J Stroke Cerebrovasc Dis. 2017 (Apr); 26 (4): 842–850 ~ FULL TEXT
J. David Cassidy, DC, PhD, DrMedSc, Eleanor Boyle, PhD, Pierre Côté, DC, PhD,
Sheilah Hogg-Johnson, PhD, Susan J. Bondy, PhD, and Scott Haldeman, MD, PhD
Department of Sports Science and Clinical Biomechanics,
Faculty of Health, University of Southern Denmark,
BACKGROUND: Chiropractic manipulation is a popular treatment for neck pain and headache, but may increase the risk of cervical artery dissection and stroke. Patients with carotid artery dissection can present with neck pain and/or headache before experiencing a stroke. These are common symptoms seen by both chiropractors and primary care physicians (PCPs). We aimed to assess the risk of carotid artery stroke after chiropractic care by comparing association between chiropractic and PCP visits and subsequent stroke.
METHODS: A population-based, case-crossover study was undertaken in Ontario, Canada. All incident cases of carotid artery stroke admitted to hospitals over a 9-year period were identified. Cases served as their own controls. Exposures to chiropractic and PCP services were determined from health billing records.
RESULTS: We compared 15,523 cases to 62,092 control periods using exposure windows of 1, 3, 7, and 14 days prior to the stroke. Positive associations were found for both chiropractic and PCP visits and subsequent stroke in patients less than 45 years of age. These associations tended to increase when analyses were limited to visits for neck pain and headache-related diagnoses. There was no significant difference between chiropractic and PCP risk estimates. We found no association between chiropractic visits and stroke in those 45 years of age or older.
CONCLUSIONS: We found no excess risk of carotid artery stroke after chiropractic care. Associations between chiropractic and PCP visits and stroke were similar and likely due to patients with early dissection-related symptoms seeking care prior to developing their strokes.
KEYWORDS: Stroke; risk factor; spinal manipulation; stroke prevention
From the FULL TEXT Article:
Neck pain and headache are common and related symptoms
in the general population and can cause considerable
health burden. [1, 2] Many individuals with these symptoms
attend chiropractors and family doctors.  Chiropractic
care usually involves manual manipulation of the cervical
spine,  and although there is evidence that
manipulation can improve neck pain and certain headaches,
there is concern that it might damage cervical arteries
and cause dissection-related stroke.  Indeed, there are multiple
case reports of both vertebrobasilar and carotid artery
dissection-related strokes occurring after cervical
manipulation.  This concern has prompted some neurologists
to warn against chiropractic manipulation of the
neck.  More recently, the American Heart and Stroke Associations
released a consensus statement concerning
cervical artery dissections associated with cervical spine
manipulation.  In their view, patients should be informed
of a statistical association between cervical
dissections and spinal manipulation prior to undergoing
manipulation of the cervical spine.
Although case reports can raise concerns and hypotheses
about risk, a study design with a control group is
required to test these hypotheses and quantify the risk.
There are several challenges in this respect. Internal carotid
artery dissection is a relatively rare event with an annual
incidence estimated at 1.72 per 100,000 population (95%
confidence interval [CI] 1.13-2.32).  Furthermore, the dissections
that are most likely to be diagnosed are those
that result in hospitalization for stroke.  To date, there
are no reported cases of stroke as an adverse event in
the published trials of cervical spine manipulation, but
these trials are too small to detect rare events. Although
about 12% of North American adults seek chiropractic
care annually,  it would require a very large cohort study
to accrue enough cases to investigate this problem. As
an alternative, the case–control study design is well suited
to address rare events, and 5 such studies have been published.
Two used Canadian health services data, [12, 13] and
the others used Californian stroke registries,  the Medicare
Advantage data from the United States,  and cervical
artery dissections seen at 18 neurology departments in
8 countries.  Three studies showed strong associations
between chiropractic care and vertebrobasilar artery (VBA)
stroke and 1 study found no association. Another study
showed an association between cervical manipulative
therapy and cervical dissections (i.e., defined as affecting
either the carotid artery, vertebral artery, or both
arteries).16 However, only 1 study included results specific
to carotid artery dissection strokes (n = 26), and they
were not associated with neck manipulation. 
