在Leiden早期关节炎人群中建立预测放射学进展的RA骨结构破坏的复合生物标记

原文

译文

Development of a Multi-Biomarker Structural Damage Score in Rheumatoid Arthritis to Predict Radiographic Progression in the Leiden Early Arthritis Cohort     Annette H.M. van der Helm-van Mil 1, Rachel Knevel1, William C. Manning2, Lyndal K. Hesterberg2, Guy Cavet2, T.W.J. Huizinga3 and Yijing Shen2, 1Leiden University Medical Center, Leiden, Netherlands, 2Crescendo Bioscience, Inc., South San Francisco, CA, 3Leiden University Medical Centre, Leiden, Netherlands   Presentation Number: 838   Background/Purpose: The ability to predict progressive structural damage has the potential to improve disease management in rheumatoid arthritis (RA) patients.  We aimed to develop a Multi-Biomarker Structural Damage (MBSD) score to predict the risk and quantity of joint damage in an individual patient over 12 months. Method: 28 serum biomarkers representing diverse biological pathways were previously selected from 93 candidates in a series of studies based on their performance at predicting joint damage. Biomarker concentrations were determined in 307 serum samples from 187 patients followed in the Leiden Early Arthritis Cohort. The concentrations of individual biomarkers were assessed for their association with change in total van der Heijde Sharp Scores (SHS) over 12 month intervals by Spearman correlation.. Multivariate statistical models were built using linear regression to predict both the rate of change in SHS (ΔSHS) and risk of progression (RP) using combinations of biomarkers and conventional variables. Independent contributions of variables to modeling were assessed in ordinary least squares (OLS) regression. Model performance for ΔSHS was evaluated in Leave One Out cross-validation by the area under the receiver operating characteristic curve (AUROC). Model performance for RP models was evaluated by specificity, sensitivity, total classification error and AUROC for classifying ≤0 versus > 0 ΔSHS. Result: At False Discovery Rate (FDR) < 0.05, 15 biomarkers were correlated with ΔSHS. Prototype MBSD models using serum markers had the highest observed performance among clinical measures with an AUROC of 0.73.The performance values of clinical measures that were significantly correlated with ΔSHS were: initial erosion score (AUROC=0.70), initial SHS (AUROC=0.67), ESR (AUROC=0.62), initial joint space narrowing score (AUROC=0.6), SJC44 ( AUROC=0.58), CRP (AUROC=0.59) and DAS44 (AUROC=0.57). MBSD had higher observed performance than a combination of clinically available variables (CRP, ESR, SJC44, CCP, Presence of Erosions, Ritchie Articular Index (RAI), Patient Global (PG) and RF; AUROC=0.64). The combination of serum biomarkers and initial erosion score gave slightly higher performance than either alone (AUROC=0.74). RF and CCP titers, PG, age, and RAI were not significantly correlated with ΔSHS. In OLS multivariate regression (with JSN and erosion scores included separately), only two measures were significant independent predictors of ΔSHS: initial erosion score (p<0.001) and MBSD (p=0.002). The sensitivity, specificity and total classification error of MBSD for risk of progression were 0.63, 0.73 and 0.32, respectively. Conclusion: Combinations of serum biomarkers had independent predictive value and higher observed performance than conventional clinical measures at predicting progressive structural damage in RA. An MBSD score generated by serum biomarkers has the potential to improve prediction of radiographic change in clinical practice.

在Leiden早期关节炎人群中建立预测放射学进展的RA骨结构破坏的复合生物标记 Annette H.M. , et al. ACR 2011. Present No: 838 背景/目的:预测进展性结构破坏对于提高类风湿关节炎治疗水平非常重要。我们的目的是建立一个结构破坏的复合生物标记，可用于预测病人个体12个月内关节破坏的风险和程度。 方法: 从一系列研究中有预测关节破坏作用 的93个候选生物标记物，挑选出代表不同生物通路的28个血清标记物。对Leiden早期关节炎人群中187例患者共307份血清标本测定其浓度。Spearman系数分析每个标记物浓度与12个月后总的van der Heide Sharp 积分(ΔSHS)变化的相关性。联合这些生物标记与传统变量应用线性回归建立多元统计模型, 以预测SHS变化速度(ΔSHS)和进展风险(RP)。普通最小二乘(OLS)回归对各独立变量对模型的作用进行评价。模型对 ΔSHS的预测作用通过留一交叉验证法ROC曲线下面积（AUROC）来评估。模型对RP的预测作用通过特异性、灵敏度、总分类误差和AUROC来评价，按照ΔSHS ≤0 和 >0分类。 结果:以假阳性率(FDR)< 0.05为界,15个生物标记物与ΔSHS相关。 应用血清标志物的MBSD原模型，在各项临床测量值中表现最好，获得最高的AUROC 0.73。与ΔSHS显著相关的临床测量指标为：初始侵蚀积分 (AUROC = 0.70),初始SHS(AUROC = 0.67),ESR(AUROC = 0.62)、初始关节间隙狭窄评分(AUROC=0.6), SJC44(AUROC=0.58), CRP(AUROC = 0.59)和DAS44(AUROC = 0.57)。相比临床指标的组合 (CRP,ESR,SJC44,CCP,侵蚀存在,Ritchie 关节指数（RAI），病人的整体评估(PG)和RF;AUROC = 0.64), MBSD的预测作用更优。联合血清生物标记物和初始侵蚀积分可以使预测性能比单独应用略微增高(AUROC = 0.74)。 RF和CCP滴度、PG、年龄、RAI都与ΔSHS 无明显相关性。OLS多元回归中 (分别包括JSN和侵蚀指数),只有两个变量是重要的ΔSHS独立预测因素:最初的侵蚀评分(p < 0.001)和MBSD(p = 0.002)。MBSD的敏感性、特异性和总分类误差风险分别为0.63,0.73和0.32。 结论:血清生物标志物的组合对RA进展性结构破坏有独立的预测价值， 比传统临床指标性能更高。由血清生物标记物衍生出的MBSD积分系统,有望更好地在临床实践中预测影像学变化。