研究目的：食管癌（Esophageal Cancer，EC）是全球致死率较高的消化道恶性肿瘤，尤其在我国部分高发地区，其发病率与死亡率持续居高不下。既往研究提示，膳食因素可能在EC的发生发展中发挥一定作用，然而现有证据多源于西方人群或非高发区，尚缺乏针对中国高发区人群膳食模式的系统研究。此外，膳食炎症指数（Dietary Inflammatory Index，DII）及其能量调整版本（Energy-Adjusted DII，E-DII）已被广泛应用于癌症流行病学研究，用以衡量膳食的炎症潜力，但其在中国高发区人群中的适用性与预测效能尚未明确验证。因此，本研究通过三种互补性研究设计，系统评估膳食模式质量及膳食炎症潜力与EC及癌前病变（Esophageal Precancerous Lesions，EPL）风险的关联，并基于本土膳食结构与炎症特征构建经验性膳食炎症指数（Empirical Dietary Inflammatory Index，EDII），以期为高发区EC的早期防控提供理论依据和实用工具。

研究方法：本研究整合横断面研究、前瞻性队列研究与病例对照研究三种互补研究设计。采用中国膳食平衡指数（Diet Balance Index-2016，DBI-16）与植物性膳食指数（Plant-based Diet Index，PDI）评估膳食模式质量；使用E-DII评估膳食的炎症潜力。横断面研究基于2017-2019年国家食管癌队列（National Cohort of Esophageal Cancer，NCEC）五个高发区的基线数据，采用多元Logistic回归模型分析DBI-16、PDI、E-DII与EPL及EC的相关性；前瞻性队列研究纳入44689名无癌个体，随访至2023年3月31日，构建Cox比例风险模型评估膳食指标与新发EC风险的关联；病例对照研究则基于2023年12月至2024年4月的筛查人群，结合膳食调查与血浆炎症标志物（C反应蛋白、白细胞介素-6、白细胞介素-10），应用降秩回归（Reduced Rank Regression，RRR）与多元回归方法构建本土化EDII，并对其诊断与预测效能进行验证和与E-DII的比较。

研究结果：横断面研究共纳入45205名参与者，诊断出1838例EPL与466例食管鳞癌（Esophageal Squamous Cell Carcinoma，ESCC）。研究发现，高发区人群普遍存在膳食结构失衡，表现为部分食物摄入不足或过量，整体膳食呈现较高的促炎倾向。

多因素Logistic回归结果表明，中高水平的摄入不足（Low Bound Score，LBS）与EPL显著相关（OR=1.39，95% CI：1.03-1.88）。在植物性膳食方面，总体PDI与EPL/ESCC合并组（OR_{Q4 vs. Q1}=0.82，95% CI：0.68-0.99）及EPL组（OR_{Q4 vs. Q1}=0.80，95% CI：0.65-0.99）呈负相关；不健康PDI（uPDI）与EPL/ESCC合并组（OR_{Q4 vs. Q1}=1.51，95% CI：1.24-1.82）、EPL组（OR_{Q4 vs. Q1}=1.44，95% CI：1.16-1.78）及ESCC组（OR_{Q4 vs. Q1}=1.82，95% CI：1.20-2.78）均显著正相关。此外，高E-DII也与EPL/ESCC合并组（OR_{Q4 vs. Q1}=1.26，95% CI：1.08-1.47）、EPL组（OR_{Q4 vs. Q1}=1.23，95% CI：1.04-1.47）和ESCC组（OR_{Q4 vs. Q1}=1.40，95% CI：1.00-1.97）三类结局表现出正相关趋势。

前瞻性队列分析显示，DBI-16中的LBS与ESCC风险增加有关，每增加1个单位，风险增加6%（HR=1.06，95% CI：1.01-1.10）。相反，摄入过量（High Bound Score，HBS）评分与较低的风险相关（HR=0.49，95% CI：0.24-0.99）。在PDI各维度中，hPDI得分升高与ESCC风险下降相关（HR_{Q4 vs. Q1}=0.51，95% CI：0.26-0.98）；而高uPDI（HR_{Q4 vs. Q1}=2.12，95% CI：1.09-4.14）和E-DII（HR_{Q4 vs. Q1}=2.12，95% CI：1.20-3.75）与ESCC风险升高显著相关。

