DOI: https://doi.org/10.21203/rs.3.rs-2112112/v1
Purpose The association of lipid metabolism linked the risk of gastric cancer (GC) was widely debated. We aimed to explore the longitudinal associations between total cholesterol, triglyceride (TG), high-density lipoprotein cholesterol (HDL-C), and low-density lipoprotein cholesterol (LDL-C) with the incident risk of GC.
Methods The serum lipids were quarterly stratified based on the distribution of GC-free populations. The Cox proportional hazard models and restricted cubic spline models were applied to estimate the hazard ratios (HRs) and dose-response association of GC under different sub-analyses. The interactions of serum lipids on GC incidence were tested by generalized additive models.
Results After average 7.2±1.2 years follow-up, 248 primary GCincident cases were collected among 45,642 cancer-free baseline individuals.In total population, the hazard risks (HRs) with 95% confidence interval (CI) of TG (HR=1.53, 95% CI: 1.02-2.29) and LDL-C (HR=2.21, 95% CI: 1.51-3.24) were significantly increased when the Q4 stratum compared with Q1. While decreased HR was found in the Q4 stratum of HDL-C (HR=0.42, 95% CI: 0.26-0.67). Further sub-analyses testified these associations in males solely. The highest GC incident risk was plainly visible when both HDL-C and LDL-C were abnormal (HR=5.38, 95% CI: 3.43-8.45), followed by excess TG and hypo-HDL-C group (HR=2.75, 95% CI: 1.89-4.00) and excess TG and LDL-C group (HR=2.55, 95% CI: 1.78- 3.64) compared with normal lipid group.
Conclusion Lipid metabolism abnormalities could be important risk factors for GC. Additionally, a combination of any abnormalities among TG, HDL-C, and LDL-C would interactively elevate the incidence risk of GC.
Gastric cancer (GC) was the fifth most prevalent malignancy and the fourth-largest cause of cancer-related death worldwide (Sung et al. 2021), which resulted in annually 1.09 million GC incident cases and 7.7 million fatalities due to GC. Accounting China for quintile of the global population, while almost half of the annual GC incidents (478,000) were discovered within Chinese population (Sung et al. 2021), the etiology of GC in China should be different. The prevalence of metabolic syndrome (MS) in Chinese adults has climbed to 24.5%, which was higher than the worldwide prevalence of 22.5% (Li et al. 2016). Thus, the substantially higher prevalence rate of MS could be a hypothetical reason.
The worldwide incidence rate of GC has decreased by 28.0% since 1990. However, the reduced trend of GC incidence rate was more evident in Europe (43.3%) and Latin America (33.2%) than in China (14.7%) (GBD 2017 Stomach Cancer Collaborators 2020). Previous studies demonstrated that multiple factors, including Helicobacter pylori (H. pylori) infection, smoking, alcohol drinking, dietary patterns, obesity, and gastric ulcer, may contribute to the incidence and development of GC (Rawla et al. 2019). However, these risk factors may correspond to an incomprehensive interpretation, particularly not fully understanding the predominant proportion of the GC in the Chinese population globally. The MS was hypothetically linked to the GC, reportedly (Li et al. 2018). While several relevant studies were conducted from a cross-sectional perspective, the conclusions regarding the prospective association of MS with GC incidence remain controversial (Lin et al. 2015; Yoo et al. 2019).
Dyslipidemia is a systemic metabolic condition, and its frequency in Chinese adults has climbed to 40.40% during the past decades (Joint committee for guideline revision 2018). Increasing evidence has suggested that dyslipidemia may also have a biological association with the development of cancer, including colon cancer, pancreatic cancer, breast cancer, and thyroid cancer (Brantley et al. 2020; Wang et al. 2021; Wu et al. 2021). Two studies showed that a population exposed to a higher level of total cholesterol (TC) would significantly increase the risk of GC incidence (Lim et al. 2022; Kitahara et al. 2011). Following evidence confirmed these preceding findings and further found elevated low-density lipoprotein cholesterol (LDL-C) (Pih et al. 2021; Zou et al. 2021) and triglyceride (TG) (Li et al. 2018) may also increase the incidence risk of GC. Furthermore, high-density lipoprotein cholesterol (HDL-C) was traditionally believed to be a protective factor for cardiovascular diseases, whereafter several prospective studies have shown that lower HDL-C would also increase the risk of GC (Li et al. 2018; Lim et al. 2022; Nam et al. 2019). Despite previous studies showing controversial or irrelevant results linking the abnormal lipid metabolism to the GC (Pih et al. 2021; Huang et al. 2016), two prospective studies have not found any association between elevated LDL-C and TG with GC (Borena et al. 2011; Wulaningsih et al. 2012). The hypothetical carcinogenicity of abnormal lipid exposure may relate to the forming free radicals and accelerating cell damage via oxidative stress and chronic inflammation mechanisms, which could be an essential trigger of tumors (Nam et al. 2019). In addition, HDL-C has been reported to exert anti-inflammatory effects through anti-inflammatory reprogramming mediated by the transcriptional regulator ATF3 (De Nardo et al. 2014). However, an animal experiment showed that TG and TC were not independent risk factors for cancer but only mediate cancer progression as mediators (Novosyadlyy et al. 2010). Furthermore, most previous studies did not adjust for H. pylori for covariates (Lim et al. 2022; Kitahara et al. 2011; Borena et al. 2011; Wulaningsih et al. 2012), which would broadly introduce potential biases.
