J Manipulative Physiol Ther 2002 (Oct); 25 (8): 504–510 ~ FULL TEXT
Bruce P. Symons, Tim Leonard, Walter Herzog
Human Performance Laboratory,
Faculty of Kinesiology,
University of Calgary, Canada
BACKGROUND: Spinal manipulative therapy (SMT) has been established as a clinically effective modality for the management of several musculoskeletal disorders. One major issue with the use of SMT is its safety, especially with respect to neck manipulation and the risk of stroke in the vertebrobasilar system.
OBJECTIVES: Our objectives were to quantify the strains and forces sustained by the vertebral artery (VA) in situ during SMT .
Study Design: This was a cadaveric study.
METHODS: Six VAs were obtained from 5 unembalmed postrigor cadavers. The cephalad/distal (C0-C1) and caudad/proximal (C6-subclavian artery) loops of the VA were carefully exposed and instrumented with a pair of piezoelectric ultrasonographic crystals. The strains between each crystal pair were recorded during range of motion testing and diagnostic tests and during a variety of SMT procedures. The VA was then dissected free and strained on a materials testing machine until mechanical failure occurred.
RESULTS: SMT performed on the contralateral side of the cervical spine resulted in an average strain of 6.2% +/- 1.3% to the distal (C0-C1) loop of the VA and a 2.1% +/- 0.4% strain to the proximal (C6) loop. These values were similar to or lower than the strains recorded during diagnostic and range of motion testing. Failure testing demonstrated that the VAs could be stretched to 139% to 162% of their resting length before mechanical failure occurred. Therefore the strains sustained by the VA during SMT represent approximately one ninth of the strain at mechanical failure.
CONCLUSIONS: SMT resulted in strains to the VA that were almost an order of magnitude (ten times) lower than the strains required to mechanically disrupt it. We conclude that under normal circumstances, a single typical (high-velocity/low-amplitude) SMT thrust is very unlikely to mechanically disrupt the VA.
From the Full-Text Article:
Spinal manipulative therapy (SMT) has now been
established as a clinically effective modality for the
treatment of patients with low back pain and other
musculoskeletal disorders. An increasing number of nonchiropractic
health care providers such as nurses, physicians,
physical therapists, and orthopedic specialists are
incorporating SMT into their treatment repertoire. However,
despite this increasing popularity of SMT, there is little
basic research into the mechanisms underlying spinal manipulation
and its pathophysiologic effects on the human
SMT typically consists of a high-velocity, low-amplitude
thrust delivered to a specific landmark on the spine in a
specific direction. We have previously characterized the
mechanics of SMT by measuring the forces exerted by
clinicians during SMT delivered to the cervical spine, [1, 2] the
thoracic spine, [3, 4] and the sacroiliac joint. [3-5] A typical SMT
treatment delivered to the cervical spine will produce peak
forces of approximately 100 to 150 N, [1, 2] whereas treatments
on the other areas of the spine are associated with average
peak forces of 400 to 500 N. [5, 6] These forces are generally
delivered within 200 ms [3, 4] and thus may produce substantial
local accelerations. However, these measurements have all
been performed on the body surface; it is not known how
these forces applied externally to the skin are transmitted
through the various soft-tissue layers and bones into the
deeper anatomic structures.
One major issue with the use of SMT is its safety,
especially with respect to neck manipulation and the risk
of stroke. Conservative estimates of the risk of a stroke
associated with SMT are on the order of 1 per million, 
but the actual number remains unknown; reports in the
literature vary between 1 in 5000 to 1 in 10 million. [7-12]
Although this risk is small, the serious and irreversible
nature of vascular accidents  makes this a material risk.
The earliest documented reports of fatal vascular accidents
after spinal manipulation can be traced back to the
case of Foster versus Thornton  in 1934 and Pratt-
Thomas and Beyer  in 1947. The vast majority of these
incidents have involved the vertebrobasilar system, specifically
the cephalad/distal loop of the vertebral artery
(VA) as it exits the foramen transversarium of C1 and
travels posteriorly into the foramen magnum.  Because
of this unique configuration of the VA, it has been
suggested that the VA experiences considerable stress
and stretch during extension and rotation of the neck,
which may lead to hemodynamic occlusion,  physical
damage, or both. Consequently, it has been hypothesized
that SMT may also cause similar types of damage because
of its high-velocity and high-acceleration nature.
