Photo Credit: lekkimly168

The following is a summary of “Validation of a Visual Field Prediction Tool for Glaucoma: A Multicenter Study Involving Patients with Glaucoma in the United Kingdom,” published in the January 2025 issue of Ophthalmology by Dean et al. 

Researchers conducted a retrospective study to examine the performance of a previously developed machine-learning approach with Kalman-filtering (KF) technology in predicting disease trajectory for patients with glaucoma using real-world data.  

They evaluated the performance of a previously validated KF model (PKF), initially trained using data from the African Descent and Glaucoma Evaluation Study and Diagnostic Innovations in Glaucoma Study, on patients with various glaucoma types and severities receiving care in the United Kingdom (UK). The predictive accuracy was compared to 2 conventional linear regression (LR) models and a newly developed KF model trained on UK patients (UK-KF).  

The results showed 3,116 patients with open-angle glaucoma or suspects were split into training (n=1,584) and testing (n=1,532) sets. At 60 months follow-up, the predictive accuracy for MD within 2.5 dB of the observed value was 75.7% for the PKF, significantly outperforming the LR models (P <0.01 for both) and being similar to the UK-KF model (75.2%, P =0.70). The proportion of MD predictions within the 95% repeatability intervals at 60 months was 67.9% for PKF, higher than for the LR models (40.2%, 40.9%) and similar to UK-KF (71.4%).  

Investigators concluded that the study validated the performance of the previously developed KF model on a real-world, multicentre patient population, demonstrating its superiority over the current clinical standard (LR) and its strong predictive capability across various glaucoma types and severities supporting further prospective studies for clinical implementation. 

Source: ajo.com/article/S0002-9394(25)00021-2/abstract