J Manipulative Physiol Ther 2007 (May); 30 (4): 279–294
Charles N.R. Henderson, DC, PhD, Gregory D. Cramer, DC, PhD, Qiang Zhang, MD, DC, James W. DeVocht, DC, PhD, and Jaeson T. Fournier, DC, MPH
Palmer Center for Chiropractic Research, Davenport, Iowa, USA. firstname.lastname@example.org
OBJECTIVES: The purpose of this study was to characterize intervertebral stiffness and alignment changes in the external link model and evaluate it as an experimental mimic for studying the chiropractic subluxation.
METHOD: A controlled test-retest design was used to evaluate rats with spine segments linked in 3 alignment configurations and controls that were never linked. Dorsal-to-ventral spine stiffness was measured with a load platform, and flexion/extension misalignment was assessed on lateral radiographs obtained with a spine extension jig. Descriptive statistics were computed for study groups, and multiple linear regression models were used to examine all potential explanatory variables for the response variables "stiffness" and "joint position."
RESULTS: Rats tested with links in place had significantly higher dorsal-to-ventral stiffness in the neutral configuration than rats in the flexed configuration. This difference remained after the links were removed. Stiffness after link removal was greater for longer linked periods. Surprisingly, stiffness after link removal was also greater with longer unlinked periods. Longer linked periods also produced greater misalignments during forced spine extension testing. Although link configuration was not a statistically significant predictor of misalignments, longer times after link removal did produce greater misalignments.
CONCLUSIONS: This study suggests that the external link model can be a valuable tool for studying the effects of spine fixation and misalignment, 2 cardinal features of what has been historically described as the chiropractic subluxation. Significant residual stiffness and misalignment remained after the links were removed. The progressive course of this lesion is consistent with subluxation theory and clinical chiropractic experience.
From the FULL TEXT Article
The findings from this study suggest that the ELM can be a valuable tool for studying biologically significant effects of spine fixation and misalignment. We showed that the external link system provided a means for producing both experimental spine fixation and misalignment, the 2 cardinal biomechanical features of subluxation. Both were marked while the external links were in place, and significant residual stiffness and misalignment remained after the links were removed. Most interesting was the apparently progressive effect of this experimental biomechanical lesion. After the links were removed, the spine segment continued to become stiffer, and misalignment observed on the stress radiographs was greater over the subsequent 12-week period of the experiments. These observations are consistent with both subluxation theory and clinical chiropractic experience.
Comparing Posterior-Anterior and Dorsal-Ventral Stiffness Testing
Manual posterior to anterior (P-A) stiffness testing is often used by chiropractors and physical therapists to identify the site and nature of biomechanical spine lesions. [10-14] Early manual methods showed poor reliability and validity; however, a newer reference-based method developed by Maher et al  showed good reliability as well as criterion-related validity (interrater ICC[2,1] = 0.78, criterion-related validity = 0.74). [10, 11, 15] Chiradejnant et al  reported that therapists could distinguish between stiffness stimuli that vary by only 9% when using this method. The validity of this reference-based manual method was determined by assessing agreement with an instrumented stiffness testing device, the stiffness assessment machine (SAM).  Stiffness testing instruments such as the SAM apply a 3-point load wherein the loading points are widely separated on the spine. The subjects lay prone on a lightly padded table as a small indenter loads the spine P-A. The spine is treated essentially as an elastic beam suspended between the thorax and pelvis. These stiffness testing instruments have become a criterion standard against which manual methods are tested. Three such instruments have been developed and validated in the past decade: (1) the spinal physiotherapy simulator (SPS),  (2) the SAM,  and (3) the spinal posteroanterior mobilizer (SPAM).  Excellent test-retest reliability have been reported for all 3 instruments (ICC[2,1] = 0.88 SPS, 0.96 SAM, and 0.98 SPAM). [7, 16, 17] In vitro accuracy has also been established for all 3 instruments using elastic beams of known stiffness. [7, 16-18]
In our study, we used a proprietary material testing instrument, the PCM-Versa Test load frame (Mecmesin), to evaluate D-V stiffness in the rat. This is comparable with P-A stiffness in humans. In vitro test-retest studies of the load frame in our laboratory demonstrated excellent instrument reliability (ICC[2,1] = 0.998) on 3 silicone rubber pads (McMaster-Carr Elmhurst, IL) that were selected because they had similar stiffness values to what we observed in the rat spine. The PCM-Versa Test load frame has a displacement resolution of 5 µm, and the 10-N load cell is accurate to 10.00 ± 0.02 N (SD). This equipment is calibrated regularly in our facility. We have also determined the intrarater and interrater reliability (ICC[2,1] = 0.815 and 0.692, respectively) for our application of this device in measuring D-V stiffness in the rat (unpublished data).
