【作者】 王雪；

【导师】 刘娅；

【作者基本信息】 吉林大学 ， 卫生毒理学， 2012， 博士

【摘要】 自上世纪50年代初期以来，油田企业为我国的经济发展做出了巨大的贡献。可持续发展战略是油田科学开发利用的长久战略，而数十万油田职工的健康状况成为实施这一战略的重要支撑内容之一。伴随着我国疾病谱的变迁和职业病相关法律法规的实施，一方面，油田职工传染病、慢性营养不良及职业病的危害得到了有效遏制，平均寿命得到延长；另一方面，由于人们的生活方式发生改变，某些疾病相关因素的微效应通过不断累积及交互作用，最终导致以慢性病高发为特点的新的疾病谱表征。这也是国内外油田职工普遍存在的问题。而我国人口老龄化的进程更凸显了整体疾病谱中慢性病趋于井喷的严重态势，因此，掌握油田职工的健康相关因素、预警机制及防治对策迫在眉睫。目的：汇总并分析2007-2009年油田职工既往体检资料结果，开展油田职工健康现状的横断面研究，建立油田职工健康电子档案，掌握油田职工的主要健康问题，建立BP神经网络结构模型和Logistic回归模型，比较并结合二者的优势，筛选危险因素，分析权重，探索提高大庆油田职工群体健康水平的实现途径。同时，应用ICP-MS对全血中Mg、Ca、Cr、Mn、Fe、Cu、Zn、Se、Mo、Al、V和Rb共12种元素进行检测，寻找与高血压病、糖尿病相关的元素及存在交互作用的元素,为提高油田职工的整体健康水平提供原始数据积累和科学依据。方法：1、按照多阶段（分层、单纯随机和整群）抽样方法，随机抽取2007～2009年化工集团、采油厂（二矿区）和钻探集团（二集团）累计17149份职工体检纸板档案，剔除新入职人员。根据检查项目和结果（血压、B超、心电、X线、电测听及其他生化检查等）建立Excel数据库，进行数据录入。利用SPSS17.0软件进行统计分析，用χ2检验进行率比较。2、按照分层和整群抽样相结合的原则，选择接触危害因素种类繁多的钻探集团、采油厂、天然气集团及不接触职业危害因素的物业集团作为抽样单位，整群抽取各自下属分公司职工作为研究对象，进行一对一访谈式问卷调查（包括职工基本情况，如年龄、性别、文化程度、婚姻及经济状况等；生活方式，如饮食状况、烟酒史、体育锻炼及睡眠情况等；既往病史和现病史等），并进行体格检查（包括身高、体重、血压、腰围、B超、心电图、幽门螺杆菌及其他生化指标等），利用EpiData3.0软件建立职工健康状况电子档案（有效记录数5953个），分析职工基本情况、行为生活方式及既往病史和现病史及异常指标的分布。3、分别利用Matlab7.0软件和SPSS17.0软件建立BP神经网络结构模型和Logistic回归模型，并结合二者优势，分析病因网中有独立作用的因素的权重。利用Matlab7.0软件，采用最小二乘法分析可控因素在疾病预防和控制过程中的可调控程度。4、应用病例对照研究，分别按照高血压病和糖尿病的诊断标准，随机抽取不接触危害因素的高血压病职工219例，糖尿病职工209例，并按年龄和性别进行1：1匹配，选取对照组职工共428例（排除心脑血管系统相关疾病、糖尿病，且无急性感染及化学毒物接触史者）。采用ICP-MS对全血中12种元素进行检测，比较病例组与对照组之间的差异，正态分布数据用t检验，非正态分布用秩和检验。结果：1、2007-2009年既往健康体格检查资料表明，脂肪肝、血压偏高、尿常规异常等的检出率均居前列。其中，钻探集团职工患病率较高。另外，该集团职工噪声性听力损伤的检出率也较高，2009年达8%，应引起足够重视。2、通过现况调查发现，受检油田职工中超重肥胖患病率较高(60.4%)，其它五种疾病依次为：高脂血症（58.4%）、脂肪肝（39.0%）、高血压病（16.1%）、心脏病（8.3%）、糖尿病（5.0%）。其中高脂血症以混合型（35.3%）和低危型（46.5%）为主。上述各疾病主要表现为男性高于女性，随年龄增加患病率增加，50岁以上人群患病率较高。3、四个厂区健康状况不完全相同，各个厂区患病居前三位的疾病相同，为超重及肥胖、高脂血症、脂肪肝。其中钻探集团这三种疾病患病率均较高，分别为69.9%、63.6%和48.0%，其肝功异常检出率也最高（37.8%）。天然气集团高血压病（19.8%）和糖尿病（6.9%）的患病率较高；采油厂职工幽门螺杆菌检出率较高（42.4%）。4、无遗传史和不良生活方式的职工生活方式病的患病率最低(29.3%)，与其相比，有遗传史和不良生活方式者患病率最高(38.6%)；仅有不良生活方式或遗传史者患病率分别是32.1%和34.7%。5、利用Logistic回归模型获得高血压病、糖尿病和心脏病的独立危险因素，在此基础上，利用BP神经网络模型对各个危险因素赋予权重，归一化后设置危险度分级。其中，高血压病的一般危险因素：有父母遗传史、不运动；中度危险因素：男性（高于女性）、体质指数高（BMI）、睡眠时间小于5小时、心电图异常、脂肪肝；高度危险因素：年龄大和高GGT者。糖尿病的一般危险因素：脂肪肝、高CH、男性（高于女性）；中度危险因素：父方遗传史、尿常规异常、年龄大、高TG者。心脏病的一般危险因素：母方遗传史；中度危险因素：年龄大、父方遗传史、睡眠时间小于5小时和心电图异常者。6、BP神经网络结构模型和Logistic回归模型预测高血压病的分类一致率分别为90.27%和84.65%；预测糖尿病的分类一致率分别为88.29%和84.40%；预测心脏病的分类一致率分别为92.70%和91.6%。7、保证睡眠时间和适当的运动是预防高血压病、糖尿病和心脏病所需要共同关注的因素。戒烟对糖尿病和心脏病有较大的干预价值，限制饮酒对预防心脏病较重要。适当运动对高血压病和糖尿病的干预作用较心脏病更明显。8、高血压病患者全血中Al含量高于对照组（中位数：1.00ug/mlvs0.75ug/ml，P<0.05），Ca含量低于对照组（均数：57.34ug/mlvs59.5ug/ml，P<0.05），Mg含量低于对照组（均数：40.22ug/mlvs43.77ug/ml，P<0.01），Cr、Mn、Fe、Cu、Zn、Se、Mo、V、Rb元素及Zn/Cu值在病例与对照组之间的差异不显著，无统计学意义（P>0.05）；糖尿病患者全血中Al（中位数：2.68ug/mlvs0.69ug/ml，P<0.01）和Cu（均数：0.99ug/mlvs0.94ug/ml，P<0.05）含量高于对照组，而V（中位数：8.26ug/Lvs17.03ug/L，P<0.05）和Se浓度（均数：103.40ug/Lvs111.56ug/L，P<0.05）低于对照组，对照组Zn/Cu值比糖尿病病例组（6.32vs5.89，P<0.05）高，且差异显著，Mg、Ca、Cr、Mn、Fe、Mo和Rb元素差异不显著，无统计学意义（P>0.05）。9、元素聚类分析结果表明（以10为分界线时）：高血压病例组与对照组中Al和Ca元素与其他元素交互作用不明显，高血压患者Fe和Cr影响了Zn和Rb的相关系数，Cr、Se、Cu、Mg共同参与了高血压的发生发展。糖尿病病例组与对照组Al和V与其他元素交互作用不明显，Zn与Rb之间的相关转化为Zn与Se、Mg和Rb的相关，同时Se与Cr的相关系数（0.61vs0.34）减小。结论：1、回顾性研究和横断面调查结果表明，受检油田职工健康现况表现为慢性病高发的趋势，如超重和肥胖、高脂血症、脂肪肝、高血压病、心脏病、糖尿病等。