Therefore, we conducted this prospective study in Jinchang Cohort after 8 years follow-up to explore the epidemiological association between abnormal serum lipids and GC incidence. The Cox proportional hazard regression models and generalized additive models were performed to further explore the variated risks of serum lipids interactions on the GC incidence.
This study was based on the Jinchang Cohort conducted in Gansu Province, China (Bai et al. 2017). The Jinchang Cohort was established on the basis of free medical examination for all employees and retirees by the Jinchuan Nonferrous Metals Corporation (JNMC) rounded every two years funded. The baseline information for 48,001 participants was collected from 2011 to 2013, including epidemiological surveys, physical examinations, laboratory biochemical tests, and bio-samples donation. The follow-ups were conducted every two years, and the phase I and phase II follow-ups were accomplished from 2014 to 2017, with a total of 46,713 and 41,888 participants being retrieved. The phase III follow-up was started in 2018 and currently retrieved 40,530 individuals. Eventually, a total of 47,367 participants who were retrieved at least once at any phase of follow-ups were identified as our study population.
The details of the inclusion and exclusion criteria were listed in Supplementary Fig. 1. More specifically, baseline cancer patients (n = 432) and participants without data of serum lipids (n = 467) were excluded from the study population. Following that, those without data on BMI, smoking status, drinking status, and physical activity (n = 826) were further excluded. Eventually, a total of 45,642 cancer-free subjects at baseline were included in the study population. Phase I and phase II follow-ups identified 83 and 90 cases of GC, respectively; 75 cases of GC were identified from 2018 to 2020.
All the data collected at follow-ups were kept identical to the baseline stage to eliminate information obliteration. Well-trained investigators with standardized questionnaires were assigned to collect health information of the participants through face-to-face interviews, including demographic characteristics, health-related lifestyle, disease history, and family history of diseases. Physical examinations of all participants were performed by workers’ hospital clinicians. 12 ml fasting venous blood was donated from the subjects for biochemical examinations. The lipoprotein and triglyceride tests were measured by an automatic biochemical analyzer (Hitachi, 7600-020). In detail, the TC and TG were measured by peroxidase method with detection kits (YZB/Shanghai 1555-40-2009), the HDL-C and LDL-C were measured by direct enzyme method by detection kits (YZB/Shanghai 0967-40-2009 and YZB/Shanghai 0972-40-2009). The GC cases were identified by dynamically reviewing the medical records from the workers’ hospital during the entire follow-ups, including the basic patient information, admission and discharge records, and International Classification of Diseases Code, 10th revision (ICD-10).
The study protocol was approved by the Ethics Committee of the School of Public Health, Lanzhou University, and the Worker Ethics Committee of JNMC Hospital, and all subjects were fully informed of the test content before entering the cohort and signed a written informed consent.
The outcome of GC in this study was determined based on the inpatient medical records of the workers’ hospital of the enterprise from 2013 to 2020, a total of 248 GC cases were identified according to ICD-10 (C16) and diagnostic name.
The follow-up time for GC incident cases was from the enrollment date to the outpatient date, inpatient date, or fatality date caused by GC, whichever came first. To quantitatively adjust the potential biases which may be introduced by smoking and alcohol consumption, the pack-year index and alcohol index were uniformly calculated in our study. Specifically, the total drinking index (Kilogram-year) was the sum of the alcohol index of liquor, beer, and red wine, the formula for calculating the drinking index: alcohol intake per day (kilogram) × time of drinking (year) × alcohol volume fraction (for liquor: 0.54, for red wine: 0.03, for beer: 0.12). The formula for calculating the smoking index: cigarettes per day (pack) × time of smoking (year).
All categorical variables were presented as numbers of cases (Column frequency), and the difference between GC and Non-GC groups was tested by chi-square tests. Abnormally distributed variables, including TC, TG, HDL-C, and LDL-C, were presented as the median (Interquartile Range), and descriptive statistics of baseline demographic characteristics were performed by Rank Sum tests.
The serum lipids were stratified into quartiles based on the frequency distribution among GC-free participants. The age-standardized incidence rate (ASIR) of GC was adjusted based on the age distribution of the sixth Chinese census. The hazard ratios (HRs) and 95% confidence intervals (CI) were calculated by the Cox proportional hazard models, and the median of each serum lipids quartile was included as a continuous variable in the regression model for linear trend test. Considering the association between blood lipids and the risk of GC may yield non-linearity, restricted cubic spline (RCS) models with 4 knots at the 5th, 25th, 75th, and 95th percentiles of serum lipids were used to fit the dose-response curves between blood lipids and GC risk.
To test the possibility of collinearity among variables we interested, numerical covariates included in our study were further tested by spearman correlation tests and variance inflation coefficient (VIF) (Supplementary Fig. 2). The stepwise regression method was used to exclude confounding factors such as physical activity and family income that affected the HR value less than 10% in the final regression models. Eventually, Cox and RCS models were adjusted by age (< 40, 40 ~ 49, 50 ~ 59, ≥ 60 years), gender (male, female), education level (Middle school or below, High school or Junior college, Bachelor’s Degree or above), smoking index (0, 0.01 ~ 6.99, 7.00 ~ 15.99, 16.00 ~ 28.99, ≥ 29.00 pack-year), drinking index (0, 0.01 ~ 67.57, 67.58 ~ 168.93, 168.94 ~ 404.05, ≥ 404.06 kilogram-year), body mass index (< 18.5, 18.5 ~ 23.9, 24.0 ~ 27.9, ≥ 28.0 kg/m2), family history of cancer within two relatives (no, yes), H. pylori infection (no, yes), fruit intake (never, < 2.5, ≥ 2.5 kg/week), fat intake (higher, moderation, lower), pickled food intake (never, occasionally, sometimes, often), high salt diet (higher, moderation, lower), intake food temperature (hot, moderation, cool).
