We repeated the meta-analysis excluding class III studies. The combined effect size was again indicative of a small association between dissection and chiropractic care, OR 3.17, 95% CI 1.30-7.74). The result was identical when using a random effects model.
Class II studies Smith et al. used a retrospective case control design, combining databases from two academic stroke centers to identify cases of arterial dissection [9] . They found 51 cases and 100 controls. Exposure to spinal manipulative therapy (SMT) was assessed by mail survey. The authors reported an association between SMT and VBA (P = .032). In multivariate analysis, chiropractor care within 30 days was associated with VBA, even when adjusting for neck pain or headache (OR 6.6, 95% CI 1.4-30). While this study controlled for possible confounders such as neck pain, there were several limitations. Head and neck pain as well as chiropractor visit were assessed in a retrospective fashion by mail survey, very possibly introducing both recall and survivor bias. The reason for reporting to the chiropractor (e.g., trauma) was not assessed. Further, there was significant variability among diagnostic procedures, which may reflect increased motivation by physicians to rule out dissection in patients with a history of SMT. Such motivation could result in interviewer bias. Dittrich et al. compared 47 patients with CAD to a control group with stroke due to etiologies other than dissection [6] . They assessed for risk factors using a face-to-face interview with blinding. These authors found no association between any individual risk factor and CAD, including cervical manipulative therapy. They blame the small sample size for the negative result, and they point out that cumulative analysis of all mechanical risk factors <24 hours prior to symptom onset showed an association (P = .01). This study is subject to recall bias. Class III studies Rothwell et al. used a retrospective case control design to test for an association between chiropractic manipulation and vertebrobasilar accidents (VBA) [8] . They reviewed Ontario hospital records for admissions for VBA from 1993–1998. There were 582 cases and 2328 matching controls. The authors report an association between VBA and visit to a chiropractor within one week (OR 5.03, 95% CI 1.32-43.87), but this was only true for young patients (<45 years). This study represented the first attempt to delineate the association between chiropractic manipulation and extremely rare VBA with controls. Limitations included requisite use of ICD-9 codes to identify cases and associated classification bias, as well as potential unmeasured confounders (e.g., neck pain). In 2008, Cassidy et al. set out to address the problem of neck pain possibly confounding the association between chiropractic care and VBA [5] . Again using a retrospective case control design, they included all residents of Ontario over a period of 9 years (1993–2002, 109,020,875 person years of observation). They identified 818 VBA strokes resulting in hospitalization and randomly selected age and sex matched controls. Next, they examined ambulatory encounters with chiropractors and primary care physicians (PCPs) in the one year preceding the stroke, limited to cervical manipulation, neck pain, and headache. Associations between chiropractor visit and VBA versus PCP visits and VBA were compared. Indeed, there were associations between both chiropractor visit and VBA (<45yrs OR 1.37, 95% CI 1.04-1.91), and PCP visit and VBA (<45 yrs OR 1.34, 95% CI .94-1.87; >45 yrs and OR 1.53, 95% CI 1.36-1.67). The association for chiropractor visit was not greater than for PCP visit. This data was interpreted as evidence that a confounder such as neck pain may account for the association between chiropractor visit and VBA. This study was subject to many of the same limitations as previous efforts. Canadian health records would not reveal whether a patient with cervical complaints underwent cervical manipulation, and the researchers could not review each chart for imaging confirming dissection. Additionally, the incidence of comorbidities (e.g., hypertension, heart disease,
2016 Church et al. Cureus 8(2): e498. DOI 10.7759/cureus.498
5 of 11
Made with FlippingBook HTML5