Construction And Validation Of A Risk Prediction Model For Non-alcoholic Fatty Liver Disease Among The Non-obese Population In The Community

ObjectiveThis study aimed to explore the influencing factors of 2-year risk of non-alcoholic fatty liver disease(NAFLD)among the non-obese population in the community,with the intention of providing evidence for community nurses to implement health education and interventions against the risk factors of NAFLD.In addition,an easy-to-calculate nomogram model for predicting the risk of NAFLD was developed and validated to provide support for community nurses to identify NAFLD risk groups among the non-obese population in the community,and also for the early prevention and control of NAFLD.Methods(1)Meta-analysisThe 8 electronic databases including CNKI,Wan Fang,VIP,CBM,Pub Med,Embase,Web of Science and Cochrane Library were searched,and the Endnote citation management software was applied to conduct the literature screening based on the specific criteria for inclusion and exclusion.The Rev Man5.3 software was used to perform the meta-analysis and extract the data about the influencing or related factors of NAFLD in the non-obese population,so as to provide the essential information for the data collection of this study.(2)Construction and validation of a risk prediction model for NAFLD in the non-obese populationBased on the baseline health examination data of permanent residents were collected by our group members through the convenience sampling method in a certain community of the Pearl River Delta region of China in 2019,1738 subjects who met the inclusion and exclusion criteria were recruited in this study.Those who did not have NAFLD at the beginning were followed up in 2021 to examine the occurrence of NAFLD over a period of 2 years.The construction of the database was carried out using Epi Data version 3.1 software,while data analyses were conducted in SPSS version 23.0 and R version 4.1.3 software.Continuous variables were reported as the mean ± standard deviation for normal distributions or the median(interquartile range)for non-normal distributions.Categorical variables were expressed as frequencies(percentages).For between-group comparisons,the Student’s t-test(normal distribution)or the non-parametric test(non-normal distribution)was employed for continuous data,and the chi-square test or Fisher’s exact test was used for comparing categorical data.The subjects were separated into a modeling group and a verification group using a 7:3 ratio in a random manner.Then,univariate analysis,lasso regression analysis and logistic regression analysis were conducted in the modeling group to screen independent risk factors used to construct nomogram model.The C-index and the receiver operating characteristic(ROC)curve were used to assess the discrimination while the Hosmer-Lemeshow goodness-of-fit test,as well as the calibration curve,were used to evaluate the calibration ability of the model.Additionally,decision curve analysis was used to evaluate the clinical applicability of the nomogram.The statistical significance level was set at α=0.05(two-tailed).ResultsIn the Meta-analysis,14 studies were selected,which identified several significant factors associated with NAFLD in non-obese individuals.These factors were age,diabetes,body mass index(BMI),diastolic blood pressure(DBP),fasting plasma glucose(FPG),alanine aminotransferase(ALT),triglyceride(TG),low-density lipoprotein cholesterol(LDL-C),high-density lipoprotein cholesterol(HDL-C).A total of 1738 community residents were included in the study at the baseline,and 1499 subjects completed the 2-year follow-up while 239(13.8%)were lost to follow-up and were excluded in the subsequent analysis.There were no statistical differences between complete follow up or subjects lost to follow up in baseline data except for age(P<0.001)and marital status(P<0.001),therefore,the bias caused by the lost to follow up was small.Of the 1499 subjects included in the analysis,236(15.7%)had a new-onset NAFLD,and there were no statistical differences between the modeling group(1049)and the verification group(450)in baseline data and 2-year incidence of NAFLD.Eventually,logistic regression analysis identified 7 independent factors(female,regular exercise,diabetes,waist circumference,ALT,TG,HDL-C)to construct the nomogram prediction model.For model validation,the C-index was 0.771(95% CI: 0.735 to 0.807)and 0.756(95% CI: 0.695 to 0.817)in the modeling and verification groups,suggesting that the prediction model developed in the present study exhibited a certain level of discrimination.The P-values obtained from the Hosmer-Lemeshow goodness-of-fit test were all above 0.05(modeling group: P=0.230;verification group: P=0.200),and the calibration curves were plotted based on the test results,indicating that the prediction model had a good agreement between the assessed grade and the actual grade.In addition,the decision curve analysis showed that the higher positive net benefits were obtained in the threshold probability range of 2%-52% in the modeling group and 6%-56% in the validation group.ConclusionsIn this study,7 significant independent influencing factors used for constructing risk predictive model were female,regular exercise,diabetes,waist circumference,ALT,TG and HDL-C.Among them,regular exercise and HDL-C were protective factors for NAFLD in the non-obese population.Community nurses can carry out health education and intervention measures on changeable risk factors.The nomogram model constructed in this study has a certain prediction ability,and can provide support for community nurses to identify NAFLD risk groups among the non-obese population.