Non-alcoholic fatty liver disease (NAFLD) is the most common chronic liver disease in the world, and the prevalence of NAFLD in China has continued to increase over the past 20 years. Some cohort studies have confirmed the causal relationship between NAFLD and dysglycaemia, but the temporal relationship between the two remains unclear.

To analyse the bidirectional time-series association between NAFLD and dysglycemia using cross-lagged panel models with the Beijing Health Management Cohort as the study population.

Based on the Beijing Health Management Cohort Study, follow-up data were collected from 2016 to 2021, with one follow-up visit per year, including questionnaire and physical examination information. According to the pre-established inclusion and exclusion criteria, 44 838 study subjects were finally included. A cross-lagged panel model was constructed using hepatic steatosis (HS) and fasting plasma glucose (FPG) as surrogate indicators of NAFLD and dysglycaemia, and stratified by gender and BMI to explore the temporal relationship between NAFLD and dysglycaemia in different populations.

(1) In the total population, the cross-lagged path coefficient β_{baseline HS→follow-up FPG} was statistically significant (P<0.05) at 0.009 (95%CI=0.002-0.016) ; however, the cross-lagged path coefficient in the opposite direction was not statistically significant (P>0.05). (2) In women, the cross-lagged path coefficients in both directions were statistically significant (P<0.05) with β_{baseline HS→follow-up FPG} of 0.025 (95%CI=0.015-0.035), and β_{baseline FPG→follow-up HS} of 0.026 (95%CI=0.014-0.038) ; in men, the cross-lagged path coefficients in both directions were not statistically significant (P>0.05). (3) Among those with BMI≥25.0 kg/m², the cross-lagged path coefficient β_{baseline HS→follow-up FPG} was statistically significant (P<0.05) at 0.114 (95%CI=0.103-0.125) ; however, the cross-lagged path coefficients in the opposite direction were not statistically significant (P>0.05) ; and among those with BMI <25.0 kg/m², the cross-lagged path coefficients in both directions were statistically significant (P>0.05). Cross-lagged path coefficients were statistically significant (P<0.05), with β_{baseline HS→follow-up FPG} of 0.101 (95%CI=0.092-0.111) and β_{baseline FPG→follow-up HS} of 0.021 (95%CI=0.012-0.031) .

There was a unidirectional temporal relationship from NAFLD to dysglycaemia in the total population and in the population with BMI≥25.0 kg/m²; and there was a bidirectional temporal relationship between NAFLD and dysglycaemia in the female population and in the population with BMI<25.0 kg/m², which provides clues for adopting the treatment idea of co-prevention of the two disorders, NAFLD and dysglycaemia, in clinical practice.

非酒精性脂肪肝病（NAFLD）是全球最常见的慢性肝病，我国NAFLD患病率在20年间持续增长。目前有队列研究已证实NAFLD和血糖异常之间存在因果关系，但二者的时序关系仍不明确。

以北京市健康管理队列为研究人群，运用交叉滞后面板模型分析NAFLD和血糖异常之间的双向时序关联。

基于北京市健康管理队列研究，收集2016—2021年的随访资料，每年随访1次，包括问卷与体格检查信息。根据纳入排除标准，最终纳入44 838名研究对象。以肝脂肪变性指数（HS）和空腹血糖（FPG）为NAFLD与血糖异常的替代指标构建交叉滞后面板模型，并按照性别与BMI进行分层，探索不同人群中NAFLD与血糖异常的时序关系。

（1）总人群中，交叉滞后路径系数β_{基线HS→随访FPG}为0.009（95%CI=0.002~0.016），有统计学意义（P<0.05）；但反方向交叉滞后路径系数无统计学意义（P>0.05）。（2）女性中双向的交叉滞后路径系数均有统计学意义（P<0.05），β_{基线HS→随访FPG}为0.025（95%CI=0.015~0.035），β_{基线FPG→随访HS}为0.026（95%CI=0.014~0.038）；而男性中两个方向的交叉滞后路径系数均无统计学意义（P>0.05）。（3）在BMI≥25.0 kg/m²的人群中，交叉滞后路径系数β_{基线HS→随访FPG}为0.114（95%CI=0.103~0.125），有统计学意义（P<0.05）；但反方向交叉滞后路径系数无统计学意义（P>0.05）；在BMI<25.0 kg/m²的人群中，双向的交叉滞后路径系数均有统计学意义（P<0.05），β_{基线HS→随访FPG}为0.101（95%CI=0.092~0.111），β_{基线FPG→随访HS}为0.021（95%CI=0.012~0.031），均有统计学意义（P<0.05）。

总人群与BMI≥25.0 kg/m²人群中存在从NAFLD到血糖异常的单向时序关系；女性人群与BMI<25.0 kg/m²人群中存在NAFLD与血糖异常的双向时序关系，本研究为临床上对NAFLD和血糖异常这两种疾病采取共同防治的治疗思路提供了线索。