Introduction
In 2020, lung cancer (LC) is the second most commonly diagnosed malignancy and the leading cause of cancer death worldwide, representing approximately 2.2 million cancers diagnosed and 1.8 million deaths [
1]. LC are categorized as small cell carcinoma (SCLC) and non-small cell carcinoma; the latter mainly represents by adenocarcinoma (LUAD) and squamous cell carcinoma (LUSC). Worldwide, more than 60% of LC-related deaths are attributable to smoking; others are caused by occupational workplace exposure, air pollution, and diet [
2]. Of note is that metabolic factors, such as high plasma glucose level and type 2 diabetes, are also associated with both LC development and prognosis [
2‐
6], indicating that metabolic factors should not be underestimated or even neglected in the prevention and management of LC. On the other hand, the prevalence of metabolic syndrome was rapidly increased over the last decades [
7,
8], suggesting a persistent increase of metabolic-related LC burden in the future.
Previous epidemiological studies have examined the association between metabolic factors and the risk of LC. For example, individuals with impaired fasting glucose or diabetes mellitus, or dyslipidemia had an increased risk of LC [
9‐
11]. However, studies reporting a null association remained [
12‐
15], indicating further investigations are warranted. The inconsistency may be attributed to the between-study heterogeneities and the inherent pitfalls of observational studies, including under-adjustment for confounders, small sample size, and reverse causality.
The accumulation of genetic information from genome-wide association study (GWAS) and the advent of genetic methods, such as Mendelian randomization (MR) analysis, provide us the opportunity to further understand the correlation between exposure and outcome. The MR estimates reveal the genetic association between exposure and outcome, which can be viewed as causation if all of the MR assumptions were met [
16]. There were several MR analyses have been performed for LC [
17‐
19]. However, the genetic associations between metabolic factors and LC have not been thoroughly studied in populations with different ethnicities. To fill this gap, we used GWAS summary data to examine the genetic associations of six metabolic biomarkers with LC and its histological subtypes in East Asians and Europeans.
Discussion
In this study, we used GWAS summary data of metabolic biomarkers and lung cancer to examine the genetic association between these two traits. In East Asians, the IVW method suggests that LDL, TC, and TG were inversely associated with LC. MVMR analysis reports a negative association of LDL and a positive association of TG, respectively, with LC. However, the negative association between LDL and LC was not replicated by either univariate or multivariable MR analysis in Europeans. Subgroup and sensitivity analyses gave similar results to the main analysis.
Blood lipids have long been reported to involve in LC development. However, evidence from epidemiological studies was scattered and inconclusive [
34,
35]. A meta-analysis showed a significant inverse association between HDL, TC, and the risk of LC, whereas reported a significantly positive association between serum TG levels and the LC risk [
35]. Chandler et al. using data from Women’s Health Study reported that HDL was negatively associated with LC risk, but a significant association was not detected for either TC or TG [
36]. In a Chinese population, Lyu et al. found that low levels of LDL were significantly associated with an increased risk of LC, whereas in subjects having high levels of serum LDL, the risk of LC was comparable with that of the reference group [
37]. No significant association was detected between HDL, TG, and LC in this study [
37]. Compared to these epidemiological studies, in our study, we reported similar findings by examining the association between genetically predicted blood lipids and LC. To our best knowledge, this is the first MR analysis to investigate the genetic association between common metabolic biomarkers and LC. Our findings not only provide complements to the previous results but also shed new light on the pathogenesis of LC.
To date, a few studies have reported a negative association between LDL and cancers. For instance, Alsheikh-Ali et al. reported an inverse association between on-treatment LDL levels and incident cancer in statin-treated patients enrolled in large randomized controlled trials [
38]. A prospective cohort study also revealed that circulating levels of LDL may be negatively associated with the risk of cancers (hazard ratio < 1), albeit the association estimates were statistically non-significant [
39]. A prospective cohort involved 68,759 Chinese male adults reported that circulating LDL levels was negatively associated with cancer risk (hazard ratio = 0.8) [
40]. However, the association between blood LDL levels and LC risk has not been thoroughly investigated in population-based studies. In the current study, we found an inverse association between LDL and LC using MR approaches. The mechanisms underlying the negative association between LDL and LC are not yet well understood, although the biological roles of LDL in carcinogenesis have been proposed [
41]. For example, low LDL has been proposed to be associated with suppressed immunity, upregulated activity or responsiveness of the mevalonate pathway, and increased activity of nuclear transcription factor NF-κB [
42], thus promoting the initiation and progression of cancer. Metabolites of cholesterol, such as bile acid, have also been implicated in cancer progression [
43]. However, the specific roles of LDL in lung tumorigenesis still need further investigations because LDL showed a distinct effect on the risk of different cancer sites, which may be driven by different mechanisms [
43]. The observed genetic association between LDL and LC may also attribute to confounders that cause both a low plasma LDL cholesterol level and an increased risk of LC. One potential confounder is smoking, which could lead to low levels of LDL and increase the LC risk [
44]. In the scenario of MR analysis, we could not completely tease out the confounding effect. Thus, concluding that LDL had a causal effect on LC should be cautious.
We also noted a significant association between genetically predicted TG and LC. In the univariate MR analysis, we found an inverse association between TG and LC. On the contrary, the MVMR analysis showed that genetically predicted TG was positively associated with LC. This finding is consistent between East Asians and Europeans and is in line with that of previous epidemiological studies [
45,
46] and suggests that estimates of univariate MR analysis may be biased by other lipids. One possible mechanism relating TG to LC is that hypertriglyceridemia is associated with the development of oxidative stress and reactive oxygen species (ROS) [
47]. On one hand, smoking is associated with elevated levels of TG [
48], which may confound the association between TG and LC. However, Ulmer et al. found that the association remained when the data set was limited to non-smokers, suggesting that factors other than smoking status may contribute to the observed association [
45].
We did not find a significant association between genetically predicted FPG and HbA1c and LC in both East Asians and Europeans, although the two diabetic factors have been shown to positively associate with LC risk in epidemiological studies [
4,
49]. Our findings are consistent with previous MR studies. Yuan et al. reported a null association between genetically predicted diabetes and LC [
50]. Torres et al. found that genetically predicted FPG was not significantly associated with LC, although there were only 24 SNPs that were used as the IVs [
51]. These MR findings suggest that FPG and HbA1c may not be independent predictors for LC, rather than reflecting a risk status predisposing to LC.
A major strength of this study is the MR study design, which diminishes confounding and reverse causality potentially biasing the results in observational studies. We conducted our analyses on East Asians and Europeans. Thus, the results are easy to be extrapolated and compared between populations. Our study also has limitations. First, although there was no horizontal pleiotropy detected by MR-Egger regression, we could not conclude that LDL and TG were causally related to LC risk because we still cannot exclude that there is any direct causal pathway from the exposure-related genetic variants to cancer. Second, our MR estimates are not strictly consistent in East Asians and Europeans, suggesting that ethnic background may play a role in the examined association that deserves further investigation.
In conclusion, our MR study provides genetic evidence that blood LDL and TG are associated with LC in different directions among East Asians. However, these associations are not observed in Europeans. We did not detect a significant association between genetically predicted HDL, TC, FPG, and HbA1c and LC and its subtypes. The associations that reported in epidemiological studies may be driven by confounders or reverse causality.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.