基于胃癌危险因素和血清胃功能建立胃癌筛查策略的全国多中心研究

发布时间：2018-08-13 15:42

【摘要】：胃癌是最常见的消化道肿瘤之一。胃癌的发生与饮食生活习惯密切相关。血清胃功能,包括血清胃泌素-17和胃蛋白酶原,在西方国家又称为GastroPanel,常用于无创的胃癌筛查,但筛查效能存在争议。通过基于院内人群的大样本、连续病例的全国多中心横断面研究,进行了胃癌危险因素的调查与分析,分析了血清胃功能与胃癌的相关性,确立了血清胃功能指标用于胃癌筛查的最佳界限值,建立了基于胃癌危险因素和血清胃功能指标的胃癌风险预测模型及评分系统,初步探索出胃癌筛查策略。一、胃癌危险因素调查及分析(一)研究目的进行胃癌危险因素调查,较全面分析饮食、生活习惯和代谢综合征等因素与胃癌的相关性,为建立胃癌发生风险预测模型提供依据。(二)研究方法基于院内人群的横断面研究。采用问卷调查方式,调查研究对象诸如饮食、生活习惯及代谢综合征等胃癌危险因素。所有研究对象均按照问卷调查、抽血化验和胃镜精查等流程操作。以胃镜及病理诊断作为最后诊断的金标准,本研究共纳入病例12961例,包括非胃癌组12637例和胃癌组324例。采用logistic回归和卡方检验等统计方法进行分析。(三)结果1、病例特征本研究共入组病例12961例,包括非胃癌组12637例和胃癌组324例。非胃癌组包括非萎缩性胃炎5401例、萎缩性胃炎2950例(其中低级别上皮内瘤变282例)和其他诊断(包括消化性溃疡、反流性食管炎和胃息肉等)4286例。胃癌组包括早期胃癌127例(39.20%)和进展期胃癌197例(60.80%)。平均年龄为57.29±9.628。女性和男性病例分别为6545例(50.5%)和6416例(49.5%)。有症状和无症状的患者分别为10571例(81.6%)和2390例(18.4%)。2、胃癌危险因素的单因素分析本研究结果表明,男性、高龄、吸烟、饮酒、经常摄入隔夜菜、饮食偏咸、经常摄入腌煎熏制食品和体重指数偏瘦是胃癌高危因素,经常摄入新鲜蔬菜水果是胃癌的保护因素。胃癌发生风险随着年龄增大而增大。饮酒习惯是胃癌发生的高危因素,且随着饮酒量和饮酒年数增加而增加。对于女性人群糖尿病和高甘油三酯人群胃癌检出率较高,但差异无统计学意义,因此代谢综合征与胃癌相关性并不显著。3、胃癌危险因素的多因素分析经多因素logistic回归分析后,性别、年龄、体重指数、饮食咸淡习惯、煎炸食品摄入、腌制食品摄入是胃癌危险因素,新鲜水果摄入为胃癌保护因素。logistic回归模型的拟合优度和区分度良好。roc曲线下面积为0.737(95%ci:0.708-0.766)(p0.001)。(四)小结本研究通过大规模流行病学调查,较全面分析了饮食、生活习惯与胃癌的相关性,初步探讨了代谢综合征与胃癌相关性。经多因素logistic回归分析可知,男性、高龄、偏瘦、饮食偏咸、经常摄入腌制食品、经常摄入煎炸食品是胃癌危险因素,经常摄入新鲜水果是胃癌保护因素。二、血清胃功能与胃癌相关性及筛查效能分析(一)研究目的探讨血清胃功能与胃癌相关性,确定血清胃功能用于胃癌筛查的界限值,评估血清胃功能用于胃癌筛查的效能。(二)研究方法基于院内人群的横断面研究。用elisa法空腹检测血清胃功能(包括血清胃泌素-17、胃蛋白酶原Ⅰ和胃蛋白酶原Ⅱ等),然后行胃镜精查及活检。以胃镜及病理诊断作为金标准,纳入病例12961例,包括非萎缩性胃炎5401例、萎缩性胃炎2950例(其中低级别上皮内瘤变282例)、早期胃癌127例、进展期胃癌197例和其他诊断4286例。采用非参数检验、卡方检验、roc曲线下面积分析和logistic回归分析等多种统计方法。(三)结果1、血清胃功能与胃癌相关性分析与非胃癌组相比,胃癌组的血清胃泌素-17水平较高,血清胃蛋白酶原Ⅱ较高,而pgr较低;与非萎缩性胃炎组相比,萎缩性胃炎组血清胃蛋白酶原Ⅰ和pgr较低。2、血清胃功能对胃癌的筛查效能分析血清胃泌素-17、胃蛋白酶原Ⅱ和pgr是筛查胃癌的可靠血清学指标,三者的roc曲线下面积分别为0.621(95%ci:0.593-0.649)、0.622(95%ci:0.592-0.652)和0.643(95%ci:0.611-0.675)(p值0.001),最佳诊断界限值分别为3.61pmol/l、12.00ug/l和9.42。血清胃功能单项诊断中,血清胃泌素-17的敏感度最高(70.1%),而pgr的特异度最高(65.3%)。血清胃泌素-17诊断胃癌的敏感度、特异度、阳性预测值、阴性预测值、准确度分别为70.1%(227/324)、51.4%(6499/12637)、3.6%(227/6365)、98.5%(6499/6596)和51.89%(6726/12961)。血清胃蛋白酶原Ⅱ诊断胃癌的敏感度、特异度、阳性预测值、阴性预测值和准确度分别为58.6%(190/324)、62.5%(7892/12637)、3.9%(190/4935)、98.3%(7892/8026)和63.96%(8082/12961)。pgr诊断胃癌的敏感度、特异度、阳性预测值、阴性预测值、准确度分别为58.3%(189/324)、65.3%(8257/12637)、4.1%(189/4569)、98.4%(8257/8392)和64.65%(8446/12961)。血清胃功能联合诊断中,血清胃泌素-17联合pgr的诊断效能较好。当s-g-17联合pgr(即s-g-173.61pmol/l且pgr9.42)诊断胃癌时,敏感度、特异度、阳性预测值、阴性预测值和准确度分别为44.1%(143/324)、77.0%(9736/12637)、4.7%(143/3044)、98.2%(9736/9917)和76.22%(9879/12961)。当s-g-17联合pgⅡ(即s-g-173.61pmol/l且pgⅡ12ug/l)诊断胃癌时,敏感度、特异度、阳性预测值、阴性预测值和准确度分别为43.5%(141/324)、73.0%(9230/12637)、4.0%(141/3548)、98.1%(9230/9413)和72.3%(9371/12961)。3、血清胃功能对预测胃癌发生风险的评价由胃泌素-17和胃蛋白酶原联合的血清胃功能abc法,可以预测胃癌发生风险,筛选出胃癌高危人群。把pgr9.42视为pgr阳性(+),pgr≥9.42视为pgr阴性(-);把s-g-173.61pmol/l视为g-17阳性(+),s-g-17≤3.61pmol/l视为g-17阴性(-)。根据上述判断标准,所有研究对象依据血清胃功能结果分为a、b和c三组,a组为血清胃功能双阴性,b组为血清胃功能单阳性,c组为血清胃功能双阳性。分析结果表明,a、b和c三组的胃癌检出率分别为1.0%(51/5071)、2.7%(130/4846)和4.7%(143/3044);各组的胃癌发生风险从a组至c组依次增加,其中a组的胃癌发生风险最低而c组的胃癌发生风险最高。与a组相比,b组、c组两组的胃癌发生风险or值分别为2.532(95%:1.823-3.515)和4.392(95%:3.171-6.084)(p值0.001)。(四)小结通过大样本、连续病例的横断面研究,确定了针对院内人群血清胃功能筛查胃癌的最佳界限值,提出了预测胃癌发生风险的血清胃功能abc法。本研究结果表明,血清胃泌素-17、胃蛋白酶原Ⅱ和pgr是筛查胃癌的可靠血清学指标,血清胃泌素-17联合pgr筛查胃癌的效能最佳;血清胃功能abc法可以筛查出胃癌高危人群,c组发生胃癌风险最大。