Supplementary Material for: Accuracy of intraocular lens power calculation formulas in cataract eyes after vitrectomy with silicone oil tamponade：a network meta-analysis.

Objective: To systematically compare the accuracy of artificial intelligence and traditional intraocular lens (IOL) calculation formulas in cataract eyes after vitrectomy with silicone oil tamponade.
Methods: PubMed, Cochrane, Embase, and Web of Science databases were searched to select relevant studies published up to October 29, 2025. 12 formulas (SRK/T, Hoffer Q, Hoffer QST, Haigis, Barrett Universal II, Holladay I, Holladay II, RBF, LSF, EVO, Kane, and Pearl-DGS) were included for final comparison. The primary outcomes were the percentage of eyes with prediction errors (PE) within ±0.50D and ±1.00D.
Results: This meta-analysis included 7 studies comprising 1060 eyes. Overall, compared to traditional formulas, newer formulas demonstrated relatively higher accuracy. The results indicated that among all the formulas being compared, the optimal formula was Kane. For the percentage of eyes with PE within ±0.50 D, Kane had the highest precision. For the IOL power percentage within ±1.00 D, EVO performed the best, with Kane also ranking highly, Furthermore, RBF also performed very well.
Conclusion: Kane and EVO appear to be better choices for silicone oil-filled eyes after vitrectomy.

研究目的：系统比较人工智能与传统人工晶状体（intraocular lens, IOL）计算公式在硅油填充玻璃体切割术后的白内障眼的准确性。
研究方法：检索PubMed、Cochrane、Embase及Web of Science数据库，筛选截至2025年10月29日发表的相关研究。最终纳入12种人工晶状体计算公式（SRK/T、Hoffer Q、Hoffer QST、Haigis、Barrett Universal II、Holladay I、Holladay II、RBF、LSF、EVO、Kane及Pearl-DGS）进行对比分析。主要结局指标为预测误差（prediction error, PE）在±0.50D及±1.00D范围内的眼数占比。
研究结果：本荟萃分析共纳入7项研究，涉及1060只术眼。总体而言，相较于传统公式，新型人工晶状体计算公式展现出相对更高的准确性。结果显示，在所有对比的公式中，最优公式为Kane。在预测误差处于±0.50D范围内的眼数占比方面，Kane的精确度最高；在人工晶状体度数误差处于±1.00D范围内的眼数占比方面，EVO表现最佳，Kane同样位列前茅，此外RBF的表现亦十分优异。
研究结论：对于玻璃体切割术后硅油填充眼，Kane与EVO或为更优选择。