OBJECTIVE:Based on different algorithms of machine learning,the prediction model of lumbar disc herniation has become a trend and hot spot in the development of precision medicine.However,there is limited evidence on the reporting quality and methodological quality of prediction models of lumbar disc herniation outcomes using machine learning.This article is aimed to explore the performance of machine learning algorithms in predicting the prognosis of lumbar disc herniation by comprehensively analyzing the report quality and risk of bias of previous studies that developed and validated prognosis prediction models based on machine learning through a comprehensive literature search,in order to explore the performance of machine learning algorithms in predicting the prognosis of lumbar disc herniation.METHODS:The databases of CNKI,WanFang,VIP,SinOMED,PubMed,Web of Science,Embase,and The Cochrane Library were searched by computer.Studies on the use of machine learning to develop(and/or validate)prognostic prediction models for lumbar disc herniation were collected from the inception of the database to December 31,2023.Two researchers independently screened the literature,extracted data,and assessed the risk of bias of the included studies.The reporting quality and risk of bias of the included studies were assessed by the Multivariable Transparent Reporting of Predictive Models(TRIPOD)statement and the Predictive Model Risk of Bias Assessment Tool(PROBAST).The results of the evaluation were analyzed using descriptive statistics and visual charts.RESULTS:(1)A total of 23 articles were included,and the TRIPOD compliance of each study ranged from 11％to 87％,with a median compliance of 54％.The quality of reporting of titles,detailed descriptions of treatment measures,blinding of predictors,handling of missing data,details of risk stratification,specific procedures for enrollment,model interpretation,and model performance was mostly poor,with TRIPOD adherence rates ranging from 4％to 35％.(2)Of all included studies,61％had a high risk of bias and 39％had an unclear overall risk of bias.The area under the curve,accuracy,sensitivity and specificity were used to evaluate the performance of the model.The areas under the curve of 20 models were reported,ranging from 0.561 to 0.999.Three models reported the accuracy of the model,ranging from 82.07％to 89.65％.(3)Among all included studies,the statistical analysis domain was most often assessed as having a high risk of bias,mainly due to the small number of valid samples,the selection of predictors based on univariate analysis and the lack of calibration and discrimination assessment of the model in the study.CONCLUSION:These results indicate that machine learning can achieve good predictive ability in the development and validation of prognostic models for lumbar disc herniation.The commonly used algorithms include regression algorithm,support vector machine,decision tree,random forest,artificial neural network,naive Bayes and other algorithms.Reasonable algorithms combined with clinical practice can improve the accuracy of prognosis prediction of lumbar disc herniation.However,the reporting and methodological quality of prognosis prediction models based on machine learning are poor,the prediction performance of different models varies greatly,and the generalization and extrapolation of research models are unclear.There is an urgent need to improve the design,implementation and reporting of such studies.To promote the application of machine learning in the clinical practice of lumbar disc herniation prediction models,it is necessary to comprehensively consider various predictors related to the prognosis of the disease before modeling,and strictly follow the relevant standards of PROBAST tool during modeling.

