Mortality Risk Prediction in Amyopathic Dermatomyositis Associated With Interstitial Lung Disease
2020; Elsevier BV; Volume: 158; Issue: 4 Linguagem: Inglês
10.1016/j.chest.2020.04.057
ISSN1931-3543
AutoresXin‐Yue Lian, Jing Zou, Qiang Guo, Sheng Chen, Liangjing Lu, Ran Wang, Mengmeng Zhou, Qiong Fu, Yan Ye, Chunde Bao,
Tópico(s)Muscle Physiology and Disorders
ResumoBackgroundThe prognosis of amyopathic dermatomyositis (ADM)-associated interstitial lung disease (ILD) is poor. A mortality risk score model is needed to predict survival in patients with ADM-ILD and to guide clinical treatment.Research QuestionHow to identify patients with ADM-ILD who are at high risk and to predict patient outcome based on a risk stratification model?Study Design and MethodsWe evaluated 207 patients with ADM-ILD in this prospective inception study. We used a multivariable Cox proportional hazards model to identify the independent prognostic risk factors and created a risk score model according to patient data from January 2012 to December 2016. We used the index of prediction accuracy that uses the Brier score to reflect both discrimination and calibration of the model. The model was validated in an independent group of patients from January 2017 to June 2018.ResultsWe developed a combined risk score, the FLAIR score, that included the following values and scores: ferritin (<636 ng/mL, 0; ≥636 ng/mL, 2), lactate dehydrogenase (<355 U/L, 0; ≥355 U/L, 2), antimelanoma differentiation-associated gene 5 antibody (negative, 0; +, 2; ++, 3; +++, 4), high-resolution CT imaging score (<133, 0; ≥133, 3), and rapidly progressive ILD (RPILD) (non-RPILD, 0; RPILD, 2). We divided patients into three risk groups according to the FLAIR score: low, 0 to 4; medium, 5 to 9; and high, 10 to 13. In both discovery and validation cohorts, high-risk patients had significantly higher mortality rates than low- and medium-risk patients (P < .001).InterpretationThe FLAIR risk score model could help to predict survival in patients with ADM-ILD and to guide further clinical research on risk-based treatment. The prognosis of amyopathic dermatomyositis (ADM)-associated interstitial lung disease (ILD) is poor. A mortality risk score model is needed to predict survival in patients with ADM-ILD and to guide clinical treatment. How to identify patients with ADM-ILD who are at high risk and to predict patient outcome based on a risk stratification model? We evaluated 207 patients with ADM-ILD in this prospective inception study. We used a multivariable Cox proportional hazards model to identify the independent prognostic risk factors and created a risk score model according to patient data from January 2012 to December 2016. We used the index of prediction accuracy that uses the Brier score to reflect both discrimination and calibration of the model. The model was validated in an independent group of patients from January 2017 to June 2018. We developed a combined risk score, the FLAIR score, that included the following values and scores: ferritin (<636 ng/mL, 0; ≥636 ng/mL, 2), lactate dehydrogenase (<355 U/L, 0; ≥355 U/L, 2), antimelanoma differentiation-associated gene 5 antibody (negative, 0; +, 2; ++, 3; +++, 4), high-resolution CT imaging score (<133, 0; ≥133, 3), and rapidly progressive ILD (RPILD) (non-RPILD, 0; RPILD, 2). We divided patients into three risk groups according to the FLAIR score: low, 0 to 4; medium, 5 to 9; and high, 10 to 13. In both discovery and validation cohorts, high-risk patients had significantly higher mortality rates than low- and medium-risk patients (P < .001). The FLAIR risk score model could help to predict survival in patients with ADM-ILD and to guide further clinical research on risk-based treatment. FOR EDITORIAL COMMENT, SEE PAGE 1307Amyopathic dermatomyositis (ADM) is a unique spectrum of classic dermatomyositis with typical skin lesions persistent for six months to two years and without clinical muscle weakness.1Dalakas M.C. Inflammatory muscle diseases.N Engl J Med. 2015; 373: 393-394Crossref PubMed Scopus (2) Google Scholar, 2Distad B.J. Amato A.A. Weiss M.D. Weiss. Inflammatory myopathies.Curr Treat Options Neurol. 