Lipid-dominated metabolites mediate the association between low body mass index and esophageal malignancy: a population-based nested case-control study

Abstract

Background

The biological mechanism underlying the association between low body mass index (BMI) and increased risk of esophageal malignancy remains uncertain. We aimed to explore the role of metabolic alterations in the effect of low BMI on esophageal malignancy.

Methods

We conducted a nested case-control study using a large-scale community-based screening cohort in a high-risk region for esophageal cancer (263 cases and 263 matched controls). Metabolomic profiling was performed based on untargeted ultra-high-performance liquid chromatography-tandem mass spectrometry, using the serum samples collected at enrollment prior to cancer diagnosis. Linear regression and conditional logistic regression were used to identify BMI-related metabolites and metabolites associated with esophageal malignancy, respectively, and the shared metabolites were included in mediation analysis.

Results

Low BMI was positively associated with esophageal malignancy (odds ratio (OR) = 1.8 (95% confidence interval: 1.1–2.8)). A total of 160 metabolites were identified to be associated with low BMI, and 107 metabolites were associated with esophageal malignancy. Among the 21 shared metabolites, four metabolites (three lipids and one amino acid) were identified as mediators for the association between low BMI and esophageal malignancy, with the mediation effect (OR) of 1.1 for each mediator. The metabolic signature constructed based on these mediators could explain 46.0% of the effect of low BMI on esophageal malignancy.

Conclusions

Lipid-dominated metabolites mediate the association between low BMI and esophageal malignancy, which has provided crucial clues for understanding the etiology of esophageal malignancy and selection of early-warning metabolic biomarkers.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12885-026-16138-4.

Background

Esophageal cancer (EC), which can be mainly classified as esophageal squamous cell carcinoma (ESCC) and esophageal adenocarcinoma (EAC), serves as one of the most frequent and lethal cancers in the world [1]. Annually, over a half of the new cases and deaths related to EC emerge in China, with the majority being ESCC. Over the past decades, a series of epidemiological researches have been conducted in high-risk regions, reporting several risk factors for ESCC, such as higher age, family history of ESCC, lower body mass index (BMI), etc. [2, 3] BMI is used to assess the nutritional status of individuals. While obesity is one of the recognized risk factors for EAC, low BMI has been consistently found to be associated with increasing risk of ESCC [2, 4, 5], and the biological mechanism behind still remains uncertain.

Cancer has been increasingly recognized to be a metabolic disease, given that dysregulated cellular proliferation might be influenced by disturbed cellular metabolism [6]. A series of serum metabolites have been observed to be associated with ESCC, providing clues that the occurrence of ESCC might be related to metabolic alteration, especially lipid metabolic pathways [7–10]. Besides, BMI has also been reported to be associated with alterations of certain metabolites, including but not limited to lipids and amino acids [11–13]. Previous studies have found BMI-related metabolites which are associated with increasing risk of breast cancer and colorectal cancer [14, 15]. However, there is a lack of evidence regarding the role of metabolic alterations in the relationship between BMI and ESCC. Given that low BMI to some extent reflects poor nutritional status and excessive energy consumption [16, 17], we hypothesized that metabolic alteration might be one of the important pathways for the effect of low BMI on esophageal malignancy.

In this study, we aimed to explore the role of metabolites in the association between low BMI and esophageal malignancy based on a large-scale community-based screening cohort.

Methods

Study participants

The flow chart of the study design and analysis is illustrated in Fig. 1. The subjects of this study were obtained from the Endoscopic Screening for Esophageal Cancer in China (ESECC) randomized controlled trial (ClinicalTrials.gov identifier: NCT01688908), which was conducted in Hua County located in the “Taihang Mountain Region” in rural China. The study design of ESECC trial has been described elsewhere [2, 18]. Briefly, from 2012 to 2016, 668 randomly selected villages in Hua County were allocated to screening arm (endoscopic examination with iodine staining) and non-screening arm at the ratio of 1:1 through blocked randomization, with a total of 33,948 village residents aged 45–69 years being enrolled.