OBJECTIVE:Based on different algorithms of machine learning,the prediction model of lumbar disc herniation has become a trend and hot spot in the development of precision medicine.However,there is limited evidence on the reporting quality and methodological quality of prediction models of lumbar disc herniation outcomes using machine learning.This article is aimed to explore the performance of machine learning algorithms in predicting the prognosis of lumbar disc herniation by comprehensively analyzing the report quality and risk of bias of previous studies that developed and validated prognosis prediction models based on machine learning through a comprehensive literature search,in order to explore the performance of machine learning algorithms in predicting the prognosis of lumbar disc herniation.METHODS:The databases of CNKI,WanFang,VIP,SinOMED,PubMed,Web of Science,Embase,and The Cochrane Library were searched by computer.Studies on the use of machine learning to develop(and/or validate)prognostic prediction models for lumbar disc herniation were collected from the inception of the database to December 31,2023.Two researchers independently screened the literature,extracted data,and assessed the risk of bias of the included studies.The reporting quality and risk of bias of the included studies were assessed by the Multivariable Transparent Reporting of Predictive Models(TRIPOD)statement and the Predictive Model Risk of Bias Assessment Tool(PROBAST).The results of the evaluation were analyzed using descriptive statistics and visual charts.RESULTS:(1)A total of 23 articles were included,and the TRIPOD compliance of each study ranged from 11％to 87％,with a median compliance of 54％.The quality of reporting of titles,detailed descriptions of treatment measures,blinding of predictors,handling of missing data,details of risk stratification,specific procedures for enrollment,model interpretation,and model performance was mostly poor,with TRIPOD adherence rates ranging from 4％to 35％.(2)Of all included studies,61％had a high risk of bias and 39％had an unclear overall risk of bias.The area under the curve,accuracy,sensitivity and specificity were used to evaluate the performance of the model.The areas under the curve of 20 models were reported,ranging from 0.561 to 0.999.Three models reported the accuracy of the model,ranging from 82.07％to 89.65％.(3)Among all included studies,the statistical analysis domain was most often assessed as having a high risk of bias,mainly due to the small number of valid samples,the selection of predictors based on univariate analysis and the lack of calibration and discrimination assessment of the model in the study.CONCLUSION:These results indicate that machine learning can achieve good predictive ability in the development and validation of prognostic models for lumbar disc herniation.The commonly used algorithms include regression algorithm,support vector machine,decision tree,random forest,artificial neural network,naive Bayes and other algorithms.Reasonable algorithms combined with clinical practice can improve the accuracy of prognosis prediction of lumbar disc herniation.However,the reporting and methodological quality of prognosis prediction models based on machine learning are poor,the prediction performance of different models varies greatly,and the generalization and extrapolation of research models are unclear.There is an urgent need to improve the design,implementation and reporting of such studies.To promote the application of machine learning in the clinical practice of lumbar disc herniation prediction models,it is necessary to comprehensively consider various predictors related to the prognosis of the disease before modeling,and strictly follow the relevant standards of PROBAST tool during modeling.

ObjectiveTo investigate the predictive factors for HBsAg clearance in chronic hepatitis B （CHB） patients with a low level of hepatitis B surface antigen （HBsAg） treated with pegylated interferon α-2b （PEG-IFN-α-2b）， to establish a combined predictive model and a nomogram based on multiple factors， and to provide a reference for formulating individualized treatment regimens and predicting treatment outcome in clinical practice. MethodsA retrospective analysis was performed for 167 CHB patients with HBsAg <1 500 IU/mL who attended The Third People’s Hospital of Kunming from January 2022 to January 2024 and were treated with PEG-IFN-α-2b. According to whether clinical cure was achieved， the patients were divided into HBsAg clearance group and HBsAg non-clearance group. Related data were collected， including general information and serological/biochemical/virological indicators at different time points during treatment. The independent samples t-test was used for comparison of normally distributed continuous data， and the Mann-Whitney U test was used for comparison of non-normally distributed continuous data； the chi-square test was used for comparison of categorical data. The multivariate logistic regression analysis was used to identify independent influencing factors. The receiver operating characteristic （ROC） curve was used to assess the value of indicators used alone or in combination in predicting clinical cure， and calibration curves were plotted to assess the risk prediction model. ResultsThe univariate analysis showed that there were significant differences between the two groups in