A model for predicting the prognosis of patients with anti-synthetase syndrome and uses thereof

Abstract

The present application relates to a kind of model for predicting the prognosis of anti-synthetase syndrome patient and its use, the model includes 5 parameters: age of onset, lactic acid dehydrogenase concentration, albumin concentration, neutrophil and lymphocyte ratio, respiratory failure;In the model, age of onset is greater than 55 years old, marked as 3 points, otherwise 0 points;Lactic acid dehydrogenase is greater than 299 U / L, marked as 1 point, otherwise 0 points;Albumin is less than 34.8g / L, marked as 1 point, otherwise 0 points;Respiratory failure occurs, marked as 2 points, otherwise 0 points;Neutrophil and lymphocyte ratio is greater than 5.6, marked as 1 point, otherwise 0 points;The total score of the above 5 parameters, 0-2 is low-risk group, 3-5 is medium-risk group, 6-8 is high-risk group.The prognosis of the anti-synthetase syndrome (ASyS) patient is to predict the survival rate of patient within 1-10 years after onset.

Description

Technical Field

[0001] This invention belongs to the field of biomedical technology, specifically, it relates to a model for predicting the prognosis of patients with antisynthetic enzyme syndrome and its applications. Background Technology

[0002] Anti-synthesizer syndrome (ASyS) is an autoimmune disease characterized by the presence of antibodies against aminoacyl-tRNA synthetase (ARS), with clinical manifestations primarily including interstitial lung disease (ILD), inflammatory myopathy, and arthritis [1, 2]. The clinical course, treatment response, and prognosis of ASyS patients are highly heterogeneous. Despite treatment, some patients may develop progressive pulmonary fibrosis with gradually deteriorating lung function, while others may develop rapidly progressive interstitial lung disease (RPILD), a type that may become resistant to treatment and lead to death within a short period [3-6]. This prognostic variability—with survival ranging from weeks to decades depending on the disease phenotype—highlights the urgent need for accurate risk stratification in clinical practice.

[0003] Currently, there are no universally accepted predictors of mortality associated with ASyS, although some studies have identified potential risk factors for adverse outcomes in ASyS, including advanced age, male sex, severe dyspnea at diagnosis, dysphagia, muscle weakness, decreased carbon monoxide diffusing capacity (DLco), elevated serum ferritin, and RPILD [4, 7-10]. However, the predictive value of these factors varies across different studies and requires further validation. The lack of reliable prognostic tools and effective treatment strategies has led to adverse outcomes in ASyS patients, with a reported standardized mortality ratio of 5.49 (95% confidence interval [CI] 4.46–6.53)

[11] . A comprehensive assessment of mortality-related risk factors and the development of a clinical model that integrates multiple clinically available parameters may improve predictive accuracy and facilitate individualized interventions for high-risk individuals, which could be an effective strategy for improving the prognosis of ASyS patients.

[0004] Therefore, we aim to systematically evaluate prognostic factors in ASyS patients and develop a practical risk stratification model using data from multicenter, large-scale, well-characterized retrospective cohorts.

[0005] Based on this, the present invention is proposed. Summary of the Invention

[0006] This invention first relates to the use of a scoring model in the preparation of a diagnostic kit for predicting the prognosis of patients with antisynthetic enzyme syndrome (ASyS), wherein the scoring model includes the following five parameters: Age of onset, serum lactate dehydrogenase concentration, serum albumin concentration, whole blood neutrophil to lymphocyte ratio, respiratory failure; The respiratory failure is defined as: the presence of obvious respiratory symptoms and arterial blood gas analysis showing PaO2 ≤60 mmHg under anaerobic support, or PaO2 / inhaled oxygen concentration (FiO2) ≤300 mmHg with assisted oxygen supply; The respiratory symptoms mentioned include, but are not limited to, cough and difficulty breathing.

[0007] In the aforementioned model, If the age of onset is >55 years old, score 3 points; otherwise, score 0 points. If lactate dehydrogenase > 299 U / L, score 1 point; otherwise, score 0 points. If albumin level is <34.8 g / L, score 1 point; otherwise, score 0 points. If respiratory failure occurs, score 2 points; otherwise, score 0 points. A neutrophil-to-lymphocyte ratio > 5.6 is scored as 1 point; otherwise, it is scored as 0 points. The total score of the above 5 parameters is 0–2 for low risk, 3–5 for medium risk, and 6–8 for high risk.

[0008] The prediction of prognosis for patients with antisynthetic enzyme syndrome (ASyS) refers to the prediction of the patient's survival rate within 1-10 years after the onset of the disease.

[0009] Specifically, (1) 1-year survival rate: 100% for low-risk patients, 96-98% for intermediate-risk patients, and 59-93% for high-risk patients; (1) 3-year survival rate: 99-100% for low-risk patients, 89-93% for intermediate-risk patients, and 55-80% for high-risk patients; (1) 5-year survival rate: 97-99% for low-risk patients, 81-88% for intermediate-risk patients, and 47-67% for high-risk patients; (2) 10-year survival rate: 76-96% for low-risk patients, 56-79% for intermediate-risk patients, and 9-24% for high-risk patients.