Fig. 1.

Flowchart of the study design and analysis. Abbreviations: BMI, body mass index; ESECC, Endoscopic Screening for Esophageal Cancer in China

Prior to baseline endoscopic examination (for the screening arm) or at enrollment (for the non-screening arm), physical examination (including height and weight measurement), fasting serum sample collection, and questionnaire interview were conducted for each participant. For the screening arm, standard upper gastrointestinal endoscopies were performed by experienced physicians, and the biopsy specimens were sent to the Anyang Cancer Hospital, which is the sole tertiary hospital specialized in cancer diagnosis and treatment in the local region. Two pathologists from the Anyang Cancer Hospital reviewed the pathologic slides without knowledge of endoscopic findings and discrepancies in histologic diagnoses were adjudicated through consultation.

Case and control selection

The cases in this study were defined as esophageal malignancy, which included severe dysplasia and above (SDA) lesions in the esophagus (severe dysplasia, carcinoma in situ, and squamous cell carcinoma). The cases were further categorized into “screening-detected cases” and “follow-up cases”, the latter of which were captured through the follow-up framework of the ESECC trial including annual active door-to-door follow-up and passive linkage with local medical insurance claims data [19, 20].

Up to November 15th, 2020 (the longest follow-up time was 9.0 years), a total of 279 SDA cases (including 134 screening-detected cases and 145 follow-up cases) were detected in the ESECC cohort. Those cases without serum samples (n = 1) or complete questionnaire information (n = 15) were excluded, and 263 SDA cases were retained for analysis (none of the cases had been previously diagnosed with other cancers). For each case, one control was randomly selected among the cohort members who did not have esophageal malignancy at the time of diagnosis of the case based on incidence density sampling, with allocated arm, gender, age at blood draw (± 1 year), and date at blood draw (± 30 days) used for matching. Finally, a total of 524 individuals were selected in this study (two cases were also selected as controls, before their diagnosis, for other two cases based on incidence density sampling).

Serum sample collection and metabolomic profiling

At baseline, each participant provided a fasting blood sample. These samples were refrigerated overnight at 4 °C before extracting serum. The serum was temporarily stored at -20 °C and later stored at -80 °C when transported to Beijing.

Metabolites of the baseline serum samples were detected by using untargeted ultra-high-performance liquid chromatography-tandem mass spectrometry (UPLC-MS/MS). For UPLC methods, both reverse phase liquid chromatography based on C18 columns and hydrophilic interaction liquid chromatography column were used to guarantee the quantity and reliability of metabolite detection. The QE mass spectrometer analysis was alternated between MS and data-dependent MS2 scans using dynamic exclusion in the m/z range of 70–1,000 with a mass resolution of 35,000. Identification of metabolites was performed by searching the in-house library which has registered over 3300 purified standard compounds, and strictly adhered to the Metabolomics Standards Initiative tier 1 standard.

Quality control samples were generated by extracting a small volume of serum from each experimental sample, and were evenly spaced among the experimental samples which were randomly arranged. Carefully selected internal standards, which had no interference with the measurement of the serum metabolites, were added into each sample for chromatographic peak alignment and instrument stability monitoring. The median relative standard deviation (SD) for the standards among the quality control samples was calculated as 4.31% (< 5%), denoting a stable LC-MS system performance during the metabolomics assay.

Several pre-processing methods were used for the raw metabolomics data, including normalization, log transformation, scaling, and imputation of missing values (Supplementary Methods). Finally, a total of 823 metabolites were included in the analysis.