age （t=-6.839， P<0.05）， history of nucleos（t）ide analogue treatment for over 1 year （χ²=59.339， P<0.05）， genotype （χ²=4.610， P<0.05）， nonalcoholic fatty liver disease （χ²=5.319， P<0.05）， hepatitis B virus DNA status before treatment （χ²=60.861， P <0.05）， compensated liver cirrhosis （χ²=10.960， P<0.05）， HBeAg status before treatment （χ²=19.060， P<0.05）， a history of interferon treatment （χ²=8.162， P<0.05）， presence of interferon antibodies after treatment （χ²=12.858， P<0.05）， HBsAg level before treatment （Z=-7.412， P<0.05）， alanine aminotransaminase （ALT） level at baseline （Z=-6.117， P<0.05）， ALT level at 12 weeks of treatment （Z=-7.171， P<0.05）， platelet count （PLT） at 24 weeks of treatment （Z=-3.622， P<0.05）， and thyroid stimulating hormone （TSH） level at 24 weeks of treatment （Z=-2.830， P<0.05）. The multivariate logistic regression analysis showed that age （odds ratio ［OR］=1.230， P=0.007）， history of nucleos（t）ide analogue treatment for over 1 year （OR=0.008， P=0.011）， HBeAg status before treatment （OR=0.003， P=0.012）， HBsAg level before treatment （OR=1.005， P=0.014）， ALT level at baseline （OR=0.949， P=0.014）， ALT level at 12 weeks of treatment （OR=0.969， P=0.016）， PLT at 24 weeks of treatment （OR=0.969， P=0.022）， and TSH level at 24 weeks of treatment （OR=3.608， P=0.045） were independent influencing factors for HBsAg clearance at 48 weeks of treatment in CHB patients with HBsAg <1 500 IU/mL. The Hosmer-Lemeshow goodness-of-fit test yielded χ²=1.398， P=0.994， indicating that the model had good fitting. The Bootstrap method was used to perform internal validation of the nomogram model， and there was a good degree of fitting between the calibration curve and the ideal curve， with a mean absolute error of 0.029. The ROC curve analysis showed that the combination of predictive factors had an area under the ROC curve of 0.982 （95% confidence interval： 0.961 — 0.999）， with a sensitivity of 94.10% and a specificity of 93.10%， suggesting that the nomogram model had a good discriminatory ability. For the CHB patients with HBsAg <1 500 IU/mL and different features， further analysis of HBsAg clearance rate at 48 weeks of treatment showed an HBsAg clearance rate of 69.60% for those with HBsAg ≤67.65 IU/mL before treatment， 58.30% for those with a baseline ALT level of ≥62.50 U/L， 68.30% for those with an ALT level of ≥92.50 U/L at 12 weeks of treatment， 42.40% for those with PLT ≥104×10⁹/L at 24 weeks of treatment， and 48.30% for those with a TSH level of ≤1.38 μIU/mL at 24 weeks of treatment， with significant differences between the two groups （all P<0.001）. ConclusionAge， history of nucleos（t）ide analogue treatment for over 1 year， HBeAg status before treatment， HBsAg level before treatment， baseline ALT level， ALT level at 12 weeks of treatment， PLT level at 24 weeks， and TSH level at 24 weeks of treatment are independent predictive factors. The combined prediction nomogram model constructed in this study has a relatively high value in predicting clinical cure at 48 weeks of PEG-IFN-α-2b treatment in CHB patients with HBsAg<1 500 IU/mL， thereby providing a reference for selecting suitable treatment population and predicting clinical cure.

Objective To investigate the expression and clinical diagnosis significance of serum long non-coding RNA PANDAR(lncRNA PANDAR)and procollagen type I C-terminal propeptide(PICP)in diabetic nephropathy(DN).Methods A total of 310 patients with type 2 diabetes mellitus(T2DM)admitted to the Third Affiliated Hospital of Chengdu Medical College from January 2022 to December 2023 were selected as the case group,and further divided into DM group(n=190)and DN group(n=120)based on the presence or absence of concomitant kidney disease.60 healthy subjects were selected as the control group.The expression level of lncRNA PANDAR was determined by real time fluorescence quantitative PCR,and the serum PICP level was determined by enzyme-linked immunosorbent assay(ELISA).The expression of lncRNA PANDAR and serum PICP in each group was compared.Logistic regression was used to analyze the influencing factors of DN.ROC curve was used to evaluate the diagnostic value of serum lncRNA PANDAR,PICP and their combination in DN.Results The levels of serum lncRNA PANDAR(4.56±1.26),PICP(147.72±19.20 μg/L)in the case group were higher than those in the control group(2.27±0.61,117.24±12.72 μg/L),and the differences were statistically significant(t=13.758,11.799,all P<0.05).With the increase of chronic kidney disease(CKD)stage,the levels of serum lncRNA PANDAR and PICP gradually increased,and the differences were statistically significant(F=16.790,286.87,all P<0.05).The levels of serum lncRNA PANDA(5.27±1.26)and PICP(159.59±17.10 μg/L)in DN group were higher than those in DM group(4.11±1.03,140.22±16.56 μg/L),the differences were statistically significant(t=-8.847,-9.905,all P<0.05).Multivariate Logistic regression analysis showed that lncRNA PANDA and PICP were independent risk factors for DN in T2DM patients(Wald χ2=11.592,7.880,all P<0.05).The areas under curve(AUC)of serum lncRNA PANDA and PICP combined detection of DN in patients with T2 DM was higher than that of single index detection,and the differences were statistically significant(Z=-2.006,-3.069,all P<0.05).Conclusion The levels of serum lncRNA PANDAR and PICP in patients with DN are increased,and their combined detection can improve the diagnostic efficiency of DN.