[0010] The present invention also relates to a diagnostic kit and / or device for predicting the prognosis of patients with antisynthetic enzyme syndrome (ASyS), the kit and / or device comprising: reagents for detecting serum lactate dehydrogenase concentration, serum albumin concentration, whole blood neutrophil to lymphocyte ratio, and reagents and / or devices for detecting respiratory failure. The respiratory failure is defined as: the presence of obvious respiratory symptoms and arterial blood gas analysis showing PaO2 ≤60 mmHg under anaerobic support, or PaO2 / inhaled oxygen concentration (FiO2) ≤300 mmHg with assisted oxygen supply; The respiratory symptoms mentioned include, but are not limited to, cough and difficulty breathing.

[0011] Preferably, serum lactate dehydrogenase concentration, serum albumin concentration, whole blood neutrophil-to-lymphocyte ratio, and respiratory failure status are measured simultaneously on the prognostic testing day.

[0012] The aforementioned detection kit and / or detection device provides a prognostic score based on the detection values ​​of serum lactate dehydrogenase concentration, serum albumin concentration, whole blood neutrophil to lymphocyte ratio, patient respiratory failure status, and age of onset. Specifically... If the age of onset is >55 years old, score 3 points; otherwise, score 0 points. If lactate dehydrogenase > 299 U / L, score 1 point; otherwise, score 0 points. If albumin level is <34.8 g / L, score 1 point; otherwise, score 0 points. If respiratory failure occurs, score 2 points; otherwise, score 0 points. A neutrophil-to-lymphocyte ratio > 5.6 is scored as 1 point; otherwise, it is scored as 0 points. The total score of the above 5 parameters is 0–2 for low risk, 3–5 for medium risk, and 6–8 for high risk.

[0013] The prediction of prognosis for patients with antisynthetic enzyme syndrome (ASyS) refers to the prediction of the patient's survival rate within 1-10 years after the onset of the disease.

[0014] Specifically, (1) 1-year survival rate: 100% for low-risk patients, 96-98% for intermediate-risk patients, and 59-93% for high-risk patients; (1) 3-year survival rate: 99-100% for low-risk patients, 89-93% for intermediate-risk patients, and 55-80% for high-risk patients; (1) 5-year survival rate: 97-99% for low-risk patients, 81-88% for intermediate-risk patients, and 47-67% for high-risk patients; (2) 10-year survival rate: 76-96% for low-risk patients, 56-79% for intermediate-risk patients, and 9-24% for high-risk patients.

[0015] The beneficial effects of this invention are as follows: The scoring model of this invention integrates five easily accessible clinical parameters, providing a simple and practical tool for predicting the mortality risk of ASyS patients and effectively assisting in the clinical management of ASyS patients. Attached Figure Description

[0016] Figure 1Risk factors for death in patients with antisynergistic syndrome were identified using univariate Cox regression analysis. The figures show the following risk factors: Female, Age at onset > 55 years, Muscle weakness, Heliotrope rash, Mechanic's hands, Gottron sign, V-sign, Shawl sign, Holster sign, Raynaud phenomenon, Arthralgia, Fever, Respiratory failure, Pulmonary hypertension, Pulmonary infection, Ro52+, ANA+, Anti-ARS antibodies, Jo1, PL7, PL12, EJ, and Lymphocyte count < 0.8 x 10^9. 9 ), NLR>5.6 (neutrophil-to-lymphocyte ratio>5.6), ALB<34.8 g / L (albumin concentration<34.8g / L), LDH>299 U / L (lactate dehydrogenase concentration>299 U / L), CK>1300 U / L (creatine kinase concentration>1300U / L), CRP>3.7 mg / dL (C-reactive protein>3.7mg / mL), ESR>26 mm / h (erythrocyte sedimentation rate>26mm / h)\Lung CT pattern (lung CT pattern), NSIP (nonspecific interstitial pneumonia), OP (organizing pneumonia), NSIP+OP (nonspecific interstitial pneumonia + organizing pneumonia), UIP (common interstitial pneumonia), Normol (normal).

[0017] Figure 2 Time-dependent AUC analysis of the risk model based on CPH; 2a. AUC analysis of the training queue; 2b. AUC analysis of the internal validation queue; Figure 3 A calibration plot of predicted mortality rate versus observed mortality rate based on a risk model using CPH; 3a. Calibration plot of predicted mortality rate versus observed mortality rate in the training cohort; 3b. Calibration plot of predicted mortality rate versus observed mortality rate in the internal validation cohort.

[0018] Figure 4 ALARN scoring method and its prediction results in the training and internal validation queues; 4a. Thresholds and scores for each predictor, and the scoring relationships among high-risk, intermediate-risk, and low-risk patients; 4b. In the derivation cohort, ALARN scores were negatively correlated with survival rates at 1, 3, and 5 years of follow-up, with survival probability decreasing as scores increased. 4c. In the training cohort, there were significant differences in survival rates among the low-risk, medium-risk, and high-risk groups. 4d. In the internal validation cohort, there were significant differences in survival rates among the low-risk, medium-risk, and high-risk groups.