Another challenge is the potential for protopathic bias. 
This occurs when an exposure (e.g., health care) is delivered
in the early prodrome of a disease (e.g., for
dissection-related neck pain or headache) before it is diagnosed
(e.g., before the dissection causes a symptomatic
ischemic event). In case–control studies, protopathic bias
can lead to the illusion that the exposure caused the
outcome, even though it is not on the causal pathway. 
Cassidy et al addressed this issue in their study by comparing
the association between both chiropractic services
and primary care physician (PCP) services prior to VBA
stroke.  They hypothesized that patients with dissectionrelated
neck pain and headache would attend both
chiropractors and PCPs prior to developing their strokes.
Furthermore, if associations were greater for chiropractors
than for PCPs, then chiropractic care would be
implicated as a cause of VBA stroke. Their results confirmed
strong associations between chiropractic services
and stroke in those less than 45 years of age, but similar
associations were seen for PCP services. They also did
an analysis limited to services coded for neck pain and/
or headaches that showed an increase in these associations.
This suggests that protopathic bias explains the link
between chiropractic care and VBA stroke.
Our study aims to investigate associations between chiropractic
exposures and carotid artery-related stroke and
compare them to PCP exposures in the same analyses. We
hypothesize that if chiropractic care increases the risk of
carotid stroke, associations between chiropractic visits and
stroke will exceed those between PCP visits and stroke. In
addition, we hypothesize that if associations between healthcare
visits and carotid stroke increase when analyses are
limited to visits provided for neck pain and headacherelated
diagnoses, protopathic bias is a likely explanation.
Study Design and Source Population
We conducted a population-based case-crossover study
using administrative healthcare data. In this design, cases
serve as their own controls by sampling control periods
before the index stroke date. This design is most appropriate
when a brief exposure (e.g., healthcare visit) causes
a transient change in risk (i.e., hazard period) of a rare
event (e.g., carotid stroke).  The within-person comparisons
provide better control for unmeasured risk factors
(e.g., potential confounding due to obesity, smoking, physical
activity, general health, etc.). The source population
was all adults, 18 years of age and older, residents in
the province of Ontario, Canada (population between 12
and 13 million during the study period), and eligible to
receive health care under the provincial health insurance
We used administrative data from the Discharge Abstract
Database (DAD) from the Canadian Institute for
Health Information, the Ontario Health Insurance Plan
(OHIP) database, and the Registered Persons Database
(RPDB). The DAD is a record of all hospital discharges,
and includes up to 8 discharge diagnoses coded using
the International Classification of Diseases, 9th Revision,
Clinical Modification (ICD-9-CM) to 4 digits only.
These diagnoses were used to identify stroke cases. OHIP
contains billing codes submitted for services rendered by
clinicians. At the time of this study, doctor of chiropractic
(DC) services were covered by OHIP. For each encounter,
clinicians submit the date, fee code(s), and an ICD code
for the responsible condition. The RPDB is a registry of
all individuals who have healthcare coverage under OHIP,
and it was used to identify the age and sex of our cases.
The University Health Network Research Ethics Board
approved our study (REB number 05-0533-AE).
All incident carotid artery stroke cases (ICD-9-CM: 433.1 — occlusion and stenosis of the carotid artery) discharged
from hospitals between April 1, 1993 and March
31, 2002 were eligible for the study. We included cases
with at least 1 year of healthcare coverage prior to the
date of the incident stroke. Cases who had a previous
hospital admission(s) with a discharge diagnosis of stroke
were excluded. We also excluded cases who had concurrent
stroke discharge diagnoses (i.e., ICD-9-CM 430,
431, 432.1, 432.9, 433.0, 433.2, and 435.0) because we could
not be sure if the main stroke was related to carotid artery
injury (Table 1). All decisions regarding codes were made
in consultation with stroke experts and epidemiologists
familiar with coding in Ontario. We further excluded cases
who were in a long-term care facility during the year prior
to their incident strokes, as these individuals would be
sicker than individuals living in the community and would
be less likely to access chiropractic services.