三、基于胃癌危险因素和血清胃功能的胃癌发生风险预测模型及其评分系统的建立(一)研究目的结合胃癌危险因素和血清胃功能指标建立胃癌发生风险预测模型及其评分系统,评估其筛查效能。(二)研究方法以胃镜及病理诊断作为金标准,本研究共纳入病例12961例,包括非胃癌组12637例和胃癌组324例。以胃癌危险因素分析为基础,结合血清胃功能指标,进行多因素logistic回归分析,建立胃癌发生风险预测模型及其评分系统,采用roc曲线下面积评估其对胃癌的筛查效能。最后通过后期入组人群对评分系统的准确性进行验证,验证人群共纳入病例918例,包括非胃癌组893例和胃癌组25例。(三)结果1、基于性别、年龄和血清胃功能的胃癌发生风险预测模型的建立由性别、年龄、s-g-17(≥3.61pmol/l)和pgr(9.42)四个因素建立的logistic回归模型,可以提高血清胃功能对于胃癌的筛查效能。此时的roc曲线下面积为0.754(95%ci:0.729-0.780),敏感度、特异度、阳性预测值、阴性预测值和准确度分别为82.4%(267/324)、56.2%(7106/12637)、4.6%(267/5798)、99.2%(7106/7163)和56.89%(7373/12961)。2、基于多个胃癌危险因素和血清胃功能的胃癌发生风险预测模型及其评分系统的建立与筛查效能评价由性别、年龄、体重指数、饮食咸淡习惯、腌制食品摄入、煎炸食品摄入、水果摄入、S-G-17(≥3.61pmol/L)和PGR(9.42)等9个胃癌危险因素建立的logistic回归胃癌发生风险预测模型,血清胃功能筛查效能又可以进一步提高。此时ROC曲线下面积为0.777(95%CI:0.751-0.802),敏感度、特异度、阳性预测值、阴性预测值和准确度分别为68.8%(223/324)、73.6%(9303/12637)、6.3%(223/3557)、98.9%(9303/9404)和73.50%(9526/12961)。根据logistic回归分析中各因素的OR值赋予分值,建立胃癌风险预测评分系统。该评分系统得分范围为0~24分,ROC曲线下面积为0.775(95%CI:0.749-0.800)。以11分作为界限值,将研究对象划分为胃癌低风险人群和高风险人群。高风险(11-24分)人群中胃癌检出率为5.5%(234/4255),显著高于低风险(0-10分)人群中胃癌检出率1.0%(90/8706)。若总评分以11分作为界限值,则评分系统对胃癌诊断的敏感度、特异度、阳性预测值、阴性预测值和准确度分别为72.2%(234/324)、68.2%(8616/12637)、5.5%(234/4255)、99.0%(8616/8706)和68.28%(7373/12961)。验证人群包括非胃癌组893例和胃癌组25例。按照胃癌发生风险评分系统标准,计算验证病例风险分值,最终总分值范围为0~20分,ROC曲线下面积为0.683(95%CI:0.583-0.783)。以11分作为界限值,验证人群分为胃癌低风险人群(0-10分)和高风险人群(11-20分),后者胃癌检出率明显高于前者(5.1%vs1.6%)。若总得分以11分作为界限值,则评分系统对胃癌诊断的敏感度、特异度、阳性预测值、阴性预测值和准确度分别为60.0%(15/25)、68.6%(613/893)、5.1%(15/295)、99.0%(613/623)和68.4%(628/918),与建模人群一致。(四)小结基于胃癌危险因素和血清胃功能指标建立的胃癌发生风险预测模型及评分系统,可以较好的预测院内人群胃癌发生风险,筛查出胃癌高危人群,节约医疗成本,提高早期胃癌诊断率。
[Abstract]:Gastric cancer is one of the most common cancers of the digestive tract.The occurrence of gastric cancer is closely related to dietary habits.Serum gastric function, including serum gastrin-17 and pepsinogen, is often used in noninvasive screening of gastric cancer in Western countries. A national multicenter cross-sectional study was conducted to investigate and analyze the risk factors of gastric cancer. The correlation between serum gastric function and gastric cancer was analyzed. The optimal limit value of serum gastric function index for screening gastric cancer was established. The risk prediction model and scoring system of gastric cancer based on the risk factors of gastric cancer and serum gastric function index were established. To find out the screening strategy for gastric cancer. First, to investigate and analyze the risk factors of gastric cancer. (1) To investigate the risk factors of gastric cancer, and to analyze the correlation between diet, lifestyle and metabolic syndrome and gastric cancer, so as to provide a basis for establishing a risk prediction model for gastric cancer. (2) Research methods based on cross-sectional study of hospital population. Methods: A questionnaire survey was conducted to investigate the risk factors of gastric cancer such as diet, living habits and metabolic syndrome. All the subjects followed the procedures of questionnaire survey, blood sampling test and gastroscopy. With gastroscopy and pathological diagnosis as the gold standard for final diagnosis, 12961 cases were included, including 12 cases of non-gastric cancer. 