[0019] Figure 5 The results of testing the risk model and ALARN scoring criteria for validating CPH in an external validation queue are shown in the figure. 5a. Perform time-dependent AUC analysis for 1 year, 3 years, 5 years and 10 years in the external validation queue; 5b. Calibration plot of predicted mortality rate versus observed mortality rate from external validation cohort.

[0020] Figure 6 Using the ALARN scoring method, patients in the external validation cohort were divided into low-risk, intermediate-risk, and high-risk groups, and there were significant differences in mortality rates among the three groups.

[0021] Figure 7 1. Variable selection using LASSO regression; 7a. LASSO coefficient curves of candidate clinical features; 7b. Selection of the optimal parameter (λ) using the minimum criterion through ten-fold cross-validation. The two vertical dashed lines represent the optimal scores obtained using the minimum criterion (left) and the one standard deviation criterion (right); the optimal λ value on the right was selected, λ = 0.069 (log(λ) = -2.67), resulting in five features with non-zero coefficients. Detailed Implementation

[0022] Anti-ARS antibody test In this study, anti-ARS antibodies (including anti-Jo1, anti-EJ, anti-PL7, anti-PL12, and anti-OJ antibodies) were detected in routine clinical practice using the EUROLINE immunoblotting kit (catalog number DL 1530-1601-4G, EUROIMMUN, Lübeck, Germany) according to the manufacturer's instructions, or retrospectively using stored serum samples. Signal intensity was measured using EURO-BlotOne (EUROIMMUN), with a cutoff threshold of 25 [19, 20]. Band intensities between 25 and 75 were considered weakly positive, and >75 was considered strongly positive. For patients with weakly positive anti-ARS antibodies, antibody status was confirmed by repeated testing and validated using a confirmatory ELISA kit (catalog number 7840R, Medical & Biological Laboratories, Nagoya, Japan).

[0023] Statistical analysis In this study, categorical variables were summarized as percentages and compared using chi-square or Fisher's exact test, as appropriate. Continuous variables were summarized as mean (standard deviation) or median (first quartile, third quartile) based on data distribution and compared using t-tests or Mann-Whitney U tests, as appropriate. To examine the impact of treatment on the predictive model, initial treatments (including high-dose glucocorticoids (GC; prednisone equivalent >1 mg / kg / day), calcineurin inhibitors (cyclosporine or tacrolimus), cyclophosphamide (CTX), mycophenolate mofetil (MMF), and intravenous immunoglobulin (IVIg)) were forcibly included as potential confounding factors in the final multivariate model. The 95% CI for C-index and IBS was calculated using 1000 repeated bootstraps. Kaplan-Meier curves were plotted for survival analysis and compared using the log-rank test.

[0024] All statistical analyses and plots were performed using IBM SPSS 26.0 (IBM, Armonk, NY, USA) and R 4.3.3 (R Foundation for Statistical Computation, Vienna, Austria). A two-sided p < 0.05 was considered statistically significant.

[0025] Example 1: Research Cohort and Selection of Variables 1. Study the cohort situation A study of patients with antisynthetic enzyme syndrome (ASyS) from three tertiary hospitals (China-Japan Friendship Hospital, Nanjing Drum Tower Hospital, and the First Affiliated Hospital of Zhengzhou University) was approved by the ethics committee of the lead center, China-Japan Friendship Hospital (2022-KY-156). All patients were of Asian descent. The diagnosis of ASyS was determined according to the criteria proposed by Connors et al.

[12] , and a total of 1194 patients met the criteria. All patients signed written informed consent forms before the data were reviewed. (1) 763 patients from the China-Japan Friendship Hospital cohort (enrollment period from April 2004 to May 2023) were used as the derivation cohort and further randomly divided into two groups. Training queue (n=531, 69.6%) were used for variable selection and model building; internal Verification queue (n=232, 30.4%) were used for optimal model selection and internal validation.

[0026] (2) The study group consisted of 220 patients from the Nanjing cohort (enrolled from January 2017 to November 2020) and 211 patients from the Zhengzhou cohort (enrolled from September 2016 to January 2024). External verification queue (n=431). This study is reported in accordance with the updated TRIPOD+AI guidelines

[13] .

[0027] Data extracted from patients' initial admission records included demographic characteristics, clinical features, laboratory parameters, imaging findings, pulmonary function tests (PFTs), and initial treatment regimen. Survival status was verified through medical records and telephone interviews. All follow-ups were completed in June 2024, and each participant's survival status was recorded at the last follow-up or at death. The last follow-up was defined as the date of the last recorded contact with the patient, which could be the last outpatient visit or the last successful telephone interview with the patient or their family. For patients lost to follow-up, the last follow-up date was the date of the last available clinical visit or the last telephone contact prior to loss to follow-up.

