diabetes) was significantly higher among cases as compared to controls, and we are concerned that these differences were non-random.
In another case control study, Thomas et al. compared the records of 47 patients with confirmed or suspected vertebral or internal carotid artery dissection with 43 controls [10] . They limited their analysis to young patients defined as <55 years. These authors report a significant association between dissection and recent head or neck trauma (OR 23.51, 95% CI 5.71-96.89) as well as neck manual therapy (OR 1.67, 95% CI 1.43-112.0). An inconsistent standard for case ascertainment (a significant number of patients lacked radiographic confirmation of dissection) and lack of blinding weaken this study. Engelter et al. evaluated data from the Cervical Artery Dissection and Ischemic Stroke Patients (CADISP) consortium, identifying 966 patients with CAD, 651 with stroke attributable to another cause, and 280 healthy controls [7] . The CADISP study involved both prospectively and retrospectively collected data at multiple centers in several countries. They assessed for prior cervical trauma within one month using questionnaires administered during clinic visits. Cervical manipulation therapy was more common for CAD versus stroke from another cause (OR 12.1, CI 4.37-33.2). The report notes that an association between any trauma and CAD was present even when restricting the analysis to prospectively recruited patients. However, in patients to whom the questionnaire was administered after dissection, recall bias may have been at work whether or not the patient was enrolled prospectively. Indeed, the frequency of prior cervical trauma in this study was substantially higher than previous reports (40% versus 12-34%). Additional weaknesses include a highly heterogeneous standard for case definition and no clear masking procedures. Body of evidence quality (GRADE rating) Having performed a systematic review and rated articles according to their individual strengths and weaknesses, we graded the overall body of evidence using the system proposed by Guyatt et al. [17-20] . The GRADE approach to rating quality of evidence proposes four categories that are applied to a body of evidence: high, moderate, low, and very low. In the setting of systematic review, a particular rating reflects the extent of confidence that the estimates of effect are correct. The GRADE approach begins with study design and sequentially examines features with the potential to enhance or diminish confidence in the meta-analytic estimate of effect size. Our final assessment of the quality of the body of evidence using these criteria was very low. The initial rating based on study design was low (observational studies). Given the controversial nature of this topic and the legal ramifications of results, there is certainly potential for bias (-1 serious). However, blinding in the Class II studies mitigated this risk to some extent. Inconsistency and imprecision did not lower our rating. Because the body of evidence is derived from measures of association, the rating was lowered for indirectness (-1 serious). Publication bias is less likely because of the impact of a negative result in this case. The funnel plot from our meta-analysis was inconclusive with regard to possible publication bias because of the small number of studies included but suggested a deficit in the publication of small negative trials. There was not a large effect size, and currently there is no evidence for a dose response gradient. Moreover, the most worrisome potential confounder (neck pain) would increase rather than reduce the hypothesized effect. Discussion The results of our systematic review and meta-analysis suggest a small association between chiropractic care and CAD. There are no class I studies addressing this issue, and this conclusion is based on five class II and III studies. Scrutiny of the quality of the body of data
2016 Church et al. Cureus 8(2): e498. DOI 10.7759/cureus.498
6 of 11
Made with FlippingBook HTML5