研究目的

从全球健康的视角出发分析全球胃癌疾病负担变化的社会经济差异。基于人群数据，探索胃癌发病风险社会不平等的成因；着眼于一级预防，量化可改变的危险因素对胃癌发病风险社会不平等的贡献；着眼于二级预防，探讨更加有效和公平的胃癌筛查目标人群选择策略。

材料与方法

基于全球疾病负担数据库，将204个国家和地区按照社会人口指数（Socio-Demographic Index, SDI）进行分组，采用Joinpoint回归计算1990-2019年间各个SDI分组的胃癌发病/死亡率的平均年度变化百分比（Average Annual Percent Change, AAPC）。采用随机截距增长混合模型拟合胃癌标化发病率和死亡率的变化轨迹。采用年龄-时期-队列模型探讨1900-1999年出生队列的胃癌发病/死亡风险的变化趋势。

基于2015-2017年一项开展于社区人群的多中心整群随机对照研究，以高发区筛查组人群为研究对象，使用修正的Poisson模型检验社会人口学因素与胃癌及各级前期病变的关联；以非高发区和高发区对照组人群为研究对象，使用中介分析计算可改变的胃癌危险因素对胃癌发病风险社会不平等的贡献；基于非高发区对照组人群数据，对“专家共识”筛查标准进行外部验证，使用Cox比例风险回归模型建立基于问卷的胃癌发病风险评分，比较二者在不同人群灵敏度的差异。

结果

高SDI国家的胃癌标化发病/死亡率下降速度最快，年龄标化发病率和死亡率的AAPC分别为-1.96 （95% CI -2.11, -1.81）和-2.48 （95% CI -2.54, -2.41），而低SDI和中低SDI国家胃癌标化发病率和死亡率下降较慢，年龄标化发病率和死亡率的AAPC分别为-1.03 （95% CI -1.08, -0.97）和-1.03 （95% CI -1.09, -0.97）。少数低和中低发展水平国家胃癌标化发病率和死亡率甚至有所上升。

社会经济变量与肿瘤性病变（上皮内瘤变和胃癌）之间存在有统计学意义的关联。未上过学的人群的胃癌发病风险是学历为大专及以上人群发病风险的2.81倍（Hazard Ratio [HR] 2.81；95% Confidence Interval [CI] 1.16, 6.81）。社会经济地位较低的人群相比较高的人群，胃癌发病风险增加78%（HR 1.78；95% CI 1.21, 2.61）。因果中介分析结果提示，对于社会经济地位与胃癌发病风险之间的关联，在可改变的危险因素中，膳食因素中介的比例最大，为33%，HR_间接效应为1.17（95% CI 1.03, 1.32）。

纳入性别、年龄、受教育水平、一级亲属患胃癌、经常持续性饮酒、水果和腌制食物摄入频次共7个变量构建胃癌发病风险预测模型。基于模型而来的风险评分取最优截断值时灵敏度为0.71，特异度为0.55，“专家共识”筛查标准的灵敏度为0.71，特异度为0.45。相比“专家共识”筛查标准，风险评分筛出同样多胃癌病例时可节省约19%的胃镜检查，并且在男性和女性之间灵敏度差异更小，在受教育水平较低群体中灵敏度有所提高。

结论

社会经济发展水平较高的国家的胃癌疾病负担下降速度更快。社会经济地位较高的群体相比较低群体的胃癌发病风险更低。就一级预防而言，膳食因素可以解释约三分之一的社会经济地位与胃癌发病风险之间的关联，对目标人群的膳食条件进行干预有望减轻胃癌发病风险的社会不平等。就二级预防而言，虽然目前“专家共识”的胃癌筛查人群选择标准有较高的灵敏度，但仍有改善的空间，风险评分有望可以节省筛查资源，促进筛查资源的公平分配。

Objectives

To analyze the socioeconomic differences in the changes in the global gastric cancer disease burden based on country-level data. We further explore the mechanisms underlying the social inequalities in gastric cancer risk based on population data. We focus on primary prevention and quantify modifiable risk factors' contribution to social disparities in gastric cancer risk. We then turn our eyes to secondary prevention and explore more effective and equitable gastric cancer screening strategies for population selection.

Materials and methods

Based on the Global Burden of Disease database, we applied the joinpoint regression to calculate the average annual percentage change of gastric cancer incidence/mortality from 1990 to 2019 for 204 countries. A random intercept growth mixed model was used to fit standardized incidence and mortality rates trajectories. We also performed the age-period-cohort model to explore trends of gastric cancer risk in the birth cohort from 1900 to 1999.

Based on a community-based multicenter cluster-randomized study, we investigated associations between sociodemographic factors and precancerous cascade lesions in the screening group living in high-risk areas. We further calculated the contribution of modifiable risk factors to the social inequality of gastric cancer risk based on the control group's data. The “Expert Consensus” criteria were externally verified. We established a questionnaire-based gastric cancer risk score. We compared it with the “Expert Consensus” criteria regarding equity and effectiveness.

Results

Significant disparities in the decline of gastric cancer burden were observed between countries with different socioeconomic development levels. A few countries with low and medium-low development levels have even experienced the increase of the standardized incidence and mortality rates.

We observed the association between socioeconomic variables and gastric neoplastic lesions. The risk of developing gastric cancer was 2.81 times higher among those with no schooling than among those with a college education or above (Hazard Ratio [HR] 2.81; 95% Confidence Interval [CI] 1.16, 6.81). Individuals with lower socioeconomic status (SES) had a 78% increased risk of gastric cancer compared with those with higher SES (HR 1.78; 95% CI 1.21, 2.61). The HR of Indirect Effect for socioeconomic status on gastric cancer risk through dietary factors was 1.17 (95% CI 1.03, 1.32), and dietary factors could explain 33% of the association between SES and gastric cancer risk.

Gender, age, education, first-degree relatives suffering from gastric cancer, frequent and persistent alcohol consumption, and frequency of fruit and preserved food intake, were included to construct a risk prediction model for gastric cancer. The "Expert Consensus" screening criteria had a sensitivity of 0.71 and a specificity of 0.45. Using the optimal cut-off value, the risk score had a sensitivity of 0.71 and a specificity of 0.55. In the same case of screening 50 gastric cancer cases, the risk score can save about 19% of gastroscopy compared with the " Expert Consensus " screening criteria. Risk scores showed more minor differences in sensitivity between men and women than and increased sensitivity in less-educated groups.

Conclusion

Countries with higher socioeconomic development levels experienced a faster decline in the gastric cancer burden. Individuals with higher SES had a lower gastric cancer risk. Dietary factors can explain about one-third of the association between SES and gastric cancer risk in terms of primary prevention. Interventions on the dietary quality in the vulnerable population are expected to reduce social inequalities in gastric cancer risk. Regarding secondary prevention, although the current "Expert Consensus" criteria have a high sensitivity, there is still room for improvement. Risk scoring is expected to save screening resources and promote fair distribution of screening resources.