The serum lipids involved in the interaction study were categorized into normal and abnormal groups based on clinical cutoff values of marginal increase : TC: (< 5.20, ≥ 5.20 ), TG: (< 1.70, ≥ 1.70), HDL-C: (< 1.00, ≥ 1.00) LDL-C: (< 3.40, ≥ 3.40) (mmol/L). The interactions between individual variables on the variation of GC incidence risks were tested by the Generalized additive models (GAM). The Cox regression models and interaction tests were performed by survival packages of ‘gam’ and ‘mgcv’ of R statistical environment (Version 4.1.2), and RCS models were operated by SAS software (Version 9.4; SAS Institute Inc, Cary, NC). All statistical analyses were set as two-sided, with ρ < 0.05 set as the significance level.
Selected demographic characteristics between GC and Non-GC groups at baseline were shown in Table 1. As expected, baseline characteristics, including age, gender, education level, BMI, smoking index, drinking index, and dyslipidemia proportion between GC and Non-GC groups, were statistically different. The mean levels of TG and LDL-C in the GC group were significantly higher than those in the non-GC group, while HDL-C was lower than that in the non-GC group. Supplementary Table 1 displayed the medians, percentiles, and geometric means of TC, TG, HDL-C, and LDL-C.
| Baseline demographic characteristics | Gastric Cancer | ρ value | |
|---|---|---|---|
| Yes (n = 248) | No (n = 45394) | ||
| Age, years | |||
| ༜55 | 50(22.16) | 327557(72.16) | < 0.001 |
| 55 ~ 64 | 84(33.87) | 6868(15.13) | |
| ≥ 65 | 114(45.97) | 5768(12.71) | |
| Sex, n (%) | |||
| Male | 214(86.29) | 27326(60.20) | < 0.001 |
| Female | 34(13.71) | 18068(39.80) | |
| Education level, n (%) | |||
| Middle school or lower | 184(72.19) | 17007(37.46) | < 0.001 |
| High school or Junior college | 55(22.18) | 21370(47.08) | |
| Bachelor’s Degree or higher | 9(3.63) | 7017(15.46) | |
| BMI (kg/m2), n (%) | |||
| < 18.5 | 6(2.42) | 2181(4.80) | < 0.001 |
| 18.5–23.9 | 101(40.73) | 23376(51.50) | |
| 24.0–27.9 | 106(42.74) | 15513(34.17) | |
| ≥ 28.0 | 35(14.11) | 4324(9.53) | |
| Smoking index (Pack-year), n (%) | |||
| 0 | 79(31.86) | 26394(55.94) | < 0.001 |
| 0.01–6.99 | 18(7.26) | 4898(10.79) | |
| 7.00–15.99 | 35(14.11) | 4784(10.54) | |
| 16.00–28.99 | 50(20.16) | 5274(11.62) | |
| ≥ 29.00 | 66(26.61) | 5044(11.11) | |
| Drinking index (Kilogram-year), n (%) | |||
| 0 | 153(61.69) | 35023(77.14) | < 0.001 |
| 0.01–67.57 | 11(4.44) | 2605(5.74) | |
| 67.58–168.93 | 9(3.63) | 2527(5.57) | |
| 168.94–404.05 | 27(10.89) | 2658(5.86) | |
| ≥ 404.06 | 48(19.35) | 2581(5.69) | |
| Family history of cancer, n (%) | 56(22.58) | 10412(22.94) | 0.753 |
| H. pylori positive, n (%) | 3(1.21) | 1094(2.41) | 0.218 |
| TC, mmol/L | 4.80(1.30) | 4.60(1.12) | 0.103 |
| TG, mmol/L | 1.90(1.30) | 1.50(1.20) | < 0.001 |
| HDL-C, mmol/L | 1.15(0.43) | 1.33(0.45) | < 0.001 |
| LDL-C, mmol/L | 3.28(1.00) | 2.97(1.02) | < 0.001 |
| Dyslipidemia, n (%) | 160(64.52) | 16490(36.33) | < 0.001 |
| Abbreviations: TC total cholesterol; TG triglyceride; HDL-C high-density lipoprotein cholesterol; LDL-C low-density lipoprotein cholesterol Continuous data were shown as mean (Interquartile Range) and categorical data as n (%) Drinking index and smoking index were stratified by the quartile distribution of drinking index and smoking index based on Non-GC group | |||
Table 2 demonstrated the IRs, ASIRs, and multi-adjusted HRs (95% CIs) of GC among different serum lipids groups in the total population. No significant variations of HRs were observed in TC stratifications compared to the Q1 strata (ρtrend = 0.83). However, statistically increased risks of GC were widely found in TG and LDL-C strata when the Q4 compared with Q1 strata, while HDL-C showed an inverse trend (ρtrend < 0.001). For TG stratifications, the risk of GC in the Q4 stratum was 1.53 times higher than the Q1 strata (HR = 1.53, 95% CI: 1.02–2.29), and the ASIR of TG stratifications (Q4) was 1.79-fold higher than the Q1 strata. Subjects in the Q4 stratum had 58% decreased risk for GC when compared to the Q1 strata of HDL-C stratifications (HR = 0.43, 95% CI: 0.26–0.67), and the ASIR of Q4 stratum was 0.30-fold lower than the Q1 strata. The adjusted HR for risk of GC was 2.21 (95% CI: 1.51–3.24) in the Q4 stratum of LDL-C stratifications, the ASIR in which was 1.30 times higher than the Q1 strata.