2011; 13: 119-130Crossref PubMed Scopus (26) Google Scholar, 3Pearson C.M. Polymyositis and dermatomyositis.in: McCarty D.J. Arthritis (and Allied Conditions). 9th ed. Lea and Febiger, Philadelphia1979: 742Google Scholar, 4Euwer R.L. Sontheimer R.D. Amyopathic dermatomyositis (dermatomyositis sine myositis): presentation of six new cases and review of the literature.J Am Acad Dermatol. 1991; 24: 959-966Abstract Full Text PDF PubMed Scopus (272) Google Scholar In patients with ADM, the levels of muscle enzymes such as creatine kinase are normal or slightly elevated with no classic myogenic damage on electroneuromyography.5Fudman E.J. Schnitzer T.J. Schnitzer Dermatomyositis without creatine kinase elevation: a poor prognostic sign.Am J Med. 1986; 80: 329-332Abstract Full Text PDF PubMed Scopus (145) Google Scholar,6Chow S.K. Yeap S.S. Amyopathic dermatomyositis and pulmonary fibrosis.Clin Rheumatol. 2000; 19: 484-485Crossref PubMed Scopus (22) Google Scholar Interstitial lung disease (ILD) is a severe complication, with a reported prevalence of 5% to 65%.7Fathi M. Lundberg I.E. Lundberg. Interstitial lung disease in polymyositis and dermatomyositis.Curr Opin Rheumatol. 2005; 17: 701-706Crossref PubMed Scopus (118) Google Scholar,8Sakamoto N. Ishimoto H. Nakashima S. et al.Clinical features of anti-MDA5 antibody-positive rapidly progressive interstitial lung disease without signs of dermatomyositis.Intern Med. 2019; 58: 837-841Crossref PubMed Scopus (19) Google Scholar Our previous retrospective cohort study showed that ILD occurred in 57% of patients with ADM.9Ye S. Chen X.X. Lu X.Y. et al.Adult clinically amyopathic dermatomyositis with rapid progressive interstitial lung disease: a retrospective cohort study.Clin Rheumatol. 2007; 26: 1647-1654Crossref PubMed Scopus (162) Google Scholar The disease course and severity of ILD is highly heterogeneous,10Fathi M. Lundberg I.E. Tornling G. Pulmonary complications of polymyositis and dermatomyositis.Semin Respir Crit Care Med. 2007; 28: 451-458Crossref PubMed Scopus (82) Google Scholar and some patients with mild ILD remain stable and responsive to treatment, while other patients undergo rapidly progressive ILD (RPILD), which is usually refractory to treatment and leads to poor outcomes.11Sato S. Kuwana M. Clinically amyopathic dermatomyositis.Curr Opin Rheumatol. 2010; 22: 639-643Crossref PubMed Scopus (78) Google Scholar,12Selva-O'Callaghan A. Pinal-Fernandez I. Trallero-Araguás E. et al.Classification and management of adult inflammatory myopathies.Lancet Neurol. 2018; 17: 816-828Abstract Full Text Full Text PDF PubMed Scopus (177) Google Scholar Despite aggressive conventional treatments, >50% of patients with ADM with RPILD die of respiratory failure within 1 year, especially in East Asia.9Ye S. Chen X.X. Lu X.Y. et al.Adult clinically amyopathic dermatomyositis with rapid progressive interstitial lung disease: a retrospective cohort study.Clin Rheumatol. 2007; 26: 1647-1654Crossref PubMed Scopus (162) Google Scholar,13Sato S. Hirakata M. Kuwana M. et al.Autoantibodiestoa140kd polypeptide, CADM-140, in Japanese patients with clinically amyopathic dermatomyositis.Arthritis Rheum. 2005; 52: 1571-1576Crossref PubMed Scopus (495) Google Scholar,14Mukae H. Ishimoto H. Sakamoto N. et al.Clinical differences between interstitial lung disease associated with clinically amyopathic dermatomyositis and classic dermatomyositis.Chest. 