目的:探讨丙氨酸氨基转移酶(ALT)水平对原发性肝癌(以下简称肝癌)发病风险的预测价值。方法:采用回顾性队列研究方法。收集2006年7月至2007年12月在华北理工大学附属开滦总医院(26 980例)、开滦林西医院(18 306例)、开滦赵各庄医院(14 469例)、开滦唐家庄医院(4 389例)、开滦范各庄医院(4 561例)、开滦荆各庄医院(7 623例)、开滦吕家坨医院(4 249例)、开滦林南仓医院(3 136例)、开滦钱家营医院(6 356例)、开滦马家沟医院(2 319例)、开滦医院分院(3 392例)行健康体检的95 780例受试者的体检资料，其中男76 491例，女19 289例；年龄为(51±11)岁，年龄范围为18～98岁。依据ALT三分位水平将受试者分为3组：T1组31 859例(ALT<14.7 U/L)，T2组30 627例(14.7 U/L≤ALT<22.0 U/L)，T3组33 294例(ALT≥22.0 U/L)。由固定医师团队于2006、2008、2010、2012、2014年在相同地点按相同健康体检顺序对受试者进行体检，收集流行病学调查内容、人体测量学及实验室检查指标。观察指标：(1)3组受试者的临床特征比较。(2)受试者随访和肝癌的发病情况。(3)影响受试者新发肝癌的危险因素分析。(4)ALT对肝癌模型预测价值的比较。采用健康体检的方式进行随访，随访内容为肝癌新发病情况和患者生存情况。以2006年首次健康体检时间作为随访起始时间，随访终止事件为发生肝癌、失访、死亡。随访时间截至2018年12月31日。正态分布的计量资料以n ±n s表示，多组间比较采用单因素方差分析。偏态分布的计量资料以n M(范围)表示，多组间比较采用Kruskal-Wallis秩和检验。计数资料以绝对数和(或)百分比表示，组间比较采用n χ2检验。采用Kaplan-Meier法计算累积发病率，累积发病率的组间比较采用Log-rank进行检验。使用限制性立方样条曲线(RCS)计算连续变化的ALT和肝癌发病风险的剂量反应关系。而后采用COX比例风险模型分析不同ALT水平分组影响肝癌发病的风险比(n HR)和95%可信区间。使用似然比检验和赤池信息量准则(AIC)计算ALT对于肝癌预测模型拟合情况的影响。使用C-统计量计算不同模型的预测能力。n 结果:(1)3组受试者的临床特征比较：T1组受试者男性、年龄、收缩压、舒张压、腰围、体质量指数、总胆固醇、甘油三酯、空腹血糖、高密度脂蛋白胆固醇、低密度脂蛋白胆固醇、乙型肝炎表面抗原(HBsAg)阳性、高血压病、糖尿病、肝硬化、脂肪肝、恶性肿瘤家族史、吸烟、饮酒、体育锻炼分别为23 516例、(53±14)岁、(130±21)mmHg(1 mmHg＝0.133 kPa)、(82±12)mmHg、(85±10)cm、(24±3)kg/mn 2、(4.9±1.1)mmol/L、1.11mmol/L(0.79～1.63 mmol/L)、(5.4±1.6)mmol/L、(1.5±0.4)mmol/L、(2.3±1.0)mmol/L、478例、12 758例、2 482例、15例、5 899例、1 071例、9 553例、5 434例、5 210例；T2组受试者上述指标分别为24 886例、(52±12)岁、(132±21)mmHg、(84±12)mmHg、(87±10)cm、(25±3)kg/mn 2、(4.9±1.2)mmol/L、1.27 mmol/L(0.91～1.90 mmol/L)、(5.5±1.7)mmol/L、(1.6±0.4)mmol/L、(2.4±0.9)mmol/L、744例、13 884例、2 677例、13例、9 365例、981例、8 935例、5 216例、5 023例；T3组受试者上述指标分别为28 089例、(50±12)岁、(132±21)mmHg、(85±12)mmHg、(89±10)cm、(26±3)kg/mn 2、(5.0±1.2)mmol/L、1.45 mmol/L(1.02～2.28 mmol/L)、(5.6±1.8)mmol/L、(1.6±0.4)mmol/L、(2.3±0.9)mmol/L、1 416例、15 471例、3 583例、70例、15 277例、1 350例、10 514例、6 184例、4 447例；3组受试者上述指标比较，差异均有统计学意义(n χ2＝1 181.88，n F＝350.52、148.39、417.75、948.63、1 917.20、111.60, n χ2＝3 747.25，n F＝133.78、18.17、14.45, n χ2＝478.33、306.60、181.38、58.19、5 649.18、38.82、45.69、35.11、152.64，n P<0.05)。(2)受试者随访和肝癌的发病情况：95 780例受试者随访时间为9.05年(8.75～9.22年)，共有新发肝癌351例，肝癌总发病密度为3.39/10 000人年，女性肝癌发病率为1.11/10 000人年，男性肝癌发病率为3.97/10 000人年。T1组、T2组、T3组肝癌的发病密度分别为2.09/10 000人年、2.54/10 000人年和5.42/10 000人年。T1组、T2组、T3组受试者的肝癌累积发病率分别为2.47‰、3.10‰和6.11‰，3组比较，差异有统计学意义(n χ2＝63.50，n P<0.05)。(3)影响受试者新发肝癌的危险因素分析。COX比例风险模型分析结果显示：校正性别、年龄、体质量指数、总胆固醇、空腹血糖、HBsAg阳性、高血压病、糖尿病、肝硬化、脂肪肝、恶性肿瘤家族史、吸烟、饮酒、体育锻炼后，与T1组比较，T3组新发肝癌的风险增加(n HR＝1.87，95%可信区间为1.40～2.49，n P<0.05)。RCS结果显示：ALT与肝癌的发病风险呈非线性相关，且曲线呈上升趋势(RCS_S1_n χ2＝7.21，n P<0.05)。(4)ALT对肝癌模型预测价值的比较：建立多因素模型，将性别、年龄、体质量指数、总胆固醇、空腹血糖、HBsAg阳性、高血压病、糖尿病、肝硬化、脂肪肝、恶性肿瘤家族史、吸烟、饮酒、体育锻炼因素共同带入模型中，计算此模型的C统计量、-2Log L值和AIC值，分别为0.791、6 313.30和6 345.30。将ALT三分位数分组带入模型中，以T1组为对照组计算此时模型的C统计量、-2Log L值和AIC值，分别为0.792、6 300.48和6 328.48。多因素＋Fbg模型的C统计量和AIC数值降低，与多因素模型的-2Log L值的差值为12.82，差异有统计学意义(n χ2＝12.82，n P<0.05)。n 结论:ALT是肝癌的独立危险因素。在建立肝癌预测模型或评估肝癌发病风险时，ALT可作为重要的参考指标。