Appraising associations between signature lipidomic biomarkers and digestive system cancer risk: novel evidences from a prospective cohort study of UK Biobank and Mendelian randomization analyses

Abstract

Background: The roles of serum lipids on digestive system cancer (DSC) risk were still inconclusive. In this study, we systematically assessed indicative effects of signature lipidomic biomarkers (high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), and triglycerides (TG)) on DSC (oesophagus, stomach, colorectal, liver, gallbladder, and pancreas cancers) risk.

Methods: HDL-C, LDL-C, and TG concentration measurements were respectively analyzed with enzyme immunoinhibition, enzymatic selective protection, and GPO-POD methods in AU5800 supplied from Beckman Coulter. The diagnoses of DSCs were coded using International Classification of Diseases, Tenth Revision (ICD-10) codes updated until October 2022 in the UK Biobank (UKB). In this study, we assessed phenotypic association patterns between signature lipidomic biomarkers and DSC risk using restricted cubic splines (RCSs) in multivariable-adjusted Cox proportional hazards regression models. Moreover, linear and nonlinear causal association patterns of signature lipidomic biomarkers with DSC risk were determined by linear and nonlinear Mendelian randomization (MR) analyses.

Results: A median follow-up time of 11.8 years was recorded for 319,568 participants including 6916 DSC cases. A suggestive independent nonlinear phenotypic association was observed between LDL-C concentration and stomach cancer risk (P_nonlinearity < 0.05, P_overall < 0.05). Meanwhile, a remarkable independent linear negative phenotypic association was demonstrated between HDL-C concentration and stomach cancer risk (P_nonlinearity > 0.05, P_overall < 0.008 (0.05/6 outcomes, Bonferroni-adjusted P)), and suggestive independent linear positive associations were observed between HDL-C concentration and colorectal cancer risk, and between TG concentration and gallbladder cancer risk (P_nonlinearity > 0.05, P_overall < 0.05). Furthermore, based on nonlinear and linear MR-based evidences, we observed an suggestive independent negative causal association (hazard ratio (HR) per 1 mmol/L increase: 0.340 (0.137-0.843), P = 0.020) between LDL-C and stomach cancer risk without a nonlinear pattern (Quadratic P = 0.901, Cochran Q P = 0.434). Meanwhile, subgroup and stratified MR analyses both supported the category of LDL-C ≥ 4.1 mmol/L was suggestively protective against stomach cancer risk, especially among female participants (HR: 0.789 (0.637-0.977), P = 0.030) and participants aged 60 years or older (HR: 0.786 (0.638-0.969), P = 0.024), and the category of TG ≥ 2.2 mmol/L concluded to be a suggestive risk factor for gallbladder cancer risk in male participants (HR: 1.447 (1.020-2.052), P = 0.038) and participants aged 60 years or older (HR: 1.264 (1.003-1.593), P = 0.047).

Conclusions: Our findings confirmed indicative roles of signature lipidomic biomarkers on DSC risk, notably detecting suggestive evidences for a protective effect of high LDL-C concentration on stomach cancer risk, and a detrimental effect of high TG concentration on gallbladder cancer risk among given participants.

Keywords: Causal associations; Digestive system cancer risk; Linear and nonlinear Mendelian randomization analysis; Phenotypic associations; Polygenic risk score; Signature lipidomic biomarkers.

Fig. 1

Flow charts of the study. RCS, restricted cubic splines; LACE, local average causal effects

Fig. 2

Phenotypic association patterns between signature lipidomic biomarkers (HDL-C (A), LDL-C (B), and TG (C)) and the risk of DSC. Distributions of signature lipidomic biomarker concentrations and RCSs (red lines) representing shapes of phenotypic associations with adjustment of age, sex, BMI, smoking status, alcohol drinking status, education qualification, employment status, TDI, physical activity level, medication use, family cancer history, ALT, AST, SBP, DBP, primary hypertension, cerebral infarction, ischaemic heart disease, HbA1c, and diabetes. P_overall and P_nonlinear values of 0.008 (0.05/6 outcomes, Bonferroni-adjusted P) were defined as the threshold of remarkable statistical significance. P_overall and P_nonlinear values between 0.008 and 0.05 were defined as suggestive statistical significance

Fig. 3

Phenotypic associations between three categories of signature lipidomic biomarkers and the risk of DSCs. Hazard ratios (HRs) (red squares) with 95% CIs (black solid lines) of three categories of HDL-C, LDL-C, and TG concentrations on the risk of oesophagus, stomach, colorectal, liver, gallbladder, and pancreas cancers with adjustment of age, sex, BMI, smoking status, alcohol drinking status, education qualification, employment status, TDI, physical activity level, medication use and family cancer history, ALT, AST, SBP, DBP, primary hypertension, cerebral infarction, ischaemic heart disease, and diabetes. P and P_trend value of 0.008 (0.05/6 ourcomes, Bonferroni-adjusted P) were defined as the threshold for remarkable statistical significance. P and P_trend values of 0.05 were defined as suggestive statistical significance

Fig. 4

Causal association patterns between signature lipidomic biomarkers and the risk of DSCs. Nonlinear MR analyses with piecewise linear method for genetically predicting the associations between (A) HDL-C, (B) LDL-C, and (C) TG concentrations and the risk of oesophagus, stomach, colorectal, liver, gallbladder, and pancreas cancers. Exposure and outcome regression stages were both adjusted with age, sex, assessment centers, genotyping array and the first 10 PCs. Each black dot and black vertical line represented the LACE with its 95% confidence interval in each stratum and red dots represented reference points

Fig. 5

Phenotypic and genetic association analyses of signature lipidomic biomarkers with the risk of DSCs. Phenotypic association analyses were adjusted with age, sex, BMI, smoking status, alcohol drinking status, education qualification, employment status, TDI, physical activity level, medication use, family cancer history, ALT, AST, SBP, DBP, primary hypertension, cerebral infarction, ischaemic heart disease, HbA1c, and diabetes. Linear one- and two-sample MR analyses were both adjusted with age, sex, assessment centers, genotyping array and the first 10 PCs. HRs (HDL-C: black squares, LDL-C: blue squares; TG: orange squares) with 95% Cis for oesophagus, stomach, colorectal, liver, gallbladder, and pancreas cancers per 1 mmol/L increase of HDL-C, LDL-C, and TG concentrations. P value of 0.008 (0.05/6 outcomes, Bonferroni-adjusted P) was defined as the threshold for remarkable statistical significance. P value between 0.008 and 0.05 was defined as suggestive statistical significance