Skip to main content

Archived Comments for: Chiropractic care and the risk of vertebrobasilar stroke: results of a case–control study in U.S. commercial and Medicare Advantage populations

Back to article

  1. Case misclassification renders this study uninformative for the manipulation/dissection debate

    David Thaler, Tufts Medical Center

    30 July 2015

    Cassidy and colleagues, studying a Canadian population, showed an association between posterior circulation stroke and spinal manipulative therapy (SMT) and between posterior circulation stroke and visits to primary care physicians (1). Kosloff et al have repeated the Cassidy study in an American population so to generalize the original Canadian observations. We have shown that the ICD-9 codes used by Cassidy to define cases have a positive predictive value for cervical arterial dissection (CAD) of only around 11% when compared to neurologist review of the medical record and neuroimaging (2). A small fraction of the “cases” identified by Cassidy’s method had the actual disease of interest (CAD, with or without stroke). Instead they likely had posterior circulation disease of a more common sort (lacunes, atherosclerosis, embolism). Cassidy’s case-identification strategy induced significant outcome misclassification that biased estimates of the SMT-CAD association towards the null.

    Kosloff and colleagues have made the same mistake but then compounded the error by systematically excluding patients with dissection from their population. In Ontario, at the time of the Cassidy study, ICD-9 codes specific for dissection (443.XX) were not in use. Patients in Canada, clinically diagnosed with dissection, would have been coded with a posterior circulation code included in the Cassidy study, and so would have met inclusion criteria for his study. In the US, the dissection-specific codes are in widespread use. Patients with clinical CAD diagnoses would have been most accurately coded with a 443.XX code and not with the anatomically-based posterior circulation codes. Kosloff did not include the dissection codes as part of the case definition and so patients with CAD were systematically excluded. It follows that the positive predictive value of the Kosloff CAD identification strategy is likely to be even lower than that which was observed by Cassidy.

    Prior studies suggest vascular risk factors are not associated with CAD risk (3). That Kosloff’s subjects had non-CAD-related stroke is suggested by the higher prevalence of cardiovascular risk factors in the case group (Table 3). The depletion of dissections from the Kosloff population also explains why no association between SMT and case-status was observed even in the younger (<45y) group – a finding which has been consistently seen in other studies, including Cassidy’s, with statistically and clinically significant odds ratios.

    Kosloff et al define their cases as “vertebrobasilar stroke” but then draw inferences about CAD. This is an error in logic that ignores the heterogeneity of stroke subtypes. In addition to outcome misclassification, epidemiologic studies of SMT as a trigger of CAD face other methodological challenges – low CAD incidence rates, confounding, reverse causation, misclassification of SMT exposure (including recall bias), and selection bias. Rigorously designed studies that address these hurdles are needed to provide quality evidence to resolve this critical public health question. Unfortunately, Kosloff’s study does not provide such data.


    (1)    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 2008; 33(4 Suppl):S176-S183.

    (2)    Cai X, Razmara A, Paulus JK, Switkowski K, Fariborz PJ, Goryachev SD, D’Avolio L, Feldmann E, Thaler DE. Case Misclassification in Studies of Spinal Manipulation and Arterial Dissection. Journal of Stroke and Cerebrovascular Diseases, 2014, 23(8): 2031-2035, doi: 10.1016/j.jstrokecerebrovasdis.2014.03.007

    (3)    Sidney M. Rubinstein, Saskia M. Peerdeman, Maurits W. van Tulder, Ingrid Riphagen, and Scott Haldeman. A Systematic Review of the Risk Factors for Cervical Artery Dissection. Stroke. 2005; 36:1575-1580, doi:10.1161/01.STR.0000169919.73219.30


    DE Thaler and J Paulus

    Competing interests

    Expert testimony.
  2. Authors response to comments

    Thomas Kosloff, Optum

    24 August 2015

    We thank Drs. Thaler and Paulus for their thoughtful comments regarding our study ‘Chiropractic care and the risk of vertebrobasilar stroke: results of a case–control study in U.S. commercial and Medicare Advantage populations’ [1]. We agree, as acknowledged in the paper, the potential for misclassification bias represents a study limitation. Although we concur the reporting of cervical artery dissection (CAD) diagnostic codes is more common in the United States, our study was focused on the clinical diagnosis – vertebrobasilar artery system (VBA) stroke – viewed as a ‘major’ adverse event in association with chiropractic care [2].

    We further agree that future studies should be designed to provide quality evidence to resolve this critical public health question. Accordingly, investigations into the associations between exposures to a wider range of health care professionals, as well as exposures to different interventions, and CAD with or without stroke are currently in progress. 


    [1] 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. Chiropr Man Therap 2015 Jun 16;23:19

    [2] Carnes D, Mullinger B, Underwood M. Defining adverse events in manual therapies: A modified Delphi consensus study. Man Ther 2010;15:2-6

    Competing interests