[0028] This study included 1194 patients with ASyS (Table 1.1), of whom 763 (63.9%) were from the China-Japan Friendship Hospital cohort as the inferred cohort, and 431 (36.1%) were from the Nanjing and Zhengzhou cohorts as external validation cohorts. The mean age at onset was 52.9 ± 12.5 years, and 858 (71.9%) were female. The inferred cohort was randomly assigned to a training cohort and an internal validation cohort, and baseline characteristics and treatment regimens were comparable after data imputation (all p > 0.05).

[0029] The median follow-up time for the entire cohort was 59 months (IQR 34–94 months). During the follow-up period, 178 patients (14.9%) died. Of the 178 deaths, 129 (72.5%) died from respiratory failure, 12 (6.7%) from cardiovascular disease, 12 (6.7%) from malignancy, and 25 (14.0%) from other causes such as old age, liver failure, or unknown causes. The cumulative survival rates at 1, 3, 5, and 10 years were 95.9%, 92.2%, 87.6%, and 73.0%, respectively.

[0030] 2. Variable Definition Blood samples were collected for laboratory analysis upon the patient's initial admission. Clinical presentations and laboratory test results were retrospectively recorded, consistent with the time of blood sample collection.

[0031] Myasthenia gravis is defined as a score of <5 for at least one muscle group, according to the UK Medical Research Council's Myasthenia Gravis Scale[7].

[0032] ILD is diagnosed through a comprehensive assessment, including respiratory symptoms, imaging findings, and available post-exposure prognoses (PFTs).

[0033] High-resolution computed tomography (HRCT) patterns are classified into four types by two specialist radiologists: nonspecific interstitial pneumonia (NSIP), organizing pneumonia (OP), NSIP combined with OP (NSIP+OP) and common interstitial pneumonia (UIP) [14, 15].

[0034] RPILD is defined as meeting ≥2 of the following criteria within 3 months: (1) progressive worsening of dyspnea; (2) accelerated deterioration of interstitial abnormalities on HRCT; (3) a decrease in forced vital capacity (FVC) of >10% or a decrease in alveolar oxygen partial pressure (PaO2) of 10 mmHg

[16] .

[0035] Respiratory failure is defined as the presence of obvious respiratory symptoms (mainly cough and dyspnea) and further assessment by arterial blood gas analysis, confirming that PaO2 ≤ 60 mmHg under anaerobic support, or PaO2 / inhaled oxygen concentration (FiO2) ≤ 300 mmHg with assisted oxygenation

[17] .

[0036] Pulmonary infection is judged based on a combination of clinical manifestations, imaging results and positive microbiological results. Pulmonary hypertension is defined as pulmonary artery systolic blood pressure >40 mmHg as assessed by echocardiography

[18] . The follow-up time for survival analysis is defined as the time from onset to death or the last follow-up.

[0037] 3. Variable selection rules Variable selection and model building were performed in the training queue. Initial correlation analysis showed no strong correlation between any two variables (all Spearman coefficients ≤ 0.7). The optimal cutoff values ​​for continuous variables were determined to identify the thresholds most significant for survival outcomes.

[0038] Given the very high incidence of ILD in the ASyS cohort (97.3%), ILD was excluded from variable selection. RPILD was also excluded because it is a retrospective parameter, typically diagnosed during severe disease progression.

[0039] In summary, the relationship between 26 routinely assessed ASyS clinical characteristics and patient mortality risk factors was evaluated in the training cohort. The results are shown in Table 1.2. Figure 1 .

[0040] Table 1.1 Clinical characteristics of patients with antisynthetic enzyme syndrome

[0041] Table Notes: Continuous data are expressed as mean (SD) or median (interquartile range), while binary data are expressed as the number of patients (percentage).

[0042] * Candidate variables used for prognostic assessment.

[0043] a) In the derivation cohort, 23 patients (3.0%) had missing data and underwent imputation, while in the external validation cohort, 18 patients (4.2%) had missing data and underwent imputation; b) In the derivation cohort, 17 patients (2.2%) had missing data and underwent imputation, while in the external validation cohort, 16 patients (3.7%) had missing data and underwent imputation; c) In the derivation cohort, 2 patients (0.3%) had missing data and underwent imputation, while in the external validation cohort, 1 patient (0.2%) had missing data and underwent imputation; d) In the derivation cohort, 53 patients (6.9%) had missing data and underwent imputation, while in the external validation cohort, 3 patients (0.7%) had missing data and underwent imputation; e) In the derivation cohort, 20 patients (2.6%) had missing data and underwent imputation, while in the external validation cohort, 17 patients (3.9%) had missing data and underwent imputation; f) In the derivation cohort, 18 patients (2.4%) had missing data and underwent imputation, while in the external validation cohort, 22 patients (5.1%) had missing data and underwent imputation; in the g derivation cohort, 9 patients (1.2%) had missing data and underwent imputation, while in the external validation cohort, 11 patients (2.6%) had missing data and underwent imputation; in the h derivation cohort, 49 patients (6.4%) had missing data and underwent imputation, while in the external validation cohort, 16 patients (3.7%) had missing data and underwent imputation; in the i derivation cohort, 78 patients (10.2%) had missing data and underwent imputation, while in the external validation cohort, 7 patients (1.6%) had missing data and underwent imputation.