The codon was optimized for the bovine nebovirus(BNeV)VP1 gene and the recombi-nant plasmid pFastBac-Dual-VP1 was constructed,and BNeV-VP1 virus-like particles(VLPs)were prepared using a baculovirus expression system,and identified by Western blot,indirect im-munofluorescence and electron microscopy.Successfully validated VLPs were mixed with MF59 adjuvant and CpG-ODG,and mice were immunised by intramuscular injection and evaluated for immunity effects.The results showed that the optimized CAI(codon adaptation index)of VP1 gene was 0.93 and the GC content was 60.4％.The constructed recombinant plasmid was trans-formed into DH10Bac for blue-white spot screening,and after successful verification,it was trans-fected into SF9 cells to obtain recombinant baculovirus Baculo-BNeV-VP1.BNeV virus-like parti-cles with diameters ranging from 35 to 40 nm were observed under the electron microscope,and both IFA and Western blot experiments proved that the target proteins were successfully ex-pressed and biologically active,and protein optimisation revealed that the highest protein expres-sion was found at the infectious dose MOI=0.5.Mice were immunized by intramuscular injection after 50 μg of VLPs were mixed with MF59 adjuvant and CpG-ODN.The results showed that the VLPs immunization group produced IgG antibodies 7 days after the first dose,and the antibody ti-ter increased gradually,reaching a maximum of 1∶102 400,and declined at 35 d,but still main-tained a high level;The blocking titer BT50 is up to 640,which can induce the production of BNeV VP1-specific blocking antibody in mice.In this study,the baculovirus expression system was used to express the VP1 protein of BNeV,and BNeV VLPs were successfully constructed,which could induce humoral immune response in mice,which provided a reference for the follow-up research of BNeV vaccine.

Codon optimization was performed for the M and HN genes of bovine parainfluenza virus type 3(BPIV3),and the recombinant shuttle plasmid Dual-M+HN was constructed.BPIV3 VLPs was prepared using the baculovirus expression system,and verified by Western blot,IFA and elec-tron microscopy.The successfully verified virus-like particle(VLPs)were mixed with MF59 adjuvant and CpG-ODN immunoenhancer to immunize mice by intramuscular-injection,and BPIV3 inactivated vaccine group and adjuvant control group were set up.The immune effect of BPIV3 VLPs was evaluated by monitoring mouse serum specific antibodies,neutralizing antibodies and hemagglutination inhibition antibodies.The results showed that the optimized codon adaptation in-dex(CAI)of the M and HN protein genes were 0.96 and 0.95,respectively,and the CG content reached 54.1％and 53.1％,respectively.The constructed recombinant plasmid was transformed in-to DHI0Bac for blue and white spot screening.The validated recombinant rod particles were trans-fected into Sf9 cells to obtain the rod-shaped virus pFastBac-M+HN.Under electron microscopy,BPIV3 VLPs with a diameter of approximately 180 nm were observed.IFA and Western blot ex-periments confirmed the successful expression and biological activity of the target protein.Through protein optimization,it was found that the protein expression was highest at an infection dose of MOI=5.After mixing 50 μg VLPs with MF59 adjuvant and CpG-ODN,mice were immunized by intramuscular injection.The results showed that the antibodies in the VLPs immunized group be-gan to rise at 2 weeks of the first immunization and reached their peak at 21 days of the second im-munization,with an average IgG antibody titer of 1∶40 228;The average titer of neutralizing anti-bodies is 1∶298;The titer of hemagglutination inhibition antibody is 1∶549,reaching the level of inactivated vaccine(P≥0.05),indicating that the VLPs prepared in this experiment can induce hu-moral immune response in the body.In summary,this study successfully prepared VLPs capable of self-assembly expression of BPIV3 HN and M proteins,and induced humoral immune response in mice,providing research basis for subsequent BPIV3 VLPs vaccine research.