其中高脂血症以混合型和低危型为主。各厂区健康状况分布不同，钻探集团职工患病率较高。各异常指标主要表现为男性高于女性，且随年龄增加患病率增加。2、性别和年龄是生活方式病的独立危险因素，且同时与生活方式之间具有交互作用，生活方式的改变对具有遗传史患者的干预作用更明显。有遗传史和不良生活方式者的患病率是无遗传史和不良生活方式者的1.32倍。3、不良的生活方式在不同疾病中所起的作用不同，同种生活方式在疾病预防和转归中作用不同，获取个人基本（性别、年龄、受教育程度等）、行为生活方式及常规体检指标等综合信息更有利于健康管理和健康促进。4、BP神经网络结构预警模型应用于预测高血压病、心脏病、糖尿病的正确率优于Logistic回归模型，二者结合可获得相关的独立危险因素及具有预警指导意义的指标权重。5、保证睡眠时间和适当的运动是预防高血压病、糖尿病和心脏病所需要共同关注的因素。戒烟对预防糖尿病和心脏病的发生具有重要作用，限制饮酒对预防心脏病较重要。适度运动对高血压病和糖尿病的干预作用较心脏病更明显。6、高血压病患者全血中Al含量高于对照组，而Ca、Mg浓度低于对照组；Cr、Zn、Rb和Fe元素可能与高血压病相关。糖尿病患者全血中Al、Cu含量高于对照组，而V、Se浓度低于对照组；Zn和Rb、Zn与Se、Mg与Rb之间的作用发生改变，Se与Cr的相关系数减小。本研究的创新点：1、利用BP神经网络结构模型，获得Logistic回归模型筛选的独立因素在病因网中的权重,并模拟正态分布对其进行预警分级，弥补了单一模型的局限性，并首次应用在健康体检中。2、利用最小二乘法对可调控因素在疾病转归及预防中的可调控程度进行数字化表达，兼顾实施预防和控制两个方向，更有利于健康促进。3、利用ICP-MS对全血中12种元素进行检测，并探讨了元素之间的交互作用，为信号通路中关键酶及信号转导途径的研究提供基础数据，同时为慢性病的营养干预提供依据。更多还原

【Abstract】 Since early-1950s the development of oilfields has brought great economic benefit to our country. As the long-term strategy of scientific development of oilfields, sustainable development largely depends on the health of hundreds of thousands of employees. On the one hand, with the change of disease pattern and implementation of occupational disease-related regulations and laws, infectious diseases, chronic malnutrition and vocational diseases have been effectively controlled, leading to prolonged average lifespan. On the other hand, with the great change in lifestyle, accumulation and interaction of illness-related subtle effect, a new spectrum of disease characterized by high incidence of chronic disease was finally produced, which is also a common problem throughout global oilfields, being aggravated by the rapidity of population aging process in our country. Therefore, it’s urgent and necessary to be in possession of the health status of oilfield employees, early warning system and the related control countermeasure.Objectives:To investigate the primary health problem of oilfield employees, explore the method of improving public health of Daqing oilfield workers, establish the back-propagation neural network model and Logistic regression forewarning model, look for hypertension-related and diabetes-related trace elements and study the interaction among them by detecting12kinds of elements (including magnesium, calcium, chromium, manganum, ferrum, copper, zinc, selenium, molybdanium, aluminium, vanadium and rubidium) in the whole blood by ICP-MS, accumulate original data of oilfield workers and provide scientific evidence for improving the health of oilfield workers.Methods:1The information on physical examination of2007-2009were randomly drawn, and Excel databases were established based on examination items. Chi-Square Test was made to analyse abnormal indexed using SPSS17.0.2According to the principle of stratified sample with cluster sample methods, four different groups-one company without occupational-disease-inductive factors and three departments exposed to various hazard factors such as drilling group, oil extraction plant, and natural group-were chosen as sampling unit. The staff and workers of subsidiary businesses were selected, all that were one-to-one