637 cases and 324 cases of gastric cancer group were analyzed by logistic regression and chi-square test. (3) Results 1. There were 12961 cases in this study, including 12637 cases in non-gastric cancer group and 324 cases in gastric cancer group. Other diagnoses (including peptic ulcer, reflux esophagitis, gastric polyp, etc.) included 4286 cases. The gastric cancer group included 127 cases (39.20%) of early gastric cancer and 197 cases (60.80%) of advanced gastric cancer. The average age was 57.29 [9.628]. The female and male cases were 6545 (50.5%) and 6416 (49.5%) respectively. 390 cases (18.4%). 2. Univariate analysis of risk factors for gastric cancer showed that male, old age, smoking, drinking, overnight vegetable intake, salty diet, frequent intake of pickled and smoked foods and lean body mass index were high risk factors for gastric cancer, and frequent intake of fresh vegetables and fruits were protective factors for gastric cancer. Drinking habit is a high risk factor for gastric cancer, and it increases with the increase of alcohol consumption and years of drinking. For women with diabetes and high triglycerides, the detection rate of gastric cancer is higher, but there is no significant difference, so the correlation between metabolic syndrome and gastric cancer is not significant. 3. Multivariate analysis of gastric cancer risk factors. After multivariate logistic regression analysis, sex, age, body mass index, salty eating habits, fried food intake, pickled food intake were risk factors for gastric cancer, fresh fruit intake was protective factors for gastric cancer. The goodness of fit and discrimination of logistic regression model were good. The area under ROC curve was 0.737 (95% ci: 0.708-0.766) (p0.001). (4) Summary Through a large-scale epidemiological survey, this study comprehensively analyzed the relationship between diet, living habits and gastric cancer, and preliminarily explored the relationship between metabolic syndrome and gastric cancer. Fresh fruit intake is a protective factor for gastric cancer. 2. Correlation between serum gastric function and gastric cancer and analysis of screening efficacy (1) To explore the relationship between serum gastric function and gastric cancer, to determine the threshold value of serum gastric function for gastric cancer screening, and to evaluate the effectiveness of serum gastric function in gastric cancer screening. (2) Research methods based on the hospital population. Cross-sectional study. Gastric function in serum (including serum gastrin-17, pepsinogen I and pepsinogen II) was measured by ELISA on an empty stomach. Gastroscopy and biopsy were performed. 12961 cases, including 5401 cases of non-atrophic gastritis and 2950 cases of atrophic gastritis (including low-grade intraepithelial neoplasia) were included by gastroscopy and pathological diagnosis as gold standard. There were 282 cases of gastric cancer, 127 cases of early gastric cancer, 197 cases of advanced gastric cancer and 4286 cases of other diagnoses. Serum pepsinogen_and PGR were lower in atrophic gastritis group than in non-atrophic gastritis group. serum pepsinogen_and PGR were lower in atrophic gastritis group. 