Statistical analysis

BMI was calculated as weight (kilograms) divided by height (meters) squared, and low BMI was defined as a BMI less than or equal to 22 kg/m². This definition is in accordance with previous studies which demonstrated that BMI ≤ 22 kg/m² was associated with the risk of esophageal malignancy [2, 21]. Differences of baseline characteristics between cases and matched healthy controls were tested using the McNemar test for categorical variables and the Wilcoxon signed-rank test for continuous variables.

The association of BMI (exposure) and esophageal malignancy (outcome) was first assessed using conditional logistic regression. For the mediation analysis, a preliminary selection of candidate metabolites was performed: (1) linear regression was used to assess the association between BMI and each metabolite; (2) conditional logistic regression was used to assess the association between each metabolite and esophageal malignancy, by calculating odds ratios (ORs) per one SD increase in metabolite abundance. Only those metabolites significantly associated with both BMI and esophageal malignancy (p < 0.05) were selected as candidates and included in the mediation analysis.

Mediation analysis was based on the following assumptions: there is no unmeasured confounding with respect to “exposure-outcome”, “mediator-outcome”, or “exposure-mediator” relationships given the adjusted covariates (age at blood draw, gender, allocated group, education level, household income per capita, family history of EC, cigarette smoking, alcohol consumption, eating speed, ingestion of leftovers, coal stove for heating, fume exposure in kitchen, and water source); there is no mediator-outcome confounder affected by the exposure. We decomposed the “total effect” of BMI on esophageal malignancy into a “natural indirect effect” (i.e., through metabolites) and a “controlled direct effect” (i.e., through other mechanisms). The mediated proportion was calculated by log (natural indirect effect) / log (total effect) ×100%. The Benjamini-Hochberg procedure was applied for multiple testing correction using false discovery rate (FDR), and those candidate metabolites with p < 0.05 and FDR-p < 0.2 were regarded as significant metabolite mediators.

In order to assess the integrated effect of the BMI on esophageal malignancy mediated through metabolic alteration, a metabolic signature was constructed by summation of the standardized abundance of each metabolite mediator multiplied by 1 (for mediators with positive association with exposure and outcome) or -1 (for mediators with negative association with exposure and outcome).

Statistical analysis was conducted using STATA version 15.0 (paramed module was used for mediation analysis). All tests were 2-sided and had a significance level of 0.05 unless otherwise specified.

Ethics statement

This study was approved by the Institutional Review Board of the Peking University School of Oncology, China (Approval number: 2011101110), and performed in accordance with the Declaration of Helsinki. All participants provided written informed consent.

Results

Participant baseline characteristics

Among the 263 cases of esophageal malignancy, severe dysplasia, carcinoma in situ, and squamous cell carcinoma accounted for 26.6%, 19.0%, and 54.4%, respectively. The distribution of age at blood draw, gender, allocated arm, education level, household income per capita, cigarette smoking, alcohol consumption, eating speed, ingestion of leftovers, coal stove for heating, fume exposure in kitchen, and source of drinking water were similar between 263 esophageal malignancy cases and 263 matched healthy controls. Cases were more likely to have lower BMI (≤ 22 kg/m²) (24.0% vs. 15.6%, p = 0.014) and a family history of EC (19.4% vs. 6.8%, p < 0.001) compared to controls (Table 1).

Table 1.

