Student’s t-tests (continuous variables) or χ2 tests (categorical variables) were performed to describe the baseline characteristics of 464 participants in each group of case and control. Multiple logistic regression was applied to evaluate the association between ever smoking (or smoking index) and ESCC risk by odds ratios (ORs) and 95% confidence intervals (CIs), by adjusting for age and gender. General linear regression model was used to assess the associations of smoking index with metabolites and metabolites with ESCC outcome. Models were adjusted for age and gender, with the false discovery rate (FDR) adjusted to account for multiple comparisons.

High dimensional mediation analyses were then performed by R package HIMA to discover potential metabolic mediators for the association between ever smoking and ESCC risk [22, 23]. The analysis steps of this method are as follows (Fig. 1): 1) The sure independence screening was used to identify a subset of metabolites that are among the top n/(2 log (n)) largest effects of ever smoking on mediators whether P value makes sense or not, where n is the sample size [24]; 2) The minimax concave penalty was performed to evaluate the effects of metabolites subset on the ESCC outcome [25]; 3) The relevant FDR of ever smoking-metabolite and metabolite-ESCC associations based on a joint significance test were used to ensure intermediate metabolites. And potential mediation effects of metabolites on the association between smoking index and ESCC risk were explored by the same method.

Flow diagram illustrating the analysis steps of high dimensional mediation analyses, where X refers to smoking exposure (ever smoking or smoking index), M refers to metabolite and Y refers to the ESCC outcome

To better understand the significance of these metabolites selected by high dimensional mediation analyses, we also performed univariate mediation analysis on selected intermediate metabolites, adjusted for age and gender, using the medflex package in R [26]. All statistical analyses were carried out using software R-project (V.3.6.0). We interpreted two-sided P values of < 0.05 as statistically significant.

Note: The content above has been extracted from a research article, so it may not display correctly.

Please log in to submit your questions online.
Your question will be posted on the Bio-101 website. We will send your questions to the authors of this protocol and Bio-protocol community members who are experienced with this method. you will be informed using the email address associated with your Bio-protocol account.

We use cookies on this site to enhance your user experience. By using our website, you are agreeing to allow the storage of cookies on your computer.