The SAUs gave direct access to the spine segments of interest (L4-L6). Therefore, we were able to focus the 3-point load to these segments by supporting the 2 outer SAU stems (L4 and L6) with small brass turnbuckles suspended from rigid supports while the indenter loaded L5. We reasoned that focusing the load in this way would minimize the dampening error that must occur when segmental stiffness is measured using a method that actually distributes the applied load over most of the spine, as in the SPS, SAM, and SPAM approaches. Suspending the region of interest with the turnbuckle supports should also have reduced, although not completely eliminated, the confounding influence of abdominal/thoracic compression. Consequently, we believe that our load-localizing approach has face validity for reducing the competing systematic errors contributed by nonfocused loading and abdominal tissue support. Reducing these confounding influences should provide more accurate measures of segmental spine stiffness.
It was interesting that the L5 D-V stiffness observed in our pooled control rats/not-yet-linked rats (D-V stiffness = 14.52 ± 4.47 N/mm [mean ± SD], n = 202) was similar to the L4 P-A stiffness reported in humans. Lee and Liversidge  reported L4 P-A stiffness values in healthy human volunteers to be 17.5 ± 4.8 N/mm and Viner et al  reported 15.5 ± 3.3 N/mm. Because the rat has 6 lumbar vertebrae, we consider the L5 vertebra in the rat to be in a similar “biomechanical position” to the L4 vertebra in humans.
Investigators have commented on the large amount of variation in P-A stiffness across populations of healthy individuals and have suggested that, when looking at treatment effects, a comparison of stiffness changes within each subject may be of greater value than comparison between subjects. [16, 20, 21] We also found great variability in D-V stiffness across rats (see standard deviations in the intervertebral fixation section of “Results”). Consequently, the stiffness assessments in this study compared stiffness changes within rats (change from baseline).
As noted in the “Methods” section, rats linked in the rotation configuration had a substantial axial (left-right) tilt of the L5 SAU stem, which converted much of the D-V load into a moment about the animal's body axis. For this reason, animals linked in the rotation configuration were excluded from D-V stiffness testing. However, preliminary work currently being conducted by our group indicates that this vertebral rotation fixation may produce unique anatomical changes of the Z joints, when evaluated with light microscopy. In future studies, we will continue to assess the features of this configuration, including stiffness using measurement methods that are not effected by SAU stem tilt.
Evaluation of Working Hypotheses
Our working hypotheses for stiffness changes were largely supported in this study. The D-V spine stiffness of control rats was the same as that of not-yet-linked experimental rats, and this did not significantly increase with age (Hypothesis 1). This finding is consistent with a clinical study by Viner et al  of 42 men and women, aged 20 to 45 years, with no history of low back pain. They also found no statistically significant correlation between age and spine stiffness. The observation in Table 4 that age at baseline (age before linking) was negatively related to linked stiffness was because younger rats were generally linked in the longest link periods, and longer link periods produced greater stiffness.
As expected, the spine stiffness of rats tested while linked was markedly greater than SAU bearing control animals (CSAU), and linked stiffness was not significantly greater for longer link periods (Hypothesis 2). However, contrary to our expectation, rats linked in the neutral link configuration had greater D-V stiffness than rats linked in a flexed configuration. We expected that, while the links were in place, the immobilization produced by all 3 link configurations would be essentially complete, with no measurable stiffness differences among them. Our data suggest that the links produced various amounts of reduced segmental mobility but not complete immobilization. This is consistent with the concept of spine fixation used by chiropractors when referring to subluxated segments. Fixation, in chiropractic use, is synonymous with spine hypomobility rather than complete immobilization.