Objective To investigate the predictive factors for the occurrence of HBsAg clearance in patients with HBeAg-negative chronic hepatitis B(CHB)receiving peginterferon alfa-2b(PEG-IFN-α-2b)treatment,analyze the effects of various indicators on the HBsAg clearance rate under different characteristics,and construct and evaluate a combined predictive model.Methods We included 125 patients with HBeAg-negative CHB at Kunming Third People's Hospital from May 2021 to May 2023.After treatment with PEG-IFN-α-2b combined with nucleoside analogues for a course of 48 weeks,they were divided into HBsAg clearance group and HBsAg non-clearance group.Their general information and serological,biochemical,and virological indicators at different time points during treatment were recorded.Continuous data in normal distribution were compared using the t test.Continuous data in non-normal distribution were compared using the Mann-Whitney U test,and comparisons across different time points were performed using the multiple paired-sample Friedman test.Categorical data were compared using the χ2 test.A Logistic regression analysis was used to select variables to establish a combined multi-parameter predictive model.Receiver operating characteristic(ROC)curves were generated to evaluate the diagnostic value of individual indicators and the combined predictive model for HBsAg clearance.Results Before treatment,there were significant differences in baseline HBsAg level(Z=-3.997,P<0.05)and treatment history(χ2=8.221,P<0.05)between the two groups.During treatment,gradually decreasing trends were observed in white blood cell count(χ2=104.944),neutrophil count(χ2=132.036),platelet count(χ2=162.881),and thyroid-stimulating hormone level(TSH,χ2=83.304,all P<0.05),while alanine aminotransferase(ALT,χ2=157.618)and alpha fetoprotein(χ2=159.472)showed gradually increasing trends(both P<0.05).At 48 weeks of treatment,treatment history(odds ratio[OR]=0.232,95%confidence interval[CI]:0.071-0.753),baseline HBsAg level(OR=13.423,95%CI:3.276-54.997),the extent of decrease in HBsAg from baseline after 12 weeks of treatment(OR=0.143,95%CI:0.040-0.515),the maximum ALT level during treatment(OR=0.986,95%CI:0.980-0.993),and the minimum TSH level during treatment(OR=3.281,95%CI:1.413-7.619)were independent factors affecting HBsAg clearance(all P<0.05).A combined predictive model for HBsAg clearance was built:Y=-1.603-1.462×treatment history+2.597×baseline HBsAg value-1.944×the extent of HBsAg reduction from baseline after 12 weeks of treatment-0.014×the maximum ALT value during treatment+1.188×the minimum TSH value during treatment.The diagnostic value of the individual indicators for HBsAg clearance from high to low was as following:the maximum ALT value during treatment(AUC=0.824),baseline HBsAg value(AUC=0.727),the minimum TSH value during treatment(AUC=0.707),the extent of HBsAg reduction from baseline after 12 weeks of treatment(AUC=0.641),and treatment history(AUC=0.636).The combined model showed better predictive performance than the individual indicators,with the AUC being 0.921(all P<0.05).Conclusion The combined model,constructed with baseline HBsAg value,the extent of HBsAg reduction from baseline after 12 weeks of treatment,the maximum ALT value during treatment,and the minimum TSH value during treatment,has high predictive value for the occurrence of HBsAg clearance in patients with HBeAg-negative CHB after 48 weeks of treatment with PEG-IFN-α-2b,which can provide a reference for identifying suitable patients for treatment and predicting clinical outcome.