The purpose of this study was to characterize the nature
and magnitude of the strains and elongations of the VA
during SMT and then to compare these values against the
ultimate failure loading strain of the VA. Based on failure
testing of the VA, we can also calculate indirectly the forces
experienced by the VA during SMT. A number of different
SMTs and procedures were investigated in this study including
range of motion (ROM) testing and vertebrobasilar
insufficiency (VBI) testing.
An ischemic event sustained in the vertebrobasilar system during SMT can arise from a variety of causes such as pinching or kinking of the VA during neck movement, vasospasm of the VA, systemic shock or hypotension, physical obstruction of the VA by a dislodged thrombus, embolus, or atherosclerotic plaque, and a traumatic tear in the VA. This study focused on the last possibility by directly evaluating the strains and forces exerted on the VA itself during SMT.
The forces applied by chiropractors were first measured by Adams and Wood [11, 17] in 1984. With an instrumented manipulation “dummy,” they compared the forces exerted by experienced and student chiropractors during SMT of the sacroiliac joints. We have previously measured the forces exerted by chiropractors and the subsequent vertebral movements for a variety of treatments. [1–19, 28] However, with the exception of the intervertebral disk,  there have been no previous studies measuring the internal forces generated during SMT. Therefore this study represents a novel approach into quantifying the biomechanical effects of SMT.
Although its efficacy is equivocal in the literature, [3, 21, 23] VBI testing by positioning the neck into extension plus rotation is currently the clinical standard for screening against potential stroke. Our data showed that VBI screening resulted in 4% and 12% strain to the ipsilateral and contralateral distal (C0-C1) loops of the VA, respectively. In addition, rotation of the neck resulted in 5% and 13% strain to the ipsilateral and contralateral distal (C0-C1) loops of the VA. In contrast, other cervical ROMs resulted in only 2% to 6% strain (Table 2). These values suggest that similar strains were placed on the VA during both procedures and support the contention that rotation may be a potential mechanism for causing VBI. 
Our results indicated that cervical SMT averaged 6% strain to the distal (C0-C1) VA loop and 2% strain to the proximal (C6-SA) segment of the VA. These values are lower than those observed during VBI screening and neck rotation. Indeed, the maximal strain produced during SMT was 11% (ipsilateral C3/C4 break) compared with the 12% and 13% strain produced during contralateral VBI screening and rotation, respectively. These values suggest that SMT results in strains that are within the range of strains produced during normal, physiologic motion of the cervical spine.
Few data are available in the literature regarding the mechanical properties of the VA. Yamada  defined the VA as a “muscular” artery and reported average longitudinal failure strains of 1.4-fold in people 20 to 39 years old. This is lower than our findings of 153% to 162% strain before mechanical failure. Johnson et al  recently reported mechanical failure of the VA in 16 cadavers 28 to 90 years old at age of death at a mean longitudinal elongation of 38.7%, which is also lower than the values we obtained in an older population (Table 1). However, 1 of the difficulties in comparing failure strains across studies is the assumption of what constitutes 0 or resting/baseline strain. Here, we defined 0 strain as the strain recorded in situ with the head in a neutral position; the length of the VA segment in this position was taken as the baseline value. In contrast, Johnson et al  defined 0 strain at the first appearance of measurable force; the length of the VA segment at that point was considered to represent baseline. During the failure testing, however, we observed that we could elongate the VA considerably from its 0 strain length before measuring any detectable force. Therefore one would expect greater failure strain values in our study as compared with that of Johnson et al.  Furthermore Johnson et al  used 2 × 20 mm strips of the VA, whereas we tested the VA intact. These differences in methods may have also contributed to the lower failure strain values they obtained compared with our results, particularly because the strip specimens used by Johnson et al  must have been prone to failure at the gripping sites. Unfortunately, they do not report the location of failure of their specimens.
