Large Cohort Study Finds No Increased Stroke Precursor Risk After Chiropractic Neck Manipulation

Adults who received chiropractor-administered spinal manipulation of the neck were not at increased risk of cervical artery dissection compared with similar patients who received an ibuprofen prescription, according to a large retrospective cohort study published in Chiropractic & Manual Therapies. The analysis drew on more than a decade of United States health records and claims data and included 431,492 patients in each of two matched cohorts, all of whom had sought care for a new episode of neck pain or headache. Cervical artery dissection, a tear in the wall of the carotid or vertebral artery that can lead to embolic stroke, has long been feared as a rare complication of neck manipulation, but the new study found the opposite direction of association: the risk of dissection within 30 days was actually lower in the manipulation group, and the authors attribute that small apparent reduction to residual confounding rather than a protective effect.

The study was designed to address a persistent methodological problem in this field, known as protopathic bias. Neck pain and headache are not only common reasons for seeking chiropractic care; they are also common early symptoms of cervical artery dissection itself. When case-control and case-crossover studies compare manipulation exposure between patients who already have a dissection and those who do not, they may mistakenly conclude that manipulation caused the dissection, when in fact the dissection may have prompted the patient to seek treatment in the first place. Earlier case series had suggested that dissection could rarely occur soon after spinal manipulative therapy, and a recent meta-analysis found no significant association between the therapy and dissection, but the authors note that the compiled evidence was judged to be of very low certainty, hampered by risk of bias, inconsistency, indirectness and imprecision.

To sidestep these biases, the research team, led by Robert J. Trager of University Hospitals in Cleveland with collaborators spanning orthopedics, physical therapy, neurosurgery, cardiology, biostatistics and chiropractic, queried the TriNetX Linked Network, a resource combining electronic health records with medical, pharmacy and mortality claims from more than 100 United States payers and 22 large healthcare organizations. The dataset covered records from 2010 to 2025 and contained nearly 20 million unique patients. The manipulation cohort consisted of adults with a new episode of neck pain or headache who received chiropractor-administered cervical spinal manipulation, identified through a combination of specific procedure codes and a diagnosis code indicating segmental dysfunction of the cervical spine. The comparison group consisted of patients with similar new complaints who received an outpatient visit with an ibuprofen prescription on the same date, chosen as an active comparator because nonsteroidal anti-inflammatory drugs are among the most commonly prescribed medications for neck pain and headache, and because ibuprofen has not been identified as an established risk factor for cervical artery dissection.

The eligibility criteria were deliberately strict. Patients with any history of cerebrovascular disease, prior cervical artery dissection or cervical artery surgery, serious spinal pathology, recent emergency care, recent anterior neck or spine surgery, critical care or inpatient care in the preceding two weeks were excluded. To define a genuinely new episode of care, patients with any eligible neck pain or headache diagnosis in the 30 days before the index visit were also excluded. Each patient was included only once, with the first qualifying encounter defining cohort assignment. Propensity score matching was then used to balance the two groups on demographics, comorbidities, medications, and markers of healthcare complexity such as inpatient and emergency visits and cardiology procedures, assessed over a one-year lookback period. After matching, both cohorts contained 431,492 patients with similar mean ages of about 43 years and a female predominance of roughly 65 to 66 percent, and all standardized mean differences for covariates fell below the 0.1 threshold for imbalance.

The primary outcome was a new diagnosis of cervical artery dissection, coded specifically as carotid or vertebral artery dissection, within a 30-day follow-up window beginning at the index date. The authors chose this window because prior research suggests symptoms of dissection after neck manipulation can begin within a week but patients may delay seeking care for up to two weeks, because diagnosis is often delayed with a reported mean delay of nine days, and because shorter windows in real-world data may miss events due to miscoding or billing lags. The positive predictive value of the diagnostic coding used has been estimated at 90 percent. Importantly, the study isolated cervical artery dissection as its own outcome rather than using composite outcomes of dissection, arterial injury and stroke, a choice the authors argue is essential because only about 56 percent of cervical artery dissections result in stroke, and the risk factors for dissection differ from those for stroke, being more associated with younger age, migraine and minor trauma, and less with traditional atherosclerotic risk factors.

The results showed that 31 patients in the manipulation cohort, or 0.007 percent, and 58 patients in the ibuprofen cohort, or 0.013 percent, had a new cervical artery dissection during the 30-day follow-up. The risk ratio comparing manipulation to ibuprofen was 0.53, with a 95 percent confidence interval of 0.35 to 0.83 and a P value of 0.004, and the absolute risk difference was minus 6.3 events per 100,000 patients, with a confidence interval of minus 10.8 to minus 2.0. Carotid artery dissection risk was significantly lower in the manipulation group, with a risk ratio of 0.40, while vertebral artery dissection risk was not significantly different, with a risk ratio of 0.67 and a confidence interval crossing the null. A post hoc time-to-event analysis using Kaplan-Meier and Cox proportional hazards methods produced hazard ratios nearly identical to the primary estimates, and no violation of the proportional hazards assumption was detected. Cumulative incidence curves for both cohorts rose sharply in the first two weeks and then began to plateau, with some events occurring on the index date itself, a pattern the authors interpret as a transiently elevated baseline risk of dissection among adults presenting with new neck pain or headache, irrespective of the treatment received.

