Statistical analysis

The study population was divided into 4 groups according to the presence or absence of hypertension and salt intake: Group 1 (participants without hypertension and high salt intake), Group 2 (participants without hypertension and with high salt intake), Group 3 (participants with hypertension and without high salt intake), and Group 4 (participants with hypertension and high salt intake). Normally distributed variables were described and compared by means ± standard deviations (SD) and one-way analysis of variance (ANOVA). Data with a skewed distribution were described as the median (IQR) and were analyzed by nonparametric tests. Categorical variables were represented as absolute values with percentages and were compared using the chi-square test. Person-years were calculated from the date of the baseline examination to the date of cancer diagnosis, death, or the end of follow-up (31 December 2019), whichever occurred first. Logistic regression analysis was used to examine the relationship between salt intake and hypertension at baseline examination. The proportional hazard assumption was checked using Schoenfeld residual test. Cox proportional hazards analysis was used to estimate the hazard ratios (HRs) and their 95% confidence intervals (CIs) to determine the effect of hypertension and/or high salt intake on the risk of incident PLC. In the subgroup analysis, participants were further stratified by sex, age, HBV infection, drinking and smoking status. The interactions were tested using multiplicative models.

Due to the close association between hypertension and salt-intake, mediation analysis was conducted to explore the relationship between high-salt intake, hypertension, and PLC risk. The mediation analyses based on the variance-covariance matrix and the maximum likelihood technique were carried out using the CAUSALMED procedure. This procedure calculated the total effect, direct effect, and indirect effect.

As a sensitivity analysis, we excluded participants with PLC that had occurred within the first year or the first five years of follow-up to eliminate the possibility of reverse causation. In addition, death (competing risk event) may occur before the diagnosis of PLC. Traditional Cox regression may overestimate the absolute risk in this competing risk setting. We further used the cause-specific hazards (CS) models to calculate HR_CS of PLC incidence.

A two-sided P value<0.05 was considered statistically significant. Statistical analyses were performed using the SAS software, version 9.4.