[0044] Abbreviations: NSIP, Nonspecific interstitial pneumonia; OP, Organizing pneumonia; RPILD, Rapidly progressive interstitial lung disease; UIP, Common interstitial pneumonia.

[0045] Table 1.2. Risk factors for death in patients with antisynthetic enzyme syndrome identified by univariate Cox regression analysis

[0046] Abbreviations: NSIP, Nonspecific interstitial pneumonia; OP, Organizing pneumonia; UIP, Common interstitial pneumonia.

[0047] Example 2: Variable Selection and Model Development Variables with missing data exceeding 20% ​​were excluded from the analysis to minimize bias from missing data. Variables excluded due to missing data >20% included serum ferritin, lymphocyte subset counts (including CD4+, CD8+, CD19+, and NK cells), and PFTs.

[0048] Residual variables were handled using multiple imputation. Chain equation multiple imputation (MICE), implemented using the `mice` package in R, was used to handle missing data. The imputation method was selected based on variable type: predicted mean matching (PMM) was used for continuous variables (e.g., neutrophil count, lymphocyte count, albumin, LDH, CK, CRP, ESR); logistic regression was used for binary variables (e.g., ANA, anti-Ro52 antibody); and multinomial logistic regression was used for unordered categorical variables (lung CT patterns). Five imputation datasets were generated, and the imputation models included all available variables as predictors. Diagnostic assessment included using trajectory plots to confirm convergence and using strip plots and density plots to compare observed and imputed distributions. Furthermore, all imputed values ​​were manually reviewed within biologically reasonable ranges to ensure clinical rationality. After imputation, a Cox proportional hazards model was fitted to each imputed dataset to assess survival predictors. The results were summarized using Rubin's rule to obtain pooled hazard ratios (HRs), 95% confidence intervals (CIs), and p-values.

[0049] Variable selection was performed in the training cohort. Pairwise correlation analysis was conducted, and when two variables showed a strong correlation (Spearman correlation coefficient > 0.7), only one of them was selected for further analysis. The decision on which variable to retain was based on: (1) the correlation pattern with other variables (preferring variables with less strong correlation), (2) clinical relevance and ease of measurement in routine practice, and (3) consistency with previous literature.

[0050] For continuous variables, the surv_cutpoint function in the 'survminer' R package is used to determine the optimal cutoff value based on the maximum selection rank statistic method to best distinguish the survival outcomes between the two groups [21, 22]. These optimal cutoff values ​​are then used to convert these continuous variables into binary variables.

[0051] 1. Mortality risk model constructed using the Cox proportional hazards (CPH) method (Model 1) The Cox proportional hazards (CPH) approach was used for variable selection and model building. First, univariate Cox regression was performed, and variables with p < 0.05 were included in the least absolute contraction and selection operator (LASSO) regression analysis with 10-fold cross-validation

[23] . Subsequently, multivariate Cox regression analysis was performed to identify independent variables associated with mortality in ASyS patients. The final CPH model was developed based on these selected variables

[24] .

[0052] To identify robust prognostic factors in the training cohort, variables with p < 0.05 in the univariate analysis ( Figure 1 ) included in LASSO regression with 10-fold cross-validation ( Figure 7Multivariate Cox regression analysis of the variables selected by LASSO revealed five independent prognostic factors: age of onset (>55 years), respiratory failure, albumin (<34.8 g / L), lactate dehydrogenase (LDH) (>299 U / L), and neutrophil-to-lymphocyte ratio (NLR) (>5.6). A CPH model based on these five identified prognostic variables was then constructed.

[0053] 2. Mortality risk model constructed using the Random Survival Forest (RSF) algorithm (Model 2) Variable selection and model development were performed using the Random Survival Forest (RSF) algorithm

[25] . Variables were selected by combining Variable Importance (VIMP) scores

[26] and minimum depth

[27] . Specifically, we retained only variables with VIMP > 0.01 and a minimum depth below a threshold (defined as the median of the average minimum depth of all candidate variables plus one standard deviation). The final RSF model was developed based on the selected variables, and hyperparameters were tuned using a grid search method [25, 28]. The discrete grid for the hyperparameter combinations was defined as follows: the number of trees (ntree) ∈ {200, 500}, the number of variables randomly sampled as candidates at each split (mtry) ∈ {3, 5}, and the size of terminal nodes (nodesize) ∈ {15, 21}. Each combination was evaluated using 5-fold cross-validation, and the combination that produced the highest average C-index was selected as the optimal combination. The optimal hyperparameters were ntree = 200, mtry = 3, nodesize = 21. Other settings (including nsplit = 10, x-weighting, and moderate conservatism) follow the default recommendations of the RSF package. The effectiveness of the internal model is evaluated using out-of-bag (OOB) prediction errors.

