A Systemic Review of the Risk Factors for Cervical Artery Dissection
To the Editor:
We read with interest the review by Rubinstein et al1 on risk factors for cervical artery dissection. In a case-control study, we had analyzed the role of recent infection on cervical artery dissection.2 This study was referenced by the authors; however, our main findings were not correctly cited by the authors. In univariate analyses, recent infection and high-social status were significantly more common, and smoking was significantly less common in patients with cervical artery dissection (CAD) than in patients with cerebral ischemia of other origin. In conditional logistic regression analysis, infection within 1 week (odds ratio, 2.87; 95% CI, 1.18 to 7.00) and high-social status (odds ratio, 6.54; 95% CI, 1.88 to 22.7) remained significantly associated with CAD. Because coughing, sneezing, and vomiting that often occur during infection could explain the association between CAD and acute infectious disease, we systematically assessed the frequency of these mechanical factors. Recent cough, sneezing, and vomiting tended to be reported more often by CAD patients (60.5%) than by control patients (41.4%; P=0.06). In multivariate analysis, infection within 1 week, (odds ratio, 2.42; 95% CI, 1.01 to 5.8; P=0.046) but not cough, sneezing, or vomiting (odds ratio, 1.60; 95% CI, 0.67 to 3.8; P=0.29), was associated with CAD. This indicates that mechanical factors during infection do not explain the association between CAD and infections.
Therefore, the data given by Rubinstein et al1 in the abstract and the text regarding infection are not correct and require revision together with conclusions in their review. From our study and the results by Guillon et al,3 so far recent infection has to be regarded as a risk factor for cervical artery dissection. Furthermore, high-socioeconomic status may be another factor that is associated with the risk of CAD, although this certainly requires additional investigations.
Rubinstein et al1 mentioned the possibility of a selection bias in our study. Our case-control study2 was not part of a population-based register; however, it was based on consecutive patients in both groups. Given the fact that almost all of the younger patients with cerebral ischemia in the catchment area are treated in the University Center, the risk of selection bias can be rated as very low. In summary, 2 well-performed case-control studies found an independent association between recent infection and CAD that was not explained by factors such as mechanical stress to cervical arteries.2,3
We would like to thank Drs Grau and Buggle for their interest in our article. They raise 2 important issues.
Their first objection may have to do with clarity of the text. We are not suggesting that the weak association refers to the relationship between dissection and the mechanical factors associated with infection, such as coughing, sneezing, or vomiting, but rather, the association between infection and dissection, even when these mechanical stressors are controlled for. Therefore, our article also agrees with the authors that recent infection may be an important risk factor, which is an association noted by others. However, this association may not be very strong. High socioeconomic status may also be an important risk factor and perhaps a subject for additional study; however, we did not identify any case-control studies that have confirmed this relationship.
Regarding the second point, case-control studies are notoriously sensitive to selection bias. Our objection was not whether patients were selectively identified (ie, consecutive), but rather that the control group chosen (ie, cerebral ischemia) may be inappropriate. Quite simply, controls must be a representative sample of the study base and must have an equal chance to develop the target disease as the cases. If not, it is a case of com-paring apples with oranges. The mechanism of cerebral ischemia is quite clearly different from dissection, and we suggest that the risk factors may also differ (eg, “vascular risk factors”), which is why we proposed that healthy subjects might be a more suitable control. Otherwise, this may result in a potential overestimation of risk.
Note: This article was originally published online as a “systemic review.” However, this was a typographic error and should have read, “a systematic review.”