantly larger in aged Fisher 344 rats, and that it decreases from C4 to C6 (Flynn and Bolton, 2007; Laing et al., 2011). However, our current study does not correspond to the general increases in SCW from C4 to C6 reported by Flynn and Bolton (2007) and Laing et al. (2011). These differences may be partly influenced by imaging modalities, as MRI is more accurate at imaging soft tissues such as the spinal cord than the calcified tissues which comprise the spinal column.

There is less consensus across studies which characterize spinal cord and canal geometry in humans. Our results concur with Ishikawa et al. (2003) (MRI) and Yu et al. (1985) (CT) observations that SCOA decreases from C4 to C6, but are at odds with Kameyama et al. (1994) who reported the opposite trend based on histological assessments. Our results align with those of Ishikawa et al. (2003) who found that the ratio of spinal cord depth to width decreased from C3 to C5/6 but was similar across ages. However, our results do not correspond with their finding that axial spinal cord area decreases with age after 20 years of age in humans.

Finally, our results correspond to reports that SCD decreases from C4 to C6 in Caucasians (Tatarek, 2005), despite opposite trends observed for those with Chinese (Lim, 2004) and African American (Tatarek, 2005) ancestry. Potential differences of cervical spinal cord and column geometry between humans and rats likely relate to the differences in how gravitational forces typically load the spinal column in bipeds (compression) versus quadrupeds (shear).

Our results have direct implications for studies that use a rat model to assess age-related differences in cervical SCI. Many contusion injury protocols involve a laminectomy followed by the use of an indentation probe to displace into and compress the spinal cord (Young, 2002). We found that the average depth of the C4–C6 spinal cord was 7.2% larger in older specimens (2.65 vs. 2.46 mm). For a probe indentation depth of 1.1 mm (Choo et al., 2007), this would represent 44.7% of spinal cord depth for younger specimens, versus only 41.5% for older specimens. Interestingly, this suggests that, if all physiological systems and material properties were similar, more pronounced tissue damage might be observed in younger specimens for contusion-type injuries. The implications are more complex for dislocation injury protocols which indirectly load the spinal cord through relative displacements of adjacent vertebrae (Fiford et al., 2004; Choo et al., 2007-2009). For example, if we consider spinal cord and spinal canal depth to grow in parallel (a reasonable assumption considering we observed no effect of age on SCOD/SCD or SCOA/SCA) then a 2.5 mm dorsal–ventral dislocation of the C4/C5 vertebrae (Choo et al., 2007) would occlude the spinal canal to 88.0% and 82.8% of total depth for our young and aged animals, respectively. The decreased strain in the older specimens might result in diminished injury severity, and if unaccounted for, could lead to the potentially erroneous conclusion that differences in injury outcomes were due to increased tissue resilience or less secondary degenerative responses to injury in the aged. For lateral dislocations, our results suggest that spinal cord width increases with age (by 5.9%) while spinal canal width remained relatively fixed. Accordingly, a given lateral dislocation may produce the same spinal canal occlusion across young adults and aged rats, but a greater compression of the wider spinal cord in the aged animals. Lau et al. (2013) accounted for the even larger morphological differences between neontal and adult rats using smaller lateral dislocation magnitudes for the neunates (ranging from 1.6 to 4.0 mm vs. 4.0 to 10.0 mm, respectively). Age-related changes in the spinal cord and spinal canal may have less influence on injury models that employ cord laceration or clip compression. However, in general the geometric differences in the spinal cord and canal we observed between young and aged adults may have implications for subsequent spinal cord severity that should be considered in parallel with potential age-related changes in the physiological injury cascades that are initiated following acute SCI.

This study was associated with several limitations. As discussed above, although the small sample size is the study's foremost limitation, it did not affect our primary aim of characterizing age and vertebral level effects on measures of spinal cord geometry. While the smaller effect sizes did limit our power to observe potential age-related effects on spinal canal geometry (our secondary aim), the variable magnitudes and trends we observed support the age-related differences in spinal canal measures observed by Laing et al. (2011) using µCT approaches. Second, our standard specimen fixation protocol resulted in a slightly lordotic curvature of the cervical spinal column. As a result, our axial images may not have been orthogonal to the posterior aspect of the vertebral bodies. However, this potential alignment issue would have primarily affected our measures of spinal canal depth. As we observed an average difference of only 4.8% between SCD and SCDsag, and a similar 4.8% difference between SCOD and SCODsag, we are confident this issue had little bearing on the results of our age or vertebral level comparisons. Third, our study included young (3 months) and aged (12 months) adult rats, but not a geriatric group. However, we feel justified in this approach as Laing et al. (2011) observed no significant differences between 18- and 30-month-old rats for morphometric measures of the spinal canal. Additional studies are required to assess whether age greater than 12 months influences spinal cord geometry in this rat model. Fourth, the resolution associated with our imaging protocol was insufficient to accurately differentiate gray and white matter within the spinal cord. Follow-up studies may want to address this limitation toward establishing in vivo evidence of potential age-related effects on spinal cord composition and distribution. Fifth, despite the benefits associated with the contrast agent we utilized, MRI is not the best tool to image bone and therefore may have influenced the accuracy of the spinal canal measurements (Norman et al., 1983). Accordingly, imaging modalities such as µCT may provide more authoritative spinal canal measures if only calcified tissues are of interest (Laing et al., 2011). If soft tissues (in isolation or relative to bone) are the primary targets, MRI is likely a more appropriate approach. Finally, this study did not assess rat strains that are most commonly used in rat models of SCI (e.g., Sprague Dawley). However, despite differences in absolute values, both Fisher 344 and Sprague Dawley females reach peak body mass at approximately 100 weeks, and 75% of peak at approximately 52 weeks (Thurman et al., 1994; Hubert et al., 2000). Accordingly, while not specifically tested in this study, there is little reason to believe the general age-related differences we observed for Fisher 344s would not apply to other rat strains.

This study characterized age-related differences in the morpometry of the cervical spinal cord and spinal canal in a rat model. This study clearly demonstrates that cervical spinal cord increases in size (by approximately 8–9%) between the ages of 3 and 12 months in Fischer 344 rats. The spinal cord increases were larger than the increased spinal canal depth in the aged versus young groups. In addition to potential age-related changes in physiological pathways, these morphometric differences could influence age-related differences in spinal cord injury severity/outcomes. Accordingly, the findings from this study can aid in the planning and interpretation of studies that use a rat model to investigate the influence of age on cervical SCI.

ACKNOWLEDGMENTS

The authors gratefully acknowledge the valuable input provided by Dr. Wolfram Tetzlaff over the course of this study.

Ancillary