非酒精性脂肪肝筛查模型与风险评估模型研究

发布时间：2018-08-11 18:52

【摘要】：非酒精性脂肪肝(NAFLD)是指排除酒精和其他明确的损肝因素所致的肝细胞内脂肪过度沉积为主要特征的临床病理综合征,是与胰岛素抵抗和遗传易感性密切相关的获得性代谢应激性肝损伤。近年来,随着生活习惯、饮食结构的改变,NAFLD发病率和患病率不断上升;并且由于其危害不仅仅累及肝脏组织,而且还是心血管疾病、代谢综合征、慢性肾病等严重慢性病的危险因素,已引起了研究者的广泛关注。其中,对NAFLD的早筛查和早诊断尤为重要。为此,本研究依托"山东多中心健康管理纵向观察队列",分别利用横断面资料和队列资料,采用Logistic回归和Cox比例风险模型构建了 NAFLD的筛查模型和3年风险评估模型,以用于人群的NAFLD筛查及3年风险评估。研究结果如下:一、NAFLD筛查模型的研究1.共有18620名符合纳入标准者进入了筛查模型的研究。其中,男性10030名,有3811名被检出NAFLD,检出率为38.00%;女性8590名,有1898名被检出NAFLD,检出率为22.10%。男性NAFLD检出率高于女性(χ2=550.27,P0.001)。男、女性NAFLD检出率均随年龄增长而增加,均呈现上升趋势。2.在NAFLD患者和非NAFLD患者比较中,男性,除年龄外,两组间各项指标的差异均具有统计学意义(P0.05);而女性,两组间所有指标的差异都具有统计学意义(P0.05)。3.采用多因素Logistic回归(后退法)进行变量筛选及建模。最终,无论男女,进入模型的9个指标完全相同,分别是年龄、体重指数、舒张压、谷丙转氨酶、谷氨酰转肽酶、空腹血糖、甘油三酯、高密度脂蛋白和低密度脂蛋白。其中,除高密度脂蛋白是NAFLD的保护因素外,其它指标均为NAFLD的危险因素。4.对所构建的NAFLD筛查模型进行评价,男女性ROC曲线下面积AUC(95%CI)分别为0.800(0.792,0.807)和0.844(0.836,0.852)。经十折交叉验证对筛查模型进行内部评价,男女性AUC(95%CI)分别为0.798(0.790,0.807)和 0.843(0.833,0.852)。男女性外部验证的 AUC(95%CI)分别为0.845(0.832,0.858)和 0.868(0.854,0.882)。二、NAFLD的3年风险评估模型的研究1.NAFLD风险评估模型队列共有3429人,男性1847名,有683名在随访期内(4514人年)发生NAFLD,发病密度为15.13/100人年,女性1582名,有431名在随访期内(4142人年)发生NAFLD,发病密度为10.14/100人年。2.对随访期内发生NAFLD者与未发生NAFLD者进行比较,男性两组间各项指标的差异均具有统计学意义(P0.05);女性,两组间除单核细胞绝对值的差异无统计学意义外,其它指标的差异均具有统计学意义(P0.05)。3.采用多因素Cox回归(后退法)构建3年NAFLD风险评估模型,最终,男性有体重指数、舒张压、谷丙转氨酶、甘油三酯、高密度脂蛋白和低密度脂蛋白等6个指标进入模型。女性指标有年龄、体重指数、舒张压、甘油三酯、高密度脂蛋白和低密度脂蛋白等6个指标进入模型。4.对所构建的3年风险评估模型进行评价,男女性ROC曲线下面积AUC(95%CI)分别为 0.724(0.703,0.744)和 0.773(0.751,0.793)。经十折交叉验证对风险评估模型进行内部评价,男女性AUC(95%CI)分别为0.718(0.695,0.742)和 0.766(0.740,0.791)。男女性外部验证的 AUC(95%CI)分别为 0.715(0.665,0.765)和 0.712(0.654,0.770)。研究结论:1.男性的NAFLD检出率高于女性。2.本研究所构建的男、女性NAFLD筛查模型均能较好地鉴别出NAFLD患者及非患者,且具有一定外推性。3.本研究所构建的男、女性3年NAFLD风险评估模型,预测效果良好,具有一定的外推性。
[Abstract]:Nonalcoholic fatty liver disease (NAFLD) is a clinicopathological syndrome characterized by excessive intracellular fat deposition excluding alcohol and other specific liver damage factors. It is an acquired metabolic stress liver injury closely related to insulin resistance and genetic susceptibility. The morbidity and morbidity of NAFLD are increasing constantly, and the risk factors of serious chronic diseases, such as cardiovascular disease, metabolic syndrome, chronic kidney disease and so on, have attracted wide attention of researchers. Among them, early screening and early diagnosis of NAFLD are particularly important. Using cross-sectional data and cohort data, a three-year risk assessment model and a logistic regression model for NAFLD screening and a three-year risk assessment model for NAFLD screening were constructed using Cox proportional hazard model. Among them, 1 030 males and 3 811 females were detected with NAFLD, the detection rate was 38.00%; 8 590 females and 1 898 females were detected with NAFLD, the detection rate was 22.10%. The detection rate of NAFLD in males was higher than that in females (2 = 550.27, P 0.001). The detection rate of NAFLD in males and females increased with age, showing an upward trend. In NAFLD patients and non-NAFLD patients, except age, the differences of all the indicators between the two groups were statistically significant (P 0.05); while in women, the differences of all the indicators between the two groups were statistically significant (P 0.05). 