Objective To investigate the influencing factors for recompensation in patients with decompensated hepatitis B cirrhosis,and to establish a predictive model.Methods A total of 517 patients who attended The Third People's Hospital of Kunming and were diagnosed with decompensated hepatitis B cirrhosis from January 1,2016 to December 31,2022 were enrolled.The clinical data of the patients were reviewed,and the 207 patients with no portal hypertension-related complications within at least 1 year were enrolled as recompensation group,while the 310 patients without recompensation were enrolled as persistent decompensation group.Related clinical data were collected,and the univariate and multivariate Cox regression analyses were performed for the factors that might affect the occurrence of recompensation.The independent-samples t test was used for comparison of normally distributed continuous data between two groups,and the Mann-Whitney U test was used for comparison of non-normally distributed continuous data between two groups;the chi-square test or the Fisher's exact test was used for comparison of categorical data between two groups.The"rms"package was used to establish a nomogram;the receiver operating characteristic(ROC)curve was plotted,and the area under the ROC curve(AUC)was calculated;the Hosmer-Lemeshow test was used to evaluate the degree of fitting of the model;the"Calibration Curves"package was used to plot the calibration curve for model assessment.Results Among the patients with decompensated hepatitis B cirrhosis,207(40.03%)had recompensation.The univariate Cox regression analysis showed that there were significant differences between the recompensation group and the persistent decompensation group in TIPS history,genotyping,portal vein thrombosis,complicated infection,Child-Pugh class,age,hemoglobin(Hb),platelet count,total protein,albumin(Alb),alanine aminotransferase,triglyceride,cholesterol,creatinine,Na,interleukin-6,high-sensitivity C-reactive protein,portal vein width,and portal vein velocity(all P<0.05).The multivariate regression analysis showed that TIPS history(hazard ratio[HR]=2.491,95%confidence interval[CI]:1.325-4.681,P=0.005),portal vein thrombosis(HR=0.345,95%CI:0.152-0.783,P=0.001),Hb(HR=1.007,95%CI:1.000-1.013,P=0.028),Alb(HR=1.048,95%CI:1.017-1.080,P=0.002),and portal vein width(HR=0.899,95%CI:0.835-0.967,P=0.004)were independent influencing factors for recompensation in patients with decompensated hepatitis B cirrhosis.A nomogram model was established based on the above five influencing factors,and the Hosmers-Lemeshow test showed that this model had a good degree of fitting(χ2=3.202,P=0.921).The nomogram model had an AUC of 0.728,a sensitivity of 50.3%,and a specificity of 85.0%,and the calibration curve showed good consistency between the actual value of this model in predicting the occurrence of recompensation and the predicted value in patients with decompensated hepatitis B cirrhosis.Conclusion Patients with decompensated hepatitis B cirrhosis who have a history of TIPS and high levels of Alb and Hb are more likely to have recompensation,and it is relatively difficult for patients with portal vein thrombosis and an increase in portal vein width to achieve recompensation.

Objective:To assess the use of R-R interval (RRI) sequence data collected via wearable devices to evaluate emotional dynamic factors in adolescents with depressive disorders, and to analyze their impact on diagnosis and severity assessment.Methods:Clinical data were prospectively collected from 154 adolescent inpatients with depressive disorders (132 females, 22 males; age 12-18, mean: 13.5±1.6 years) treated at the Child Psychiatry Ward of Beijing Anding Hospital, Capital Medical University between January 2023 and January 2024.A control group of 152 healthy adolescents (62 females, 90 males; age 12-18, mean 14.5±1.3 years) was recruited during the same period. RRI data were obtained using the built-in photoplethysmography (PPG) sensor in the HUAWEI Band 7 wearable device. The device′s integrated emotion evaluation system extracted arousal and emotional valence (as indicators of emotional dynamics) from the collected RRIs. A Long Short-Term Memory (LSTM) network was employed to develop a model for depression diagnosis and severity prediction, while a random forest model was applied to generate receiver operating characteristic (ROC) curves to evaluate model performance. Binary Logistic regression was conducted to investigate the influence of arousal and emotional valence on depression diagnosis and severity.Results:A total of 429 records were collected and analyzed from 306 participants. The LSTM-based diagnosis and severity assessment models achieved area under the curve (AUC) of 0.896 9 and 0.715 3, respectively, indicating good model performance. Binary Logistic regression analysis showed that arousal and emotional valence had significant effects on diagnosis and severity. Specially, lower arousal in both the first 4 hours ( β=-8.906, 95%CI:-17.497 to -0.315) and second 4 hours ( β=-3.033, 95%CI:-5.109 to -0.957) significantly predicted positive depression diagnosis ( β=-1.219, 95%CI:-2.205 to -0.233), while emotional valence in the second 4 hours showed a trend toward a positive association ( β=0.675, 95%CI:-0.107-1.457). First 4-hour emotional valence: significantly positive association with severity ( β=0.322, 95%CI: 0.067-0.577), second 4-hour arousal level: negative association with severity ( β=-0.258, 95%CI:-0.527 to 0.011), whereas arousal in the second 4 hours had a marginal negative effect (β=-0.258, 95% CI:-0.527 to 0.011). Conclusion:RRI may serve as a useful auxiliary measure in diagnosing depressive disorders and predicting severity among adolescents. Wearable smart devices offer promising potential for screening emotional dynamic factors related to adolescent depression.

