Logistic回归分析判断光学相干断层成像、共焦激光断层扫描和视网膜厚度分析在老年性黄斑变性检查中的价值

Using Logistic regression analysis to value the three technologies during the early diagnosis of AMD

目的 探讨光学相干断层成像（optic coherence tomography，OCT）、共焦激光断层扫描（Heidelberg retinal tomography，HRT）、视网膜厚度分析（retinal thickness analyzer，RTA）检测老年性黄斑变性（age-related macular degeneration，AMD）不同病变时各参数价值，为寻找敏感的指标进行AMD早期诊断提供依据。方法 选取AMD患者共63例97只眼（萎缩型AMD65只眼、渗出型AMD32只眼），年龄52～80岁，其中玻璃膜疣41只眼，地图状萎缩24只眼，渗出型AMD32只眼。进行OCT神经上皮厚度值、HRT-Z轴面信号宽度、容积值、RTA中心凹视网膜平均厚度（foveal average thickneAss，FAT）、中心凹旁视网膜平均厚度（peri-foveal average thickness，PFAT）、后极视网膜平均厚度（posterior-pole average thickness，PPAT）值。以Logistic回归对OCT、HRT、RTA检测结果进行统计学分析。结果（1）以玻璃膜疣组及地图状萎缩组分类，对各参数进行二元单因素Logistic回归检验，发现具有意义的自变量是：OCT神经上皮厚度（B=2．252，P=0．000）；HRT中，HRT容积（B：1．270，P=0．001）、Z信号宽度（13=1．172，P=0．001）；RTA中，PFAT（B=1．959，P：0．000）、FAT（β=1．077，P=0．013）、PPAT（B=0．890，P=0．010）。（2）以玻璃膜疣组及渗出型AMD组分类，对各参数进行二元单因素Logistic回归检验，发现具有意义的自变量是：OCT神经上皮厚度（B=9．882，P=0．003）；HRT中，HRTZ信号宽度（B：1．070，P=0．001），容积对反映此情况已不具有统计学意义（P=0．991）；RTA中，PPAT（B=2．936，P=0．000）、PFAT（B=2．413，P=0．000）、FAT（B=2．250，P=0．001）。（3）经单因素Logistic回归检验后，选取OCT神经上皮厚度、HRT容积、RTA-PFAT及其三者交互作用进入以玻璃膜疣组及地图状萎缩组分类的二元多因素Logistic回归检验，具有意义的自变量依次为OCT（B=1．695，P=0．008）、HRT（p=1．304，P=0．037）及RTA（B=0．936，P=0．141），三者交互作用不具统计学意义（P=0．414）。（4）选取OCT神经上皮厚度、HRT-Z信号宽度、RTA-PPAT及其三者交互作用进入以玻璃膜疣组及渗出型AMD组分类的二元多因素Logistic回归检验，具有意义的自变量依次为OCT（B=16．629，P=0．103）、PPAT（B=5．145，P=0．140）。结论 OCT神经上皮厚度可作为早期发现AMD、鉴别AMD类型分期的工具。HRT-黄斑区容积和RTA-PFAV可作为鉴别玻璃膜疣和萎缩型AMD的指标，HRT-Z轴面信号宽度和RTA-PPAV可作为鉴别玻璃膜疣和渗出型型AMD的指标。

Objective To show the value of diagnosis of AMD in order to supply evidences which index can be used to detect the certain AMD pathological changes and be helpful to illustrate the process of AMD. Methods Select the 63 patients （ total 97 eyes） who were diagnosed for AMD in Tianjin eye hospital from Nov. ,2004 to Apr. ,2005, age from 52 ～ 80 years old, visual activity from hand moving before eyes to 1.0. Then by turns to receive the examinations through OCT/HRT/RTA and use the statistical treatment of χ^2 to survey and evaluate the vary index results. Results （ 1 ） TO distinguish drusen group and dry AMD group with Logistic regression analysis, OCT＇s neural epithelium thickness has the ability to do that （ β = 2. 252, P = 0.000）. HRT＇s macular volume sensitivity（ β = 1. 270, P = 0. 001 ） is better than HRT＇s Z signal width（ β = 1. 172, P =0. 001 ） in HRT. RTA＇s PFAT has the best sensitive（ β = 1. 959, P = 0. 000）, FAT is the second （ β = 1. 077, P = 0. 013 ） and PPAT is the third （ β = 0. 890, P = 0.010） in RTA. （2） To distinguish drusen group and wet AMD group with Logistic regression analysis, OCT＇s neural epithelium thickness has the ability to do that （ β = 9. 882, P=0.003）. HRT＇s Z signal width sensitivity can also do that （ β = 1.070 , P =0.001 ） HRT＇s volume is no statisti cal significance （ P = 0.991 ）. RTA＇s PPAT has the best sensitive（ β= 2.936, P = 0. 000）, PFAT is the second （ β = 2. 413, P = 0. 000） and FAT is the third （ β = 2. 250, P = 0.001 ） in RTA. 3. To compare the three technologies, OCT＇s neural epithelium thickness has the best sensitivity（ β = 1. 695, P = 0.008） , HRT＇s volume is the second（ β= 1. 304, P = 0.037） and RTA＇s PFAT is the third（ β= 0.936, P = 0. 141 ）in distinguishing drusen group and dry AMD group（ RTA- PFAT＇s P value is near to 0.1, so it could be seem as statistical significance in considering from clinical meaning）. The interaction among three machines has no statistical significance（ P = 0. 414）. （4）OCT＇s neural epithelium thickness has the best sensitivity （ β= 16. 629, P = 0. 103） and RTA＇s PPAT is the second （β = 5. 145, P = 0. 140） in distinguishing drusen group and wet AMD group OCT＇s neural epithelium thickness and RTA-PPAT＇s P value is near to 0. 1, so it could be seem as statistical significance in considering from clinical meaning）. HRTZ signal width P = 0.567） and the interaction among three machines （ P =0.386） has no statistical significance. Conclusions OCT＇s neural epithelium thickness can be used a tool to diagnosis earlier and distinguish AMD. HRT＇s macular volume sensitivity and RTA＇s PFAT can be used to distinguish drusen group and dry AMD group. HRT＇s Z signal width and RTA＇s PPAT can be used to distinguish drusen group and wet AMD group.