For descriptive purposes, we used the OHIP database
to extract all services related to comorbid conditions that
might be related to stroke during the year prior to the stroke
(i.e., hypertension, heart disease, peripheral vascular diseases,
diabetes, hypercholesterolemia, cerebrovascular
diseases, overweight, and addiction) and history of stroke
or transient ischemic attack.
We extracted all ambulatory DC and PCP fee codes from
OHIP for the year prior to the index stroke. We further
identified specific visits for neck pain or headacherelated
services and redid our analyses to investigate
protopathic bias. For chiropractors, we identified neck or
headache-related encounters using their unique diagnostic
codes (i.e., C01-C06 cervical and cervicothoracic
subluxation, C13-C15 multiple site subluxation, C30 cervical
sprain/strain, C40 cervical neuritis/neuralgia, C44
arm neuritis/neuralgia, C50 brachial radiculitis, C51 cervical
radiculitis, and C60 headache). For the PCP visits,
we identified neck or headache-related encounters by ICD-9
codes (i.e., 307 tension headaches; 346 migraine headaches;
722 intervertebral disc disorders; 780 headache except
tension headache and migraine; 729 fibrositis, myositis,
and muscular rheumatism; and 847 whiplash sprain/
strain and other traumas associated with neck).
Using a time-stratified approach, 4 control periods were
randomly chosen during the year prior to the stroke for
each case.  The year was divided into disjoint strata with
2-week intervals between strata because chiropractic care
is often delivered in episodes of care, and a 2-week separation
in sampling would limit overlap bias associated
with time trends in this exposure.  The control periods
were matched to exposure windows of 1, 3, 7, and 14
days, depending on the hazard period under examination.
We used conditional logistic regression to estimate associations
between stroke and healthcare visits. Separate
models were built for all visits and for neck or headacherelated
visits for each different hazard period. A 4-level
exposure variable was created: no exposure as the referent,
DC only visit, PCP only visit, and both DC and
PCP visits. We excluded PCP visits that occurred on the
day of the stroke because individuals may have seen their
PCPs after their strokes, but prior to hospital admission.
Conversely, DC visits on the day of the stroke were
included in our analyses.
We built separate models for younger (age <45 years)
and older (age ?45 years) cases based on previous studies
indicating an increased risk of stroke after chiropractic
care in those less than 45 years of age. [12, 13] Separate models
were also built for 1-, 3-, 7-, and 14-day exposure periods.
We report our results as odds ratios (ORs) with their 95%
CIs and bias-corrected bootstrapped CI. We performed
contrasts to test whether the ORs for DC visits were significantly
different from the comparable ORs for PCP visits
(i.e., null hypothesis of equality). All analyses were done
using STATA/SE version 12.1. 
A total of 15,523 carotid artery stroke cases met our
inclusion/exclusion criteria. The mean age of the cases
was 69.9 years and 61.4% were male. Only 214 (1.4%)
of cases occurred in those less than 45 years of age. There
were a lower proportion of men in the younger age group
compared with the older age group (48.6% versus 61.6%,
respectively). In the 14 days preceding their strokes, 5,433
cases (35%) had received only PCP services, 186 cases
(1.2%) had received only DC services, and 116 cases (.7%)
had received both services. As expected, older cases had
received more services coded for comorbid conditions,
and those who saw a DC tended to have fewer
comorbidities (Table 2).
Overall, there were few cases exposed to chiropractic
care prior to their strokes. Six cases less than 45 years
of age saw a chiropractor within 14 days of their strokes,
compared to 70 cases who saw a PCP (Table 3). With such
few exposed cases in the DC group, it was not possible
to bootstrap 95% CI for all estimates (Tables 4 and 5).
When considering all DC and PCP visits in those less
than 45 years of age, there is an association with stroke
for both groups (Table 4). However, there is no significant
difference between PCP and DC estimates (i.e., fail
to reject the null hypothesis of similar effects for either
provider type). There were no younger cases who saw
both practitioners less than 7 days before their strokes,
and too few cases saw both practitioners in the 7- and
14-day exposure periods to calculate the estimate. For stroke
cases 45 years of age and older, there is no positive association
with chiropractic care, or with the combination
of DC and PCP care. However, there is a consistent but
weak association with PCP care in the older group.