| Serum lipid types | n/N | IR | ASIR | Crude | Model 1 | Model 2 |
|---|---|---|---|---|---|---|
| (/100k person-years) | HR (95%CI) | HR (95%CI) | HR (95%CI) | |||
| TC a | ||||||
| Q1(< 4.10) | 60/11090 | 88.91 | 73.26 | 1.00 (Ref) | 1.00 (Ref) | 1.00(Ref) |
| Q2(4.10 ~ 4.59) | 45/10326 | 71.27 | 76.16 | 0.79(0.53,1.16) | 0.80(0.54,1.17) | 0.79(0.53,1.16) |
| Q3(4.60 ~ 5.21) | 69/12803 | 87.87 | 57.41 | 0.96(0.67,1.35) | 0.89(0.63,1.26) | 0.88(0.62,1.25) |
| Q4(≥ 5.22) | 74/11423 | 107.65 | 59.09 | 1.18(0.85,1.67) | 0.98(0.69,1.40) | 1.01(0.71,1.44) |
| ρ for trend | 0.189 | 0.891 | 0.831 | |||
| TG b | ||||||
| Q1(< 1.10) | 37/10932 | 56.24 | 50.25 | 1.00(Ref) | 1.00(Ref) | 1.00(Ref) |
| Q2(1.10 ~ 1.49) | 30/10263 | 48.39 | 33.44 | 0.87(0.54,1.41) | 0.63(0.39,1.02) | 0.63(0.39,1.02) |
| Q3(1.50 ~ 2.29) | 87/12929 | 110.48 | 69.35 | 1.95(1.32,2.87) | 1.23(0.83,1.83) | 1.23(0.82,1.82) |
| Q4(≥ 2.30) | 94/11518 | 131.73 | 89.92 | 2.25(1.53,3.30) | 1.52(1.01,2.28) | 1.53(1.02,2.29) |
| ρ for trend | < 0.001 | < 0.001 | < 0.001 | |||
| HDL-C c | ||||||
| Q1(< 1.12) | 103/11003 | 154.92 | 90.10 | 1.00(Ref) | 1.00(Ref) | 1.00(Ref) |
| Q2(1.12 ~ 1.32) | 73/11579 | 103.80 | 70.14 | 0.67(0.49,0.90) | 0.84(0.61,1.14) | 0.83(0.61,1.14) |
| Q3(1.33 ~ 1.56) | 46/11403 | 66.24 | 48.26 | 0.43(0.31,0.61) | 0.65(0.44,0.94) | 0.65(0.45,0.95) |
| Q4(≥ 1.57) | 26/11657 | 36.29 | 27.16 | 0.24(0.15,0.37) | 0.41(0.26,0.66) | 0.42(0.26,0.67) |
| ρ for trend | < 0.001 | < 0.001 | < 0.001 | |||
| LDL-C d | ||||||
| Q1(< 2.48) | 44/11401 | 68.02 | 58.42 | 1.00(Ref) | 1.00(Ref) | 1.00(Ref) |
| Q2(2.48 ~ 2.96) | 31/11212 | 44.43 | 34.64 | 0.56(0.35,0.89) | 0.66(0.41,1.06) | 0.68(0.42,1.09) |
| Q3(2.97 ~ 3.48) | 74/11534 | 102.23 | 74.78 | 1.42(0.98,2.01) | 1.61(1.10,2.36) | 1.59(1.08,2.33) |
| Q4(≥ 3.49) | 99/11495 | 139.34 | 75.72 | 1.91(1.34,2.73) | 2.23(1.52,3.27) | 2.21(1.51,3.24) |
| ρ for trend | < 0.001 | < 0.001 | < 0.001 | |||
| Abbreviations: TC total cholesterol; TG triglyceride; HDL-C high-density lipoprotein cholesterol; LDL-C low-density lipoprotein cholesterol; IR Incidence Rate; ASIR age-standardized incidence rate; HR hazard ratio; CI confidence interval Crude: Unadjusted covariates Model 1: Adjusted for sex, age, education level, smoking index, drinking index, BMI, family history of cancer, H. pylori infection a further adjusted for TG; b adjusted for TC; c adjusted for TG and LDL-C; d adjusted for TG and HDL-C. Model 2: Further adjusted for fruit intake, fat intake, pickled food intake, high salt diet, intake food temperature ρ for trend: The median of each serum lipid group was treated as a continuous variable into the model for the trend test | ||||||
In the female population, no significant associations were found between serum lipids and GC incidence (Table 3). While in the male population, we discovered a significant 1.73-fold increased risk of GC in Q4 stratum for TG stratifications with an adjusted HR of 1.73 (95% CI: 1.14–2.61), and the ASIR of the Q4 stratum was 1.79 times that of the Q1 strata. For HDL-C stratifications, the risk of GC was significantly lower in the Q2, Q3, and Q4 stratums compared with Q1 strata (HR for Q2 = 0.64, 95% CI: 0.45–0.90, HR for Q3 = 0.42, 95% CI: 0.28–0.63, HR for Q4 = 0.35, 95% CI: 0.22–0.55), and ASIR was decreased by 0.58, 0.54, 0.30 times compared to the Q1 strata, respectively. We found that the risk of GC in the Q3 and Q4 stratum of LDL-C stratifications was 1.61 times (HR = 1.61, 95% CI: 1.10–2.37) and 2.23 times (HR = 2.23, 95% CI: 1.52–3.27) higher than that in the Q1 strata, the ASIR was 1.19-fold and 1.42-fold higher than that of the Q1 strata, respectively.
