Purchase this article with an account.
F.A. Medeiros, P.A. Sample, L.M. Zangwill, C.F. Gomi, C. Bowd, J.G. Crowston, R.N. Weinreb; Validation of a Predictive Model for the Conversion From Ocular Hypertension to Glaucoma . Invest. Ophthalmol. Vis. Sci. 2005;46(13):2471.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Purpose: To assess the external validity of a predictive model for the risk of conversion from ocular hypertension to glaucoma. Methods: Predictive models were derived from the Ocular Hypertension Treatment Study (OHTS) results. The performance of these models was assessed in an independent population from a longitudinal study (Diagnostic Innovations in Glaucoma Study – DIGS). Ocular hypertensive patients included in DIGS had elevated intraocular pressure (IOP) in both eyes (>22 mmHg), normal visual fields and normal optic discs. Glaucoma conversion was defined as development of reproducible visual field loss or progression of optic disc damage. Data on baseline risk factors for development of glaucoma, including age, IOP, central corneal thickness (CCT), vertical cup disc ratio (VCD), pattern standard deviation (PSD) and presence of diabetes mellitus were assessed for each patient. A predictive model was derived from the DIGS population and compared to the models derived from the OHTS results. For each risk factor, regression coefficients (hazard ratios) from the DIGS– and OHTS–derived models were compared using a z–test. The ability of the risk models to discriminate patients who developed glaucoma during follow–up from those who did not was evaluated by calculating the c statistic (ROC curve area). Model calibration was evaluated by comparing predicted and observed glaucoma events. Results: 37 (24%) of 153 patients developed glaucoma during follow–up. Mean follow–up time was 8 years (range: 1 to 17 years). 20 (54%) patients developed progression of optic disc, 15 (40%) developed abnormal visual fields and 2 (5%) developed both. Hazard ratios for DIGS– and OHTS–derived predictive models were similar for CCT (1.64 vs. 1.82 [per 40µm thinner]), IOP (1.10 vs. 1.11 [per 1 mmHg higher]), age (1.35 vs. 1.25 [per decade]), VCD (1.29 vs. 1.32 [per 0.1 higher]) and PSD (1.07 vs 1.25 [per 0.2 dB higher]). Hazard ratios were significantly different for the presence of diabetes (1.17 vs. 0.35). Similar results were obtained when risk models were developed without PSD and VCD. ROC curve areas for the discriminatory abilities of the OHTS risk models when applied to the DIGS population ranged from 0.68 to 0.71 and were similar to those of DIGS–derived models. Calibrations of models without VCD and PSD were better than those of models incorporating these variables. Conclusions: Risk models derived from the OHTS results performed reasonably well to predict risk of conversion from ocular hypertension to glaucoma in an independent population.
This PDF is available to Subscribers Only