PURPOSE To identify predictive factors for visual field (VF) fluctuation in glaucoma patients. DESIGN Retrospective cohort study. PARTICIPANTS A total of 1392 eyes (816 patients) with 6 or more VFs and 3 years or more of follow-up. METHODS For each eye, the VF mean deviation (MD) and the pointwise sensitivities were regressed against time to model the series trend, and the root mean square error (RMSE) was estimated as a measure of variability. Potential predictors were selected with least absolute shrinkage and selection operator regression and included eye laterality, ethnicity, glaucoma type, intraocular pressure (IOP) fluctuation, baseline best corrected-visual acuity, intervening cataract or glaucoma surgery, length of follow-up, frequency of testing, baseline MD, rates of VF progression, and median false positive (FP) and false negative (FN) responses. MAIN OUTCOME MEASURES Predictors of global and pointwise VF long-term fluctuation. RESULTS In the global model, left eye (0.063 dB; P = 0.022), Asian descent (0.265 dB; P = 0.006), larger IOP fluctuation (0.051 dB; P < 0.001), intervening cataract surgery (0.090 dB; P = 0.023), longer follow-up (0.130 dB; P < 0.001), worse baseline MD (-0.145 dB; P < 0.001), faster VF decay rate (-0.090 dB; P < 0.001), and higher FP rate (0.145 dB; P < 0.001) and FN rate (0.220 dB; P < 0.001) were predictors of VF fluctuation. In the pointwise model, larger IOP fluctuation (0.039 dB; P = 0.022), longer follow-up (0.340 dB; P < 0.001), higher VF frequency (0.238 dB; P = 0.002), intervening glaucoma surgery (0.190 dB; P = 0.01), worse baseline MD (-0.535 dB; P < 0.001), faster VF decay rate (-0.340 dB; P < 0.001), and higher FP rate (0.255 dB; P < 0.001) and FN rate (0.395 dB; P < 0.001) were associated with increased fluctuation. The multivariable model explained 57% and 28% of the pointwise and global variances, respectively. CONCLUSIONS This study identified novel predictors of VF fluctuation, and explains nearly 60% of the pointwise variance. In the presence of factors predictive of high fluctuation, increased frequency of testing and better analytics will help to identify VF progression more accurately.

中文翻译：

---

青光眼患者长期视野波动的预测因素。

目的确定青光眼患者视野（VF）波动的预测因素。设计回顾性队列研究。参与者共有1392眼（816例患者），其VF≥6，且随访时间≥3年。方法对于每只眼睛，将VF平均偏差（MD）和点敏感度随时间进行回归，以模拟级数趋势，并估计均方根误差（RMSE）作为变异性的量度。选择了具有最小绝对收缩和选择算子回归最小的潜在预测因素，包括眼球偏侧性，种族，青光眼类型，眼内压（IOP）波动，基线最佳矫正视力，介入性白内障或青光眼手术，随访时间，测试，基线MD，VF进展率，以及假阳性（FP）和假阴性（FN）响应的中位数。主要观察指标总体和逐点VF长期波动的预测指标。结果在整体模型中，左眼（0.063 dB; P = 0.022），亚洲人后裔（0.265 dB; P = 0.006），IOP波动较大（0.051 dB; P <0.001），白内障介入手术（0.090 dB; P = 0.023） ），更长的随访时间（0.130 dB; P <0.001），较差的基线MD（-0.145 dB; P <0.001），更快的VF衰减率（-0.090 dB; P <0.001）和更高的FP率（0.145 dB; P <0.001）和FN率（0.220 dB； P <0.001）是VF波动的预测指标。在点状模型中，较大的IOP波动（0.039 dB； P = 0.022），较长的随访（0.340 dB； P <0.001），较高的VF频率（0.238 dB； P = 0.002），青光眼介入治疗（0.190 dB； P = 0.01），更差的基线MD（-0.535 dB； P <0.001），更快的VF衰减率（-0.340 dB; P <0.001），更高的FP率（0.255 dB; P <0.001）和FN率（0.395 dB; P <0.001）与波动增加相关。多变量模型分别解释了点状和全局方差的57％和28％。结论本研究确定了VF波动的新型预测因子，并解释了近60％的点状方差。在存在可预测高波动的因素的情况下，增加的检测频率和更好的分析将有助于更准确地确定室颤进展。结论本研究确定了VF波动的新型预测因子，并解释了近60％的点状方差。在存在可预测高波动的因素的情况下，增加的检测频率和更好的分析将有助于更准确地确定室颤进展。结论本研究确定了VF波动的新型预测因子，并解释了近60％的点状方差。在存在可预测高波动的因素的情况下，增加的检测频率和更好的分析将有助于更准确地确定室颤进展。