[0054] The RSF algorithm was applied to identify risk factors associated with mortality in the training cohort, resulting in nine variables: age at onset (>55 years), respiratory failure, albumin (<34.8 g / L), LDH (>299 U / L), lung CT pattern of UIP, NLR (>5.6), C-reactive protein (CRP) (>3.7 mg / dL), mechanic's hand, and female sex. An RSF model incorporating these nine variables was then developed.

[0055] Example 3: Comparison and Evaluation of Models The predictive performance of the CPH and RSF models was compared by evaluating overall discrimination and calibration in an internal validation queue. The overall discrimination ability of each model was evaluated using the consistency index (C-index) [29, 30], which ranges from 0.5 to 1, with higher values ​​indicating better performance. The Integrated Brier score (IBS) was used as a measure of overall model performance, reflecting calibration and discrimination

[31] . To obtain the 95% confidence intervals (CI) for the C-index and IBS, we performed a bootstrap procedure, refitting the CPH and RSF models in each bootstrap iteration.

[0056] After determining the optimal model, other metrics were used to assess predictive accuracy. Discrimination was assessed by generating time-dependent receiver operating characteristic (ROC) curves and calculating the area under the curve (AUC) to assess the model’s ability to distinguish between different outcomes

[30] . Calibration plots were also generated to assess the consistency between predicted and observed outcomes

[32] .

[0057] The results are shown in Table 2. Table 2. Performance of the CPH and RSF models in predicting mortality risk in the internal validation cohort.

[0058] Note: The CPH model variables include age of onset, LDH, ALB, respiratory failure, and NLR. The RSF model includes 9 variables, including age of onset (>55 years), respiratory failure, albumin (<34.8 g / L), LDH (>299 U / L), lung CT pattern UIP, NLR (>5.6), C-reactive protein (CRP) (>3.7 mg / dL), mechanic's hand, and female sex.

[0059] Abbreviations: ALB, albumin; CI, confidence interval; CPH, Cox proportional hazards; C-index, concordance index; CRP, C-reactive protein; IBS, Brier score; LDH, lactate dehydrogenase; NLR, neutrophil-to-lymphocyte ratio; RSF, random survival forest.

[0060] (1) In predicting the risk of death in the internal validation cohort, the CPH model achieved a C-index (0.809; 95% CI: 0.735-0.873) comparable to that of the RSF model (0.808; 95% CI: 0.732-0.877). (2) The IBS (Comprehensive Brier Score) of the CPH model (0.135; 95% CI: 0.094-0.180) was lower than that of the RSF model (0.142; 95% CI: 0.099-0.195), indicating that its overall model performance was better.

[0061] These results indicate that the CPH model has better predictive performance than the RSF model. .

[0062] Given its superior performance, as well as the advantages of requiring fewer variables and better clinical interpretability, the CPH model was chosen for final implementation.

[0063] Further sensitivity analysis showed that the statistical significance of all five variables remained consistent after adjusting the initial treatment regimen (Table 3), indicating that the model is robust.

[0064] Table 3. Risk factors for predicting all-cause mortality after adjusting for treatment status.

[0065] Note: Treatment includes high doses of GC, CTX, CNI, MMF, IVIg, and JAKi.

[0066] Abbreviations: ALB, albumin; CI, confidence interval; CTX, cyclophosphamide; CNI, calcineurin inhibitor; GC, glucocorticoid; HR, hazard ratio; IVIg, intravenous immunoglobulin; JAKis, Janus kinase inhibitor; LASSO, minimum absolute contraction and selection operator; LDH, lactate dehydrogenase; MMF, mycophenolate mofetil; NLR, neutrophil-to-lymphocyte ratio.

[0067] The predictive performance of the CPH-based risk model was further evaluated using time-dependent AUC analysis. Training queue and internal validation queue The study demonstrated strong discriminative power in predicting mortality rates over 1, 3, 5, and 10 years, with AUC values ​​consistently above 0.75 at all time points. Figure 2 a, 2b). The calibration plot also shows a high degree of agreement between predicted mortality and observed mortality ( Figure 3 a, 3b) This indicates that the model has robust calibration in both cohorts. These results confirm that... CPH model for ASYS patients The researchers provided a reliable and well-calibrated estimate of mortality risk. .

[0068] Example 4: Development of CPH Model Interpretation and Scoring System 1. Establishing an ALARN scoring system based on the CPH model After determining the optimal model based on performance comparisons in the internal validation cohort, we focused on its clinical applicability. To simplify the CPH model into a score-based system, we proportionally assigned risk scores based on the individual β coefficients (regression coefficients from Cox multivariate analysis) of each predictor variable. The CPH model scores were calculated using the methods described previously [33-35]. The reference category (assigned 0 points) for each predictor variable was defined as the clinical state associated with the lowest risk of death. To obtain intuitive integer scores for clinical use, we scaled the β coefficients so that the smallest coefficient value was 1; the remaining coefficients were proportionally scaled and then rounded to the nearest integer. The total risk score was calculated by summing the scores of all existing predictors.

