Purchase this article with an account.
irene c. floriani, Sonia Jain, Feng He, Valter Torri, Davide Poli, Mae O. Gordon, Stefano Miglior, Norbert Pfeiffer, Julie Huecker, Linda M. Zangwill; Validation Of Data Pooling Of The Confocal Scanning Laser Ophthalmoscopy (cslo) As Predictor Of The Onset Of Primary Open Angle Glaucoma (poag). Invest. Ophthalmol. Vis. Sci. 2011;52(14):4134.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Pooled analyses (PA) are an important tool for investigating the role of therapies and predictive factors in a specific disease. Although not as exhaustive as systematic reviews, by combining data from different studies with similar aims and design, PA allows meaningful interpretation of results once cross-validation has been performed and evaluated. This project describes the pooled analysis methodology for OHTS and EGPS to assess the adequacy of pooling CSLO data to predict the onset of POAG.
To assess the appropriateness of pooling OHTS and EGPS CSLO data, 5 separate OHTS multivariable prediction models (adjusting for baseline IOP, baseline age, and central corneal thickness) that included one HRT parameter (neuroretinal rim area, neuroretinal rim volume, mean height contour, rim to disc ratio or Moorfields Regression Analysis result) were tested on EGPS and OHTS data using the 1) z-test statistic to compare Cox proportional hazard ratios (HR), 2) c statistic to discriminate between participants who did and did not develop POAG, and 3) the calibration chi-square for assessing the overestimation/underestimation of predicted risk.
The 5 pooled multivariable models for the development of POAG had good discrimination (c statistics all greater than 0.75) and accurate estimate of POAG risk (calibration Chi square always less than 17.8), and similar hazard ratios (z-test p-values all greater than 0.3). Pooling was determined to be appropriate, and the two separate study samples were combined without reference to study stratification.
The described approach outlines methods that can be used to determine whether studies can be pooled together for data analysis. These results indicate that the OHTS and EGPS study samples can be pooled thus increasing precision and generalizability of the POAG predictive model.
This PDF is available to Subscribers Only