Assessment of the Role of Carotid Atherosclerosis in the Association Between Major Cardiovascular Risk Factors and Ischemic Stroke Subtypes

Key Points Question What is the role of carotid atherosclerosis in the association between major cardiovascular risk factors and different ischemic stroke subtypes? Findings In this population-based cohort study of Chinese adults with subtyping of incident ischemic strokes, carotid artery ultrasonographic measurements were recorded in 23 973 participants after 8 years’ follow-up. Blood pressure was associated with all ischemic stroke subtypes independently of carotid plaque burden, but after adjustment for blood pressure, carotid plaque burden was associated with probable large artery and lacunar stroke but not with probable cardioembolic stroke. Meaning Drug treatments targeting atherosclerosis may affect the risk of ischemic stroke subtypes to different extents.


eMethods 1: Further details of study measurements Carotid artery measurements
Automated B-mode ultrasound screening of the extra-cranial carotid arteries (using a Panasonic Cardio-Health Station implementing edge-detection software) was undertaken following a standard protocol consistent with the Mannheim consensus, and involved scanning both carotid arteries with automated measurements of carotid intima-media thickness (cIMT) and semi-automated recording of plaques. 1 cIMT was measured in the distal 1cm of the common carotid artery (CCA) just before the bifurcation at four predefined angles (two on each side) based on the Meijer Carotid Arc, using an inbuilt electronic transducer position guidance, including the right CCA at 150º and 120º and the left CCA at 210º and 240º. Hence, mean cIMT was estimated as the mean of four measurements per person. The number of carotid plaques (defined as focal thickenings of intima-media >1.5 mm) and the thickness of the largest plaque within four segments of the carotid arteries were recorded. 1 Carotid plaque burden was derived by first standardising the plaque number and maximum thickness (i.e., dividing each by its standard deviation [SD]) and estimating the average, then multiplying the average value by the SD of the maximum plaque thickness to provide a plaque burden recorded in millimetre units (i.e. interpretable as an enhanced estimate of the maximum plaque thickness).

Other measurements
Blood pressure was measured twice using a Omron UA-779 digital sphygmomanometer after participants had remained at rest in a seated position for at least 5 minutes. If the difference between the two measurements was more than 10 mmHg for SBP, a third measurement was made and the last two measurements were recorded. The mean of the two recorded values was used for analysis.
Except in one study area, where the protocol specified fasting by all participants, initial screening for hyperglycaemia involved immediate on-site testing of non-fasting blood glucose using the SureStep Plus meter (LifeScan, Milpitas, CA, USA). Participants with nonfasting glucose levels ≥7.8 and <11.1 mmol/l were invited to return for a fasting blood glucose test the next day. Screen-detected diabetes was defined as no prior history of diabetes and any of: (1) a random blood glucose level ≥7.0 mmol/l and a fasting time >8 h; (2) a random blood glucose level ≥11.1mmol/l and a fasting time <8 h; or (3) a fasting blood glucose level ≥7.0 mmol/l. Low and high density lipoprotein cholesterol measurements in baseline samples were available for a subset of participants overlapping a nested case-control study of the association of genome-wide panel and biochemistry markers with cardiovascular disease. Data were available for 2899 participants in the present study (256 with an ischemic stroke and 2643 without).

eMethods 2: Adjudication and subtyping of incident stroke events
Incident cases of stroke were identified using the linked electronic health records. To confirm the accuracy of pathological stroke types and to classify strokes into subtypes, medical records of all reported stroke cases were sought for independent verification and diagnosis by a panel of trained neurologists in China.
All hospital admissions with reported stroke episodes to be verified were collated in a central data repository in the co-ordinating centre at Oxford, UK prior to loading onto a portable computer tablet Organization criteria for stroke (defined as "rapidly developing clinical signs of focal or global disturbance of cerebral function, lasting more than 24 hours or leading to death due a vascular cause"). Adjudicators used findings in radiological reports on brain imaging (including the presence of hemorrhage or ischemia, laterality of lesions, and location in the brain) and other relevant findings in the patient's medical records to classify strokes into the pathological types: ischemic stroke, intracranial hemorrhage or subarachnoid hemorrhage. Confirmed ischemic strokes were further classified into the subtypes: lacunar stroke if the radiological report stated an infarct < 15 mm in diameter diagnosed as a lacunar infarct; 2 or non-lacunar stroke if not. All verified stroke cases that remained unconfirmed after adjudication were referred for final review by both Chinese-and Englishspeaking clinicians in the study coordinating centre in Oxford.
The most recent data release available (Release 15, August 2018) provides reliable confirmation and subtyping of a high proportion of strokes but has not yet consolidated information to reliably refute that a participant had a stroke, as this requires the further complex process of ascertainment that all potential strokes for a participant have been adjudicated. (Medical records for patients in China are not centrally consolidated and may be spread across several hospitals.)

Further details on adjustments
Analyses of associations between carotid measures and cardiovascular risk factors included basic adjustment of age-at-resurvey (5 year groups from <45 to ≥ 80, as a categorical variable) x sex x region. Analyses of associations between carotid measures and cardiovascular events included basic adjustment of age-at-resurvey (<5 year groups from <45 to ≥ 80, as a categorical variable) x sex + region. (The slightly more limited adjustment for the interactions with age in the outcome analyses was chosen to avoid over-adjustment in analyses with more limited power.) Additional adjustment for baseline blood pressure included systolic blood pressure (SBP), SBP x age-at-resurvey (<60, 60-69, ≥70 years) diastolic blood pressure (DBP), DBP x age-at-resurvey (<60, 60-69, ≥70 years) and diagnosed-hypertension at baseline.

Grouping of measures
For examination of the shape of the relationship between cardiovascular risk factors and the carotid measures, continuous valued risk factors were divided into 7 groups defined by the 10 th , 20 th , 40 th , 60 th , 80 th and 90 th percentiles of their respective distributions. For examination of the shape of the relationship between the carotid measures and stroke risk, participants were ranked by the respective carotid measure and divided into 6 groups so that an approximately equal number (~158) of strokes was included in each group.

Correspondence of presented results to a mediation framework
Following the casual steps approach summarised by McKinnon et al. 3 , 4 are steps are required to assess mediation between an independent variable and a dependent variable, i.e., in the present context between cardiovascular risk factors and stroke by plaque burden. The correspondence between the presented analyses and the 4 steps is indicated below. In interpretting these regression associations, based on measurements of risk factors and carotid parameters at a single time, consideration needs to be given to the impact of measurement error and short-term variability, as has been done in the Discussion.  2. "A significant relation of the independent variable to the hypothesized mediating variable". 3 Figure 1 showing the association of cardiovascular risk factors with plaque burden.
3. "The mediating variable must be significantly related to the dependent variable when both the independent variable and mediating variable are predictors of the dependent variable". 3 Plaque burden must be significantly related to stroke when cardiovascular risk factors are also in the model, which is shown in Figures 3 and 4, in the rows for the association of plaque burden with strokes after adjustment for cardiovascular risk factors.

"
The coefficient relating the independent variable to the dependent variable must be larger (in absolute value) than the coefficient relating the independent variable to the dependent variable in the regression model with both the independent variable and the mediating variable predicting the dependent variable". 3 In Figures 3 and 4 the odds ratios for the association of cardiovascular risk factors with strokes are larger when plaque burden is not adjusted for, than when it is adjusted for.