2009; 136: 1341-1347Abstract Full Text Full Text PDF PubMed Scopus (158) Google Scholar FOR EDITORIAL COMMENT, SEE PAGE 1307 Previous studies explored baseline parameters for guiding clinical diagnosis and predicting survival in patients with ADM-ILD. Overall high-resolution CT (HRCT) imaging score, anti-melanoma differentiation-associated gene 5 (MDA5) antibody, RPILD, neutrophil lymphocyte ratio (NLR), Krebs von den Lungen-6, soluble IL-2 receptor, IL-6, and ferritin are associated with disease activity and could predict a patient's therapeutic response.15Matsushita T. Mizumaki K. Kano M. et al.Antimelanoma differentiation-associated protein 5 antibody level is a novel tool for monitoring disease activity in rapidly progressive interstitial lung disease with dermatomyositis.Br J Dermatol. 2017; 176: 395-402Crossref PubMed Scopus (95) Google Scholar, 16Yasuda H. Ikeda T. Hamaguchi Y. et al.Clinically amyopathic dermatomyositis with rapidly progressive interstitial pneumonia: the relation between the disease activity and the serum interleukin-6 level.J Dermatol. 2017; 44: 1164-1167Crossref PubMed Scopus (11) Google Scholar, 17Osawa T. Morimoto K. Sasaki Y. et al.The serum ferritin level is associated with the treatment responsivity for rapidly progressive interstitial lung disease with amyopathic dermatomyositis, irrespective of the anti-MDA5 antibody level.Intern Med. 2018; 57: 387-391Crossref PubMed Scopus (11) Google Scholar RPILD and positive anti-MDA5 antibody are independent prognostic factors for the prediction of lung progression and death in patients with ADM. However, the described results were derived from several case reports or studies with small samples. Moreover, the use of a single biomarker to predict a patient's survival would be biased because of disease heterogeneity. Combining biomarkers would consider a patient's comprehensive clinical features and may be more suitable to evaluate disease severity, to predict outcomes, and to guide individualized treatment. Therefore, it is clinically important to explore outcome-relevant parameters and to establish a reliable outcome-predicting tool for ADM-ILD. In the current study, we aimed to identify baseline clinical indicators that could predict the prognosis of patients with ADM-ILD during follow up in the discovery cohort and to establish a simple and practical score model by combining independent predictors and verifying the derived prediction model in a validation cohort. The prospective inception cohort study has been set up since 2012. All the patient with ADM-ILD who were hospitalized at the Department of Rheumatology, Renji Hospital, China, from June 2012 to June 2018 were included and prospectively followed thereafter. One hundred forty-two consecutive patients from June 2012 to December 2016 were included in a discovery cohort to establish the risk score model. We validated the model in an independent group of 65 consecutive patients who were hospitalized from January 2017 to June 2018. The study temporal end point was December 2018. All of the patients with ADM were diagnosed according to Sontheimer's criteria and had no other autoimmune diseases.18Sontheimer R.D. Would a new name hasten the acceptance of amyopathic dermatomyositis (dermatomyositis sine' myositis) as a distinctive subset within the idiopathic inflammatory dermatomyopathies spectrum of clinical illness?.J Am Acad Dermatol. 