| Serum lipid types | Female population | Male population | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| n/N | IR | ASIR | Multivariate- adjusted HR (95%CI) | n/N | IR | ASIR | Multivariate- adjusted HR (95%CI) | |||
| (/100k person-years) | (/100k person-years) | |||||||||
| TC a | ||||||||||
| Q1 | 8/4064 | 32.16 | 50.53 | 1.00 (Ref) | 51/6879 | 121.71 | 79.24 | 1.00 (Ref) | ||
| Q2 | 6/4743 | 24.73 | 19.02 | 0.62 (0.21,1.81) | 43/6503 | 108.65 | 70.56 | 0.88 (0.59,1.33) | ||
| Q3 | 6/4400 | 19.39 | 9.89 | 0.41 (0.13,1.26) | 52/6837 | 123.52 | 74.92 | 0.92 (0.62,1.36) | ||
| Q4 | 14/4895 | 47.98 | 25.70 | 0.65 (0.27,1.59) | 68/7321 | 150.93 | 90.85 | 1.13 (0.78,1.64) | ||
| ρ for trend | 0.517 | 0.526 | ||||||||
| TG b | ||||||||||
| Q1 | 4/4511 | 14.37 | 10.02 | 1.00 (Ref) | 39/6502 | 100.85 | 64.08 | 1.00 (Ref) | ||
| Q2 | 5/3839 | 21.88 | 18.63 | 1.51 (0.35,6.44) | 38/6844 | 91.14 | 55.15 | 0.81 (0.51,1.26) | ||
| Q3 | 11/5095 | 35.97 | 23.38 | 1.87 (0.49,7.08) | 61/6942 | 142.54 | 88.33 | 1.26 (0.83,1.91) | ||
| Q4 | 14/4657 | 50.02 | 30.33 | 2.27 (0.59,8.75) | 76/7252 | 167.14 | 122.30 | 1.73 (1.14,2.61) | ||
| ρ for trend | 0.137 | < 0.001 | ||||||||
| HDL-C c | ||||||||||
| Q1 | 10/4444 | 38.62 | 37.88 | 1.00 (Ref) | 93/6869 | 222.03 | 125.68 | 1.00 (Ref) | ||
| Q2 | 11/4539 | 40.38 | 21.87 | 1.47 (0.58,3.73) | 56/6849 | 132.51 | 79.38 | 0.64 (0.45,0.90) | ||
| Q3 | 7/4581 | 25.01 | 15.06 | 1.22 (0.42,3.46) | 35/6896 | 82.93 | 59.07 | 0.42 (0.28,0.63) | ||
| Q4 | 6/4538 | 21.32 | 36.99 | 1.18 (0.38,3.70) | 30/6926 | 70.96 | 48.27 | 0.35 (0.22,0.55) | ||
| ρ for trend | 0.886 | < 0.001 | ||||||||
| LDL-C d | ||||||||||
| Q1 | 5/4455 | 20.80 | 19.56 | 1.00 (Ref) | 39/6794 | 98.29 | 76.71 | 1.00 (Ref) | ||
| Q2 | 6/4498 | 21.50 | 18.49 | 1.51 (0.46,5.00) | 25/6866 | 58.36 | 40.39 | 0.70 (0.43,1.13) | ||
| Q3 | 8/4557 | 27.84 | 21.17 | 1.67 (0.54,5.12) | 64/6919 | 147.75 | 91.58 | 1.61 (1.10,2.37) | ||
| Q4 | 15/4592 | 52.48 | 30.09 | 2.35 (0.80,6.89) | 86/6961 | 200.92 | 109.20 | 2.23 (1.52,3.27) | ||
| ρ for trend | 0.137 | < 0.001 | ||||||||
| Abbreviations: TC total cholesterol; TG triglyceride; HDL-C high-density lipoprotein cholesterol; LDL-C low-density lipoprotein cholesterol; IR Incidence Rate; ASIR age-standardized incidence rate; HR hazard ratio; CI confidence interval Multivariate-adjusted for age, education level, smoking index, drinking index, BMI, family history of cancer, H. pylori infection, fruit intake, fat intake, pickled food intake, high salt diet, intake food temperature a further adjusted for TG; b adjusted for TC; c adjusted for TG and LDL-C; d adjusted for TG and HDL-C ρ for trend: The median of each serum lipid group was treated as a continuous variable into the model for the trend test | ||||||||||
Figure 1 indicated a linear dose-response relationship between TG with HR of GC (TG: ρoverall = 0.011, ρnon−linear = 0.152), and non-linear dose-response relationships between HDL-C and LDL-C with HRs of GC in the male population (HDL-C: ρoverall < 0.001, ρnon−linear = 0.047, LDL-C: ρoverall < 0.001, ρnon−linear = 0.038).