95% ci: 0.593-0.649, 0.622 (95% ci: 0.592-0.652) and 0.643 (95% ci: 0.611-0.675) (p value 0.001), respectively. The best diagnostic limits were 3.61 pmol/l, 12.00 ug/l and 9.42. The sensitivity of serum gastrin-17 was the highest (70.1%) and the specificity of PGR was the highest (65.3%). The positive predictive value and negative predictive value were 70.1% (227/324), 51.4% (6499/12637), 3.6% (227/6365), 98.5% (6499/6596) and 51.89% (6726/12961), respectively. 3% (7892 / 8026) and 63.96% (8082 / 12961). The sensitivity, specificity, positive predictive value and negative predictive value of PGR were 58.3% (189 / 324), 65.3% (8257 / 12637), 4.1% (189 / 4569), 98.4% (8257 / 8392) and 64.65% (8446 / 12961) respectively. In the combined diagnosis of serum gastrin-17 and pgr, the diagnostic efficacy was better. Sensitivity, specificity, positive predictive value, negative predictive value and accuracy were 44.1% (143 / 324), 77.0% (9736 / 12637), 4.7% (143 / 3044), 98.2% (9736 / 9917) and 76.22% (9879 / 12961) respectively in the diagnosis of gastric cancer. Heterogeneity, positive predictive value, negative predictive value and accuracy were 43.5% (141/324), 73.0% (9230/12637), 4.0% (141/3548), 98.1% (9230/9413) and 72.3% (9371/12961) respectively. Serum gastric function ABC combined with gastrin-17 and pepsinogen could predict the risk of gastric cancer. PGR 9.42 was regarded a s PGR positive (+), PGR > 9.42 a s PGR negative (-), s-g-173.61 pmol / L A s G-17 positive (+), S-G-17 < 3.61 pmol / L A s G-17 negative (-). According to the above criteria, a l l the subjects were divided into three groups according to the results of serum gastric function: a, B and c, group A was serum gastric function double negative, group B was serum gastric function negative. The results showed that the detection rates of gastric cancer in group a, B and C were 1.0% (51/5071), 2.7% (130/4846) and 4.7% (143/3044), respectively. The risk of gastric cancer in each group increased from group A to group c, with the lowest risk of gastric cancer in group A and the highest risk of gastric cancer in group C. The OR values of gastric cancer risk in the two groups were 2.532 (95%:1.823-3.515) and 4.392 (95%:3.171-6.084) (p value 0.001). (4) Summarize the best threshold of serum gastric function screening for gastric cancer in hospital population by cross-sectional study of large sample and consecutive cases. The results showed that serum gastrin-17, pepsinogen II and PGR were reliable serum markers for screening gastric cancer, and serum gastrin-17 combined with PGR was the best for screening gastric cancer; ABC method of serum gastric function could screen high-risk group of gastric cancer, and group C had the greatest risk of gastric cancer. 