Methods: Secondary analyses of a prospective, multicenter diagnostic study (GATE) referred to hospital eye services in the United Kingdom. We included data from 899 of 966 participants, referred to hospital eye services with suspected glaucoma or ocular hypertension. We used both eyes’ data and logistic
regression-based Receiver Operator Characteristics (ROC) analysis to build a set of models to measure the sensitivity and specificity of the average and inferior quadrant RNFL thickness data of OCT. The reference standard was expert clinician examination including automated perimetry. The main outcome measures were sensitivity at 0.95 specificity and specificity at 0.95
sensitivity and the corresponding RNFL thickness thresholds. We explored the possibility of accuracy improvement by adding measures of within-eye and between-eye variation, scan quality, intraocular pressure, and age.
Results: Glaucoma was diagnosed in at least one eye in 17% of participants. Areas under the curve were between 0.83 and 0.88. When specificity was fixed at 0.95, the sensitivity was between 0.38 and 0.55, the highest values were reached with models including the inferior quadrant rather than the average RNFL thickness. Fixing specificity at 0.95 the sensitivity was between 0.36 and 0.58. The addition of age, refractive error, IOP, or within-subject variation did not improve the accuracy.
Conclusion: RNFL thickness data of OCT can be used as a diagnostic test but accuracy estimates remain moderate even in exploratory multivariable modelling of aiming to improve accuracy.
Bibliographical noteFinancial Support:
- The GATE study was funded by the National Institute of Health Research HTA
programme (09/22/111). The views and opinions expressed are those of the authors and do not necessarily reflect those of the HTA programme, NIHR, NHS or the Department of Health.
- The contribution of Manuele Michelessi of the IRCCS Fondazione Bietti in this paper was supported by the Italian Ministry of Health and by Fondazione Roma
- Diagnostic tests/Investigation