The authors are careful about how the apparent risk reduction should be read. They frame the findings as conceptually more consistent with a non-inferiority framework than a superiority framework, since the study provides no evidence of increased risk rather than evidence of genuine benefit. Unmeasured confounding, they argue, likely explains the small apparent reduction. Chiropractors may selectively avoid neck manipulation in patients they perceive to be at higher cerebrovascular risk based on clinical findings or family histories that are not well captured by diagnosis codes. Genetic predispositions linked to dissection, such as certain methylenetetrahydrofolate reductase polymorphisms, were not reliably coded in the data, and fibromuscular dysplasia, which was matched on, is often underdiagnosed. Minor undiagnosed cervical trauma, higher education levels linked to greater dissection diagnosis, and differences in workup or referral strategies are among other potential confounders the authors enumerate. An E-value analysis suggested that an unmeasured confounder associated with both the exposure and the outcome by a risk ratio of 3.18 could move the upper confidence limit to include the null.

Several data quality checks supported the internal validity of the analysis. Loss to follow-up was minimal, at less than 3 percent, and complete follow-up was achieved for 98.5 percent of the manipulation cohort and 97.1 percent of the ibuprofen cohort. Negative control outcomes, procedures selected because they have no plausible link to either manipulation or dissection but share a similar confounding structure of general healthcare engagement, were partially balanced: colonoscopy and lower extremity magnetic resonance imaging showed balanced risk ratios, while appendectomy and foreign body removal were imbalanced, which the authors caution should not be over-interpreted given their exploratory nature. The manipulation cohort averaged 3.6 chiropractic visits over the 30-day window, consistent with utilization patterns documented in large national cohorts and with guideline recommendations for headache care, and crossover between treatments was modest, with 5.11 percent of the ibuprofen cohort receiving manipulation after day 7 and 1.87 percent of the manipulation cohort receiving ibuprofen.

The study also carries important limitations that the authors discuss at length. As an observational analysis of administrative data, it cannot fully exclude an association between manipulation and dissection, and the accuracy of the exposure definition could not be validated without chart review, since some chiropractors may have documented the cervical dysfunction code while manipulating only non-cervical regions, or non-manipulative procedures may have been miscoded as manipulation. The dataset’s aggregate nature prevented formal imputation of missing values, dose-response analysis, and stratification by manipulation technique, force or patient positioning. Over-the-counter ibuprofen use was not captured, and the design did not restrict inclusion to patients who were naive to either treatment. The 30-day washout for prior neck pain and headache diagnoses cannot guarantee the complete absence of antecedent symptoms, and some patients may have had subclinical or unreported neck pain before the index visit.

The authors conclude that among adults seeking care for a new episode of neck pain or headache, there is no evidence of an increased risk of cervical artery dissection within 30 days following chiropractor-administered cervical spinal manipulation compared with an ibuprofen prescription, and that observed incidence patterns tentatively suggest an elevated incidence of dissection among these patients irrespective of the care they receive. They nonetheless emphasize that all clinicians, including chiropractors, should maintain vigilance for symptoms of cervical artery dissection, and they call for further research on higher-risk subgroups such as patients with migraine, connective tissue disorders or pregnancy, on the development of validated prediction models to help clinicians distinguish dissection from benign causes of neck pain and headache, and on whether spinal manipulation is associated with stroke, an outcome that could not be reliably assessed in this study.

Subject of Research: The association between chiropractor-administered cervical spinal manipulation and cervical artery dissection risk in adults with new neck pain or headache

Article Title: Association between spinal manipulative therapy and cervical artery dissection in patients with neck pain or headache: a retrospective cohort study

Article References: Trager, R. J., Baumann, A. N., Labak, C. M., Lewis, K. L., Perez, J. A., Goertz, C. M., & Schuster, A. (2026). Association between spinal manipulative therapy and cervical artery dissection in patients with neck pain or headache: a retrospective cohort study. Chiropractic & Manual Therapies, 34(1), Article 53. https://doi.org/10.1186/s12998-026-00678-4

Image Credits: AI Generated

DOI: 10.1186/s12998-026-00678-4

Keywords: cervical artery dissection, spinal manipulative therapy, chiropractic, neck pain, headache, stroke, propensity score matching, retrospective cohort study, TriNetX, ibuprofen, vertebral artery dissection, protopathic bias