[0069] To enhance clinical applicability, we simplified the CPH model into a score-based system called the ALARN score. For each predictor, the reference class (assigned 0 points) was defined as the clinical state associated with the lowest risk of death. Since the smallest β coefficient among the five independent predictors is 0.481 of the NLR, we multiplied each β coefficient by 2 and then rounded the product to the nearest integer to obtain a score for each predictor. Figure 4 (a, Table 4). Variables with the smallest β coefficient are rounded to 1. The total risk score, calculated by summing the weighted values ​​of all identified prognostic variables, ranges from 0 to 8, with a higher total score indicating a greater risk of death.

[0070] Table 4. Thresholds and scores for each predictor (0 points are awarded if the threshold is not met).

[0071] Figure 4 b shows that in the derived cohort, ALARN score was negatively correlated with survival at 1, 3, and 5 years of follow-up, with survival probability decreasing as the score increased. At 10 years of follow-up, this trend persisted in patients with scores of 0–5; however, patients with scores of 6–8 did not show a declining trend, although their survival was still significantly lower compared to those with scores of 0–5. In assessing the discriminative power of ALARN score in the derived cohort, we found a C-index of 0.826 (95% CI: 0.791–0.860). The time-dependent AUCs for predicting 1-, 3-, 5-, and 10-year mortality were 0.914 (95% CI: 0.866–0.955), 0.867 (95% CI: 0.829–0.906), 0.839 (95% CI: 0.789–0.885), and 0.854 (95% CI: 0.789–0.909), respectively.

[0072] Using X-tile software

[36] , patients were divided into three mortality risk groups: Low-risk group (ALARN score 0–2), intermediate-risk group Group 3 (ALARN score 3–5) and high-risk group (ALARN score 6–8) Kaplan-Meier analysis shows that in both the training and internal validation queues, There were significant differences in survival rates among the low, medium, and high-risk groups (all log-rank p < 0.0001); Figure 4 c, 4d) Pairwise comparisons further confirmed the significant differences in survival rates among all risk groups (log-rank p < 0.001).

[0073] Specifically, (1) 1-year survival rate: 100% for low-risk patients, 96-98% for intermediate-risk patients, and 59-93% for high-risk patients; (1) 3-year survival rate: 99-100% for low-risk patients, 89-93% for intermediate-risk patients, and 55-80% for high-risk patients; (1) 5-year survival rate: 97-99% for low-risk patients, 81-88% for intermediate-risk patients, and 47-67% for high-risk patients; (2) 10-year survival rate: 76-96% for low-risk patients, 56-79% for intermediate-risk patients, and 9-24% for high-risk patients.

[0074] 2. Internal validation of the ALARN model To assess the robustness of the risk groups defined by the ALARN score, we conducted a comprehensive subgroup analysis using different partitioning methods in the derived cohort.

[0075] First, we performed a resampling method with five-fold cross-validation for 200 iterations in the derivation queue, and the time dependency AUCs for 1 year, 3 years, 5 years and 10 years were 0.916 (95% CI: 0.893-0.93), 0.867 (95% CI: 0.856-0.878), 0.839 (95% CI: 0.826-0.85) and 0.854 (95% CI: 0.841-0.87), respectively.

[0076] Secondly, the derived cohort was divided into early (admitted before January 2020, n=382) and recent (admitted after January 2020, n=381) sub-cohorts based on admission time. The C-index for the difference in survival probability between scores in the early and recent sub-cohorts was 0.827 (95% CI: 0.786–0.869) and 0.827 (95% CI: 0.772–0.882), respectively. The ALARN score maintained a significant ability to distinguish mortality in both cohorts (p<0.0001).

[0077] 3. External validation of the ALARN model Finally, the ALARN score was validated in an external cohort of 431 patients from two independent institutions (Nanjing and Zhengzhou). In the external validation cohort, the time-dependent AUCs of the ALARN score at 1, 3, 5, and 10 years were 0.912 (95% CI: 0.835–0.989), 0.857 (95% CI: 0.787–0.927), 0.854 (95% CI: 0.779–0.928), and 0.638 (95% CI: 0.420–0.855), respectively. Figure 5 a). The calibration plot further confirmed good agreement between predicted and observed outcomes at 1, 3, and 5 years. Figure 5 b). However, we note that the AUC and calibration accuracy at 10 years are relatively low, which may be attributed to the limited sample size at this extended time point. Furthermore, Kaplan-Meier analysis shows that... ALARN The scoring system effectively categorized patients in the external validation cohort into low-, intermediate-, and high-risk subgroups, with significant differences in mortality rates among these groups. Difference (log-rank p < 0.0001); Figure 6 ) .

[0078] Finally, it should be noted that the above embodiments are only used to help those skilled in the art understand the essence of the present invention, and are not intended to limit the scope of protection of the present invention.