2002; 46: 626-636Abstract Full Text Full Text PDF PubMed Scopus (355) Google Scholar Extensive examinations were performed to exclude malignancy; we recorded the following clinical manifestations on admission: heliotrope rash, Gottron papules, ulcerative Gottron papules, periungual capillary changes, arthritis, and hoarseness. Laboratory results at baseline were also measured and recorded. Patients were reassessed once a month during the first three months and every three to six months, thereafter. Treatment regimens included mainly glucocorticoids or glucocorticoids combined with oral or IV immunosuppressive agents such as cyclosporine, tacrolimus, cyclophosphamide, or mycophenolate mofetil. In addition, some patients were treated with IV immunoglobulin or the biologic agent, basiliximab.19Zou J. Li T. Huang X. et al.Basiliximab may improve the survival rate of rapidly progressive interstitial pneumonia in patients with clinically amyopathic dermatomyositis with anti-MDA5 antibody.Ann Rheum Dis. 2014; 73: 1591-1593Crossref PubMed Scopus (41) Google Scholar Written informed consent from study participants was obtained, and the study was approved by the Ethics Committee of Renji Hospital (ID: 2013-126), Shanghai, China. The diagnosis of ILD was made according to respiratory symptoms, pulmonary function test results, and HRCT imaging.20Akira M. Hamada H. Sakatani M. et al.CT findings during phase of accelerated deterioration in patients with idiopathic pulmonary fibrosis.AJR Am J Roentgenol. 1997; 168: 79-83Crossref PubMed Scopus (260) Google Scholar In some patients, progressive dyspnea occurs within one month after the onset of respiratory symptoms, and chest radiography shows new interstitial abnormalities. This type of ILD is defined as RPILD.13Sato S. Hirakata M. Kuwana M. et al.Autoantibodiestoa140kd polypeptide, CADM-140, in Japanese patients with clinically amyopathic dermatomyositis.Arthritis Rheum. 2005; 52: 1571-1576Crossref PubMed Scopus (495) Google Scholar Given that our department is a tertiary referral center, patient HRCT images were acquired in the following manner: (1) For those who were admitted and were diagnosed for the first time, we used the first HRCT image as baseline data. (2) For those who had had a CT scan and had been diagnosed as ADM-ILD within three months, if they had not had progressive respiratory manifestations and had stable pulmonary function, we would use his/her previous HRCT scan as baseline data. (3) For those who had worsening respiratory symptoms and therefore been referred to our hospital, we used his/her HRCT images that had been performed within five days after the onset of worsening respiratory symptoms as baseline data. Two independent radiologists evaluated the images in a randomized, single-blind manner, and the final HRCT score for each patient was determined by the average of the score calculated by each radiologist. If there was a significant difference in HRCT score between the two radiologists, the HRCT films were regraded. HRCT imaging scores were evaluated based on the method described by Ichikado et al21Ichikado K. Suga M. Müller N.L. et al.Acute interstitial pneumonia: comparison of high-resolution computed tomography findings between survivors and nonsurvivors.Am J Respir Crit Care Med. 2002; 165: 1551-1556Crossref PubMed Scopus (164) Google Scholar: The lungs were divided into six zones (upper, middle, and lower on both sides), and each zone was assessed separately. The HRCT score was graded on the percentage of abnormal lung parenchyma and was estimated to the nearest 5% of parenchymal involvement. The following formula was use: overall HRCT score (%) = average score for normal attenuation × 1 + average score for ground-glass attenuation without traction bronchiectasis or bronchiolectasis × 2 + average score for consolidation without traction bronchiectasis × 3 + average score for ground-glass attenuation with traction bronchiectasis × 4 + average score for consolidation with traction bronchiectasis × 5 + average score for honeycombing × 6.22Ichikado K. Suga M. Muranaka H. et al.Prediction of prognosis for acute respiratory distress syndrome with thin-section CT: validation in 44 cases.Radiology. 