Baseline characteristics of esophageal malignancy cases and matched controls

Variables	Control n (%)	Case n (%)	p valuea
n	263	263	NA
Age at blood draw
Mean (SD)	62.3 (4.3)	62.3 (4.3)	0.544
Gender
Female	108 (41.1)	108 (41.1)	NA
Male	155 (58.9)	155 (58.9)
Allocated arm
Non-screening arm	77 (29.3)	77 (29.3)	NA
Screening arm	186 (70.7)	186 (70.7)
Stage
Severe dysplasia	—	70 (26.6)	NA
Carcinoma in situ	—	50 (19.0)
Squamous cell carcinoma	—	143 (54.4)
Body mass index
> 22 kg/m2	222 (84.4)	200 (76.0)	0.014
≤ 22 kg/m2	41 (15.6)	63 (24.0)
Education level
Middle school or above	78 (29.7)	83 (31.6)	0.604
Primary school or below	185 (70.3)	180 (68.4)
Household income per capita (¥)
>2000	83 (31.6)	82 (31.2)	0.922
≤ 2000	180 (68.4)	181 (68.8)
Family history of esophageal cancer
No	245 (93.2)	212 (80.6)	< 0.001
Yes	18 (6.8)	51 (19.4)
Cigarette smoking
No	187 (71.1)	184 (70.0)	0.753
Yes	76 (28.9)	79 (30.0)
Alcohol consumption
No	226 (85.9)	217 (82.5)	0.249
Yes	37 (14.1)	46 (17.5)
Eating rapidly
No	48 (18.3)	36 (13.7)	0.157
Yes	215 (81.7)	227 (86.3)
Ingestion of leftovers
No	169 (64.3)	155 (58.9)	0.186
Yes	94 (35.7)	108 (41.1)
Coal stove for heating
No	149 (56.7)	164 (62.4)	0.162
Yes	114 (43.3)	99 (37.6)
Fume exposure in kitchen
No	82 (31.2)	71 (27.0)	0.238
Yes	181 (68.8)	192 (73.0)
Water source
Shallow well	171 (65.0)	171 (65.0)	1.000
Deep motor-pumped well	92 (35.0)	92 (35.0)

Abbreviations: SD Standard deviation, NA Not applicable

^ap values were derived using the McNemar test (categorical variables) or the Wilcoxon signed-rank test (continuous variables)

Association between low BMI and esophageal malignancy

Based on conditional logistic regression, low BMI was positively associated with esophageal malignancy with an OR of 1.8 (95% confidence interval (CI): 1.1–2.8), before and after the adjustment of covariates.

Identification of metabolites associated with both low BMI and esophageal malignancy

A total of 160 metabolites were identified to be associated with low BMI, including 103 lipids (64.4%), 38 amino acids (23.8%), seven xenobiotics (4.4%), and 12 other classes (7.4%) (Supplementary Table S1). In addition, a total of 107 metabolites were identified to be associated with esophageal malignancy, including 53 lipids (49.5%), 19 amino acids (17.8%), 14 peptides (13.1%), six cofactors and vitamins (5.6%), six xenobiotics (5.6%), and nine other classes (8.4%) (Supplementary Table S2).

As shown in Table 2, a total of 21 metabolites were associated with both low BMI (13 positive associations, beta coefficients: 0.2–0.5; eight negative associations, beta coefficients: -0.4–-0.2) and esophageal malignancy (14 positive associations, ORs: 1.2–1.7; seven negative associations, ORs: 0.8), including 16 lipids (76.2%), three amino acids (14.3%), one carbohydrate (4.8%), and one xenobiotic (4.8%). These 21 metabolites were preliminarily selected as candidate mediators for the association between low BMI and esophageal malignancy.

Table 2.

Serum metabolites significantly associated with both low body mass index and esophageal malignancy

