Supplementary Material for: Risk of Ovarian Malignancy Algorithm versus Risk Malignancy Index-I for Preoperative Assessment of Adnexal Masses: A Systematic Review and Meta-Analysis

Purpose: To perform a systematic review and meta-analysis of studies comparing the diagnostic accuracy of Risk of Ovarian Malignancy Algorithm (ROMA) and risk of malignancy index (RMI) for detecting ovarian cancer. Methods: A systematic review and meta-analysis was performed according to PRISMA statement. A search for studies evaluating the diagnostic performance of ROMA and RMI-I indices for detecting ovarian malignancy from January 2010 to October 2018 was performed in the PubMed/MEDLINE and Web of Science databases. The quality of the studies was evaluated by the Quality Assessment of Diagnostic Accuracy Studies 2. Results: Sixty-six citations were identified. After exclusions, 8 papers comprising 2,662 women (1,319 premenopausal and 1,343 postmenopausal) were ultimately included. The mean prevalence of ovarian malignancy was 29.0% in premenopausal women and 51.0% in postmenopausal women. High risk of bias for patient selection was observed for most studies. ROMA and RMI-I had a similar diagnostic performance in postmenopausal women (pooled sensitivity [87 vs. 77%] and specificity [75 vs. 85%], respectively. p = 0.29). In premenopausal women, RMI-I showed better specificity than ROMA (89 vs. 78%, p = 0.022) with similar sensitivity (73 vs. 80%, p= 0.27). Significant heterogeneity was found for sensitivity and specificity in comparisons of both groups. Conclusions: ROMA and RMI-I have similar diagnostic performance for detecting ovarian cancer in women presenting with an adnexal mass. However, RMI-I showed a higher specificity than ROMA in premenopausal women. Notwithstanding, as the risk of bias is high in most studies, our results should be interpreted with caution.

**研究目的**：本研究旨在针对比较卵巢恶性肿瘤风险算法（Risk of Ovarian Malignancy Algorithm, ROMA）与恶性肿瘤风险指数（risk of malignancy index, RMI）诊断卵巢癌准确性的相关研究，开展系统评价与Meta分析。
**研究方法**：本研究遵循《系统评价和Meta分析优先报告条目》（PRISMA）规范开展。我们于2010年1月至2018年10月期间，在PubMed/MEDLINE及Web of Science数据库中，检索评估ROMA与RMI-I指数诊断卵巢恶性病变性能的相关研究。采用诊断准确性研究质量评价工具2（QUADAS-2）对纳入研究的方法学质量进行评价。
**研究结果**：初检出66篇文献，经筛选排除后最终纳入8项研究，共纳入2662名女性患者（其中绝经前1319例，绝经后1343例）。绝经前女性卵巢恶性病变的平均患病率为29.0%，绝经后女性为51.0%。多数研究在患者选择环节存在高偏倚风险。在绝经后女性中，ROMA与RMI-I的诊断性能相当：合并灵敏度分别为87%与77%，合并特异度分别为75%与85%（p=0.29）。在绝经前女性中，RMI-I的特异度显著高于ROMA（89% vs. 78%，p=0.022），二者的灵敏度则无显著差异（73% vs. 80%，p=0.27）。两组分析的灵敏度与特异度均存在显著异质性。
**研究结论**：对于因附件包块就诊的女性，ROMA与RMI-I在卵巢癌检测中的诊断性能相当。但在绝经前女性中，RMI-I的特异度高于ROMA。尽管如此，由于多数研究存在较高偏倚风险，本研究结果需谨慎解读。