2006; 238: 321-329Crossref PubMed Scopus (144) Google Scholar Myositis-specific autoantibodies and myositis-associated autoantibodies were detected with immunoblot testing (Euroimmun, Lubeck, Germany) based on the manufacturer's instructions. The gray-scale value of the antibody band was scanned to obtain semiquantitative results, and gray-scale values of 0 to 5 units/L were defined in the following manner: 11 to 25 units/L as +, 26 to 50 units/L as ++, and >50 units/L as +++. Double-stranded DNA antibody was measured with a Farr array (Anti-DNA Antibody Radioimmunoassay Kit, Northern Technology, Beijing, China). Antinuclear antibody and extractable nuclear antigen were identified with the Nova Lite Hep-2 ANA kit (Inova Diagnostics, Inc, San Diego, CA) and Quanta Lite ENA 6 kit (Inova Diagnostics Inc). Statistical analysis was performed with the use of the SPSS software package (version 20.0; IBM Corp, Armonk, NY) and the R statistical package (version 3.5.1; R Foundation for Statistical Computing, Vienna, Austria; http://www.r-project.org). We used the χ2 test or Fisher exact test to compare the categoric variables between different groups and the Mann-Whitney U test to compare differences between the two groups for continuous variables. The optimal cutoff value was determined with the use of receiver operating characteristic curve analysis, then each continuous parameter was converted into a classification variable. We used a multivariable Cox proportional hazards model to identify independent prognostic risk factors and calculate their HRs, 95% CI, and β regression coefficients. The decision curve analysis (DCA) was used to verify whether these cutoff values were the optimal threshold of clinical significance. The DCA was performed with the source file "stdca.r," which was downloaded from decisioncurveanalysis.org. To select the optimized model, the Index of Predictive Accuracy (IPA) was applied to evaluate the different risk prediction models by with the use of the IPA function in the "RiskRegression" package in R.23Kattan M.W. Gerds T.A. The index of prediction accuracy: an intuitive measure useful for evaluating risk prediction models.Diagn Progn Res. 2018; 2: 7Crossref PubMed Google Scholar For time-event analysis, we used the Kaplan-Meier method to calculate the cumulative survival rates during follow up and the logarithmic rank test to compare different groups. The predictive accuracy of the risk model was assessed by calculating the Harrell's concordance index in the "Hmisc" package in R and by determining the difference among different risk groups in the probability of survival within one year.24Altman D.G. Royston P. What do we mean by validating a prognostic model?.Stat Med. 2000; 19: 453-473Crossref PubMed Scopus (1131) Google Scholar All statistical analyses were performed in both the discovery and validation groups. For all analyses, statistical significance was calculated with the use of two-tailed probability values; P < .05 was considered significant. In the discovery cohort, the overall mean age of the 142 patients was 50.1 ± 10.9 years. The age in the nonsurvivor group was significantly higher than the survivor group (P = .007). There were more women than men (94 vs 48, respectively), but the difference was not statistically significant (Table 1). Patients with RPILD generally had higher mortality rates compared with patients with non-RPILD (P < .001). HRCT imaging scores in the nonsurvivor group were significantly higher than those in the survivor group (P < .001), and the incidence