Diagnosis of liver disease at earlier stages can improve outcomes and reduce the risk of progression to malignancy. Liver biopsy is the gold standard for diagnosis of liver disease, but is invasive and sample acquisition errors are common. Serum biomarkers for liver function and fibrosis, combined with patient factors, may allow for noninvasive detection of liver disease. In this pilot study, we tested and validated the performance of an algorithm that combines GP73 and LG2m serum biomarkers with age and sex (GLAS) to differentiate between patients with liver disease and healthy individuals in two independent cohorts. To develop the algorithm, prototype immunoassays were used to measure GP73 and LG2m in residual serum samples collected between 2003 and 2016 from patients with staged fibrosis and cirrhosis of viral or non-viral etiology (n = 260) and healthy subjects (n = 133). The performance of five predictive models using combinations of age, sex, GP73, and/or LG2m from the development cohort were tested. Residual samples from a separate cohort with liver disease (fibrosis, cirrhosis, or chronic liver disease; n = 395) and healthy subjects (n = 106) were used to validate the best performing model. GP73 and LG2m concentrations were higher in patients with liver disease than healthy controls and higher in those with cirrhosis than fibrosis in both the development and validation cohorts. The best performing model included both GP73 and LG2m plus age and sex (GLAS algorithm), which had an AUC of 0.92 (95% CI: 0.90–0.95), a sensitivity of 88.8%, and a specificity of 75.9%. In the validation cohort, the GLAS algorithm had an estimated an AUC of 0.93 (95% CI: 0.90–0.95), a sensitivity of 91.1%, and a specificity of 80.2%. In both cohorts, the GLAS algorithm had high predictive probability for distinguishing between patients with liver disease versus healthy controls. GP73 and LG2m serum biomarkers, when combined with age and sex (GLAS algorithm), showed high sensitivity and specificity for detection of liver disease in two independent cohorts. The GLAS algorithm will need to be validated and refined in larger cohorts and tested in longitudinal studies for differentiating between stable versus advancing liver disease over time.

中文翻译：

---

验证新型 GLAS 算法有助于根据 GP73、LG2m、年龄和性别检测肝纤维化和肝硬化

早期诊断肝病可以改善预后并降低进展为恶性肿瘤的风险。肝活检是诊断肝病的金标准，但具有侵入性，且样本采集错误很常见。肝功能和纤维化的血清生物标志物与患者因素相结合，可以实现肝病的无创检测。在这项试点研究中，我们测试并验证了一种算法的性能，该算法将 GP73 和 LG2m 血清生物标志物与年龄和性别 (GLAS) 结合起来，以区分两个独立队列中的肝病患者和健康个体。为了开发该算法，使用原型免疫测定法测量了 2003 年至 2016 年间从病毒或非病毒病因的分期纤维化和肝硬化患者 (n = 260) 和健康受试者 (n = 133) 收集的残留血清样本中的 GP73 和 LG2m 。测试了使用来自开发队列的年龄、性别、GP73 和/或 LG2m 组合的五种预测模型的性能。来自患有肝病（纤维化、肝硬化或慢性肝病；n = 395）和健康受试者（n = 106）的单独队列的剩余样本被用来验证最佳表现模型。在开发和验证队列中，肝病患者的 GP73 和 LG2m 浓度高于健康对照，肝硬化患者的浓度高于纤维化患者。表现最佳的模型包括 GP73 和 LG2m 加上年龄和性别（GLAS 算法），其 AUC 为 0.92（95% CI：0.90-0.95），敏感性为 88.8%，特异性为 75.9%。在验证队列中，GLAS 算法的估计 AUC 为 0.93（95% CI：0.90-0.95），敏感性为 91.1%，特异性为 80.2%。在这两个队列中，GLAS 算法在区分肝病患者与健康对照方面具有很高的预测概率。GP73 和 LG2m 血清生物标志物与年龄和性别（GLAS 算法）相结合，在两个独立队列中显示出检测肝病的高敏感性和特异性。GLAS 算法需要在更大的队列中进行验证和完善，并在纵向研究中进行测试，以区分随着时间的推移稳定的肝病与进展的肝病。