One of the limitations of this study was the use of cadaveric specimens. Panjabi et al  reported that the biomechanical properties of cadaveric spinal specimens did not alter significantly even after 232 days of storage at –20°C, and Yamada  showed that the tensile properties of common carotid arteries harvested from cattle did not change appreciably after 4 days of refrigeration in normal saline solution. Another consideration is the lack of muscle tone in cadavers. However, in our experience most patients are relaxed before SMT, and there is a force-time delay of at least 150 to 300 ms after the onset of the treatment before the muscles begin to respond to the manipulation.  Because most SMTs are completed within 150 ms, and because the muscular forces evoked by SMT start at the earliest at 150 ms after the onset of the thrust, the muscular forces opposing the treatment are of little or no concern, except in those patients who have spasticity in the muscles near the treatment area. This contention is also supported by observations from other investigators. [13–30]
It is important to note that another factor in the use of cadaveric specimens is the condition of the individual. In these experiments it is likely that the failure strains we obtained for the VA were disproportionately low (and thus the relative strains during SMT high) for the following reasons. First, we studied an older population (average: 86.4 ± 7.3 years old; range: 80 to 99 years), most of whom died of cardiovascular diseases (Table 1). Indeed, in cadaver number 4, there was a large aneurysm in the left VA between C3-C4, which did not rupture during the experiment or during the mechanical failure testing (data not shown). Second, although great care was taken during the dissection, it is possible that the VA was nicked or otherwise compromised during the experiment or during removal for the failure testing. Third, the connective tissues, fascial layers, and supporting ligaments may have been separated or removed during the dissection, thus rendering the environment of the VA less stable. Fourth, despite the discussion in the preceding paragraph, there may have been some decay in the VA over the 72 hours after death that weakened the biomechanical integrity of the VA. Fifth, up to 40 trials of ROM/diagnostic testing plus SMT were performed while the VA was exposed and perhaps desiccating, before the VA was removed for failure testing, which may have also compromised the blood vessel. Finally, our cadavers were generally quite thin, with little musculature around the cervical spine. However, all of these potential errors tend to make our specimens a “worst-case scenario,” and we are confident that our estimates of the forces and strains sustained by the VA during SMT are very conservative.
We observed large individual variations in the behavior of the VA. For example, we recorded absolute failure forces for individual VAs (proximal and distal segments) ranging from 4.2 N to 18 N and at strains ranging from 31% to 75% above the resting, neutral strain (data not shown; only mean values are presented in , ). The raw strain values measured during SMT ranged from 0.5% (contralateral C1/C2 break) to a maximum of 14.7% (ipsilateral C3/C4 break) for individual VAs (data not shown; only mean values are presented in , ). This large variability was also echoed by Johnson et al. 
Another assumption in this study is that the forces acting on the VA during SMT occurred in a linear, longitudinal fashion as opposed to a radial, transverse, or other 3-dimensional (eg, spiral) fashion. Because the VA loops backwards around the C1 transverse process in vivo, the results from failing it in a longitudinal direction ex vivo should be interpreted with caution. Johnson et al  tested their VA specimens both radially and longitudinally. They observed that the strains were significantly lower longitudinally (38.7%) than radially (59.4%). Based on those results, they speculated that longitudinal extension as measured in this study would likely be the primary cause of VA injury.
The clinical relevance of these results is equivocal, mainly because these were single, manipulative thrusts in a non-living subject. Although we can comment on the biomechanical properties of the VA, we cannot interpolate these results into a living system. For example, we cannot predict the results of repetitively stretching the VA in vivo to 6% strain over a period of time, nor can we comment on the development of microtears and so on in the walls of the VA. These questions are currently being pursued in our laboratory with the use of an animal model. [17, 23]
SMT resulted in strains sustained internally by the VA that were similar to those experienced during neck ROM testing and VBI screening. These strains were almost an order of magnitude lower than those required to mechanically disrupt the VA.
We conclude that under normal circumstances, a single, typical (high-velocity/low-amplitude) SMT thrust is very unlikely to tear or otherwise mechanically disrupt the VA.