When we restricted our analyses to visits that were coded
for neck pain and/or headache, the ORs increased considerably
for all DC and PCP exposures in those less than
45 years of age, with 1 exception (Table 5). For the exposure
period of 3 days before the stroke, both the DC
and the PCP estimates were about the same as those for
all services for the same exposure period. Again, there
is no significant difference between DC and PCP estimates.
There were no younger cases who saw both a DC
and a PCP. For those more than 45 years of age, the estimates
were very similar to those seen when all service
visits were considered (Table 5). The odds of seeing a
chiropractor prior to stroke remained below 1 across exposure
periods whereas the odds of seeing a PCP were
above 1. There were no older cases seen by both practitioners
within 1 day of their stroke, and for the other
exposure periods, the odds of seeing both practitioners
indicate no association with stroke.
Our study is the first population-based, controlled study
to address the risk of carotid artery strokes after chiropractic
care. Using a case-crossover methodology, we have
shown an increased association between DC and PCP visits
and subsequent hospitalization for strokes coded as occlusion
and stenosis of the carotid artery in those less
than 45 years of age. Although the point estimates are
different for DC and PCP visits, there is no statistical difference
between them. Furthermore, these associations
increased when analyses were limited to service codes
for neck pain and headache-related diagnoses. Taken together,
our results suggest that the association between
chiropractic care and carotid artery stroke is explained
by protopathic bias. In other words, younger patients with
an impending carotid artery stroke could be seeking care
for dissection-related pain in the head and neck prior to
developing stroke. Under this scenario, any care provided
by chiropractors or PCPs is coincidental to the stroke
and not on the causal pathway.
Another potential explanation of our results is that both
DC and PCP care increase the risk of these strokes.
However, because PCPs do not usually manipulate the
cervical spine, or provide other care that significantly increases
the risk of stroke in young people, this explanation
is unlikely. It is likely that the weak associations seen
between PCP care and stroke in older individuals are due
to comorbid disease, as sicker, older individuals are more
likely to consult a PCP than a DC.
To our knowledge, there is one other controlled study
assessing the association between chiropractic care and
carotid dissection-related stroke. Smith et al used 2 Californian
academic stroke registries to investigate the
relationship between chiropractic treatment and both
vertebrobasilar and carotid artery dissection-related
strokes. [14, 23] Twenty-six carotid dissection-related strokes
or transient ischemic attacks were compared to 100 other
non–dissection-related strokes using a case–control study
design. Although cases were more likely to complain of
pain before their ischemic event (OR 4.7; 95% CI 1.7-
13.0), the authors reported no significant association with
previous cervical spine manipulation. However, the small
sample size would have limited their ability to test this
Headache and neck pain are common presenting symptoms
in patients with cervical artery dissection, [9, 24] and
in some cases are the only presenting symptoms.  They
are also common and recurrent in the general
population. [26, 27] Although some ischemic events are preceded
by sudden intense neck and/or head pain, in many
cases it is less sudden and severe and likely indistinguishable
from less serious causes. [24, 28] In the absence of
neurological signs and symptoms, there are no practical,
clinically valid screening tests to identify underlying
dissections in patients with head or neck pain.  This leaves
clinicians who treat these conditions vulnerable to misdiagnosis,
providing inappropriate treatment and
subsequent malpractice lawsuits.  Fortunately, internal
carotid artery dissection is rare,  but this makes it difficult
to study and a challenge to identify in the absence
of neurological signs.
A strength of our study is the case-crossover design,
which allows better control of time-independent confounding
factors, both known and unknown, than the
traditional case–control method. As many stroke risk
factors are not captured in health administrative data
(e.g., smoking, obesity, physical inactivity, genetic susceptibility,
and undiagnosed hypertension and connective
tissue disorders), our self-controlled design compensates
for this. We also measured DC and PCP exposures
independent of case definition by merging separate databases,
which would limit diagnostic selection bias. 
