Preliminary Construction And Analysis Of Prognostic Models For Systemic Sclerosis

ObjectiveSystemic sclerosis/scleroderma（SSc）,also known as scleroderma,is a diffuse connective tissue disease characterized by localized or diffuse skin thickening or fibrosis,often involving multiple organs and tissues such as the heart,lungs,kidneys,and digestive system,with a high fatality rate and poor prognosis.Among them,systemic sclerosis-pulmonary arterial hypertension（SSc-PAH）and systemic sclerosis-associated interstitial lung disease（SSc-ILD）are the most common fatal complications of SSc.However,due to the heterogeneity of clinical manifestations and prognosis of SSc patients,the clinical diagnosis and evaluation of this disease have always been an urgent problem in clinical medicine.Therefore,based on the clinical data of SSc patients,this project explores the risk factors affecting the prognosis of patients,exploratorily constructs the SSc prognostic model,SSc-PAH prediction model and SSc-ILD prediction model,aiming to provide a reference for improving the prognosis of SSc patients and early identification of Important complications of SSc.MethodsThe relevant medical records of patients with SSc who were inpatient and outpatient in the Department of Clinical Immunology of the First Affiliated Hospital of Air Force Medical University from January 2009 to November 2021 were collected,and the patients were divided into Limited cutaneous systemic sclerosis（lc SSc）,Diffuse cutaneous sclerosis（Dc SSc）and overlap syndrome syndrome（OS）groups.General and demographic characteristics of the patient,clinical presentation,imaging data,and laboratory data were gathered.Based on outpatient follow-up and telephone follow-up,the patient’s survival information and long-term prognosis were generated.We performed a descriptive analysis of the collected patient data and compared differences between characteristics of patients with each SSc subtype.Depending on the characteristics of patient survival and prognosis,it was divided SSc patient survival cohort,SSc-PAH cohort and SSc-ILD cohort.Univariate Cox regression and multivariate Cox regression models were used to screen the factors affecting survival,SSc-PAH and SSc-ILD,the smallest variable group of AIC（Akaike information criterion）was used to construct a predictive model through all-subsets regression.Variables meaningful from the univariate analysis were incorporated into and random survival forest（RSF）and neural network-based Cox-Time models were constructed.In this study,the degree of differentiation of the three models was measured by the consistency index（C-index）or time-dependent C-index(C^td),and the prediction error of different models was quantified by the combined Brier score of 3 years and 5 years.For the Cox and RSF models,the accuracy of the model prediction was further evaluated by the area under the curve（AUC）of the time-dependent receiver operating characteristic curve（td ROC）at 3 and 5 years.Results1.A total of 517 eligible cases was included,including 437 women（84.5%）,80 men（15.5%）,141（27.3%）patients with dc SSc,239（46.2%）patients with lc SSc,and 137patients with OS type（26.5%）.2.There was no significant difference in the occurrence of important organ involvement such as SSc-PAH,SSc-ILD and gastrointestinal involvement in each group.However,patients with OS subtype are more likely to have symptoms such as muscle-joint involvement and dry mouth and eyes,and are more likely to have SSA and Ro-52antibodies,and the white blood cells,red blood cells,hemoglobin,and platelets in this group are significantly lower than those in other groups（P<0.05）.3.In the SSc prognostic model,the variables included in the Cox regression model were Modified Rodnan skin score（m RSS）,Digital Ulcers（DU）,and SSc-PAH.The hazard rate（HR）and 95%confidence interval（CI）corresponding to these variables were 1.06（1.01,1.10）,1.87（1.01,3.48）,and 3.80（2.07,6.98）,respectively.The model C index is0.77,and the IBS of 3 years and 5 years is 0.15 and 0.18.The AUC of td ROC is 0.73 and0.69,respectively.The C index of the Cox-Time model is 0.79,and the IBS is 0.09 and0.11.The C index of the random survival forest model was 0.79,the IBS was 0.07 and0.11,and the AUC of td ROC was 0.74 and 0.70.4.In the SSc-PAH prediction model,the variables included in the Cox regression model were OS subtype,combined DU,NTpro-BNP level,24-hour urine protein quantification,and ESR.The HR and 95%CIs corresponding to the above factors were1.73（1.07,2.81）,1.90（1.20,3.02）,3.57（2.16,5.92）,1.47（1.01,2.14）,and 1.01（1.00,1.02）,respectively.The C index of the model is 0.83,and the IBS for 3 years and 5 years is 0.13and 0.12.The AUC of td ROC is 0.82,0.77.The C index of the Cox-Time model is 0.79,and the IBS is 0.17 and 0.14.The C index of the random survival forest model was 0.82,and the IBS was 0.12 and 0.16.The AUC of 3-year and 5-year td ROC is 0.86 and 0.83.5.In the SSc-ILD prediction model,the variables included OS subtype,m RSS had significant effects on the development of SSc-ILD.ConclusionsPatients with different subtypes of SSc have strong heterogeneity in clinical symptoms and autoantibody expression,and OS subtypes may be more likely to require timely diagnosis and more systematic monitoring and follow-up.MRSS score,DU,and PAH were significant factors affecting the prognosis of SSc patients.OS subtype,Ntpro-BNP,24-hour urine protein quantification,elevated ESR,and DU were significant influencing factors of SSc-PAH.The OS subtype and m RSS were significant factors affecting the development of SSc-ILD.The RSF model can improve the prognosis of SSc patients and the accuracy of SSc-PAH prediction,provide a reference for clinical diagnosis and treatment.The prognostic model and exploration of analytical methods preliminaries constructed in this paper provide the necessary research base and convenient conditions for in-depth study of the clinical diagnosis,evaluation and risk factors affecting prognosis of this disease.