Background & aims

The accuracy of non-invasive tests (NITs) should be ≥80% (EASL recommendation). We aimed to compare the accuracies of the recommended NITs for advanced fibrosis in MASLD and improve NIT accuracy.

Methods

1051 MASLD patients were allocated to derivation (n=637) and validation (n=414) sets. The main outcome (Kleiner F3+F4) was primarily evaluated by accuracy. Conventional NITs were FIB-4, Fibrotest, FibroMeter, liver stiffness measurement (LSM by Fibroscan), Elasto-FibroMeter (FibroMeter-LSM combination), and ELF in 396 patients. We used machine-learning-optimized multitargeting to develop new NITs: FIB-9 (including 9 usual biomarkers), FIB-11 (adding 2 specialized blood markers) and FIB-12 (adding LSM).

Results

In the whole population, the accuracies of recommended NITs were insufficient, Fibrotest: 68.0%, FIB-4: 71.2%, FibroMeter: 75.1%, LSM: 75.9%, Elasto-FibroMeter: 78.6%. Therefore, new NITs (FIB-9, FIB-11, FIB-12) were developed in the derivation set. In the validation set, AUROCs were, FIB-4: 0.757, Fibrotest: 0.766, FibroMeter: 0.850, LSM: 0.852, FIB-9: 0.863, FIB-11: 0.880, Elasto-FibroMeter: 0.894, FIB-12: 0.912 (p<0.001). The FIB-12 AUROC was superior to the ELF AUROC (0.906 vs 0.865, p=0.039). Accuracies were, FIB-4: 68.8%, Fibrotest: 68.6%, LSM: 75.4%, FibroMeter: 76.3%, FIB-9: 78.7%, Elasto-FibroMeter: 79.7%, FIB-11: 80.2%, FIB-12: 83.3% (p<0.001 between all NITs). Scores were segmented by ≥90% sensitivity and specificity cut-offs or NIT match, which individualized subgroups with NIT accuracies ≥80%, e.g. for FIB-9: 85.8% in 68.1% of patients using two cut-offs and 83.2% in 71.7% of patients where FIB-9 agreed with FIB-4.

Conclusions

Recommended NITs had accuracies <80% for advanced fibrosis in MASLD. Several NIT segmentations individualized subgroups with accuracies ≥80%. New NITs further improved accuracy. The simple FIB-9 (available via a free calculator) provided accuracy equaling or surpassing recommended NITs. FIB-12 outperformed other NITs.

Impact and implications

Currently recommended non-invasive tests (NITs) have insufficient accuracy (<80%) for the diagnosis of advanced fibrosis in MASLD. Therefore, we developed three new NITs with new statistical techniques. Thus, FIB-9 (available via a free calculator), including nine usual blood markers, equaled the performance of patented NITs. FIB-11, adding two specialized blood markers, and FIB-12, adding liver stiffness, had accuracy >80%. FIB-12, outperformed all other NITs. FIB-9 is suitable for screening and FIB-11 or FIB-12 for diagnosis.

中文翻译：

---

新一代无创肝纤维化检测，提高 MASLD 的准确性

 背景和目标

无创检测 （NIT） 的准确性应为 ≥80%（EASL 推荐）。我们旨在比较推荐的 NIT 对 MASLD 晚期纤维化的准确性，并提高 NIT 准确性。

 方法

1051 例 MASLD 患者被分配到衍生集 （n=637） 和验证集 （n=414）。主要结局 （Kleiner F3+F4） 主要通过准确性进行评估。常规 NIT 为 FIB-4 、 Fibrotest 、 FibroMeter 、 肝脏硬度测量 （LSM by Fibroscan） 、 Elasto-FibroMeter （FibroMeter-LSM 组合） 和 ELF 在 396 例患者中。我们使用机器学习优化的多目标来开发新的 NIT：FIB-9 （包括 9 个常用生物标志物）、FIB-11 （添加 2 个专门的血液标志物）和 FIB-12 （添加 LSM）。

 结果

在整个人群中，推荐的 NIT 的准确性不足，Fibrotest：68.0%，FIB-4：71.2%，FibroMeter：75.1%，LSM：75.9%，Elasto-FibroMeter：78.6%。因此，在推导集中开发了新的 NIT （FIB-9、FIB-11、FIB-12）。在验证集中，AUROC 是，FIB-4：0.757，Fibrotest：0.766，FibroMeter：0.850，LSM：0.852，FIB-9：0.863，FIB-11：0.880，Elasto-FibroMeter：0.894，FIB-12：0.912 （p<0.001）。FIB-12 AUROC 优于 ELF AUROC （0.906 vs 0.865，p=0.039）。精度为，FIB-4：68.8%，纤维测试：68.6%，LSM：75.4%，纤维计：76.3%，FIB-9：78.7%，弹性纤维计：79.7%，FIB-11：80.2%，FIB-12：83.3%（所有 NIT 之间的 p<0.001）。分数按 ≥90% 的敏感性和特异性临界值或 NIT 匹配进行细分，其中将 NIT 准确度≥ 80% 的个体化亚组，例如 FIB-9：使用两个临界值在 68.1% 的患者中为 85.8%，在 FIB-9 与 FIB-4 一致的患者中为 83.2% 在 71.7% 的患者中。

 结论

推荐的 NIT 对 MASLD 晚期纤维化的准确率为 <80%。几个 NIT 细分对亚组进行了个体化处理，准确率为 ≥80%。新的 NIT 进一步提高了准确性。简单的 FIB-9（可通过免费计算器获得）提供等于或超过推荐的 NIT 的精度。FIB-12 的表现优于其他 NIT。

 影响和影响

目前推荐的无创检测 （NIT） 对 MASLD 晚期纤维化的诊断准确性不足 （<80%）。因此，我们开发了三种具有新统计技术的新 NIT。因此，FIB-9（可通过免费计算器获得），包括 9 种常见的血液标志物，与获得专利的 NIT 的性能相当。FIB-11 增加了两个专门的血液标志物，FIB-12 增加了肝脏硬度，准确率 >80%。FIB-12 的表现优于所有其他 NIT。FIB-9 适用于筛查，FIB-11 或 FIB-12 用于诊断。