We addressed the issue of protopathic bias by including
PCP exposures in our analysis as a measure of the
background risk of becoming a case. Our subgroup analysis,
which limited visits to diagnostic codes related to
conditions that would cause head or neck pain, suggests
the presence of protopathic bias. We also found
very few cases who had seen both DCs and PCPs during
exposure periods, thus limiting misclassification of exposures.
We also excluded from our analyses cases who
presented to PCPs the day of their stroke hospitalization,
but included those who had been seen by a
chiropractor. As a result, our findings include acute onset
stroke after seeing a chiropractor, but exclude cases seen
by a PCP on the day of their strokes. Finally, our study
base includes the entire population of Ontario, Canada,
over a 9-year period, representing 109,020,875 person-years
of observation, and the results should be generalizable
to other populations where chiropractic treatment is offered.
Nevertheless, because of the small number of exposed
cases, we could not bootstrap all our CIs.
There are also limitations with using administrative data.
We can not be certain that all chiropractic visits resulted
in manipulation to the cervical spine. We were able
to exclude visits coded for radiographic examination, and
more than 80% of patients seeing chiropractors in Ontario
receive spinal manipulation.  However, there is potential
for visits to be misclassified as cervical spinal
manipulation exposures that are not, which could result
in an underestimation of the DC–stroke association. In
addition, the positive predictive value of our case definition
of carotid stroke is not known, raising the potential
of misclassification of strokes.  However, any
misclassification would be nondifferential across exposure
groups, which would equally attenuate risk estimates
in the PCP and DC groups.  Several limitations might
bias our estimates in favor of an increased association
between stroke and chiropractic care. These include excluding
PCP visits on the same day of the stroke from
our analysis and diagnostic work-up bias in cases that
present to hospital after chiropractic care.  Alternatively,
there is some evidence that recent infection is
associated with cervical artery dissection, [33, 34] and patients
with infection might be more likely to consult their
PCPs than a DC. If this were the case, our estimates of
the association between stroke and PCPs would be elevated
by selection bias. Finally, our results are based on
a small proportion of exposed cases, and this is reflected
in our wide CIs.
In conclusion, our study suggests that the association
between chiropractic care and carotid artery stroke could
be due to care being delivered for dissection-related
neck pain and/or headache, prior to the ischemic event.
However, stroke is a serious disorder, and all practitioners
treating patients with neck pain and headache
should be aware that it could occur. Although these
events are rare, they can result in serious impairment
Hagen K, Einarsen C, Zwart JA, et al.
The co-occurrence of headache and musculoskeletal symptoms amongst 51,050 adults in Norway.
Eur J Neurol 2002;9:527-533.
Martin BI, Deyo RA, Mirza SK, et al.
Expenditures and Health Status Among Adults With Back and Neck Problems
JAMA 2008 (Feb 13); 299 (6): 656–664
Chevan J, Riddle DL.
Factors associated with care seeking from physicians, physical therapists, or chiropractors by persons with spinal pain: a population-based study.
J Orthop Sports Phys Ther 2011;41:467-476.
Hurwitz EL, Coulter ID, Adams AH, et al.
Use of chiropractic services from 1985 through 1991 in the United States and Canada.
Am J Public Health 1998;88:771-776.
Cassidy JD, Bronfort G, Hartvigsen J.
Should we abandon cervical spine manipulation for mechanical neck pain? No.
Haldeman S, Carey P, Townsend M, Papadopoulos C.
Arterial Dissections Following Cervical Manipulation: The Chiropractic Experience
Canadian Medical Association Journal (CMAJ) 2001 2001 (Oct 2); 165: 905–906
Neurologists warn about link between chiropractic, stroke.
Biller J, Sacco RL, Albuquerque FC, et al.:
Cervical Arterial Dissections and Association With Cervical Manipulative Therapy
A Statement for Healthcare Professionals From the American Heart Association/
American Stroke Association
Stroke. 2014 (Oct); 45 (10): 3155–3174
Lee VH, Brown RD Jr, Mandrekar JN, et al.
Incidence and outcome of cervical artery dissection: a populationbased study.
Spontaneous dissection of the carotid and vertebral arteries.
N Engl J Med 2001;344:898-906.