“,”Objective:To explore the prognostic value of serum alanine aminotransferase (ALT) on morbidity of primary liver cancer (hereinafter referred to as liver cancer).Methods:The retrospective cohort study was conducted. The data of 95 780 participants including 26 980 cases in the Kailuan General Hospital of North China University of Science and Technology, 18 306 cases in Kailuan Linxi Hospital, 14 469 cases in Kailuan Zhaogezhuang Hospital, 4 389 cases in Kailuan Tangjiazhuang Hospital, 4 561 cases in Kailuan Fan′gezhuang Hospital, 7 623 cases in Kailuan Jinggezhuang Hospital, 4 249 cases in Kailuan Lyujiatuo Hospital, 3 136 cases in Kailuan Linnancang Hospital, 6 356 cases in Kailuan Qianjiaying Hospital, 2 319 cases in Kailuan Majiagou Hospital and 3 392 cases in Kailuan Branch Hospital who participated health examination from July 2006 to December 2007 were collected. There were 76 491 males and 19 289 females, aged (51±11)years, with a range of 18-98 years. All the subjects were allocated into 3 groups according to tertiles of ALT, including 31 859 with ALT<14.7 U/L in the T1 group, 30 627 with 14.7 U/L≤ALT<22.0 U/L in the T2 group and 33 294 with ALT≥22.0 U/L in the T3 group. All participants received the same-order health examinations by the fixed team of doctors in 2006, 2008, 2010, 2012 and 2014 at the same place. Epidemiological investigation, anthropometric parameters and biochemical indicators were collected. Observation indicators: (1) comparisons of clinical characteristics among the 3 groups; (2) follow-up and morbidity of liver cancer; (3) risk factors analysis affecting new-onset liver cancer; (4) comparisons of the prognostic value of ALT on liver cancer model. Follow-up using physical examination was performed to detect new-onset liver cancer and survival up to December 31, 2015. The start time of follow-up was the first health examination in 2006 and the terminal event was new-onset liver cancer, loss of follow-up or death. Measurement data with normal distribution were expressed asn Mean±n SD, and comparisons among groups were analyzed using the one-way ANOVA. Measurement data with skewed distribution were described as n M (range), and comparisons among groups were analyzed using the Kruskal-Wallis rank sum test. Count data were described as absolute numbers and (or) percentages, and comparisons among groups were analyzed using the chi-square test. The cumulative incidence of new-onset liver cancer was calculated by the Kaplan-Meier method, and comparisons of incidences among groups were done by the Log-rank test. Restrictive cubic spline regression was used to calculate the dose-response relation between the continuous ALT and the risks of new-onset liver cancer. The hazard ratio (n HR) and 95% confidence interval (n CI) of different levels of ALT on new-onset liver cancer were estimated by the COX proportional hazards regression models. The fitting degree of ALT on new-onset liver cancer model was calculated by the likelihood ratio test and akaike information criterion (AIC). The predictive power of different models was calculated using the C-statistics.n Results:(1) Comparisons of clinical characteristics among the 3 groups: gender(male), age, systolic pressure, diastolic pressure, waistline, body mass index (BMI), total cholesterol (TC), triglyceride (TG), fasting blood glucose, high density lipoprotein cholesterol, low density lipoprotein, cases with positive Hepatitis B surface antigen (HBsAg), cases with hypertension, cases with diabetes, cases with cirrhosis, cases with fatty liver, cases with malignant tumor in immediate family, cases with smoking, cases with drinking, cases with physical exercise were 23 516, (53±14)years, (130±21)mmHg (1 mmHg=0.133 kPa), (82±12)mmHg, (85±10)cm, (24±3)kg/mn 2, (4.9±1.1)mmol/L, 1.11 mmol/L (range, 0.79-1.63 