Claims

1. The application of a scoring model in the preparation of a diagnostic kit for predicting the prognosis of patients with antisynthetic enzyme syndrome (ASyS), wherein the scoring model includes the following 5 parameters: Age of onset, serum lactate dehydrogenase concentration, serum albumin concentration, whole blood neutrophil to lymphocyte ratio, respiratory failure; The respiratory failure is defined as: the presence of obvious respiratory symptoms, and arterial blood gas analysis showing PaO2 ≤ 60 mmHg under anaerobic support, or PaO2 / inhaled oxygen concentration (FiO2) ≤ 300 mmHg with assisted oxygen supply; The respiratory symptoms mentioned include, but are not limited to, cough and difficulty breathing.

2. The application according to claim 1, characterized in that, In the aforementioned model, If the age of onset is >55 years old, score 3 points; otherwise, score 0 points. If lactate dehydrogenase > 299 U / L, score 1 point; otherwise, score 0 points. If albumin level is <34.8 g / L, score 1 point; otherwise, score 0 points. If respiratory failure occurs, score 2 points; otherwise, score 0 points. A neutrophil-to-lymphocyte ratio > 5.6 is scored as 1 point; otherwise, it is scored as 0 points. The total score of the above 5 parameters is 0–2 for low risk, 3–5 for medium risk, and 6–8 for high risk.

3. The application according to claim 1 or 2, characterized in that, The prediction of prognosis for patients with antisynthetic enzyme syndrome (ASyS) refers to the prediction of the patient's survival rate within 1-10 years after the onset of the disease.

4. The application according to claim 3, characterized in that, The survival rate within 1-10 years is: (1) 1-year survival rate: 100% for low-risk patients, 96-98% for intermediate-risk patients, and 59-93% for high-risk patients; (1) 3-year survival rate: 99-100% for low-risk patients, 89-93% for intermediate-risk patients, and 55-80% for high-risk patients; (1) 5-year survival rate: 97-99% for low-risk patients, 81-88% for intermediate-risk patients, and 47-67% for high-risk patients; (2) 10-year survival rate: 76-96% for low-risk patients, 56-79% for intermediate-risk patients, and 9-24% for high-risk patients.

5. A diagnostic kit and / or device for predicting the prognosis of patients with antisynthetic enzyme syndrome (ASyS), characterized in that, The kit and / or testing device include: Reagents for detecting serum lactate dehydrogenase concentration, serum albumin concentration, and whole blood neutrophil-to-lymphocyte ratio; and Reagents and / or equipment for detecting respiratory failure; The respiratory failure is defined as: the presence of obvious respiratory symptoms and arterial blood gas analysis showing PaO2 ≤60 mmHg under anaerobic support, or PaO2 / inhaled oxygen concentration (FiO2) ≤300 mmHg with assisted oxygen supply; The respiratory symptoms mentioned include, but are not limited to, cough and difficulty breathing.

6. The detection kit and / or detection device according to claim 5, characterized in that, On the prognostic testing day, serum lactate dehydrogenase concentration, serum albumin concentration, whole blood neutrophil to lymphocyte ratio, and respiratory failure status were also measured.

7. The detection kit and / or detection device according to claim 5 or 6, characterized in that, The aforementioned detection kit and / or detection device provides a prognostic score based on the detection values ​​of serum lactate dehydrogenase concentration, serum albumin concentration, whole blood neutrophil to lymphocyte ratio, patient respiratory failure status, and age of onset. Specifically... If the age of onset is >55 years old, score 3 points; otherwise, score 0 points. If lactate dehydrogenase > 299 U / L, score 1 point; otherwise, score 0 points. If albumin level is <34.8 g / L, score 1 point; otherwise, score 0 points. If respiratory failure occurs, score 2 points; otherwise, score 0 points. A neutrophil-to-lymphocyte ratio > 5.6 is scored as 1 point; otherwise, it is scored as 0 points. The total score of the above 5 parameters is 0–2 for low risk, 3–5 for medium risk, and 6–8 for high risk.

8. The detection kit and / or detection device according to any one of claims 5-7, characterized in that, The prediction of prognosis for patients with antisynthetic enzyme syndrome (ASyS) refers to the prediction of the patient's survival rate within 1-10 years after the onset of the disease.

9. The detection kit and / or detection device according to claim 8, characterized in that, The survival rate within 1-10 years is: (1) 1-year survival rate: 100% for low-risk patients, 96-98% for intermediate-risk patients, and 59-93% for high-risk patients; (1) 3-year survival rate: 99-100% for low-risk patients, 89-93% for intermediate-risk patients, and 55-80% for high-risk patients; (1) 5-year survival rate: 97-99% for low-risk patients, 81-88% for intermediate-risk patients, and 47-67% for high-risk patients; (2) 10-year survival rate: 76-96% for low-risk patients, 56-79% for intermediate-risk patients, and 9-24% for high-risk patients.

Images