The combination of any abnormalities among TG, HDL-C, and LDL-C interactively elevates the incidence risk of GC (Table 4). The highest GC incident risk was plainly visible when both HLD-C and LDL-C were abnormal (HR = 5.38, 95% CI: 3.43–8.45). Compared to the normal TG and HDL-C group, the adjusted HR was 2.75 (95% CI: 1.89-4.00) for those with abnormal TG and HDL-C. The adjusted HR was 2.55 (95% CI: 1.78–3.64) in the abnormal TG and HDL-C group compared normal TG and LDL-C strata. The interaction effect of abnormal TC and TG on the risk of GC was the lowest (HR = 1.82, 95% CI: 1.28–2.60). The interaction tests by the GAM were shown in. In the tests of interaction by generalized additive model, linear predictors represent trends in outcome probabilities, so the probability of GC outcome increases with increasing linear predictors. The results of the interaction test showed that a combination of any abnormalities among TG, HDL-C, and LDL-C would interactively elevate the incidence risk of GC (Interaction tset ρ < 0.05). However, there was no interaction between abnormal TC and TG on the risk of GC. (Interaction tset ρ༞0.05 ) (Fig. 2).
| Groups of Interaction | n/N | IR (/100k person-years) | Multivariate- adjusted HR (95%CI) | Interaction tset* ρ value | |
|---|---|---|---|---|---|
| TG (mmol/L) | TC (mmol/L) | 0.626 | |||
| < .1.70 | < 5.20 | 67/20780 | 53.26 | 1.00 (Ref) | |
| < 1.70 | ≥ 5.20 | 24/4808 | 83.59 | 1.37 (0.86,2.20) | |
| ≥ 1.70 | < 5.20 | 98/12033 | 131.47 | 1.81 (1.32,2.48) | |
| ≥ 1.70 | ≥ 5.20 | 59/8021 | 120.82 | 1.82 (1.28,2.60) | |
| TG (mmol/L) | HDL-C (mmol/L) | 0.002 | |||
| < 1.70 | ≥ 1.00 | 65/24117 | 44.49 | 1.00 (Ref) | |
| < 1.70 | < 1.00 | 26/1471 | 309.01 | 3.17 (1.99,5.04) | |
| ≥ 1.70 | ≥ 1.00 | 101/15863 | 103.07 | 1.90 (1.38,2.60) | |
| ≥ 1.70 | < 1.00 | 56/4191 | 220.59 | 2.75 (1.89,4.00) | |
| TG (mmol/L) | LDL-C (mmol/L) | 0.022 | |||
| < 1.70 | < 3.40 | 52/17910 | 44.02 | 1.00 (Ref) | |
| < 1.70 | ≥ 3.40 | 39/5878 | 107.15 | 1.98 (1.31,3.01) | |
| ≥ 1.70 | < 3.40 | 81/12670 | 104.37 | 1.85 (1.30,2.63) | |
| ≥ 1.70 | ≥ 3.40 | 76/7384 | 166.06 | 2.55 (1.78,3.64) | |
| HDL-C (mmol/L) | LDL-C (mmol/L) | 0.131 | |||
| ≥ 1.00 | < 3.40 | 78/27456 | 46.86 | 1.00 (Ref) | |
| ≥ 1.00 | ≥ 3.40 | 88/12524 | 113.33 | 2.13 (1.49,3.04) | |
| < 1.00 | < 3.40 | 55/4924 | 187.82 | 1.81 (1.33,2.46) | |
| < 1.00 | ≥ 3.40 | 27/738 | 597.87 | 5.38(3.43,8.45) | |
| Abbreviations: TC total cholesterol; TG triglyceride; HDL-C high-density lipoprotein cholesterol; LDL-C low-density lipoprotein cholesterol; IR Incidence Rate; HR hazard ratio; CI confidence interval Multivariate-adjusted for age, education level, smoking index, drinking index, BMI, family history of cancer, H. pylori infection, fruit intake, fat intake, pickled food intake, high salt diet, intake food temperature *Interaction test by generalized additive models | |||||
We conducted sensitivity analyses to establish the robustness of the associations between lipid metabolism and GC (Supplementary Table 2). The results were nearly identical after excluding participants were diagnosed with obesity and Helicobacter pylori infection. Furthermore, we found that in different heavy metal exposure statuses and age stratifications, higher TG and higher LDL-C increase the risk of GC, and lower HDL-C decreases the risk of GC (Supplementary Tables 3 and 4). In addition, hyperlipidemia consistently revealed a significant positive relationship with GC, whereas hyperbilirubinemia, hyperuricemia, diabetes, hypertension, and obesity did not show any independent association with GC (Supplementary Table 5).
In this prospective cohort study, evidential results were found after average 8 years follow-up that higher TG, LDL-C, and lower HDL-C may substantially increase the incidence risk of GC. These associations were only observed in males with significant dose-response trends. Moreover, interactions among TG (beyond normal) / HDL-C (below normal) / LDL-C (beyond normal) widely correspondingly affect the incident risk of GC. Thereinto, the interaction between abnormal HDL-C and LDL-C could more effectively increase the incidence risk of GC, which was 5.38 times higher than the normal group.