Objective To investigate the influencing factors for recompensation in patients with decompensated hepatitis B cirrhosis,and to establish a predictive model.Methods A total of 517 patients who attended The Third People's Hospital of Kunming and were diagnosed with decompensated hepatitis B cirrhosis from January 1,2016 to December 31,2022 were enrolled.The clinical data of the patients were reviewed,and the 207 patients with no portal hypertension-related complications within at least 1 year were enrolled as recompensation group,while the 310 patients without recompensation were enrolled as persistent decompensation group.Related clinical data were collected,and the univariate and multivariate Cox regression analyses were performed for the factors that might affect the occurrence of recompensation.The independent-samples t test was used for comparison of normally distributed continuous data between two groups,and the Mann-Whitney U test was used for comparison of non-normally distributed continuous data between two groups;the chi-square test or the Fisher's exact test was used for comparison of categorical data between two groups.The"rms"package was used to establish a nomogram;the receiver operating characteristic(ROC)curve was plotted,and the area under the ROC curve(AUC)was calculated;the Hosmer-Lemeshow test was used to evaluate the degree of fitting of the model;the"Calibration Curves"package was used to plot the calibration curve for model assessment.Results Among the patients with decompensated hepatitis B cirrhosis,207(40.03%)had recompensation.The univariate Cox regression analysis showed that there were significant differences between the recompensation group and the persistent decompensation group in TIPS history,genotyping,portal vein thrombosis,complicated infection,Child-Pugh class,age,hemoglobin(Hb),platelet count,total protein,albumin(Alb),alanine aminotransferase,triglyceride,cholesterol,creatinine,Na,interleukin-6,high-sensitivity C-reactive protein,portal vein width,and portal vein velocity(all P<0.05).The multivariate regression analysis showed that TIPS history(hazard ratio[HR]=2.491,95%confidence interval[CI]:1.325-4.681,P=0.005),portal vein thrombosis(HR=0.345,95%CI:0.152-0.783,P=0.001),Hb(HR=1.007,95%CI:1.000-1.013,P=0.028),Alb(HR=1.048,95%CI:1.017-1.080,P=0.002),and portal vein width(HR=0.899,95%CI:0.835-0.967,P=0.004)were independent influencing factors for recompensation in patients with decompensated hepatitis B cirrhosis.A nomogram model was established based on the above five influencing factors,and the Hosmers-Lemeshow test showed that this model had a good degree of fitting(χ2=3.202,P=0.921).The nomogram model had an AUC of 0.728,a sensitivity of 50.3%,and a specificity of 85.0%,and the calibration curve showed good consistency between the actual value of this model in predicting the occurrence of recompensation and the predicted value in patients with decompensated hepatitis B cirrhosis.Conclusion Patients with decompensated hepatitis B cirrhosis who have a history of TIPS and high levels of Alb and Hb are more likely to have recompensation,and it is relatively difficult for patients with portal vein thrombosis and an increase in portal vein width to achieve recompensation.