Retention of considerable mobility, even in surgically fused spines, has been reported in both animal and human studies. In an analysis of single-level (L5-L6), bilateral, posterolateral intertransverse process fusion in a New Zealand white rabbit model, Erulkar et al  reported statistically (P = .01) and biologically significant decreases in flexion (81%) and extension (61%) at 5 weeks postsurgery. They noted that, although posterolateral fusion reduced spine motion, it did not eliminate it. Similarly, Lee and Langrana  simulated posterolateral, posterior, and anterior fusion in human cadavers with transfixion wires and polymethylmethacrylate. They reported that posterolateral fusion increased spine stiffness by 48%, whereas posterior fusion and anterior fusion increased stiffness by 92% and 47%, respectively. Cadaver studies have also showed increased stiffness without complete rigidity immediately after implantation of fixator instruments (time 0 studies). [24-26]
We do not know why rats linked in a neutral configuration (EN) were stiffer than those linked in the flexed configuration (EF), but we note that the normal posture of the L4-L6 region in the resting (crouched) rat is moderately flexed.  The neutral link configuration in this study was obtained by yoking the 3 SAU stems of the external linking system in the position they assumed when the anesthetized animal was placed prone on a flat surface. Therefore, it could be argued that our neutral link configuration actually placed the L4-L6 region of the spine in a more extended position than that assumed when the animal is in its normal crouched posture. By contrast, the flexed link position probably placed the lumbar spine in a posture that is much closer to its resting position. Therefore, linking the L4-L6 region in neutral position may actually have placed that spine segment in an abnormal (ie, extended) posture that was closer to its end range of motion. At end range, the bony “stops” of the inferior articular processes meet the pars interarticularis of the vertebra below. 
As anticipated, we found that D-V stiffness immediately after the links were removed (immediate residual stiffness) was much less than while linked (linked stiffness), but greater than baseline (prelinked stiffness). Also, immediate residual stiffness was greater for longer linked periods and was influenced by the link configuration (Hypothesis 3). This effect of link configuration on immediate residual stiffness was similar to, but was less marked than, that observed during the link period. The opportunity for spontaneous recovery was examined by observing stiffness changes within subgroups of rats at 1, 2, 4, 8, or 12 weeks after link removal (long-term residual stiffness). As expected, D-V stiffness measured weeks after link removal was greater for longer link periods (Hypothesis 4). But, surprisingly, long-term residual stiffness also increased with greater weeks unlinked, rather than decreasing toward baseline as we anticipated (Table 4). However, this effect was less than that observed for longer link periods (weeks linked). It appears that a progressive stiffening of the spine was initiated by the experimental fixation. If this phenomenon is also present in human populations with naturally occurring spine fixations, it suggests that such lesions are progressive and have increasing biomechanical consequences.
Spine stiffness is of interest to patients and clinicians because it is widely thought that abnormal stiffness (increased or decreased) is associated with pain, degenerative change, or reduced function. [29-32] Grob et al  reported that external fixation relieved pain in 89% of patients with suspected cervical spine instability, and it has been shown that manual P-A stiffness testing can identify the level of intervertebral joint dysfunction. [34, 35] Latimer et al  reported that, in 25 patients with low back pain, P-A stiffness at the most symptomatic level decreased by a mean 8% as the level of pain decreased to at least an 80% improvement. This suggests that an 8% change in stiffness may be clinically significant. Nine of the 25 low back pain subjects in the Latimer et al  study demonstrated a 14% to 37% reduction in stiffness. Consequently, the immediate residual stiffness increases of linked animals over control or not-yet-linked animals observed in our ELM may well be biologically significant. Immediate residual stiffness increases above baseline differed with the configuration and with number of weeks linked (Table 4). Moreover, long-term residual stiffness also substantially increased with increased number of weeks linked. These observations are consistent with both subluxation theory and clinical chiropractic experience. 
Our recently published work on degenerative spine changes after experimental fixation showed one way this model may be used to explore the biologic effects of spinal fixation (hypomobility, subluxation).  Consistent with the stiffness data reported here, we observed degenerative changes in zygapophysial joints (Z-joints), and these changes were greater for animals experimentally linked for longer periods. In addition, like long-term residual stiffness, we observed progressive degenerative changes after the links were removed. However, it was interesting that spontaneous improvement in Z-joint articular surface degeneration and osteophyte formation appeared to occur in some rats after the external links were removed. When present, this improvement was associated with a link-time threshold. Some rats linked for 1 to 4 weeks, and then observed weeks after the links were removed showed improvement in articular surface degeneration. Similarly, osteophytes were less frequent and/or severe in some rats linked for 4 to 8 weeks and then observed for weeks after link removal. It was noted that rats linked for more than 4 weeks showed only progressive articular surface degeneration, and rats linked for more than 8 weeks developed only more (or larger) osteophytes during the subsequent unlinked period. These observations suggest 2 very interesting questions for future study: (1) How are D-V stiffness and degenerative Z-joint changes correlated? (2) Can therapeutic intervention stop or even reverse increased spine stiffness and/or degenerative spine changes? We are currently examining these questions with the ELM.
