Abstract
Objectives: To elucidate the influence of eye selection criteria (ESC) on the reliability of biomarkers in diagnosis and prediction using pre-surgical parameters, assessments were undertaken as the subject of analysis. Methods: Pituitary macroadenoma (PMA) diagnosis and postsurgical visual function recovery biomarker analysis was used as the subject to illustrate the point. Six datasets (right, left, best, worst, random and both eyes), derived from a longitudinal study that involved 42 PMA patients and age-matched healthy volunteers, were generated. A comparison of the diagnostic efficacies of the amplitude of pattern visual evoked potentials (pVEP) and bi-nasal sector thickness in the ganglion cells complex plus the inner plexiform layer was performed using ESC. Afterwards, multivariate models for PMA diagnosis and the prediction of postsurgical visual function recovery, using Stable Sparse Biomarkers Detection methodology, were developed. A comprehensive evaluation was performed once for controls and in pre-surgical PMA patients at 3 and 12 months after transsphenoidal tumor removal. Results: The proposed biomarkers displayed specificity and sensibility ≥ 0.74 and AUC ≥ 0.87. The diagnostic values derived were ESC-dependent. All the prediction models had accuracies over 0.96, and the proposed biomarkers had stability ≥ 99% and the highest β values. Conclusions: Although the diagnostic values of the proposed biomarkers are affected by ESC, they exhibit equal accuracy for the same eye. Worse eye data represent the best choice for the analysis. Further studies are needed to validate the models for use in the prediction of the 12-month postsurgical restoration of parvocellular traffic.