of hoarseness in the nonsurvival group was higher than in the survival group (P = .034). Patients in the nonsurvival group had higher NLR values and higher ferritin and lactate dehydrogenase (LDH) levels than those in the survival group (P < .001). In the nonsurvival group, the level of anti-MDA5 antibody was higher compared with that in the survival group. Other clinical manifestations and laboratory features showed no significant differences between the two groups. Detailed patient demographic baseline data in the discovery cohort are summarized in Table 1.Table 1Comparison of Patients Clinical Manifestations and Laboratory Features at Baseline Between Survivors and Nonsurvivors in the Discovery and Validation CohortsPatient ParametersDiscovery Cohort (n = 142)Survivor (n = 101)Nonsurvivor (n = 41)P ValueValidation Cohort (n = 65)Survivor (n = 41)Nonsurvivor (n = 24)P ValueP ValueaA comparison of the discovery and validation cohorts showed that there was no heterogeneity between the two groups.Median age, y (range)51 (16-70)50 (22-68)53 (16-70).00752 (21-73)52 (21-73)53 (36-66).965.437Sex (male/female)48/9433/6815/26.43324/4116/258/16.791.753RPILD/non- RPILD, No.64/7830/7134/7< .00125/405/3620/4< .001.450High-resolution CT score, median (range)143 (102-260)132 (102-216)178 (113-260)< .001134 (102-258)121 (106-168)198 (134-258)< .001.140Heliotrope rash, No. (%)117 (83.0)84 (83.2)33 (80.5).80840 (61.5)25 (61.0)15 (62.5)1.000.157Gottron sign, No. (%)131 (92.3)93 (92.1)38 (92.7)1.00061 (93.8)38 (92.7)23 (95.8)1.000.780Ulcerative Gottron, No. (%)35 (24.6)22 (21.8)13 (31.7).28213 (20)8 (19.5)5 (20.8)1.0001.000Arthritis, No. (%)37 (26.1)28 (27.7)9 (22.0).53319 (29.2)14 (34.1)5 (20.8).397.736Hoarseness, No. (%)10 (7.1)4 (4.0)6 (14.6).0345 (7.7)2 (4.9)3 (12.5).3501.000Tumor, No. (%)2 (1.4)1 (1.0)1 (2.4).4961 (1.5)1 (2.4)0 (0)1.0001.000NLR, median (range), L5.4 (0.4-50.3)4.4 (1.1-50.3)8.7 (0.4-46.6)< .0015.8 (2.1-22.8)5.5 (2.1-22.2)8.5 (2.2-22.8).005.240CRP, median (range), mg/L3.5 (0.2-98.6)3.5 (0.2-52.1)3.4 (2.6-98.6).4413.1 (0.3-162.6)3.1 (0.3-45.2)3.3 (0.9-162.6).184.387ESR, median (range), mm/hr29 (5-95)27 (5-95)35 (7-95).06132 (5-92)32 (9-92)37 (5-82).322.226Ferritin, median (range), ng/mL700.3 (6.1-2,270)560 (6.1-2,200)1,207 (173.6-2,270)< .0011,002.4 (15.9-1,500)862.7 (15.9-1,500)1,117.3 (29.7-1,500).012.304Albumin, mean ± SD, g/L34.2 ± 4.935.5 ± 4.630.9 ± 4.3.06731.8 ± 4.932.6 ± 4.930.3 ± 4.5.0571.000ALT, median (range), units/L42 (10-791)35 (10-400)59 (18-791).06553 (5-420)54 (5-420)49.5 (9-386).992.748AST, median (range), units/L42 (12-1,044)31 (12-396)62 (21-1,044).07342 (9-279)38 (9-225)52 (15-279).107.755CK, median (range), units/L51 (10-810)51 (10-506)52 (16-810).39146 (14-340)44 (14-340)51 (15-316).654.386LDH, median (range), units/L275 (128-791)259 (128-586)380 (181-791)< .001313 (142-1,110)246 (142-507)359 (181-1,110).002.276SCr, median (range), μmol/L46 (22-94)47.9 (26-94)44 (22-65).07346 (27-122)44 (27-79)48 (29-122).663.872BUN, median (range), mmol/L5.1 (2.1-35)4.97 (2.1-35)5.6 (2.6-10.2).0605.4 (1.9-10.3)5 (2.8-9.6)5.8 (1.9-10.3).718.185BNP, median (range), pg/mL27.5 (11-637)27 (11-317)37 (25-637).11832 (5-319)28 (11-266)40 (5-319).443.331FVC%, mean ± SD, L (btps)53.0 ± 11.259.1 ± 6.249.0 ± 8.6.08453.9 ± 18.254.3 ± 9.649.8 ± 15.1.065.741DLCO%, mean ± SD, mL/min/mm Hg36.1 ± 13.640.4 ± 14.231.0 ± 10.7.07841.2 ± 15.243.5 ± 13.936.6 ± 8.4.061.592ANA (≥1:80), No. (%)48 (33.8)35 (34.7)13 (31.7).84628 (43.1)17 (41.5)11 (45.8).798.495Anti-MDA5 antibody, No. (%)< .001.041.226 -36 (25.4)35 (34.6)1 (2.4)9 (13.8)9 (21.9)0 (0) +29 (20.4)23 (22.8)6 (14.6)7 (10.8)4 (9.8)3 (12.5) ++28 (19.7)18 (17.8)10 (24.4)14 (21.5)9 (22.0)5 (20.8) +++49 (34.5)25 (24.8)24 (58.6)35 (53.9)19 (46.3)16 (66.7)ALT = alanine