Misalignment is a frequently cited component of subluxation theory. [1, 37, 39] Historically, it has often been referred to as “a bone out of place.” The early notion was that subluxation was a static mechanical lesion, often demonstrable on a neutral radiograph wherein 1 vertebra was out of normal alignment with adjacent segments. This misalignment was asserted to be less than a dislocation, hence, the term sub (less than) luxation (dislocation). From its earliest presentation by the founders of chiropractic, this static perspective of the subluxation has been vigorously challenged within the chiropractic profession. [39, 40] The current view of vertebral subluxation describes a dynamic biomechanical lesion wherein an error of movement is present that may, or may not, be associated with static misalignment. [29, 41]
We expected that rats examined radiographically in the extension jig while linked (during the link period) would have decreased extension. We also thought that a stable misalignment in the sagittal plane would be achieved with the links in place and, therefore, anticipated that this decreased extension would be correlated with link configuration, but not with the length of the link period (Hypothesis 5). Our data support this hypothesis (Table 5). We anticipated that some residual of this dynamic misalignment would remain immediately after the links were removed—with weeks linked and link configuration influencing the magnitude of the PBA (Hypothesis 6). Although there was a residual misalignment immediately after the links were removed, which was significantly greater for longer link periods, the link configuration did not have a statistically significant effect (Table 6). It is possible that a link configuration effect was hidden in this small sample because the residual PBAs were small compared with the linked PBAs. Lastly, we thought that the misalignment created while linked would decrease after the links were removed, possibly to stabilize at a minimal misaligned state in the extension jig or even approximate control values (Hypothesis 7). We further expected that the magnitude of this long-term residual misalignment would be different for the 3 link configurations and would be greater for longer link periods but smaller for longer unlinked periods. As expected, misalignment was greater for longer link periods, but as with the stiffness data, misalignment was also greater for longer unlinked periods. These data further support the notion that a progressively developing process affected the biomechanics of the experimentally fixed spine segments (L4-L6).
This minimal residual radiographic misalignment in the presence of increased intervertebral stiffness suggests that malposition (misalignment) is a dynamic phenomenon that cannot be simply characterized as a bone out of place on static films. Gatterman  commented on the assertion by some that the term subluxation should be reserved for radiographically measurable disrelationships of joint surfaces. She argued that some misalignments may be insufficient to be readily discernible by current radiographic technologies, and therefore, radiographically shown misalignment should not be the sole criterion for detecting subluxation. Rather, she emphasized the role of fixation in identifying the subluxation, “It seems probable that subluxation refers to impaired mobility with or without positional alteration …. Logic leads us to conclude that we are dealing with a functional entity involving restricted vertebral movement, because it is the movement-restriction component of manipulable subluxation that responds to thrust procedures.”
Quadruped vs Biped
It is widely accepted that quadruped spines are subjected to different loads than upright human spines. Consequently, the observations made in this quadruped model must be interpreted cautiously when making inferences about bipedal spine function. This consideration was discussed in our first article on this topic.  We concluded that fundamental biomechanical similarities between the rat and human spines permit critical studies of basic biomechanical mechanisms that are common to both species. Moreover, such studies can benefit from both the similarities and differences between species. Much of the work presented here is currently being replicated in “behaviorally induced” biped rats. Biped rats spend more time on their hind legs than typical quadruped rats and have been found by investigators to more closely mimic the biomechanical features of bipeds, such as humans. [42-45] Thoughtful application of the ELM to the study of spine function, with due consideration to important experimental advantages and limitations, offers great promise. Further study is also needed to determine if the experimental hypomobility and misalignment produced in the ELM has a homologue in either the rat or the human population. Consequently, we stress that although this study supports the theoretical construct that has been historically identified as the chiropractic subluxation, it does not establish it as a lesion that actually exists in either rats or people.
The findings from this study suggest that the ELM can be a valuable tool for studying biologically significant effects of spine fixation and misalignment. We showed that the external link system provided a means for producing both experimental spine fixation and misalignment, the 2 cardinal biomechanical features of subluxation. Both spine fixation and misalignment were marked while the external links were in place, and significant residual stiffness and misalignment remained after the links were removed. Most interesting was the apparently progressive effect of this experimental biomechanical lesion. After the links were removed, the spine segment continued to become stiffer over the subsequent 12-week period of the experiments. These observations are consistent with both subluxation theory and clinical chiropractic experience. In addition to studying the biologically significant effects of subluxation, the controlled application of specific treatment interventions to this model may also allow us to study the mechanisms by which spine treatments produce their effects.