Hurwitz EL, Chiang LM.
A comparative analysis of chiropractic and general practitioner patients in North America: findings from the joint Canada/United States Survey of Health, 2002-03.
BMC Health Serv Res 2006; 6:49.
Rothwell DM, Bondy SJ, Williams JI.
Chiropractic manipulation and stroke: a population-based case-control study.
Cassidy JD, Boyle E, Cote P, et al.
Risk of Vertebrobasilar Stroke and Chiropractic Care: Results of a Population-based
Case-control and Case-crossover Study
Spine (Phila Pa 1976) 2008 (Feb 15); 33 (4 Suppl): S176–183
Smith WS, Johnston SC, Skalabrin EJ, et al.
Spinal manipulative therapy is an independent risk factor for vertebral artery dissection.
Kosloff TM, Elton D, Tao J, Bannister WM.
Chiropractic Care and the Risk of Vertebrobasilar Stroke: Results of a Case-control Study
in U.S. Commercial and Medicare Advantage Populations
Chiropractic & Manual Therapies 2015 (Jun 16); 23: 19
Engelter ST, Grond-Ginsbach C, Metso TM, et al.
Cervical artery dissection: trauma and other potential mechanical trigger events.
Horwitz RI, Feinstein AR.
The problem of “protopathic bias” in case-control studies.
Am J Med 1980;68: 255-258.
Korhonen MJ, Huupponen R, Ruokoniemi P, et al.
Protopathic bias in observational studies on statin effectiveness.
Eur J Clin Pharmacol 2009;65:1167-1168.
The case-crossover design: a method for studying transient effects on the risk of acute events.
Am J Epidemiol 1991;133:144-153.
Mittleman MA, Maclure M, Robins JM.
Control sampling strategies for case-crossover studies: an assessment of relative efficiency.
Am J Epidemiol 1995;142:91-98.
Janes H, Sheppard L, Lumley T.
Case-crossover analyses of air pollution exposure data: referent selection strategies and their implications for bias.
Stata base reference manual. 12th ed.
College Station, Texas: Stata Press, 2011.
Chung CL, Cote P, Stern P, et al.
The association between cervical spine manipulation and carotid artery dissection: a systematic review of the literature.
J Manipulative Physiol Ther 2015;38:672-676.
Debette S, Leys D.
Cervical-artery dissections: predisposing factors, diagnosis, and outcome.
Lancet Neurol 2009;8:668-678.
Arnold M, Cumurciuc R, Stapf C, et al.
Pain as the only symptom of cervical artery dissection.
J Neurol Neurosurg Psychiatry 2006;77:1021-1024.
Côté P, Cassidy JD, Carroll LJ, et al.
The annual incidence and course of neck pain in the general population: a population-based cohort study.
Boardman HF, Thomas E, Croft PR, et al.
Epidemiology of headache in an English district.
Thomas LC, Rivett DA, Attia JR, et al.
Risk factors and clinical features of craniocervical arterial dissection.
Man Ther 2011;16:351-356.
Kerry R, Taylor AJ, Mitchell J, et al.
Cervical arterial dysfunction and manual therapy: a critical literature review to inform professional practice.
Man Ther 2008;13:278-288.
Boyle, E, Grier, AR, and Cassidy, JD.
Examining Vertebrobasilar Artery Stroke in Two Canadian Provinces
Spine (Phila Pa 1976). 2008 (Feb 15); 33 (4 Suppl): S170–175
Cai X, Razmara A, Paulus JK, et al.
Case misclassification in studies of spinal manipulation and arterial dissection.
J Stroke Cerebrovasc Dis 2014;23:2031-2035.
Cassidy JD, Cote P. Re: Cai et al,
Case misclassification in studies of spinal manipulation and arterial dissection (letter).
J Stroke Cerebrovasc Dis 2015;24:901-902.
Guillon B, Berthet K, Benslamia L, et al.
Infection and the risk of spontaneous cervical artery dissection: a case-control study.
Rubinstein SM, Peerdeman SM, van Tulder MW, et al.
A systematic review of the risk factors for cervical artery dissection.
Return to STROKE AND CHIROPRACTIC