TC did not show any independent association with GC incidence after adjusting for multivariate (HR for Q4 = 1.01, 95%CI: 0.71–1.44). A case-control study in Korea also supported no association between TC (OR = 1.28, 95% CI: 0.82–1.99) with GC (Pih et al. 2021). Replicated with Wulaningsih W et al., we observed similar results but from a prospective perspective that no association between TC and GC either according to quartiles (HR = 1.03, 95% CI: 0.80–1.32) or dichotomized values (HR = 1.04, 95% CI: 0.89–1.22) (2012). The reason was probably that the beneficial effects of HDL-C and the harmful effects of LDL-C on the risk of GC were offset.
Evidence of the link from serum TG to the GC prevalence has been widely demonstrated. However, the association from abnormal TG exposure to the GC incidence remained spare. Our study implicated a positive association between elevated TG and increased GC incidence (HR for Q4 = 1.53, 95% CI: 1.02–2.29). Similar to a national-based cohort study, Lim JH et al. recruited 2,722,614 participants after 8.26 years follow-up found that higher TG may increase the incidence risk of GC (HR = 1.70, 95% CI: 1.62–1.77) (2022). A Mendelian randomization study also provided evidence for a causal relationship between TG and GC (IVW: OR = 1.33, ρ = 0.024; ρhet = 0.378) (Wu et al. 2022). Results from studies on metabolic syndrome and GC further supported the perspectives listed above (Li et al. 2018). Contradicted results from cohort studies showed that higher TG might increase the risks of cancers, including lung, rectal, and thyroid cancer, but overall, no association has been found between TG and GC (Borena et al. 2011; Ulmer et al. 2009). The differences in conclusions may be due to ethnic differences and differences in dietary factors associated with GC. For those patients under susceptible state of gastric cancer, toxic products generated by the decomposition of superfluous TG could quickly reach through the gastric epithelium by H. pylori-damaged blood vessels, further accelerating the development of local inflammation, inflammatory infiltration develops with the release of related cytokines and eventually leads to GC (Wu et al. 2022). Besides, excessive lipolysis increased harmful free fatty acid synthetase (FAs) in the cytoplasm and promoted FA oxidation in mitochondria, which leads to the production of ROS that promote cancer progression (Osumi et al. 2016).
For cholesterol, HDL-C and LDL-C demonstrated an opposite effect on the trigging of GC. More specifically, higher HDL-C illustrated a protective effect on the incidence of GC (HR for Q4 = 0.42, 95% CI: 0.26–0.67). On the other hand, the increased incidence risk of GC among higher LDL-C was plainly visible (HR for Q4 = 2.21, 95% CI: 1.51–3.24) in our study. A similar result by Joo Hyun Lim et al. found that lower HDL-C and higher LDL-C increased the risk of GC incidence (Lim et al. 2022). Su Youn Nam et al. and Jung MK et al. further supported the results shown above that lower HDL-C (HR = 2.67 95% CI: 1.14–6.16) and higher LDL-C (HR = 1.74, 95% CI: 1.06–2.85) may have a dependent and positive correlation with GC development (Nam et al. 2019; Jung et al. 2008). Experimental studies have also shown that abnormal cholesterol levels accelerate cellular damage via oxidative stress and chronic inflammatory mechanisms, which may play an important role in triggering the initiation of tumors (Nam et al. 2019). Meanwhile, accumulating evidence has established a direct role of HDL in suppressing inflammation (De Nardo et al. 2014; Namiri-Kalantari et al. 2015). HDL-C scavenges cholesterol from macrophages through lipid transporters such as ATP-binding cassette transporter A1 (ABCA1) ATP-binding cassette transporter G1 (ABCG1), and scavenger receptor class B type1(SR-B1) (Wang et al. 2007), which was thought to be part of the anti-inflammatory mechanism of HDL-C. Furthermore, previous studies suggested that exogenous LDL-C may promote tumor cell growth through LDL receptors on tumor cells (Kalaivani et al. 2014). Chushi L et al. showed that elevated cholesterol levels could induce the increased activity of the cholesterol biosynthesis rate-limiting enzymes (HMGCR) and squalene monooxygenase (SQLE), which was proved to be a promotor of the growth and migration of tumor cells, including gastric cancer (Chushi et al. 2016). These may reveal the potential mechanism that abnormal HDL-C and LDL-C could promote the progression of GC by stimulating malignant transformation via proinflammatory, oxidative stress or suppressing the immune system.
The significant association between abnormal lipid metabolism (HR of Q4 TG: 1.73, 95% CI: 1.14–2.61), (HR of Q4 HDL-C: 0.35, 95% CI: 0.22–0.55), and (HR of Q4 LDL-C: 2.23, 95% CI: 1.52–3.27) and risk of GC were only observed in males. The reasons for these ‘specifical significances in males’ may underlie the following mechanisms. Firstly, the association of GC with serum lipids differed between intestinal-type and diffuse-type gastric cancer. A previous study found that most GC cases were intestinal-type in Chinese men, while women were mostly diffuse-type (Qiu et al. 2013). Secondly, the correlations were not easily detected due to the limited number of cases in females. Although the incidence of GC has decreased in recent years, the incidence remains higher in males than females globally (Yang et al. 2021). Furthermore, female sex hormones were considered to be protective factors for GC incidence (Nie et al. 2018). Moreover, it has been reported that higher serum levels are generally associated with a higher risk of H. pylori infection, which occurs more frequently in males (Gerig et al. 2013). Thus, our results may provide a potential approach for implementing gender-specific strategies to prevent GC.