Metabolite	Class of metabolism	Subclass of metabolism	Associations between metabolites and low body mass index		Associations between metabolites and esophageal malignancy
			Beta (SE)	p value	OR per SD (95% CI)	p value
4-methyl-2-oxopentanoate	Amino Acid	Leucine, Isoleucine and Valine Metabolism	-0.3 (0.109)	1.4E-02	0.8 (0.6,1.0)	2.8E-02
S-methylmethionine	Amino Acid	Methionine, Cysteine, SAM and Taurine Metabolism	-0.2 (0.109)	2.3E-02	0.8 (0.6,0.9)	1.2E-02
pro-hydroxy-pro	Amino Acid	Urea cycle; Arginine and Proline Metabolism	0.5 (0.108)	1.8E-05	1.3 (1.1,1.6)	1.3E-02
ribose	Carbohydrate	Pentose Metabolism	-0.2 (0.109)	2.2E-02	1.4 (1.1,1.7)	2.1E-03
linoleoyl-docosahexaenoyl-glycerol (18:2/22:6) [2]	Lipid	Diacylglycerol	-0.4 (0.108)	5.2E-05	0.8 (0.7,1.0)	1.8E-02
linoleoyl-docosahexaenoyl-glycerol (18:2/22:6) [1]	Lipid	Diacylglycerol	-0.3 (0.109)	1.8E-03	0.8 (0.7,1.0)	3.2E-02
linoleoyl-linolenoyl-glycerol (18:2/18:3) [2]	Lipid	Diacylglycerol	-0.3 (0.109)	2.0E-03	0.8 (0.7,1.0)	2.4E-02
2-aminoheptanoate	Lipid	Fatty Acid, Amino	-0.2 (0.109)	2.3E-02	1.2 (1.0,1.4)	3.1E-02
sebacate (C10-DC)	Lipid	Fatty Acid, Dicarboxylate	0.3 (0.109)	1.8E-02	0.8 (0.7,1.0)	3.4E-02
1-oleoyl-GPC (18:1)	Lipid	Lysophospholipid	0.4 (0.108)	2.5E-04	1.3 (1.1,1.5)	7.8E-03
1-palmitoyl-GPE (16:0)	Lipid	Lysophospholipid	0.3 (0.109)	2.2E-02	1.3 (1.1,1.5)	1.0E-02
1-oleoylglycerophosphate (18:1)	Lipid	Lysophospholipid	0.2 (0.109)	3.4E-02	1.7 (1.3,2.1)	1.1E-04
1-(1-enyl-stearoyl)-GPE (P-18:0)	Lipid	Lysoplasmalogen	0.3 (0.109)	2.1E-03	1.5 (1.2,1.9)	8.7E-05
1-(1-enyl-palmitoyl)-GPE (P-16:0)	Lipid	Lysoplasmalogen	0.3 (0.109)	4.3E-03	1.6 (1.3,2.0)	9.1E-05
1-(1-enyl-oleoyl)-GPE (P-18:1)	Lipid	Lysoplasmalogen	0.3 (0.109)	9.5E-03	1.4 (1.1,1.7)	4.9E-03
1-(1-enyl-palmitoyl)-GPC (P-16:0)	Lipid	Lysoplasmalogen	0.3 (0.109)	1.8E-02	1.3 (1.1,1.6)	1.2E-02
1,2-dipalmitoyl-GPC (16:0/16:0)	Lipid	Phosphatidylcholine (PC)	0.4 (0.108)	4.2E-04	1.3 (1.1,1.5)	5.8E-03
1-stearoyl-2-oleoyl-GPS (18:0/18:1)	Lipid	Phosphatidylserine (PS)	0.2 (0.109)	4.4E-02	1.3 (1.1,1.5)	8.1E-03
glycerophosphoethanolamine	Lipid	Phospholipid Metabolism	0.3 (0.109)	1.9E-02	1.7 (1.4,2.1)	4.0E-06
glycerophosphoserine	Lipid	Phospholipid Metabolism	0.2 (0.109)	2.5E-02	1.5 (1.2,2.0)	1.2E-03
thioproline	Xenobiotics	Chemical	-0.2 (0.109)	4.8E-02	0.8 (0.6,1.0)	3.5E-02

Metabolites significantly associated with low body mass index (not exceeding 22 kg/m²) were determined by using univariate linear regression with p < 0.05. Metabolites significantly associated with esophageal malignancy were determined by using univariate conditional logistic regression with p < 0.05

Abbreviations: CI Confidence interval, OR Odds ratio, SD Standard deviation, SE Standard error

Mediation effect of metabolites on the association between low BMI and esophageal malignancy

Through mediation analysis, four of the 21 candidate mediators, including pro-hydroxy-pro (amino acid), linoleoyl-docosahexaenoyl-glycerol (18:2/22:6) [2] (lipid), 1-(1-enyl-stearoyl)-GPE (P-18:0) (lipid), and 1,2-dipalmitoyl-GPC (16:0/16:0) (lipid), were finally identified as mediators for the association between low BMI and esophageal malignancy. For these mediators, the indirect effect (OR) was 1.1, with the proportion of mediation ranging from 12.6% to 15.7%.