在病例对照研究中，通过RRR提取11个与炎症标志物（C反应蛋白、白细胞介素-6、白细胞介素-10）高度相关的食物组，构建了本土化的EDII。与E-DII相比，EDII在训练集中表现出较高的判别效能，EDII的AUC为0.64，明显高于E-DII的0.50（P<0.001）；在纳入年龄与地区等关键变量后，EDII的AUC进一步提升至0.79，在测试集中也达到0.80。加权Logistic模型分析显示，EDII最高四分位（Q4）人群的疾病风险显著高于最低组（Q1），OR为4.16（95% CI：1.92-9.01）；在测试集中，经Firth校正后亦观察到类似结果（OR=3.02，95% CI：1.06-9.73）。相比之下，E-DII在同样模型中未见显著关联。

研究结论：本研究本研究基于多中心数据，系统评估了中国EC高发区人群膳食模式质量与膳食炎症潜力在EPL及EC风险中的关联。结果提示，健康植物性膳食（hPDI）与适度营养摄入可能与较低的风险水平相关，而不健康植物性膳食（uPDI）与促炎饮食（高E-DII）则与风险升高有关。基于本土膳食结构和炎症特征构建的EDII在风险识别与预测效能方面优于传统E-DII，显示出良好的应用价值，未来可作为高危人群筛查与营养干预策略的实用工具。建议今后结合纵向追踪与组学数据，进一步探究饮食–炎症–癌变之间的机制联系，助力精准预防策略的制定与公共卫生政策的优化。

Background and Objective: Esophageal cancer (EC) is a gastrointestinal malignancy with a high global mortality rate, particularly in certain high-risk areas of China where incidence and death rates remain elevated. Although dietary factors have been implicated in the development of EC, most available evidence is derived from Western populations or non-high-risk areas, and there is a lack of systematic studies on the dietary patterns of populations in high-risk populations in China. Moreover, the Dietary Inflammatory Index (DII) and its energy-adjusted version (E-DII) have been widely applied to quantify the inflammatory potential of diets in cancer epidemiology, yet their applicability and predictive performance in Chinese high-risk populations have not been clearly established. This study aimed to comprehensively evaluate the associations of dietary pattern quality and dietary inflammatory potential with the risk of EC and esophageal precancerous lesions (EPL), using three complementary study designs. Additionally, we developed a localized Empirical Dietary Inflammatory Index (EDII) based on regional dietary and inflammatory characteristics to provide practical tools for early prevention and control of EC in high-risk areas.

Methods: This study integrated three complementary study designs: cross-sectional study, prospective cohort study and case-control study. The quality of dietary patterns was assessed using the Chinese Diet Balance Index-2016 (DBI-16) and Plant-based Diet Index (PDI); the inflammatory potential of diets was assessed using the E-DII. 

The cross-sectional analysis used baseline data from five high-risk regions in the National Cohort of Esophageal Cancer (NCEC, 2017–2019), applying multivariable logistic regression to examine associations between DBI-16, PDI, E-DII and the risk of EPL and EC. The prospective cohort study included 44,564 cancer-free individuals with follow-up until March 31, 2023, and constructed Cox proportional risk models to assess the association between dietary indicators and the risk of new-onset EC. And the case-control study was based on the screening population from December 2023 to April 2024, combining dietary surveys and plasma inflammatory markers (C-reactive protein, Interleukin-6, Interleukin-10), and applying the Reduced Rank Regression (RRR) method and multiple regression methods to construct a localized EDII, and its diagnostic and predictive efficacy was validated and compared with the E-DII.

