Response to Letter by Paraskevas et al
In response to our article,1 Paraskevas et al suggest that “severe, but not moderate, carotid atherosclerosis may predict concomitant vascular disease in other arterial beds.” In their study,2 they found a significant association between symptomatic (stroke, transient ischemic attack, amaurosis fugax) internal carotid artery occlusion and symptomatic atherosclerotic disease in other vascular beds, including the coronary, vertebral, aortic, and peripheral arteries. In contrast, in our study, we found that in patients with acute stroke symptoms, significant atherosclerotic disease in the carotid arteries (>50% stenosis imaged on CT angiography) did not predict significant disease in the coronary, aortic, or vertebral arteries. Our findings challenge the need for extensive investigation into the presence of coronary disease in patients with stroke symptoms as well as the general concept that atherosclerotic disease tends most frequently to be a systemic disease.
There are a number of key differences between our studies explaining our very different results. Our study was prospective in design, enrolling consecutive patients from 2006 to 2008 who had acute stroke symptoms. In contrast, Paraskevas et al conducted a retrospective chart review of nonconsecutive patients who had internal carotid artery occlusion. Their study was conducted over a large period of time (1987 to 2006), and patient inclusion was dependent on them being referred for a carotid ultrasound study. Because their study spanned such a long period, it is likely that the standard workup of the patients varied over this period, leading to significant selection bias. One of the many illustrations of this selection bias is that “only” 120 patients were included in a study for a period spanning 20 years, which seems low considering the prevalence of carotid occlusion and the number of patients who likely underwent ultrasound during that period (this number is by the way not reported by Paraskevas et al).
Another key difference between our studies lies in the way atherosclerotic disease was assessed in the different vascular beds. In the study by Paraskevas et al, “The presence of clinically evident vascular disease in the aorta, coronary and peripheral arteries was recorded,” which we assume means was looked up from the medical records. “The presence of coronary artery disease was not verified by stress ECG or angiography,” which is a serious issue because it is well known that coronary symptoms do not necessarily correlate with the presence and severity of underlying atherosclerotic disease.3 “For PAD, the Ankle Brachial Pressure Index (ABPI) was measured for most patients,” but not all patients. “For aortic disease, all patients underwent a physical examination, and in the case of suspicion of abdominal aortic aneurysm, an ultrasound examination was performed,” again opening the door to selection bias. In contrast, our assessment for atherosclerosis in the different vascular beds was very systematic, using 1 single imaging technique (CT angiography) obtained in all patients without exception at 1 single time point for all vascular beds.
We were surprised by our results. Originally, we thought that we would find that atherosclerotic disease was more often present simultaneously in multiple vascular beds as has been reported in multiple studies before. Our study showed that this is not the case and that significant atherosclerotic disease is usually confined to 1 vascular bed. This was true for stenosis >50% and would have been even more pronounced if we would have looked at rarer complete vascular occlusion. It is possible that some of the studies that previously showed that atherosclerosis was a systemic process shared some of the limitations presented by the study reported by Paraskevas et al.
Our study was not completely free of selection bias. Our results apply to a population of patients presenting with stroke symptoms (not all of them ended up diagnosed with a stroke). It is likely that the prevalence of atherosclerosis and the association between the vascular beds would have been lower/weaker in a general population of patients without stroke symptoms and higher/stronger in a population of patients with a confirmed diagnosis of stroke.