Study populations
The GBCS is a three-way collaboration among the Guangzhou Number12 Hospital in Guangzhou, China, and the Universities of Hong Kong and Birmingham. The study aims to examine the effects of genetic, environmental, lifestyle, and occupational factors on health problems and chronic diseases. The Guangzhou Biobank Cohort Study-Cardiovascular Disease Sub-cohort (GBCS-CVD), nested in the GBCS, included 1,996 participants (992 (49.7%) men and 1,004 (50.3%) women) examined during September 2006 to December 2007. A detailed description of the CVD Sub-cohort has been published elsewhere [22, 23].
Ethical approval
The study received ethical approval from the Guangzhou Medical Ethics Committee of the Chinese Medical Association, Guangzhou, China. All participants provided written informed consent before participation. All methods were performed in accordance with the relevant guidelines and regulations.
Exposure measurement
A computer-assisted standardized questionnaire was used to collect information of demographic characteristics, socioeconomic position, occupational exposure, family and personal disease history, and lifestyle factors including smoking status, alcohol use, and physical activity measured by the International Physical Activity Questionnaire (IPAQ). Reliability of the questionnaire was evaluated by recalling 200 randomly selected participants for re-interview andathe results were satisfactory [24]. Body height, weight, waist and hip circumference, and blood pressure were measured following standardized procedure. Body mass index (BMI) was calculated by weight (kg) divided by the square of body height (m2). Venous blood samples were collected after fasting for at least 8 hours for assay of lipids, fibrinogen and high-sensitivity C-reactive protein (hs-CRP) levels. Assay of complete blood count including leukocyte count and differential (granulocyte and lymphocyte count) was performed in an automated hematology analyzer (KX-21, SYSMEX, Japan).
Study outcome
Carotid intima-media thickness (IMT) were measured by carotid B-mode color ultrasonography using ALT HDI 3000 mainframe with a high-resolution, linear array scanner (medium frequency 7.5MHz) by a specialist physician [25]. The operators were registered ultrasound doctors who had a professional certificate for color Doppler ultrasound measurement awarded by the Ministry of Health of China. All scans were performed following a predetermined, standardized scanning protocol for the right and left carotid arteries using images of the far wall of the distal 10mm of the common carotid arteries. Three scanning angles, with the image focused on the posterior wall, were recorded from the angle showing the greatest distance between the lumen-intima interface and the media-adventitia interface. Carotid IMT measurements were performed off-line with the use of automated image analysis software. All scans were analyzed by the same physician, blinded to participants’ information. Details of our carotid IMT measurement and research findings have been reported elsewhere [26, 27].
Mediation analysis
To estimate the contribution of inflammation to the association between smoking and IMT, we used causal mediation analysis under the counterfactual framework, which can decompose the averaged total effects (ATE) into indirect (average causal mediation effect, ACME) and direct effects (average direct effect, ADE) [28]. Let Mi (t) denotes the potential value of a mediator of interest for unit i under the positive exposure status Ti = t, Yi (t, m) denotes the potential outcome if the positive exposure status and mediating variables equal to t and m. As only one outcome was considered in the current study, the observed mediating and outcome variables were represented as Mi = Mi (Ti) and Yi = Yi {Ti, Mi (Ti)}, respectively. Following the potential outcomes notation, the indirect effects (δi) and direct effects () for each unit i and each treatment status t = (0, 1) were defined as (t) =Yi {t, Mi (1)} - Yi {t, Mi (0)} and (t) = Yi {1, Mi(t)} - Yi {0, Mi (t)}, respectively, and the ACME was defined as =E [Yi {t, Mi (1)} - Yi {t, Mi (0)}]. Similarly, the ADE was defined as = E [Yi {1, Mi(t)} - Yi {0, Mi(t)}]. As it is implausible to observe a counterfactual outcome Yi{t, Mi(1 - t)} with one observational unit, the estimation of ACME and ADE requires an additional assumption known as sequential ignorability (SI) [21]. The SI assumes that, firstly, given the observed pretreatment confounders, the treatment assignment is ignorable, and sequentially, given the actual treatment status and pretreatment confounders, the observed mediating status is ignorable. The causal effects could be estimated as function of the sensitivity parameter ρ [20, 29]. The equation between ρ and the coefficients of determination as below, ρ2= , where and represent the percentage of residual variance that is explained by the unmeasured confounders in the mediator and in the outcome, and represent the proportion of variance that is explained by the unobserved confounder in the mediator and in the outcome. Values of ρ different from zero may imply that the SI assumption is violated, indicating that estimation of the mediation effect may be biased. In this study, we used the mediation package in Stata for the mediation analysis [30].
Statistical analysis
Continuous variables were presented as means (standard deviations (SD), or medians (25th to 75th quartiles) for variables that were not normally distributed. Categorical variables were presented as frequencies (percentages). Chi-square tests or analysis of variance (ANVOA) were used to compare participants’ baseline characteristics by cigarette smoking status (never, former, and current smoking). The average IMT of the left and right common carotid artery was used for data analysis. Cigarette smoking was considered as the exposure variable, inflammation indicated by leukocyte and differential count (granulocyte and lymphocyte count), hs-CRP and fibrinogen were considered as potential mediators, and IMT was considered as the study outcome. All tests of significance were two-tailed, with p <0.05 as statistically significant.