Rozema Jos J, Lechat Momo, Tassignon Marie-José
*PhD †MaO ‡MD, PhD Department of Ophthalmology (JJR, M-JT), Antwerp University Hospital, Edegem; and Departments of Medicine and Health Science (JJR, M-JT) and Sciences (ML), University of Antwerp, Wilrijk, Belgium.
Optom Vis Sci. 2013 Nov;90(11):1342-8. doi: 10.1097/OPX.0000000000000071.
To reconstruct and correct four historical biometry data sets.
The data sets measured by Stenström (1946), Sorsby (1957), Zeeman (1911), and Awerbach (1900) are of great historical interest but have issues that need to be dealt with before they may be of practical use in comparative studies. These issues range from only publishing statistical descriptors instead of the numerical data (Stenström) to assuming an equivalent refractive index of the crystalline lens that was constant with age (Sorsby) and calculation errors (Awerbach and Zeeman). These issues were corrected by only considering subjects with ages ranging between 20 and 35 years and refraction within ±8D. This left 86 eyes for Sorsby, 73 for Zeeman, and 54 for Awerbach that could be used for further study. For the Stenström data set, the statistical descriptors of 983 eyes were used to generate data that are statistically indistinguishable from the original data through multivariate Gaussian modeling.
After verification that all biometric parameters, except refraction, could be well approximated by a Gaussian distribution, a statistical eye model was defined for the Stenström data set. By means of the "two-one-sided t test" procedure, it was demonstrated that the data generated by the model are statistically indistinguishable from the original data at a confidence level of 99%.
These results provide the best possible correction and reconstruction of the historical data sets. Provided one acknowledges the limitations of the data acquisition at the time, these data may be of use in studies on how ocular biometry has changed in the past century.
重建并校正四个历史生物测量数据集。
由斯特伦斯特伦（1946年）、索斯比（1957年）、塞曼（1911年）和奥韦尔巴赫（1900年）测量的数据集具有重大历史意义，但在可用于比较研究之前存在一些需要处理的问题。这些问题包括只公布统计描述符而非数值数据（斯特伦斯特伦）、假设晶状体的等效折射率随年龄恒定（索斯比）以及计算误差（奥韦尔巴赫和塞曼）。通过仅考虑年龄在20至35岁之间且屈光度在±8D范围内的受试者来校正这些问题。这使得索斯比数据集剩下86只眼睛、塞曼数据集剩下73只眼睛、奥韦尔巴赫数据集剩下54只眼睛可用于进一步研究。对于斯特伦斯特伦数据集，利用983只眼睛的统计描述符通过多元高斯建模生成在统计上与原始数据无法区分的数据。
在验证除屈光度外所有生物测量参数都能很好地用高斯分布近似后，为斯特伦斯特伦数据集定义了一个统计眼模型。通过“双单侧t检验”程序表明，在99%的置信水平下，该模型生成的数据在统计上与原始数据无法区分。
这些结果对历史数据集进行了尽可能好的校正和重建。只要认识到当时数据采集的局限性，这些数据可用于研究过去一个世纪眼生物测量是如何变化的。