Results: A total of 45,205 participants were enrolled in the cross-sectional study, and 1,838 cases of EPL with 466 cases of esophageal squamous cell carcinoma (ESCC) were diagnosed. The study found that dietary imbalance was common in the population of high prevalence areas, which was characterized by insufficient or excessive intake of some foods, and the overall diet showed a high pro-inflammatory tendency.

The results of multifactorial logistic regression showed that moderate to high levels of inadequate intake (Low Bound Score, LBS) were significantly associated with EPL (OR=1.39, 95% CI: 1.03-1.88). For plant-based diets, overall-PDI was negatively associated with EPL/ESCC combined group (OR_{Q4 vs. Q1}=0.82, 95% CI: 0.68-0.99) and EPL (OR_{Q4 vs. Q1}=0.80, 95% CI: 0.65-0.99); unhealthy PDI (uPDI) was significantly positively associated with EPL/ESCC combined group (OR_{Q4 vs. Q1}=1.51, 95% CI: 1.24-1.82), EPL (OR_{Q4 vs. Q1}=1.44, 95% CI: 1.16-1.78), and ESCC (OR_{Q4 vs. Q1}=1.82, 95% CI: 1.20-2.78). In addition, high E-DII was also associated with the combined EPL/ESCC group (OR_{Q4 vs. Q1}=1.26, 95% CI: 1.08-1.47), EPL (OR_{Q4 vs. Q1}=1.23, 95% CI: 1.04-1.47), and ESCC (OR_{Q4 vs. Q1}=1.40, 95% CI: 1.00-1.97) categories of outcomes showed positive correlation trends.

Prospective cohort analysis showed that higher LBS in the DBI-16 was associated with an increased risk of ESCC, with a 6% increase in risk for each unit increase in LBS (HR=1.06, 95% CI: 1.01-1.10). Conversely, a higher High Bound Score (HBS), which represents excessive intake, was associated with a lower risk of ESCC (HR=0.49, 95% CI: 0.24-0.99). Among the PDI dimensions, an elevated hPDI score was associated with a decreased risk of ESCC (HR_{Q4 vs. Q1}=0.51, 95% CI: 0.26-0.98); whereas high uPDI (HR_{Q4 vs. Q1}=2.12, 95% CI: 1.09-4.14) and E-DII (HR_{Q4 vs. Q1}=2.12, 95% CI: 1.20-3.75) were significantly associated with elevated ESCC risk.

In the case-control study, a localized EDII was constructed by RRR extracting 11 food groups highly correlated with inflammatory markers (C-reactive protein, Interleukin-6, Interleukin-10). Compared with the E-DII, the EDII showed higher discriminant efficacy in the training set, with an AUC of the EDII of 0.64, which was significantly higher than that of the E-DII of 0.50 (P<0.001); after the inclusion of key variables such as age and region, the AUC of EDII further increased to 0.79, and also reached 0.80 in the test set. Weighted logistic model analysis showed that the risk of EPL/ESCC was significantly higher in the highest quartile (Q4) of the EDII than in the lowest group (Q1), with an OR of 4.16 (95% CI: 1.92-9.01). Similar results were observed in the test set after Firth correction (OR=3.02, 95% CI: 1.06-9.73). In contrast, E-DII did not show a significant association in the same model.

Conclusion: This study systematically assessed the association between the quality of dietary patterns and the inflammatory potential of diets in EPL and EC risk in a population from high-risk areas in China based on multicenter data. The results suggested that a healthy plant-based diet (hPDI) with moderate nutrient intake may be associated with lower risk of EC, whereas an unhealthy plant-based diet (uPDI) with a diet with high inflammatory potential (high E-DII) was associated with elevated risk. The EDII constructed on the basis of indigenous dietary structure and inflammatory characteristics outperformed the traditional E-DII in terms of risk identification and prediction efficacy, showing good application value, and can be used as a practical tool for screening and nutritional intervention strategies for high-risk populations in the future. It is recommended to further investigate the mechanism of diet-inflammation-cancer linkage by combining longitudinal tracking and histological data, which will help to develop precise prevention strategies and optimize public health policies.