3. gastric cancer based on risk factors of gastric cancer and serum gastric function had wind Establishment of risk prediction model and scoring system (1) Objective To establish a risk prediction model and scoring system for gastric cancer in combination with risk factors of gastric cancer and serum gastric function indicators, and to evaluate its screening effectiveness. (2) Gastroscopy and pathological diagnosis were used as gold standard in the study. A total of 12961 cases, including 12637 cases of non-gastric cancer, were included. Based on the analysis of risk factors of gastric cancer, combined with serum gastric function index, a multivariate logistic regression analysis was conducted to establish a risk prediction model and a scoring system for gastric cancer. The area under the ROC curve was used to evaluate the screening efficacy for gastric cancer. Finally, the accuracy of the scoring system was assessed by the later grouping population. Results 1. Establishment of a logistic regression model based on sex, age, S-G-17 (> 3.61 pmol/l) and PGR (9.42) could improve the serum gastric function. The area under the ROC curve was 0.754 (95% ci: 0.729-0.780), sensitivity, specificity, positive predictive value, negative predictive value and accuracy were 82.4% (267/324), 56.2% (7106/12637), 4.6% (267/5798), 99.2% (7106/7163) and 56.89% (7373/12961) respectively, based on multiple risk factors and serum gastric function. Establishment of risk prediction model and scoring system and evaluation of screening efficacy were based on nine risk factors including sex, age, body mass index, salty eating habits, pickled food intake, fried food intake, fruit intake, S-G-17 (>3.61 pmol/L) and PG (9.42) for gastric cancer, serum gastric function. The area under the ROC curve was 0.777 (95% CI: 0.751-0.802), sensitivity, specificity, positive predictive value, negative predictive value and accuracy were 68.8% (223/324), 73.6% (9303/12637), 6.3% (223/3557), 98.9% (9303/9404) and 73.50% (9526/12961) respectively. The scoring system ranged from 0 to 24 and the area under the ROC curve was 0.775 (95% CI: 0.749-0.800). The subjects were divided into low-risk group and high-risk group with 11 points as the threshold value. The detection rate of gastric cancer in high-risk group (11-24 points) was 5.5% (234/4255), significantly higher than that in low-risk group (95% CI: 0.749-0.800). The sensitivity, specificity, positive predictive value, negative predictive value and accuracy of the scoring system for gastric cancer were 72.2% (234/324), 68.2% (8616/12637), 5.5% (234/4255), 99.0% (8616/8706) and 68.28% (7373/12961) respectively. According to the criteria of gastric cancer risk scoring system, the final total score ranged from 0 to 20, and the area under the ROC curve was 0.683 (95% CI: 0.583-0.783). With 11 as the threshold, the validated population was divided into low-risk group (0-10 points) and high-risk group (11-20 points). The latter was screened for gastric cancer. The sensitivity, specificity, positive predictive value, negative predictive value and accuracy of the scoring system were 60.0% (15/25), 68.6% (613/893), 5.1% (15/295), 99.0% (613/623) and 68.4% (628/918), respectively, if the total score was 11 as the threshold value. The risk prediction model and scoring system of gastric cancer based on risk factors and serum gastric function index can better predict the risk of gastric cancer in hospital population, screen out the high risk group of gastric cancer, save medical cost and improve the diagnostic rate of early gastric cancer.
【学位授予单位】：第二军医大学
【学位级别】：博士
【学位授予年份】：2017
【分类号】：R735.2

【相似文献】