We also observed substantial interactions between different lipids in trigging GC (Fig. 2), the effect of HDL-C and LDL-C on the risk of GC changed with the increasing TG level, and elevated TG contributed to an increased risk of gastric cancer caused by hypo-HDL-C or excess LDL-C. Particularly, the risk of gastric cancer was greatest with lower HDL-C and higher LDL-C. For excess TG with hypo-HDL-C (HR = 2.75, 95% CI: 1.89–4.99), excess TG with excess LDL-C (HR = 2.55, 95% CI: 1.78–3.64), and hypo-HDL-C with excess LDL-C (HR = 5.38, 95% CI: 3.43–8.45). Studies have shown that TG-rich lipoproteins lead to the activation of nuclear factor-kappa B (NF-Kβ), vascular cell adhesion molecule 1 (VCAM-1), and other inflammatory mediators (Welty 2013). For the downstream responders of LDL-C, LDL modifiers, acetylated LDL, and NO2-LDL, may participate in the toll-like receptors (TLRs) pathway by binding to CD36 and scavenger receptor type A (SR-A) to activate the NF-Kβ factor and further lead to inflammation (Kim et al. 2011). HDL-C prevents the pro-inflammatory effects of OX-LDL by inhibiting NF-Kβ activation (Matsunaga et al. 2003). In addition, inflammatory factors, namely interleukin-1β (IL-1β) and tumor necrosis factor-α (TNF-α), can disrupt the negative feedback regulation mechanism of LDLR (Chen et al. 2007), which leads to the persistent expression of LDLR and the simultaneous increase of TG and LDL-C levels (Khovidhunkit et al. 2004). Therefore, a combination of any abnormalities among TG, HDL-C, and LDL-C would interactively elevate the incidence risks of GC. Thereinto, the interaction between HDL-C (below normal) and LDL-C (beyond normal) could more effectively increase the incidence risk of GC. However, its biological and physiological carcinogenic effects need to be clarified by experimental studies.
The strength of our study was based on a large cohort of 48,001 participants with 8 years follow-up to prospectively investigate the epidemiologic relationships between lipid metabolism with incidence risk of GC, taking into account the impact of life-behavior factors such as smoking, drinking, dietary habits, etc. and H. pylori, on GC. Several limitations of the present study must be mentioned. Firstly, Given the within-individual variability in serum cholesterol during long-term follow-up, a single measurement of baseline lipid levels may lead to misclassification of individual serum lipid levels. Secondly, since clinical GC cases were collected from the medical record, there may be some selection bias. Finally, there was a lack of participant information on lipid-lowering medication. In general, any prospective study, randomized clinical trials, or mechanistic studies that can replicate or contradict our study were strongly warranted to fill the understanding gap between a complex association of serum lipids and GC development.
In conclusion, lipid metabolism abnormalities could be significant risk factors for GC. A combination of any abnormalities among TG, HDL-C, and LDL-C would interactively elevate the incidence risk of GC. Our results will provide strong evidence for the relationship between abnormal lipid metabolism with GC.
Acknwledgements We are extremely grateful the researchers of the Lanzhou University School of Public Health and the clinicians from the Workers’ Hospital of Jinchuan Group Co., Ltd. who have made significant contribution to this research.
This work was supported by Natural Science Foundation of China (No. 81673248). We are extremely grateful the researchers of the Lanzhou University School of Public Health and the clinicians from the Workers’ Hospital of Jinchuan Group Co., Ltd. who have made significant contribution to this research.
Acknwledgements We are extremely grateful the researchers of the Lanzhou University School of Public Health and the clinicians from the Workers’ Hospital of Jinchuan Group Co., Ltd. who have made significant contribution to this research.
Author contribution The work reported in the paper has been performed by the authors，the specific work assignments were as follows：study conception and design, acquisition of data: Yana Bai; Analysis, or interpretation of data: Zhiyuan Cheng and Jing Li; Drafting of the manuscript: Zhiyuan Cheng and Jing Li; Statistical analysis and discussion: Siyu Li, Desheng Zhang, Jingli Yang, Yarong Chen, Yujia Hu, Lulu Xu, Lizhen Zhang, Zhongge Wang, Ruirui Chen; Obtained funding: Yana Bai and Zhiyuan Cheng; Yana Bai had full access to all the data in the study. The first author takes responsibility for the integrity of the data and the accuracy of the data analysis. The final version for submission was approved by all co-authors.
Funding This study was funded by Natural Science Foundation of China (grant number: 81673248).
Data availability Under the premise of not infringing on the interests of researchers and the privacy rights of cohort members, the datasets analysed during the current study are available from the co-author Yana Bai on reasonable request.
Competing interests The authors declare no competing interests.
Conflict of interest The authors have no relevant fnancial or non-fnancial interests to disclose.
Ethics approval The study protocol was approved by the Ethics Committee of the School of Public Health, Lanzhou University, and the Worker Ethics Committee of JNMC Hospital upon the initiation of this study. All participants were fully informed of the research content before entering the cohort and signed informed consent.