The metabolic signature was then calculated by summation of the standardized abundance of each metabolite mediator multiplied by the respective weight (1 and − 1 for positive and negative associations respectively between mediator and exposure / outcome), using the following formula:

The mediation effect (i.e., indirect effect) of the metabolic signature was 1.4 (95% CI: 1.2–1.6), and the metabolic signature could explain 46.0% of the effect of low BMI on esophageal malignancy (Table 3; Fig. 2).

Table 3.

Mediation effect of the association between low body mass index and esophageal malignancy through metabolites and integrated metabolic signature

Mediator	Class of metabolism	Total effect, OR (95% CI)	Direct effect, OR (95% CI)	Indirect effect, OR (95% CI)	Mediated proportion
pro-hydroxy-proa	Amino Acid	1.9 (1.2 ,3.0)	1.7 (1.1 ,2.8)	1.1 (1.0 ,1.2)	15.7%
linoleoyl-docosahexaenoyl-glycerol (18:2/22:6) [2]a	Lipid	1.9 (1.2 ,3.1)	1.8 (1.1 ,2.8)	1.1 (1.0 ,1.2)	13.8%
1-(1-enyl-stearoyl)-GPE (P-18:0)a	Lipid	1.9 (1.2 ,3.1)	1.8 (1.1 ,2.8)	1.1 (1.0 ,1.2)	13.4%
1,2-dipalmitoyl-GPC (16:0/16:0)a	Lipid	1.9 (1.2 ,3.0)	1.8 (1.1 ,2.8)	1.1 (1.0 ,1.2)	12.6%
Metabolic signature	NA	2.0 (1.2 ,3.2)	1.4 (0.9 ,2.3)	1.4 (1.2 ,1.6)	46.0%
Abbreviations: CI, confidence interval; FDR, false discovery rate; NA, not applicable; OR, odds ratio.

^aThese metabolites were selected from the 21 candidate mediators given that the mediation effect was statistically significant, with p < 0.05 and FDR-p < 0.2

Fig. 2.

Mediation effect of the metabolic signature on the association between low body mass index and esophageal malignancy. Abbreviations: OR, odds ratio

We further conducted subgroup analysis based on stage of esophageal malignancy. For severe dysplasia, the mediation effect of the metabolic signature was 1.3 (95% CI: 1.0-1.6), with a mediation proportion of 21.7%; for carcinoma in situ, the mediation effect was 1.4 (95% CI: 1.1–1.9), with a mediation proportion of 56.8%; for squamous cell carcinoma, the mediation effect was 1.5 (95% CI: 1.2–1.8), with a mediation proportion of 64.2%.

Discussion

Low BMI has been consistently reported to be a risk factor for ESCC [2, 4, 5], but the underlying mechanism is unclear. In addition to ESCC, previous studies have also reported that low BMI is one of the risk factors for gastric cancer, which is also a malignancy of the upper digestive tract [22–25]. Given that low BMI reflects poor nutritional status and excessive energy consumption to a certain extent [16], metabolic abnormalities might play an important role in the effect of low BMI on ESCC. In this study, we adopted nested case-control design on the basis of a community-based screening cohort, and first identified four lipid-dominated metabolites mediating nearly a half of the effect of low BMI on esophageal malignancy, which has shed light on the etiologic prevention for ESCC and selection of early-warning metabolic biomarkers.

