A pooled analysis of individual patient data from trials of endarterectomy for symptomatic carotid stenosis: efficacy of surgery in important subgroups
Benefit from carotid endarterectomy (CEA) depends on the degree of symptomatic stenosis, but is also likely to be influenced by other clinical and angiographic characteristics. However, individual trials have been too small to allow reliable subgroup analysis. We therefore studied pooled individual patient data from the European Carotid Surgery Trial, the North American Symptomatic Carotid Endarterectomy Trial and the Veterans Administration trial #309. We determined the effect of CEA in 11 predefined subgroups: age (<65, 65–74, 75+), sex, type of presenting event (cerebral vs ocular; TIA vs stroke; lacunar vs non-lacunar), side of presenting event, months since last event (<1, 2–3, 4+), diabetes, plaque surface irregularity, near-occlusion, and contralateral carotid occlusion. There were statistically significant interactions between the risk of ipsilateral ischaemic stroke in the medical group and 8 of the 11 subgroup variables. In the surgery group, there were interactions between the operative risk of stroke and death and 6 subgroup variables. We therefore assessed heterogeneity of overall treatment effect (any ipsilateral ischaemic stroke and surgical stroke/death) in these subgroups. There was clinically and statistically significant heterogeneity within 5 subgroups: benefit from surgery increased with age, was greater in men than women; decreased with time since presenting event; was greater after stroke than TIA, and was absent in cases of near-occlusion. For example, in patients with 50–69% stenosis, the 5 yr absolute risk reduction was 10% (95% CI = 3–10, P=0.0005) in men and -3% (95% CI = -8 - 2, P=0.8) in women (overall interaction, P=0.003). There were also important differences in the effect of surgery for lacunar and non-lacunar stroke. Patients who are most likely to benefit from CEA cannot be identified using the degree of symptomatic carotid stenosis alone. Several other clinical and angiographic characteristics influence the efficacy of surgery. Optimal selection of patients will require a risk-modelling approach using multiple baseline characteristics.