Objective: To compare the accuracy of three different formulas for intraocular lens power calculation in high myopic eyes with cataract and analyze their influencial factors. Methods: One hundred and three high myopic patients of cataract (103 eyes), with average age of 60.2±8.8 years old (39.0-77.0), including 45 male and 54 female and with axial length ≥ 26 mm were enrolled in this retrospective case-series study. All of them underwent routine ocular examination and IOLMastermeasurement preoperatively and then underwent phacoemulsification through temporal clear-corneal incision with implantation of HumanOptic posterior chamber Intraocular lens (IOL). All analyses were conducted using SPSS version 19.0. Repeated-measures analysis of variance was applied to compare the refractive results one month postoperatively with the predicted IOL powers calculated by SRK/T, Holladay 1, or Haigis formula before surgery. The differences were further compared based on different grouping of axial length (AXL), corneal curvature (K) and corneal astigmatism (CA). The accuracies of the three formulas were analyzed using Bland-Altman analyses and the possible error sources of each formula were analyzed using multiple regression model. Results: The majority of patients enrolled had hyperopic shift after cataract surgery. The mean errors (ME) of the three formulas were SRK/T: 0.70±0.89D, Holladay 1: (1.20±0.88) D and Haigis: (0.60±0.88) D; the mean absolute errors (MAE) of the three formulas were (0.81±0.79) D, (1.23±0.84) D and (0.76±0.74) D, respectively. Both ME and MAE of Holladay formula were significantly greater than the other two formulas (F=86.31, P<0.01). Besides, the proportion of patients having a prediction error within 0.50 D was lower in those using Holladay formula (20.4%, 21/103) than the other two (SRK/T: 38.8%, 40/103, χ(2)=8.41, P<0.01, Haigis: 45.6%, 47/103, χ(2)=14.84, P<0.01). Bland-Altman analyses showed that the accuracies of all the three formulas were acceptable in patients of cataract with high myopia in clinical practice. ME and MAE tended to be larger with longer axial length, larger corneal curvature and astigmatism of the patients in all three formulas. However, in eyes with axial length> 30 mm or corneal curvature ≤43.00 D, the MAE of Haigis formula was lower than that of SRK/T formula (F=63.26,63.94, both P<0.01). The prediction error of SRK/T formula was positively correlated with axial length and corneal astigmatism (F=33.97, r=0.66, β=0.48, P<0.01 and β=0.42, P<0.01), while for Holladay and Haigis formulas, in addition to the previous two factors, the errors were also positively correlated with mean corneal curvature (Holladay 1: F=31.26, r=0.72, AXL: β=0.52, P<0.01, K: β=0.20, P<0.05 and CA: β=0.37, P<0.01; Haigis: F=30.96, r=0.72, AXL: β=0.33, P<0.01, K: β=0.40, P<0.01 and CA: β=0.37, P<0.01). Conclusions: In the selection of IOL formula for high myopic patients with cataract, Haigis or SRK/T would reduce the prediction error and serve as the more accurate formulas than Holladay 1. Haigis formula may be more accurate than SRK/T formula in case of AXL>30 mm or K≤43.00 D. (Chin J Ophthalmol, 2017, 53: 260-265).
目的： 研究3种IOL计算公式预测高度近视眼并发性白内障患者白内障摘除术后屈光状态的误差比较。 方法： 回顾性系列病例研究。收集复旦大学附属眼耳鼻喉科医院眼科2014年12月至2015年4月的高度近视眼并发性白内障患者的临床资料，共103例（103只眼），眼轴长度均≥26 mm，年龄39.0~77.0岁，平均（60.2±8.8）岁；男性45例；女性58例。所有患者均行常规术前检查以及IOLMaster测量，采用颞侧透明角膜切口超声乳化白内障吸除并植入HumanOptics后房型IOL。使用SPSS19.0软件进行统计学分析，采用重复测量的方差分析比较术后1个月实际屈光度数与应用SRK/T，Holladay 1及Haigis公式计算所植入IOL的理论屈光度数的差别，以及按照眼轴长度、角膜屈光力（K值）及角膜散光度数大小分组比较，并用Bland-Altman图分析3种公式的预测一致性。后用多元线性回归分析各公式的误差来源。 结果： 3种计算公式预测中绝大多数患者均发生了远视漂移现象。SRK/T、Holladay 1及Haigis公式平均误差分别为（0.70±0.89）D、（1.20±0.88）D及（0.60±0.88）D，平均绝对误差分别为（0.81±0.79）D、（1.23±0.84）D、（0.76±0.74）D。Holladay 1公式的平均误差及绝对误差均远高于其他两者，差异有统计学意义（F=86.31，P<0.01）。Holladay 1公式误差在±0.50 D范围的患者比例（20.4%，21/103）也低于SRK/T公式（38.8%，40/103）及Haigis公式（45.6%，47/103），差异有统计学意义（χ(2)=8.41，P<0.01；χ(2)=14.84，P<0.01）。Bland-Altman分析显示3种公式在长眼轴患者中预测一致性均在临床可接受范围。随眼轴长度增加，K值增大及角膜散光增大，平均误差及绝对误差在3种公式中均增大，而在眼轴长度>30 mm及K值≤43.00 D时Haigis公式的平均绝对误差小于SRK/T公式，差异有统计学意义（F=63.26、63.94，P均<0.01）。在误差来源方面，SRK/T公式误差与眼轴长度及角膜散光度数正相关（F=33.97，r=0.66；β=0.48，P<0.01；β=0.42，P<0.01），而Holladay 1及Haigis公式误差除外这两者还与K值正相关（Holladay 1：F=31.26，r=0.72，眼轴长度：β=0.52，P<0.01，K值：β=0.20，P<0.05及角膜散光：β=0.37，P<0.01；Haigis：F=30.96，r=0.72，眼轴长度：β=0.33，P<0.01，K值：β=0.40，P<0.01及角膜散光：β=0.37，P<0.01）。 结论： Haigis及SRK/T公式较Holladay 1公式在高度近视眼并发性白内障术后屈光度数预测中准确性较高，而在眼轴长度>30 mm及K值≤43.00 D时Haigis公式可能更为准确。（中华眼科杂志，2017，53：260-265）.
Keywords: Algorithms; Cataract; Forecasting; Lenses, intraocular; Myopia.