In this study, we primarily identified 107 serum metabolites associated with esophageal malignancy, the majority of which belonged to lipid and amino acid. Some of these metabolites have been reported in other ESCC metabolomics researches using serum samples. For example, in the serum of ESCC patients, the levels of glutamate [26], 1-oleoyl-GPC (18:1), and 1,2-dipalmitoyl-GPC (16:0/16:0) [8] were observed to significantly increase, while the levels of glutamine, pyruvate, and alpha-tocopherol [27] significantly decreased. In addition, we identified a total of 160 BMI-related serum metabolites, the majority of which also belonged to lipid and amino acid, and ~ 30% of these metabolites have been reported previously [14, 15, 28–30]. For instance, alanine, tyrosine, valine [29], and sphingomyelins [28, 29] have been found to be positively associated with BMI, while a negative association has also been reported for guanidinosuccinate, hypotaurine, sebacate, and hexadecanedioate [28]. Among the BMI-related metabolites, 21 metabolites were also associated with esophageal malignancy, including three previously reported differential metabolites for ESCC (ribose, 1-oleoyl-GPC (18:1), and 1,2-dipalmitoyl-GPC (16:0/16:0)) [8, 31].

We further identified four metabolic mediators for the association between low BMI and esophageal malignancy, including three lipids and one amino acid. Among these, 1-(1-enyl-stearoyl)-GPE (P-18:0) and 1,2-dipalmitoyl-GPC (16:0/16:0) belong to glycerophospholipid metabolism, and linoleoyl-docosahexaenoyl-glycerol (18:2/22:6) [2] (a diacylglycerol) and pro-hydroxy-pro belong to glycerolipid metabolism and arginine and proline metabolism respectively. To integrate the mediation effects of metabolites, we constructed a metabolic signature based on these metabolic mediators. It was shown that the metabolic signature mediated nearly a half of the effect of low BMI on esophageal malignancy. This result signifies that lipid-dominated metabolic alterations might play an important role in the biological pathways from low BMI to the carcinogenic process of ESCC, and these metabolic mediators have potential to serve as warning biomarkers for early detection of esophageal malignancy. The subgroup analysis results suggested that the metabolic signature exerted a mediating effect in the relationship between low BMI and different stages of esophageal malignancy, supporting the robustness of the findings. It also further suggested that lipid-dominated metabolic alterations may have varying levels of contribution to different stages of esophageal malignancy, providing clues for subsequent exploration of etiological mechanisms.

To further ensure that the controls were appropriately representative, we also conducted a subgroup analysis that included only 186 cases and 186 controls in the screening arm (with no cancerous lesions detected by endoscopic examination in the controls). Analysis showed that the metabolic signature still significantly mediated the effect of low BMI on ESCC risk (mediation effect: 1.3; 95% CI: 1.1–1.6), with a mediation proportion of 38.4%. This result is close to the mediation proportion of 46.0% observed in the full sample, further supporting the robustness of the findings.

Glycerophospholipids are the main components of cell membrane lipids. The observed positive association between increased glycerophospholipids in blood samples and decreased BMI might indicate the release of tissue lipids due to negative energy balance (i.e., energy expenditure > energy intake) [32]. In addition, glycerolipid metabolism and diacylglycerols (precursors to triglycerides) are identified to be correlated to adipose tissue fat stores, and alterations in diacylglycerol composition might be indicative of lipolytic or lipogenic activity [33].

Several of these metabolite mediators and corresponding pathways have been reported in various cancers [34–37]. Metabolomics studies by Zang et al. and by Li et al. showed that arginine and proline metabolism and glycerolipid metabolism were two of the abnormal pathways for ESCC [34, 35]. Shu et al. observed that 1-(1-enyl-stearoyl)-GPE (P-18:0) was positively associated with gastric cancer [36], and Kamarajan et al. found that pro-hydroxy-pro could serve as one of the metabolic biomarkers to distinguish head and neck squamous cell carcinoma and healthy controls [37]. To facilitate growth and proliferation, cancer cells preferentially utilize glycolysis to generate a series of glycerophospholipids to meet cellular requirements for synthesis of new biomass and membrane mediated signaling [38]. In addition, the alteration of proline metabolism was observed to be associated with the oral cancer overexpressed 1 (ORAOV1) gene, which is frequently amplified in ESCC. During ORAOV1 overexpression, intracellular proline level increases and production of reactive oxygen species decreases, further inhibiting apoptosis and autophagy and promoting tumorigenicity [39].