interviewed by using questionnaire, and received physical checkup. The records of employee’s health were established using Epi Data3.0and analysed by SPSS17.0based on employee’s basic information, life style, illness history, and abnormal indexes.3Back-propagation neural network model and Logistic regression model were set up by Matlab7.0and SPSS17.0respectively. Combination of them were used to analyse the weight of independent factors. The least square method was used with the software of Matlab7.0, to analyse the regulation role of controllable factors in health promotion for patients and illness prevention for healthy population.4A case-control study was adopted, randomly selecting219cases of workers with hypertension,209cases of workers with diabetes without exposure to risk factors, according to the diagnostic criteria of hypertension and diabetes, respectively. Meanwhile,428workers were selected as control group,1:1matched by age and gender with the case group (exclude the cardiovascular system-related diseases, diabetes, acute infection and chemical toxicant exposure history).12kinds of elements in the whole blood of these workers were analysed by ICP-MS, and data were expressed by mean±standard deviation. T test or mann-whitney-wilcoxon test were used for the data which fits the normal distribution or not, respectively.Results:1The prevalence of chronic diseases ranked top five in the statistic based on the physical checkup information files from2007to2009, and employees with the high prevalence were distributed in drilling group and oil extraction plant. Furthermore, the positive rate of noise-induced hearing impairment was relatively higher, especially for drilling group at8%in2009, worthy of being paid enough attention.2For oilfield employees undergoing physical checkup, overweight (60.4%) ranked top of prevalence, followed by hyperlipidemia (58.4%), fatty liver (39.0%), hypertension (16.1%), heart disease (8.3%) and diabetes (5.0%). The primary forms of abnormal blood fat are mixed type and low-risk type. All abnormal indexes are characterized by the incidence higher rate of men than that of women. There is a trend of increased prevalence of all diseases with age, especially among people above50. 3There are some differences among the results of health investigation in four units but on the top three of prevalence are same-overweight,hyperlipidemia and fatty liver-which are69.9%,63.6%and48.0%respectively for drilling group. Positive rate of abnormal liver function is also higher (37.8%) in drilling group. The prevalence of hypertension and diabetes are19.8%and6.9%respectively in natural gas group. The positive rate of helicobacter pylori in oil extraction plant is the highest when compared with other plants.4The prevalence of people without inheritance and unhealthy lifestyles is lowest (29.3%), about38.6%of that with inheritance and unhealthy lifestyles,32.1%of that with unhealthy lifestyles alone, and34.7%of that with inheritance alone.5The lower risk factors of hypertension contain parental inheritance and lack of exercise; gender difference (male