aminotransferase; ANA = antinuclear antibody; AST = aspartate aminotransferase; BNP = brain natriuretic peptide; CK = creatine kinase; CRP = C-reactive protein; Dlco% = estimated diffusing capacity of lung carbon monoxide; ESR = erythrocyte sedimentation rate; FVC% = estimated FVC; LDH = lactate dehydrogenase; MDA5 = melanoma differentiation-associated gene 5; NLR = neutrophil-lymphocyte ratio; RPILD = rapidly progressive interstitial lung disease; Scr = serum creatinine concentration.a A comparison of the discovery and validation cohorts showed that there was no heterogeneity between the two groups. Open table in a new tab ALT = alanine aminotransferase; ANA = antinuclear antibody; AST = aspartate aminotransferase; BNP = brain natriuretic peptide; CK = creatine kinase; CRP = C-reactive protein; Dlco% = estimated diffusing capacity of lung carbon monoxide; ESR = erythrocyte sedimentation rate; FVC% = estimated FVC; LDH = lactate dehydrogenase; MDA5 = melanoma differentiation-associated gene 5; NLR = neutrophil-lymphocyte ratio; RPILD = rapidly progressive interstitial lung disease; Scr = serum creatinine concentration. Similar to the discovery cohort, ILD, HRCT scores and NLR, ferritin, LDH, and anti-MDA5-antibody values showed distinct differences in the validation cohort. Patients in the nonsurvivor group had a higher proportion of RPILD than in the survivor group, and patients in the nonsurvivor group had higher values in HRCT scores, NLR, ferritin, LDH, and anti-MDA5-antibody as compared with those in the survivor group. However, there was no significant difference for age between the survivor and nonsurvivor groups. Furthermore, there were no differences in patients baseline demographics and laboratory values between discovery and validation cohorts (Table 1). Patient baseline clinical and laboratory findings are shown in Table 1. There was no difference regarding the treatment regimens between survivors and nonsurvivors in both the discovery and validation cohorts (e-Table 1). According to the cutoff value, which was determined by receiver operating characteristic curve analysis, each continuous parameter was converted into a classification variable (e-Fig 1). Univariate and multivariable Cox analyses were then used to explore the independent risk factors for prognosis in patients with ADM-ILD. Variables with P < .05 in the univariate analysis were included in the multivariable cox analysis as the following covariates: age (<45 vs ≥45 years), ILD (RPILD vs non-RPILD), hoarseness (yes vs no), HRCT imaging score (<133 vs ≥133), anti-MDA5 antibody (+++ vs ++ vs + vs −), NLR (<6.06 vs ≥6.06), ferritin (<636 ng/mL vs ≥636 ng/mL), and LDH (<355 units/L vs ≥355 units/L). The results of the multivariable Cox analysis indicated that RPILD, HRCT imaging score, and anti-MDA5 antibody and ferritin and LDH levels were independent risk factors for prognosis in patients with ADM-ILD (RPILD: HR, 2.55 [95% CI, 1.04-6.22; P = .040]; HRCT imaging score: HR, 6.24 [95% CI, 1.47-12.56; P = .013]; anti-MDA5 antibody: HR, 1.48 [95% CI, 1.01-2.16; P = .042; ferritin: HR, 2.62 [95% CI, 1.18-5.83; P = .018]; LDH: HR, 3.59 [95% CI, 1.83-7.01; P = .001) (Fig 1). We then selected the five independent risk factors to create a new mortality prediction model, the FLAIR score. The score was calculated based on the five prognostic predictors weighted by regression coefficients, which were rounded into integer values (Tables 2 and 3). The total risk score was calculated by adding the weighted values of all identified prognostic variables. Patients were then divided into three risk groups according to their FLAIR score: low risk, 0 to 4; medium risk, 5 to 9, and high risk, 10 to 13 (e-Fig 2). Then, DCA was used to assess whether