3. Multivariate logistic regression (regression) was used to screen variables and model. Finally, both men and women entered the model. Nine indexes were identical, including age, body mass index, diastolic blood pressure, glutamic-alanine aminotransferase, glutamyl transpeptidase, fasting blood glucose, triglyceride, high density lipoprotein and low density lipoprotein. The ROC curve area under AUC (95% CI) was 0.800 (0.792,0.807) and 0.844 (0.836,0.852) for men and women, respectively. The internal evaluation of the screening model was carried out by 10-fold cross-validation. AUC (95% CI) for men and women were 0.798 (0.790,0.807) and 0.843 (0.833,0.852). AUC (95% CI) for men and women were 0.845 (0.832,0.858) and 0.854 (0.854, respectively). NAFLD risk assessment model cohort of 3 429 people, 1 847 men, 683 in the follow-up period (4 514 person-years) NAFLD incidence density of 15.13/100 person-years, 1582 women, 431 in the follow-up period (4 142 person-years) NAFLD incidence density of 10.14/100 person-years. Comparing the patients with NAFLD with those without NAFLD, the difference of each index between male and female groups was statistically significant (P 0.05); the difference of other indexes between the two groups except the absolute value of monocytes was statistically significant (P 0.05). 3. Finally, men entered the model with six indicators, including body mass index, diastolic blood pressure, alanine aminotransferase, triglyceride, high-density lipoprotein and low-density lipoprotein. Women's indicators included age, body mass index, diastolic blood pressure, triglyceride, high-density lipoprotein and low-density lipoprotein. AUC (95% CI) was 0.724 (0.703, 0.744) and 0.773 (0.751, 0.793) for men and women, respectively. The risk assessment model was evaluated internally by 10-fold cross-validation. AUC (95% CI) for men and women were 0.718 (0.695, 0.742) and 0.766 (0.740, 0.791), respectively. 765) and 0.712 (0.654, 0.770). Conclusion: 1. The detection rate of NAFLD in males is higher than that in females. 2. The NAFLD screening models of males and females can distinguish NAFLD patients and non-patients well, and have extrapolation. 3. The three-year NAFLD risk assessment model of males and females constructed in this study has good predictive effect and has certain extrapolation. Push sex.
【学位授予单位】：山东大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：R575.5

【相似文献】