| 【Background】Myasthenia gravis(MG)is an organ-specific autoimmune disease with acquired neuromuscular junction transmission disorders mediated by autoantibodies.The overall global prevalence of MG is(150-250)/million,and the annual incidence rate is(4-10)/million.Its repeated and protracted course of disease,complex and changeable conditions,high disability rate,and high incidence of depression,high suicide rate and long-term high medical expenses have brought great burdens to families and society.The clinical symptoms of MG are fluctuating muscle weakness and fatigue,which are often aggravated by activity and improved after rest.Currently,it is recognized that the pathogenesis of MG is closely related to the abnormal immune response of the thymus,and thymectomy has become an important treatment for MG.A series of complications may occur after thymectomy,including myasthenic crisis,phrenic nerve palsy,hemorrhage,and infection,among which myasthenic crisis is the most serious complication and increases the risk of death.Actively looking for risk factors related to myasthenic crisis after thymectomy and establishing a prediction model will help medical staff to intervene in advance and minimize the occurrence of myasthenic crisis.【Aims】Postoperative myasthenic crisis(POMC)is the most serious complication after thymectomy in patients with MG.The purpose of this study was to analyze the demographic and clinical characteristics of MG patients with POMC,and to find the relevant risk factors for POMC.Based on logistic regression analysis,a logistic regression model and a decision tree model were established to predict the myasthenic crisis occurring within 30 days after thymectomy in patients with MG.【Methods】(1)A retrospective cohort study was conducted to investigate the demographic information and clinical data of MG patients who underwent thymectomy in our hospital from January 2015 to October 2020.MG patients were divided into POMC group and non-POMC group according to whether POMC occurred within 30 days after operation,and then the demographic and clinical characteristics of the two groups were descriptively analyzed,and the characteristics of the two groups were compared at baseline.(2)The correlation between gender,age of onset,ethnicity,province of origin,hypertension,diabetes,smoking,drinking,first symptom,preoperative course of disease,preoperative bulbar symptoms,preoperative MGFA type,preoperative intravenous immunoglobulin(IVIG),preoperative medication regimen,preoperative thymus status,history of myasthenic crisis and surgical method and POMC was analyzed by univariate logistic regression analysis.The factors with P<0.05 in the univariate analysis results were included in the multivariate logistic regression analysis to obtain the independent risk factors of POMC.(3)All MG patients were randomly divided into modeling group and verification group according to 7:3 to establish and verify the risk prediction model.The factors with P<0.2 in the single factor analysis results were used as initial variables,and the stepwise regression backward method was adopted.According to the Akaike information criterion(AIC),the regression was stopped when the minimum AIC was reached,and the model with the minimum AIC value was obtained and listed.The line diagram is displayed to obtain the logistic regression model.The logistic regression model was evaluated in the modeling group and the validation group by area under curve(AUC),calibration curve,and decision curve analysis(DCA).Similarly,the factors with P<0.2 in the single factor analysis results were used as initial variables,and the "rpart" package in the R language software was used to draw the decision tree model,and the confusion matrix,accuracy rate,precision rate,the recall rate,F1 value and AUC value were used to evaluate the decision tree model.【Results】(1)Among the 522 MG patients who underwent thymectomy,POMC occurred in 52 cases,with an incidence rate of 9.96%.The descriptive analysis of the demographic characteristics of the two groups showed that the mean age of onset in the POMC group was49.27±14.11 years old,and that in the non-POMC group was 43.56±14.62 years old.Male patients accounted for 55.80% in the POMC group,and the male and female patients in the non-POMC group were basically equal.The provinces of origin of the two groups of patients were mainly Shaanxi and Gansu.Patients with hypertension and diabetes accounted for 21.20%and 9.60% of the POMC group,respectively,while the two chronic diseases accounted for16.40% and 6.40% of the non-POMC group.The smoking percentages of the two groups were basically the same,while the number of alcohol drinkers in the POMC group accounted for7.70%,which was higher than the 2.60% in the non-POMC group.Descriptive analysis of the clinical characteristics of the two groups of patients showed that first symptom in the POMC group was generalized,accounting for 80.80%,which was higher than that in the non-POMC group.In the POMC group,48.10% of patients had bulbar symptoms before operation,which was higher than that in the non-POMC group.The percentage of preoperative MGFA type Ⅰpatients in the POMC group was lower than that of the non-POMC group,the percentage of type Ⅱ patients was similar,and the percentage of MGFA type Ⅲ-Ⅳ patients was higher than that of the non-POMC group.In terms of preoperative medication regimen,the percentage of patients in the POMC group who received no treatment or pyridostigmine bromide alone was lower than that in the non-POMC group,and the percentage of patients who used steroids combined with non-steroidal immunosuppressants was higher than that in the non-POMC group.The percentages of preoperative IVIG,preoperative thymus status as thymoma,history of myasthenic crisis,and thoracotomy in POMC group were higher than those in non-POMC group.The characteristics of the two groups were compared at baseline: age of onset,first symptom,preoperative bulbar symptoms,preoperative MGFA type,preoperative IVIG,preoperative medication regimen,preoperative thymus status,preoperative myasthenic crisis history and surgical methods showed a statistical difference(P <0.05).(2)Univariate logistic regression analysis showed that: older onset age,generalized symptoms at first symptom,preoperative bulbar symptoms,preoperative MGFA type Ⅲ-Ⅳ,preoperative IVIG,preoperative medication(steroids combined with non-steroidal immunosuppressants),preoperative thymus status as thymoma,history of myasthenic crisis,and thoracotomy were high risk factors for POMC in MG patients(P<0.05).Further multivariate logistic regression analysis showed that older age at onset,generalized symptoms at first symptom,preoperative MGFA type Ⅲ-Ⅳ,and thoracotomy were independent risk factors for POMC(P<0.05).(3)A logistic regression model was established,and the final included variables were age of onset,alcohol consumption,first symptom,preoperative MGFA type,and surgical method.After the model was verified in the modeling group and the verification group,the AUCs were obtained as 0.794,0.791.The calibration curve showed that the predicted probability of POMC in the modeling group and the verification group by the logistic regression model was close to the actual probability.The clinical decision curve shows that the model has a better net benefit rate.A decision tree model was constructed,and the final included variables were age of onset,preoperative MGFA type,first symptom,and preoperative thymus status.After the decision tree model was verified by the modeling group and the verification group,the accuracy rates were 0.923,0.924,the precision rates were 0.595,0.647,the recall rates were 0.629,0.647,the F1 values were 0.611,0.647,and the AUC values were 0.888 and 0.761,respectively.【Conclusions】(1)In terms of clinical characteristics,there were statistical differences between the two groups in first symptoms,preoperative bulbar symptoms,preoperative MGFA types,preoperative IVIG,preoperative medication regimens,preoperative thymus status,myasthenic crisis history and surgical methods,indicating that these factors may have an impact on the occurrence of POMC.(2)Multivariate logistic regression analysis showed that older age at onset,generalized symptoms,MGFA type Ⅲ-Ⅳ,and thoracotomy were independent risk factors for POMC.(3)The constructed logistic regression model and decision tree model have a certain predictive ability for the occurrence of POMC,and have certain application value in clinical practice. |
PDF Full Text Request