selected cutoffs were the optimal thresholds of clinical significance. The clinical value of each cutoff was assessed by clinical net benefit. The results showed that net benefits of ferritin (636 ng/mL), LDH (355 units/L), and HRCT score (133) were the highest in the DCA (e-Fig 1). Therefore, we confirmed ferritin (636 ng/mL), LDH (355 units/L), and HRCT score (133) to be the clinical optimal thresholds. For the one-year prediction horizon, the Brier scores of the final model (LDH + HRCT score + RPILD + ferritin + MDA5) of the discovery and validation cohorts were 10.6 and 6.7, respectively. These corresponded to IPA values of 47.0% and 69.3%. Table 4 shows the changes in Brier score and IPA for the one-year prediction horizon when each of the five variables was dropped from the model. These results indicated that the existence of these five variables in the final model was beneficial to improve the prediction accuracy of the FLAIR model.Table 2Calculation of the FLAIR Score for Risk Stratification in the Discovery and Validation CohortsVariableHR (95% CI)P Valueβ CoefficientScoreILD< .001 Non-RPILDReference group… RPILD7.5 (3.318-16.949)2.0152HRCT score< .001 <133Reference group… ≥13314.826 (3.576-61.459)2.6963Anti-MDA5-antibody.006 -Reference group… +8.640 (1.040-71.776).0462.1562 ++15.318 (1.958-119.824).0092.7293 +++22.136 (2.992-163.77).0023.6974Ferritin< .001 <636 ng/mLReference group… ≥636 ng/mL4.73 (2.18-10.264)1.5542LDH< .001 <355 units/LReference group… ≥355 units/L6.533 (3.469-12.302)1.8772See Table 1 legend for expansion of abbreviations. Open table in a new tab Table 3Prognostic Score System for Patients With Amyopathic Dermatomyositis-Interstitial Lung DiseaseVariableClinical Evaluation ParametersScoreaOverall score: 0 to 4 low risk; 5 to 9 medium risk; 10 to 13 high risk.ILDNon-RPILD0RPILD2HRCT score<1330≥1333Anti-MDA5 antibody-0+2++3+++4Ferritin<636 ng/mL0≥636 ng/mL2LDH<355 units/L0≥355 units/L2See Table 1 legend for expansion of abbreviations.a Overall score: 0 to 4 low risk; 5 to 9 medium risk; 10 to 13 high risk. Open table in a new tab Table 4Comparison of the Different Models Regarding Overall Survival in the Discovery and Validation CohortsNo. of VariablesVariables Included in the ModelBrier ScoreIPA at 1 y, %IPA DropDiscovery cohort 1 (final model)LDH + HRCT score + RPILD + ferritin + MDA510.6 (4.8-16.8)47.00.0 2LDH + HRCT score + RPILD + ferritin11.2 (4.8-17.6)44.03.0 3LDH + HRCT score + RPILD + MDA511.1 (4.6-17.6)44.12.9 4LDH + HRCT score + ferritin + MDA511.4 (4.9-17.9)42.64.4 5LDH + RPILD + ferritin + MDA511.3 (4.5-18.1)43.23.8 6HRCT score + RPILD + ferritin + MDA512.0 (5.8-18.1)39.97.1Validation cohort 1 (final model)LDH + HRCT score + RPILD + ferritin + MDA56.7 (2.4-9.0)69.30 2LDH + HRCT score + RPILD + ferritin7.3 (2.3-8.9)66.82.5 3LDH + HRCT score + RPILD + MDA57.3 (2.3-9.1)66.92.4 4LDH + HRCT score + ferritin + MDA58.5 (3.4-12.9)61.28.1 5LDH + RPILD + ferritin + MDA59.0 (1.2-13.5)59.110.2 6HRCT score + RPILD + ferritin + MDA58.0 (1.7-10.6)63.55.8IPA = index of predictive accuracy.See Table 1 legend for expansion of the other abbreviations. Open table in a new tab See Table 1 legend for expansion of abbreviations. See Table 1 legend for expansion of abbreviations. IPA = index of predictive accuracy. See Table 1 legend for expansion of the other abbreviations. After classification of the discovery cohort according to the risk score, 33.1% of the patients were assigned to the low-risk group; 38.7% were assigned to the medium-risk group, and 28.2% were assigned to the high-risk group. The results of the validation cohort were similar: 24.6% of the patients were in the low-risk group, 41.5% in the medium-risk group, and 33.8% in the high-risk group (Table 5). Th
Referência(s)