mmol/L), (5.4±1.6)mmol/L, (1.5±0.4)mmol/L, (2.3±1.0)mmol/L, 478, 12 758, 2 482, 15, 5 899, 1 071, 9 553, 5 434, 5 210 in the T1 group, 24 886, (52±12) years, (132±21)mmHg, (84±12)mmHg, (87±10)cm, (25±3)kg/mn 2, (4.9±1.2)mmol/L, 1.27 mmol/L(range, 0.91-1.90 mmol/L), (5.5±1.7)mmol/L, (1.6± 0.4)mmol/L, (2.4±0.9)mmol/L, 744, 13 884, 2 677, 13, 9 365, 981, 8 935, 5 216, 5 023 in the T2 group and 28 089, (50±12)years, (132±21)mmHg, (85±12)mmHg, (89±10)cm, (26±3)kg/mn 2, (5.0±1.2)mmol/L, 1.45 mmol/L (range, 1.02-2.28 mmol/L), (5.6±1.8)mmol/L, (1.6±0.4)mmol/L, (2.3±0.9)mmol/L, 1 416, 15 471, 3 583, 70, 15 277, 1 350, 10 514, 6 184, 4 447 in the T3 group, showing significant differences among the three groups (n χ2=1 181.88, n F=350.52, 148.39, 417.75, 948.63, 1 917.20, 111.60, n χ2=3 747.25, n F=133.78, 18.17, 14.45, n χ2=478.33, 306.60, 181.38, 58.19, 5 649.18, 38.82, 45.69, 35.11, 152.64, n P<0.05). (2) Follow-up and morbidity of liver cancer: all 95 780 participants were followed up for 9.05 years (range, 8.75-9.22 years). During the follow-up, there were 351 cases diagnosed with liver cancer. The total incidence density, incidence density in the female and male participants of liver cancer were 3.39/10 000 person-year, 1.11/10 000 person-year and 3.97/10 000 person-year, respectively. The incidence density of liver cancer was 2.09/10 000 person-year, 2.54/10 000 person-year, 5.42/10 000 person-year in the T1, T2, T3 groups, respectively. The cumulative incidence was 2.47‰, 3.10‰, 6.11‰ in the T1, T2, T3 groups, respectively, showing significant differences among the three groups (n χ2=63.50, n P<0.05). (3) Risk factors analysis affecting new-onset liver cancer. Results of COX proportional hazards regression model analysis showed that comparison with T1 group, the incidence density of liver cancer was increased in the T3 group (n HR=1.87, 95%n CI: 1.40-2.49, n P<0.05) after adjustment of gender, age, BMI, TC, fasting blood glucose, positive HBsAg, hypertension, diabetes, liver cirrhosis, fatty liver, malignant tumor in immediate family, smoking, drinking, physical exercise. Results of restrictive cubic spline regression showed that continous ALT was nonlinear correlated with incidence of liver cancer, but showed a trend of rising curve (RCS_S1_n χ2=7.21, n P<0.05). (4) Comparisions of the prognostic value of ALT on liver cancer model: a traditional multivariate model was established to incorporate factors such as gender, age, BMI, TC, fasting blood glucose, positive HBsAg, hypertension, diabetes, liver cirrhosis, fatty liver, malignant tumor in immediate family, smoking, drinking, physical exercise. The C statistic, -2Log L value and AIC value of this model were 0.791, 6 313.30 and 6 345.30 respectively. Then the ALT tertiles were grouped into the model to calculate the C statistic, -2Log L value and AIC value of the model at this time, which were 0.792, 6 300.48 and 6 328.48, respectively. The C statistic and AIC value of the multivariate + Fbg model decreased, and the difference between the -2Log L value of the multivariate model was 12.82, and the difference was statistically significant (n χ2=12.82, n P<0.05).n Conclusion:ALT is an independent risk factor for liver cancer and is an important reference indicator when establishing a predictive model of liver cancer or assessing the risk of liver cancer.