Previous ESCC metabolomics studies were mostly conducted in clinical settings, where patients were mainly advanced-stage ESCC cases and often suffered from dysphagia, which might affect their energy intake and further lead to systemic metabolic disorders (i.e., alterations of metabolic levels) and malnutrition (i.e., low BMI), resulting in reverse causality. This study relies on the nested case-control design and the community-based screening cohort, which can ensure that the collection of serum samples and BMI data was prior to cancer diagnosis. Thus, it can largely reveal the effect of low BMI on early-stage carcinogenic process of ESCC as well as the role of metabolic alterations in it. Future studies in clinical settings could further investigate the effect of the observed lipid-dominated metabolites on prognosis of ESCC patients.

However, several limitations should also be noted. Firstly, this is a single-center study conducted in a high-risk region. The main findings need to be further validated among diverse populations in other high-risk and non-high-risk regions for ESCC. Secondly, physical examination and serum collection were conducted almost simultaneously at enrollment, limiting causal inference in terms of the relationship between low BMI and altered metabolic levels. Thirdly, this study lacks temporal data on individual metabolite levels. Since metabolic profiles are dynamic, it is suggested to involve repeated measurements of metabolite levels to provide further evidence for exploring the metabolic mechanisms underlying the etiology of ESCC. Furthermore, given that metabolite levels may be influenced by changes in dietary habits, it is recommended that longitudinal data on dietary habits be also collected to further control for the influence of confounding factors.

Conclusions

In summary, we have for the first time uncovered that lipid-dominated metabolites mediate the association between low BMI and esophageal malignancy to a significant extent. This study has provided crucial clues for understanding the etiology of esophageal malignancy, thus facilitating the establishment of tailored strategies for prevention and control of ESCC.

Abbreviations

BMI

Body mass index

CI

Confidence interval

EAC

Esophageal adenocarcinoma

EC

Esophageal cancer

ESCC

Esophageal squamous cell carcinoma

ESECC

Endoscopic Screening for Esophageal Cancer in China

FDR

False discovery rate

OR

Odds ratio

SD

Standard deviation

SDA

Severe dysplasia and above

UPLC-MS/MS

Ultra-high-performance liquid chromatography-tandem mass spectrometry

Authors’ contributions

YK, ZHe, ML: Conceptualization; ZHe, ML, HT: Formal analysis; ZHe, ML, HT, MW, CG, RX, FLi, AL, HY, LD, LS, QW, ZL, YL, FLiu, YP, ZHu, HChen, HCai: Investigation; HT, ML, ZHe, YK: Writing - original draft; HT, ML, ZHe, YK: Writing - review & editing; and all authors: read and approved the final manuscript.

Funding

This work was supported by the Key R&D Program of Ningxia Hui Autonomous Region (grant number 2025BEG01011), the Science and Technology Planning Project of Tibet Autonomous Region (grant number XZ202501JD0021), and the Beijing Natural Science Foundation (grant number 7222243). The funders of the study had no role in study design, collection, analysis, and interpretation of data, or writing of the report.

Data availability

Data used in this study are available from the corresponding authors upon reasonable request.

Declarations

Ethics approval and consent to participate

This study was approved by the Institutional Review Board of the Peking University School of Oncology, China (Approval number: 2011101110), and performed in accordance with the Declaration of Helsinki. All participants provided written informed consent.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.