higher than female), sleep time less than5hours, abnormal changes of electrocardiogram, and patients with fatty liver come within the middle risk factors, and old age and high level of GGT are belong to the relative higher risk factors.The lower risk factors of diabetes patients are always with fatty liver, high level of CH, especially male. And the moderate risk factors always contain paternal inheritance, abnormal routine urine test, old age and high level of TG.Low risk factors of heart disease are always relevant to maternal inheritance; middle risk factors contain old age, paternal inheritance, sleep time less than5hours and abnormal changes of electrocardiogram.6Consistency rate of prediction of hypertension using BP neural network model and Logistic regression model was90.27%and84.65%respectively;88.29%and84.40%for diabetes;92.70%and91.6%for heart disease.7Getting enough sleep and good exercise are associated with the prevention of hypertension, diabetes and heart disease. There is an interference effect of smoking on the occurrence of diabetes and heart disease. Limitation to drinking is important to prevent heart disease, and the effect of sport on hypertension and diabetes is more greater than that on heart disease.8The blood aluminium levels among patients with hypertension were higher than that of control group (median:1.00ug/ml vs0.75ug/ml), but levels of calcium and magnesium were lower than that of control group (average:59.5ug/ml vs57.34ug/ml and40.22ug/ml vs43.77ug/ml, respectively). Meanwhile the results showed that the blood aluminium levels (median:2.68ug/ml vs0.69ug/ml) and copper levels (average:0.99ug/ml vs0.94ug/ml) among patients with diabetes were higher than that of control group, but vitriol levels (median:8.26ug/L vs17.03ug/L) and selenium levels (average:103.40ug/L vs111.56ug/L) were lower than that of control group; the Zn/Cu value of control group was significantly higher than that of the case group (6.32vs5.89). There were no significant differences between the case group and control group for other detected elements.9The result of elements cluster analysis in hypertension case group and control group showed no obvious interaction between aluminium, calcium and other elements when10was assigned as the boundary. The relative coefficient of zinc and rubidium was affected by the blood ferrum, chromium levels among patients with hypertension. Chromium and selenium may play a role in the development of hypertension together with copper and magnesium.There was no obvious interaction between aluminium, vitriol and other elements in diabetes case group and control group. The correlation between zinc and rubidium was changed into the correlation among selenium, magnesium and rubidium. Meanwhile, there was a marked reduction in correlation coefficient (0.61vs0.34) between selenium and chromium.Conclusions:1The results of retrospective study and Cross-section investigation suggest that there is a trend of steady prevalence of occupational disease and high incidence of chronic disease among oilfield employees in the investigation. The primary threat to health is lifestyle-related disease, including overweight, abnormal blood fat, fatty liver, hypertension, heart disease and diabetes ranked in descending order. As for The primary forms of abnormal blood lipids levels, mixed type and low-risk type reach to the majority. All abnormal indexes are characterized by the higher incidence anmog male employees than the female, and the trend of prevalence always increases with age.2As independent risk factors of lifestyle-related disease, both sex and age interact with life style. The healthy lifestyle was more important to patients who report a family history of hereditary. The synergistic effect between inheritance and the unhealthy lifestyle may be more harmful, leading to1.32-fold increased risk of illness.3Most of the diseases suffer from varying degrees of different unhealthy lifestyles. The change of some lifestyles plays different roles in the prevention and outcome of disease. It’s more favorable to consolidated information system for health administration consisting of lifestyle, individual attribute (the societal attribute and the natural attribute) and routine indexes of physical checkup.4When applied in the prediction of hypertension, heart disease and diabetes, the accuracy of back-propagation neural network model is better than Logistic regression model. By using the combination of the two models, associated independent risk factors and the meaningful index weight on prediction can be obtained.5Getting enough sleep and good exercise are associated with the prevention of hypertension, diabetes and heart disease. There is an interference effect of smoking on the occurrence of diabetes and heart disease. Limitation to drinking is important to prevent heart disease, and the effect of sport on hypertension and diabetes is more greater than on heart disease.6Among patients with hypertension, the blood aluminium levels were higher, but calcium levels and magnesium levels were lower than that of control group. There was an interaction among magnesium, chromium, zinc, rubidium, ferrum during the process of the hypertension. Among patients with diabetes, the levels of aluminium and copper were higher, while vanadium and selenium levels were lower than that of control group. There was a change in the relations between zinc and rubidium, zinc and selenium, magnesium and rubidium, respectively. The correlation between selenium and chromium markedly decreased.Innovation:1Combined with the results of Logistic regression, independent factors were endowed with weight by the help of Bp neural network model, and were layered pre-warning was made after mimicking normal distribution.2The regulation extent of changeable behaviors in the disease prevention and control was expressed for arithmetic figure by using least square method. 3Twelve kinds of elements in the whole blood of oil workers were analysed by ICP-MS and the interaction among them were studied, providing both the fundamental data for the study on signal transduction pathway and key enzymes, and the foundation for the nutrition intervention on the chronic diseases.更多还原